1 2018-09-05 Martin Liska <mliska@suse.cz>
3 * doc/gcov.texi: Update documentation of humar
5 * gcov.c (format_count): Print one decimal place, it provides
6 more fine number of situations like '1G' vs. '1.4G'.
8 2018-09-05 Martin Liska <mliska@suse.cz>
11 * config/rs6000/rs6000.opt: Mark the option as Deprecated.
12 * optc-gen.awk: Allow 'Var' for Deprecated options in order
13 to generate a MASK value.
15 2018-09-04 H.J. Lu <hongjiu.lu@intel.com>
18 * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
19 if frame pointer isn't used.
20 (compute_frame_pointer_to_fb_displacement): Likewise.
22 2018-09-04 Jakub Jelinek <jakub@redhat.com>
25 * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
26 OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
27 OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
28 (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
29 and OPTION_MASK_ISA_XSAVEC_UNSET.
31 2018-09-04 Max Filippov <jcmvbkbc@gmail.com>
33 * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
34 XOR operations in NAND case.
36 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
38 * wide-int-range.cc (wide_int_range_convert): New.
39 * wide-int-range.h (wide_int_range_convert): New.
40 * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
41 code into wide_int_range_convert.
42 (extract_range_into_wide_ints): Do not munge anti range constants
43 into the entire domain. Just return the range back.
45 2018-09-04 Martin Liska <mliska@suse.cz>
47 * genmatch.c (output_line_directive): Add new argument
49 (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
51 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
53 * doc/invoke.texi (Option Summary): Add whitespace.
55 * doc/invoke.texi (Option Summary): Add -Waligned-new.
57 2018-09-04 Richard Biener <rguenther@suse.de>
59 PR tree-optimization/87211
60 * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
61 backedge value we're supposed to treat as VARYING also number
62 the PHI to VARYING in case it got a different value-number already.
64 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
66 * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
67 (extract_range_from_binary_expr_1): Do not call
68 vrp_can_optimize_bit_op.
69 * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
71 (wide_int_range_get_mask_and_bounds): New.
72 (wide_int_range_optimize_bit_op): New.
73 (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
74 (wide_int_range_bit_and): Same.
75 * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
76 (wide_int_range_optimize_bit_op): New.
77 (wide_int_range_get_mask_and_bounds): New.
79 2018-09-04 Richard Biener <rguenther@suse.de>
81 PR tree-optimization/87176
82 * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
83 variable. When value-numbering a virtual PHI node make sure
84 to not value-number to the backedge value.
86 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
88 * doc/extend.texi (Long Long, Hex Floats): Document support for
89 long long and hex floats in more recent versions of ISO C++.
91 2018-09-03 Richard Biener <rguenther@suse.de>
93 PR tree-optimization/87177
94 * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
97 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
99 * bb-reorder.c (edge_order): Convert to C-qsort-style
100 tri-state comparator.
101 (reorder_basic_blocks_simple): Change std::stable_sort to gcc_stablesort.
103 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
105 * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
106 tri-state comparator.
107 (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
109 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
111 * sort.cc (struct sort_ctx): New field 'nlim'. Use it...
112 (mergesort): ... here as maximum count for using netsort.
113 (gcc_qsort): Set nlim to 3 if stable sort is requested.
114 (gcc_stablesort): New.
115 * system.h (gcc_stablesort): Declare.
117 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
119 * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
120 * system.h (qsort): Always redirect to gcc_qsort. Update comment.
121 * vec.c (qsort_chk): Do not call gcc_qsort. Update comment.
123 2018-09-03 Segher Boessenkool <segher@kernel.crashing.org>
125 * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
126 lxsdx and stxsdx alternatives.
127 (*mov<mode>_hardfloat64): Ditto.
128 * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
130 2018-09-03 Richard Biener <rguenther@suse.de>
132 PR tree-optimization/87200
133 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
136 2018-09-03 Martin Liska <mliska@suse.cz>
138 PR tree-optimization/87201
139 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
140 Fix parenthesis in an expression.
142 2018-09-03 Richard Biener <rguenther@suse.de>
144 PR tree-optimization/87197
145 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
146 visited. CSE the VN_INFO hashtable lookup.
148 PR tree-optimization/87169
149 * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
150 iterating make sure there's no extra backedges from irreducible
151 regions feeding the header. Mark the destination block
154 2018-09-03 Martin Liska <mliska@suse.cz>
157 * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
158 * common/common-targhooks.c (default_get_valid_option_values):
160 * common/common-targhooks.h (default_get_valid_option_values):
162 * common/config/i386/i386-common.c: Move processor_target_table
164 (ix86_get_valid_option_values): New function.
165 (TARGET_GET_VALID_OPTION_VALUES): New macro.
166 * config/i386/i386.c (struct ptt): Move to i386-common.c.
167 (PTA_*): Move all defined masks into i386-common.c.
168 (ix86_function_specific_restore): Use new processor_cost_table.
169 * config/i386/i386.h (struct ptt): Moved from i386.c.
170 (struct pta): Likewise.
171 * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
172 * doc/tm.texi.in: Likewise.
173 * opt-suggestions.c (option_proposer::suggest_option):
174 Pass prefix to build_option_suggestions.
175 (option_proposer::get_completions): Likewise.
176 (option_proposer::build_option_suggestions): Use the new target
178 * opts.c (struct option_help_tuple): New struct.
179 (print_filtered_help): Use the new target hook.
181 2018-09-03 Martin Liska <mliska@suse.cz>
184 * predict.c (set_even_probabilities): Add likely_edges
185 argument and handle cases where we have precisely one
187 (combine_predictions_for_bb): Catch also likely_edges.
188 (tree_predict_by_opcode): Handle gswitch statements.
189 * tree-cfg.h (find_case_label_for_value): New declaration.
190 (find_taken_edge_switch_expr): Likewise.
191 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
192 Find pivot in decision tree based on probabily, not by number of
195 2018-09-02 Gerald Pfeifer <gerald@pfeifer.com>
197 * doc/standards.texi (Standards): Update Objective-C reference.
199 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
201 * doc/install.texi (Prerequisites): Update link for MPC.
203 2018-09-01 Michael Matz <matz@suse.de>
205 PR tree-optimization/87074
206 * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
207 PHIs for outer-loop uses.
209 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
211 * doc/generic.texi (OpenMP): Adjust link to openmp.org.
212 * doc/invoke.texi (C Dialect Options): Ditto.
214 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
216 * doc/install.texi (Prerequisites): Adjust link mpfr.org.
218 2018-08-31 Richard Biener <rguenther@suse.de>
220 PR tree-optimization/87168
221 * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
222 (rpo_elim::eliminate_avail): When OP was not visited it must
225 2018-08-31 David Malcolm <dmalcolm@redhat.com>
227 * tree-vrp.c (copy_value_range): Convert param "from" from
228 "value_range *" to "const value_range *".
229 (range_is_null): Likewise for param "vr".
230 (range_int_cst_p): Likewise.
231 (range_int_cst_singleton_p): Likewise.
232 (symbolic_range_p): Likewise.
233 (value_ranges_intersect_p): Likewise for both params.
234 (value_range_nonnegative_p): Likewise for param "vr".
235 (value_range_constant_singleton): Likewise.
236 (vrp_set_zero_nonzero_bits): Likewise for param "ar".
237 (extract_range_into_wide_ints): Likewise for param "vr".
238 (extract_range_from_multiplicative_op): Likewise for params "vr0"
240 (vrp_can_optimize_bit_op): Likewise.
241 (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
243 (extract_range_from_unary_expr): Likewise.
244 (debug_value_range): Likewise for param "vr".
245 (value_range::dump): Add "const" qualifier.
246 (vrp_prop::check_array_ref): Convert local "vr" from
247 "value_range *" to "const value_range *".
248 (vrp_prop::check_mem_ref): Likewise.
249 (vrp_prop::visit_stmt): Likewise for local "old_vr".
250 (vrp_intersect_ranges_1): Likewise for param "vr_1".
251 (vrp_intersect_ranges): Likewise.
252 (simplify_stmt_for_jump_threading): Likewise for local "vr".
253 (vrp_prop::vrp_finalize): Likewise.
254 * tree-vrp.h (value_range::dump): Add "const" qualifier.
255 (vrp_intersect_ranges): Add "const" qualifier to params as above.
256 (extract_range_from_unary_expr): Likewise.
257 (value_range_constant_singleton): Likewise.
258 (symbolic_range_p): Likewise.
259 (copy_value_range): Likewise.
260 (extract_range_from_binary_expr_1): Likewise.
261 (range_int_cst_p): Likewise.
262 (vrp_set_zero_nonzero_bits): Likewise.
263 (range_int_cst_singleton_p): Likewise.
265 2018-08-31 Vlad Lazar <vlad.lazar@arm.com>
267 * config/aarch64/arm_neon.h (vabsd_s64): New.
268 (vnegd_s64): Likewise.
270 2018-08-31 Martin Jambor <mjambor@suse.cz>
272 * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
274 2018-08-31 Martin Liska <mliska@suse.cz>
276 * ipa-icf.c (sem_item::add_type): Use
277 sem_item::m_type_hash_cache.
278 * ipa-icf.h: Move the cache from sem_item_optimizer
281 2018-08-31 Nathan Sidwell <nathan@acm.org>
283 * doc/extend.texi (Backwards Compatibility): Remove implicit
284 extern C leeway of () being (...).
286 2018-08-31 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
288 * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
290 2018-08-31 Segher Boessenkool <segher@kernel.crashing.org>
294 * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
296 2018-08-31 Jakub Jelinek <jakub@redhat.com>
299 * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
300 gen_int_mode. Formatting fixes.
302 2018-08-30 Sandra Loosemore <sandra@codesourcery.com>
304 * target.def (custom_function_descriptors): Improve documentation.
305 * doc/tm.texi.in (Trampolines): Expand discussion of function
306 descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
307 beginning of the section.
308 * doc/tm.texi: Regenerated.
310 2018-08-30 Jose E. Marchesi <jose.marchesi@oracle.com>
312 * cfg.h (class auto_edge_flag): Spell out the template-id of the
313 base class in the initializer list. This is a workaround for
314 building with older GCC.
315 (class auto_bb_flag): Likewise.
317 2018-08-30 Aaron Sawdey <acsawdey@linux.ibm.com>
319 * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
320 (altivec_vcmpequ<VI_char>_p): Remove star.
321 * config/rs6000/rs6000-string.c (do_load_for_compare): Support
323 (expand_strncmp_vec_sequence): New function.
324 (emit_final_str_compare_vec): New function.
325 (expand_strn_compare): Add support for vector strncmp.
326 * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
327 length specification to bytes.
328 * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
329 (vcmpnezb_p): New pattern.
330 * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
331 for option -mstring-compare-inline-limit.
333 2018-08-30 Thiago Macieira <thiago.macieira@intel.com>
335 * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
336 (PTA_SKYLAKE): Add PTA_AES.
337 (PTA_GOLDMONT): Likewise.
339 2018-08-29 Jan Hubicka <jh@suse.cz>
342 * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
343 * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
345 2018-08-29 Jan Hubicka <jh@suse.cz>
347 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
349 (hash_tree): Do not visit TYPE_STUB_DECL.
350 * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
351 stream TYPE_STUB_DECL.
352 * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
353 * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
354 after free_lang_data.
355 (type_in_anonymous_namespace_p): Likewise.
357 2018-08-29 Jan Hubicka <jh@suse.cz>
359 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
360 comment that it has to be even number.
361 (class sreal): Change m_sig type to int32_t.
362 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
363 int64_t for temporary calculations.
364 (sreal_verify_basics): Drop one bit from minimum and maximum.
366 2018-08-30 Richard Biener <rguenther@suse.de>
368 PR tree-optimization/87147
369 * tree-ssa-sccvn.c (SSA_VISITED): New function.
370 (visit_phi): When the degenerate result is from the backedge and
371 we didn't visit its definition yet drop to VARYING.
372 (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
374 2018-08-29 Jan Hubicka <jh@suse.cz>
376 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
378 (hash_tree): Likewise.
380 2018-08-29 Jan Hubicka <jh@suse.cz>
382 * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
383 and TYPE_NEXT_REF_TO.
385 2018-08-29 Jan Hubicka <jh@suse.cz>
387 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
388 comment that it has to be even number.
389 (class sreal): Change m_sig type to int32_t.
390 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
391 int64_t for temporary calculations.
392 (sreal_verify_basics): Drop one bit from minimum and maximum.
394 2018-08-30 Tamar Christina <tamar.christina@arm.com>
396 * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
398 2018-08-30 Vlad Lazar <vlad.lazar@arm.com>
401 * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
402 if to_add is negative.
404 2018-08-29 Bernd Edlinger <bernd.edlinger@hotmail.de>
407 * builtins.c (c_strlen): Improve range checks.
409 2018-08-29 Martin Sebor <msebor@redhat.com>
410 Jeff Law <law@redhat.com>
412 PR tree-optimization/86714
413 PR tree-optimization/86711
414 * builtins.c (c_strlen): Add arguments to call to string_constant.
415 * expr.c (string_constant): Add argument. Detect missing nul
416 terminator and outermost declaration it's missing in.
417 * expr.h (string_constant): Add argument.
418 * fold-const.c (read_from_constant_string): Add arguments to call to
420 (c_getstr): Likewise.
421 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
423 * tree-ssa-strlen.c (get_stridx): Likewise.
425 2018-08-29 Jan Hubicka <jh@suse.cz>
427 * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
428 Do not stream DECL_VINDEX.
429 * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
430 * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
431 (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
433 2018-08-29 Richard Biener <rguenther@suse.de>
435 * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
436 virtual operands that are not default defs to honor region
438 (rpo_vn_valueize): Remove ineffective code here.
440 2018-08-29 Richard Biener <rguenther@suse.de>
442 PR tree-optimization/87132
443 * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
444 when skipping defs reachable over backedges.
446 2018-08-29 Richard Biener <rguenther@suse.de>
448 * tree-core.h: Document use of deprecated_flag in SSA_NAME.
449 * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
450 * tree-into-ssa.c (pass_build_ssa::execute): Initialize
451 function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
452 * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
453 (vn_reference_lookup_3): Remove use of const_parms.
454 (free_rpo_vn): Do not free const_parms.
455 (do_rpo_vn): Do not call init_const_parms.
456 * tree-ssa-alias.c (refs_may_alias_p_1): Honor
457 SSA_NAME_POINTS_TO_READONLY_MEMORY.
458 (call_may_clobber_ref_p_1): Likewise.
460 2018-08-29 Alexander Monakov <amonakov@ispras.ru>
463 * invoke.texi (Optimization Options): List -ftree-scev-cprop.
465 (-ftree-scev-cprop): Document.
467 2018-08-29 Jan Hubicka <jh@suse.cz>
469 * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
471 (sreal constructor): Update.
472 * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
473 sreal:operator/): Update.
475 2018-08-29 Martin Liska <mliska@suse.cz>
477 * tree-switch-conversion.c (switch_conversion::expand):
478 Strenghten assumption about gswitch statements.
480 2018-08-29 Richard Biener <rguenther@suse.de>
482 PR tree-optimization/87117
483 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
484 re-value-number released SSA VDEFs.
486 2018-08-29 Richard Biener <rguenther@suse.de>
488 PR tree-optimization/87126
489 * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
491 2018-08-28 Jim Wilson <jimw@sifive.com>
493 * config/riscv/pic.md: Rewrite.
494 * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
496 * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
497 (SOFTF, default_load, softload, softstore): New.
499 2018-08-28 Jeff Law <law@redhat.com>
501 * fold-const.c (fold_binary_loc): Remove recently added assert.
503 2018-08-28 Joern Rennecke <joern.rennecke@riscy-ip.com>
505 * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
506 to OP parmeter of generated function.
508 2018-08-28 MCC CS <deswurstes@users.noreply.github.com>
510 PR tree-optimization/87009
511 * match.pd: Add boolean optimizations.
513 2018-08-28 Martin Sebor <msebor@redhat.com>
516 * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
517 * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
518 (pass_walloca::gate): Use it.
519 (alloca_call_type): Same.
520 (pass_walloca::execute): Same.
521 * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
523 2018-08-28 David Malcolm <dmalcolm@redhat.com>
525 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
526 GCC_VERSION for usage of "__gcc_dump_printf__" format from
529 2018-08-28 Richard Biener <rguenther@suse.de>
531 PR tree-optimization/87124
532 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
533 constants before looking up avail.
535 2018-08-28 Jakub Jelinek <jakub@redhat.com>
538 * calls.c (maybe_warn_nonstring_arg): Punt early if
539 warn_stringop_overflow is zero. Don't call get_range_strlen
540 on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
541 Swap comparison operands to have constants on rhs. Only use
542 lenrng[1] if non-NULL and INTEGER_CST. Don't uselessly
545 2018-08-28 Richard Sandiford <richard.sandiford@arm.com>
547 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
548 use of tree_to_shwi. Remove duplicated test for the size being
549 a whole number of bytes.
551 2018-08-28 Richard Biener <rguenther@suse.de>
553 PR tree-optimization/87117
554 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
555 Handle removed stmt without LHS (GIMPLE_NOP).
557 2018-08-28 Richard Biener <rguenther@suse.de>
559 PR tree-optimization/87117
560 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
561 void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
563 2018-08-28 Richard Biener <rguenther@suse.de>
565 PR tree-optimization/87117
566 * tree-ssa-pre.c (compute_avail): Do not make expressions
567 with predicated values available.
568 (get_expr_value_id): Assert we do not run into predicated value
571 2018-08-28 Richard Biener <rguenther@suse.de>
573 PR tree-optimization/87117
574 * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
575 get virtual operands.
576 (get_expr_operands): Handle STRING_CST like other decls.
578 2018-08-28 Martin Liska <mliska@suse.cz>
580 * tree.h: Update documentation of fndecl_built_in_p
584 2018-08-27 Jeff Law <law@redhat.com>
585 PR tree-optimization/87110
586 * tree-ssa-dse.c (compute_trims): Handle non-constant
589 2018-08-27 Martin Sebor <msebor@redhat.com>
591 PR tree-optimization/86914
592 * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
594 2018-08-27 Martin Sebor <msebor@redhat.com>
596 PR tree-optimization/87112
597 * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
598 the type of the bound argument.
600 2018-08-27 Jeff Law <law@redhat.com>
602 * tree-ssa-dse.c (compute_trims): Handle case where the reference's
603 type does not have a TYPE_SIZE_UNIT.
605 2018-08-27 Steve Ellcey <sellcey@cavium.com>
607 * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
608 with include of backend.h.
610 2018-08-27 Richard Biener <rguenther@suse.de>
612 PR tree-optimization/86927
613 * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
614 use const cond reduction code.
616 2018-08-27 Alexander Monakov <amonakov@ispras.ru>
618 PR tree-optimization/85758
619 * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
621 2018-08-27 David Malcolm <dmalcolm@redhat.com>
624 * diagnostic-show-locus.c (class layout_range): Update for
625 conversion of show_caret_p to a tri-state.
626 (layout_range::layout_range): Likewise.
627 (make_range): Likewise.
628 (layout::maybe_add_location_range): Likewise.
629 (layout::should_print_annotation_line_p): Don't show annotation
630 lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
631 (layout::get_state_at_point): Update for conversion of
632 show_caret_p to a tri-state. Bail out early for
633 SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
634 underlining or source colorization.
635 (gcc_rich_location::add_location_if_nearby): Update for conversion
636 of show_caret_p to a tri-state.
637 (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
638 (selftest::test_one_liner_fixit_replace_equal_secondary_range):
640 (selftest::test_one_liner_labels): Likewise.
641 * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
642 conversion of show_caret_p to a tri-state.
643 * pretty-print.c (text_info::set_location): Likewise.
644 * pretty-print.h (text_info::set_location): Likewise.
645 * substring-locations.c (format_warning_n_va): Likewise.
646 * tree-diagnostic.c (default_tree_printer): Likewise.
647 * tree-pretty-print.c (newline_and_indent): Likewise.
649 2018-08-27 David Malcolm <dmalcolm@redhat.com>
652 * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
653 line above for line-insertion fix-it hints.
654 (selftest::test_fixit_insert_containing_newline): Update the
655 expected results, and add a test with line-numbering enabled.
657 2018-08-27 Martin Liska <mliska@suse.cz>
660 * sanopt.c (sanitize_rewrite_addressable_params): Ignore
661 params with DECL_HAS_VALUE_EXPR_P.
663 2018-08-27 Martin Liska <mliska@suse.cz>
665 * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
666 selected expansion strategy.
668 2018-08-27 Martin Liska <mliska@suse.cz>
670 * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
671 * builtins.c (is_builtin_fn): Likewise.
672 * attribs.c (diag_attr_exclusions): Use new function
673 fndecl_built_in_p and remove check for FUNCTION_DECL if
675 (builtin_mathfn_code): Likewise.
676 (fold_builtin_expect): Likewise.
677 (fold_call_expr): Likewise.
678 (fold_builtin_call_array): Likewise.
679 (fold_call_stmt): Likewise.
680 (set_builtin_user_assembler_name): Likewise.
681 (is_simple_builtin): Likewise.
682 * calls.c (gimple_alloca_call_p): Likewise.
683 (maybe_warn_nonstring_arg): Likewise.
684 * cfgexpand.c (expand_call_stmt): Likewise.
685 * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
686 (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
687 (cgraph_node::verify_node): Likewise.
688 * cgraphclones.c (build_function_decl_skip_args): Likewise.
689 (cgraph_node::create_clone): Likewise.
690 * config/arm/arm.c (arm_insert_attributes): Likewise.
691 * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
692 * dse.c (scan_insn): Likewise.
693 * expr.c (expand_expr_real_1): Likewise.
694 * fold-const.c (operand_equal_p): Likewise.
695 (fold_binary_loc): Likewise.
696 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
697 * gimple-low.c (lower_stmt): Likewise.
698 * gimple-pretty-print.c (dump_gimple_call): Likewise.
699 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Likewise.
700 * gimple.c (gimple_build_call_from_tree): Likewise.
701 (gimple_call_builtin_p): Likewise.
702 (gimple_call_combined_fn): Likewise.
703 * gimplify.c (gimplify_call_expr): Likewise.
704 (gimple_boolify): Likewise.
705 (gimplify_modify_expr): Likewise.
706 (gimplify_addr_expr): Likewise.
707 * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
708 * ipa-cp.c (determine_versionability): Likewise.
709 * ipa-fnsummary.c (compute_fn_summary): Likewise.
710 * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
711 * ipa-split.c (visit_bb): Likewise.
712 (split_function): Likewise.
713 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
714 * lto-cgraph.c (input_node): Likewise.
715 * lto-streamer-out.c (write_symbol): Likewise.
716 * omp-low.c (setjmp_or_longjmp_p): Likewise.
717 (lower_omp_1): Likewise.
718 * predict.c (strip_predict_hints): Likewise.
719 * print-tree.c (print_node): Likewise.
720 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
721 * trans-mem.c (is_tm_irrevocable): Likewise.
722 (is_tm_load): Likewise.
723 (is_tm_simple_load): Likewise.
724 (is_tm_store): Likewise.
725 (is_tm_simple_store): Likewise.
726 (is_tm_abort): Likewise.
727 (tm_region_init_1): Likewise.
728 * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
729 * tree-cfg.c (verify_gimple_call): Likewise.
730 (move_stmt_r): Likewise.
731 (stmt_can_terminate_bb_p): Likewise.
732 * tree-eh.c (lower_eh_constructs_2): Likewise.
733 * tree-if-conv.c (if_convertible_stmt_p): Likewise.
734 * tree-inline.c (remap_gimple_stmt): Likewise.
736 (estimate_num_insns): Likewise.
737 (fold_marked_statements): Likewise.
738 * tree-sra.c (scan_function): Likewise.
739 * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
740 (optimize_stack_restore): Likewise.
741 (pass_fold_builtins::execute): Likewise.
742 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
743 (mark_all_reaching_defs_necessary_1): Likewise.
744 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
745 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
746 (pass_forwprop::execute): Likewise.
747 * tree-ssa-loop-im.c (stmt_cost): Likewise.
748 * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
749 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
750 * tree-ssa-strlen.c (get_string_length): Likewise.
751 * tree-ssa-structalias.c (handle_lhs_call): Likewise.
752 (find_func_aliases_for_call): Likewise.
753 * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
754 * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
755 * tree-tailcall.c (find_tail_calls): Likewise.
756 * tree.c (need_assembler_name_p): Likewise.
757 (free_lang_data_in_decl): Likewise.
758 (get_call_combined_fn): Likewise.
759 * ubsan.c (is_ubsan_builtin_p): Likewise.
760 * varasm.c (incorporeal_function_p): Likewise.
761 * tree.h (DECL_BUILT_IN): Remove and replace with
763 (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
764 (fndecl_built_in_p): New.
766 2018-08-27 Martin Liska <mliska@suse.cz>
768 PR tree-optimization/86847
769 * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
770 Dump also subtree probability.
771 (switch_decision_tree::do_jump_if_equal): New function.
772 (switch_decision_tree::emit_case_nodes): Handle special
773 situations in balanced tree that can be emitted much simpler.
774 Fix calculation of probabilities that happen in tree expansion.
775 * tree-switch-conversion.h (struct cluster): Add
777 (struct simple_cluster): Likewise.
778 (struct case_tree_node): Add new function has_child.
779 (do_jump_if_equal): New.
781 2018-08-27 Martin Liska <mliska@suse.cz>
783 * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
784 Add new argument to bit_test_cluster constructor.
785 (bit_test_cluster::emit): Set bits really number of values
787 (bit_test_cluster::hoist_edge_and_branch_if_true): Add
788 probability argument.
789 * tree-switch-conversion.h (struct bit_test_cluster):
790 Add m_handles_entire_switch.
792 2018-08-27 Martin Liska <mliska@suse.cz>
794 PR tree-optimization/86702
795 * tree-switch-conversion.c (jump_table_cluster::emit):
796 Make probabilities even for values in jump table
797 according to number of cases handled.
798 (switch_decision_tree::compute_cases_per_edge): Pass
799 argument to reset_out_edges_aux function.
800 (switch_decision_tree::analyze_switch_statement): Likewise.
801 * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
804 2018-08-27 Martin Liska <mliska@suse.cz>
806 * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
807 cfun argument explicitly.
808 * gimple-pretty-print.c (dump_gimple_switch): Likewise.
809 * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
810 function gimple_switch_default_bb.
811 (convert_switch_statements):
813 * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
814 * stmt.c (label_to_block_fn): Use label_to_block and pass
815 cfun argument explicitly and use gimple_switch_label_bb.
816 (expand_case): Likewise.
817 * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
818 cfun argument explicitly. Likewise.
819 (make_edges_bb): Likewise.
820 (make_cond_expr_edges): Likewise.
821 (get_cases_for_edge): Likewise.
822 (make_gimple_switch_edges): Likewise.
823 (label_to_block_fn): Likewise.
824 (label_to_block): Likewise.
825 (make_goto_expr_edges): Likewise.
826 (make_gimple_asm_edges): Likewise.
827 (main_block_label): Likewise.
828 (group_case_labels_stmt): Likewise.
829 (find_taken_edge_computed_goto): Likewise.
830 (find_taken_edge_switch_expr): Likewise.
831 (gimple_verify_flow_info): Likewise.
832 (gimple_redirect_edge_and_branch): Likewise.
833 (gimple_switch_label_bb): New function.
834 (gimple_switch_default_bb): Likewise.
835 (gimple_switch_edge): Likewise.
836 (gimple_switch_default_edge): Likewise.
837 * tree-cfg.h (label_to_block_fn): Remove and replace ...
838 (label_to_block): ... with this.
839 (gimple_switch_label_bb): New.
840 (gimple_switch_default_bb): Likewise.
841 (gimple_switch_edge): Likewise.
842 (gimple_switch_default_edge): Likewise.
843 * tree-cfgcleanup.c (convert_single_case_switch): Use
844 new gimple functions and pass new argument to label_to_block.
845 (cleanup_control_flow_bb):
846 * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
847 cfun argument explicitly.
848 (make_eh_edges): Likewise.
849 (redirect_eh_dispatch_edge): Likewise.
850 (lower_resx): Likewise.
851 (lower_eh_dispatch): Likewise.
852 (maybe_remove_unreachable_handlers): Likewise.
853 (unsplit_eh): Likewise.
854 (cleanup_empty_eh): Likewise.
855 (verify_eh_edges): Likewise.
856 (verify_eh_dispatch_edge): Likewise.
857 * tree-ssa-dom.c (record_edge_info): Likewise.
858 * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
859 * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
860 (thread_through_normal_block): Likewise.
861 * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
862 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
863 * tree-switch-conversion.c (switch_conversion::collect): Use new
865 (switch_conversion::check_final_bb): Likewise.
866 (switch_conversion::gather_default_values): Pass new argument
868 (switch_conversion::build_constructors): Likewise.
869 (switch_decision_tree::compute_cases_per_edge): Use new
870 gimple_switch_edge function.
871 (switch_decision_tree::analyze_switch_statement): Pass new argument
873 (switch_decision_tree::try_switch_expansion): Use
874 gimple_switch_default_edge.
875 * tree-vrp.c (find_switch_asserts): Pass new argument
877 * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
878 (vr_values::simplify_switch_using_ranges): Likewise.
880 2018-08-27 Richard Biener <rguenther@suse.de>
882 * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
883 * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
885 * tree-ssa-sccvn.h (struct vn_pval): New structure.
886 (struct vn_nary_op_s): Add unwind_to member. Add
887 predicated_values flag and put result into a union together
888 with a linked list of vn_pval.
889 (struct vn_ssa_aux): Add name member to make maintaining
890 a map of SSA name to vn_ssa_aux possible. Remove no longer
891 needed info, dfsnum, low, visited, on_sccstack, use_processed
892 and range_info_anti_range_p members.
893 (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
894 (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
896 (vn_valueize): New global.
897 (vn_context_bb): Likewise.
898 (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
899 VN_INFO_PTR_INFO): Remove.
900 * tree-ssa-sccvn.c: ... (rewrite)
901 (pass_fre::execute): For -O2+ initialize loops and run
902 RPO VN in optimistic mode (iterating). For -O1 and -Og
903 run RPO VN in non-optimistic mode.
904 * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
905 (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
906 * doc/invoke.texi (sccvn-max-scc-size): Remove.
907 (rpo-vn-max-loop-depth): Document.
908 * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
909 when valuezing the VUSE signals we walked out of the region.
910 * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
911 (phi_translate): Set VN context block to use for availability
913 (compute_avail): Likewise.
914 (pre_valueize): New function.
915 (pass_pre::execute): Adjust to the RPO VN API.
917 * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
918 (propagate_constants_for_unrolling): Remove.
919 (tree_unroll_loops_completely): Perform value-numbering
920 on the unrolled bodies loop parent.
922 2018-08-27 Richard Biener <rguenther@suse.de>
924 * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
925 for partial antic compute.
927 2018-08-27 Jakub Jelinek <jakub@redhat.com>
929 PR rtl-optimization/87065
930 * combine.c (simplify_if_then_else): Formatting fix.
931 (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
933 (known_cond): Don't return const_true_rtx for vector modes. Use
934 CONST0_RTX instead of const0_rtx. Formatting fixes.
936 2018-08-27 Martin Liska <mliska@suse.cz>
938 PR gcov-profile/87069
939 * gcov.c (process_file): Record files already processed
940 and warn about a file being processed multiple times.
942 2018-08-27 Martin Liska <mliska@suse.cz>
945 * config/aarch64/aarch64.c (aarch64_override_options_internal):
946 Set default values for x_aarch64_*_string strings.
947 * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
948 prefix. For -mabi do not print '=ABI' in help and use
949 <option_value> format for -msve-vector-bits and -moverride
952 2018-08-26 Jeff Law <law@redhat.com>
954 * config/mips/frame-header-opt.c: Include "backend.h" rather than
957 2018-08-26 Marek Polacek <polacek@redhat.com>
959 PR c++/87029, Implement -Wredundant-move.
960 * doc/invoke.texi: Document -Wredundant-move.
962 2018-08-25 Martin Sebor <msebor@redhat.com>
964 PR tree-optimization/87059
965 * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
966 to the same type as the other.
967 * fold-const.c (fold_binary_loc): Assert expectation.
969 2018-08-25 Iain Sandoe <iain@sandoe.co.uk>
971 * config/darwin.c (machopic_legitimize_pic_address): Clean up
972 extraneous parentheses, dead code section and formatting.
974 2018-08-24 David Malcolm <dmalcolm@redhat.com>
977 * diagnostic-show-locus.c (layout::layout): Ensure the margin is
978 wide enough for jumps in the line-numbering to be visible.
979 (layout::print_gap_in_line_numbering): New member function.
980 (layout::calculate_line_spans): When using line numbering, merge
981 line spans that are only 1 line apart.
982 (diagnostic_show_locus): When printing line numbers, show gaps in
983 line numbering directly, rather than printing headers.
984 (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
985 line-numbering with multiple line spans.
986 (selftest::test_fixit_insert_containing_newline_2): Add test of
987 line-numbering, in which the spans are close enough to be merged.
989 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
991 * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
992 to range_includes_zero_p. Do not special case VR_ANTI_RANGE.
993 * tree-vrp.c (range_is_nonnull): Remove.
994 (range_includes_zero_p): Accept value_range instead of min/max.
995 (extract_range_from_binary_expr_1): Do not early bail on
997 Use range_includes_zero_p instead of range_is_nonnull.
998 (extract_range_from_unary_expr): Use range_includes_zero_p instead
1000 (vrp_meet_1): Pass value_range to range_includes_zero_p. Do not
1001 special case VR_ANTI_RANGE.
1002 (vrp_finalize): Same.
1003 * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
1005 (range_is_nonnull): Remove.
1006 * vr-values.c (vrp_stmt_computes_nonzero): Use
1007 range_includes_zero_p instead of range_is_nonnull.
1008 (extract_range_basic): Pass value_range to range_includes_zero_p
1009 instead of range_is_nonnull.
1011 2018-08-24 Uros Bizjak <ubizjak@gmail.com>
1013 * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
1014 * emit-rtl.h (rtl_data): Remove return_bnd.
1015 * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
1016 * function.c (diddle_return_value): Do not handle crtl->return_bnd.
1017 * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
1018 (POINTER_BOUNDS_MODE): Remove definition.
1019 (make_pointer_bounds_mode): Remove.
1020 (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
1021 * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
1022 (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
1023 * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
1024 * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
1025 * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
1026 * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
1028 * config/i386/i386-modes.def (BND32, BND64): Remove.
1029 * config/i386/i386.c (dbx_register_map): Remove bound registers.
1030 (dbx64_register_map): Ditto.
1031 (svr4_dbx_register_map): Ditto.
1032 (indirect_thunk_bnd_needed): Remove.
1033 (indirect_thunks_bnd_used): Ditto.
1034 (indirect_return_bnd_needed): Ditto.
1035 (indirect_return_via_cx_bnd): Ditto.
1036 (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
1037 (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
1038 (output_indirect_thunk): Ditto. Remove need_prefix argument.
1039 (output_indirect_thunk_function): Remove handling of
1040 indirect_return_bnd_needed, indirect_return_via_cx_bnd,
1041 indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
1042 (ix86_save_reg): Remove handling of crtl->return_bnd.
1043 (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
1044 (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
1045 and UNSPEC_BNDLX_ADDR.
1046 (ix86_output_indirect_branch_via_reg): Remove handling of
1047 indirect_thunk_prefix_bnd.
1048 (ix86_output_indirect_branch_via_push): Ditto.
1049 (ix86_output_function_return): Ditto.
1050 (ix86_output_indirect_function_return): Ditto.
1051 (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
1052 * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
1053 (CALL_USED_REGISTERS): Ditto.
1054 (REG_ALLOC_ORDER): Update for removal of bound registers.
1055 (HI_REGISTER_NAMES): Ditto.
1056 * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
1057 (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
1058 (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
1059 (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
1060 (FIRST_PSEUDO_REG): Update.
1061 (BND): Remove mode iterator.
1062 * config/i386/predicates.md (bnd_mem_operator): Remove.
1064 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
1066 * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
1069 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
1071 * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
1072 the case in which the permute needs only a single element and
1073 repeats for every vector of the result. Extend that case to
1074 handle variable-length vectors.
1075 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
1077 2018-08-24 H.J. Lu <hongjiu.lu@intel.com>
1080 * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
1081 on DW_MACINFO_start_file for -gsplit-dwarf -g3.
1083 2018-08-24 Richard Biener <rguenther@suse.de>
1085 * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
1086 bb_flags_allocated members.
1087 (auto_flag): New RAII class for allocating flags.
1088 (auto_edge_flag): New RAII class for allocating edge flags.
1089 (auto_bb_flag): New RAII class for allocating bb flags.
1090 * cfgloop.c (verify_loop_structure): Allocate temporary edge
1092 * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
1093 in favor of temporarily allocated BB flag.
1094 * hsa-brig.c: Re-order includes.
1095 * hsa-dump.c: Likewise.
1096 * hsa-regalloc.c: Likewise.
1097 * print-rtl.c: Likewise.
1098 * profile-count.c: Likewise.
1100 2018-08-24 Segher Boessenkool <segher@kernel.crashing.org>
1103 * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
1106 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
1109 * wide-int-range.cc (wide_int_range_div): Do not ignore result
1110 from wide_int_range_multiplicative_op.
1112 2018-08-23 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1114 * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
1117 2018-08-23 Giuliano Belinassi <giuliano.belinassi@usp.br>
1119 * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
1122 2018-08-23 Alexander Monakov <amonakov@ispras.ru>
1124 * tree-scalar-evolution.c (final_value_replacement_loop): Dump
1125 full GENERIC expression used for replacement.
1127 2018-08-23 Aldy Hernandez <aldyh@redhat.com>
1129 * tree-vrp.c (abs_extent_range): Remove.
1130 (extract_range_into_wide_ints): Pass wide ints by reference.
1131 (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
1132 Pass wide ints by reference in all calls to
1133 extract_range_into_wide_ints.
1134 * wide-int-range.cc (wide_int_range_div): New.
1135 * wide-int-range.h (wide_int_range_div): New.
1136 (wide_int_range_includes_zero_p): New.
1137 (wide_int_range_zero_p): New.
1139 2018-08-23 Matthew Malcomson <matthew.malcomson@arm.com>
1141 * config/aarch64/aarch64.md (arches): New enum.
1142 (arch): New enum attr.
1143 (arch_enabled): New attr.
1144 (enabled): Now uses arch_enabled only.
1145 (simd, sve, fp16): Removed attribute.
1146 (fp): Attr now defined in terms of 'arch'.
1147 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
1148 *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
1149 <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
1150 <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
1151 attributes into 'arch'.
1152 (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
1153 subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
1154 *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
1155 *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
1156 *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
1157 'simd' attribute into 'arch'.
1158 (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
1159 store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
1160 Convert use of 'fp' attribute to 'arch'.
1161 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
1162 move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
1164 (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
1165 (different modes) Merge 'fp' and 'simd' into 'arch'.
1166 (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
1169 2018-08-23 Segher Boessenkool <segher@kernel.crashing.org>
1171 PR rtl-optimization/87026
1172 * expmed.c (canonicalize_comparison): If we can no longer create
1173 pseudoregisters, don't.
1175 2018-08-23 Richard Earnshaw <rearnsha@arm.com>
1178 * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
1180 * config/arm/arm.c (speculation_barrier_libfunc): New static
1182 (arm_init_libfuncs): Initialize it.
1183 (arm_emit_speculation_barrier): New function.
1184 * config/arm/arm.md (speculation_barrier): Call
1185 arm_emit_speculation_barrier for architectures that do not have
1187 (speculation_barrier_insn): Only match on Armv7 or later.
1189 2018-08-23 Richard Biener <rguenther@suse.de>
1192 * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
1195 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
1197 * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
1198 of single-vector TBLs.
1199 (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
1202 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
1205 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
1206 aarch64_evpc_tbl guard.
1208 2018-08-22 Bernd Edlinger <bernd.edlinger@hotmail.de>
1210 * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
1213 2018-08-22 Martin Sebor <msebor@redhat.com>
1216 * tree-pretty-print.c (pretty_print_string): Add argument.
1217 (dump_generic_node): Call to pretty_print_string with string size.
1219 2018-08-22 Segher Boessenkool <segher@kernel.crashing.org>
1221 PR rtl-optimization/86771
1222 * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
1223 of two SETs into those two SETs, one to be placed at i2, if that SETs
1224 destination is modified between i2 and i3.
1226 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
1228 PR tree-optimization/86725
1229 * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
1231 (vect_analyze_scalar_cycles_1): Check it.
1233 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
1235 PR tree-optimization/86725
1236 * tree-vect-loop.c (vect_is_simple_reduction): When treating
1237 an outer loop phi as a double reduction, make sure that the
1238 single user of the phi result is an inner loop phi.
1240 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
1242 * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
1243 grouped stores with gaps to a strided group.
1245 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
1247 * tree-vect-stmts.c (get_group_load_store_type)
1248 (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
1249 first statement in a group.
1251 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
1253 * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
1254 the sequence used in gcc/gcc.c.
1256 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
1259 * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
1262 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
1264 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
1265 Darwin10-specific unwinder-shim.
1266 * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
1267 * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
1268 New to cater for Darwin10 Rosetta.
1270 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
1272 * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
1273 specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
1275 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
1280 * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
1281 (dwarf2out_switch_text_section): Generate a local label for the
1282 second function sub-section and apply it as the second FDE start
1284 * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
1285 second sub-section start.
1287 2018-08-22 Richard Biener <rguenther@suse.de>
1289 PR tree-optimization/86988
1290 * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
1292 2018-08-22 Richard Biener <rguenther@suse.de>
1294 PR tree-optimization/86945
1295 * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
1297 2018-08-22 Alexandre Oliva <oliva@adacore.com>
1299 * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
1300 a comment about how uses of r2 for .sdata2 come about.
1302 2018-08-22 Alexandre Oliva <aoliva@redhat.com>
1304 * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
1306 2018-08-21 Marek Polacek <polacek@redhat.com>
1308 PR c++/86981, Implement -Wpessimizing-move.
1309 * doc/invoke.texi: Document -Wpessimizing-move.
1311 2018-08-21 Jan Hubicka <jh@suse.cz>
1313 * tree.c (find_decls_types_r): Do not check for redundant typedefs.
1314 * tree.h (is_redundant_typedef): Remove.
1315 * dwarf2out.c (is_redundant_typedef): Turn into static function.
1317 2018-08-21 Jan Hubicka <jh@suse.cz>
1319 * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
1322 2018-08-21 Tamar Christina <tamar.christina@arm.com>
1324 * expmed.c (extract_low_bits): Reject invalid subregs early.
1326 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
1329 * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
1332 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
1334 * config/vxworks.h: Guard vxworks_asm_out_constructor and
1335 vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
1336 * config/vxworks.c: Likewise.
1338 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
1340 * config/vxworks.c: Set targetm.have_ctors_dtors
1341 if HAVE_INITFINI_ARRAY_SUPPORT.
1342 * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
1343 if HAVE_INITFINI_ARRAY_SUPPORT.
1345 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
1347 * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
1348 default search path for VxWorks < 7.
1350 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
1352 * gimple-ssa-sprintf.c (decl_constant_value): Remove.
1353 (get_format_string): Refer to c_getstr.
1355 2018-08-21 Tom de Vries <tdevries@suse.de>
1357 * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
1358 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
1359 (debuginfo_early_stop): Declare.
1360 * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
1361 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
1362 (debuginfo_early_stop): New function.
1363 (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
1364 and debuginfo_early_stop.
1365 * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
1366 * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
1367 (general_init): Call debuginfo_early_init.
1368 (finalize): Call debuginfo_fini.
1369 (do_compile): Call debuginfo_init.
1370 * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
1372 (@item -fdump-debug, @item -fdump-earlydebug): Add.
1374 2018-08-21 Tom de Vries <tdevries@suse.de>
1376 * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
1377 flag_dump_noaddr and flag_dump_unnumbered.
1379 2018-08-21 Aldy Hernandez <aldyh@redhat.com>
1381 * wide-int-range.cc (wide_int_range_abs): New.
1382 (wide_int_range_order_set): Rename from wide_int_range_min_max.
1383 * wide-int-range.h (wide_int_range_abs): New.
1384 (wide_int_range_min_max): New.
1385 * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
1386 case to call wide_int_range_abs.
1387 Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
1388 (extract_range_from_abs_expr): Delete.
1390 2018-08-20 Michael Meissner <meissner@linux.ibm.com>
1393 * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
1394 from 'Y' to 'YZ' to enable the LWAX instruction to be generated
1397 2018-08-20 Nathan Sidwell <nathan@acm.org>
1398 Jeff Law <law@redhat.com>
1400 * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
1401 * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
1403 2018-08-20 David Malcolm <dmalcolm@redhat.com>
1406 * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
1407 (decl_attributes): Likewise.
1408 * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
1410 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
1411 * diagnostic-core.h (class auto_diagnostic_group): New class.
1412 * diagnostic.c (diagnostic_initialize): Initialize the new fields.
1413 (diagnostic_report_diagnostic): Handle the first diagnostics within
1415 (emit_diagnostic): Add auto_diagnostic_group instance.
1417 (inform_n): Likewise.
1418 (warning): Likewise.
1419 (warning_at): Likewise.
1420 (warning_n): Likewise.
1421 (pedwarn): Likewise.
1422 (permerror): Likewise.
1424 (error_n): Likewise.
1425 (error_at): Likewise.
1427 (fatal_error): Likewise.
1428 (internal_error): Likewise.
1429 (internal_error_no_backtrace): Likewise.
1430 (auto_diagnostic_group::auto_diagnostic_group): New ctor.
1431 (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
1432 * diagnostic.h (struct diagnostic_context): Add fields
1433 "diagnostic_group_nesting_depth",
1434 "diagnostic_group_emission_count", "begin_group_cb",
1436 * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
1437 Add auto_diagnostic_group instance(s).
1438 (find_explicit_erroneous_behavior): Likewise.
1439 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
1440 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
1441 * gimplify.c (warn_implicit_fallthrough_r): Likewise.
1442 (gimplify_va_arg_expr): Likewise.
1443 * hsa-gen.c (HSA_SORRY_ATV): Likewise.
1444 (HSA_SORRY_AT): Likewise.
1445 * ipa-devirt.c (compare_virtual_tables): Likewise.
1446 (warn_odr): Likewise.
1447 * multiple_target.c (expand_target_clones): Likewise.
1448 * opts-common.c (cmdline_handle_error): Likewise.
1449 * reginfo.c (globalize_reg): Likewise.
1450 * substring-locations.c (format_warning_n_va): Likewise.
1451 * tree-inline.c (expand_call_inline): Likewise.
1452 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
1453 * tree-ssa-loop-niter.c
1454 (do_warn_aggressive_loop_optimizations): Likewise.
1455 * tree-ssa-uninit.c (warn_uninit): Likewise.
1456 * tree.c (warn_deprecated_use): Likewise.
1458 2018-08-20 H.J. Lu <hongjiu.lu@intel.com>
1461 * config/i386/i386.md (eh_return): Always update EH return
1462 address in word_mode.
1464 2018-08-20 Chung-Lin Tang <cltang@codesourcery.com>
1466 * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
1467 TARGET_SPLIT_COMPLEX_ARG is defined.
1469 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
1471 * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
1473 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
1476 * expr.c (expand_assignment): Assert that bitpos is positive.
1477 (store_field): Likewise
1478 (expand_expr_real_1): Make sure that bitpos is positive.
1479 * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
1482 2018-08-20 Nathan Sidwell <nathan@acm.org>
1484 * Makefile.in (CPP_ID_DATA_H): Delete.
1485 (CPP_INTERNAL_H): Don't add it.
1486 (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
1487 * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
1489 2018-08-20 Richard Biener <rguenther@suse.de>
1491 PR tree-optimization/78655
1492 * tree-vrp.c (extract_range_from_binary_expr_1): Make
1493 pointer + offset nonnull if either operand is nonnull work.
1495 2018-08-20 Tom de Vries <tdevries@suse.de>
1497 * dwarf2out.c (add_scalar_info): Don't add reference to existing die
1498 unless the referenced die describes the added property using
1499 DW_AT_location or DW_AT_const_value. Fall back to exprloc case.
1500 Otherwise, add a DW_AT_location to the referenced die.
1502 2018-08-19 Uros Bizjak <ubizjak@gmail.com>
1505 * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
1506 register_operand when calling ix86_set_reg_reg_cost.
1507 [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
1508 Set *total to 0 for operands that satisfy x86_64_immediate_operand
1509 predicate and to 1 otherwise.
1511 2018-08-18 Iain Sandoe <iain@sandoe.co.uk>
1513 * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
1514 emit a diagnostic that it is not supported and reset the option.
1515 * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
1516 supported and consume the option. (ASM_FINAL_SPEC): New.
1518 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
1520 * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
1523 2018-08-17 Sandra Loosemore <sandra@codesourcery.com>
1525 C-SKY port: Documentation
1527 * doc/extend.texi (C-SKY Function Attributes): New section.
1528 * doc/invoke.texi (Option Summary): Add C-SKY options.
1529 (C-SKY Options): New section.
1530 * doc/md.texi (Machine Constraints): Document C-SKY constraints.
1532 2018-08-17 Jojo <jijie_rong@c-sky.com>
1533 Huibin Wang <huibin_wang@c-sky.com>
1534 Sandra Loosemore <sandra@codesourcery.com>
1535 Chung-Lin Tang <cltang@codesourcery.com>
1537 C-SKY port: Backend implementation
1539 * config/csky/*: New.
1540 * common/config/csky/*: New.
1542 2018-08-17 Jojo <jijie_rong@c-sky.com>
1543 Huibin Wang <huibin_wang@c-sky.com>
1544 Sandra Loosemore <sandra@codesourcery.com>
1545 Chung-Lin Tang <cltang@codesourcery.com>
1546 Andrew Jenner <andrew@codesourcery.com>
1548 C-SKY port: Configury
1550 * config.gcc (csky-*-*): New.
1551 * configure.ac: Add csky to targets for dwarf2 debug_line support.
1552 * configure: Regenerated.
1554 2018-08-17 David Malcolm <dmalcolm@redhat.com>
1556 * dump-context.h: Include "dumpfile.h".
1557 (dump_context::dump_printf_va): Convert final param from va_list
1558 to va_list *. Convert from ATTRIBUTE_PRINTF to
1559 ATTRIBUTE_GCC_DUMP_PRINTF.
1560 (dump_context::dump_printf_loc_va): Likewise.
1561 * dumpfile.c: Include "stringpool.h".
1562 (make_item_for_dump_printf_va): Delete.
1563 (make_item_for_dump_printf): Delete.
1564 (class dump_pretty_printer): New class.
1565 (dump_pretty_printer::dump_pretty_printer): New ctor.
1566 (dump_pretty_printer::emit_items): New member function.
1567 (dump_pretty_printer::emit_any_pending_textual_chunks): New member
1569 (dump_pretty_printer::emit_item): New member function.
1570 (dump_pretty_printer::stash_item): New member function.
1571 (dump_pretty_printer::format_decoder_cb): New member function.
1572 (dump_pretty_printer::decode_format): New member function.
1573 (dump_context::dump_printf_va): Reimplement in terms of
1574 dump_pretty_printer.
1575 (dump_context::dump_printf_loc_va): Convert final param from va_list
1577 (dump_context::begin_scope): Reimplement call to
1578 make_item_for_dump_printf.
1579 (dump_printf): Update for change to dump_printf_va.
1580 (dump_printf_loc): Likewise.
1581 (selftest::test_capture_of_dump_calls): Convert "stmt" from
1582 greturn * to gimple *. Add a test_decl. Add tests of dump_printf
1583 with %T, %E, and %G.
1584 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
1585 (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
1586 ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
1587 (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
1588 ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
1589 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
1590 use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
1591 within a dump_printf_loc call to "%wu".
1592 (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
1593 converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd". Add a
1594 missing space after "=".
1595 * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
1596 call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
1597 * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
1598 convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
1599 * tree-vectorizer.c (try_vectorize_loop_1): Likewise. Remove
1600 duplicate "vectorized" from message.
1602 2018-08-17 Szabolcs Nagy <szabolcs.nagy@arm.com>
1604 * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
1605 polyNxK_t element's TYPE_STRING_FLAG.
1607 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
1609 * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
1610 (they were unnamed before). Fix comments.
1612 2018-08-17 Nathan Sidwell <nathan@acm.org>
1614 * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
1616 2018-08-17 Richard Biener <rguenther@suse.de>
1618 PR tree-optimization/86841
1619 * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
1621 2018-08-17 Martin Liska <mliska@suse.cz>
1623 * common.opt: Remove Warn, Init and Report for options with
1624 Ignore/Deprecated flag. Warning is done automatically for
1626 * config/i386/i386.opt: Likewise.
1627 * config/ia64/ia64.opt: Likewise.
1628 * config/rs6000/rs6000.opt: Likewise.
1629 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
1630 Remove usage of flag_check_pointer_bounds.
1631 * lto-wrapper.c (merge_and_complain): Do not handle
1632 OPT_fcheck_pointer_bounds.
1633 (append_compiler_options): Likewise.
1634 * opt-functions.awk: Do not handle Deprecated.
1635 * optc-gen.awk: Check that Var, Report and Init are not
1636 used for an option with Ignore/Deprecated flag.
1637 * opts-common.c (decode_cmdline_option): Do not report
1639 (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
1641 * opts.h (struct cl_option): Remove cl_deprecated flag.
1642 (CL_ERR_DEPRECATED): Remove error enum value.
1644 2018-08-17 Richard Biener <rguenther@suse.de>
1647 * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
1648 across a va-arg-pack using call adjust its return value accordingly.
1650 2018-08-16 Martin Sebor <msebor@redhat.com>
1652 PR tree-optimization/86853
1653 * gimple-ssa-sprintf.c (struct format_result): Rename member.
1654 (struct fmtresult): Add member and initialize it in ctors.
1655 (format_character): Handle %C. Extend range to NUL. Set MAYFAIL.
1656 (format_string): Handle %S the same as %ls. Set MAYFAIL.
1657 (format_directive): Set POSUNDER4K when MAYFAIL is set.
1658 (parse_directive): Handle %C same as %c.
1659 (sprintf_dom_walker::compute_format_length): Adjust.
1660 (is_call_safe): Adjust.
1662 2018-08-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
1664 * builtins.c (c_strlen): Add new parameter eltsize. Use it
1665 for determining how to count the elements.
1666 * builtins.h (c_strlen): Adjust prototype.
1667 * expr.c (string_constant): Add new parameter mem_size.
1668 Set *mem_size appropriately.
1669 * expr.h (string_constant): Adjust protoype.
1670 * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
1671 * gimple-fold.h (get_range_strlen): Adjust prototype.
1672 * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
1673 (format_string): Call get_string_length with eltsize.
1675 2018-08-16 David Malcolm <dmalcolm@redhat.com>
1677 * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
1678 to emit the span, rather than setting it as the prefix.
1680 2018-08-16 David Malcolm <dmalcolm@redhat.com>
1682 * diagnostic-show-locus.c (layout::start_annotation_line): Add
1683 "margin_char" parameter, defaulting to space. Use it in place
1684 of pp_space for the initial part of the margin.
1685 (layout::print_leading_fixits): Use '+' when filling the margin
1686 of line-insertion fix-it hints.
1688 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
1690 * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
1693 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
1695 * config/rs6000/altivec.md: Don't set length attribute to the default
1697 * config/rs6000/darwin.md: Ditto.
1698 * config/rs6000/dfp.md: Ditto.
1699 * config/rs6000/htm.md: Ditto.
1700 * config/rs6000/rs6000.md: Ditto.
1701 * config/rs6000/sync.md: Ditto.
1702 * config/rs6000/vsx.md: Ditto.
1704 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
1706 * config/rs6000/altivec.md: Don't set length attribute to the default
1707 value, for branch instructions.
1708 * config/rs6000/darwin.md: Ditto.
1709 * config/rs6000/rs6000.md: Ditto.
1711 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
1713 * config/rs6000/rs6000.md (length): Always define as const_int 4.
1714 (unnamed conditional branch define_insn): Set length to 4 or 8
1715 depending on offset.
1716 (<bd>_<mode>): Similar, for alternative 0.
1717 (<bd>tf_<mode>): Ditto.
1719 2018-08-16 Tamar Christina <tamar.christina@arm.com>
1721 * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
1723 2018-08-16 Matthew Malcomson <matthew.malcomson@arm.com>
1725 * doc/rtl.texi: Replace old RTX class names with new names.
1728 2018-08-16 Vlad Lazar <vlad.lazar@arm.com>
1730 * expmed.h (canonicalize_comparison): New declaration.
1731 * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
1732 * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
1733 * optabs.c (prepare_cmp_insn): Likewise.
1734 * rtl.h (unsigned_condition_p): New function which checks if a
1735 comparison operator is unsigned.
1737 2018-08-16 Nathan Sidwell <nathan@acm.org>
1739 * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
1740 * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
1742 2018-08-16 Tamar Christina <tamar.christina@arm.com>
1745 * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
1746 * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
1747 (mov<mov>): ..this and enable unconditionally.
1749 2018-08-16 Tamar Christina <tamar.christina@arm.com>
1751 * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
1753 2018-08-16 Sam Tebbs <sam.tebbs@arm.com>
1755 * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
1756 (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
1757 "Common" with "Target".
1759 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
1761 * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
1762 * doc/invoke.texi (mmitigate-rop): Remove.
1763 * config/i386/i386.c: Do not include "regrename.h".
1764 (ix86_rop_should_change_byte_p, reg_encoded_number)
1765 (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
1767 (ix86_reorg): Remove call to ix86_mitigate_rop.
1768 * config/i386/i386.md (attr "modrm_class"): Remove.
1769 (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
1770 (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
1771 (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
1773 2018-08-15 Will Schmidt <will_schmidt@vnet.ibm.com>
1775 * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
1776 allow folding of mergeh() and mergel() for the float and double types.
1777 (fold_mergehl_helper): Rework to handle building a permute tree
1780 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
1782 * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
1785 2018-08-15 David Malcolm <dmalcolm@redhat.com>
1787 * common.opt (fdiagnostics-show-labels): New option.
1788 * diagnostic-show-locus.c (class layout_range): Add field
1790 (class layout): Add field "m_show_labels_p".
1791 (layout_range::layout_range): Add param "label" and use it to
1793 (make_range): Pass in NULL for new "label" param of layout_range's
1795 (layout::layout): Initialize m_show_labels_p.
1796 (layout::maybe_add_location_range): Pass in loc_range->m_label
1797 when constructing layout_range instances.
1798 (struct line_label): New struct.
1799 (layout::print_any_labels): New member function.
1800 (layout::print_line): Call it if label-printing is enabled.
1801 (selftest::test_one_liner_labels): New test.
1802 (selftest::test_diagnostic_show_locus_one_liner): Call it.
1803 * diagnostic.c (diagnostic_initialize): Initialize
1804 context->show_labels_p.
1805 * diagnostic.h (struct diagnostic_context): Add field
1807 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
1808 -fno-diagnostics-show-labels.
1809 * dwarf2out.c (gen_producer_string): Add
1810 OPT_fdiagnostics_show_labels to the ignored options.
1811 * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
1813 (gcc_rich_location::maybe_add_expr): Likewise.
1814 * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
1815 label" param, defaulting to NULL.
1816 (gcc_rich_location::add_expr): Add "label" param.
1817 (gcc_rich_location::maybe_add_expr): Likewise.
1818 (class text_range_label): New class.
1819 (class range_label_for_type_mismatch): New class.
1820 * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
1821 of format_warning_va.
1822 (fmtwarn_n): Likewise for new params of format_warning_n_va.
1823 * lto-wrapper.c (merge_and_complain): Add
1824 OPT_fdiagnostics_show_labels to the "pick one setting" options.
1825 (append_compiler_options): Likewise to the dropped options.
1826 (append_diag_options): Likewise to the passed-on options.
1827 * opts.c (common_handle_option): Handle the new option.
1828 * selftest-diagnostic.c
1829 (test_diagnostic_context::test_diagnostic_context): Enable
1831 * substring-locations.c: Include "gcc-rich-location.h".
1832 (format_warning_n_va): Add "fmt_label" and "param_label" params
1833 and use them as appropriate.
1834 (format_warning_va): Add "fmt_label" and "param_label" params,
1835 passing them on to format_warning_n_va.
1836 (format_warning_at_substring): Likewise.
1837 (format_warning_at_substring_n): Likewise.
1838 * substring-locations.h (format_warning_va): Add "fmt_label" and
1839 "param_label" params.
1840 (format_warning_n_va): Likewise.
1841 (format_warning_at_substring): Likewise.
1842 (format_warning_at_substring_n): Likewise.
1843 * toplev.c (general_init): Initialize global_dc->show_labels_p.
1845 2018-08-15 Qing Zhao <qing.zhao@oracle.com>
1848 * builtins.c (expand_builtin_memcmp): Do not expand the call
1849 when overflow is detected.
1851 2018-08-15 Martin Sebor <msebor@redhat.com>
1853 PR tree-optimization/71625
1854 * config/aarch64/aarch64-builtins.c
1855 (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
1857 2018-08-15 Ilya Leoshkevich <iii@linux.ibm.com>
1859 * config/s390/s390.c (s390_reorg): Remove loop.
1861 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
1864 (darwin_function_switched_text_sections): Delete.
1865 * gcc/config/darwin.h
1866 (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
1868 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
1871 * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
1872 DEBUG_RNGLISTS_SECTION) new macros. (DEBUG_PUBNAMES_SECTION,
1873 DEBUG_PUBTYPES_SECTION) update to include GNU variant.
1875 2018-08-15 Martin Liska <mliska@suse.cz>
1877 PR tree-optimization/86925
1878 * predict.c (expr_expected_value_1): When taking
1879 later predictor, assign also probability.
1880 Use fold_build2_initializer_loc in order to fold
1881 the expression in -frounding-math.
1883 2018-08-14 Allan Sandfeld Jensen <allan.jensen@qt.io>
1885 * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
1887 (expand_vec_perm_1): Try the new method.
1889 2018-08-14 Ilya Leoshkevich <iii@linux.ibm.com>
1892 * lra-lives.c (remove_some_program_points_and_update_live_ranges):
1893 Check whether lra_live_max_point is 0 before dividing.
1895 2018-08-14 Martin Sebor <msebor@redhat.com>
1897 PR tree-optimization/86650
1898 * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
1899 (vrp_prop::check_mem_ref): Same.
1901 2018-08-13 Liu Hao <lh_mouse@126.com>
1903 * pretty-print.c (eat_esc_sequence): Swap the foreground and
1904 background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
1905 and clear it thereafter, as it only works for DBCS.
1907 2018-08-13 Liu Hao <lh_mouse@126.com>
1909 * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
1910 handle returned by _get_osf_handle().
1912 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
1914 * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
1915 for folding vec_perm.
1917 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
1919 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
1920 Add support for gimple-folding of vec_pack() and vec_unpack()
1923 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
1925 * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
1926 vec_xst variants to the list.
1927 (rs6000_gimple_fold_builtin): Add support for folding unaligned
1928 vector loads and stores.
1930 2018-08-13 David Edelsohn <dje.gcc@gmail.com>
1932 * config.gcc (rs6000-ibm-aix4.x): Delete.
1933 (rs6000-ibm-aix5.1): Delete.
1934 (rs6000-ibm-aix5.2): Delete.
1935 (rs6000-ibm-aix5.3): Delete.
1936 * config/rs6000/aix43.h: Delete.
1937 * config/rs6000/aix51.h: Delete.
1938 * config/rs6000/aix52.h: Delete.
1939 * config/rs6000/t-aix43: Delete.
1941 2018-08-13 Ilya Leoshkevich <iii@linux.ibm.com>
1943 * config/s390/s390.c (s390_decompose_constant_pool_ref):
1945 (s390_decompose_address): Factor out constant pool ref
1948 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
1950 * config/nds32/nds32-predicates.c
1951 (nds32_can_use_bclr_p): Change return type as bool.
1952 (nds32_can_use_bset_p): Ditto.
1953 (nds32_can_use_btgl_p): Ditto.
1954 (nds32_can_use_bitci_p): Ditto.
1955 * config/nds32/nds32-protos.h
1956 (nds32_can_use_bclr_p): Change declaration.
1957 (nds32_can_use_bset_p): Ditto.
1958 (nds32_can_use_btgl_p): Ditto.
1959 (nds32_can_use_bitci_p): Ditto.
1961 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
1963 * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
1964 Support -msched-prolog-epilog option.
1965 * config/nds32/nds32.opt (msched-prolog-epilog): New option.
1967 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
1969 * common/config/nds32/nds32-common.c
1970 (nds32_option_optimization_table): Enalbe -malways-align.
1972 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
1974 * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
1976 * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
1977 OPT_misr_secure_ case.
1978 * config/nds32/nds32-isr.c: Implementation of backward compatibility.
1979 * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
1980 * config/nds32/nds32.c (nds32_attribute_table): Add critical and
1982 * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
1983 (nds32_isr_info): New field security_level.
1984 (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
1985 * config/nds32/nds32.md (return_internal): Consider critical attribute.
1986 * config/nds32/nds32.opt (misr-secure): New option.
1987 * config/nds32/nds32_init.inc: New file.
1988 * config/nds32/nds32_isr.h: New file.
1990 2018-08-11 John David Anglin <danglin@gcc.gnu.org>
1992 * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
1993 Update comment for atomic instructions.
1994 (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
1995 atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
1997 (atomic_loaddi): Revise fence expansion to only emit fence prior to
1998 load for __ATOMIC_SEQ_CST model.
1999 (atomic_loaddi_1): Remove float register target.
2000 (atomic_storedi): Handle CONST_INT values.
2001 (atomic_storedi_1): Remove float register source. Add special case
2003 (memory_barrier): New expander and insn.
2005 2018-08-11 Jakub Jelinek <jakub@redhat.com>
2007 PR tree-optimization/86835
2008 * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
2009 new_stmt after def_gsi, make sure to insert new_square_stmt after
2010 that stmt, not 2 stmts before it.
2012 2018-08-10 Alexander Monakov <amonakov@ispras.ru>
2015 * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
2018 2018-08-10 Martin Liska <mliska@suse.cz>
2021 * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
2023 * builtins.c (expand_builtin_expect_with_probability):
2025 (expand_builtin_expect_with_probability): New function.
2026 (build_builtin_expect_predicate): Add new argumnet probability
2027 for BUILT_IN_EXPECT_WITH_PROBABILITY.
2028 (fold_builtin_expect):
2031 * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
2032 * builtins.h (fold_builtin_expect): Set new argument.
2033 * doc/extend.texi: Document __builtin_expect_with_probability.
2034 * doc/invoke.texi: Likewise.
2035 * gimple-fold.c (gimple_fold_call): Pass new argument.
2036 * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
2037 also BUILT_IN_EXPECT_WITH_PROBABILITY.
2038 * predict.c (get_predictor_value): New function.
2039 (expr_expected_value): Add new argument probability. Assume
2040 that predictor and probability are always non-null.
2041 (expr_expected_value_1): Likewise. For __builtin_expect and
2042 __builtin_expect_with_probability set probability. Handle
2043 combination in binary expressions.
2044 (tree_predict_by_opcode): Simplify code by simply calling
2045 get_predictor_value.
2046 (pass_strip_predict_hints::execute): Add handling of
2047 BUILT_IN_EXPECT_WITH_PROBABILITY.
2048 * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
2050 * tree.h (DECL_BUILT_IN_P): New function.
2052 2018-08-10 Martin Liska <mliska@suse.cz>
2054 PR tree-optimization/85799
2055 * passes.def: Add argument for pass_strip_predict_hints.
2056 * predict.c (class pass_strip_predict_hints): Add new argument
2058 (strip_predictor_early): New function.
2059 (pass_strip_predict_hints::execute): Call the function to
2061 (strip_predict_hints): New function.
2062 * predict.def: Fix comment.
2064 2018-08-10 Thomas Preud'homme <thomas.preudhomme@linaro.org>
2066 * Makefile.in: Clarify which tm.texi to copy over to assert the
2067 right to grant a GFDL license for all.
2069 2018-08-09 Jeff Law <law@redhat.com>
2071 * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
2074 2018-08-09 Andreas Schwab <schwab@linux-m68k.org>
2076 * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
2079 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
2081 * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
2082 reductions for variable-length vectors.
2084 2018-08-09 David Malcolm <dmalcolm@redhat.com>
2087 * common.opt (fdiagnostics-show-line-numbers): New option.
2088 * diagnostic-show-locus.c (class layout): Add fields
2089 "m_show_line_numbers_p" and "m_linenum_width";
2090 (num_digits): New function.
2091 (test_num_digits): New function.
2092 (layout::layout): Initialize new fields. Update m_x_offset
2093 logic to handle any left margin.
2094 (layout::print_source_line): Print line number when requested.
2095 (layout::start_annotation_line): New member function.
2096 (layout::print_annotation_line): Call it.
2097 (layout::print_leading_fixits): Likewise.
2098 (layout::print_trailing_fixits): Likewise. Update calls to
2099 move_to_column for new parameter.
2100 (layout::get_x_bound_for_row): Add "add_left_margin" param and use
2101 it to potentially call start_annotation_line.
2102 (layout::show_ruler): Call start_annotation_line.
2103 (selftest::test_line_numbers_multiline_range): New selftest.
2104 (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
2105 and selftest::test_line_numbers_multiline_range.
2106 * diagnostic.c (diagnostic_initialize): Initialize
2107 show_line_numbers_p.
2108 * diagnostic.h (struct diagnostic_context): Add field
2109 "show_line_numbers_p".
2110 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
2111 -fno-diagnostics-show-line-numbers.
2112 * dwarf2out.c (gen_producer_string): Add
2113 OPT_fdiagnostics_show_line_numbers to the ignored options.
2114 * lto-wrapper.c (merge_and_complain): Likewise to the "pick
2115 one setting" options.
2116 (append_compiler_options): Likewise to the dropped options.
2117 (append_diag_options): Likewise to the passed-on options.
2118 * opts.c (common_handle_option): Handle the new option.
2119 * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
2121 2018-08-09 Kelvin Nilsen <kelvin@gcc.gnu.org>
2123 * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
2124 ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub. Add
2125 third argument of type "const signed char" to descriptions of
2126 __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
2127 __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
2128 __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
2129 __builtin_bcdsub_ov functions.
2131 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
2133 PR tree-optimization/86858
2134 * tree-vect-loop.c (vect_is_simple_reduction): Restore
2135 flow_bb_inside_loop_p calls.
2137 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
2139 PR tree-optimization/86871
2140 * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
2141 instead of gimple_assign_lhs.
2143 2018-08-09 Richard Earnshaw <rearnsha@arm.com>
2146 * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
2147 register constraint to operand 0.
2148 (add<mode>3_carryinC): Likewise.
2149 (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
2151 2018-08-09 Martin Liska <mliska@suse.cz>
2154 * common.opt: Remove extra line.
2156 2018-08-09 Martin Liska <mliska@suse.cz>
2158 * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
2159 at the end of a line, make first letter capital and end up
2160 a sentence with a dot.
2161 (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
2162 (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
2163 (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
2164 (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
2165 (PARAM_MAX_ISL_OPERATIONS): Likewise.
2166 (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
2167 (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
2168 (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
2169 (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
2170 (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
2171 (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
2172 (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
2173 (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
2174 (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
2175 (PARAM_TREE_REASSOC_WIDTH): Likewise.
2176 (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
2177 (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
2178 (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
2180 2018-08-09 Andreas Krebbel <krebbel@linux.ibm.com>
2183 * config/s390/s390.c (s390_option_override_internal): Reduce the
2184 stack-clash-protection-probe-interval param if it would be too big
2187 2018-08-08 Andreas Schwab <schwab@linux-m68k.org>
2190 * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
2191 * config/m68k/m68k.c (handle_move_double): Don't call
2192 m68k_final_prescan_insn.
2193 (m68k_adjust_decorated_operand): Renamed from
2194 m68k_final_prescan_insn, remove first and third operand and
2196 (print_operand): Call it.
2197 (print_operand_address): Call it.
2199 2018-08-08 Nathan Sidwell <nathan@acm.org>
2201 * diagnostic.c (diagnostic_report_current_module): Use
2202 linemap_included_from & linemap_included_from_linemap.
2204 2018-08-08 Hongbo Zhang <hongbo.zhang@linaro.org>
2206 * config/aarch64/aarch64-cores.def: Add phecda core.
2207 * config/aarch64/aarch64-tune.md: Regenerate.
2208 * doc/invoke.texi: Add phecda core.
2210 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
2213 * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
2215 * config/s390/s390.md ("movti"): Add more alternatives for
2216 constant to GPR copies.
2218 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
2220 * config/s390/s390.c: Fix whitespace damage throughout the file.
2221 * config/s390/s390.h: Likewise.
2222 * config/s390/tpf.h: Likewise.
2224 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
2226 * config/s390/s390.c (s390_loadrelative_operand_p):
2227 Remove TARGET_CPU_ZARCH usages.
2228 (s390_rtx_costs): Likewise.
2229 (s390_legitimate_constant_p): Likewise.
2230 (s390_cannot_force_const_mem): Likewise.
2231 (legitimate_reload_constant_p): Likewise.
2232 (s390_preferred_reload_class): Likewise.
2233 (legitimize_pic_address): Likewise.
2234 (legitimize_tls_address): Likewise.
2235 (s390_split_branches): Removed.
2236 (s390_add_execute): Removed.
2237 (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
2238 (s390_mainpool_start): Likewise.
2239 (s390_mainpool_finish): Likewise.
2240 (s390_mainpool_cancel): Removed.
2241 (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
2242 (s390_chunkify_cancel): Likewise.
2243 (s390_return_addr_rtx): Likewise.
2244 (s390_register_info): Remove split_branches_pending_p uages.
2245 (s390_optimize_register_info): Likewise.
2246 (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
2247 split_branches_pending_p usages.
2248 (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
2249 (s390_load_got): Likewise.
2250 (s390_expand_split_stack_prologue): Likewise.
2251 (output_asm_nops): Likewise.
2252 (s390_function_profiler): Likewise.
2253 (s390_emit_call): Likewise.
2254 (s390_conditional_register_usage): Likewise.
2255 (s390_optimize_prologue): Likewise.
2256 (s390_reorg): Remove TARGET_CPU_ZARCH and
2257 split_branches_pending_p usages.
2258 (s390_option_override_internal): Remove TARGET_CPU_ZARCH
2260 (s390_output_indirect_thunk_function): Likewise.
2261 * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
2262 (TARGET_CPU_ZARCH_P): Removed.
2263 (struct machine_function): Remove split_branches_pending_p.
2264 * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
2266 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
2268 * common/config/s390/s390-common.c (processor_flags_table):
2270 * config.gcc: Remove with_arch/with_tune support.
2271 * config/s390/2064.md: Remove cpu attribute comparisons.
2272 * config/s390/driver-native.c (s390_host_detect_local_cpu):
2274 * config/s390/linux.h (ASM_SPEC):
2275 Remove -march support.
2276 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
2277 Use a table to get an arch level.
2278 * config/s390/s390-opts.h (enum processor_type):
2280 * config/s390/s390.c
2281 (processor_table): Remove entries, add arch_level values.
2282 (s390_issue_rate): Remove cases.
2283 (s390_option_override): Adjust
2284 s390_option_override_internal() call.
2285 (s390_option_override_internal): Remove deprecation warning.
2286 (s390_valid_target_attribute_tree): Adjust
2287 s390_option_override_internal() call.
2288 * config/s390/s390.h (struct s390_processor):
2289 Share with s390-c.c, add arch_level field.
2290 * config/s390/s390.md:
2291 Remove occurrences in cpu attribute.
2292 * config/s390/s390.opt: Remove -march/-mtune support.
2293 * config/s390/tpf.h (ASM_SPEC): Remove -march support.
2294 * doc/invoke.texi: Remove deprecation warning.
2296 2018-08-08 Luis Machado <luis.machado@linaro.org>
2298 * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
2300 (qdf24xx_tunings): Set vector cost structure to
2301 qdf24xx_vector_cost.
2303 * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
2304 <register_sextend>: Set to 3.
2306 2018-08-07 Richard Sandiford <richard.sandiford@arm.com>
2309 * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
2310 * config/aarch64/aarch64-simd.md
2311 (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
2312 (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
2313 (aarch64_frecpx<mode>): ...this new pattern.
2314 * config/aarch64/aarch64-simd-builtins.def: Remove comment
2315 about aarch64_frecp<FRECP:frecp_suffix><mode>.
2317 2018-08-07 Martin Liska <mliska@suse.cz>
2320 * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
2321 BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
2322 * predict.def (PRED_MALLOC_NONNULL): New predictor.
2323 * doc/extend.texi: Document that malloc attribute adds
2326 2018-08-06 John David Anglin <danglin@gcc.gnu.org>
2329 * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2330 Define to speculation_safe_value_not_needed.
2332 2018-08-06 Jeff Law <law@redhat.com>
2334 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
2335 the vr_values instance to cprop_into_stmt.
2336 (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
2337 (cprop_operand): Also query EVRP to determine if OP is a constant.
2339 2018-08-06 Nathan Sidwell <nathan@acm.org>
2341 * diagnostic.c (diagnostic_report_current_module): Reroll
2342 included-at loop. Translate text.
2344 2018-08-06 David Malcolm <dmalcolm@redhat.com>
2346 * function-tests.c (selftest::test_expansion_to_rtl): Call
2347 free_after_compilation.
2349 2018-08-06 Alan Hayward <alan.hayward@arm.com>
2351 * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
2353 2018-08-06 Andreas Krebbel <krebbel@linux.ibm.com>
2355 * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
2356 loops with memory block operations from getting unrolled.
2358 2018-08-06 Ulrich Weigand <uweigand@de.ibm.com>
2361 * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2362 Define to speculation_safe_value_not_needed.
2364 2018-08-06 Jeff Law <law@redhat.com>
2366 * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
2369 2018-08-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2372 * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
2373 with all enabled __intN types.
2375 * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
2377 2018-08-06 Alan Hayward <alan.hayward@arm.com>
2379 * alias.c (record_set): Check for clobber high.
2380 * cfgexpand.c (expand_gimple_stmt): Likewise.
2381 * combine-stack-adj.c (single_set_for_csa): Likewise.
2382 * combine.c (find_single_use_1): Likewise.
2383 (set_nonzero_bits_and_sign_copies): Likewise.
2384 (get_combine_src_dest): Likewise.
2385 (is_parallel_of_n_reg_sets): Likewise.
2386 (try_combine): Likewise.
2387 (record_dead_and_set_regs_1): Likewise.
2388 (reg_dead_at_p_1): Likewise.
2389 (reg_dead_at_p): Likewise.
2390 * dce.c (deletable_insn_p): Likewise.
2391 (mark_nonreg_stores_1): Likewise.
2392 (mark_nonreg_stores_2): Likewise.
2393 * df-scan.c (df_find_hard_reg_defs): Likewise.
2394 (df_uses_record): Likewise.
2395 (df_get_call_refs): Likewise.
2396 * dwarf2out.c (mem_loc_descriptor): Likewise.
2397 * haifa-sched.c (haifa_classify_rtx): Likewise.
2398 * ira-build.c (create_insn_allocnos): Likewise.
2399 * ira-costs.c (scan_one_insn): Likewise.
2400 * ira.c (equiv_init_movable_p): Likewise.
2401 (rtx_moveable_p): Likewise.
2402 (interesting_dest_for_shprep): Likewise.
2403 * jump.c (mark_jump_label_1): Likewise.
2404 * postreload-gcse.c (record_opr_changes): Likewise.
2405 * postreload.c (reload_cse_simplify): Likewise.
2406 (struct reg_use): Add source expr.
2407 (reload_combine): Check for clobber high.
2408 (reload_combine_note_use): Likewise.
2409 (reload_cse_move2add): Likewise.
2410 (move2add_note_store): Likewise.
2411 * print-rtl.c (print_pattern): Likewise.
2412 * recog.c (decode_asm_operands): Likewise.
2413 (store_data_bypass_p): Likewise.
2414 (if_test_bypass_p): Likewise.
2415 * regcprop.c (kill_clobbered_value): Likewise.
2416 (kill_set_value): Likewise.
2417 * reginfo.c (reg_scan_mark_refs): Likewise.
2418 * reload1.c (maybe_fix_stack_asms): Likewise.
2419 (eliminate_regs_1): Likewise.
2420 (elimination_effects): Likewise.
2421 (mark_not_eliminable): Likewise.
2422 (scan_paradoxical_subregs): Likewise.
2423 (forget_old_reloads_1): Likewise.
2424 * reorg.c (find_end_label): Likewise.
2425 (try_merge_delay_insns): Likewise.
2426 (redundant_insn): Likewise.
2427 (own_thread_p): Likewise.
2428 (fill_simple_delay_slots): Likewise.
2429 (fill_slots_from_thread): Likewise.
2430 (dbr_schedule): Likewise.
2431 * resource.c (update_live_status): Likewise.
2432 (mark_referenced_resources): Likewise.
2433 (mark_set_resources): Likewise.
2434 * rtl.c (copy_rtx): Likewise.
2435 * rtlanal.c (reg_referenced_p): Likewise.
2436 (single_set_2): Likewise.
2437 (noop_move_p): Likewise.
2438 (note_stores): Likewise.
2439 * sched-deps.c (sched_analyze_reg): Likewise.
2440 (sched_analyze_insn): Likewise.
2442 2018-08-06 Alan Hayward <alan.hayward@arm.com>
2444 * cse.c (invalidate_reg): New function extracted from...
2445 (invalidate): ...here.
2446 (canonicalize_insn): Check for clobber high.
2447 (invalidate_from_clobbers): invalidate clobber highs.
2448 (invalidate_from_sets_and_clobbers): Likewise.
2449 (count_reg_usage): Check for clobber high.
2450 (insn_live_p): Likewise.
2451 * cselib.c (cselib_expand_value_rtx_1):Likewise.
2452 (cselib_invalidate_regno): Check for clobber in setter.
2453 (cselib_invalidate_rtx): Pass through setter.
2454 (cselib_invalidate_rtx_note_stores):
2455 (cselib_process_insn): Check for clobber high.
2456 * cselib.h (cselib_invalidate_rtx): Add operand.
2458 2018-08-06 Alan Hayward <alan.hayward@arm.com>
2460 * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
2461 (mark_not_eliminable): Likewise.
2462 * lra-int.h (struct lra_insn_reg): Add clobber high marker.
2463 * lra-lives.c (process_bb_lives): Check for clobber high.
2464 * lra.c (new_insn_reg): Remember clobber highs.
2465 (collect_non_operand_hard_regs): Check for clobber high.
2466 (lra_set_insn_recog_data): Likewise.
2467 (add_regs_to_insn_regno_info): Likewise.
2468 (lra_update_insn_regno_info): Likewise.
2470 2018-08-06 Alan Hayward <alan.hayward@arm.com>
2472 * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
2473 * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
2475 2018-08-06 Alan Hayward <alan.hayward@arm.com>
2477 * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
2478 (copy_insn_1): Likewise.
2479 (gen_hard_reg_clobber_high): New gen function.
2480 * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
2481 * genemit.c (gen_exp): Likewise.
2482 (gen_emit_seq): Pass through info.
2483 (gen_insn): Check for CLOBBER_HIGH.
2484 (gen_expand): Pass through info.
2485 (gen_split): Likewise.
2486 (output_add_clobbers): Likewise.
2487 * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
2488 (remove_clobbers): Likewise.
2489 * rtl.h (gen_hard_reg_clobber_high): New declaration.
2491 2018-08-06 Alan Hayward <alan.hayward@arm.com>
2493 * doc/rtl.texi (clobber_high): Add.
2494 (parallel): Add in clobber high
2495 * rtl.c (rtl_check_failed_code3): Add function.
2496 * rtl.def (CLOBBER_HIGH): Add expression.
2497 * rtl.h (RTL_CHECKC3): Add macro.
2498 (rtl_check_failed_code3): Add declaration.
2499 (XC3EXP): Add macro.
2501 2018-08-05 H.J. Lu <hongjiu.lu@intel.com>
2504 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
2505 cfun->machine->max_used_stack_alignment if needed.
2507 2018-08-04 Martin Sebor <msebor@redhat.com>
2509 PR tree-optimization/86571
2510 * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
2513 2018-08-03 Sandra Loosemore <sandra@codesourcery.com>
2516 * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2519 2018-08-03 Jeff Law <law@redhat.com>
2522 * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2523 Define to speculation_safe_value_not_needed.
2525 2018-08-03 David Malcolm <dmalcolm@redhat.com>
2527 * doc/gcov.texi (-x): Remove duplicate "to".
2528 * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
2529 (-Wif-not-aligned): Remove duplicate "is".
2530 (-flto): Remove duplicate "the".
2531 (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
2532 duplicate "v5.00.b".
2533 (MSP430 Options): Remove duplicate "and" from the description
2534 of "-mgprel-sec=regexp".
2535 (x86 Options): Remove duplicate copies of "vmldLog102" and
2536 vmlsLog104 from description of "-mveclibabi=type".
2538 2018-08-03 Richard Sandiford <richard.sandiford@arm.com>
2540 * internal-fn.h (first_commutative_argument): Declare.
2541 * internal-fn.c (first_commutative_argument): New function.
2542 * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
2543 restrictions for pattern statements. Use first_commutative_argument
2544 to look for commutative operands in calls to internal functions.
2546 2018-08-03 Aldy Hernandez <aldyh@redhat.com>
2548 * Makefile.in (wide-int-range.o): New.
2549 * tree-vrp.c: Move all the wide_int_* functions to...
2550 * wide-int-range.cc: ...here.
2551 * tree-vrp.h: Move all the wide_int_* prototypes to...
2552 * wide-int-range.h: ...here.
2554 2018-08-03 Tom de Vries <tdevries@suse.de>
2556 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
2558 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
2559 * except.c (output_function_exception_table): Do early exit if
2560 targetm_common.except_unwind_info (&global_options) == UI_NONE.
2562 2018-08-03 Martin Liska <mliska@suse.cz>
2564 * predict.c (dump_prediction): Change to 2 digits
2567 2018-08-03 Siddhesh Poyarekar <siddhesh@sourceware.org>
2569 * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
2571 (falkor_am_1_gtov_gtov): ... a new insn reservation.
2573 2018-07-19 Ilya Leoshkevich <iii@linux.ibm.com>
2575 * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
2576 * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
2577 * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
2578 * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
2579 * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
2581 2018-08-02 David Malcolm <dmalcolm@redhat.com>
2583 * diagnostic-show-locus.c (diagnostic_show_locus): Use
2584 pp_take_prefix when saving the existing prefix.
2585 * diagnostic.c (diagnostic_append_note): Likewise.
2586 * langhooks.c (lhd_print_error_function): Likewise.
2587 * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
2588 param's type. Free the existing prefix.
2589 (pp_take_prefix): New function.
2590 (pretty_printer::pretty_printer): Drop the prefix parameter.
2591 Rename the length parameter to match the comment.
2592 (pretty_printer::~pretty_printer): Free the prefix.
2593 * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
2595 (struct pretty_printer): Drop the "const" from "prefix" field's
2596 type and clarify memory management.
2597 (pp_set_prefix): Drop the "const" from the 2nd param.
2598 (pp_take_prefix): New decl.
2600 2018-08-02 Aaron Sawdey <acsawdey@linux.ibm.com>
2602 * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
2603 for word_mode_ok here instead of passing as argument.
2604 (expand_block_compare): Change select_block_compare_mode() call.
2605 (expand_strncmp_gpr_sequence): New function.
2606 (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
2608 2018-08-02 Jeff Law <law@redhat.com>
2611 * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2612 Define to speculation_safe_value_not_needed.
2615 * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2616 Define to speculation_safe_value_not_needed.
2618 2018-08-02 Tom de Vries <tdevries@suse.de>
2621 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
2622 function. Return UI_TARGET unconditionally.
2623 (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
2624 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
2626 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
2628 * genemit.c (print_overload_test): Fix typo.
2630 2018-08-02 Richard Biener <rguenther@suse.de>
2632 PR tree-optimization/86816
2633 * tree-ssa-tail-merge.c (tail_merge_valueize): New function
2634 which checks for value availability before querying it.
2635 (gvn_uses_equal): Use it.
2636 (same_succ_hash): Likewise.
2637 (gimple_equal_p): Likewise.
2639 2018-08-02 Nick Clifton <nickc@redhat.com>
2642 * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2643 Define to speculation_safe_value_not_needed.
2646 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2647 Define to speculation_safe_value_not_needed.
2650 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2651 Define to speculation_safe_value_not_needed.
2654 * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2655 Define to speculation_safe_value_not_needed.
2658 * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2659 Define to speculation_safe_value_not_needed.
2662 * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2663 Define to speculation_safe_value_not_needed.
2666 * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2667 Define to speculation_safe_value_not_needed.
2670 * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2671 Define to speculation_safe_value_not_needed.
2674 * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
2675 speculation_safe_value_not_needed.
2678 * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
2679 to speculation_safe_value_not_needed.
2681 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
2683 * doc/md.texi: Expand the documentation of instruction names
2684 to mention port-local uses. Document '@' in pattern names.
2685 * read-md.h (overloaded_instance, overloaded_name): New structs.
2687 (md_reader::handle_overloaded_name): New member function.
2688 (md_reader::get_overloads): Likewise.
2689 (md_reader::m_first_overload): New member variable.
2690 (md_reader::m_next_overload_ptr): Likewise.
2691 (md_reader::m_overloads_htab): Likewise.
2692 * read-md.c (md_reader::md_reader): Initialize m_first_overload,
2693 m_next_overload_ptr and m_overloads_htab.
2694 * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
2695 (get_mode_token, get_code_token, get_int_token): New functions.
2696 (map_attr_string): Add an optional argument that passes back
2697 the associated iterator.
2698 (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
2699 (md_reader::handle_overloaded_name, add_overload_instance): New
2701 (apply_iterators): Handle '@' names. Report an error if '@'
2702 is used without iterators.
2703 (initialize_iterators): Initialize the new iterator_group fields.
2704 * genopinit.c (handle_overloaded_code_for)
2705 (handle_overloaded_gen): New functions.
2706 (main): Use them to print declarations of maybe_code_for_* and
2707 maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
2708 * genemit.c (print_overload_arguments, print_overload_test)
2709 (handle_overloaded_code_for, handle_overloaded_gen): New functions.
2710 (main): Use it to print definitions of maybe_code_for_* and
2711 maybe_gen_* functions.
2712 * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
2713 gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
2714 instead of explicit mode checks.
2715 (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
2716 (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
2717 (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
2718 (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
2719 (aarch64_expand_compare_and_swap): Likewise
2720 gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
2721 (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
2722 (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
2723 (aarch64_constant_pool_reload_icode): Delete.
2724 (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
2725 instead of aarch64_constant_pool_reload_icode. Use
2726 code_for_aarch64_reload_mov instead of explicit mode checks.
2727 (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
2728 (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
2729 get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
2730 (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
2731 (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
2732 get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
2733 (aarch64_atomic_load_op_code): Delete.
2734 (aarch64_emit_atomic_load_op): Likewise.
2735 (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
2736 aarch64_atomic_load_op_code. Use gen_aarch64_atomic_load
2737 instead of aarch64_emit_atomic_load_op.
2738 * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
2739 (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
2740 (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
2741 (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
2742 character before the pattern name.
2743 * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
2744 (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
2745 (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
2746 (aarch64_frecps<mode>): Likewise.
2747 * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
2748 (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
2749 (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
2750 (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
2751 (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
2753 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
2755 * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
2756 Allow HFmode constants if TARGET_FP_F16INST.
2758 2018-08-02 Jackson Woodruff <jackson.woodruff@arm.com>
2761 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
2762 No longer check last store for clobber of address register.
2764 2018-08-02 Martin Liska <mliska@suse.cz>
2766 PR gcov-profile/86817
2767 * gcov.c (process_all_functions): New function.
2769 (process_file): Move functions processing to
2770 process_all_functions.
2772 2018-08-02 David Malcolm <dmalcolm@redhat.com>
2774 * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
2775 "const" to the "gimple *" and "rtx_insn *" parameters.
2776 * dumpfile.h (dump_user_location_t::dump_user_location_t):
2778 (dump_location_t::dump_location_t): Likewise.
2780 2018-08-01 Martin Sebor <msebor@redhat.com>
2782 PR tree-optimization/86650
2783 * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
2784 rather than a "gcall *". Directly pass the data of interest
2785 to percent_K_format, rather than building a temporary CALL_EXPR
2787 * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
2788 (gimple_fold_builtin_strncat): Adjust.
2789 * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
2790 gcall* argument with gimple*.
2791 * gimple-ssa-warn-restrict.c (check_call): Same.
2792 (wrestrict_dom_walker::before_dom_children): Same.
2793 (builtin_access::builtin_access): Same.
2794 (check_bounds_or_overlap): Same
2795 (maybe_diag_overlap): Same.
2796 (maybe_diag_offset_bounds): Same.
2797 * tree-diagnostic.c (default_tree_printer): Move usage of
2798 EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
2800 * tree-pretty-print.c (percent_K_format): Add argument.
2801 * tree-pretty-print.h: Add argument.
2802 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
2803 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
2804 (maybe_diag_stxncpy_trunc): Same.
2805 (handle_builtin_stxncpy): Same.
2806 (handle_builtin_strcat): Same.
2808 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
2810 * match.pd: Optimise pointer range checks.
2812 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
2814 PR tree-optimization/86758
2815 * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
2816 to remove pattern statements.
2818 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
2820 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
2821 result of dfs_enumerate_from when constructing stmt_vec_infos,
2822 instead of additionally calling get_loop_body.
2824 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
2826 * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
2828 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
2829 When creating an iv, assert that the step is not known to be zero.
2830 (vect_setup_realignment): Update call accordingly.
2831 * tree-vect-stmts.c (vectorizable_store): Likewise.
2832 (vectorizable_load): Likewise. Handle VMAT_INVARIANT separately.
2834 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
2836 * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
2837 * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
2838 (vectorizable_reduction): Likewise.
2839 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
2840 (vect_detect_hybrid_slp_stmts): Likewise.
2841 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
2843 2018-08-01 Aldy Hernandez <aldyh@redhat.com>
2845 * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
2846 (wide_int_set_zero_nonzero_bits): ...this.
2847 (zero_nonzero_bits_from_vr): Rename to...
2848 (vrp_set_zero_nonzero_bits): ...this.
2849 (extract_range_from_multiplicative_op_1): Abstract wide int
2851 (wide_int_range_multiplicative_op): ...here.
2852 (extract_range_from_binary_expr_1): Extract wide int binary
2853 operations into their own functions.
2854 (wide_int_range_lshift): New.
2855 (wide_int_range_can_optimize_bit_op): New.
2856 (wide_int_range_shift_undefined_p): New.
2857 (wide_int_range_bit_xor): New.
2858 (wide_int_range_bit_ior): New.
2859 (wide_int_range_bit_and): New.
2860 (wide_int_range_trunc_mod): New.
2861 (extract_range_into_wide_ints): New.
2862 (vrp_shift_undefined_p): New.
2863 (extract_range_from_multiplicative_op): New.
2864 (vrp_can_optimize_bit_op): New.
2865 * tree-vrp.h (value_range::dump): New.
2866 (wide_int_range_multiplicative_op): New.
2867 (wide_int_range_lshift):New.
2868 (wide_int_range_shift_undefined_p): New.
2869 (wide_int_range_bit_xor): New.
2870 (wide_int_range_bit_ior): New.
2871 (wide_int_range_bit_and): New.
2872 (wide_int_range_trunc_mod): New.
2873 (zero_nonzero_bits_from_bounds): Rename to...
2874 (wide_int_set_zero_nonzero_bits): ...this.
2875 (zero_nonzero_bits_from_vr): Rename to...
2876 (vrp_set_zero_nonzero_bits): ...this.
2877 (range_easy_mask_min_max): Rename to...
2878 (wide_int_range_can_optimize_bit_op): this.
2879 * vr-values.c (simplify_bit_ops_using_ranges): Rename
2880 zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
2882 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
2884 * tree-vectorizer.h (vect_orig_stmt): New function.
2885 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
2886 * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
2887 (vect_create_epilog_for_reduction): Likewise.
2888 (vectorizable_live_operation): Likewise.
2889 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
2890 (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
2891 * tree-vect-stmts.c (vectorizable_call): Likewise.
2892 (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
2894 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
2896 * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
2898 * tree-vect-stmts.c (vect_transform_stmt): Likewise.
2899 * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
2900 (vect_transform_loop): Likewise.
2901 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
2903 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
2905 * tree-vectorizer.h (vect_schedule_slp): Return void.
2906 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
2907 (vect_schedule_slp): Likewise.
2909 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
2911 * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
2913 (vect_transform_loop): Update calls accordingly. Schedule SLP
2914 instances before the main loop, if any exist.
2916 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
2918 PR tree-optimization/86749
2919 * tree-vect-patterns.c (vect_determine_min_output_precision_1):
2920 If the lhs is used in a COND_EXPR, check that it is being used
2921 as the "then" or "else" value.
2923 2018-08-01 Tom de Vries <tdevries@suse.de>
2926 * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
2927 speculation_safe_value_not_needed.
2929 2018-08-01 Richard Biener <rguenther@suse.de>
2931 * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
2934 2018-08-01 Martin Liska <mliska@suse.cz>
2936 * value-prof.c (gimple_divmod_fixed_value_transform): Unify
2937 format how successful transformation is dumped.
2938 (gimple_mod_pow2_value_transform): Likewise.
2939 (gimple_mod_subtract_transform): Likewise.
2940 (gimple_stringops_transform): Likewise.
2942 2018-08-01 Martin Liska <mliska@suse.cz>
2945 * value-prof.c (interesting_stringop_to_profile_p):
2946 Simplify the code and add BUILT_IN_MEMMOVE.
2947 (gimple_stringops_transform): Likewise.
2949 2018-08-01 Sam Tebbs <sam.tebbs@arm.com>
2951 * config/aarch64/aarch64-simd.md
2952 (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
2953 (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
2954 use GPI iterator instead of SI mode.
2956 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
2958 * config/rs6000/rs6000.md (speculation_barrier): Renamed from
2959 rs6000_speculation_barrier.
2960 * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
2961 new barrier pattern name.
2963 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
2965 * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
2966 (speculation_barrier): New insn.
2968 2018-08-01 Richard Biener <rguenther@suse.de>
2971 * graphite.h: Include isl/id.h and isl/space.h to allow build
2974 2018-08-01 Jan Willem Jagersma <jwjagersma@gmail.com>
2977 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
2978 mode for COFF targets.
2979 * defaults.h (TARGET_COFF): Define.
2980 * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
2981 TARGET_COFF): Define.
2982 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
2983 * config/i386/djgpp.c (saved_debug_info_level): New static variable.
2984 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
2986 2018-07-31 Alexandre Oliva <oliva@adacore.com>
2988 * gimple-streamer-in.c (input_bb): Restore BB discriminator.
2989 * gimple-streamer-out.c (output_bb): Save it.
2990 * lto-streamer-in.c (input_struct_function_base): Restore
2991 instance discriminator if available. Create map on demand.
2992 * lto-streamer-out.c (output_struct_function_base): Save it if
2994 * final.c (decl_to_instance_map): Document LTO strategy.
2996 2018-07-31 Alexandre Oliva <oliva@adacore.com>
2997 Olivier Hainque <hainque@adacore.com>
2999 * debug.h (decl_to_instance_map_t): New type.
3000 (decl_to_instance_map): Declare.
3001 (maybe_create_decl_to_instance_map): New inline function.
3002 * final.c (bb_discriminator, last_bb_discriminator): New statics,
3003 to track basic block discriminators.
3004 (final_start_function_1): Initialize them.
3005 (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
3007 (decl_to_instance_map): New variable.
3008 (map_decl_to_instance, maybe_set_discriminator): New functions.
3009 (notice_source_line): Set discriminator.
3011 2018-07-31 Ian Lance Taylor <iant@golang.org>
3013 * targhooks.c (default_have_speculation_safe_value): Add
3016 2018-07-31 David Malcolm <dmalcolm@redhat.com>
3018 * dump-context.h: Include "pretty-print.h".
3019 (dump_context::refresh_dumps_are_enabled): New decl.
3020 (dump_context::emit_item): New decl.
3021 (class dump_context): Add fields "m_test_pp" and
3023 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
3024 (temp_dump_context::get_dumped_text): New decl.
3025 (class temp_dump_context): Add field "m_pp".
3026 * dumpfile.c (refresh_dumps_are_enabled): Convert to...
3027 (dump_context::refresh_dumps_are_enabled): ...and add a test for
3029 (set_dump_file): Update for above change.
3030 (set_alt_dump_file): Likewise.
3031 (dump_loc): New overload, taking a pretty_printer *.
3032 (dump_context::dump_loc): Call end_any_optinfo. Dump the location
3033 to any test pretty-printer.
3034 (make_item_for_dump_gimple_stmt): New function, adapted from
3035 optinfo::add_gimple_stmt.
3036 (dump_context::dump_gimple_stmt): Call it, and use the result,
3037 eliminating the direct usage of dump_file and alt_dump_file in
3038 favor of indirectly using them via emit_item.
3039 (make_item_for_dump_gimple_expr): New function, adapted from
3040 optinfo::add_gimple_expr.
3041 (dump_context::dump_gimple_expr): Call it, and use the result,
3042 eliminating the direct usage of dump_file and alt_dump_file in
3043 favor of indirectly using them via emit_item.
3044 (make_item_for_dump_generic_expr): New function, adapted from
3046 (dump_context::dump_generic_expr): Call it, and use the result,
3047 eliminating the direct usage of dump_file and alt_dump_file in
3048 favor of indirectly using them via emit_item.
3049 (make_item_for_dump_printf_va): New function, adapted from
3050 optinfo::add_printf_va.
3051 (make_item_for_dump_printf): New function.
3052 (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
3053 and use the result, eliminating the direct usage of dump_file and
3054 alt_dump_file in favor of indirectly using them via emit_item.
3055 (make_item_for_dump_dec): New function.
3056 (dump_context::dump_dec): Call it, and use the result,
3057 eliminating the direct usage of dump_file and alt_dump_file in
3058 favor of indirectly using them via emit_item.
3059 (make_item_for_dump_symtab_node): New function, adapted from
3060 optinfo::add_symtab_node.
3061 (dump_context::dump_symtab_node): Call it, and use the result,
3062 eliminating the direct usage of dump_file and alt_dump_file in
3063 favor of indirectly using them via emit_item.
3064 (dump_context::begin_scope): Reimplement, avoiding direct usage
3065 of dump_file and alt_dump_file in favor of indirectly using them
3067 (dump_context::emit_item): New member function.
3068 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
3069 Set up test pretty-printer on the underlying context. Call
3070 refresh_dumps_are_enabled.
3071 (temp_dump_context::~temp_dump_context): Call
3072 refresh_dumps_are_enabled.
3073 (temp_dump_context::get_dumped_text): New member function.
3074 (selftest::verify_dumped_text): New function.
3075 (ASSERT_DUMPED_TEXT_EQ): New macro.
3076 (selftest::test_capture_of_dump_calls): Run all tests twice, with
3077 and then without optinfo enabled. Add uses of
3078 ASSERT_DUMPED_TEXT_EQ to all tests. Add test of nested scopes.
3079 * dumpfile.h: Update comment for the dump_* API.
3080 * optinfo-emit-json.cc
3081 (selftest::test_building_json_from_dump_calls): Update for new
3082 param for temp_dump_context ctor.
3083 * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
3084 and "m_owned" field.
3085 (optinfo_item::~optinfo_item): Likewise.
3086 (optinfo::add_item): New member function.
3087 (optinfo::emit): Update comment.
3088 (optinfo::add_string): Delete.
3089 (optinfo::add_printf): Delete.
3090 (optinfo::add_printf_va): Delete.
3091 (optinfo::add_gimple_stmt): Delete.
3092 (optinfo::add_gimple_expr): Delete.
3093 (optinfo::add_tree): Delete.
3094 (optinfo::add_symtab_node): Delete.
3095 (optinfo::add_dec): Delete.
3096 * optinfo.h (class dump_context): New forward decl.
3097 (optinfo::add_item): New decl.
3098 (optinfo::add_string): Delete.
3099 (optinfo::add_printf): Delete.
3100 (optinfo::add_printf_va): Delete.
3101 (optinfo::add_gimple_stmt): Delete.
3102 (optinfo::add_gimple_expr): Delete.
3103 (optinfo::add_tree): Delete.
3104 (optinfo::add_symtab_node): Delete.
3105 (optinfo::add_dec): Delete.
3106 (optinfo::add_poly_int): Delete.
3107 (optinfo_item::optinfo_item): Remove "owned" param.
3108 (class optinfo_item): Remove field "m_owned".
3110 2018-07-31 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3113 * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
3114 MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
3115 requested variable alignment.
3116 (expand_one_ssa_partition): Likewise.
3117 (expand_one_var): Likewise.
3119 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3121 * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
3122 to speculation_safe_value_not_needed.
3124 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3126 * targhooks.h (speculation_safe_value_not_needed): New prototype.
3127 * targhooks.c (speculation_safe_value_not_needed): New function.
3128 * target.def (have_speculation_safe_value): Update documentation.
3129 * doc/tm.texi: Regenerated.
3131 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3133 * config/aarch64/iterators.md (ALLI_TI): New iterator.
3134 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
3136 (despeculate_copy<ALLI:mode>_insn): New insn.
3137 (despeculate_copyti_insn): New insn.
3138 (despeculate_simple<ALLI:mode>): New insn
3139 (despeculate_simpleti): New insn.
3140 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
3142 (TARGET_SPECULATION_SAFE_VALUE): Redefine to
3143 aarch64_speculation_safe_value.
3144 (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
3146 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3148 * config/aarch64/aarch64-speculation.cc: New file.
3149 * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
3150 before pass_reorder_blocks.
3151 * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
3153 * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
3154 X14 and X15 when tracking speculation.
3155 * config/aarch64/aarch64.md (register name constants): Add
3156 SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
3157 (unspec): Add UNSPEC_SPECULATION_TRACKER.
3158 (speculation_barrier): New insn attribute.
3159 (cmp<mode>): Allow SP in comparisons.
3160 (speculation_tracker): New insn.
3161 (speculation_barrier): Add speculation_barrier attribute.
3162 * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
3163 * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
3164 * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
3166 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3168 * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
3169 aarch64_track_speculation is true.
3170 (tb<optab><mode>1): Likewise.
3171 * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
3172 generate CB[N]Z when tracking speculation.
3173 (aarch64_split_compare_and_swap): Likewise.
3174 (aarch64_split_atomic_op): Likewise.
3176 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3178 * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
3180 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3182 * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
3183 (speculation_barrier): New insn.
3185 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3187 * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
3188 * config/arm/arm.md (speculation_barrier): New expand.
3189 (speculation_barrier_insn): New pattern.
3191 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3193 * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
3194 (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
3195 (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
3196 * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
3198 * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
3199 (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
3200 (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
3201 (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
3202 (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
3203 (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
3204 (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
3205 * builtins.c (expand_speculation_safe_value): New function.
3206 (expand_builtin): Call it.
3207 * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
3208 * doc/extend.texi: Document __builtin_speculation_safe_value.
3209 * doc/md.texi: Document "speculation_barrier" pattern.
3210 * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
3211 TARGET_HAVE_SPECULATION_SAFE_VALUE.
3212 * doc/tm.texi: Regenerated.
3213 * target.def (have_speculation_safe_value, speculation_safe_value): New
3215 * targhooks.c (default_have_speculation_safe_value): New function.
3216 (default_speculation_safe_value): New function.
3217 * targhooks.h (default_have_speculation_safe_value): Add prototype.
3218 (default_speculation_safe_value): Add prototype.
3220 2018-07-31 David Malcolm <dmalcolm@redhat.com>
3222 * dump-context.h (dump_context::dump_loc): New decl.
3223 * dumpfile.c (dump_context::dump_loc): New member function.
3224 (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
3225 and dump_gimple_stmt.
3226 (dump_context::dump_gimple_expr_loc): Likewise, using
3228 (dump_context::dump_generic_expr_loc): Likewise, using
3230 (dump_context::dump_printf_loc_va): Likewise, using
3232 (dump_context::begin_scope): Explicitly using the global function
3233 "dump_loc", rather than the member function.
3235 2018-07-31 Martin Sebor <msebor@redhat.com>
3237 PR tree-optimization/86741
3238 * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
3240 2018-07-31 Andreas Krebbel <krebbel@linux.ibm.com>
3242 * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
3243 depend on whether prefetch instructions will be emitted or not.
3244 Use TARGET_SETMEM_PFD for checking whether prefetch instructions
3245 will be emitted or not.
3246 * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
3247 (TARGET_SETMEM_PFD): New macros.
3249 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3251 * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
3252 (NULL_STMT_VEC_INFO): Delete.
3253 (stmt_vec_info::operator*): Likewise.
3254 (stmt_vec_info::operator gimple *): Likewise.
3255 * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
3256 of NULL_STMT_VEC_INFO.
3257 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
3258 (vect_reassociating_reduction_p): Likewise.
3259 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
3260 (vectorizable_store): Likewise.
3261 * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
3262 (vec_info::free_stmt_vec_infos): Likewise.
3264 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3266 * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
3267 * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
3269 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3271 * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
3272 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
3273 (vec_info::free_stmt_vec_info): New private member functions.
3274 (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
3275 (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
3276 * tree-parloops.c (gather_scalar_reductions): Remove calls to
3277 set_stmt_vec_info_vec and free_stmt_vec_infos.
3278 * tree-vect-loop.c (_loop_vec_info): Remove call to
3279 set_stmt_vec_info_vec.
3280 * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
3281 (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
3282 * tree-vectorizer.c (vec_info::new_stmt_vec_info)
3283 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
3284 (vec_info::free_stmt_vec_info): ...these new functions. Remove
3285 assignments in {vec_info::,}new_stmt_vec_info that are redundant
3286 with the clearing in the xcalloc.
3287 (stmt_vec_info_vec): Delete.
3288 (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
3289 (vectorize_loops): Likewise.
3290 (vec_info::~vec_info): Remove argument from call to
3291 free_stmt_vec_infos.
3292 (vec_info::add_stmt): Remove vinfo argument from call to
3295 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3297 * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
3298 rather than a gimple stmt.
3299 * tree-vect-stmts.c (free_stmt_vec_info): Likewise. Don't free
3300 information for pattern statements when passed the original
3301 statement; instead wait to be passed the pattern statement itself.
3302 Don't call set_vinfo_for_stmt here.
3303 (free_stmt_vec_infos): Update call to free_stmt_vec_info.
3304 * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
3305 stmt_vec_infos here.
3306 * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
3307 * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
3308 stmt_vec_infos entry.
3310 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3312 * tree-vectorizer.h (vec_info::replace_stmt): Declare.
3313 * tree-vectorizer.c (vec_info::replace_stmt): New function.
3314 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
3315 * tree-vect-stmts.c (vectorizable_call): Likewise.
3316 (vectorizable_simd_clone_call): Likewise.
3318 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3320 * tree-vectorizer.h (vec_info::remove_stmt): Declare.
3321 * tree-vectorizer.c (vec_info::remove_stmt): New function.
3322 * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
3323 * tree-vect-loop.c (vect_transform_loop): Likewise.
3324 * tree-vect-slp.c (vect_schedule_slp): Likewise.
3325 * tree-vect-stmts.c (vect_remove_stores): Likewise.
3327 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3329 * tree-vectorizer.h (vec_info::lookup_dr): New member function.
3330 (vect_dr_stmt): Delete.
3331 * tree-vectorizer.c (vec_info::lookup_dr): New function.
3332 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
3334 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
3335 (vect_analyze_data_ref_dependence, vect_record_base_alignments)
3336 (vect_verify_datarefs_alignment, vect_peeling_supportable)
3337 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
3338 (vect_analyze_data_refs): Likewise.
3339 (vect_slp_analyze_data_ref_dependence): Likewise. Take a vec_info
3341 (vect_find_same_alignment_drs): Likewise.
3342 (vect_slp_analyze_node_dependences): Update calls accordingly.
3343 (vect_analyze_data_refs_alignment): Likewise. Use vec_info::lookup_dr
3344 instead of DR_VECT_AUX.
3345 (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
3346 of a vector data references. Use vec_info::lookup_dr instead of
3348 (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
3349 (vect_enhance_data_refs_alignment): Likewise. Use vec_info::lookup_dr
3350 instead of DR_VECT_AUX.
3352 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3354 * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
3356 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
3358 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
3359 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
3360 (vect_gen_prolog_loop_niters): Likewise.
3362 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3364 * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
3365 (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
3366 (vect_known_alignment_in_bytes, vect_dr_behavior)
3367 (vect_get_scalar_dr_size): Take references as dr_vec_infos
3368 instead of data_references. Update calls to other routines for
3369 which the same change has been made.
3370 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
3371 dr_vec_infos instead of stmt_vec_infos.
3372 (vect_analyze_data_ref_dependence): Update call accordingly.
3373 (vect_slp_analyze_data_ref_dependence)
3374 (vect_record_base_alignments): Use DR_VECT_AUX.
3375 (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
3376 (vect_update_misalignment_for_peel, verify_data_ref_alignment)
3377 (vector_alignment_reachable_p, vect_get_data_access_cost)
3378 (vect_peeling_supportable, vect_analyze_group_access_1)
3379 (vect_analyze_group_access, vect_analyze_data_ref_access)
3380 (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
3381 (vect_compile_time_alias, vect_small_gap_p)
3382 (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
3383 (vect_supportable_dr_alignment): Take references as dr_vec_infos
3384 instead of data_references. Update calls to other routines for
3385 which the same change has been made.
3386 (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
3387 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
3388 (vect_slp_analyze_and_verify_node_alignment)
3389 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
3390 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
3391 (vect_setup_realignment): Use dr_vec_infos. Update calls after
3393 (_vect_peel_info::dr): Replace with...
3394 (_vect_peel_info::dr_info): ...this new field.
3395 (vect_peeling_hash_get_most_frequent)
3396 (vect_peeling_hash_choose_best_peeling): Update accordingly.
3397 (vect_peeling_hash_get_lowest_cost):
3398 (vect_enhance_data_refs_alignment): Likewise. Update calls to other
3399 routines for which the same change has been made.
3400 (vect_peeling_hash_insert): Likewise. Take a dr_vec_info instead of a
3402 * tree-vect-loop-manip.c (get_misalign_in_elems)
3403 (vect_gen_prolog_loop_niters): Use dr_vec_infos. Update calls after
3405 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
3406 * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
3407 (vect_truncate_gather_scatter_offset, compare_step_with_zero)
3408 (get_group_load_store_type, get_negative_load_store_type)
3409 (vect_get_data_ptr_increment, vectorizable_store)
3410 (vectorizable_load): Likewise.
3411 (ensure_base_align): Take a dr_vec_info instead of a data_reference.
3412 Update calls to other routines for which the same change has been made.
3414 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3416 * tree-vectorizer.h (vec_info::move_dr): New member function.
3417 (dataref_aux): Rename to...
3418 (dr_vec_info): ...this and add "dr" and "stmt" fields.
3419 (_stmt_vec_info::dr_aux): Update accordingly.
3420 (_stmt_vec_info::data_ref_info): Delete.
3421 (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
3422 (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
3423 (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
3424 (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
3426 (STMT_VINFO_DATA_REF): Likewise. Turn into an lvalue.
3427 (STMT_VINFO_DR_INFO): New macro.
3428 (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
3429 (set_dr_misalignment): Update after rename of dataref_aux.
3430 (vect_dr_stmt): Move earlier in file. Return dr_aux.stmt.
3431 * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
3432 initialization of STMT_VINFO_DATA_REF.
3433 * tree-vectorizer.c (vec_info::move_dr): New function.
3434 * tree-vect-patterns.c (vect_recog_bool_pattern)
3435 (vect_recog_mask_conversion_pattern)
3436 (vect_recog_gather_scatter_pattern): Use it.
3437 * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
3438 the "dr" and "stmt" fields of dr_vec_info instead of
3439 STMT_VINFO_DATA_REF.
3441 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3443 * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
3444 (is_pattern_stmt_p): Use it.
3445 * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
3446 on pattern statements.
3448 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3450 * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
3451 original stmt as a stmt_vec_info rather than a gimple stmt.
3452 (vect_pattern_recog_1): Take the statement directly as a
3453 stmt_vec_info, rather than via a gimple_stmt_iterator.
3454 Update call to vect_mark_pattern_stmts.
3455 (vect_pattern_recog): Update calls accordingly.
3457 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3459 * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
3460 (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
3461 a vect_def_type for the first argument.
3462 * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
3463 (vect_get_vec_def_for_stmt_copy): Likewise. Return the original
3464 operand if it isn't defined by a vectorized statement.
3465 (vect_build_gather_load_calls): Remove the mask_dt argument and
3466 update calls to vect_get_vec_def_for_stmt_copy.
3467 (vectorizable_bswap): Likewise the dt argument.
3468 (vectorizable_call): Update calls to vectorizable_bswap and
3469 vect_get_vec_def_for_stmt_copy.
3470 (vectorizable_simd_clone_call, vectorizable_assignment)
3471 (vectorizable_shift, vectorizable_operation, vectorizable_condition)
3472 (vectorizable_comparison): Update calls to
3473 vect_get_vec_def_for_stmt_copy.
3474 (vectorizable_store): Likewise. Remove now-unnecessary calls to
3476 (vect_get_loop_based_defs): Remove dt argument and update call
3477 to vect_get_vec_def_for_stmt_copy.
3478 (vectorizable_conversion): Update calls to vect_get_loop_based_defs
3479 and vect_get_vec_def_for_stmt_copy.
3480 (vectorizable_load): Update calls to vect_build_gather_load_calls
3481 and vect_get_vec_def_for_stmt_copy.
3482 * tree-vect-loop.c (vect_create_epilog_for_reduction)
3483 (vectorizable_reduction, vectorizable_live_operation): Update calls
3484 to vect_get_vec_def_for_stmt_copy.
3486 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3488 * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
3489 and gimple stmt arguments with a stmt_vec_info.
3490 (vect_record_base_alignments): Update calls accordingly.
3491 * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
3492 and gimple stmt arguments with a stmt_vec_info.
3493 (vect_build_slp_tree_1): Remove vinfo argument and update call
3494 to vect_record_max_nunits.
3495 (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
3496 and vect_record_max_nunits.
3498 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3500 * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
3501 file and take a stmt_vec_info instead of a gimple stmt.
3502 (supportable_widening_operation, vect_finish_replace_stmt)
3503 (vect_finish_stmt_generation, vect_get_store_rhs)
3504 (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
3505 (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
3506 (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
3507 (vect_get_smallest_scalar_type, vect_check_gather_scatter)
3508 (vect_create_data_ref_ptr, bump_vector_ptr)
3509 (vect_permute_store_chain, vect_setup_realignment)
3510 (vect_transform_grouped_load, vect_record_grouped_load_vectors)
3511 (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
3512 (vectorizable_reduction, vectorizable_induction)
3513 (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
3514 (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
3515 than gimple stmts as arguments.
3516 * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
3517 (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
3518 (can_group_stmts_p, vect_check_gather_scatter)
3519 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
3520 (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
3521 (vect_permute_load_chain, vect_shift_permute_load_chain)
3522 (vect_transform_grouped_load)
3523 (vect_record_grouped_load_vectors): Likewise.
3524 * tree-vect-loop.c (vect_fixup_reduc_chain)
3525 (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
3526 (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
3527 (vectorizable_reduction, vectorizable_induction)
3528 (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
3529 * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
3530 (vect_get_load_store_mask): Likewise.
3531 * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
3532 (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
3533 * tree-vect-stmts.c (vect_mark_relevant)
3534 (is_simple_and_all_uses_invariant)
3535 (exist_non_indexing_operands_for_use_p, process_use)
3536 (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
3537 (vect_get_vec_def_for_operand, vect_get_vec_defs)
3538 (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
3539 (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
3540 (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
3541 (get_negative_load_store_type, get_load_store_type)
3542 (vect_check_load_store_mask, vect_check_store_rhs)
3543 (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
3544 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
3545 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
3546 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
3547 (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
3548 (vectorizable_load, vectorizable_condition, vectorizable_comparison)
3549 (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
3550 (supportable_widening_operation): Likewise.
3552 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3554 * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
3555 a stmt_vec_info instead of a gcall.
3556 (vect_check_gather_scatter): Update call accordingly.
3557 * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
3559 (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
3560 (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
3561 * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
3562 instead of a gimple stmt.
3563 (vect_transform_loop): Update calls accordingly.
3564 * tree-vect-slp.c (vect_split_slp_store_group): Take and return
3565 stmt_vec_infos instead of gimple stmts.
3566 (vect_analyze_slp_instance): Update use accordingly.
3567 * tree-vect-stmts.c (read_vector_array, write_vector_array)
3568 (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
3569 (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
3570 (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
3571 (vect_gen_widened_results_half, vect_get_loop_based_defs)
3572 (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
3573 Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
3574 down to subroutines.
3576 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3578 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
3579 of the worklist from a vector of gimple stmts to a vector of
3581 * tree-vect-stmts.c (vect_mark_relevant, process_use)
3582 (vect_mark_stmts_to_be_vectorized): Likewise
3584 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3586 * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
3587 statement before passing it to vect_analyze_stmt.
3588 (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
3589 the chain of phi vector definitions. Track the exit phi via its
3591 (vectorizable_reduction): Set cond_stmt_vinfo directly from the
3592 STMT_VINFO_REDUC_DEF.
3593 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
3594 stmt_vec_infos to handle the statement chains.
3595 (vect_get_slp_defs): Record the first statement in the node
3596 using a stmt_vec_info.
3597 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
3598 statements here and pass their stmt_vec_info down to subroutines.
3599 (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
3600 down to vect_finish_stmt_generation.
3601 (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
3602 (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
3603 stmt_vec_infos to subroutines.
3604 (vect_remove_stores): Use stmt_vec_infos to handle the statement
3607 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3609 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
3610 (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
3611 (vect_permute_store_chain, vect_setup_realignment)
3612 (vect_permute_load_chain, vect_shift_permute_load_chain)
3613 (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
3614 stmts internally, and when passing values to other vectorizer routines.
3615 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
3616 * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
3617 (vect_analyze_loop_operations, get_initial_def_for_reduction)
3618 (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
3619 (vectorizable_reduction, vectorizable_induction)
3620 (vectorizable_live_operation, vect_transform_loop_stmt)
3621 (vect_transform_loop): Likewise.
3622 * tree-vect-patterns.c (vect_reassociating_reduction_p)
3623 (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
3624 (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
3625 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
3626 (vect_slp_analyze_node_operations_1): Likewise.
3627 * tree-vect-stmts.c (vect_mark_relevant, process_use)
3628 (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
3629 (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
3630 (vect_finish_stmt_generation_1, get_group_load_store_type)
3631 (get_load_store_type, vect_build_gather_load_calls)
3632 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
3633 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
3634 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
3635 (vectorizable_store, vectorizable_load, vectorizable_condition)
3636 (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
3637 (supportable_widening_operation): Likewise.
3638 (vect_get_vector_types_for_stmt): Likewise.
3639 * tree-vectorizer.h (vect_dr_behavior): Likewise.
3641 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3643 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
3644 (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
3645 (vect_permute_store_chain, vect_permute_load_chain)
3646 (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
3647 repeated stmt_vec_info lookups.
3648 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
3649 (vect_update_ivs_after_vectorizer): Likewise.
3650 * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
3651 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
3652 * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
3653 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
3654 (vect_bb_slp_scalar_cost): Likewise.
3655 * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
3657 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3659 * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
3660 gcall rather than the generic gimple stmt to gimple_call_internal_fn.
3661 (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
3662 to get gassigns and gcalls, rather than operating on generc gimple
3664 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
3665 (vect_mark_stmts_to_be_vectorized, vectorizable_store)
3666 (vectorizable_load, vect_analyze_stmt): Likewise.
3667 * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
3669 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3671 * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
3672 return stmt_vec_infos rather than gimple stmts. Do not accept
3674 (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
3676 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
3677 Update use of get_later_stmt.
3678 (vect_get_constant_vectors): Update call accordingly.
3679 (vect_schedule_slp_instance): Likewise
3680 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
3681 (vect_slp_analyze_instance_dependence): Likewise.
3682 (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
3684 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3686 * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
3687 (stmt_info_for_cost::stmt_info): ...this new field.
3688 (add_stmt_costs): Update accordingly.
3689 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
3690 (vect_get_known_peeling_cost): Likewise.
3691 (vect_estimate_min_profitable_iters): Likewise.
3692 * tree-vect-stmts.c (record_stmt_cost): Likewise.
3694 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3696 * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
3697 from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
3698 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
3700 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
3702 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3704 * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
3705 a gimple stmt to a stmt_vec_info.
3706 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
3708 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3710 * tree-vectorizer.h (vec_info::grouped_stores): Change from
3711 an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
3712 (_loop_vec_info::reduction_chains): Likewise.
3713 * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
3715 * tree-vect-slp.c (vect_analyze_slp): Likewise.
3717 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3719 * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
3720 a gimple stmt to a stmt_vec_info.
3721 (_stmt_vec_info::next_element): Likewise.
3722 * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
3723 (vect_slp_analyze_and_verify_node_alignment)
3724 (vect_analyze_group_access_1, vect_analyze_group_access)
3725 (vect_small_gap_p, vect_prune_runtime_alias_test_list)
3726 (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
3727 (vect_supportable_dr_alignment): Update accordingly.
3728 * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
3729 (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
3730 (vect_is_simple_reduction, vectorizable_reduction): Likewise.
3731 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
3732 * tree-vect-slp.c (vect_build_slp_tree_1)
3733 (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
3734 (vect_split_slp_store_group, vect_analyze_slp_instance)
3735 (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
3736 * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
3737 (get_group_load_store_type, get_load_store_type)
3738 (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
3739 (vect_transform_stmt, vect_remove_stores): Likewise.
3741 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3743 * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
3745 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
3746 (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
3747 (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
3748 (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
3749 (vector_alignment_reachable_p, vect_get_data_access_cost)
3750 (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
3751 (vect_peeling_supportable, vect_enhance_data_refs_alignment)
3752 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
3753 (vect_analyze_group_access_1, vect_analyze_group_access)
3754 (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
3755 (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
3756 (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
3757 result of vect_dr_stmt and use the stmt_vec_info instead of
3758 the associated gimple stmt.
3759 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
3760 (vect_gen_prolog_loop_niters): Likewise.
3761 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
3763 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3765 * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
3766 to a vec<stmt_vec_info>.
3767 * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
3768 (vect_create_new_slp_node): Take a vec<gimple *> instead of a
3770 (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
3771 to a vec<stmt_vec_info>.
3772 (bst_traits::value_type, bst_traits::value_type): Likewise.
3773 (bst_traits::hash): Update accordingly.
3774 (vect_get_and_check_slp_defs): Change the stmts parameter from
3775 a vec<gimple *> to a vec<stmt_vec_info>.
3776 (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
3777 (vect_build_slp_tree): Likewise.
3778 (vect_build_slp_tree_2): Likewise. Update uses of
3779 SLP_TREE_SCALAR_STMTS.
3780 (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
3781 (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
3782 (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
3783 (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
3784 (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
3785 (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
3786 (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
3787 (vect_get_constant_vectors, vect_get_slp_defs)
3788 (vect_transform_slp_perm_load, vect_schedule_slp_instance)
3789 (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
3790 (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
3791 instead of gimple stmts.
3792 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
3793 the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
3794 (vect_slp_analyze_instance_dependence): Update uses of
3795 SLP_TREE_SCALAR_STMTS.
3796 (vect_slp_analyze_and_verify_node_alignment): Likewise.
3797 (vect_slp_analyze_and_verify_instance_alignment): Likewise.
3798 * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
3799 (get_initial_defs_for_reduction): Likewise.
3800 (vect_create_epilog_for_reduction): Likewise.
3801 (vectorize_fold_left_reduction): Likewise.
3802 * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
3803 (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
3804 (can_vectorize_live_stmts): Likewise.
3806 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3808 * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
3809 auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
3810 (vect_force_simple_reduction): Take and return stmt_vec_infos rather
3812 * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
3814 (gather_scalar_reductions): Update after above interface changes.
3815 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
3816 (vect_is_simple_reduction): Take and return stmt_vec_infos rather
3818 (vect_force_simple_reduction): Likewise.
3819 * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
3820 LOOP_VINFO_REDUCTIONS.
3821 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
3823 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3825 * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
3826 a gimple stmt to a stmt_vec_info.
3827 * tree-vect-loop.c (vect_active_double_reduction_p)
3828 (vect_force_simple_reduction, vectorizable_reduction): Update
3831 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3833 * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
3834 vec<gimple *> to a vec<stmt_vec_info>.
3835 * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
3836 the reduction_phis argument from a vec<gimple *> to a
3838 (vectorizable_reduction): Likewise the phis local variable that
3839 is passed to vect_create_epilog_for_reduction. Update for new type
3840 of SLP_TREE_VEC_STMTS.
3841 (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
3842 (vectorizable_live_operation): Likewise.
3843 * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
3844 (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
3846 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3848 * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
3849 a gimple stmt to a stmt_vec_info.
3850 (vectorizable_condition, vectorizable_live_operation)
3851 (vectorizable_reduction, vectorizable_induction): Pass back the
3852 vectorized statement as a stmt_vec_info.
3853 * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
3854 use of STMT_VINFO_VEC_STMT.
3855 * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
3856 accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
3857 as stmt_vec_infos rather than gimple stmts.
3858 (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
3860 (vectorizable_live_operation): Likewise.
3861 (vectorizable_reduction, vectorizable_induction): Likewise,
3862 updating use of STMT_VINFO_VEC_STMT.
3863 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
3864 of STMT_VINFO_VEC_STMT.
3865 (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
3866 (vectorizable_simd_clone_call, vectorizable_conversion)
3867 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
3868 (vectorizable_store, vectorizable_load, vectorizable_condition)
3869 (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
3870 from a gimple stmt to a stmt_vec_info.
3871 (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT. Pass a
3872 pointer to a stmt_vec_info to the vectorizable_* routines.
3874 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3876 * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
3877 a gimple stmt to a stmt_vec_info.
3878 (is_pattern_stmt_p): Update accordingly.
3879 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
3880 (vect_record_grouped_load_vectors): Likewise.
3881 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
3882 (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
3883 (vect_model_reduction_cost): Likewise.
3884 (vect_create_epilog_for_reduction): Likewise.
3885 (vectorizable_reduction, vectorizable_induction): Likewise.
3886 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
3887 Return the stmt_vec_info for the pattern statement.
3888 (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
3889 (vect_split_statement, vect_mark_pattern_stmts): Likewise.
3890 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
3891 (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
3892 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
3893 (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
3894 (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
3895 (free_stmt_vec_info, vect_is_simple_use): Likewise.
3897 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3899 * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
3900 (vect_finish_stmt_generation): Likewise.
3901 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
3902 (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
3903 (vect_build_gather_load_calls): Use the return value of the above
3904 functions instead of a separate call to vinfo_for_stmt. Use narrow
3905 scopes for the input gimple stmt and wider scopes for the associated
3906 stmt_vec_info. Use vec_info::lookup_def when setting these
3907 stmt_vec_infos from an SSA_NAME definition.
3908 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
3909 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
3910 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
3911 (vectorizable_store, vectorizable_load, vectorizable_condition)
3912 (vectorizable_comparison): Likewise.
3913 * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
3914 (vectorizable_reduction): Likewise.
3916 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3918 * tree-vectorizer.h (vect_is_simple_use): Add an optional
3919 stmt_vec_info * parameter before the optional gimple **.
3920 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
3921 (process_use, vect_get_vec_def_for_operand_1): Update callers.
3922 (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
3923 * tree-vect-loop.c (vectorizable_reduction): Likewise.
3924 (vectorizable_live_operation): Likewise.
3925 * tree-vect-patterns.c (type_conversion_p): Likewise.
3926 (vect_look_through_possible_promotion): Likewise.
3927 (vect_recog_rotate_pattern): Likewise.
3928 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
3930 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3932 * tree-vectorizer.h (stmt_vec_info): Temporarily change from
3933 a typedef to a wrapper class.
3934 (NULL_STMT_VEC_INFO): New macro.
3935 (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
3936 (stmt_vec_info::operator*): New function.
3937 (stmt_vec_info::operator gimple *): Likewise.
3938 (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
3939 (add_stmt_costs): Likewise.
3940 * tree-vect-loop-manip.c (iv_phi_p): Likewise.
3941 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
3942 (vect_get_known_peeling_cost): Likewise.
3943 (vect_estimate_min_profitable_iters): Likewise.
3944 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
3945 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
3946 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
3947 (vectorizable_store, free_stmt_vec_infos): Likewise.
3948 (new_stmt_vec_info): Change return type of xcalloc to
3951 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3953 * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
3954 * tree-vectorizer.c (vec_info::lookup_single_use): New function.
3955 * tree-vect-loop.c (vectorizable_reduction): Use it instead of
3956 a single_imm_use-based sequence.
3957 * tree-vect-stmts.c (supportable_widening_operation): Likewise.
3959 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3961 * tree-vectorizer.h (vec_info::lookup_def): Declare.
3962 * tree-vectorizer.c (vec_info::lookup_def): New function.
3963 * tree-vect-patterns.c (vect_get_internal_def): Use it.
3964 (vect_widened_op_tree): Likewise.
3965 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
3966 * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
3967 (vectorizable_reduction): Likewise.
3968 (vect_valid_reduction_input_p): Take a stmt_vec_info instead
3970 (vect_is_slp_reduction): Update calls accordingly. Use
3971 vec_info::lookup_def.
3972 (vect_is_simple_reduction): Likewise
3973 * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
3975 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3977 * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
3978 * tree-vectorizer.c (vec_info::lookup_stmt): New function.
3979 * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
3981 (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
3982 (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
3983 (vect_update_vf_for_slp, vect_analyze_loop_operations)
3984 (vect_is_slp_reduction, vectorizable_induction)
3985 (vect_transform_loop_stmt, vect_transform_loop): Likewise.
3986 * tree-vect-patterns.c (vect_init_pattern_stmt):
3987 (vect_determine_min_output_precision_1, vect_determine_precisions)
3988 (vect_pattern_recog): Likewise.
3989 * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
3990 * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
3991 * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
3992 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
3993 (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
3994 (vect_detect_hybrid_slp): Likewise. Change the walk_stmt_info
3995 info field from a loop to a loop_vec_info.
3997 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3999 * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
4000 (vec_info::add_stmt): Declare.
4001 * tree-vectorizer.c (vec_info::add_stmt): New function.
4002 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
4003 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
4004 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
4005 (vectorizable_induction): Likewise.
4006 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
4007 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
4008 (vectorizable_simd_clone_call, vectorizable_store): Likewise.
4009 (vectorizable_load): Likewise.
4010 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
4011 (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
4012 (vect_recog_gather_scatter_pattern): Likewise.
4013 (append_pattern_def_seq): Likewise. Remove a check that is
4014 performed by add_stmt itself.
4016 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4018 * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
4019 which make_ssa_name was called with new_stmt before new_stmt
4022 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4024 * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
4026 (vect_is_slp_reduction): ...here...
4027 (vect_is_simple_reduction): ...and here. Remove repetition of tests
4028 that are already known to be false.
4030 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4032 * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
4033 * tree-vect-slp.c (vect_free_slp_tree): Likewise. Don't update
4034 STMT_VINFO_NUM_SLP_USES when it's true.
4035 (vect_free_slp_instance): Add a final_p parameter and pass it to
4037 (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
4038 (vect_analyze_slp_instance): Likewise.
4039 (vect_slp_analyze_operations): Likewise.
4040 (vect_slp_analyze_bb_1): Likewise.
4041 * tree-vectorizer.c (vec_info): Likewise.
4042 * tree-vect-loop.c (vect_transform_loop): Likewise.
4044 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4046 * tree-vect-loop.c (vectorizable_reduction): Assert that the
4047 function is not called for second and subsequent members of
4050 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4052 * tree-vect-loop.c (get_initial_def_for_reduction): Move special
4053 cases for nested loops from here to ...
4054 (vect_create_epilog_for_reduction): ...here. Only call
4055 vect_is_simple_use for inner-loop reductions.
4057 2018-07-31 Martin Liska <mliska@suse.cz>
4059 PR gcov-profile/85338
4060 PR gcov-profile/85350
4061 PR gcov-profile/85372
4062 * profile.c (struct location_triplet): New.
4063 (struct location_triplet_hash): Likewise.
4064 (output_location): Do not output a BB that
4065 is already recorded for a line.
4066 (branch_prob): Use streamed_locations.
4068 2018-07-31 Martin Liska <mliska@suse.cz>
4070 PR gcov-profile/85370
4071 * coverage.c (coverage_begin_function): Do not mark target
4072 clones as artificial functions.
4074 2018-07-31 Martin Liska <mliska@suse.cz>
4076 PR gcov-profile/83813
4077 PR gcov-profile/84758
4078 PR gcov-profile/85217
4079 PR gcov-profile/85332
4080 * profile.c (branch_prob): Do not record GOTO expressions
4081 for GIMPLE statements which locations are already streamed.
4083 2018-07-31 Olivier Hainque <hainque@adacore.com>
4085 * gcc.c (handle_spec_function): Accept a soft_matched_part
4086 argument, as do_spec_1. Pass it down to ...
4087 (eval_spec_function): Accept a soft_matched_part argument,
4088 and pass it down to ...
4089 (do_spec_2): Accept a soft_matched_part argument, and pass
4090 it down to do_spec_1.
4091 (do_spec_1): Pass soft_matched_part to handle_spec_function.
4092 (handle_braces): Update call to handle_spec_function.
4093 (driver::set_up_specs): Update calls to do_spec_2.
4094 (compare_debug_dump_opt_spec_function): Likewise.
4095 (compare_debug_self_opt_spec_function): Likewise.
4097 2018-07-31 Olivier Hainque <hainque@adacore.com>
4099 * common.opt (nolibc): New option.
4100 * doc/invoke.texi (Link Options): Document it.
4101 * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
4102 * config/alpha/linux.h: Likewise.
4103 * config/arc/elf.h: Likewise.
4104 * config/arm/uclinux-elf.h: Likewise.
4105 * config/arm/unknown-elf.h: Likewise.
4106 * config/avr/avrlibc.h: Likewise.
4107 * config/bfin/bfin.h: Likewise.
4108 * config/bfin/linux.h: Likewise.
4109 * config/bfin/uclinux.h: Likewise.
4110 * config/darwin.h: Likewise.
4111 * config/darwin10.h: Likewise.
4112 * config/darwin12.h: Likewise.
4113 * config/gnu-user.h: Likewise.
4114 * config/lm32/uclinux-elf.h: Likewise.
4115 * config/pa/pa-hpux11.h: Likewise.
4116 * config/pa/pa64-hpux.h: Likewise.
4117 * config/sparc/sparc.h: Likewise.
4119 2018-07-31 Olivier Hainque <hainque@adacore.com>
4121 * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
4122 undefined variables.
4124 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
4127 * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
4130 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
4132 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
4133 terminated string literal.
4135 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
4137 PR rtl-optimization/85160
4138 * combine.c (is_just_move): New function.
4139 (try_combine): Allow combining two instructions into two if neither of
4140 the original instructions was a move.
4142 2018-07-30 Alexander Monakov <amonakov@ispras.ru>
4145 * doc/extend.texi (Global Register Variables): Discourage use of type
4147 (Local Register Variables): Likewise.
4149 2018-07-30 Richard Sandiford <richard.sandiford@arm.com>
4151 PR tree-optimization/86506
4152 * hwint.h (ceil_log2): Resync with hwint.c implementation.
4154 2018-07-30 Ilya Leoshkevich <iii@linux.ibm.com>
4157 * lra-constraints.c (spill_hard_reg_in_range): When selecting the
4158 hard_regno, make sure no insn between `from` and `to` clobbers it.
4160 2018-07-30 Cesar Philippidis <cesar@codesourcery.com>
4161 Tom de Vries <tdevries@suse.de>
4163 * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
4164 (PTX_DEFAULT_RUNTIME_DIM): ... this.
4165 (nvptx_goacc_validate_dims): Set default worker and gang dims to
4166 PTX_DEFAULT_RUNTIME_DIM.
4167 (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
4169 2018-07-29 John David Anglin <danglin@gcc.gnu.org>
4171 * config/pa/pa.c (pa_output_addr_vec): Align address table.
4172 * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
4173 * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
4175 2018-07-27 Michael Meissner <meissner@linux.ibm.com>
4177 * config/rs6000/constraints.md (wG constraint): Delete, no longer
4179 * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
4180 predicate to reflect toc fusion has been deleted.
4181 (toc_fusion_mem_raw): Delete, no longer used.
4182 (toc_fusion_mem_wrapped): Likewise.
4183 * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
4185 * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
4186 Delete, no longer used.
4187 * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
4188 meant to be used for toc fusion.
4189 (rs6000_debug_print_mode): Delete toc fusion debugging.
4190 (rs6000_debug_reg_global): Likewise.
4191 (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
4192 fusion and secondary reload support that were never used.
4193 (rs6000_option_override_internal): Delete TOC fusion, that was only
4194 partially defined, and it did not work unless you also used the
4196 (rs6000_legitimate_address_p): Delete TOC fusion support.
4197 (rs6000_opt_masks): Likewise.
4198 (fusion_wrap_memory_address): Delete function, no longer used.
4199 (fusion_split_address); Delete TOC fusion support.
4200 * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
4201 longer used with toc fusion being deleted.
4202 (TARGET_TOC_FUSION_FP): Likewise.
4203 * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
4205 (toc fusion spliter): Delete TOC fusion support.
4206 (toc_fusionload_<mode>): Likewise.
4207 (toc_fusionload_di): Likewise.
4208 (fusion_gpr_load_<mode>): Delete generator function, this insn no
4209 longer needs to be named. Rename predicate to delete TOC fusion.
4210 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
4211 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
4212 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
4213 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
4214 (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
4216 2018-07-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
4218 * doc/extend.texi (Basic PowerPC Built-in Functions Available on
4219 ISA 2.05): Replace __uint128_t with __uint128 and __int128_t with
4220 __int128 in built-in function prototypes.
4221 (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
4222 (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
4224 2018-07-27 Martin Sebor <msebor@redhat.com>
4226 PR tree-optimization/86696
4227 * tree-ssa-strlen.c (get_min_string_length): Handle all integer
4228 types, including enums.
4229 (handle_char_store): Be prepared for the above function to fail.
4231 2018-07-26 Qing Zhao <qing.zhao@oracle.com>
4233 * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
4234 when optimization level is lower than 2 or optimize for size.
4236 2018-07-26 Martin Sebor <msebor@redhat.com>
4238 PR tree-optimization/86043
4239 PR tree-optimization/86042
4240 * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
4241 (get_string_cst_length): Rename...
4242 (get_min_string_length): ...to this. Add argument.
4243 (handle_char_store): Extend to handle multi-character stores by
4245 * tree.c (initializer_zerop): Use new argument. Handle MEM_REF.
4246 * tree.h (initializer_zerop): Add argument.
4248 2018-07-26 Jakub Jelinek <jakub@redhat.com>
4251 * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
4252 declare target to variables if they have always,{to,from,tofrom} map
4255 2018-07-26 Martin Liska <mliska@suse.cz>
4258 * lto-wrapper.c: Add linker_output as prefix
4259 for ltrans_output_file.
4261 2018-07-26 Segher Boessenkool <segher@kernel.crashing.org>
4263 PR rtl-optimization/85805
4264 * combine.c (reg_nonzero_bits_for_combine): Only use the last set
4265 value for hard registers if that was written in the same mode.
4267 2018-07-26 Martin Liska <mliska@suse.cz>
4269 PR gcov-profile/86536
4270 * gcov.c (format_gcov): Use printf format %.*f directly
4271 and do not handle special values.
4273 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
4275 * common/config/arc/arc-common.c (arc_option_optimization_table):
4276 Update default optimizations for size.
4278 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
4280 * config/arc/arc.md (movsf_insn): Add short instruction selection.
4281 * config/arc/constraints.md (CfZ): New constraint.
4282 * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
4283 (subsf3_fpu): Likewise.
4284 (cmpsf_fpu): Likewise.
4285 (cmpsf_fpu_uneq): Likewise.
4287 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
4289 * config/arc/arc.c (compact_memory_operand_p): Check for uncached
4291 (arc_is_uncached_mem_p): uncached applies to both the variable and
4294 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
4296 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
4299 2018-07-25 David Malcolm <dmalcolm@redhat.com>
4301 * optinfo-emit-json.cc (class optrecord_json_writer): Convert
4302 field "m_scopes" from vec to auto_vec.
4304 2018-07-25 Martin Liska <mliska@suse.cz>
4306 * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
4309 2018-07-25 Richard Biener <rguenther@suse.de>
4312 * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
4313 special wrt context_die late.
4314 (gen_subprogram_die): Re-use DIEs in local scope.
4316 2018-07-25 Richard Sandiford <richard.sandiford@arm.com>
4318 PR tree-optimization/86644
4319 * hwint.c (ceil_log2): Fix comment. Return 0 for 0.
4321 2018-07-25 Martin Liska <mliska@suse.cz>
4324 * dumpfile.c: And excluded values with TDF_ALL_VALUES.
4325 * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
4327 2018-07-25 Martin Liska <mliska@suse.cz>
4330 * params.def: Explain ASan abbreviation and provide
4331 a documentation link.
4333 2018-07-24 Martin Sebor <msebor@redhat.com>
4335 PR tree-optimization/86622
4336 PR tree-optimization/86532
4337 * builtins.h (string_length): Declare.
4338 * builtins.c (c_strlen): Correct handling of non-constant offsets.
4339 (check_access): Be prepared for non-constant length ranges.
4340 (string_length): Make extern.
4341 * expr.c (string_constant): Only handle the minor non-constant
4342 array index. Use string_constant to compute the length of
4343 a generic string constant.
4345 2018-07-24 Richard Sandiford <richard.sandiford@arm.com>
4347 PR tree-optimization/86618
4348 * tree-vect-stmts.c (vectorizable_call): Don't take the address
4349 of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
4351 2018-07-24 David Malcolm <dmalcolm@redhat.com>
4353 PR tree-optimization/86636
4354 * json.cc (json::object::set): Fix comment. Add assertions.
4355 (json::array::append): Move here from json.h. Add comment and an
4357 (json::string::string): Likewise.
4358 * json.h (json::array::append): Move to json.cc.
4359 (json::string::string): Likewise.
4360 * optinfo-emit-json.cc
4361 (optrecord_json_writer::impl_location_to_json): Assert that we
4362 aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
4363 wrapper around it. Expand the location once, rather than three
4365 (optrecord_json_writer::inlining_chain_to_json): Fix the check for
4366 UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
4368 (optrecord_json_writer::optinfo_to_json): Likewise, in four
4369 places. Fix some overlong lines.
4371 2018-07-24 Matthew Malcomson <matthew.malcomson@arm.com>
4373 * config/aarch64/aarch64-simd.md
4374 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
4375 (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
4376 (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
4377 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
4378 (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
4379 (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
4380 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
4381 (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
4382 (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
4384 2018-07-24 Jakub Jelinek <jakub@redhat.com>
4387 * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
4388 and size > HOST_BITS_PER_WIDE_INT. For size > HOST_BITS_PER_WIDE_INT
4389 and abs_d == d, do the power of two handling if profitable.
4391 2018-07-24 Richard Biener <rguenther@suse.de>
4393 * match.pd: Add BIT_FIELD_REF canonicalizations.
4395 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
4398 * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
4400 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
4402 * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
4403 terminated STRING_CST object.
4405 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
4407 hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
4409 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
4411 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
4412 * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
4413 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
4414 the elements into a register.
4415 (rs6000_split_v4si_init_di_reg): Delete.
4416 (rs6000_split_v4si_init): Delete.
4417 * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
4418 (vsx_init_v4si): Rewrite as a define_expand.
4420 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
4422 * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
4423 zero_extend argument from memory): New.
4425 2018-07-22 Martin Sebor <msebor@redhat.com>
4428 * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
4429 diagnosing calls with unknown arguments unless -Walloca-larger-than
4430 is restricted to less than PTRDIFF_MAX bytes.
4432 2018-07-22 Gerald Pfeifer <gerald@pfeifer.com>
4434 * doc/gcov.texi (Invoking Gcov): Editorial changes.
4436 2018-07-20 David Malcolm <dmalcolm@redhat.com>
4438 * pretty-print.c (text_info::set_location): Remove redundant
4439 "line_table" parameter from call to rich_location::set_range.
4441 2018-07-20 Martin Sebor <msebor@redhat.com>
4444 * builtins.c (expand_builtin_alloca): Adjust.
4445 * calls.c (alloc_max_size): Simplify.
4446 * cgraphunit.c (cgraph_node::expand): Adjust.
4447 * common.opt (larger_than_size, warn_frame_larger_than): Remove
4449 (frame_larger_than_size): Same.
4450 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
4451 to take a HOST_WIDE_INT argument and accept a byte-size suffix.
4453 * doc/invoke.texi (GCC Command Options): Document option arguments.
4454 Explain byte-size arguments and suffixes.
4455 (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
4456 (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
4457 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
4458 * doc/options.texi (UInteger): Expand.
4459 (Host_Wide_Int, ByteSize): Document new properties.
4460 * final.c (final_start_function_1): Include sizes in an error message.
4461 * function.c (frame_offset_overflow): Same.
4462 * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
4463 (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
4464 Diagnose unbounded alloca calls only for limits of less than
4466 (alloca_call_type): Adjust. Diagnose possibly out-of-bounds alloca
4467 calls and VLA size only for limits of less than PTRDIFF_MAX. Same
4469 (pass_walloca::execute): Adjust. Diagnose alloca calls in loops
4470 only for limits of less than PTRDIFF_MAX.
4471 * langhooks-def.h (lhd_handle_option): Change function argument
4473 * langhooks.c (lhd_handle_option): Same.
4474 * langhooks.h (handle_option): Same.
4475 * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
4477 (var_type, var_type_struct): Same.
4478 (var_set): Handle ByteSize flag.
4479 * optc-gen.awk: Add comments to output to ease debugging. Make
4480 use of HOST_WIDE_INT where appropriate.
4481 * opts-gen-save.awk: Use %lx to format unsigned long.
4482 * opth-gen.awk: Change function argument to HOST_WIDE_INT.
4483 * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
4484 arguments. Parse bytes-size suffixes.
4485 (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
4486 (enum_value_to_arg): Same.
4487 (decode_cmdline_option): Handle cl_host_wide_int. Adjust.
4488 (handle_option): Adjust.
4489 (generate_option): Change function argument to HOST_WIDE_INT.
4490 (cmdline_handle_error): Adjust.
4491 (read_cmdline_option): Change function argument to HOST_WIDE_INT.
4492 (set_option): Change function argument to HOST_WIDE_INT.
4493 (option_enabled): Handle cl_host_wide_int.
4494 (get_option_state): Handle CLVC_SIZE.
4495 (control_warning_option): Same.
4496 * opts.c (common_handle_option): Change function argument to
4497 HOST_WIDE_INT. Remove handling of OPT_Walloca_larger_than_ and
4498 OPT_Wvla_larger_than_.
4499 * opts.h (enum cl_var_type): Add an enumerator.
4500 * stor-layout.c (layout_decl): Print a more meaningful warning.
4501 * toplev.c (output_stack_usage): Adjust.
4503 2018-07-20 Qing Zhao <qing.zhao@oracle.com>
4505 * builtins.c (expand_builtin_memcmp): Delete the last parameter for
4506 call to inline_expand_builtin_string_cmp.
4507 (expand_builtin_strcmp): Likewise.
4508 (expand_builtin_strncmp): Likewise.
4509 (inline_string_cmp): Delete the last parameter, change char_type_node
4510 to unsigned_char_type_node for strcmp/strncmp, add conversions to the
4512 (inline_expand_builtin_string_cmp): Delete the last parameter, give up
4513 the inlining expansion on target where the type of the call has same or
4514 narrower precision than unsigned char.
4516 2018-07-20 David Malcolm <dmalcolm@redhat.com>
4518 * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
4519 (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
4520 * common.opt (fsave-optimization-record): New option.
4521 * coretypes.h (struct kv_pair): Move here from dumpfile.c.
4522 * doc/invoke.texi (-fsave-optimization-record): New option.
4523 * dumpfile.c: Include "optinfo-emit-json.h".
4524 (struct kv_pair): Move to coretypes.h.
4525 (optgroup_options): Make non-static.
4526 (dump_context::end_scope): Call
4527 optimization_records_maybe_pop_dump_scope.
4528 * dumpfile.h (optgroup_options): New decl.
4529 * json.cc: New file.
4531 * optinfo-emit-json.cc: New file.
4532 * optinfo-emit-json.h: New file.
4533 * optinfo.cc: Include "optinfo-emit-json.h".
4534 (optinfo::emit): Call optimization_records_maybe_record_optinfo.
4535 (optinfo_enabled_p): Check optimization_records_enabled_p.
4536 (optinfo_wants_inlining_info_p): Likewise.
4537 * optinfo.h: Update comment.
4538 * profile-count.c (profile_quality_as_string): New function.
4539 * profile-count.h (profile_quality_as_string): New decl.
4540 (profile_count::quality): New accessor.
4541 * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
4542 and optinfo_emit_json_cc_tests.
4543 * selftest.h (selftest::json_cc_tests): New decl.
4544 (selftest::optinfo_emit_json_cc_tests): New decl.
4545 * toplev.c: Include "optinfo-emit-json.h".
4546 (compile_file): Call optimization_records_finish.
4547 (do_compile): Call optimization_records_start.
4548 * tree-ssa-live.c: Include optinfo.h.
4549 (remove_unused_scope_block_p): Retain inlining information if
4550 optinfo_wants_inlining_info_p returns true.
4552 2018-07-20 Richard Biener <rguenther@suse.de>
4555 * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
4556 to cover -flto-partition=none.
4558 2018-07-20 Martin Liska <mliska@suse.cz>
4560 * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
4561 (get_decl_source_range): Remove unused function.
4563 2018-07-20 Richard Biener <rguenther@suse.de>
4565 * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
4566 (struct vn_phi_s): Likewise.
4567 (struct vn_reference_s): Likewise.
4568 * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
4569 for searching the slot of an entry known to be in the hash itself.
4570 (vn_phi_hasher::equal): Likewise.
4571 (vn_reference_hasher::equal): Likewise.
4572 (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
4574 (optimistic_info, current_info): Remove, keeping only valid_info.
4575 (vn_reference_lookup_1): Remove fallback lookup.
4576 (vn_reference_lookup_2): Likewise.
4577 (vn_nary_op_lookup_1): Likewise.
4578 (vn_phi_lookup): Likewise.
4579 (vn_nary_build_or_lookup_1): Make sure to not chain the built
4581 (vn_reference_insert): Adjust, chain the inserted hash element
4582 at last_inserted_ref.
4583 (vn_reference_insert_pieces): Likewise.
4584 (visit_reference_op_call): Likewise.
4585 (vn_nary_op_insert_into): Chain the inserted hash element at
4587 (vn_nary_op_insert_pieces): Adjust.
4588 (vn_nary_op_insert): Likewise.
4589 (vn_nary_op_insert_stmt): Likewise.
4590 (vn_phi_insert): Adjust, chain the inserted hash element at
4592 (process_scc): Remove clearing and copying the optimistic
4593 table. Instead remove elements inserted during an optimistic
4594 iteration from the single table we maintain.
4595 (init_scc_vn): Adjust.
4596 (free_scc_vn): Likewise.
4597 (sccvn_dom_walker::record_cond): Likewise.
4598 (sccvn_dom_walker::after_dom_children): Likewise.
4600 2018-07-19 Martin Sebor <msebor@redhat.com>
4602 PR tree-optimization/84047
4603 PR tree-optimization/83776
4604 * tree-vrp.c (vrp_prop::check_mem_ref): New function.
4605 (check_array_bounds): Call it.
4607 2018-07-19 Martin Sebor <msebor@redhat.com>
4609 * align.h (align_flags): Use member initialization.
4611 2018-07-19 David Malcolm <dmalcolm@redhat.com>
4613 * Makefile.in (OBJS): Add optinfo.o.
4614 * coretypes.h (class symtab_node): New forward decl.
4615 (struct cgraph_node): New forward decl.
4616 (class varpool_node): New forward decl.
4617 * dump-context.h: New file.
4618 * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
4620 (refresh_dumps_are_enabled): Use optinfo_enabled_p.
4621 (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
4622 (set_alt_dump_file): Likewise.
4623 (dump_context::~dump_context): New dtor.
4624 (dump_gimple_stmt): Move implementation to...
4625 (dump_context::dump_gimple_stmt): ...this new member function.
4626 Add the stmt to any pending optinfo, creating one if need be.
4627 (dump_gimple_stmt_loc): Move implementation to...
4628 (dump_context::dump_gimple_stmt_loc): ...this new member function.
4629 Start a new optinfo and add the stmt to it.
4630 (dump_gimple_expr): Move implementation to...
4631 (dump_context::dump_gimple_expr): ...this new member function.
4632 Add the stmt to any pending optinfo, creating one if need be.
4633 (dump_gimple_expr_loc): Move implementation to...
4634 (dump_context::dump_gimple_expr_loc): ...this new member function.
4635 Start a new optinfo and add the stmt to it.
4636 (dump_generic_expr): Move implementation to...
4637 (dump_context::dump_generic_expr): ...this new member function.
4638 Add the tree to any pending optinfo, creating one if need be.
4639 (dump_generic_expr_loc): Move implementation to...
4640 (dump_context::dump_generic_expr_loc): ...this new member
4641 function. Add the tree to any pending optinfo, creating one if
4643 (dump_printf): Move implementation to...
4644 (dump_context::dump_printf_va): ...this new member function. Add
4645 the text to any pending optinfo, creating one if need be.
4646 (dump_printf_loc): Move implementation to...
4647 (dump_context::dump_printf_loc_va): ...this new member function.
4648 Start a new optinfo and add the stmt to it.
4649 (dump_dec): Move implementation to...
4650 (dump_context::dump_dec): ...this new member function. Add the
4651 value to any pending optinfo, creating one if need be.
4652 (dump_context::dump_symtab_node): New member function.
4653 (dump_context::get_scope_depth): New member function.
4654 (dump_context::begin_scope): New member function.
4655 (dump_context::end_scope): New member function.
4656 (dump_context::ensure_pending_optinfo): New member function.
4657 (dump_context::begin_next_optinfo): New member function.
4658 (dump_context::end_any_optinfo): New member function.
4659 (dump_context::s_current): New global.
4660 (dump_context::s_default): New global.
4661 (dump_scope_depth): Delete global.
4662 (dumpfile_ensure_any_optinfo_are_flushed): New function.
4663 (dump_symtab_node): New function.
4664 (get_dump_scope_depth): Reimplement in terms of dump_context.
4665 (dump_begin_scope): Likewise.
4666 (dump_end_scope): Likewise.
4667 (selftest::temp_dump_context::temp_dump_context): New ctor.
4668 (selftest::temp_dump_context::~temp_dump_context): New dtor.
4669 (selftest::verify_item): New function.
4670 (ASSERT_IS_TEXT): New macro.
4671 (ASSERT_IS_TREE): New macro.
4672 (ASSERT_IS_GIMPLE): New macro.
4673 (selftest::test_capture_of_dump_calls): New test.
4674 (selftest::dumpfile_c_tests): Call it.
4675 * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
4676 (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
4677 (dump_gimple_stmt, dump_dec): Gather these related decls and add a
4678 descriptive comment.
4679 (dump_function, print_combine_total_stats, enable_rtl_dump_file)
4680 (dump_node, dump_bb): Move these unrelated decls.
4681 (class dump_manager): Add leading comment.
4682 * optinfo.cc: New file.
4683 * optinfo.h: New file.
4685 2018-07-19 Michael Collison <michael.collison@arm.com>
4686 Richard Henderson <rth@redhat.com>
4688 * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
4689 (subti): Handle op1 zero.
4690 (subvti4, usub4ti4): New.
4691 (*sub<GPI>3_compare1_imm): New.
4692 (sub<GPI>3_carryinCV): New.
4693 (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
4694 (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
4696 2018-07-19 Michael Collison <michael.collison@arm.com>
4697 Richard Henderson <rth@redhat.com>
4699 * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
4700 (addti3): Create simpler code if low part is already known to be 0.
4701 (addvti4, uaddvti4): New.
4702 (*add<GPI>3_compareC_cconly_imm): New.
4703 (*add<GPI>3_compareC_cconly): New.
4704 (*add<GPI>3_compareC_imm): New.
4705 (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
4706 handle constants within this pattern..
4707 (*add<GPI>3_compareV_cconly_imm): New.
4708 (*add<GPI>3_compareV_cconly): New.
4709 (*add<GPI>3_compareV_imm): New.
4710 (add<GPI>3_compareV): New.
4711 (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
4712 (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
4713 (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
4714 ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
4715 with 'comparison' operator.
4716 (*add<GPI>3_compareV_cconly_imm): Ditto.
4717 (*add<GPI>3_compareV_cconly): Ditto.
4718 (*add<GPI>3_compareV_imm): Ditto.
4719 (add<GPI>3_compareV): Ditto.
4720 (add<mode>3_carryinC): Ditto.
4721 (*add<mode>3_carryinC_zero): Ditto.
4722 (*add<mode>3_carryinC): Ditto.
4723 (add<mode>3_carryinV): Ditto.
4724 (*add<mode>3_carryinV_zero): Ditto.
4725 (*add<mode>3_carryinV): Ditto.
4727 2018-07-19 Michael Collison <michael.collison@arm.com>
4728 Richard Henderson <rth@redhat.com>
4730 * config/aarch64/aarch64-modes.def (CC_V): New.
4731 * config/aarch64/aarch64-protos.h
4732 (aarch64_addti_scratch_regs): Declare
4733 (aarch64_subvti_scratch_regs): Declare.
4734 (aarch64_expand_subvti): Declare.
4735 (aarch64_gen_unlikely_cbranch): Declare
4736 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
4737 for signed overflow using CC_Vmode.
4738 (aarch64_get_condition_code_1): Handle CC_Vmode.
4739 (aarch64_gen_unlikely_cbranch): New function.
4740 (aarch64_addti_scratch_regs): New function.
4741 (aarch64_subvti_scratch_regs): New function.
4742 (aarch64_expand_subvti): New function.
4744 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
4746 * config/aarch64/aarch64-option-extensions.def: New entry for profile
4748 * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
4749 * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
4752 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
4755 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
4756 address check not strict.
4758 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
4760 * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
4762 (store_pair_lanes<mode>): Likewise.
4763 * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
4764 enum value 'ADDR_QUERY_LDP_STP_N'.
4765 * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
4766 (aarch64_print_address_internal): Add declaration.
4767 (aarch64_print_ldpstp_address): Remove.
4768 (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
4769 (aarch64_print_operand): Change printing of 'y'.
4770 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
4771 new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
4772 'true' rather than '1'.
4773 * config/aarch64/constraints.md (Uml): Likewise.
4774 (Uml): Rename to Umn.
4777 2018-07-19 Richard Biener <rguenther@suse.de>
4779 * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
4781 * tree-ssa-sccvn.c: Remove alloc-pool.h use.
4782 (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
4783 (vn_reference_hasher): Likewise.
4784 (struct vn_tables_s): Remove obstack and alloc-pool members.
4785 (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
4786 (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
4787 (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
4788 (vn_reference_insert_pieces): Likewise.
4789 (alloc_vn_nary_op_noinit): Adjust.
4790 (vn_nary_op_insert_stmt): Allocate phiargs in-place.
4791 (vn_phi_eq): Adjust.
4792 (shared_lookup_phiargs): Remove.
4793 (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
4794 (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
4795 (visit_reference_op_call): Likewise.
4796 (copy_nary, copy_phi, copy_reference): Remove.
4797 (process_scc): Rewind the obstack when iterating. Do not
4798 copy the elements to valid_info but just move them from one
4799 hashtable to the other.
4800 (allocate_vn_table): Adjust.
4801 (free_vn_table): Likewise.
4802 (init_scc_vn): Likewise.
4803 (free_scc_vn): Likewise.
4805 2018-07-19 H.J. Lu <hongjiu.lu@intel.com>
4808 * config/i386/i386.c (rest_of_insert_endbranch): Lookup
4809 indirect_return as function type attribute.
4810 (ix86_attribute_table): Change indirect_return to function
4812 * doc/extend.texi: Update indirect_return attribute.
4814 2018-07-19 Aldy Hernandez <aldyh@redhat.com>
4816 * wide-int.h (widest2_int): New.
4817 * gimple-fold.c (arith_overflowed_p): Use it.
4818 * tree.h (widest2_int_cst): New.
4819 * tree-vrp.c (wide_int_binop_overflow): Rename from
4820 vrp_int_const_binop.
4821 Rewrite to work on trees.
4822 (extract_range_from_multiplicative_op_1): Abstract code to...
4823 (wide_int_range_min_max): ...here.
4824 (wide_int_range_cross_product): ...and here.
4825 (extract_range_from_binary_expr_1): Abstract overflow code to...
4826 (wide_int_range_mult_wrapping): ...here.
4827 * tree-vrp.h (wide_int_range_cross_product): New.
4828 (wide_int_range_mult_wrapping): New.
4830 2018-07-19 Andrew Senkevich <andrew.senkevich@intel.com>
4831 Julia Koval <julia.koval@intel.com>
4833 * config/i386/x86-tune-costs.h (skylake_memcpy,
4834 skylake_memset): Replace rep_prefix with unrolling for size 512.
4836 2018-07-18 Kugan Vivekanandarajah <kuganv@linaro.org>
4839 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
4840 comparision with EQ_EXPR in last stmt.
4842 2018-07-18 Kelvin Nilsen <kelvin@gcc.gnu.org>
4844 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
4845 this subsection to "PowerPC AltiVec/VSX Built-in Functions".
4846 (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
4847 previously known as "PowerPC AltiVec Built-in Functions". Move
4848 some material to new subsubsections "PowerPC AltiVec Built-in
4849 Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
4851 (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
4852 (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
4853 (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
4854 (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
4856 2018-07-18 Richard Biener <rguenther@suse.de>
4858 PR tree-optimization/86557
4859 * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
4862 2018-07-18 Ilya Leoshkevich <iii@linux.ibm.com>
4864 * config/s390/s390.c (s390_function_profiler): Generate CFI.
4866 2018-07-17 Jeff Law <law@redhat.com>
4868 * config/arm/arm.c (get_label_padding): Update for recent
4869 changes to label_to_alignment.
4871 PR tree-optimization/86010
4872 * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
4874 * config/mips/mips.c (vr4130_align_insns): Update for recent
4875 changes to label_to_alignment.
4877 * config/frv/frv.c (frv_label_align): Update for recent changes
4878 to label_to_alignment.
4880 * config/nios2/nios2.c (nios2_label_align): Update for recent
4881 changes which dropped ALIGN_LABELS_LOG.
4883 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
4885 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
4886 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
4888 2018-07-17 Claudiu Zissulescu <claziss@synopsys.com>
4890 * config/arc/arc.c (arc_label_align): Use align_labels instead of
4891 deprecated align_labels_log.
4893 2018-07-17 Richard Biener <rguenther@suse.de>
4896 * dwarf2out.c (init_sections_and_labels): Always generate
4897 a debug_line_str_section for early LTO debug.
4898 (dwarf2out_finish): Reset debug_line_str_hash output early.
4899 Bump counter for extra dwarf5 .debug_loc labels to not conflict
4901 (dwarf2out_early_finish): Output debug_line_str.
4903 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
4905 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
4906 index register on z196 or later.
4908 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
4910 * config/s390/s390.c (s390_default_align): Set default function
4912 (s390_override_options_after_change): Call s390_default align.
4913 (s390_option_override_internal): Call s390_default_align.
4914 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
4916 2018-07-17 Jakub Jelinek <jakub@redhat.com>
4919 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
4920 to _looptemp_ clauses, other than the first two.
4922 2018-07-17 Martin Liska <mliska@suse.cz>
4924 * opts.c: Do not enable OPT_falign_* for -Os.
4926 2018-07-17 Martin Liska <mliska@suse.cz>
4928 * align.h (MAX_CODE_ALIGN): New.
4929 (MAX_CODE_ALIGN_VALUE): New.
4930 * common/config/i386/i386-common.c (ix86_handle_option):
4931 (MAX_CODE_ALIGN): Moved to align.h.
4932 * final.c (MAX_CODE_ALIGN): Likewise.
4933 * opts.c (parse_and_check_align_values):
4934 (MAX_CODE_ALIGN): Likewise.
4935 (MAX_CODE_ALIGN_VALUE): Likewise.
4937 2018-07-17 Martin Liska <mliska@suse.cz>
4939 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
4940 in order to fulfil coding style.
4941 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
4942 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4943 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
4944 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
4945 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
4946 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
4947 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
4948 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
4949 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4951 2018-07-17 Martin Liska <mliska@suse.cz>
4953 * align.h: New file.
4954 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
4956 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
4957 align_flags of label_to_alignment.
4958 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
4960 * config/m68k/m68k.c: Do not use removed align_labels_value and
4962 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
4963 (LOOP_ALIGN): Likewise.
4964 (LABEL_ALIGN): Likewise.
4965 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
4966 Remove not used macro.
4967 (rs6000_loop_align): Change return type to align_flags.
4968 (rs6000_loop_align_max_skip): Remove.
4969 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
4970 Change return type to align_flags.
4971 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
4972 Remove not used macro.
4973 (rs6000_loop_align): Change return type to align_flags.
4974 (rs6000_loop_align_max_skip): Remove.
4975 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
4976 * config/rx/rx-protos.h (rx_align_for_label): Make it
4978 * config/rx/rx.c (rx_align_for_label): Change return type
4980 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
4982 into align_flags class.
4983 (LABEL_ALIGN): Likewise.
4984 (LOOP_ALIGN): Likewise.
4985 * config/s390/s390.c (s390_label_align): Use align_flags
4987 (s390_asm_output_function_label): Likewise.
4988 * config/sh/sh.c (sh_override_options_after_change):
4989 Use align_flags class directly without macros.
4990 (find_barrier): Likewise.
4991 (barrier_align): Likewise.
4992 (sh_loop_align): Likewise.
4993 * config/spu/spu.c (spu_option_override):
4994 Use align_flags_tuple::get_value instead of removed macros.
4995 (spu_sched_init): Likewise.
4996 * config/spu/spu.h (GTY): Likewise.
4997 * config/visium/visium.c (visium_option_override):
4998 Set "8" as default secondary alignment.
4999 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
5000 in order to guarantee secondary alignment of 8.
5001 * coretypes.h: Include align.h header file.
5002 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
5003 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
5004 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
5005 * doc/tm.texi.in: Likewise.
5006 * final.c (struct label_alignment): Remove not used structure.
5007 (LABEL_ALIGN): Change type to align_flags.
5008 (LOOP_ALIGN): Likewise.
5009 (JUMP_ALIGN): Likewise.
5010 (default_loop_align_max_skip): Remove.
5011 (default_label_align_max_skip): Likewise.
5012 (default_jump_align_max_skip): Likewise.
5013 (default_label_align_after_barrier_max_skip):
5014 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
5015 (LABEL_TO_MAX_SKIP): Remove.
5016 (label_to_alignment): Return align_flags type instead of integer.
5017 (label_to_max_skip): Remove.
5018 (align_fuzz): Use align_flags type.
5019 (compute_alignments): Use align_flags type and use align_flags::max
5020 to combine multiple alignments.
5021 (grow_label_align): Grow vec instead of C array.
5022 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
5023 (shorten_branches): Use align_flags type and use align_flags::max
5024 to combine multiple alignments.
5025 (final_scan_insn_1): Remove usage of secondary alignment that comes
5026 from label alignment, but instead use proper secondary alignment
5027 which is computed in grow_label_align.
5028 * flags.h (struct align_flags_tuple): Move to align.h.
5029 (struct align_flags): Likewise.
5030 (state_align_loops): Rename to align_loops.
5031 (state_align_jumps): Rename to align_jumps.
5032 (state_align_labels): Rename to align_labels.
5033 (state_align_functions): Rename to align_functions.
5034 (align_loops_log): Remove.
5035 (align_jumps_log): Remove.
5036 (align_labels_log): Remove.
5037 (align_functions_log): Remove.
5038 (align_loops_max_skip): Remove.
5039 (align_jumps_max_skip): Remove.
5040 (align_labels_max_skip): Remove.
5041 (align_functions_max_skip): Remove.
5042 (align_loops_value): Remove.
5043 (align_jumps_value): Remove.
5044 (align_labels_value): Remove.
5045 (align_functions_value): Remove.
5046 * output.h (label_to_alignment): Change return type to align_flags.
5047 (label_to_max_skip): Remove.
5048 * target.def: Remove loop_align_max_skip, label_align_max_skip,
5049 jump_align_max_skip macros.
5050 * targhooks.h (default_loop_align_max_skip): Remove.
5051 (default_label_align_max_skip): Likewise.
5052 (default_jump_align_max_skip): Likewise.
5053 (default_label_align_after_barrier_max_skip): Remove.
5054 * toplev.c (read_log_maxskip): Use ::normalize function.
5055 (parse_N_M): Remove not used argument and also call ::normalize.
5056 (parse_alignment_opts): Do not pass unused arguments.
5057 * varasm.c (assemble_start_function): Use directly align_functions
5058 instead of removed macros.
5059 * system.h: Do not poison removed macros.
5061 2018-07-17 Jakub Jelinek <jakub@redhat.com>
5064 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
5065 and cond temporaries don't have reference type if iterator has
5066 pointer type. For init use &for_pre_body instead of pre_p if
5067 for_pre_body is non-empty.
5069 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
5071 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
5072 double-double modes to SFmode directly directly.
5073 (trunc<mode>sf2_fprs): Delete.
5075 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
5077 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
5078 for conversions between IFmode and the decimal floating point modes.
5079 (init_float128_ieee): Use the correct names for conversions between
5080 KFmode and the decimal floating point modes.
5082 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
5084 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
5085 for the conversions between TDmode and IFmode.
5086 (init_float128_ieee): Use more correct names for the conversions
5087 between TDmode and KFmode.
5089 2018-07-16 Jakub Jelinek <jakub@redhat.com>
5091 PR tree-optimization/86526
5092 * builtins.c (expand_builtin_memcmp): Formatting fixes.
5093 (inline_expand_builtin_string_cmp): Likewise.
5094 (inline_string_cmp): Likewise. Use c_readstr instead of
5095 builtin_memcpy_read_str. Add unit_mode temporary.
5097 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
5100 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
5101 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
5103 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
5105 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
5106 Alphabetize prototypes of built-in functions, separating out
5107 built-in functions that are listed in this section but should be
5108 described elsewhere.
5110 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
5113 * expmed.c (emit_store_flag): Do not emit setcc followed by a
5114 conditional move when trapping comparison was split to a
5115 non-trapping one (and vice versa).
5117 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
5119 * config/s390/s390.c (s390_function_profiler): Generate nops
5120 instead of profiler call sequences.
5121 * config/s390/s390.opt: Add the new option.
5123 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
5125 * config/s390/s390.c (s390_function_profiler): Generate
5126 __mcount_loc section.
5127 * config/s390/s390.opt: Add the new option.
5129 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
5131 * common.opt: Add the new warning.
5132 * config/s390/s390.c (s390_function_profiler): Emit "brasl
5133 %r0,__fentry__" when -mfentry is specified.
5134 (s390_option_override_internal): Disallow -mfentry for 31-bit
5136 * config/s390/s390.opt: Add the new option.
5138 2018-07-16 Richard Biener <rguenther@suse.de>
5141 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
5142 for function-local FUNCTION_DECL and RESULT_DECL immediately.
5144 2018-07-16 Martin Liska <mliska@suse.cz>
5147 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
5150 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
5152 * config/arc/arcHS.md: Update ARCHS scheduling rules.
5154 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
5156 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
5158 * config/arc/arc-cpus.def (hs4x): New cpu.
5160 * config/arc/arc-tables.opt: Regenerate.
5161 * config/arc/arc.c (arc_sched_issue_rate): New function.
5162 (TARGET_SCHED_ISSUE_RATE): Define.
5163 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
5164 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
5166 (attr tune): Add ARCHS4x tune values.
5167 (attr tune_dspmpy): Define.
5168 (*tst): Correct instruction type.
5169 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
5170 * config/arc/arcHS4x.md: New file.
5171 * config/arc/fpu.md: Update instruction type attributes.
5172 * config/arc/t-multilib: Regenerate.
5174 2018-07-16 Tom de Vries <tdevries@suse.de>
5177 * var-tracking.c (vt_initialize): Fix pre_dec handling.
5179 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5181 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
5184 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
5186 PR tree-optimization/86514
5187 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
5188 conversion to a boolean type from a type with greater precision.
5190 2018-07-16 Tom de Vries <tdevries@suse.de>
5192 * var-tracking.c (vt_initialize): Print adjusted insn slim if
5193 dump_flags request TDF_SLIM.
5195 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
5197 * fold-const.c (int_const_binop_1): Abstract...
5198 (wide_int_binop): ...wide int code here.
5199 (poly_int_binop): ...poly int code here.
5200 Abstract the rest of int_const_binop_1 into int_const_binop.
5201 * fold-const.h (wide_int_binop): New.
5202 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
5203 Remove useless PLUS/MINUS_EXPR case.
5204 (zero_nonzero_bits_from_vr): Move wide int code...
5205 (zero_nonzero_bits_from_bounds): ...here.
5206 (extract_range_from_binary_expr_1): Move mask optimization code...
5207 (range_easy_mask_min_max): ...here.
5208 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
5209 (range_easy_mask_min_max): New.
5211 2018-07-15 Jeff Law <law@redhat.com>
5214 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
5217 2018-07-14 Jim Wilson <jimw@sifive.com>
5219 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
5221 2018-07-14 Paul Koning <ni1d@arrl.net>
5223 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
5225 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
5227 * lto-streamer-out.c (copy_function_or_variable): Dump info about
5230 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
5231 Steve Munroe <munroesj52@gmail.com>
5233 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
5234 (_mm_andnot_si128): Likewise.
5235 (_mm_or_si128): Likewise.
5236 (_mm_xor_si128): Likewise.
5238 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
5241 * builtins.c (expand_builtin_memcmp): Inline the calls first
5242 when result_eq is false.
5243 (expand_builtin_strcmp): Inline the calls first.
5244 (expand_builtin_strncmp): Likewise.
5245 (inline_string_cmp): New routine. Expand a string compare
5246 call by using a sequence of char comparison.
5247 (inline_expand_builtin_string_cmp): New routine. Inline expansion
5248 a call to str(n)cmp/memcmp.
5249 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
5251 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
5253 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
5255 * config/arm/driver-arm.c: Include arm-native.h.
5256 (host_detect_local_cpu): Use auto-generated data tables.
5257 (vendors, arm_cpu_table): Delete. Move part information to ...
5258 * config/arm/arm-cpus.in: ... here.
5259 * config/arm/parsecpu.awk (gen_native): New function.
5260 (vendor, part): New CPU fields.
5261 (END): Add support for building the native CPU detection tables.
5262 * config/arm/t-arm (arm-native.h): Add build rule.
5263 (driver-arm.o): Add dependency on arm-native.h.
5265 2018-07-13 Richard Biener <rguenther@suse.de>
5268 * match.pd (addr1 - addr2): Allow either of the operand to
5271 2018-07-13 Tom de Vries <tdevries@suse.de>
5273 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
5276 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
5278 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
5279 arrays instead of numbered variables.
5281 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
5283 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
5284 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
5286 2018-07-13 Richard Biener <rguenther@suse.de>
5289 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
5290 instead of get_context_die.
5292 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
5293 Richard Biener <rguenther@suse.de>
5296 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
5297 that the loop latch destination where phi is defined.
5299 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
5301 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
5302 (riscv_expand_epilogue): Add assertion to check interrupt mode.
5303 (riscv_set_current_function): Extract getting interrupt type to new
5305 (riscv_get_interrupt_type): New function.
5306 (riscv_merge_decl_attributes): New function, checking interrupt type is
5308 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
5310 2018-07-12 Paul Koning <ni1d@arrl.net>
5312 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
5315 2018-07-12 Paul Koning <ni1d@arrl.net>
5317 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
5318 zero reference, add doloop_end instead.
5319 * doc/md.texi (decrement_and_branch_until_zero): Remove.
5320 (Looping patterns): Remove decrement_and_branch_until_zero. Add
5321 detail for doloop_end.
5323 2018-07-12 Martin Sebor <msebor@redhat.com>
5326 * attribs.c (decl_attributes): Reject conflicting attributes before
5327 calling attribute handlers.
5329 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
5331 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
5333 (gcc::dump_manager::get_dump_file_name): likewise.
5334 (dump_begin): Likewise.
5335 * dumpfile.h (dump_begin): Update prototype.
5336 (gcc::dump_manager::get_dump_file_name,
5337 gcc::dump_manager::get_dump_file_name): Update prototype.
5339 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
5341 * internal-fn.h (vectorizable_internal_fn_p): New function.
5342 * tree-vect-slp.c (compatible_calls_p): Likewise.
5343 (vect_build_slp_tree_1): Remove nops argument. Handle calls
5344 to internal functions.
5345 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
5347 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
5349 * fold-const.h (inverse_conditions_p): Declare.
5350 * fold-const.c (inverse_conditions_p): New function.
5351 * match.pd: Use inverse_conditions_p. Add folds of view_converts
5352 that test the inverse condition of a conditional internal function.
5353 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
5354 * internal-fn.c (internal_fn_mask_index): Handle conditional
5356 (vectorized_internal_fn_supported_p): New function.
5357 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
5358 (any_pred_load_store): Replace with...
5359 (need_to_predicate): ...this new variable.
5360 (redundant_ssa_names): New variable.
5361 (ifcvt_can_use_mask_load_store): Move initial checks to...
5362 (ifcvt_can_predicate): ...this new function. Handle tree codes
5363 for which a conditional internal function exists.
5364 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
5365 instead of ifcvt_can_use_mask_load_store. Update after variable
5367 (predicate_load_or_store): New function, split out from
5368 predicate_mem_writes.
5369 (check_redundant_cond_expr): New function.
5370 (value_available_p): Likewise.
5371 (predicate_rhs_code): Likewise.
5372 (predicate_mem_writes): Rename to...
5373 (predicate_statements): ...this. Use predicate_load_or_store
5374 and predicate_rhs_code.
5375 (combine_blocks, tree_if_conversion): Update after above name changes.
5376 (ifcvt_local_dce): Handle redundant_ssa_names.
5377 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
5378 general conditional functions.
5379 * tree-vect-stmts.c (vectorizable_call): Likewise.
5381 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
5382 Alan Hayward <alan.hayward@arm.com>
5383 David Sherwood <david.sherwood@arm.com>
5385 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
5386 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
5387 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
5388 plus and minus and convert them into IFN_COND_FMA-based sequences.
5389 (convert_mult_to_fma): Handle conditional plus and minus.
5391 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
5393 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
5394 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
5395 (cond_fnms_optab): New optabs.
5396 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
5398 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
5399 * internal-fn.h (get_conditional_internal_fn): Declare.
5400 (get_unconditional_internal_fn): Likewise.
5401 * internal-fn.c (cond_ternary_direct): New macro.
5402 (expand_cond_ternary_optab_fn): Likewise.
5403 (direct_cond_ternary_optab_supported_p): Likewise.
5404 (FOR_EACH_COND_FN_PAIR): Likewise.
5405 (get_conditional_internal_fn): New function.
5406 (get_unconditional_internal_fn): Likewise.
5407 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
5408 (gimple_match_op::gimple_match_op): Add a new overload for 5
5410 (gimple_match_op::set_op): Likewise.
5411 (gimple_resimplify5): Declare.
5412 * genmatch.c (decision_tree::gen): Generate simplifications for
5414 * gimple-match-head.c (gimple_simplify): Define an overload for
5415 5 operands. Handle calls with 5 arguments in the top-level overload.
5416 (convert_conditional_op): Handle conversions from unconditional
5417 internal functions to conditional ones.
5418 (gimple_resimplify5): New function.
5419 (build_call_internal): Pass a fifth operand.
5420 (maybe_push_res_to_seq): Likewise.
5421 (try_conditional_simplification): Try converting conditional
5422 internal functions to unconditional internal functions.
5423 Handle 3-operand unconditional forms.
5424 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
5425 Define ternary equivalents of the current rules for binary conditional
5427 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
5429 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
5430 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
5431 (optab): Handle them.
5432 (SVE_COND_FP_TERNARY): New int iterator.
5433 (sve_fmla_op, sve_fmad_op): New int attributes.
5434 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
5435 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
5436 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
5438 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
5440 * target.def (preferred_else_value): New target hook.
5441 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
5442 * doc/tm.texi: Regenerate.
5443 * targhooks.h (default_preferred_else_value): Declare.
5444 * targhooks.c (default_preferred_else_value): New function.
5445 * internal-fn.h (conditional_internal_fn_code): Declare.
5446 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
5447 (get_conditional_internal_fn): Use it.
5448 (conditional_internal_fn_code): New function.
5449 * gimple-match.h (gimple_match_cond): New struct.
5450 (gimple_match_op): Add a cond member function.
5451 (gimple_match_op::gimple_match_op): Update all forms to take a
5453 * genmatch.c (expr::gen_transform): Use the same condition as res_op
5454 for the suboperation, but don't specify a particular else_value.
5455 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
5456 (visit_nary_op, visit_reference_op_load): Pass
5457 gimple_match_cond::UNCOND to the gimple_match_op constructor.
5458 * gimple-match-head.c: Include tree-eh.h
5459 (convert_conditional_op): New function.
5460 (maybe_resimplify_conditional_op): Likewise.
5461 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
5462 (gimple_resimplify2): Likewise.
5463 (gimple_resimplify3): Likewise.
5464 (gimple_resimplify4): Likewise.
5465 (maybe_push_res_to_seq): Return null for conditional operations.
5466 (try_conditional_simplification): New function.
5467 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
5469 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
5471 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
5473 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
5475 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
5477 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
5479 (hash_tree): Do not hash DECL_FCONTEXT
5480 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
5481 Do not stream DECL_FCONTEXT.
5482 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
5483 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
5485 2018-07-12 Richard Biener <rguenther@suse.de>
5488 * dwarf2out.c (gen_block_die): Only output blocks when they have
5489 at least one !DECL_IGNORED_P variable.
5491 2018-07-12 Richard Biener <rguenther@suse.de>
5494 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
5495 Remove -mieee-fp handling.
5497 2018-07-12 Richard Biener <rguenther@suse.de>
5499 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
5500 left-over from last patch.
5502 2018-07-12 Jakub Jelinek <jakub@redhat.com>
5504 PR tree-optimization/86492
5505 * gimple-ssa-store-merging.c
5506 (imm_store_chain_info::coalesce_immediate_stores): Call
5507 check_no_overlap even for the merge_overlapping case. Formatting fix.
5509 2018-07-12 Richard Biener <rguenther@suse.de>
5512 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
5513 move possibly trapping operations into the conditional.
5515 2018-07-12 Richard Biener <rguenther@suse.de>
5517 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
5518 (vn_lookup_simplify_result): Remove recursion limit applied
5520 (vn_nary_build_or_lookup_1): Adjust.
5521 (try_to_simplify): Likewise.
5522 * gimple-match-head.c (gimple_resimplify1): Instead apply one
5524 (gimple_resimplify2): Likewise.
5525 (gimple_resimplify3): Likewise.
5526 (gimple_resimplify4): Likewise.
5528 2018-07-11 Jakub Jelinek <jakub@redhat.com>
5530 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
5531 Use __mmask64 type instead of __mmask8 for __M argument.
5532 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
5533 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
5535 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
5536 __mmask16 for __M argument.
5537 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
5538 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
5539 to __mmask16 instead of __mmask8.
5540 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
5541 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
5542 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
5543 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
5544 instead of __mmask16 for __U argument.
5545 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
5546 __mmask16 instead of __mmask8 for __U argument.
5547 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
5549 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
5551 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
5553 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
5555 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
5557 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
5559 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
5561 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
5563 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
5565 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
5567 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
5568 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
5569 return type as well as __M argument type and all casts from __mmask8
5571 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
5572 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
5573 return type as well as __M argument type and all casts from __mmask8
5575 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
5576 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
5577 return type as well as __M argument type and all casts from __mmask8
5579 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
5580 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
5581 return type as well as __M argument type and all casts from __mmask8
5583 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
5584 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
5587 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
5589 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
5590 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
5593 2018-07-11 Paul Koning <ni1d@arrl.net>
5595 * doc/md.texi (define_subst): Document how multiple occurrences of
5596 the same argument in the replacement pattern are handled.
5598 2018-07-11 Paul Koning <ni1d@arrl.net>
5600 * doc/extend.texi (Common Variable Attributes): Move "mode" into
5602 (Common Type Attributes): Add "mode" attribute.
5604 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
5606 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
5607 stream DECL_ORIGINAL_TYPE.
5608 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
5609 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
5610 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
5611 Do not walk original type.
5612 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
5614 (write_ts_decl_non_common_tree_pointers): Do not stream
5616 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
5617 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
5619 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
5621 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
5622 thread twice from the same starting edge.
5624 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
5626 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
5627 * gimple.c (gimple_call_nonnull_result_p): ...here...
5628 (gimple_call_nonnull_arg): ...and here.
5629 * gimple.h (gimple_call_nonnull_result_p): New.
5630 (gimple_call_nonnull_arg): New.
5632 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
5634 * config/arm/arm-cpus.in: Move information from fpu field of each
5635 cpu definition to the isa field.
5636 * config/arm/parsecpu.awk (fpu): Delete match rule.
5637 (gen_comm_data): Don't add bits from the CPU's FPU entry.
5639 2018-07-11 Richard Biener <rguenther@suse.de>
5642 * dwarf2out.c (init_sections_and_labels): Use
5643 output_asm_line_debug_info consistently.
5644 (dwarf2out_early_finish): Likewise.
5645 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
5648 2018-07-11 Richard Biener <rguenther@suse.de>
5650 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
5651 Rework father_bb setting in a way to avoid propagating constants
5652 multiple times on a loop body.
5654 2018-07-10 Mark Wielaard <mark@klomp.org>
5657 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
5659 2018-07-10 Richard Biener <rguenther@suse.de>
5661 * hash-map.h (hash_map::iterator::operator*): Return
5662 references to key and value.
5664 2018-07-10 Jakub Jelinek <jakub@redhat.com>
5667 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
5668 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
5669 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
5670 (gimplify_omp_for): For composite loops, move outer
5671 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
5672 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
5673 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
5674 TREE_LIST for both the original class iterator and the "last" helper
5675 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
5676 loop, remember has_decl_expr from outer composite loops for the
5677 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
5679 2018-07-09 Martin Sebor <msebor@redhat.com>
5683 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
5684 accessing implicitly initialized array elements.
5685 * expr.c (string_constant): Handle string initializers of
5686 character arrays within aggregates.
5687 * gimple-fold.c (fold_array_ctor_reference): Add argument.
5688 Store element offset. As a special case, handle zero size.
5689 (fold_nonarray_ctor_reference): Same.
5690 (fold_ctor_reference): Add argument. Store subobject offset.
5691 * gimple-fold.h (fold_ctor_reference): Add argument.
5693 2018-07-09 Paul Koning <ni1d@arrl.net>
5695 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
5696 (pdp11_insn_cost): New function.
5697 (pdp11_md_asm_adjust): New function.
5698 (TARGET_INVALID_WITHIN_DOLOOP): Define.
5699 (pdp11_rtx_costs): Update to match machine better.
5700 (output_addr_const_pdp11): Correct format mismatch warnings.
5701 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
5702 * config/pdp11/pdp11.md: General change to add base_cost and/or
5703 length attributes for use by new pdp11_insn_cost function.
5704 (MIN_BRANCH): Correct definition.
5706 (doloop_end): Use standard pattern name for looping pattern.
5707 (doloop_end_nocc): New.
5708 (movsf): Add another constraint alternative.
5709 (zero_extendqihi2): Add constraint alternatives for not in place
5711 (zero_extendhisi2): Remove.
5712 (shift patterns): Add CC handling variants.
5716 (define_peephole2): New peephole to recognize mov that sets CC for
5719 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5721 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
5722 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
5725 2018-07-09 Jakub Jelinek <jakub@redhat.com>
5728 * real.c (real_nextafter): Return true if result is denormal.
5730 2018-07-09 Martin Liska <mliska@suse.cz>
5732 * common.opt: Add back wrongly removed attribute.
5734 2018-07-09 Richard Biener <rguenther@suse.de>
5737 * dwarf2out.c (gen_block_die): For an early generated DIE
5738 always output high/low PC attributes.
5740 2018-07-09 Tom de Vries <tdevries@suse.de>
5742 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
5743 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
5744 onto VAR_DECL with abstract origin.
5746 2018-07-07 Jim Wilson <jimw@sifive.com>
5748 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
5750 2018-07-07 Tom de Vries <tdevries@suse.de>
5752 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
5754 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
5756 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
5758 (combine_bound): Use wide-int overflow calculation instead of
5760 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
5762 * fold-const.c (int_const_binop_2): Same.
5763 (extract_muldiv_1): Same.
5764 (fold_div_compare): Same.
5765 (fold_abs_const): Same.
5767 * poly-int.h (add): Same.
5771 * predict.c (predict_iv_comparison): Same.
5772 * profile-count.c (slow_safe_scale_64bit): Same.
5773 * simplify-rtx.c (simplify_const_binary_operation): Same.
5774 * tree-chrec.c (tree_fold_binomial): Same.
5775 * tree-data-ref.c (split_constant_offset_1): Same.
5776 * tree-if-conv.c (idx_within_array_bound): Same.
5777 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
5778 * tree-ssa-phiopt.c (minmax_replacement): Same.
5779 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
5780 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
5781 * vr-values.c (vr_values::adjust_range_with_scev): Same.
5782 * wide-int.cc (wi::add_large): Same.
5783 (wi::mul_internal): Same.
5784 (wi::sub_large): Same.
5785 (wi::divmod_internal): Same.
5786 * wide-int.h: Change overflow type to overflow_type for neg, add,
5787 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
5788 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
5789 mul_internal, divmod_internal.
5790 (overflow_type): New enum.
5791 (accumulate_overflow): New.
5793 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
5795 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
5796 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
5798 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
5800 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
5801 argument is checked for zero before entering loop, avoid checking again.
5803 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
5805 * gimplify.h (generic_expr_could_trap_p): Set as global function.
5806 * gimplify.h (generic_expr_could_trap_p): Likwise.
5807 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
5809 2018-07-06 Jakub Jelinek <jakub@redhat.com>
5811 PR tree-optimization/86401
5812 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
5813 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
5814 (fold_bit_and_mask): ... here. New helper function for match.pd.
5815 * fold-const.h (fold_bit_and_mask): Declare.
5816 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
5818 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
5821 * target.def (translate_mode_attribute): New hook.
5822 * targhooks.h (default_translate_mode_attribute): Declare.
5823 * targhooks.c (default_translate_mode_attribute): New function.
5824 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
5825 * doc/tm.texi: Regenerate.
5826 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
5827 (rs6000_translate_mode_attribute): New function.
5829 2018-07-06 Paul Koning <ni1d@arrl.net>
5831 * doc/md.texi (define_split): Document DONE and FAIL.
5832 (define_peephole2): Ditto.
5834 2018-07-05 Jeff Law <law@redhat.com>
5836 PR tree-optimization/86010
5837 * tree-ssa-dse.c (compute_trims): More aggressively trim at
5838 both the head and tail of mem* and str* calls.
5840 2018-07-05 Jim Wilson <jimw@sifive.com>
5842 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
5844 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
5846 * config/aarch64/aarch64-simd.md: correct flags text for
5847 MIN_EXPR replacement.
5849 2018-07-05 James Clarke <jrtc27@jrtc27.com>
5851 * configure: Regenerated.
5853 2018-07-05 Carl Love <cel@us.ibm.com>
5855 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
5856 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
5857 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
5858 VSX_BUILTIN_DOUBLEL_V4SF.
5860 2018-07-05 Martin Sebor <msebor@redhat.com>
5863 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
5864 than its domain to compute its the upper bound of a char array.
5866 2018-07-05 Nathan Sidwell <nathan@acm.org>
5868 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
5869 * doc/cpp.texi: Update comment.
5870 * doc/tm.texi: Rebuilt.
5871 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
5872 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
5873 * doc/extend.texi (Backwards Compatibility): Clarify it is system
5874 headers affected by extern "C".
5875 * system.h: Poison NO_IMPLICIT_EXTERN_C.
5876 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
5877 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
5878 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
5879 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
5880 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
5881 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
5882 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
5883 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
5884 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
5885 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
5886 config/sparc/sp64-elf.h, config/spu/spu.h,
5887 config/stormy16/stormy16.h, config/v850/v850.h,
5888 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
5889 define NO_IMPLICIT_EXTERN_C.
5890 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
5892 2018-07-05 Tamar Christina <tamar.christina@arm.com>
5895 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
5896 instead of GET_MODE_SIZE when comparing Units.
5898 2018-07-05 Tamar Christina <tamar.christina@arm.com>
5901 * rtlanal.c (set_noop_p): Constrain on mode change,
5902 include hard-reg-set.h
5904 2018-07-05 Tamar Christina <tamar.christina@arm.com>
5906 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
5908 2018-07-05 Jakub Jelinek <jakub@redhat.com>
5911 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
5914 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
5916 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
5918 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
5921 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
5923 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
5925 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
5928 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
5930 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
5932 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
5933 neg_*_op* variables.
5935 2018-07-04 Martin Liska <mliska@suse.cz>
5937 * tree-switch-conversion.c: Define
5938 max_ratio_for_speed and max_ratio_for_size constants.
5940 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
5941 Martin Liska <mliska@suse.cz>
5946 * common.opt: Rename align options with 'str_' prefix.
5947 * common/config/i386/i386-common.c (set_malign_value): New
5949 (ix86_handle_option): Use it to set -falign-* options/
5950 * config/aarch64/aarch64-protos.h (struct tune_params): Change
5951 type from int to string.
5952 * config/aarch64/aarch64.c: Update default values from int
5954 * config/alpha/alpha.c (alpha_override_options_after_change):
5956 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
5957 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
5958 max skip conditionally.
5959 * config/i386/freebsd.h (SUBALIGN_LOG): New.
5960 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
5961 max skip conditionally.
5962 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
5963 max skip conditionally.
5964 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
5965 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
5966 max skip conditionally.
5967 * config/i386/i386.c (struct ptt): Change type from int to
5969 (ix86_default_align): Set default values.
5970 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
5971 max skip conditionally.
5972 * config/i386/iamcu.h (SUBALIGN_LOG): New.
5973 (ASM_OUTPUT_MAX_SKIP_ALIGN):
5974 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
5975 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
5976 max skip conditionally.
5977 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
5978 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
5979 * config/i386/x86-64.h (SUBALIGN_LOG): New.
5980 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
5981 max skip conditionally.
5982 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
5983 * config/ia64/ia64.c (ia64_option_override): Set default values
5984 for alignment options.
5985 * config/m68k/m68k.c: Handle new str_align_* options.
5986 * config/mips/mips.c (mips_set_compression_mode): Change
5988 (mips_option_override): Set default values for options.
5989 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
5991 * config/rs6000/rs6000.c (rs6000_option_override_internal):
5993 * config/rx/rx.c (rx_option_override): Likewise.
5994 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
5995 (LABEL_ALIGN): Use align_labels_log.
5996 (LOOP_ALIGN): Use align_loops_align.
5997 * config/s390/s390.c (s390_asm_output_function_label): Use new
5999 * config/sh/sh.c (sh_override_options_after_change):
6000 Change type of constants.
6001 * config/spu/spu.c (spu_sched_init): Likewise.
6002 * config/sparc/sparc.c (sparc_option_override): Set default
6004 * config/visium/visium.c (visium_option_override): Likewise.
6005 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
6006 emit p2align format with last argument if it's not needed.
6007 * doc/invoke.texi: Document extended format of -falign-*.
6008 * final.c: Use align_labels alignment.
6009 * flags.h (struct target_flag_state): Change type to use
6011 (struct align_flags_tuple): New.
6012 (struct align_flags): Likewise.
6013 (align_loops_log): Redefine macro to use new types.
6014 (align_loops_max_skip): Redefine macro to use new types.
6015 (align_jumps_log): Redefine macro to use new types.
6016 (align_jumps_max_skip): Redefine macro to use new types.
6017 (align_labels_log): Redefine macro to use new types.
6018 (align_labels_max_skip): Redefine macro to use new types.
6019 (align_functions_log): Redefine macro to use new types.
6020 (align_loops): Redefine macro to use new types.
6021 (align_jumps): Redefine macro to use new types.
6022 (align_labels): Redefine macro to use new types.
6023 (align_functions): Redefine macro to use new types.
6024 (align_functions_max_skip): Redefine macro to use new types.
6025 (align_loops_value): New macro.
6026 (align_jumps_value): New macro.
6027 (align_labels_value): New macro.
6028 (align_functions_value): New macro.
6029 * function.c (invoke_set_current_function_hook): Propagate
6030 alignment values from flags to global variables default in
6032 * ipa-icf.c (sem_function::equals_wpa): Use
6033 cl_optimization_option_eq instead of memcmp.
6034 * lto-streamer.h (cl_optimization_stream_out): Support streaming
6036 (cl_optimization_stream_in): Likewise.
6037 * optc-save-gen.awk: Support strings in cl_optimization.
6038 * opth-gen.awk: Likewise.
6039 * opts.c (finish_options): Remove error checking of invalid
6041 (MAX_CODE_ALIGN): Remove.
6042 (MAX_CODE_ALIGN_VALUE): Likewise.
6043 (parse_and_check_align_values): New function.
6044 (check_alignment_argument): Likewise.
6045 (common_handle_option): Use check_alignment_argument.
6046 * opts.h (parse_and_check_align_values): Declare.
6047 * toplev.c (init_alignments): Remove.
6048 (read_log_maxskip): New.
6049 (parse_N_M): Likewise.
6050 (parse_alignment_opts): Likewise.
6051 (backend_init_target): Remove usage of init_alignments.
6052 * toplev.h (parse_alignment_opts): Declare.
6053 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
6055 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
6056 * tree.c (cl_option_hasher::equal): New.
6057 * varasm.c: Use new global macros.
6059 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
6061 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
6062 Use a simpler align directive also if MAXSKIP = ALIGN-1.
6063 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6064 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6065 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6066 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
6067 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
6068 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
6069 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
6070 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6071 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6072 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6073 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6075 2018-07-04 Martin Liska <mliska@suse.cz>
6076 Jonathan Wakely <jwakely@redhat.com>
6078 * coverage.c: Use correct type.
6079 * doc/invoke.texi: Language correction.
6081 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
6084 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
6085 ENDBRANCH for non-tail call which may return via indirect branch.
6086 * doc/extend.texi: Document indirect_return attribute.
6088 2018-07-03 Martin Sebor <msebor@redhat.com>
6090 PR tree-optimization/86274
6091 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
6093 (format_floating): Correct handling of infinities and NaNs.
6095 2018-07-03 Martin Sebor <msebor@redhat.com>
6097 * print-tree.c (print_real_cst): New function.
6098 (print_node_brief): Call it.
6099 (print_node): Ditto.
6101 2018-07-03 Jeff Law <law@redhat.com>
6103 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
6104 into a single pattern.
6106 * config/h8300/h8300.md (ors code_iterator): New.
6107 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
6108 a single pattern and single splitter.
6109 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
6110 (iorqi3_1, xorqi3_1): Likewise.
6111 (iorqi3, xorqi3 expanders): Similarly.
6113 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
6114 (movmd_internal) into a single pattern using the P mode iterator.
6115 (movmd splitters): Similarly.
6116 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
6117 (movsd splitters): Similarly.
6119 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
6120 ADDB, ADDW and ADDL into a single ADD attribute which selects the
6121 right table based on the size of the operand.
6122 * config/h8300/h8300.md (length_table): Corresponding changes. All
6123 references to "addb", "addw" and "addl" changed to "add".
6124 (btst patterns): Merge two variants into a single pattern.
6125 (tstqi, tsthi): Likewise.
6126 (addhi3_incdec, addsi3_incdec): Likewise.
6127 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
6128 (mulhi3, mulsi3): Likewise.
6129 (udivhi3, udivsi3): Likewise.
6130 (divhi3, divsi3): Likewise.
6131 (andorqi3, andorhi3, andorsi3): Likewise.
6133 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
6136 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
6137 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
6139 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
6141 PR tree-optimization/85694
6142 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
6143 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
6145 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
6146 (<u>avg<mode>3_ceil): New patterns.
6148 2018-07-03 David Malcolm <dmalcolm@redhat.com>
6150 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
6151 scan-tree-dump directive.
6152 * gcc.dg/vect/slp-perm-2.c: Likewise.
6153 * gcc.dg/vect/slp-perm-3.c: Likewise.
6154 * gcc.dg/vect/slp-perm-5.c: Likewise.
6155 * gcc.dg/vect/slp-perm-6.c: Likewise.
6156 * gcc.dg/vect/slp-perm-7.c: Likewise.
6157 * gcc.dg/vect/slp-perm-8.c: Likewise.
6159 2018-07-03 Marek Polacek <polacek@redhat.com>
6162 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
6164 2018-07-03 Richard Biener <rguenther@suse.de>
6167 * tree-ssa-structalias.c (find_func_clobbers): Properly
6168 handle indirect calls.
6170 2018-07-03 Jeff Law <law@redhat.com>
6172 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
6173 (shifts): New code iterator.
6174 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
6175 expander. Fix HImode handling on H8/SX.
6176 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
6177 (subqi3, subhi3, subsi3 expanders): Likewise.
6178 (andqi3, andhi3, andsi3 expanders): Likewise.
6179 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
6180 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
6181 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
6182 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
6183 (zero_extendqihi2, zero_extendqisi2): Likewise.
6184 (extendqihi2, extendqisi2): Likewise.
6185 (rotlqi3, rotlhi3, rotlsi3): Likewise.
6186 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
6187 (rotlqi3_1, rotlhi3_1): Likewise.
6188 (logicalhi3_sn, logicalsi3_sn): Likewise.
6189 (logicalhi3, logicalsi3): Likewise.
6191 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
6193 * tree-vect-patterns.c (vect_recog_rotate_pattern)
6194 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
6195 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
6196 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
6197 type to append_pattern_def_seq instead of creating a stmt_vec_info
6199 (build_mask_conversion): Likewise. Remove vinfo argument.
6200 (vect_add_conversion_to_patterm): Likewise, renaming to...
6201 (vect_add_conversion_to_pattern): ...this.
6202 (vect_recog_mask_conversion_pattern): Update call to
6203 build_mask_conversion. Pass the vector type to
6204 append_pattern_def_seq here too.
6205 (vect_recog_gather_scatter_pattern): Update call to
6206 vect_add_conversion_to_pattern.
6208 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
6210 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
6211 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
6212 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
6213 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
6214 STMT_VINFO_PATTERN_DEF_SEQ to null here.
6215 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
6216 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
6217 append_pattern_def_seq instead of new_pattern_def_seq.
6218 (vect_recog_divmod_pattern): Do both of the above.
6219 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
6222 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
6224 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
6225 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
6226 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
6227 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
6228 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
6229 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
6230 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
6231 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
6232 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
6233 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
6234 parameter with a single stmt_vec_info.
6235 (vect_recog_func_ptr): Likewise.
6236 (vect_recog_gather_scatter_pattern): Likewise, folding in...
6237 (vect_try_gather_scatter_pattern): ...this.
6238 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
6239 the stmt_vec_info of the statement to be matched. Don't clear
6240 STMT_VINFO_RELATED_STMT.
6241 (vect_pattern_recog): Update call accordingly.
6243 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
6245 PR tree-optimization/85694
6246 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
6247 (uavgM3_ceil): Document new optabs.
6248 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
6249 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
6251 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
6252 (savg_ceil_optab): New optabs.
6253 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
6254 (vect_vect_recog_func_ptrs): Add it.
6255 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
6256 constant directly from the associated lhs.
6258 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
6260 * tree-vect-patterns.c (vect_split_statement): New function.
6261 (vect_convert_input): Use it to try to split an existing cast.
6263 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
6265 * poly-int.h (print_hex): New function.
6266 * dumpfile.h (dump_dec, dump_hex): Declare.
6267 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
6268 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
6269 min_input_precision, operation_precision and operation_sign.
6270 * tree-vect-patterns.c (vect_get_range_info): New function.
6271 (vect_same_loop_or_bb_p, vect_single_imm_use)
6272 (vect_operation_fits_smaller_type): Delete.
6273 (vect_look_through_possible_promotion): Add an optional
6274 single_use_p parameter.
6275 (vect_recog_over_widening_pattern): Rewrite to use new
6276 stmt_vec_info infomration. Handle one operation at a time.
6277 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
6278 (vect_truncatable_operation_p, vect_set_operation_type)
6279 (vect_set_min_input_precision): New functions.
6280 (vect_determine_min_output_precision_1): Likewise.
6281 (vect_determine_min_output_precision): Likewise.
6282 (vect_determine_precisions_from_range): Likewise.
6283 (vect_determine_precisions_from_users): Likewise.
6284 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
6285 (vect_vect_recog_func_ptrs): Put over_widening first.
6287 (vect_pattern_recog): Call vect_determine_precisions.
6289 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
6291 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
6292 statements that have been replaced by further pattern statements.
6293 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
6295 2018-07-03 Richard Biener <rguenther@suse.de>
6297 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
6298 always set *dt. Dump vectype in vectype overload.
6299 * dumpfile.h (dump_gimple_expr): New function.
6300 (dump_gimple_expr_loc): Likewise.
6301 * dumpfile.c (dump_gimple_expr): New function.
6302 (dump_gimple_expr_loc): Likewise.
6304 2018-07-02 Jeff Law <law@redhat.com>
6306 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
6307 the H8/300, H8/300H and H8/S variants into a single pattern.
6308 (movhi_h8300, movqi_h8300hs): Similarly.
6309 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
6310 (QHI mode iterator): New.
6312 * config/h8300/h8300.md: Remove trailing whitespace.
6314 2018-07-02 Jim Wilson <jimw@sifive.com>
6316 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
6317 instead of emit_insn for interrupt returns.
6318 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
6319 (riscv_sret, riscv_uret): Likewise.
6321 2018-07-02 David Malcolm <dmalcolm@redhat.com>
6323 * pretty-print.c (selftest::test_pp_format): Move save and restore
6324 of quotes to class auto_fix_quotes, and add an instance.
6325 * selftest.c: Include "intl.h".
6326 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
6327 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
6328 * selftest.h (selftest::auto_fix_quotes): New class.
6330 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
6332 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
6333 (aarch64_sve_prepare_conditional_op): Remove.
6334 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
6335 Allow aarch64_simd_reg_or_zero as select operand; remove
6336 the aarch64_sve_prepare_conditional_op call.
6337 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
6338 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
6339 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
6340 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
6341 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
6342 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
6343 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
6344 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
6345 and a splitters to match all of the *_any patterns.
6346 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
6348 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
6349 (SVE_COND_FP_BINARY_REV): Remove.
6350 (sve_int_op_rev, sve_fp_op_rev): New.
6351 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
6352 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
6353 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
6354 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
6355 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
6356 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
6357 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
6358 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
6359 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
6361 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
6362 Remove match_dup 1 from the inner unspec.
6363 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
6365 * config/aarch64/aarch64.md (movprfx): New attr.
6366 (length): Default movprfx to 8.
6367 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
6368 (*madd<SVE_I>, *msub<SVE_I): Likewise.
6369 (*<su>mul<SVE_I>3_highpart): Likewise.
6370 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
6371 (*v<ASHIFT><SVE_I>3): Likewise.
6372 (*<su><MAXMIN><SVE_I>3): Likewise.
6373 (*<su><MAXMIN><SVE_F>3): Likewise.
6374 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
6375 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
6376 (*div<SVE_F>4): Likewise.
6378 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
6380 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
6383 2018-07-02 Richard Biener <rguenther@suse.de>
6385 PR tree-optimization/86363
6386 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
6387 memset argument refers to a non-variable address.
6389 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
6391 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
6392 {PLUS,MINUS}_EXPR code to...
6393 (adjust_symbolic_bound): ...here,
6394 (combine_bound): ...here,
6395 (set_value_range_with_overflow): ...and here.
6397 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
6399 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
6401 (extract_range_from_abs_expr): ...here.
6403 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
6405 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
6406 -fno-omit-frame-pointer when not optimizing.
6408 2018-07-02 Martin Liska <mliska@suse.cz>
6411 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
6412 (propagate_nothrow): Likewise.
6414 2018-07-02 Martin Liska <mliska@suse.cz>
6417 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
6419 2018-07-02 David Malcolm <dmalcolm@redhat.com>
6421 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
6422 function in r262149, changing "loc" param from source_location to
6423 const dump_location_t &.
6424 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
6425 declaration, as above.
6427 2018-07-01 Paul Koning <ni1d@arrl.net>
6429 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
6430 -munit-asm, -mgnu-asm, -mdec-asm.
6431 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
6432 (pdp11_output_labelref): New.
6433 (pdp11_output_def): New.
6434 (pdp11_output_addr_vec_elt): New.
6435 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
6436 %# and %@ format codes.
6437 (pdp11_option_override): New.
6438 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
6439 (pdp11_output_ident): New.
6440 (pdp11_asm_named_section): New.
6441 (pdp11_asm_init_sections): New.
6442 (pdp11_file_start): New.
6443 (pdp11_file_end): New.
6444 (output_ascii): Use .ascii/.asciz for -mdec-asm.
6445 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
6446 %o, like %c but octal.
6447 (pdp11_option_override): New.
6448 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
6450 (DATA_SECTION_ASM_OP): Ditto.
6451 (READONLY_DATA_SECTION_ASM_OP): New.
6452 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
6453 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
6454 (ASM_OUTPUT_LABELREF): Ditto.
6455 (ASM_OUTPUT_DEF): Ditto.
6456 (ASM_OUTPUT_EXTERNAL): New.
6457 (ASM_OUTPUT_SOURCE_FILENAME): New.
6458 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
6459 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
6460 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
6461 %# and %@ format codes.
6462 * config/pdp11/pdp11.opt (mgnu-asm): New.
6463 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
6464 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
6465 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
6467 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
6469 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
6470 dereferencing path[] beyond its length.
6472 (debug_all_paths): New.
6473 (rewire_first_differing_edge): New.
6474 (adjust_paths_after_duplication): New.
6475 (duplicate_thread_path): Call adjust_paths_after_duplication.
6477 (thread_through_all_blocks): Add new argument to
6478 duplicate_thread_path.
6480 2018-06-30 Jim Wilson <jimw@sifive.com>
6482 * config/riscv/predicates.md (p2m1_shift_operand): New.
6483 (high_mask_shift_operand): New.
6484 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
6485 pattern using p2m1_shift_operand.
6486 (lshsi3_zero_extend_3+2): New combiner pattern using
6487 high_mask_shift_operand.
6489 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
6491 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
6493 (vect_recog_rotate_pattern): ...here.
6494 (vect_convert_input): Try to insert casts of invariants in the
6496 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
6497 preheader to be empty.
6499 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
6501 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
6502 vector type. If given, install it in the new statement's
6504 (vect_element_precision): New function.
6505 (vect_unpromoted_value): New struct.
6506 (vect_unpromoted_value::vect_unpromoted_value): New function.
6507 (vect_unpromoted_value::set_op): Likewise.
6508 (vect_look_through_possible_promotion): Likewise.
6509 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
6510 (vect_widened_op_tree, vect_convert_input): Likewise.
6511 (vect_convert_inputs, vect_convert_output): Likewise.
6512 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
6513 to handle the optional cast of the multiplication result and
6514 vect_widened_op_tree to detect the widened multiplication itself.
6515 Do not require the input and output of promotion casts to have
6516 the same sign, but base the signedness of the operation on the
6517 input rather than the result. If the pattern includes two
6518 promotions, check that those promotions have the same sign.
6519 Do not restrict the MULT_EXPR handling to a double-width result;
6520 handle quadruple-width results and wider. Use vect_convert_inputs
6521 to convert the inputs to the common type.
6522 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
6523 to handle the optional cast of the ABS result. Also allow a sign
6524 change or a sign extension between the ABS and MINUS.
6525 Use vect_widened_op_tree to detect the widened subtraction and use
6526 vect_convert_inputs to convert the inputs to the common type.
6527 (vect_handle_widen_op_by_const): Delete.
6528 (vect_recog_widen_op_pattern): New function.
6529 (vect_recog_widen_mult_pattern): Use it.
6530 (vect_recog_widen_shift_pattern): Likewise.
6531 (vect_recog_widen_sum_pattern): Use
6532 vect_look_through_possible_promotion to handle the promoted
6535 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
6537 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
6538 the containing gimple_seq *.
6539 * gimple-iterator.h (gsi_for_stmt): Declare it.
6540 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
6541 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
6542 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
6543 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
6544 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
6546 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
6548 (vect_mark_pattern_stmts): ...here. Handle cases in which the
6549 statement being replaced is part of an existing pattern
6550 definition sequence, inserting the new pattern statements before
6552 (vect_pattern_recog_1): Don't return a bool. If the statement
6553 is already part of a pattern, instead apply pattern matching
6554 to the pattern definition statements. Don't clear the
6555 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
6556 (vect_pattern_recog): Don't break after the first match;
6557 continue processing the pattern definition statements instead.
6558 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
6560 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
6562 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
6563 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
6564 (vect_recog_widen_sum_pattern): Use it.
6566 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
6568 * tree-vect-loop.c (vectorizable_reduction): Assert that the
6569 phi is not a pattern statement and has not been replaced by
6570 a pattern statement.
6571 * tree-vect-patterns.c (type_conversion_p): Don't check
6572 STMT_VINFO_IN_PATTERN_P.
6573 (vect_recog_vector_vector_shift_pattern): Likewise.
6574 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
6575 the pattern statement rather than the original statement; check
6576 directly for a WIDEN_MULT_EXPR here.
6577 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
6578 vect_is_simple_use to return the pattern statement rather
6579 than the original statement; use is_pattern_stmt_p to check
6580 for such a pattern statement.
6581 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
6582 to return the pattern statement rather than the original statement;
6583 don't do the same transformation here.
6584 (vect_is_simple_use): If the defining statement has been replaced
6585 by a pattern statement, return the pattern statement instead.
6586 Remove the corresponding (local) transformation from the vectype
6589 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
6591 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
6592 end and default to null.
6593 * tree-vect-loop.c (vect_create_epilog_for_reduction)
6594 (vectorizable_reduction): Update calls accordingly, dropping the
6595 gimple ** argument if the passed-back statement isn't needed.
6596 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
6597 (vect_recog_rotate_pattern): Likewise.
6598 (vect_recog_mask_conversion_pattern): Likewise.
6599 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
6600 (vect_mask_constant_operand_p): Likewise.
6601 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
6602 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
6603 (get_group_load_store_type, get_load_store_type): Likewise.
6604 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
6605 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
6606 (vectorizable_conversion, vectorizable_assignment): Likewise.
6607 (vectorizable_shift, vectorizable_operation): Likewise.
6608 (vectorizable_store, vect_is_simple_cond): Likewise.
6609 (vectorizable_condition, vectorizable_comparison): Likewise.
6610 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
6611 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
6612 and move it to the end. Cope with null def_stmt_outs.
6614 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
6616 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
6618 2018-06-29 Jeff Law <law@redhat.com>
6620 * config/v850/v850.c (v850_legitimate_address_p): Handle large
6621 displacements for TARGET_V850E2V3 and newer.
6622 (TARGET_LRA_P): Remove. Defaults to LRA now.
6623 * config/v850/v850.md (sign23byte_load): Remove.
6624 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
6625 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
6627 2018-06-29 Martin Liska <mliska@suse.cz>
6630 * coverage.c (coverage_init): Mangle full path name.
6631 * doc/invoke.texi: Document the change.
6632 * gcov-io.c (mangle_path): New.
6633 * gcov-io.h (mangle_path): Likewise.
6634 * gcov.c (mangle_name): Use mangle_path for path mangling.
6636 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6638 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
6639 if starting source register is not even.
6641 2018-06-29 Martin Liska <mliska@suse.cz>
6643 PR tree-optimization/86263
6644 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
6645 Make edge redirection.
6647 2018-06-29 David Malcolm <dmalcolm@redhat.com>
6649 * dumpfile.c (dump_loc): Add indentation based on scope depth.
6650 (dump_scope_depth): New variable.
6651 (get_dump_scope_depth): New function.
6652 (dump_begin_scope): New function.
6653 (dump_end_scope): New function.
6654 * dumpfile.h (get_dump_scope_depth): New declaration.
6655 (dump_begin_scope): New declaration.
6656 (dump_end_scope): New declaration.
6657 (class auto_dump_scope): New class.
6658 (AUTO_DUMP_SCOPE): New macro.
6659 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
6662 2018-06-29 Richard Biener <rguenther@suse.de>
6664 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
6665 compute_all_dependences succeeds.
6666 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
6667 exceed --param loop-max-datarefs-for-datadeps.
6669 2018-06-29 Jakub Jelinek <jakub@redhat.com>
6671 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
6673 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
6676 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
6677 alternative 0 in preferred_for_speed attribute.
6679 2018-06-28 Paul Koning <ni1d@arrl.net>
6681 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
6682 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
6683 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
6684 * config/pdp11/pdp11.md: Correct "length" attribute calculation
6685 for shift insn patterns.
6687 2018-06-28 David Malcolm <dmalcolm@redhat.com>
6689 * cgraph.c (cgraph_node::get_body): Replace assignments to
6690 "dump_file" with calls to set_dump_file.
6691 * dumpfile.c (alt_dump_file): Make static, and group with...
6692 (alt_flags): ...this definition.
6693 (dumps_are_enabled): New variable.
6694 (refresh_dumps_are_enabled): New function.
6695 (set_dump_file): New function.
6696 (set_alt_dump_file): New function.
6697 (gcc::dump_manager::dump_start): Replace assignments to
6698 "dump_file" and "alt_dump_file" with calls to set_dump_file and
6700 (gcc::dump_manager::dump_finish): Likewise.
6701 * dumpfile.h (alt_dump_file): Delete decl.
6702 (dumps_are_enabled): New variable decl.
6703 (set_dump_file): New function decl.
6704 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
6706 * tree-nested.c (lower_nested_functions): Replace assignments to
6707 "dump_file" with calls to set_dump_file.
6709 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
6711 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
6712 goto_locus of each outgoing edge of each basic block.
6714 2018-06-28 Richard Biener <rguenther@suse.de>
6716 * dwarf2out.c (decl_scope_table): Remove.
6717 (push_decl_scope): Likewise.
6718 (pop_decl_scope): Likewise.
6719 (gen_type_die_for_member): Do not call push/pop_decl_scope.
6720 (gen_struct_or_union_type_die): Likewise.
6721 (gen_tagged_type_die): Likewise.
6722 (dwarf2out_init): Do not initialize decl_scope_table.
6723 (dwarf2out_c_finalize): Do not free it.
6725 2018-06-28 Richard Biener <rguenther@suse.de>
6727 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
6728 deciding whether to not re-use a DIE.
6730 2018-06-28 Richard Biener <rguenther@suse.de>
6732 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
6733 DW_AT_abstract_origin attribute.
6735 2018-06-28 Martin Liska <mliska@suse.cz>
6737 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
6738 Use newly introduced constants.
6739 * tree-switch-conversion.h (struct jump_table_cluster):
6740 Define max_ratio_for_size and max_ratio_for_speed.
6742 2018-06-28 Martin Liska <mliska@suse.cz>
6744 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
6745 Add new checking assert to catch invalid state.
6746 (jump_table_cluster::can_be_handled): Handle single case
6748 (jump_table_cluster::is_beneficial): Bail out for such case.
6749 (bit_test_cluster::find_bit_tests):
6750 Add new checking assert to catch invalid state.
6751 (bit_test_cluster::can_be_handled): Handle single case
6753 (bit_test_cluster::is_beneficial): Bail out for such case.
6754 (switch_decision_tree::analyze_switch_statement):
6757 2018-06-28 Martin Liska <mliska@suse.cz>
6759 * common.opt: Introduce -completion option.
6760 * gcc.c (driver_handle_option): Handle it.
6761 (driver::main): Print completions if completion
6763 * opt-suggestions.c (option_proposer::get_completions):
6765 (option_proposer::suggest_completion): Likewise.
6766 (option_proposer::find_param_completions): Likewise.
6767 (verify_autocompletions): Likewise.
6768 (test_completion_valid_options): Likewise.
6769 (test_completion_valid_params): Likewise.
6770 (in_completion_p): Likewise.
6771 (empty_completion_p): Likewise.
6772 (test_completion_partial_match): Likewise.
6773 (test_completion_garbage): Likewise.
6774 (opt_proposer_c_tests): Likewise.
6775 * opt-suggestions.h: Declare new functions.
6776 * opts.c (common_handle_option): Handle OPT__completion_.
6777 * selftest-run-tests.c (selftest::run_tests): Add
6778 opt_proposer_c_tests.
6779 * selftest.c (assert_str_startswith): New.
6780 * selftest.h (assert_str_startswith): Likewise.
6781 (opt_proposer_c_tests): New.
6782 (ASSERT_STR_STARTSWITH): Likewise.
6784 2018-06-28 Martin Liska <mliska@suse.cz>
6786 * Makefile.in: Add opt-suggestions.o.
6787 * gcc-main.c: Include opt-suggestions.h.
6788 * gcc.c (driver::driver): Likewise.
6789 (driver::~driver): Remove m_option_suggestions.
6790 (driver::build_option_suggestions): Moved to option_proposer.
6791 (driver::suggest_option): Likewise.
6792 (driver::handle_unrecognized_options): Use option_proposer.
6793 * gcc.h (class driver): Add new memver m_option_proposer.
6794 * opt-suggestions.c: New file.
6795 * opt-suggestions.h: New file.
6797 2018-06-28 Martin Liska <mliska@suse.cz>
6799 * vec.h (class auto_string_vec): New (moved from auto_argvec).
6800 (auto_string_vec::~auto_string_vec): Likewise.
6802 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
6804 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
6805 prevent_decl_creation_for_types fields up and add reset_location field.
6806 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
6807 statement if id->reset_location is true.
6808 (copy_edges_for_bb): Do not set goto_locus on the new edges if
6809 id->reset_location is true.
6810 (copy_phis_for_bb): Force input_location on the arguments if
6811 id->reset_location is true.
6812 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
6813 is set on the function to be inlined.
6815 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
6817 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
6819 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
6821 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
6822 registers for Pmode.
6823 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
6824 hard registers for the clobbered pseudo.
6826 2018-06-27 Paul Koning <ni1d@arrl.net>
6828 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
6829 mutually exclusive options.
6830 * config/pdp11/constraints.md (h): New constraint.
6831 (O): Update definition to match shift code generation.
6832 (D): New constraint.
6833 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
6835 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
6837 (output_jump): Change arguments.
6838 (pdp11_fixed_cc_regs): New function.
6839 (pdp11_cc_mode): Ditto.
6840 (pdp11_expand_shift): Ditto.
6841 (pdp11_assemble_shift): Ditto.
6842 (pdp11_small_shift): Ditto.
6843 (pdp11_branch_cost): Remove.
6844 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
6846 (pdp11_register_move_cost): Update for CC registers.
6847 (pdp11_rtx_costs): Add case for LSHIFTRT.
6848 (pdp11_output_jump): Add CCNZ mode conditional branches.
6849 (notice_update_cc_on_set): Remove.
6850 (pdp11_cc_mode): New function.
6851 (simple_memory_operand): Correct pre/post decrement case.
6852 (no_side_effect_operand): New function.
6853 (pdp11_regno_reg_class): Add CC_REGS class.
6854 (pdp11_fixed_cc_regs): New function.
6855 (pdp11_small_shift): New function.
6856 (pdp11_expand_shift): New function to expand shift insns.
6857 (pdp11_assemble_shift): New function to output shifts.
6858 (pdp11_branch_cost): Remove.
6859 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
6860 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
6861 (WCHAR_TYPE): Ditto.
6862 (PTRDIFF_TYPE): Ditto.
6863 (ADJUST_INSN_LENGTH): New macro.
6864 (FIXED_REGISTERS): Add CC registers.
6865 (CALL_USED_REGISTERS): Ditto.
6867 (REG_CLASS_NAMES): Ditto.
6868 (REG_CLASS_CONTENTS): Ditto.
6869 (SELECT_CC_MODE): Use new function.
6870 (TARGET_FLAGS_REGNUM): New macro.
6871 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
6872 (cc0_reg_rtx): Remove.
6873 (CC_STATUS_MDEP): Remove.
6874 (CC_STATUS_MDEFP_INIT): Remove.
6875 (CC_IN_FPU): Remove.
6876 (NOTICE_UPDATE_CC): Remove.
6877 (REGISTER_NAMES): Add CC registers.
6878 (BRANCH_COST): Change to constant 1.
6879 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
6881 * config/pdp11/pdp11.opt (mbcopy): Remove.
6882 (mbcopy-builtin): Remove.
6883 (mbranch-cheap): Remove.
6884 (mbranch-expensive): Remove.
6885 * config/pdp11/predicates.md (expand_shift_operand): Update to
6886 match shift code generation.
6887 (ccnz_operator): New predicate.
6888 * doc/invoke.texi (PDP-11 Options): Remove deleted options
6889 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
6890 Remove non-existent option -mabshi, -mno-abshi. Document mutually
6892 * doc/md.texi (PDP-11): Document new D and h constraints. Update
6893 description of O constraint.
6895 2018-06-27 Jeff Law <law@redhat.com>
6896 Austin Law <austinklaw@gmail.com>
6898 * config/v850/v850.md (addsi3_set_flags): New pattern.
6899 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
6900 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
6901 (zero_extendhisi2_v850_set_flags): Likewise.
6902 (zero_extendqisi2_v850_set_flags): Likewise.
6903 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
6904 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
6905 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
6907 * config/v850/v850-protos.h (notice_update_cc): Remove.
6908 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
6909 (v850_print_operand): Handle 'D' and "d".
6910 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
6911 Add handling of arithmetic/logical operations compared against zero.
6912 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
6913 Do not look at v850_compare_op, instead get mode from last argument.
6914 (v850_gen_compare): Remove
6915 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
6916 after reload for prologue insns.
6917 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
6919 (construct_save_jarl): Likewise.
6920 (TARGET_FLAGS_REGNUM): Define.
6921 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
6922 (NOTICE_UPDATE_CC): Remove.
6923 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
6924 than cc0. Conditionalize on reload_completed.
6925 (cmpsi_insn, setfcc_insn): Likewise.
6926 (tst1 splitter): Turn into define_and_split which sets the flags
6928 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
6929 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
6930 (cstoresf4, cstoredf4): Clobber the flags.
6931 (cmpsi, cmpsf, cmpdf): Remove expanders.
6932 (setf_insn): Remove pattern.
6933 (addsi3): Turn into define_and_split which clobbers the flags after
6934 reload and a suitable pattern (addsi3_clobber_flags) for use after
6936 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
6937 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
6938 (ashrsi3, ashrsi3_v850e2): Likewise.
6939 (bins): Clobber the flags.
6940 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
6941 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
6942 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
6943 (call_value_internal_short, call_value_internal_long): Likewise.
6944 (callt_save_interrupt, callt_return_interrupt): Likewise.
6945 (save_interrupt, return_interrupt): Likewise.
6946 (callt_save_all_interrupt, save_all_interrupt): Likewise.
6947 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
6948 (restore_all_interrupt, _restore_all_interrupt): Likewise.
6949 (All FP comparisons): Only allow after reload has completed.
6951 (divh, divhu): Tweak output template.
6952 (branch_z_normal, branch_z_invert): Remove
6953 (branch_nz_normal, branch_nz_invert): Likewise.
6954 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
6956 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
6957 * config/v850/v850.c (notice_update_cc): Remove.
6958 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
6959 (CC_NO_CARRY): Likewise.
6960 (NOTICE_UPDATE_CC): Define to nothing.
6961 * config/v850/v850.md: Remove block comment on cc0 handling
6962 Remove "cc" attribute from all patterns. Remove cc_status handling
6963 from all patterns. Minor formatting fixes.
6965 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6967 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
6968 (cortex-a76.cortex-a55): Likewise.
6969 * config/aarch64/aarch64-tune.md: Regenerate.
6970 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
6971 cortex-a76.cortex-a55.
6973 2018-06-27 Jeff Law <law@redhat.com>
6975 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
6976 (MULTILIB_DIRNAMES): Similarly.
6978 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
6980 * gimple.h (gimple_return_retbnd): Delete.
6981 (gimple_return_set_retbnd): Likewise.
6982 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
6983 gimple_return_set_retbnd.
6984 * gimple-pretty-print.c (dump_gimple_return): Remove call to
6985 gimple_return_retbnd and adjust.
6986 * tree-inline.h (struct copy_body_data): Remove retbnd field.
6987 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
6988 Explicitly return NULL in a couple more cases. Move assertion
6989 on debug statements and remove unreachable code.
6990 (reset_debug_binding): Do not test id->retbnd.
6991 (expand_call_inline): Do not set it.
6993 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
6995 * configure.ac: Add --disable-gcov option.
6996 * configure: Regenerate.
6997 * Makefile.in: Honour @enable_gcov@.
6998 * doc/install.texi: Document --disable-gcov.
7000 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7002 * config/arm/arm-cpus.in (cortex-a76): New entry.
7003 (cortex-a76.cortex-a55): Likewise.
7004 * config/arm/arm-tables.opt: Regenerate.
7005 * config/arm/arm-tune.md: Likewise.
7006 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
7007 * doc/invoke.texi (ARM Options): Document cortex-a76 and
7008 cortex-a76.cortex-a55.
7010 2018-06-27 Tamar Christina <tamar.christina@arm.com>
7013 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
7015 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
7017 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
7019 (EPILOGUE_USES): Likewise.
7021 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
7023 * tree-inline.c (remap_location): New function extracted from...
7024 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
7025 (copy_phis_for_bb): ...here. Call remap_location.
7026 (copy_cfg_body): Adjust call to copy_edges_for_bb.
7028 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
7030 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
7031 unaligned vsx for 16B memset.
7033 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
7036 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
7037 ieee128_float_type_node to long_double_type_node unless
7038 TARGET_LONG_DOUBLE_128 is set.
7040 2018-06-26 David Malcolm <dmalcolm@redhat.com>
7042 * cfgloop.c (get_loop_location): Convert return type from
7043 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
7044 by implicit construction from rtx_insn *, and using
7045 dump_user_location_t::from_function_decl for the fallback case.
7046 * cfgloop.h (get_loop_location): Convert return type from
7047 location_t to dump_user_location_t.
7048 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
7049 dump_printf_loc to pass in a dump_location_t rather than a
7050 location_t, via the gimple stmt.
7051 * coverage.c (get_coverage_counts): Update calls to
7052 dump_printf_loc to pass in dump_location_t rather than a
7054 * doc/optinfo.texi (Dump types): Convert example of
7055 dump_printf_loc from taking "locus" to taking "insn". Update
7056 description of the "_loc" calls to cover dump_location_t.
7057 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
7059 (dump_user_location_t::dump_user_location_t): New constructors,
7060 from gimple *stmt and rtx_insn *.
7061 (dump_user_location_t::from_function_decl): New function.
7062 (dump_loc): Make static.
7063 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
7064 const dump_location_t &.
7065 (dump_generic_expr_loc): Delete.
7066 (dump_printf_loc): Convert param "loc" from location_t to
7067 const dump_location_t &.
7068 (selftest::test_impl_location): New function.
7069 (selftest::dumpfile_c_tests): New function.
7070 * dumpfile.h: Include "profile-count.h".
7071 (class dump_user_location_t): New class.
7072 (struct dump_impl_location_t): New struct.
7073 (class dump_location_t): New class.
7074 (dump_printf_loc): Convert 2nd param from source_location to
7075 const dump_location_t &.
7076 (dump_generic_expr_loc): Delete.
7077 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
7078 const dump_location_t &.
7079 * gimple-fold.c (fold_gimple_assign): Update call to
7080 dump_printf_loc to pass in a dump_location_t rather than a
7081 location_t, via the gimple stmt.
7082 (gimple_fold_call): Likewise.
7083 * gimple-loop-interchange.cc
7084 (loop_cand::analyze_iloop_reduction_var): Update for change to
7085 check_reduction_path.
7086 (tree_loop_interchange::interchange): Update for change to
7088 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
7089 change in return-type of find_loop_location.
7090 (graphite_regenerate_ast_isl): Likewise.
7091 * graphite-optimize-isl.c (optimize_isl): Likewise.
7092 * graphite.c (graphite_transform_loops): Likewise.
7093 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
7094 pass in a dump_location_t rather than a location_t, via the
7096 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
7097 * ipa.c (walk_polymorphic_call_targets): Likewise.
7098 * loop-unroll.c (report_unroll): Convert "locus" param from
7099 location_t to dump_location_t.
7100 (decide_unrolling): Update for change to get_loop_location's
7102 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
7103 location_t to dump_user_location_t.
7104 (grid_find_single_omp_among_assignments_1): Updates calls to
7105 dump_printf_loc to pass in a dump_location_t rather than a
7106 location_t, via the gimple stmt.
7107 (grid_parallel_clauses_gridifiable): Convert "tloc" from
7108 location_t to dump_location_t. Updates calls to dump_printf_loc
7109 to pass in a dump_location_t rather than a location_t, via the
7111 (grid_inner_loop_gridifiable_p): Likewise.
7112 (grid_dist_follows_simple_pattern): Likewise.
7113 (grid_gfor_follows_tiling_pattern): Likewise.
7114 (grid_target_follows_gridifiable_pattern): Likewise.
7115 (grid_attempt_target_gridification): Convert initialization
7116 of local "grid" from memset to zero-initialization; FIXME: does
7117 this require C++11? Update call to dump_printf_loc to pass in a
7118 optinfo_location rather than a location_t, via the gimple stmt.
7119 * profile.c (read_profile_edge_counts): Updates call to
7120 dump_printf_loc to pass in a dump_location_t rather than a
7122 (compute_branch_probabilities): Likewise.
7123 * selftest-run-tests.c (selftest::run_tests): Call
7125 * selftest.h (dumpfile_c_tests): New decl.
7126 * tree-loop-distribution.c (pass_loop_distribution::execute):
7127 Update for change in return type of find_loop_location.
7128 * tree-parloops.c (parallelize_loops): Likewise.
7129 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
7130 "locus" from location_t to dump_user_location_t.
7131 (canonicalize_loop_induction_variables): Likewise.
7132 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
7133 for change in return type of find_loop_location.
7134 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
7135 to dump_printf_loc to pass in a dump_location_t rather than a
7136 location_t, via the stmt.
7137 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
7139 * tree-vect-loop-manip.c (find_loop_location): Convert return
7140 type from source_location to dump_user_location_t.
7141 (vect_do_peeling): Update for above change.
7142 (vect_loop_versioning): Update for change in type of
7144 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
7145 from location_t to dump_user_location_t.
7146 (vect_estimate_min_profitable_iters): Update for change in type
7148 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
7149 location_t to dump_location_t.
7150 (vect_slp_bb): Update for change in type of vect_location.
7151 * tree-vectorizer.c (vect_location): Convert from source_location
7152 to dump_user_location_t.
7153 (try_vectorize_loop_1): Update for change in vect_location's type.
7154 (vectorize_loops): Likewise.
7155 (increase_alignment): Likewise.
7156 * tree-vectorizer.h (vect_location): Convert from source_location
7157 to dump_user_location_t.
7158 (find_loop_location): Convert return type from source_location to
7159 dump_user_location_t.
7160 (check_reduction_path): Convert 1st param from location_t to
7161 dump_user_location_t.
7162 * value-prof.c (check_counter): Update call to dump_printf_loc to
7163 pass in a dump_user_location_t rather than a location_t; update
7164 call to error_at for change in type of "locus".
7165 (check_ic_target): Update call to dump_printf_loc to
7166 pass in a dump_user_location_t rather than a location_t, via the
7169 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
7171 * config/s390/s390.h (enum processor_flags): Do not use
7172 default tune parameter when -march was specified.
7174 2018-06-26 Jakub Jelinek <jakub@redhat.com>
7177 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
7178 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
7181 2018-06-26 Richard Biener <rguenther@suse.de>
7183 PR tree-optimization/86287
7185 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
7186 (vect_analyze_loop): Initialize n_stmts.
7188 2018-06-26 Richard Biener <rguenther@suse.de>
7191 * fold-const.c (fold_convertible_p): Pointer extension
7194 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
7197 * dwarf2out.c (loc_list_has_views): Adjust comments.
7198 (dw_loc_list): Split single cross-partition range with
7201 2018-06-25 Jeff Law <law@redhat.com>
7203 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
7204 on -mbig-switch by default.
7206 * config/v850/predicates.md (const_float_1_operand): Fix match_code
7208 (const_float_0_operand): Remove unused predicate.
7209 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
7210 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
7211 (recipsf2): New expander. Original pattern now called
7213 (recipdf2, recipdf2_insn): Similarly.
7214 (rsqrtsf2, rsqrtsf2_insn): Similarly
7215 (rsqrtdf2, rsqrtdf2_insn): Similarly
7217 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
7219 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
7220 Simplify logic for FreeBSD (twice).
7222 2018-06-25 Martin Sebor <msebor@redhat.com>
7224 PR tree-optimization/86204
7225 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
7226 a strnlen result if it's less than the length of the string.
7228 2018-06-25 Martin Sebor <msebor@redhat.com>
7230 PR tree-optimization/85700
7231 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
7232 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
7233 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
7235 2018-06-25 Martin Sebor <msebor@redhat.com>
7237 * doc/extend.texi (Zero-length arrays): Update and clarify.
7239 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
7241 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
7242 added IEEE/IBM long double multilib support on PowerPC little
7243 endian Linux systems.
7244 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
7245 (MULTILIB_DEFAULTS): Likewise.
7246 * config/rs6000/rs6000.c (rs6000_option_override_internal):
7248 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
7249 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
7250 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
7252 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
7255 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
7256 (REORDER_45): Likewise.
7258 2018-06-25 Jeff Law <law@redhat.com>
7260 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
7261 dividend to 32 bits. Adjust length.
7262 (udivmodhi4): Cleanup output template. Fix length.
7264 2018-06-25 Carl Love <cel@us.ibm.com>
7266 * config/rs6000/vsx.md: Change word selector to prefered location.
7268 2018-06-25 Richard Biener <rguenther@suse.de>
7270 PR tree-optimization/86304
7271 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
7272 epilogue-if-converted loops as well.
7274 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
7276 * lto-section-out.c (lto_begin_section): Do not print section
7277 name for noaddr and unnumbered dumps.
7279 2018-06-25 Richard Biener <rguenther@suse.de>
7281 * tree-vectorizer.h (struct vec_info_shared): New structure
7282 with parts split out from struct vec_info and loop_nest from
7283 struct _loop_vec_info.
7284 (struct vec_info): Adjust accordingly.
7285 (struct _loop_vec_info): Likewise.
7286 (LOOP_VINFO_LOOP_NEST): Adjust.
7287 (LOOP_VINFO_DATAREFS): Likewise.
7288 (LOOP_VINFO_DDRS): Likewise.
7289 (struct _bb_vec_info): Likewise.
7290 (BB_VINFO_DATAREFS): Likewise.
7291 (BB_VINFO_DDRS): Likewise.
7292 (struct _stmt_vec_info): Add dr_aux member.
7293 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
7294 (DR_MISALIGNMENT_UNINITIALIZED): New.
7295 (set_dr_misalignment): Adjust.
7296 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
7297 (vect_analyze_loop): Adjust prototype.
7298 (vect_analyze_loop_form): Likewise.
7299 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
7300 Compute dependences lazily.
7301 (vect_record_base_alignments): Use shared datarefs/ddrs.
7302 (vect_verify_datarefs_alignment): Likewise.
7303 (vect_analyze_data_refs_alignment): Likewise.
7304 (vect_analyze_data_ref_accesses): Likewise.
7305 (vect_analyze_data_refs): Likewise.
7306 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
7307 constructor parameter for shared part.
7308 (vect_analyze_loop_form): Pass in shared part and adjust.
7309 (vect_analyze_loop_2): Pass in storage for the number of
7310 stmts. Move loop nest finding to the caller. Compute
7312 (vect_analyze_loop): Pass in shared part.
7313 (vect_transform_loop): Verify shared datarefs are unchanged.
7314 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
7315 constructor parameter for shared part.
7316 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
7317 (vect_slp_bb): Verify shared datarefs are unchanged before
7319 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
7321 (new_stmt_vec_info): Initialize DR_AUX misalignment to
7322 DR_MISALIGNMENT_UNINITIALIZED.
7323 * tree-vectorizer.c (vec_info::vec_info): Add constructor
7324 parameter for shared part.
7325 (vec_info::~vec_info): Adjust.
7326 (vec_info_shared::vec_info_shared): New.
7327 (vec_info_shared::~vec_info_shared): Likewise.
7328 (vec_info_shared::save_datarefs): Likewise.
7329 (vec_info_shared::check_datarefs): Likewise.
7330 (try_vectorize_loop_1): Construct shared part live for analyses
7331 of a single loop for multiple vector sizes.
7332 * tree-parloops.c (gather_scalar_reductions): Adjust.
7334 2018-06-25 Richard Biener <rguenther@suse.de>
7336 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
7337 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
7338 (vect_analyze_data_refs): Remove similar code from here and
7339 simplify accordingly.
7341 2018-06-25 Richard Biener <rguenther@suse.de>
7343 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
7344 for reverse storage order accesses rather than asserting
7345 they cannot happen here.
7347 2018-06-25 Tom de Vries <tdevries@suse.de>
7350 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
7351 Use data16 instead of .byte for insn prefix.
7353 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
7356 * parser.c (make_char_string_pack): Pass this literal chars
7357 through cpp_interpret_string.
7358 (cp_parser_userdef_numeric_literal): Check the result of
7359 make_char_string_pack.
7361 2018-06-24 Maya Rashish <coypu@sdf.org>
7363 * ginclude/stddef.h: Simplify conditions around avoiding
7364 re-definition of __size_t.
7366 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
7368 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
7369 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
7371 2018-06-22 Maya Rashish <coypu@sdf.org>
7373 * doc/invoke.texi (mno-fancy-math-387): Update for changes
7374 made to OpenBSD and NetBSD through the years.
7376 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
7378 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
7379 behavior of vec_pack (vector double, vector double) to match
7380 behavior of vec_float2 (vector double, vector double).
7382 2018-06-22 Olivier Hainque <hainque@adacore.com>
7384 * gimplify.c (gimplify_function_tree): Prevent creation
7385 of a trampoline for the address of the current function
7386 passed to entry/exit instrumentation hooks.
7388 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
7391 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
7394 2018-06-22 Martin Liska <mliska@suse.cz>
7396 PR tree-optimization/86263
7397 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
7398 Bail out if is_enabled is false.
7399 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
7401 (jump_table_cluster::is_enabled): New function.
7403 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
7405 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
7406 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
7407 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
7408 (lto_input_ts_binfo_tree_pointers): Likewise.
7409 * tree-streamer-out.c (streamer_write_tree_bitfields,
7410 write_ts_binfo_tree_pointers): Likewise.
7411 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
7413 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
7415 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
7417 2018-06-22 Martin Liska <mliska@suse.cz>
7419 * symbol-summary.h (get): Make it pure and inline move
7420 functionality from ::get function.
7421 (get): Remove and inline into ::get and ::get_create.
7422 (get_create): Move code from ::get function.
7424 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7427 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
7428 -x assembler-with-cpp.
7430 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7432 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
7433 _FILE_OFFSET_BITS=64 for C++.
7435 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
7437 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
7438 conversion insn that shows up when pr85657-3.c is compiled using
7439 IEEE 128-bit long double.
7440 (neg<mode>2_internal): Use the correct mode to check whether the
7441 mode is IBM extended.
7442 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
7443 multiply and divide external functions from being created more
7446 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
7448 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
7450 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
7451 the edge can be forwarded.
7452 (cfg_layout_merge_blocks): Likewise.
7454 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
7456 * except.c (finish_eh_generation): Commit edge insertions only after
7457 the EH edges have been redirected from post-landing to landing pads.
7459 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
7461 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
7462 create_tmp_var_for to create the FRAME decl.
7463 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
7465 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
7467 * tree-inline.c (copy_edges_for_bb): Minor tweak.
7468 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
7469 debug statement when resetting its value.
7470 (expand_call_inline): Copy the locus of the call onto the assignment
7471 of the return value, if any. Use local variable in more cases.
7473 2018-06-21 Martin Liska <mliska@suse.cz>
7475 * ipa-pure-const.c (propagate_nothrow): Use
7476 funct_state_summaries->get.
7477 (dump_malloc_lattice): Likewise.
7478 (propagate_malloc): Likewise.
7480 2018-06-21 Richard Biener <rguenther@suse.de>
7482 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
7483 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
7484 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
7486 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
7487 BLOCK_ABSTRACT_ORIGIN unconditionally.
7489 2018-06-21 David Malcolm <dmalcolm@redhat.com>
7491 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
7493 * ipa-reference.c (ipa_reference_c_finalize): Delete
7494 ipa_ref_opt_sum_summaries and set it to NULL.
7496 2018-06-21 Tom de Vries <tdevries@suse.de>
7498 PR tree-optimization/85859
7499 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
7500 test with comment from bb_no_side_effects_p.
7502 2018-06-21 Richard Biener <rguenther@suse.de>
7504 PR tree-optimization/86232
7505 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
7506 max for constant niter.
7508 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
7510 * config/aarch64/aarch64-simd.md
7511 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
7513 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
7515 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
7516 Make opernads of the unspec commutative.
7518 2018-06-21 Richard Biener <rguenther@suse.de>
7520 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
7521 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
7522 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
7523 (vect_analyze_data_ref_dependence): Re-order checks to deal with
7525 (vect_record_base_alignments): Do not record base alignment
7526 for gathers or scatters.
7527 (vect_compute_data_ref_alignment): Drop return value that is always
7528 true. Bail out early for gathers or scatters.
7529 (vect_enhance_data_refs_alignment): Bail out early for gathers
7531 (vect_find_same_alignment_drs): Likewise.
7532 (vect_analyze_data_refs_alignment): Remove dead code.
7533 (vect_slp_analyze_and_verify_node_alignment): Likewise.
7534 (vect_analyze_data_refs): For possible gathers or scatters do
7535 not create an alternate DR, just check their possible validity
7536 and mark them. Adjust DECL_NONALIASED handling to not rely
7538 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
7539 update inits of gathers or scatters.
7540 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
7541 Also copy gather/scatter flag to pattern vinfo.
7543 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
7545 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
7546 behavior of vec_packsu (vector unsigned long long, vector unsigned
7547 long long) to match behavior of vec_packs with same signature.
7549 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
7550 Thomas Schwinge <thomas@codesourcery.com>
7551 Cesar Philippidis <cesar@codesourcery.com>
7553 * gimplify.c (gimplify_scan_omp_clauses): Add support for
7554 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
7555 (gimplify_adjust_omp_clauses): Likewise.
7556 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
7557 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
7558 (gimplify_omp_target_update): Update handling of acc update and
7560 * omp-low.c (install_var_field): Remove unused parameter
7561 base_pointers_restrict.
7562 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
7563 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
7565 (omp_target_base_pointers_restrict_p): Delete.
7566 (scan_omp_target): Update call to scan_sharing_clauses.
7567 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
7569 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
7570 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
7571 (convert_local_omp_clauses): Likewise.
7572 * tree-pretty-print.c (dump_omp_clause): Likewise.
7573 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
7575 (omp_clause_code_name): Likewise.
7577 2018-06-20 Jakub Jelinek <jakub@redhat.com>
7580 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
7583 PR tree-optimization/86231
7584 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
7585 anti-range don't overwrite *vr0min before using it to compute *vr0max.
7587 2018-06-20 Tom de Vries <tdevries@suse.de>
7589 PR tree-optimization/86097
7590 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
7591 iv type if signedness of iv type is not the same as that of *nit.
7593 2018-06-20 Jakub Jelinek <jakub@redhat.com>
7595 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
7596 EDGE_EH edges, verify they are all EDGE_EH.
7598 2018-06-20 Maya Rashish <coypu@sdf.org>
7600 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
7602 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7604 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
7605 * config/aarch64/aarch64.c (xgene1_tunings): Add
7606 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
7607 (aarch64_mode_valid_for_sched_fusion_p):
7608 Allow 16-byte modes.
7609 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
7610 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
7612 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
7614 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
7615 * config/aarch64/iterators.md (VQ2): New mode iterator.
7617 2018-06-20 Martin Liska <mliska@suse.cz>
7619 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
7620 Change default ratio from 10 to 8.
7622 2018-06-20 Martin Liska <mliska@suse.cz>
7624 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
7626 (bit_test_cluster::find_bit_tests): Likewise.
7627 (switch_decision_tree::analyze_switch_statement): Find clusters.
7628 * tree-switch-conversion.h (struct jump_table_cluster): Document
7631 2018-06-20 Martin Liska <mliska@suse.cz>
7633 * tree-switch-conversion.c (switch_conversion::collect):
7634 Record m_uniq property.
7635 (switch_conversion::expand): Bail out for special conditions.
7636 (group_cluster::~group_cluster): New.
7637 (group_cluster::group_cluster): Likewise.
7638 (group_cluster::dump): Likewise.
7639 (jump_table_cluster::emit): New.
7640 (switch_decision_tree::fix_phi_operands_for_edges): New.
7641 (struct case_node): Remove struct.
7642 (jump_table_cluster::can_be_handled): New.
7643 (case_values_threshold): Moved to header.
7644 (reset_out_edges_aux): Likewise.
7645 (jump_table_cluster::is_beneficial): New.
7646 (bit_test_cluster::can_be_handled): Likewise.
7647 (add_case_node): Remove.
7648 (bit_test_cluster::is_beneficial): New.
7649 (case_bit_test::cmp): New.
7650 (bit_test_cluster::emit): New.
7651 (expand_switch_as_decision_tree_p): Remove.
7652 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
7653 (fix_phi_operands_for_edge): Likewise.
7654 (switch_decision_tree::analyze_switch_statement): New.
7655 (compute_cases_per_edge): Move ...
7656 (switch_decision_tree::compute_cases_per_edge): ... here.
7657 (try_switch_expansion): Likewise.
7658 (switch_decision_tree::try_switch_expansion): Likewise.
7659 (record_phi_operand_mapping): Likewise.
7660 (switch_decision_tree::record_phi_operand_mapping): Likewise.
7661 (emit_case_decision_tree): Likewise.
7662 (switch_decision_tree::emit): Likewise.
7663 (balance_case_nodes): Likewise.
7664 (switch_decision_tree::balance_case_nodes): Likewise.
7665 (dump_case_nodes): Likewise.
7666 (switch_decision_tree::dump_case_nodes): Likewise.
7667 (emit_jump): Likewise.
7668 (switch_decision_tree::emit_jump): Likewise.
7669 (emit_cmp_and_jump_insns): Likewise.
7670 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
7671 (emit_case_nodes): Likewise.
7672 (switch_decision_tree::emit_case_nodes): Likewise.
7673 (conditional_probability): Remove.
7674 * tree-switch-conversion.h (enum cluster_type): New.
7676 (struct cluster): Likewise.
7677 (cluster::cluster): Likewise.
7678 (struct simple_cluster): Likewise.
7679 (simple_cluster::simple_cluster): Likewise.
7680 (struct group_cluster): Likewise.
7681 (struct jump_table_cluster): Likewise.
7682 (struct bit_test_cluster): Likewise.
7683 (struct min_cluster_item): Likewise.
7684 (struct case_tree_node): Likewise.
7685 (case_tree_node::case_tree_node): Likewise.
7686 (jump_table_cluster::case_values_threshold): Likewise.
7687 (struct case_bit_test): Likewise.
7688 (struct switch_decision_tree): Likewise.
7689 (struct switch_conversion): Likewise.
7690 (switch_decision_tree::reset_out_edges_aux): Likewise.
7692 2018-06-20 Martin Liska <mliska@suse.cz>
7694 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
7695 (hoist_edge_and_branch_if_true): Likewise.
7696 (expand_switch_using_bit_tests_p): Likewise.
7697 (struct case_bit_test): Likewise.
7698 (case_bit_test_cmp): Likewise.
7699 (emit_case_bit_tests): Likewise.
7700 (switch_conversion::switch_conversion): New class.
7701 (struct switch_conv_info): Remove old struct.
7702 (collect_switch_conv_info): More to ...
7703 (switch_conversion::collect): ... this.
7704 (check_range): Likewise.
7705 (switch_conversion::check_range): Likewise.
7706 (check_all_empty_except_final): Likewise.
7707 (switch_conversion::check_all_empty_except_final): Likewise.
7708 (check_final_bb): Likewise.
7709 (switch_conversion::check_final_bb): Likewise.
7710 (create_temp_arrays): Likewise.
7711 (switch_conversion::create_temp_arrays): Likewise.
7712 (free_temp_arrays): Likewise.
7713 (gather_default_values): Likewise.
7714 (switch_conversion::gather_default_values): Likewise.
7715 (build_constructors): Likewise.
7716 (switch_conversion::build_constructors): Likewise.
7717 (constructor_contains_same_values_p): Likewise.
7718 (switch_conversion::contains_same_values_p): Likewise.
7719 (array_value_type): Likewise.
7720 (switch_conversion::array_value_type): Likewise.
7721 (build_one_array): Likewise.
7722 (switch_conversion::build_one_array): Likewise.
7723 (build_arrays): Likewise.
7724 (switch_conversion::build_arrays): Likewise.
7725 (gen_def_assigns): Likewise.
7726 (switch_conversion::gen_def_assigns): Likewise.
7727 (prune_bbs): Likewise.
7728 (switch_conversion::prune_bbs): Likewise.
7729 (fix_phi_nodes): Likewise.
7730 (switch_conversion::fix_phi_nodes): Likewise.
7731 (gen_inbound_check): Likewise.
7732 (switch_conversion::gen_inbound_check): Likewise.
7733 (process_switch): Use the newly created class.
7734 (switch_conversion::expand): New.
7735 (switch_conversion::~switch_conversion): New.
7736 * tree-switch-conversion.h: New file.
7738 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
7740 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
7741 tree-vect-patterns.c.
7742 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
7743 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
7744 (vect_recog_sad_pattern): Likewise.
7745 (vect_recog_widen_sum_pattern): Likewise.
7746 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
7747 (vect_recog_widen_shift_pattern): Remove the type_in argument.
7748 (vect_recog_rotate_pattern): Likewise.
7749 (vect_recog_mult_pattern): Likewise.
7750 (vect_recog_vector_vector_shift_pattern): Likewise.
7751 (vect_recog_divmod_pattern): Likewise.
7752 (vect_recog_mixed_size_cond_pattern): Likewise.
7753 (vect_recog_bool_pattern): Likewise.
7754 (vect_recog_mask_conversion_pattern): Likewise.
7755 (vect_try_gather_scatter_pattern): Likewise.
7756 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
7757 (vect_recog_over_widening_pattern): Likewise.
7758 (vect_recog_gather_scatter_pattern): Likewise.
7759 (vect_recog_func_ptr): Move from tree-vectorizer.h
7760 (vect_vect_recog_func_ptrs): Move further down the file.
7761 (vect_recog_func): Likewise. Remove the third argument.
7762 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
7763 (vect_pattern_recog_1): Expect the pattern function to do any
7764 necessary target tests. Also expect it to provide a vector type.
7765 Remove the type_in handling.
7767 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
7769 * tree-vect-patterns.c (vect_pattern_detected): New function.
7770 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
7771 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
7772 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
7773 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
7774 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
7775 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
7776 (vect_recog_mask_conversion_pattern)
7777 (vect_try_gather_scatter_pattern): Likewise.
7779 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
7781 * tree-vect-patterns.c (vect_get_internal_def): New function.
7782 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
7783 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
7784 (search_type_for_mask_1): Use it.
7786 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
7788 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
7789 redundant WIDEN_SUM_EXPR handling.
7790 (vect_recog_sad_pattern): Likewise.
7792 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
7794 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
7795 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
7796 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
7797 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
7798 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
7800 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
7802 * tree-vect-stmts.c (vectorizable_call): Make sure that we
7803 use the stmt_vec_info of the original bb statement for the
7804 new zero assignment, even if the call is part of a pattern.
7806 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
7808 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
7809 that the sequence is attached to the original statement rather
7810 than the pattern statement.
7811 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
7812 PATTERN_DEF_SEQ from the original statement rather than
7813 the main pattern statement.
7814 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
7815 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
7816 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
7818 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
7820 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
7821 definition statements before the early exit for statements that aren't
7823 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
7825 (vect_transform_loop): ...here. Process pattern definition
7826 statements without first checking whether the main pattern
7827 statement is live or relevant.
7829 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
7831 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
7832 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
7834 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
7836 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
7837 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
7838 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
7839 (expand_block_compare): Change select_block_compare_mode call.
7840 (expand_strncmp_align_check): Use new functions, fix comment.
7841 (emit_final_str_compare_gpr): New function.
7842 (expand_strn_compare): Refactor and clean up code.
7843 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
7845 2018-06-19 Tony Reix <tony.reix@atos.com>
7846 Damien Bergamini <damien.bergamini@atos.com>
7847 David Edelsohn <dje.gcc@gmail.com>
7849 * collect2.c (static_obj): New variable.
7850 (static_libs): New variable.
7851 (is_in_list): Uncomment declaration.
7852 (main): Track AIX libraries linked statically.
7853 (is_in_list): Uncomment definition.
7854 (scan_prog_file): Don't add AIX shared libraries initializer
7855 to constructor list if linking statically.
7857 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
7859 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
7861 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
7863 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
7865 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
7868 2018-06-19 Martin Liska <mliska@suse.cz>
7870 * config/i386/i386.c (ix86_can_inline_p): Do not use
7871 ipa_fn_summaries::get_create.
7872 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
7874 (devirtualization_time_bonus): Likewise.
7875 (ipcp_propagate_stage): Likewise.
7876 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
7877 (edge_set_predicate): Likewise.
7878 (evaluate_conditions_for_known_args): Likewise.
7879 (evaluate_properties_for_edge): Likewise.
7880 (ipa_call_summary::reset): Tranform to ...
7881 (ipa_call_summary::~ipa_call_summary): ... this.
7882 (ipa_fn_summary::reset): Transform to ...
7883 (ipa_fn_summary::~ipa_fn_summary): ... this.
7884 (ipa_fn_summary_t::remove): Rename to ...
7885 (ipa_fn_summary_t::remove_callees): ... this.
7886 (ipa_fn_summary_t::duplicate): Use placement new
7887 instead of memory copy.
7888 (ipa_call_summary_t::duplicate): Likewise.
7889 (ipa_call_summary_t::remove): Remove.
7890 (dump_ipa_call_summary): Change get_create to get.
7891 (ipa_dump_fn_summary): Dump only when summary exists.
7892 (analyze_function_body): Use symbol_summary::get instead
7894 (compute_fn_summary): Likewise.
7895 (estimate_edge_devirt_benefit): Likewise.
7896 (estimate_edge_size_and_time): Likewise.
7897 (inline_update_callee_summaries): Likewise.
7898 (remap_edge_change_prob): Likewise.
7899 (remap_edge_summaries): Likewise.
7900 (ipa_merge_fn_summary_after_inlining): Likewise.
7901 (write_ipa_call_summary): Likewise.
7902 (ipa_fn_summary_write): Likewise.
7903 (ipa_free_fn_summary): Likewise.
7904 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
7905 (struct ipa_call_summary): Likewise.
7906 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
7908 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
7909 (estimate_size_after_inlining): Likewise.
7910 (estimate_growth): Likewise.
7911 (growth_likely_positive): Likewise.
7912 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
7913 (inline_call): Likewise.
7914 * ipa-inline.c (caller_growth_limits): Likewise.
7915 (can_inline_edge_p): Likewise.
7916 (can_inline_edge_by_limits_p): Likewise.
7917 (compute_uninlined_call_time): Likewise.
7918 (compute_inlined_call_time): Likewise.
7919 (want_inline_small_function_p): Likewise.
7920 (edge_badness): Likewise.
7921 (update_caller_keys): Likewise.
7922 (update_callee_keys): Likewise.
7923 (inline_small_functions): Likewise.
7924 (inline_to_all_callers_1): Likewise.
7925 (dump_overall_stats): Likewise.
7926 (early_inline_small_functions): Likewise.
7927 (early_inliner): Likewise.
7928 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
7929 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
7930 * ipa-pure-const.c (malloc_candidate_p): Likewise.
7931 * ipa-split.c (execute_split_functions): Likewise.
7932 * symbol-summary.h: Likewise.
7933 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
7935 2018-06-19 Richard Biener <rguenther@suse.de>
7937 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
7938 (vectorize_loops): ... here. Fix dbgcnt handling.
7939 (try_vectorize_loop): Wrap try_vectorize_loop_1.
7941 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
7944 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
7945 ieee128 argument takes up only one (vector) register, not two (floating
7948 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
7950 * gimplify.c (gimplify_init_constructor): Really never clear for an
7951 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
7953 2018-06-19 Richard Biener <rguenther@suse.de>
7955 PR tree-optimization/86179
7956 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
7957 after failed recognition.
7959 2018-06-18 Martin Sebor <msebor@redhat.com>
7962 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
7963 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
7964 Handle integer subtraction.
7965 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
7966 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
7968 2018-06-18 David Malcolm <dmalcolm@redhat.com>
7970 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
7971 param from rtx to rtx_insn *.
7972 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
7974 (frv_ifcvt_modify_insn): Likwise.
7975 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
7976 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
7977 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
7978 as_a <rtx_insn *> cast to local "unprotected_region" once
7979 it's been established that it's not NULL or pc_rtx.
7980 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
7981 param "sethi" from rtx to rtx_insn *.
7982 (nds32_group_float_insns): Likewise for param "insn".
7983 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
7985 (vax_output_int_subtract): Likewise.
7986 * config/vax/vax.c (vax_output_int_add): Likewise for param
7988 (vax_output_int_subtract): Likewise.
7989 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
7990 (emit_pattern_after): Likewise for param "after".
7991 (emit_insn_after): Likewise.
7992 (emit_jump_insn_after): Likewise.
7993 (emit_call_insn_after): Likewise.
7994 (emit_debug_insn_after): Likewise.
7995 (emit_pattern_before): Likewise for param "before".
7996 (emit_insn_before): Likewise.
7997 (emit_jump_insn_before): Likewise.
7998 * final.c (get_insn_template): Likewise for param "insn", removing
8000 * output.h (get_insn_template): Likewise for 2nd param.
8001 * rtl.h (emit_insn_before): Likewise.
8002 (emit_jump_insn_before): Likewise.
8003 (emit_debug_insn_before_noloc): Likewise.
8004 (emit_insn_after): Likewise.
8005 (emit_jump_insn_after): Likewise.
8006 (emit_call_insn_after): Likewise.
8007 (emit_debug_insn_after): Likewise.
8008 (set_insn_deleted): Likewise for param.
8010 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
8013 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
8014 floating point modes, so that IFmode is numerically greater than
8015 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
8016 to declare the ordering. This prevents IFmode from being
8017 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
8018 machine. Include rs6000-modes.h to share the fractional values
8019 between genmodes* and the rest of the compiler.
8023 * config/rs6000/rs6000-modes.h: New file.
8024 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
8025 meaning of rs6000_long_double_size so that 126..128 selects an
8026 appropriate 128-bit floating point type.
8027 (rs6000_option_override_internal): Likewise.
8028 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
8029 (TARGET_LONG_DOUBLE_128): Change the meaning of
8030 rs6000_long_double_size so that 126..128 selects an appropriate
8031 128-bit floating point type.
8032 (LONG_DOUBLE_TYPE_SIZE): Update comment.
8033 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
8034 source and destination to match the standard usage.
8035 (truncifkf2): Likewise.
8036 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
8037 ISA 2.07 to use an explicit clobber, instead of passing in a
8039 (copysign<mode>3_soft): Likewise.
8041 2018-06-18 David Malcolm <dmalcolm@redhat.com>
8043 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
8044 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
8045 (vect_slp_analyze_instance_dependence): Likewise.
8046 (vect_enhance_data_refs_alignment): Likewise.
8047 (vect_analyze_data_refs_alignment): Likewise.
8048 (vect_slp_analyze_and_verify_instance_alignment
8049 (vect_analyze_data_ref_accesses): Likewise.
8050 (vect_prune_runtime_alias_test_list): Likewise.
8051 (vect_analyze_data_refs): Likewise.
8052 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
8053 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
8054 (vect_analyze_scalar_cycles_1): Likewise.
8055 (vect_get_loop_niters): Likewise.
8056 (vect_analyze_loop_form_1): Likewise.
8057 (vect_update_vf_for_slp): Likewise.
8058 (vect_analyze_loop_operations): Likewise.
8059 (vect_analyze_loop): Likewise.
8060 (vectorizable_induction): Likewise.
8061 (vect_transform_loop): Likewise.
8062 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
8063 * tree-vect-slp.c (vect_analyze_slp): Likewise.
8064 (vect_make_slp_decision): Likewise.
8065 (vect_detect_hybrid_slp): Likewise.
8066 (vect_slp_analyze_operations): Likewise.
8067 (vect_slp_bb): Likewise.
8068 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
8069 (vectorizable_bswap): Likewise.
8070 (vectorizable_call): Likewise.
8071 (vectorizable_simd_clone_call): Likewise.
8072 (vectorizable_conversion): Likewise.
8073 (vectorizable_assignment): Likewise.
8074 (vectorizable_shift): Likewise.
8075 (vectorizable_operation): Likewise.
8076 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
8078 2018-06-18 Martin Sebor <msebor@redhat.com>
8080 PR tree-optimization/81384
8081 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
8082 * builtins.c (expand_builtin_strnlen): New function.
8083 (expand_builtin): Call it.
8084 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
8085 * builtins.def (BUILT_IN_STRNLEN): New.
8086 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
8087 Warn for bounds in excess of maximum object size.
8088 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
8089 single-value ranges. Handle strnlen.
8090 (handle_builtin_strlen): Handle strnlen.
8091 (strlen_check_and_optimize_stmt): Same.
8092 * doc/extend.texi (Other Builtins): Document strnlen.
8094 2018-06-18 Maya Rashish <coypu@sdf.org>
8096 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
8097 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
8098 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
8100 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
8102 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
8104 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8106 * tree.c (escaped_string::escape): Replace cast to char * by
8107 const_cast<char *> (unescaped).
8109 2018-06-18 Nick Clifton <nickc@redhat.com>
8112 * tree.c (escaped_string): New class. Converts an unescaped
8113 string into its escaped equivalent.
8114 (warn_deprecated_use): Use the new class to convert the
8115 deprecation message, if present.
8116 (test_escaped_strings): New self test.
8117 (test_c_tests): Add test_escaped_strings.
8118 * doc/extend.texi (deprecated): Add a note that the
8119 deprecation message is affected by the -fmessage-length
8120 option, and that control characters will be escaped.
8121 (#pragma GCC error): Document this pragma.
8122 (#pragma GCC warning): Likewise.
8123 * doc/invoke.texi (-fmessage-length): Document this option's
8124 effect on the #warning and #error preprocessor directives and
8125 the deprecated attribute.
8127 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
8129 * tree.c (decl_value_expr_lookup): Revert latest change.
8130 (decl_value_expr_insert): Likewise.
8132 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
8134 * gimplify.c (nonlocal_vlas): Delete.
8135 (nonlocal_vla_vars): Likewise.
8136 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
8138 (gimplify_body): Do not create and destroy nonlocal_vlas.
8139 * tree-nested.c: Include diagnostic.h.
8140 (use_pointer_in_frame): Tweak.
8141 (lookup_field_for_decl): Add assertion and declare the transformation.
8142 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
8143 internal error when the reference is in a wrong context. Do not
8144 create a debug decl by default.
8145 (note_nonlocal_block_vlas): Delete.
8146 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
8147 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
8148 create a debug decl by default.
8149 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
8150 call is in a wrong context.
8151 (fixup_vla_decls): New function.
8152 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
8153 debug variables were created.
8154 * tree.c (decl_value_expr_lookup): Add checking assertion.
8155 (decl_value_expr_insert): Likewise.
8157 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
8160 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
8161 * tree-scalar-evolution.c (interpret_expr): Likewise.
8162 (expression_expensive_p): Likewise.
8163 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
8164 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
8165 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
8166 (ssa_defined_by_minus_one_stmt_p): New.
8168 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
8171 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
8172 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
8173 * dojump.c (do_jump): Likewise.
8174 * expr.c (expand_expr_real_2): Check operand type's sign.
8175 * fold-const.c (const_unop): Handle ABSU_EXPR.
8176 (fold_abs_const): Likewise.
8177 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
8178 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
8179 (strip_sign_op_1): Likesise.
8180 * match.pd: Add new pattern to generate ABSU_EXPR.
8181 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
8182 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
8183 * tree-eh.c (operation_could_trap_helper_p): Likewise.
8184 * tree-inline.c (estimate_operator_cost): Likewise.
8185 * tree-pretty-print.c (dump_generic_node): Likewise.
8186 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
8187 * tree.def (ABSU_EXPR): New.
8189 2018-06-16 Jakub Jelinek <jakub@redhat.com>
8192 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
8193 documented as preserved for backward compatibility only.
8194 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
8196 PR rtl-optimization/86108
8197 * bb-reorder.c (create_forwarder_block): Renamed to ...
8198 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
8199 jump from new landing pad to the second part.
8200 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
8203 2018-06-15 Jakub Jelinek <jakub@redhat.com>
8206 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
8207 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
8208 Only call store_expr for halves if the mode is the same.
8211 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
8212 Fix up comment formatting.
8214 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
8216 * typed-splay-tree.h (typed_splay_tree::remove): New function.
8217 (typed_splay_tree::closure,
8218 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
8219 (typed_splay_tree::typed_splay_tree,
8220 typed_splay_tree::operator =): Declared private.
8221 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
8222 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
8223 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
8224 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
8225 typed_splay_tree::splay_tree_splay,
8226 typed_splay_tree::splay_tree_foreach_helper,
8227 typed_splay_tree::splay_tree_insert,
8228 typed_splay_tree::splay_tree_remove,
8229 typed_splay_tree::splay_tree_lookup,
8230 typed_splay_tree::splay_tree_predecessor,
8231 typed_splay_tree::splay_tree_successor,
8232 typed_splay_tree::splay_tree_min,
8233 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
8234 (typed_splay_tree::root, typed_splay_tree::comp,
8235 typed_splay_tree::delete_key,
8236 typed_splay_tree::delete_value): New data members.
8237 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
8238 typed_splay_tree::remove.
8240 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
8242 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
8243 -mginv and -mno-ginv to the assembler.
8244 * config/mips/mips.opt (-mcrc): New option.
8246 * doc/invoke.text (-mcrc): Document.
8249 2018-06-15 Nick Clifton <nickc@redhat.com>
8252 * tree.c (escaped_string): New class. Converts an unescaped
8253 string into its escaped equivalent.
8254 (warn_deprecated_use): Use the new class to convert the
8255 deprecation message, if present.
8256 (test_escaped_strings): New self test.
8257 (test_c_tests): Add test_escaped_strings.
8258 * doc/extend.texi (deprecated): Add a note that the
8259 deprecation message is affected by the -fmessage-length
8260 option, and that control characters will be escaped.
8261 (#pragma GCC error): Document this pragma.
8262 (#pragma GCC warning): Likewise.
8263 * doc/invoke.texi (-fmessage-length): Document this option's
8264 effect on the #warning and #error preprocessor directives and
8265 the deprecated attribute.
8267 2018-06-15 Richard Biener <rguenther@suse.de>
8269 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
8270 here, also noting vector size used.
8271 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
8272 size used in MSG_OPTIMIZED_LOCATIONS dump.
8273 (pass_slp_vectorize::execute): Adjust.
8275 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
8278 * config/arc/arc.c (arc_return_address_register): Fix
8281 2018-06-15 Richard Biener <rguenther@suse.de>
8284 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
8285 leave useless conversion stripping to force_gimple_operand_gsi.
8286 (gimplify_build2): Likewise.
8287 (gimplify_build1): Likewise.
8289 2018-06-15 Richard Biener <rguenther@suse.de>
8292 * tree-cfg.c (move_stmt_op): unshare invariant addresses
8293 before adjusting their block.
8295 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
8297 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
8298 multilibs for *-*-rtems*.
8299 * config/riscv/t-rtems: New file.
8301 2018-06-14 Jakub Jelinek <jakub@redhat.com>
8304 * match.pd ((A +- CST1) +- CST2): Punt if last resort
8305 unsigned_type_for returns NULL.
8308 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
8309 subregs of multi-word pseudos unless the float mode has word size.
8311 2018-06-14 Richard Biener <rguenther@suse.de>
8314 * tree-vect-generic.c (build_word_mode_vector_type): Remove
8315 duplicate and harmful type_hash_canon.
8316 * tree.c (type_hash_canon): Assert we didn't find ourselves.
8318 2018-06-14 Richard Biener <rguenther@suse.de>
8321 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
8324 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
8326 * config/rtems.h (STDINT_LONG32): Define.
8328 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
8329 Prachi Godbole <prachi.godbole@imgtec.com>
8331 * config/mips/mips-cpus.def: Define P6600.
8332 * config/mips/mips-tables.opt: Regenerate.
8333 * config/mips/mips.c (mips_ucbranch_type): New enum.
8334 (mips_rtx_cost_data): Add support for P6600.
8335 (mips_issue_rate): Likewise.
8336 (mips_multipass_dfa_lookahead): Likewise.
8337 (mips_avoid_hazard): Likewise.
8338 (mips_reorg_process_insns): Likewise.
8339 (mips_classify_branch_p6600): New function.
8340 * config/mips/mips.h (TUNE_P6600): New define.
8341 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
8342 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
8343 * config/mips/mips.md: Include p6600.md.
8344 (processor): Add p6600.
8345 * config/mips/p6600.md: New file.
8346 * doc/invoke.texi: Add p6600 to supported architectures.
8348 2018-06-13 Martin Sebor <msebor@redhat.com>
8350 PR tree-optimization/86114
8351 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
8353 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
8355 2018-06-13 Richard Biener <rguenther@suse.de>
8357 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
8358 Properly set vector type of the intermediate stmt.
8359 * tree-vect-stmts.c (vectorizable_operation): The destination
8360 var always has vectype_out type.
8362 2018-06-13 Jeff Law <law@redhat.com>
8364 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
8365 integer 0 for argument to print_rtl_with_bb.
8366 (rl78_reorg): Likewise.
8368 2018-06-13 David Malcolm <dmalcolm@redhat.com>
8370 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
8371 from rtx to rtx_insn *.
8372 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
8374 (add_sched_insns_for_speculation): Likewise for local "target",
8375 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
8376 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
8377 from rtx_insn ** to rtx_code_label **.
8378 (reorg_emit_nops): Likewise.
8379 (c6x_reorg): Likewise for local "call_labels".
8380 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
8382 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
8383 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
8384 the loops over LABEL_REFS.
8385 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
8387 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
8388 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
8389 (split_branches): Strengthen local "olabel" from rtx to
8390 rtx_insn *, adding a safe_as_a cast.
8391 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
8393 (add_insn_after): Likewise for first two params.
8394 (add_insn_before): Likewise.
8395 (remove_insn): Likewise for param.
8396 (emit_pattern_before_noloc): Likewise for second and third params.
8397 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
8398 (emit_call_insn_before_noloc): Likewise.
8399 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
8401 (emit_barrier_before): Likewise.
8402 (emit_label_before): Strengthen "label" param from "rtx" to
8403 "rtx_code_label *". Strengthen "before" param from "rtx" to
8405 (emit_insn_after_1): Strengthen "after" param from "rtx" to
8407 (emit_pattern_after_noloc): Likewise.
8408 (emit_insn_after_noloc): Likewise.
8409 (emit_jump_insn_after_noloc): Likewise.
8410 (emit_call_insn_after_noloc): Likewise.
8411 (emit_debug_insn_after_noloc): Likewise.
8412 (emit_barrier_after): Likewise.
8413 (emit_label_after): Likewise for both params.
8414 (emit_pattern_after_setloc): Likewise for "after" param. Convert
8415 "loc" param from "int" to "location_t".
8416 (emit_insn_after_setloc): Likewise.
8417 (emit_jump_insn_after_setloc): Likewise.
8418 (emit_call_insn_after_setloc): Likewise.
8419 (emit_debug_insn_after_setloc): Likewise.
8420 (emit_pattern_before_setloc): Likewise for "before" param. Convert
8421 "loc" param from "int" to "location_t".
8422 (emit_pattern_before): Convert NULL_RTX to NULL.
8423 (emit_insn_before_setloc): Convert "loc" param from "int" to
8425 (emit_jump_insn_before_setloc): Likewise.
8426 (emit_call_insn_before_setloc): Likewise.
8427 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
8428 rtx_insn *. Convert "loc" param from "int" to "location_t".
8429 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
8430 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
8431 Convert 3rd param from "int" to "location_t".
8432 (emit_barrier_before, emit_barrier_after, next_real_insn):
8433 Strengthen param from rtx to rtx_insn *.
8434 (emit_label_before): Strengthen 1st param from "rtx" to
8435 "rtx_code_label *". Strengthen 2nd param from "rtx" to
8437 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
8438 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
8439 Strengthen 2nd param from "rtx" to "rtx_insn *".
8440 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
8441 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
8442 Likewise. Convert 3rd param from "int" to "location_t".
8443 (emit_label_after): Strengthen 1st param from "rtx" to
8445 (next_real_insn, remove_insn): Strengthen param from "rtx" to
8447 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
8448 from "rtx" to "rtx_insn *".
8450 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
8452 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
8453 bodies streamed in with -Q.
8454 * dumpfile.c (dump_files): Add lto-stream-out dump file.
8455 * dumpfile.h (tree_dump_index): Add lto_stream_out.
8456 * gimple-streamer-out.c: Include gimple-pretty-print.h
8457 (output_bb): Dump stmts streamed.
8458 * lto-section-out.c: Include print-tree.h
8459 (lto_begin_section): Dump sections created.
8460 (lto_output_decl_index): Dump decl encoded.
8461 * lto-streamer-out.c: Include print-tree.h
8462 (create_output_block): Dump output block created.
8463 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
8464 (output_function): Dump function output.
8465 (output_constructor): Dump constructor streamed.
8466 (write_global_stream): Output indexes encoded.
8467 (produce_asm_for_decls): Dump streams encoded.
8468 * lto-streamer.c (streamer_dump_file): New global var.
8469 * lto-streamer.h (streamer_dump_file): Declare.
8470 * passes.c (ipa_write_summaries): Initialize streamer dump.
8471 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
8474 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
8477 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
8478 offsets for register save directives. Emit a second batch of save
8479 directives, if need be, when the function accesses prior frames.
8481 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
8483 * config/arc/fpu.md (fmasf4): Force operand to register.
8484 (fnmasf4): Likewise.
8486 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
8488 * config/arc/arc-protos.h (arc_pad_return): Remove.
8489 * config/arc/arc.c (machine_function): Remove force_short_suffix
8491 (arc_print_operand): Adjust printing of '&'.
8492 (arc_verify_short): Remove conditional printing of short suffix.
8493 (arc_final_prescan_insn): Remove reference to size_reason.
8494 (pad_return): New function.
8495 (arc_reorg): Call pad_return.
8496 (arc_pad_return): Remove.
8497 (arc_init_machine_status): Remove reference to force_short_suffix.
8498 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
8499 (attr length): When attribute iscompact is true force to 2
8500 regardless; in the case of maybe check if we want to force the
8501 instruction to have 4 bytes length.
8502 (nopv): Change it to generate 4 byte long nop as well.
8503 (blockage): New pattern.
8504 (simple_return): Remove call to arc_pad_return.
8505 (p_return_i): Likewise.
8507 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
8509 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
8511 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
8513 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
8516 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
8518 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
8519 for ARC700 and ARCv2.
8521 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
8524 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
8525 operands[2] instead of operands[1].
8528 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
8530 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
8531 case, check whether the outer register overlaps an unallocatable
8532 register, not just whether it fits the required class.
8534 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
8536 * poly-int.h (can_div_trunc_p): Add new overload in which all values
8538 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
8539 (memrefs_conflict_p): Likewise.
8540 (init_alias_analysis): Likewise.
8541 * cfgexpand.c (expand_debug_expr): Likewise.
8542 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
8543 * cse.c (fold_rtx): Likewise.
8544 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
8545 * expr.c (emit_block_move_hints): Likewise.
8546 (clear_storage_hints, push_block, emit_push_insn): Likewise.
8547 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
8548 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
8549 (emit_group_store): Likewise.
8550 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
8551 to read the PRE/POST_MODIFY increment.
8552 * calls.c (store_one_arg): Use strip_offset.
8553 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
8555 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
8557 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
8558 (simplify_binary_operation_1): Extend CONST_INT handling to
8560 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
8561 than a HOST_WIDE_INT.
8562 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
8564 (adjust_mems, add_stores): Update accodingly.
8565 (vt_canonicalize_addr): Track polynomial offsets.
8566 (emit_note_insn_var_location): Likewise.
8567 (vt_add_function_parameter): Likewise.
8568 (vt_initialize): Likewise.
8570 2018-06-12 Jeff Law <law@redhat.com>
8572 * config.gcc (alpha*-*-freebsd*): Remove.
8573 * config/alpha/freebsd.h: Remove.
8575 2018-06-12 David Malcolm <dmalcolm@redhat.com>
8578 * spellcheck-tree.c (levenshtein_distance): Rename to...
8579 (get_edit_distance): ...this, and update for underlying renaming.
8580 * spellcheck-tree.h (levenshtein_distance): Rename to...
8581 (get_edit_distance): ...this.
8582 * spellcheck.c (levenshtein_distance): Rename to...
8583 (get_edit_distance): ...this. Convert from Levenshtein distance
8584 to Damerau-Levenshtein distance by supporting transpositions of
8585 adjacent characters. Rename "v1" to "v_next" and "v0" to
8587 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
8588 (selftest::test_edit_distance_unit_test_oneway): ...this, and
8589 update for underlying renaming.
8590 (selftest::levenshtein_distance_unit_test): Rename to...
8591 (selftest::test_get_edit_distance_unit): ...this, and update for
8592 underlying renaming.
8593 (selftest::test_find_closest_string): Add example from PR 69968
8594 where transposition helps
8595 (selftest::test_metric_conditions): Update for renaming.
8596 (selftest::test_metric_conditions): Likewise.
8597 (selftest::spellcheck_c_tests): Likewise.
8598 * spellcheck.h (levenshtein_distance): Rename both overloads to...
8599 (get_edit_distance): ...this.
8600 (best_match::consider): Update for renaming.
8602 2018-06-12 Martin Sebor <msebor@redhat.com>
8604 PR tree-optimization/85259
8605 * builtins.c (compute_objsize): Handle constant offsets.
8606 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
8607 true iff a warning has been issued.
8608 * gimple.h (gimple_nonartificial_location): New function.
8609 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
8610 gimple_nonartificial_location and handle -Wno-system-headers.
8611 (handle_builtin_stxncpy): Same.
8613 2018-06-12 Martin Sebor <msebor@redhat.com>
8616 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
8618 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
8620 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
8621 BUILTIN_VEC_XST entries for pointer to double and long long.
8623 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
8626 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
8628 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
8631 2018-06-12 Martin Liska <mliska@suse.cz>
8633 * doc/options.texi: Document IntegerRange.
8635 2018-06-12 Martin Liska <mliska@suse.cz>
8637 * config/i386/i386.opt: Make MPX-related options as Deprecated.
8638 * opt-functions.awk: Handle Deprecated flag.
8639 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
8641 (read_cmdline_option): Report warning for a deprecated option.
8642 * opts.h (struct cl_option): Add new field cl_deprecated.
8643 (CL_ERR_DEPRECATED): New.
8645 2018-06-12 Martin Liska <mliska@suse.cz>
8647 * doc/options.texi: Document Deprecated option flag.
8649 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
8651 * config/arc/arc-arch.h (arc_extras): New enum.
8652 (arc_cpu_t):Add field extra.
8653 (arc_cpu_types): Consider the extras.
8654 * config/arc/arc-cpus.def: Add extras info.
8655 * config/arc/arc-opts.h (processor_type): Consider extra field.
8656 * config/arc/arc.c (arc_override_options): Handle extra field.
8658 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
8660 * config/arc/arc-arch.h: Update ARC_OPTX macro.
8661 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
8663 * config/arc/arc.c (arc_init): Update pic warning.
8664 (irq_range): Update irq range parsing warnings.
8665 (arc_override_options): Update various warning messages.
8666 (arc_handle_aux_attribute): Likewise.
8668 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
8670 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
8672 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
8674 * doc/sourcebuild.texi: Document usage of line number 0 in verify
8675 compiler messages directives.
8677 2018-06-12 Matthew Fortune <mfortune@gmail.com>
8679 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
8680 * config/mips/mips-tables.opt: Regenerate.
8681 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
8683 * doc/invoke.texi: Document -march=i6500.
8685 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
8687 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
8688 (i6400_gpmul): Add cpu_unit.
8689 (i6400_gpdiv): Likewise.
8690 (i6400_msa_add_d): Update reservations.
8691 (i6400_msa_int_add) Likewise.
8692 (i6400_msa_short_logic3) Likewise.
8693 (i6400_msa_short_logic2) Likewise.
8694 (i6400_msa_short_logic) Likewise.
8695 (i6400_msa_move) Likewise.
8696 (i6400_msa_cmp) Likewise.
8697 (i6400_msa_short_float2) Likewise.
8698 (i6400_msa_div_d) Likewise.
8699 (i6400_msa_long_logic1) Likewise.
8700 (i6400_msa_long_logic2) Likewise.
8701 (i6400_msa_mult) Likewise.
8702 (i6400_msa_long_float2) Likewise.
8703 (i6400_msa_long_float4) Likewise.
8704 (i6400_msa_long_float5) Likewise.
8705 (i6400_msa_long_float8) Likewise.
8706 (i6400_fpu_fadd): Include frint type.
8707 (i6400_fpu_store): New define_insn_reservation.
8708 (i6400_fpu_load): Likewise.
8709 (i6400_fpu_move): Likewise.
8710 (i6400_fpu_fcmp): Likewise.
8711 (i6400_fpu_fmadd): Likewise.
8712 (i6400_int_mult): Include imul3nc type and update reservation.
8713 (i6400_int_div): Include idiv3 type and update reservation.
8714 (i6400_int_load): Update to check type not move_type.
8715 (i6400_int_store): Likewise.
8716 (i6400_int_prefetch): Set zero latency.
8718 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
8720 * gcc.c: Document new %@{...} sequence.
8721 (LINK_COMMAND_SPEC): Use it for the -L switches.
8722 (cpp_unique_options): Use it for the -I switches.
8723 (at_file_argbuf): New global variable.
8724 (in_at_file): Likewise.
8725 (alloc_args): Create at_file_argbuf.
8726 (clear_args): Truncate at_file_argbuf.
8727 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
8728 (open_at_file): New function.
8729 (close_at_file): Likewise.
8730 (create_at_file): Delete.
8731 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
8734 (validate_switches_from_spec): Deal with %@{...} sequence.
8735 (validate_switches): Likewise.
8736 (driver::finalize): Call clear_args.
8738 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
8740 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
8742 2018-06-11 Martin Sebor <msebor@redhat.com>
8744 * doc/invoke.texi (-Wall): List -Wc++17-compat.
8745 (Wno-class-memaccess): Add @opindex.
8746 (Wno-templates, Wno-multiple-inheritance): Same.
8747 (Wno-virtual-inheritance, Wno-namespaces): Same.
8748 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
8749 (Wno-format-overflow, Wno-format-truncation): Same.
8750 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
8751 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
8752 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
8753 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
8754 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
8755 (Wno-misspelled-isr): Same.
8757 2018-06-11 Martin Sebor <msebor@redhat.com>
8759 * PR tree-optimization/86083
8760 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
8762 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
8764 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
8766 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
8769 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
8770 on the correct operand.
8771 (*movdi_internal64): Ditto.
8773 2018-06-11 Martin Liska <mliska@suse.cz>
8775 PR tree-optimization/86089
8776 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
8778 2018-06-11 Julia Koval <julia.koval@intel.com>
8780 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
8781 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
8782 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
8784 2018-06-11 Olivier Hainque <hainque@adacore.com>
8786 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
8787 for Ada with strict dwarf2.
8789 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
8792 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
8795 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
8797 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
8799 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
8801 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
8804 2018-06-08 Martin Liska <mliska@suse.cz>
8806 * tree-cfg.h (debug_function): Fix argument type to match
8809 2018-06-08 Martin Liska <mliska@suse.cz>
8811 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
8812 Remove usage of MPX-related (and removed) fields.
8813 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
8815 2018-06-08 David Malcolm <dmalcolm@redhat.com>
8817 * cfg.c (debug): Use TDF_NONE rather than 0.
8818 * cfghooks.c (debug): Likewise.
8819 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
8820 (struct dump_option_value_info): Convert to...
8821 (struct kv_pair): ...this template type.
8822 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
8824 (optinfo_verbosity_options): Likewise.
8825 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
8827 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
8828 than int for "optgroup_flags" param.
8829 (dump_generic_expr_loc): Use dump_flags_t rather than int for
8831 (dump_dec): Likewise.
8832 (dump_finish): Use TDF_NONE rather than 0.
8833 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
8834 rather than int for "optgroup_flags" param. Use TDF_NONE rather
8835 than 0. Update for change to option_ptr.
8836 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
8837 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
8838 0. Update for changes to optinfo_verbosity_options and
8840 (opt_info_switch_p): Convert optgroup_flags from int to
8842 (dump_basic_block): Use dump_flags_t rather than int
8843 for "dump_kind" param.
8844 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
8845 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
8846 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
8847 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
8848 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
8849 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
8850 TDF_NONE): Convert from macros to...
8851 (enum dump_flag): ...this new enum.
8852 (dump_flags_t): Update to use enum.
8853 (operator|, operator&, operator~, operator|=, operator&=):
8854 Implement for dump_flags_t.
8855 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
8856 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
8857 Convert from macros to...
8858 (enum optgroup_flag): ...this new enum.
8859 (optgroup_flags_t): New typedef.
8860 (operator|, operator|=): Implement for optgroup_flags_t.
8861 (struct dump_file_info): Convert field "alt_flags" to
8862 dump_flags_t. Convert field "optgroup_flags" to
8864 (dump_basic_block): Use dump_flags_t rather than int for param.
8865 (dump_generic_expr_loc): Likewise.
8866 (dump_dec): Likewise.
8867 (dump_register): Convert param "optgroup_flags" to
8869 (opt_info_enable_passes): Likewise.
8870 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
8872 * gimple-pretty-print.c (debug): Likewise.
8873 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
8874 (merged_store_group::apply_stores): Likewise.
8875 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
8876 * gimple.c (verify_gimple_pp): Likewise.
8877 * graphite-poly.c (print_pbb_body): Likewise.
8878 * passes.c (pass_manager::register_one_dump_file): Convert
8879 local "optgroup_flags" to optgroup_flags_t.
8880 * print-tree.c (print_node): Use TDF_NONE rather than 0.
8882 (debug_body): Likewise.
8883 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
8884 to optgroup_flags_t.
8885 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
8887 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
8888 (convert_mult_to_fma): Likewise.
8889 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
8890 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
8891 * tree-vect-data-refs.c (dump_lower_bound): Convert param
8892 "dump_kind" to dump_flags_t.
8894 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
8896 * config/rs6000/rs6000.c (min, max): Delete.
8898 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
8900 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
8901 -mabi=spe and -mabi=no-spe.
8903 2018-06-08 Martin Liska <mliska@suse.cz>
8905 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
8906 where we expect an existing summary.
8908 2018-06-08 Martin Liska <mliska@suse.cz>
8910 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
8911 * ipa-inline.h (estimate_edge_growth): Likewise.
8913 2018-06-08 Martin Liska <mliska@suse.cz>
8915 * cgraph.c (function_version_hasher::hash): Use
8916 cgraph_node::get_uid ().
8917 (function_version_hasher::equal):
8918 * cgraph.h (cgraph_node::get_uid): New method.
8919 * ipa-inline.c (update_caller_keys): Use
8920 cgraph_node::get_uid ().
8921 (update_callee_keys): Likewise.
8922 * ipa-utils.c (searchc): Likewise.
8923 (ipa_reduced_postorder): Likewise.
8924 * lto-cgraph.c (input_node): Likewise.
8925 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
8926 * symbol-summary.h (symtab_insertion): Likewise.
8927 (symtab_removal): Likewise.
8928 (symtab_duplication): Likewise.
8929 * tree-pretty-print.c (dump_function_header): Likewise.
8930 * tree-sra.c (convert_callers_for_node): Likewise.
8932 2018-06-08 Martin Liska <mliska@suse.cz>
8934 * cgraph.c (symbol_table::create_edge): Always assign a new
8936 (symbol_table::free_edge): Do not recycle numbers.
8937 * cgraph.h (cgraph_edge::get): New method.
8938 * symbol-summary.h (symtab_removal): Use it.
8939 (symtab_duplication): Likewise.
8940 (call_summary::hashable_uid): Remove.
8942 2018-06-08 Martin Liska <mliska@suse.cz>
8944 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
8945 (initialize_growth_caches): Remove.
8946 (free_growth_caches): Likewise.
8947 (do_estimate_edge_time): Use edge_growth_cache.
8948 (do_estimate_edge_size): Likewise.
8949 (do_estimate_edge_hints): Likewise.
8950 * ipa-inline.c (reset_edge_caches): Likewise.
8951 (recursive_inlining): Likewise.
8952 (inline_small_functions): Likewise.
8953 * ipa-inline.h (initialize_growth_caches): Remove.
8954 (estimate_edge_size): Likewise.
8955 (estimate_edge_time): Likewise.
8956 (estimate_edge_hints): Likewise.
8957 (reset_edge_growth_cache): Likewise.
8958 * symbol-summary.h (call_summary::remove): New method.
8960 2018-06-08 Martin Liska <mliska@suse.cz>
8962 * ipa-cp.c (class edge_clone_summary): New summary.
8963 (grow_edge_clone_vectors): Remove.
8964 (ipcp_edge_duplication_hook): Remove.
8965 (class edge_clone_summary_t): New call_summary class.
8966 (ipcp_edge_removal_hook): Remove.
8967 (edge_clone_summary_t::duplicate): New function.
8968 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
8969 (create_specialized_node): Likewise.
8970 (ipcp_driver): Initialize edge_clone_summaries and do not
8973 2018-06-08 Martin Liska <mliska@suse.cz>
8975 * symbol-summary.h (get): New function.
8976 (call_summary::m_initialize_when_cloning): New class member.
8978 2018-06-08 Martin Liska <mliska@suse.cz>
8980 * cgraph.c (cgraph_node::remove): Do not recycle uid.
8981 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
8982 (symbol_table::allocate_cgraph_symbol): Do not set uid.
8983 * passes.c (uid_hash_t): Record removed_nodes by their uids.
8984 (remove_cgraph_node_from_order): Use the removed_nodes set.
8985 (do_per_function_toporder): Likwise.
8986 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
8987 instead of summary_uid.
8988 (symtab_removal): Likewise.
8989 (symtab_duplication): Likewise.
8991 2018-06-08 Martin Liska <mliska@suse.cz>
8993 * ipa-cp.c (ipcp_store_bits_results): Use
8994 ipcp_transformation_sum.
8995 (ipcp_store_vr_results): Likewise.
8996 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
8998 (ipcp_transformation_initialize): ... this.
8999 (ipa_set_node_agg_value_chain):
9000 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
9001 (write_ipcp_transformation_info): Likewise.
9002 (read_ipcp_transformation_info): Likewise.
9003 (ipcp_update_bits): Likewise.
9004 (ipcp_update_vr): Likewise.
9005 (ipcp_transform_function): Likewise.
9006 * ipa-prop.h: Rename ipcp_transformation_summary to
9007 ipcp_transformation.
9008 (class ipcp_transformation_t): New function summary.
9009 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
9010 (ipa_get_agg_replacements_for_node): Likewise.
9012 2018-06-08 Martin Liska <mliska@suse.cz>
9014 * ipa-pure-const.c (struct funct_state_d): Do it class instead
9016 (class funct_state_summary_t): New function_summary class.
9017 (has_function_state): Remove.
9018 (get_function_state): Likewise.
9019 (set_function_state): Likewise.
9020 (add_new_function): Likewise.
9021 (funct_state_summary_t::insert): New function.
9022 (duplicate_node_data): Remove.
9023 (remove_node_data): Remove.
9024 (funct_state_summary_t::duplicate): New function.
9025 (register_hooks): Create new funct_state_summaries.
9026 (pure_const_generate_summary): Use it.
9027 (pure_const_write_summary): Likewise.
9028 (pure_const_read_summary): Likewise.
9029 (propagate_pure_const): Likewise.
9030 (propagate_nothrow): Likewise.
9031 (dump_malloc_lattice): Likewise.
9032 (propagate_malloc): Likewise.
9033 (execute): Do not register hooks, just remove summary
9035 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
9038 2018-06-08 Martin Liska <mliska@suse.cz>
9040 * ipa-reference.c (remove_node_data): Remove.
9041 (duplicate_node_data): Likewise.
9042 (class ipa_ref_var_info_summary_t): New class.
9043 (class ipa_ref_opt_summary_t): Likewise.
9044 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
9045 (get_reference_optimization_summary): Use
9046 ipa_ref_opt_sum_summaries.
9047 (set_reference_vars_info): Remove.
9048 (set_reference_optimization_summary): Likewise.
9049 (ipa_init): Create summaries.
9050 (init_function_info): Use function summary.
9051 (ipa_ref_opt_summary_t::duplicate): New function.
9052 (ipa_ref_opt_summary_t::remove): New function.
9053 (get_read_write_all_from_node): Fix GNU coding style.
9054 (propagate): Use function summary.
9055 (write_node_summary_p): Fix GNU coding style.
9056 (stream_out_bitmap): Likewise.
9057 (ipa_reference_read_optimization_summary): Use function summary.
9058 (ipa_reference_c_finalize): Do not release hooks.
9060 2018-06-08 Martin Liska <mliska@suse.cz>
9062 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
9063 (analyze_function_body): Extract multiple calls of get_create.
9064 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
9065 * ipa-inline.c (recursive_inlining): Use ::get method.
9066 * ipa-inline.h (estimate_edge_growth): Likewise.
9068 2018-06-08 Martin Liska <mliska@suse.cz>
9070 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
9072 (hsa_function_summary::hsa_function_summary): Use the new enum
9074 (hsa_gpu_implementation_p): Use hsa_summaries::get.
9075 * hsa-gen.c (hsa_get_host_function): Likewise.
9076 (get_brig_function_name): Likewise.
9077 * ipa-hsa.c (process_hsa_functions): Likewise.
9078 (ipa_hsa_write_summary): Likewise.
9079 * symbol-summary.h (symtab_duplication): Use ::get function/
9080 (get): New function.
9082 2018-06-08 Martin Liska <mliska@suse.cz>
9084 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
9086 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
9087 (hsa_register_kernel): Likewise.
9088 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
9089 * hsa-gen.c (hsa_get_host_function): Likewise.
9090 (get_brig_function_name): Likewise.
9091 (generate_hsa): Likewise.
9092 (pass_gen_hsail::execute): Likewise.
9093 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
9094 (devirtualization_time_bonus): Likewise.
9095 (ipcp_propagate_stage): Likewise.
9096 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
9097 (edge_set_predicate): Likewise.
9098 (evaluate_conditions_for_known_args): Likewise.
9099 (evaluate_properties_for_edge): Likewise.
9100 (ipa_fn_summary::reset): Likewise.
9101 (ipa_fn_summary_t::duplicate): Likewise.
9102 (dump_ipa_call_summary): Likewise.
9103 (ipa_dump_fn_summary): Likewise.
9104 (analyze_function_body): Likewise.
9105 (compute_fn_summary): Likewise.
9106 (estimate_edge_devirt_benefit): Likewise.
9107 (estimate_edge_size_and_time): Likewise.
9108 (estimate_calls_size_and_time): Likewise.
9109 (estimate_node_size_and_time): Likewise.
9110 (inline_update_callee_summaries): Likewise.
9111 (remap_edge_change_prob): Likewise.
9112 (remap_edge_summaries): Likewise.
9113 (ipa_merge_fn_summary_after_inlining): Likewise.
9114 (ipa_update_overall_fn_summary): Likewise.
9115 (read_ipa_call_summary): Likewise.
9116 (inline_read_section): Likewise.
9117 (write_ipa_call_summary): Likewise.
9118 (ipa_fn_summary_write): Likewise.
9119 (ipa_free_fn_summary): Likewise.
9120 * ipa-hsa.c (process_hsa_functions): Likewise.
9121 (ipa_hsa_write_summary): Likewise.
9122 (ipa_hsa_read_section): Likewise.
9123 * ipa-icf.c (sem_function::merge): Likewise.
9124 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
9125 (do_estimate_edge_time): Likewise.
9126 (estimate_size_after_inlining): Likewise.
9127 (estimate_growth): Likewise.
9128 (growth_likely_positive): Likewise.
9129 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
9130 (inline_call): Likewise.
9131 * ipa-inline.c (caller_growth_limits): Likewise.
9132 (can_inline_edge_p): Likewise.
9133 (can_inline_edge_by_limits_p): Likewise.
9134 (compute_uninlined_call_time): Likewise.
9135 (compute_inlined_call_time): Likewise.
9136 (want_inline_small_function_p): Likewise.
9137 (edge_badness): Likewise.
9138 (update_caller_keys): Likewise.
9139 (update_callee_keys): Likewise.
9140 (recursive_inlining): Likewise.
9141 (inline_small_functions): Likewise.
9142 (inline_to_all_callers_1): Likewise.
9143 (dump_overall_stats): Likewise.
9144 (early_inline_small_functions): Likewise.
9145 (early_inliner): Likewise.
9146 * ipa-inline.h (estimate_edge_growth): Likewise.
9147 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
9148 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
9149 * ipa-prop.h (IPA_NODE_REF): Likewise.
9150 (IPA_EDGE_REF): Likewise.
9151 * ipa-pure-const.c (malloc_candidate_p): Likewise.
9152 (propagate_malloc): Likewise.
9153 * ipa-split.c (execute_split_functions): Likewise.
9154 * symbol-summary.h: Rename get to get_create.
9156 (get_create): Likewise.
9157 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
9159 2018-06-08 Martin Liska <mliska@suse.cz>
9161 * symbol-summary.h (release): Move definition out of class
9163 (symtab_removal): Likewise.
9164 (symtab_duplication): Likewise.
9166 2018-06-08 Martin Liska <mliska@suse.cz>
9168 * symbol-summary.h (function_summary): Move constructor
9169 implementation out of class declaration.
9170 (release): Likewise.
9171 (symtab_insertion): Likewise.
9172 (symtab_removal): Likewise.
9173 (symtab_duplication): Likewise.
9176 2018-06-08 Martin Liska <mliska@suse.cz>
9178 * Makefile.in: Remove support for MPX (macros, related functions,
9179 fields in cgraph_node, ...).
9180 * builtin-types.def (BT_BND): Likewise.
9181 (BT_FN_BND_CONST_PTR): Likewise.
9182 (BT_FN_CONST_PTR_BND): Likewise.
9183 (BT_FN_VOID_PTR_BND): Likewise.
9184 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
9185 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
9186 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
9187 (expand_builtin_mempcpy_with_bounds): Likewise.
9188 (expand_builtin_memset_with_bounds): Likewise.
9189 (expand_builtin_memset_args): Likewise.
9190 (std_expand_builtin_va_start): Likewise.
9191 (expand_builtin): Likewise.
9192 (expand_builtin_with_bounds): Likewise.
9193 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
9194 (DEF_LIB_BUILTIN_CHKP): Likewise.
9195 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
9196 (DEF_CHKP_BUILTIN): Likewise.
9197 (BUILT_IN_MEMCPY): Likewise.
9198 (BUILT_IN_MEMMOVE): Likewise.
9199 (BUILT_IN_MEMPCPY): Likewise.
9200 (BUILT_IN_MEMSET): Likewise.
9201 (BUILT_IN_STPCPY): Likewise.
9202 (BUILT_IN_STRCAT): Likewise.
9203 (BUILT_IN_STRCHR): Likewise.
9204 (BUILT_IN_STRCPY): Likewise.
9205 (BUILT_IN_STRLEN): Likewise.
9206 (BUILT_IN_MEMCPY_CHK): Likewise.
9207 (BUILT_IN_MEMMOVE_CHK): Likewise.
9208 (BUILT_IN_MEMPCPY_CHK): Likewise.
9209 (BUILT_IN_MEMSET_CHK): Likewise.
9210 (BUILT_IN_STPCPY_CHK): Likewise.
9211 (BUILT_IN_STRCAT_CHK): Likewise.
9212 (BUILT_IN_STRCPY_CHK): Likewise.
9213 * calls.c (store_bounds): Likewise.
9214 (emit_call_1): Likewise.
9215 (special_function_p): Likewise.
9216 (maybe_warn_nonstring_arg): Likewise.
9217 (initialize_argument_information): Likewise.
9218 (finalize_must_preallocate): Likewise.
9219 (compute_argument_addresses): Likewise.
9220 (expand_call): Likewise.
9221 * cfgexpand.c (expand_call_stmt): Likewise.
9222 (expand_return): Likewise.
9223 (expand_gimple_stmt_1): Likewise.
9224 (pass_expand::execute): Likewise.
9225 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
9226 (cgraph_node::remove): Likewise.
9227 (cgraph_node::dump): Likewise.
9228 (cgraph_node::verify_node): Likewise.
9229 * cgraph.h (chkp_function_instrumented_p): Likewise.
9230 (symtab_node::get_alias_target): Likewise.
9231 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
9232 (cgraph_local_p): Likewise.
9233 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
9234 (cgraph_edge::rebuild_references): Likewise.
9235 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
9236 (walk_polymorphic_call_targets): Likewise.
9237 (cgraph_node::expand_thunk): Likewise.
9238 (symbol_table::output_weakrefs): Likewise.
9239 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
9240 (ix86_handle_option): Likewise.
9241 * config/i386/constraints.md: Likewise.
9242 * config/i386/i386-builtin-types.def (BND): Likewise.
9246 * config/i386/i386-builtin.def (BDESC_END): Likewise.
9247 (BDESC_FIRST): Likewise.
9249 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
9250 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
9251 * config/i386/i386.c (enum reg_class): Likewise.
9252 (ix86_target_string): Likewise.
9253 (ix86_option_override_internal): Likewise.
9254 (ix86_conditional_register_usage): Likewise.
9255 (ix86_valid_target_attribute_inner_p): Likewise.
9256 (ix86_set_indirect_branch_type): Likewise.
9257 (ix86_set_current_function): Likewise.
9258 (ix86_function_arg_regno_p): Likewise.
9259 (init_cumulative_args): Likewise.
9260 (ix86_function_arg_advance): Likewise.
9261 (ix86_function_arg): Likewise.
9262 (ix86_pass_by_reference): Likewise.
9263 (ix86_function_value_regno_p): Likewise.
9264 (ix86_function_value_1): Likewise.
9265 (ix86_function_value_bounds): Likewise.
9266 (ix86_return_in_memory): Likewise.
9267 (ix86_setup_incoming_vararg_bounds): Likewise.
9268 (ix86_va_start): Likewise.
9269 (indirect_thunk_need_prefix): Likewise.
9270 (print_reg): Likewise.
9271 (ix86_print_operand): Likewise.
9272 (ix86_expand_call): Likewise.
9273 (ix86_output_function_return): Likewise.
9274 (reg_encoded_number): Likewise.
9275 (BDESC_VERIFYS): Likewise.
9276 (ix86_init_mpx_builtins): Likewise.
9277 (ix86_init_builtins): Likewise.
9278 (ix86_emit_cmove): Likewise.
9279 (ix86_emit_move_max): Likewise.
9280 (ix86_expand_builtin): Likewise.
9281 (ix86_builtin_mpx_function): Likewise.
9282 (ix86_get_arg_address_for_bt): Likewise.
9283 (ix86_load_bounds): Likewise.
9284 (ix86_store_bounds): Likewise.
9285 (ix86_load_returned_bounds): Likewise.
9286 (ix86_store_returned_bounds): Likewise.
9287 (ix86_class_likely_spilled_p): Likewise.
9288 (ix86_hard_regno_mode_ok): Likewise.
9289 (x86_order_regs_for_local_alloc): Likewise.
9290 (ix86_mitigate_rop): Likewise.
9291 (ix86_bnd_prefixed_insn_p): Likewise.
9292 (ix86_mpx_bound_mode): Likewise.
9293 (ix86_make_bounds_constant): Likewise.
9294 (ix86_initialize_bounds): Likewise.
9295 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
9296 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
9297 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
9298 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
9299 (TARGET_CHKP_BOUND_MODE): Likewise.
9300 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
9301 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
9302 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
9303 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
9304 * config/i386/i386.h (TARGET_MPX): Likewise.
9305 (TARGET_MPX_P): Likewise.
9306 (VALID_BND_REG_MODE): Likewise.
9307 (FIRST_BND_REG): Likewise.
9308 (LAST_BND_REG): Likewise.
9309 (enum reg_class): Likewise.
9310 (BND_REG_P): Likewise.
9311 (BND_REGNO_P): Likewise.
9312 (BNDmode): Likewise.
9313 (ADJUST_INSN_LENGTH): Likewise.
9314 * config/i386/i386.md: Likewise.
9315 * config/i386/i386.opt: Likewise.
9316 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
9317 (defined): Likewise.
9318 (LINK_MPX): Likewise.
9319 (MPX_SPEC): Likewise.
9320 (LIBMPX_SPEC): Likewise.
9321 (LIBMPXWRAPPERS_SPEC): Likewise.
9322 (CHKP_SPEC): Likewise.
9323 * config/i386/predicates.md: Likewise.
9324 * dbxout.c (dbxout_type): Likewise.
9325 * doc/extend.texi: Likewise.
9326 * doc/invoke.texi: Likewise.
9327 * doc/md.texi: Likewise.
9328 * doc/tm.texi: Likewise.
9329 * doc/tm.texi.in: Likewise.
9330 * dwarf2out.c (is_base_type): Likewise.
9331 (gen_formal_types_die): Likewise.
9332 (gen_subprogram_die): Likewise.
9333 (gen_type_die_with_usage): Likewise.
9334 (gen_decl_die): Likewise.
9335 (dwarf2out_late_global_decl): Likewise.
9336 * expr.c (expand_assignment): Likewise.
9337 (emit_storent_insn): Likewise.
9338 (store_expr_with_bounds): Likewise.
9339 (store_expr): Likewise.
9340 (expand_expr_real_1): Likewise.
9341 * expr.h (store_expr_with_bounds): Likewise.
9342 * function.c (use_register_for_decl): Likewise.
9343 (struct bounds_parm_data): Likewise.
9344 (assign_parms_augmented_arg_list): Likewise.
9345 (assign_parm_find_entry_rtl): Likewise.
9346 (assign_parm_is_stack_parm): Likewise.
9347 (assign_parm_load_bounds): Likewise.
9348 (assign_bounds): Likewise.
9349 (assign_parms): Likewise.
9350 (expand_function_start): Likewise.
9351 * gcc.c (CHKP_SPEC): Likewise.
9352 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
9353 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
9354 (wrestrict_dom_walker::check_call): Likewise.
9355 * gimple.c (gimple_build_call_from_tree): Likewise.
9356 * gimple.h (enum gf_mask): Likewise.
9357 (gimple_call_with_bounds_p): Likewise.
9358 (gimple_call_set_with_bounds): Likewise.
9359 * gimplify.c (gimplify_init_constructor): Likewise.
9360 * ipa-cp.c (initialize_node_lattices): Likewise.
9361 (propagate_constants_across_call): Likewise.
9362 (find_more_scalar_values_for_callers_subset): Likewise.
9363 * ipa-hsa.c (process_hsa_functions): Likewise.
9364 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
9365 * ipa-icf.c (sem_function::merge): Likewise.
9366 * ipa-inline.c (early_inliner): Likewise.
9367 * ipa-pure-const.c (warn_function_noreturn): Likewise.
9368 (warn_function_cold): Likewise.
9369 (propagate_pure_const): Likewise.
9370 * ipa-ref.h (enum GTY): Likewise.
9371 * ipa-split.c (find_retbnd): Likewise.
9372 (consider_split): Likewise.
9373 (split_function): Likewise.
9374 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
9375 * ipa.c (walk_polymorphic_call_targets): Likewise.
9376 (symbol_table::remove_unreachable_nodes): Likewise.
9377 (process_references): Likewise.
9378 (cgraph_build_static_cdtor_1): Likewise.
9379 * lto-cgraph.c (lto_output_node): Likewise.
9380 (output_refs): Likewise.
9381 (compute_ltrans_boundary): Likewise.
9382 (input_overwrite_node): Likewise.
9383 (input_node): Likewise.
9384 (input_cgraph_1): Likewise.
9385 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
9386 * passes.c (pass_manager::execute_early_local_passes): Likewise.
9387 (class pass_chkp_instrumentation_passes): Likewise.
9388 (make_pass_chkp_instrumentation_passes): Likewise.
9389 * passes.def: Likewise.
9390 * rtl.h (struct GTY): Likewise.
9391 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
9392 * stor-layout.c (layout_type): Likewise.
9393 * symtab.c: Likewise.
9394 * target.def: Likewise.
9395 * targhooks.c (default_chkp_bound_type): Likewise.
9396 (default_chkp_bound_mode): Likewise.
9397 (default_builtin_chkp_function): Likewise.
9398 (default_chkp_function_value_bounds): Likewise.
9399 (default_chkp_make_bounds_constant): Likewise.
9400 (default_chkp_initialize_bounds): Likewise.
9401 * targhooks.h (default_chkp_bound_type): Likewise.
9402 (default_chkp_bound_mode): Likewise.
9403 (default_builtin_chkp_function): Likewise.
9404 (default_chkp_function_value_bounds): Likewise.
9405 (default_chkp_make_bounds_constant): Likewise.
9406 (default_chkp_initialize_bounds): Likewise.
9407 * toplev.c (compile_file): Likewise.
9408 (process_options): Likewise.
9409 * tree-core.h (DEF_BUILTIN): Likewise.
9410 (DEF_BUILTIN_CHKP): Likewise.
9411 * tree-inline.c (declare_return_variable): Likewise.
9412 (remap_gimple_stmt): Likewise.
9413 (copy_bb): Likewise.
9414 (initialize_inlined_parameters): Likewise.
9415 (expand_call_inline): Likewise.
9416 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
9417 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
9418 (make_pass_ipa_chkp_produce_thunks): Likewise.
9419 (make_pass_chkp): Likewise.
9420 (make_pass_chkp_opt): Likewise.
9421 (make_pass_chkp_instrumentation_passes): Likewise.
9422 * tree-pretty-print.c (dump_generic_node): Likewise.
9423 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
9424 * tree-ssa-dce.c (propagate_necessity): Likewise.
9425 (eliminate_unnecessary_stmts): Likewise.
9426 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
9427 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
9428 * tree-ssa-sccvn.h: Likewise.
9429 * tree-ssa-strlen.c (get_string_length): Likewise.
9430 (valid_builtin_call): Likewise.
9431 (adjust_last_stmt): Likewise.
9432 (handle_builtin_strchr): Likewise.
9433 (handle_builtin_strcpy): Likewise.
9434 (handle_builtin_stxncpy): Likewise.
9435 (handle_builtin_memcpy): Likewise.
9436 (handle_builtin_strcat): Likewise.
9437 (strlen_check_and_optimize_stmt): Likewise.
9438 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
9439 * tree-streamer-in.c: Likewise.
9440 * tree-streamer.c (record_common_node): Likewise.
9441 * tree.c (tree_code_size): Likewise.
9442 (wide_int_to_tree_1): Likewise.
9443 (type_contains_placeholder_1): Likewise.
9444 (build_common_tree_nodes): Likewise.
9445 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
9446 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
9447 (POINTER_BOUNDS_P): Likewise.
9448 (BOUNDED_TYPE_P): Likewise.
9449 (BOUNDED_P): Likewise.
9450 (CALL_WITH_BOUNDS_P): Likewise.
9451 (pointer_bounds_type_node): Likewise.
9452 * value-prof.c (gimple_ic): Likewise.
9453 * var-tracking.c (vt_add_function_parameters): Likewise.
9454 * varasm.c (make_decl_rtl): Likewise.
9455 (assemble_start_function): Likewise.
9456 (output_constant): Likewise.
9457 (maybe_assemble_visibility): Likewise.
9458 * varpool.c (ctor_for_folding): Likewise.
9459 * chkp-builtins.def: Remove.
9460 * ipa-chkp.c: Remove.
9461 * ipa-chkp.h: Remove.
9462 * rtl-chkp.c: Remove.
9463 * rtl-chkp.h: Remove.
9464 * tree-chkp-opt.c: Remove.
9465 * tree-chkp.c: Remove.
9466 * tree-chkp.h: Remove.
9468 2018-06-07 Carl Love <cel@us.ibm.com>
9470 * config/rs6000/vsx.md (vextract_fp_from_shorth,
9471 vextract_fp_from_shortl): Add BE support.
9473 2018-06-07 Paul Koning <ni1d@arrl.net>
9475 * compare-elim.c (try_merge_compare): Don't merge compare if
9476 address contains a side effect.
9477 (try_eliminate_compare): Likewise.
9479 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
9481 * config.gcc: Support "tremont".
9482 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
9483 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
9485 * config/i386/i386.c (m_TREMONT): Define.
9486 (processor_target_table): Add "tremont".
9487 (PTA_TREMONT): Define.
9488 (ix86_lea_outperforms): Add TARGET_TREMONT.
9489 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
9490 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
9491 and M_INTEL_GOLDMONT_PLUS.
9492 (fold_builtin_cpu): Add "tremont".
9493 (ix86_add_stmt_cost): Add TARGET_TREMONT.
9494 (ix86_option_override_internal): Add "tremont".
9495 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
9496 (processor_type): Add PROCESSOR_TREMONT.
9497 * config/i386/x86-tune.def: Add m_TREMONT.
9498 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
9500 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
9502 * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
9503 symbol defined for msp430i* devices to be lower case.
9505 2018-06-07 Richard Biener <rguenther@suse.de>
9507 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
9508 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
9509 Properly wrap signed arithmetic if overflow wraps.
9511 2018-06-07 Jakub Jelinek <jakub@redhat.com>
9513 PR tree-optimization/69615
9514 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
9515 of a cast from a same precision integral SSA_NAME in a bb dominated
9516 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
9517 cast to utype if rhs2 has already a compatible type.
9519 2018-06-07 Richard Biener <rguenther@suse.de>
9521 PR tree-optimization/85935
9522 * graphite-scop-detection.c (find_params_in_bb): Analyze
9523 condition operands with respect to the correct loop. Assert
9524 the analysis doesn't fail.
9526 2018-06-04 Carl Love <cel@us.ibm.com>
9528 * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
9529 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
9530 as it is slightly cheaper.
9531 (first_match_or_eos_index_<mode>):
9532 Calculate index using natural element order.
9533 (first_match_index_<mode>):
9534 Calculate index using natural element order.
9535 (first_match_or_eos_index_<mode>):
9536 Calculate index using natural order.
9537 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
9538 for BE and LE modes.
9539 * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
9540 P9V_BUILTIN_VCLZLSBB_V16QI.
9541 * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
9544 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
9546 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
9547 indentation and line wrap for many prototypes. Add missing
9548 @smallexample directives around block of prototypes for vec_xl and
9551 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
9553 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
9554 track if we pass or return IEEE 128-bit floating point.
9555 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
9556 C++ mangling that is compatible with GCC 8.1.
9557 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
9558 (init_cumulative_args): Note if we pass or return IEEE 128-bit
9559 floating point types.
9560 (rs6000_function_arg_advance_1): Likewise.
9561 (rs6000_mangle_type): Optionally generate mangled names that match
9562 what GCC 8.1 generated for IEEE 128-bit floating point types.
9563 (rs6000_globalize_decl_name): If we have an external function that
9564 passes or returns IEEE 128-bit types, generate a weak reference
9565 from the mangled name used in GCC 8.1 to the current mangled
9567 (rs6000_init_builtins): Make __ibm128 use the long double type if
9568 long double is IBM extended double. Make __float128 use the long
9569 double type if long double is IEEE 128-bit.
9572 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
9573 macro for __ibm128 built-in functions.
9574 (PACK_IF): Add __ibm128 pack/unpack functions.
9575 (UNPACK_IF): Likewise.
9576 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
9577 enable long double built-in functions if long double is IEEE
9578 128-bit floating point.
9579 (rs6000_invalid_builtin): Update long double built-in function
9581 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
9582 functions, adjust the built-in function to use the long double
9583 built-in function if __ibm128 and long double are the same type.
9584 * doc/extend.texi (PowerPC builtins): Update documention for
9585 __builtin_{,un}pack_longdouble. Add documentation for
9586 __builtin_{,un}pack_ibm128.
9588 2018-06-06 Jim Wilson <jimw@sifive.com>
9590 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
9591 (struct machine_function): New field interrupt_mode.
9592 (riscv_handle_type_attribute): New function. Add forward declaration.
9593 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
9594 (riscv_expand_epilogue): Check interrupt_mode field.
9595 (riscv_set_current_function): Check interrupt attribute args and
9596 set interrupt_mode field.
9597 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
9598 (riscv_sret, riscv_uret): New.
9599 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
9600 new arguments to interrupt attribute.
9602 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
9605 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
9606 Don't handle -mcpu=power8 if -mpower9-vector is also used.
9608 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
9610 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
9611 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
9612 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
9613 several redundant entries.
9615 2018-06-06 David Malcolm <dmalcolm@redhat.com>
9617 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
9618 type from "rtx" to "rtx_insn *".
9619 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
9620 for local "call_insn", removing cast.
9621 (ix86_expand_call): Likewise, introducing a "call_insn" local.
9623 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
9625 PR tree-optimization/86066
9626 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
9627 for BIT_INSERT_EXPR stores.
9629 2018-06-06 Richard Biener <rguenther@suse.de>
9631 PR tree-optimization/86062
9632 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
9633 component refs ontop
9634 of to be offsetted base.
9636 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
9638 * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
9639 to be static and remove check on interrupt attribute name.
9641 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
9643 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
9644 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
9646 2018-06-05 Steve Ellcey <sellcey@cavium.com>
9649 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
9651 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
9652 Remove second argument, change how error is called.
9653 (aarch64_layout_arg): Remove second argument from
9654 aarch64_err_no_fpadvsimd call.
9655 (aarch64_init_cumulative_args): Ditto.
9656 (aarch64_gimplify_va_arg_expr): Ditto.
9657 * config/aarch64/aarch64.md (mov<mode>): Ditto.
9659 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
9661 * config/i386/i386.md (simple_return_indirect_internal): New expander.
9662 (*simple_return_indirect_internal<mode>): Rename from
9663 simple_return_indirect_internal. Use W mode iterator.
9664 (rstorssp): New expander.
9665 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
9666 (clrssbsy): New expander.
9667 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
9669 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
9671 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
9673 (cmse_check_pointed_object): Likewise.
9675 2018-06-05 Martin Liska <mliska@suse.cz>
9677 PR gcov-profile/47618
9678 * doc/invoke.texi: Document how -fprofile-dir format
9681 2018-06-05 Richard Biener <rguenther@suse.de>
9683 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
9684 removal pretend DOM info isn't available so we do not update
9685 it and only remove edges, not dominated blocks. Actually free
9686 DOM info in case we removed something. Remove unreachable blocks.
9687 (mfb_keep_latches): Work with either DOM info or marked backedges.
9688 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
9689 first. Mark backedges if DOM info isn't available.
9690 (Re-)compute DOM info after cleanup_control_flow_pre.
9692 2018-06-05 Richard Biener <rguenther@suse.de>
9694 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
9695 (locus_discrim_hasher::hash): Adjust.
9696 (locus_discrim_hasher::equal): Likewise.
9697 (next_discriminator_for_locus): Work on line directly.
9698 (same_line_p): Pass in expanded locus1 as well.
9699 (assign_discriminators): Avoid redundant location expansions.
9701 2018-06-05 Richard Biener <rguenther@suse.de>
9703 PR tree-optimization/86046
9704 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
9705 if required after clearing TREE_ADDRESSABLE.
9707 2018-06-05 Richard Biener <rguenther@suse.de>
9709 PR tree-optimization/86047
9710 * tree-ssa-loop.c (for_each_index): Glob handling of all
9711 decls and constants and really handle all of them.
9713 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9716 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
9717 qualifier_void_pointer and qualifier_const_void_pointer.
9718 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
9719 (arm_init_builtins): Handle the above.
9720 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
9721 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
9722 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
9725 2018-06-05 Martin Liska <mliska@suse.cz>
9727 * auto-profile.c (read_autofdo_file): Do not use
9728 gcov_ctr_summary struct.
9729 (afdo_callsite_hot_enough_for_early_inline): Likewise.
9730 * coverage.c (struct counts_entry): Likewise.
9731 (read_counts_file): Read just single summary entry.
9732 (get_coverage_counts): Use gcov_summary struct.
9733 * coverage.h (get_coverage_counts): Likewise.
9734 * gcov-dump.c (dump_working_sets): Likewise.
9735 (tag_summary): Dump just single summary.
9736 * gcov-io.c (gcov_write_summary): Write just histogram
9738 (gcov_read_summary): Read just single summary.
9739 (compute_working_sets): Use gcov_summary struct.
9740 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
9741 of GCOV_COUNTERS_SUMMABLE.
9742 (GCOV_COUNTERS_SUMMABLE): Remove.
9743 (GCOV_FIRST_VALUE_COUNTER): Replace with
9744 GCOV_COUNTER_V_INTERVAL.
9745 (struct gcov_ctr_summary): Remove.
9746 (struct gcov_summary): Directly use fields of former
9748 (compute_working_sets): Use gcov_summary struct.
9749 * gcov.c (read_count_file): Do not use ctrs fields.
9750 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
9752 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
9754 * profile.c: Likewise.
9755 * profile.h: Likewise.
9757 2018-06-05 Martin Liska <mliska@suse.cz>
9759 PR gcov-profile/84846
9760 * gcov.c (output_lines): Print working directory only
9761 in intermediate format.
9763 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
9765 * config/s390/s390-builtin-types.def: Add void function type.
9766 * config/s390/s390-builtins.def: Use the function type for the
9769 2018-06-04 Jim Wilson <jimw@sifive.com>
9771 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
9773 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
9774 and maybe_eh_return. Change regno to unsigned int. Use new args to
9775 handle EH_RETURN_DATA_REGNO registers properly.
9776 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
9777 (riscv_expand_epilogue): Update comment. Change argument name and
9778 type. Update code to use new name and type. Pass new args to
9779 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
9781 * config/riscv/riscv.md (NORMAL_RETURN): New.
9782 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
9783 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
9784 (eh_return): Call gen_eh_return_internal and emit barrier.
9785 (eh_return_internal): Call riscv_expand_epilogue.
9787 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
9789 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
9790 bit_insertion field and declare can_be_merged_into method.
9791 (merged_store_group::can_be_merged_into): New method.
9792 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
9793 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
9794 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
9796 2018-06-04 Richard Biener <rguenther@suse.de>
9798 PR tree-optimization/85955
9799 * builtins.c (fold_builtin_sincos): Convert pointers to
9800 destination to appropriate type before dereferencing.
9802 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
9804 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
9806 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
9808 * expr.c (expand_expr_real_1): Force the operand into memory if
9809 its TYPE_MODE is BLKmode and if there is no integer mode for
9810 the number of bits being extracted.
9812 2018-06-04 Jakub Jelinek <jakub@redhat.com>
9816 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
9817 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
9819 2018-06-04 Richard Biener <rguenther@suse.de>
9821 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
9822 (cleanup_tree_cfg_noloop): ... single caller. Do
9823 start_recording_case_labels later.
9825 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
9827 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
9828 to _IMMINTRIN_H_INCLUDED.
9829 * config/i386/pconfigintrin.h: Ditto.
9830 * config/i386/waitpkgintrin.h: Ditto.
9831 * config/i386/immintrin.h: Add includes for sgxintrin.h,
9832 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
9833 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
9834 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
9835 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
9836 waitpkgintrin.h and cldemoteintrin.h.
9838 2018-06-04 Richard Biener <rguenther@suse.de>
9840 PR tree-optimization/86038
9841 * tracer.c (find_best_successor): Check probability for
9842 being initialized, bail out if not.
9844 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
9847 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
9848 of bits to ignore when comparing architectures.
9850 2018-06-04 Jakub Jelinek <jakub@redhat.com>
9852 PR tree-optimization/69615
9853 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
9854 maximum or minimum of the type, try to merge it also as if
9855 range1 is + [-, x - 1] or + [x + 1, -].
9858 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
9860 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
9862 PR tree-optimization/86034
9863 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
9864 the unsigned bitfield type in a bit insertion sequence if it does not
9865 have a larger precision than the bitfield size.
9866 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
9868 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
9870 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
9872 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
9874 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
9875 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
9876 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
9877 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
9879 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
9881 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
9882 Disable -fdelete-null-pointer-checks for ELF toolchain.
9884 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
9885 Kito Cheng <kito.cheng@gmail.com>
9887 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
9888 (nds32le-*-*, nds32be-*-*): Integrate checking process.
9889 (nds32*-*-*): Add glibc and uclibc conditions.
9890 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
9891 (TARGET_EXCEPT_UNWIND_INFO): Define.
9892 * config/nds32/elf.h: New file.
9893 * config/nds32/linux.h: New file.
9894 * config/nds32/nds32-elf.opt: New file.
9895 * config/nds32/nds32-linux.opt: New file.
9896 * config/nds32/nds32-fp-as-gp.c
9897 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
9898 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
9900 (nds32_asm_file_end): Ditto.
9901 (nds32_print_operand): Ditto.
9902 (nds32_insert_attributes): Ditto.
9903 (nds32_init_libfuncs): New function.
9904 (TARGET_HAVE_TLS): Define.
9905 (TARGET_INIT_LIBFUNCS): Define.
9906 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
9908 (TARGET_ELF): Apply different mcmodel setting.
9909 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
9910 been migrated into elf.h and linux.h files.
9911 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
9912 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
9913 (mcmodel): The content has been migrated into nds32-elf.opt and
9914 nds32-linux.opt files.
9915 * config/nds32/t-elf: New file.
9916 * config/nds32/t-linux: New file.
9918 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
9919 Shiva Chen <shiva0217@gmail.com>
9921 * config/nds32/constants.md (unspec_volatile_element): Add
9922 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
9923 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
9925 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
9926 (make_pass_nds32_fp_as_gp): Declare.
9927 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
9929 (nds32_asm_function_end_prologue): Remove unused asm output.
9930 (nds32_asm_function_begin_epilogue): Remove unused asm output.
9931 (nds32_asm_file_start): Output necessary fp_as_gp information.
9932 (nds32_option_override): Adjust register usage.
9933 (nds32_expand_prologue): Consider fp_as_gp situation.
9934 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
9935 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
9938 (simple_return): Ditto.
9939 (omit_fp_begin): Output special directive for fp_as_gp.
9940 (omit_fp_end): Output special directive for fp_as_gp.
9941 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
9942 mforbid-fp-as-gp): New options.
9944 2018-06-01 Mark Wielaard <mark@klomp.org>
9946 * dwarf2out.c (dwarf2out_finish): Remove generation of
9947 DW_AT_loclists_base.
9949 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
9951 * gimple-ssa-store-merging.c: Include gimple-fold.h.
9952 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
9953 (struct merged_store_group): Add bit_insertion field.
9954 (dump_char_array): Use standard hexadecimal format.
9955 (merged_store_group::merged_store_group): Set bit_insertion to false.
9956 (merged_store_group::apply_stores): Use optimal buffer size. Deal
9957 with BIT_INSERT_EXPR stores. Move up code updating the mask and
9958 also print the mask in the dump file.
9959 (pass_store_merging::gate): Minor tweak.
9960 (imm_store_chain_info::coalesce_immediate): Fix wrong association
9961 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
9962 stores with INTEGER_CST stores.
9963 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
9964 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
9965 and use it throughout. Generate bit insertion sequences if need be.
9966 (pass_store_merging::process_store): Remove redundant condition.
9967 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
9969 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
9971 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
9972 the 128-bit floating point types. Fix function comment.
9974 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9976 * config/aarch64/aarch64-simd.md
9977 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
9979 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
9982 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
9984 PR tree-optimization/85989
9985 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
9987 (backprop::intersect_uses): Check it when deciding whether this
9988 is a backedge reference.
9989 (backprop::process_block): Add each phi to m_visited_phis
9990 after visiting it, then clear it at the end.
9992 2018-06-01 Richard Biener <rguenther@suse.de>
9994 * tree-vectorizer.h (vect_dr_stmt): New function.
9995 (vect_get_load_cost): Adjust.
9996 (vect_get_store_cost): Likewise.
9997 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
9998 Use vect_dr_stmt instead of DR_SMTT.
9999 (vect_record_base_alignments): Likewise.
10000 (vect_calculate_target_alignment): Likewise.
10001 (vect_compute_data_ref_alignment): Likewise and make static.
10002 (vect_update_misalignment_for_peel): Likewise.
10003 (vect_verify_datarefs_alignment): Likewise.
10004 (vector_alignment_reachable_p): Likewise.
10005 (vect_get_data_access_cost): Likewise. Pass down
10006 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
10007 (vect_get_peeling_costs_all_drs): Likewise.
10008 (vect_peeling_hash_get_lowest_cost): Likewise.
10009 (vect_enhance_data_refs_alignment): Likewise.
10010 (vect_find_same_alignment_drs): Likewise.
10011 (vect_analyze_data_refs_alignment): Likewise.
10012 (vect_analyze_group_access_1): Likewise.
10013 (vect_analyze_group_access): Likewise.
10014 (vect_analyze_data_ref_access): Likewise.
10015 (vect_analyze_data_ref_accesses): Likewise.
10016 (vect_vfa_segment_size): Likewise.
10017 (vect_small_gap_p): Likewise.
10018 (vectorizable_with_step_bound_p): Likewise.
10019 (vect_prune_runtime_alias_test_list): Likewise.
10020 (vect_analyze_data_refs): Likewise.
10021 (vect_supportable_dr_alignment): Likewise.
10022 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
10023 (vect_gen_prolog_loop_niters): Likewise.
10024 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
10025 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
10027 (vect_recog_mask_conversion_pattern): Likewise.
10028 (vect_try_gather_scatter_pattern): Likewise.
10029 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
10030 to vect_get_store_cost.
10031 (vect_get_store_cost): Get stmt_info instead of DR.
10032 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
10033 (vect_get_load_cost): Get stmt_info instead of DR.
10035 2018-06-01 Richard Biener <rguenther@suse.de>
10037 PR middle-end/86017
10038 * gimple-fold.c (var_decl_component_p): Also allow offsetted
10039 vars wrapped in MEM_REFs.
10041 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
10043 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
10044 Fix subreg tests so that we only return a choice between
10045 GENERAL_REGS and FP_REGS if the original classes included both.
10047 2018-06-01 Richard Biener <rguenther@suse.de>
10050 * tree-ssa-structalias.c (get_function_part_constraint):
10051 Handle NULL fi->decl.
10052 (find_func_aliases_for_call): Properly handle indirect
10053 fi from direct call.
10054 (find_func_clobbers): Likewise.
10055 (ipa_pta_execute): Likewise.
10056 (create_variable_info_for): For functions that are ifunc_resolver
10057 resolve to a varinfo that contains the result of the resolver call.
10058 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
10061 2018-05-31 Michael Collison <michael.collison@arm.com>
10063 * config/aarch64/aarch64.md:
10064 (*fix_to_zero_extenddfdi2): New pattern.
10065 * gcc.target/aarch64/fix_extend1.c: New testcase.
10067 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
10069 PR middle-end/78809
10070 PR middle-end/83026
10071 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
10072 and BUILT_IN_STRNCMP_EQ.
10073 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
10074 BUILT_IN_STRNCMP_EQ.
10075 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
10076 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
10077 (gimple_fold_builtin): Likewise.
10078 * tree-ssa-strlen.c (compute_string_length): New function.
10079 (determine_min_obsize): New function.
10080 (handle_builtin_string_cmp): New function to handle calls to
10081 string compare functions.
10082 (strlen_optimize_stmt): Add handling to builtin string compare
10084 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
10085 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
10086 * tree.c (build_common_builtin_nodes): Add new defines of
10087 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
10089 2018-05-31 Jakub Jelinek <jakub@redhat.com>
10092 * bb-reorder.c (pass_partition_blocks::gate): Return false for
10093 functions with naked attribute.
10095 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
10097 * config/i386/sse.md (avx_vec_concat<mode>):
10098 Substitute concat_tg_mode mode attribute with xtg_mode.
10099 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
10100 (concat_tg_mode): Remove mode attribute.
10102 2018-05-31 Martin Sebor <msebor@redhat.com>
10105 * calls.c (alloc_max_size): Correct a logic error/typo.
10106 Treat excessive arguments as infinite. Warn for invalid arguments.
10107 * doc/invoke.texi (-Walloc-size-larger-than): Update.
10109 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
10112 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
10113 and movx for Haswell.
10115 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
10116 Cesar Philippidis <cesar@codesourcery.com>
10118 PR middle-end/85879
10119 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
10120 when emitting error on private/firstprivate reductions.
10121 * omp-low.c (lower_omp_target): Avoid reference-type processing
10122 on pointers for firstprivate clause.
10124 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
10126 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
10129 * config/aarch64/aarch64-simd.md
10130 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
10131 (aarch64_ld1_x3_<mode>): Likewise
10132 (aarch64_st1x2<VALLDIF:mode>): Likewise
10133 (aarch64_st1_x2_<mode>): Likewise
10134 (aarch64_st1x3<VALLDIF:mode>): Likewise
10135 (aarch64_st1_x3_<mode>): Likewise
10136 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
10137 (vld1_s8_x3): Likewise.
10138 (vld1_u16_x3): Likewise.
10139 (vld1_s16_x3): Likewise.
10140 (vld1_u32_x3): Likewise.
10141 (vld1_s32_x3): Likewise.
10142 (vld1_u64_x3): Likewise.
10143 (vld1_s64_x3): Likewise.
10144 (vld1_f16_x3): Likewise.
10145 (vld1_f32_x3): Likewise.
10146 (vld1_f64_x3): Likewise.
10147 (vld1_p8_x3): Likewise.
10148 (vld1_p16_x3): Likewise.
10149 (vld1_p64_x3): Likewise.
10150 (vld1q_u8_x3): Likewise.
10151 (vld1q_s8_x3): Likewise.
10152 (vld1q_u16_x3): Likewise.
10153 (vld1q_s16_x3): Likewise.
10154 (vld1q_u32_x3): Likewise.
10155 (vld1q_s32_x3): Likewise.
10156 (vld1q_u64_x3): Likewise.
10157 (vld1q_s64_x3): Likewise.
10158 (vld1q_f16_x3): Likewise.
10159 (vld1q_f32_x3): Likewise.
10160 (vld1q_f64_x3): Likewise.
10161 (vld1q_p8_x3): Likewise.
10162 (vld1q_p16_x3): Likewise.
10163 (vld1q_p64_x3): Likewise.
10164 (vst1_s64_x2): Likewise.
10165 (vst1_u64_x2): Likewise.
10166 (vst1_f64_x2): Likewise.
10167 (vst1_s8_x2): Likewise.
10168 (vst1_p8_x2): Likewise.
10169 (vst1_s16_x2): Likewise.
10170 (vst1_p16_x2): Likewise.
10171 (vst1_s32_x2): Likewise.
10172 (vst1_u8_x2): Likewise.
10173 (vst1_u16_x2): Likewise.
10174 (vst1_u32_x2): Likewise.
10175 (vst1_f16_x2): Likewise.
10176 (vst1_f32_x2): Likewise.
10177 (vst1_p64_x2): Likewise.
10178 (vst1q_s8_x2): Likewise.
10179 (vst1q_p8_x2): Likewise.
10180 (vst1q_s16_x2): Likewise.
10181 (vst1q_p16_x2): Likewise.
10182 (vst1q_s32_x2): Likewise.
10183 (vst1q_s64_x2): Likewise.
10184 (vst1q_u8_x2): Likewise.
10185 (vst1q_u16_x2): Likewise.
10186 (vst1q_u32_x2): Likewise.
10187 (vst1q_u64_x2): Likewise.
10188 (vst1q_f16_x2): Likewise.
10189 (vst1q_f32_x2): Likewise.
10190 (vst1q_f64_x2): Likewise.
10191 (vst1q_p64_x2): Likewise.
10192 (vst1_s64_x3): Likewise.
10193 (vst1_u64_x3): Likewise.
10194 (vst1_f64_x3): Likewise.
10195 (vst1_s8_x3): Likewise.
10196 (vst1_p8_x3): Likewise.
10197 (vst1_s16_x3): Likewise.
10198 (vst1_p16_x3): Likewise.
10199 (vst1_s32_x3): Likewise.
10200 (vst1_u8_x3): Likewise.
10201 (vst1_u16_x3): Likewise.
10202 (vst1_u32_x3): Likewise.
10203 (vst1_f16_x3): Likewise.
10204 (vst1_f32_x3): Likewise.
10205 (vst1_p64_x3): Likewise.
10206 (vst1q_s8_x3): Likewise.
10207 (vst1q_p8_x3): Likewise.
10208 (vst1q_s16_x3): Likewise.
10209 (vst1q_p16_x3): Likewise.
10210 (vst1q_s32_x3): Likewise.
10211 (vst1q_s64_x3): Likewise.
10212 (vst1q_u8_x3): Likewise.
10213 (vst1q_u16_x3): Likewise.
10214 (vst1q_u32_x3): Likewise.
10215 (vst1q_u64_x3): Likewise.
10216 (vst1q_f16_x3): Likewise.
10217 (vst1q_f32_x3): Likewise.
10218 (vst1q_f64_x3): Likewise.
10219 (vst1q_p64_x3): Likewise.
10221 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
10223 * config/msp430/msp430.c (msp430_output_labelref): Prepend
10224 user_label_prefix to name.
10226 * tree-core.h: Update comment about the format of NAME string
10227 passed to handler in attribute_spec.
10229 * config/msp430/msp430.md: Remove erroneous subreg expression from
10230 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
10231 zero_extend{q,h}isi2.
10233 2018-05-30 Borislav Petkov <bp@suse.de>
10235 * doc/extend.texi: Document some architecture specific
10236 constraints and sort entries.
10238 2018-05-30 Martin Sebor <msebor@redhat.com>
10240 PR middle-end/85369
10241 * builtins.c (expand_builtin_stpcpy_1): New function.
10242 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
10243 only if the former succeeds.
10245 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
10247 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
10250 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
10252 * doc/invoke.texi (-flinker-output): Document
10254 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
10256 * passes.c (ipa_write_summaries): Only modify statements if body
10258 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
10259 incrementally linking.
10260 (ipa_passes): Likewise.
10261 * lto-cgraph.c (lto_output_node): When incrementally linking do not
10262 pass down resolution info.
10263 * common.opt (flag_incremental_link): Update info.
10264 * gcc.c (plugin specs): Turn flinker-output=* to
10265 -plugin-opt=-linker-output-known
10266 * toplev.c (compile_file): Also cut compilation when doing incremental
10268 * flag-types. (enum lto_partition_model): Add
10269 LTO_LINKER_OUTPUT_NOLTOREL.
10270 (invoke.texi): Add -flinker-output docs.
10271 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
10272 link same way as WPA; do not stream in dead initializers.
10274 * dwarf2out.c (dwarf2out_die_ref_for_decl,
10275 darf2out_register_external_decl): Support incremental link.
10277 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
10279 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
10281 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
10283 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
10284 it down to simple_object_copy_lto_debug_sections.
10285 (run_gcc): Determine incremental LTO link time and configure
10286 lto1 into non-wpa mode, disable renaming of debug sections.
10288 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
10290 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
10291 descriptions of various incorrectly documented functions.
10293 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
10296 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
10297 address check not strict.
10299 2018-05-30 Richard Biener <rguenther@suse.de>
10301 PR tree-optimization/85964
10302 * tracer.c (better_p): Drop initialized count check, we only
10303 call the function with initialized counts now.
10304 (find_best_successor): Do find a best edge if one
10305 has uninitialized count.
10306 (find_best_predecessor): Likewise. Do BB frequency check only
10307 if count is initialized.
10309 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
10311 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
10312 (aarch64_ldrstr_offset_compare): New.
10313 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
10314 load/store orderings.
10315 (aarch64_gen_adjusted_ldpstp): Likewise.
10317 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
10319 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
10320 Check for subset of GENERAL_REGS and FP_REGS.
10321 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
10324 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
10326 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
10327 and wi::to_poly_offset. Add the current offset and then check
10328 whether the sum fits, rather than using an unchecked addition of
10329 a checked term. Check for a shwi rather than a uhwi.
10330 * expr.c (get_bit_range): Use tree_to_poly_uint64.
10331 (store_constructor): Use poly_int_tree_p.
10332 (expand_expr_real_1): Likewise.
10333 * function.c (assign_temp): Likewise.
10334 * fold-const.c (const_binop): Use poly_int_tree_p and
10335 wi::to_poly_offset.
10336 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
10338 * ipa-icf-gimple.c (func_checker::compare_operand): Use
10339 to_poly_offset for MEM offsets.
10340 * ipa-icf.c (sem_variable::equals): Likewise.
10341 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
10342 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
10343 wi::to_poly_offset for BIT_FIELD_REF offsets.
10344 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
10345 wi::to_poly_offset.
10346 * var-tracking.c (emit_note_insn_var_location): Use
10347 tree_to_poly_uint64.
10349 2018-05-29 Jim Wilson <jimw@sifive.com>
10351 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
10353 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
10356 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
10357 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
10359 (sse4_1_round<mode>2): Use nonimmediate_operand
10360 for operand 1 predicate.
10362 2018-05-29 Martin Sebor <msebor@redhat.com>
10363 Richard Biener <rguenther@suse.de>
10366 * calls.c (get_size_range): Call determine_value_range instead
10367 of get_value_range..
10368 * tree-vrp.h (determine_value_range): Declared new function.
10369 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
10371 2018-05-29 Richard Biener <rguenther@suse.de>
10373 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
10374 sure to use non-pattern stmts for get_earlier_stmt arguments.
10375 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
10376 called on pattern stmts.
10377 (get_later_stmt): Likewise.
10379 2018-05-29 Martin Liska <mliska@suse.cz>
10381 PR gcov-profile/85759
10382 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
10385 2018-05-29 Jakub Jelinek <jakub@redhat.com>
10387 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
10389 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
10393 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
10394 VEC_PACK_FLOAT_EXPR): New tree codes.
10395 * tree-pretty-print.c (op_code_prio): Handle
10396 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
10397 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
10398 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
10399 * tree-inline.c (estimate_operator_cost): Likewise.
10400 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
10401 * fold-const.c (const_binop): Likewise.
10402 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
10403 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
10404 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
10405 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
10406 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
10407 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
10408 * expr.c (expand_expr_real_2): Likewise.
10409 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
10410 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
10411 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
10413 * optabs.c (expand_widen_pattern_expr): For
10414 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
10415 sign from result type rather than operand's type.
10416 (expand_binop_directly): For vec_packu_float_optab and
10417 vec_packs_float_optab allow result type to be different from operand's
10419 * optabs-tree.c (optab_for_tree_code): Handle
10420 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
10421 VEC_PACK_FLOAT_EXPR. Formatting fixes.
10422 * tree-vect-generic.c (expand_vector_operations_1): Handle
10423 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
10424 VEC_PACK_FLOAT_EXPR.
10425 * tree-vect-stmts.c (supportable_widening_operation): Handle
10427 (supportable_narrowing_operation): Handle FLOAT_EXPR.
10428 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
10429 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
10430 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
10431 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
10433 (vec_pack<floatprefix>_float_<mode>): New expander.
10434 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
10436 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
10437 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
10438 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
10439 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
10440 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
10442 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
10443 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
10444 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
10445 VEC_PACK_FLOAT_EXPR): Document.
10447 2018-05-29 Richard Biener <rguenther@suse.de>
10449 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
10451 (stmt_vec_info_vec): Make pointer.
10452 (init_stmt_vec_info_vec): Remove.
10453 (free_stmt_vec_info_vec): Likewise.
10454 (set_stmt_vec_info_vec): New function.
10455 (free_stmt_vec_infos): Likewise.
10456 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
10457 (set_vinfo_for_stmt): Likewise.
10458 (get_earlier_stmt): Likewise.
10459 (get_later_stmt): Likewise.
10460 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
10461 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
10462 (vec_info::~vec_info): Free stmt_vec_infos.
10463 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
10464 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
10465 (pass_slp_vectorize::execute): Likewise.
10466 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
10467 (free_stmt_vec_info_vec): Likewise.
10468 (set_stmt_vec_info_vec): New function.
10469 (free_stmt_vec_infos): Likewise.
10470 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
10471 the global stmt_vec_info_vec.
10472 * tree-parloops.c (gather_scalar_reductions): Use
10473 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
10476 2018-05-29 Richard Biener <rguenther@suse.de>
10478 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
10480 2018-05-29 Martin Liska <mliska@suse.cz>
10481 David Malcolm <dmalcolm@redhat.com>
10483 * vec.c (test_reverse): New.
10484 (vec_c_tests): Add new test.
10485 * vec.h (vl_ptr>::reverse): New function.
10487 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
10489 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
10491 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
10494 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
10496 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
10498 2018-05-28 Richard Biener <rguenther@suse.de>
10500 PR tree-optimization/85933
10501 * tree-vect-data-refs.c (vect_record_base_alignments): Only
10502 look at stmts marked as vectorizable.
10504 2018-05-28 Richard Biener <rguenther@suse.de>
10506 PR tree-optimization/85934
10507 * tree-vect-generic.c (expand_vector_operations_1): Hoist
10508 vector boolean check before scalar optimization.
10510 2018-05-28 Jakub Jelinek <jakub@redhat.com>
10512 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
10515 2018-05-28 Mark Wielaard <mark@klomp.org>
10517 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
10518 if it is an expression containing a minus sign.
10520 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
10522 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
10524 2018-05-27 Paul Koning <ni1d@arrl.net>
10526 * config/pdp11/pdp11.md (truncsihi2): Remove.
10528 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
10529 Chung-Ju Wu <jasonwucj@gmail.com>
10531 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
10533 (unaligned_store_dw): Ditto.
10534 * config/nds32/nds32-memory-manipulation.c
10535 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
10536 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
10537 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
10538 (emit_setmem_word_loop): Rename to ...
10539 (emit_setmem_doubleword_loop): ... this.
10540 (nds32_gen_dup_4_byte_to_word_value): New function.
10541 (nds32_gen_dup_8_byte_to_double_word_value): New function.
10542 (nds32_expand_setmem_loop): Refine implementation.
10543 (nds32_expand_setmem_loop_v3m): Ditto.
10544 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
10547 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
10549 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
10551 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
10553 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
10554 (nds32_init_machine_status): Initialize machine->attr_naked_p and
10555 machine->attr_no_prologue_p.
10556 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
10557 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
10558 (nds32_expand_epilogue): Consider attr_naked_p.
10559 (nds32_expand_epilogue_v3pop): Likewise.
10560 (nds32_can_use_return_insn): Likewise.
10561 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
10562 attr_no_prologue_p fields.
10563 * config/nds32/nds32.opt (mret-in-naked-func): New option.
10565 2018-05-27 Jakub Jelinek <jakub@redhat.com>
10568 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
10570 * config/i386/sse.md
10571 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
10573 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
10575 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
10577 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
10579 (*<floatsuffix>floatv2div2sf2): Rename to ...
10580 (*float<floatunssuffix>v2div2sf2): ... this.
10581 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
10582 (float<floatunssuffix>v2div2sf2_mask): ... this.
10583 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
10584 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
10585 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
10587 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
10589 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
10591 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
10593 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
10595 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
10597 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
10598 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
10599 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
10600 gen_ufix_truncv8dfv8si2.
10601 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
10602 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
10603 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
10604 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
10605 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
10606 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
10607 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
10608 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
10610 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
10614 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
10616 2018-05-25 Jim Wilson <jimw@sifive.com>
10618 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
10619 * config/riscv/riscv.c (struct machine_function): Add
10620 interrupt_handler_p and attribute_checked_p fields.
10621 (riscv_attribute_table): Add interrupt.
10622 (riscv_interrupt_type_p): New.
10623 (riscv_save_reg_p): Save extra regs for interrupt handler.
10624 (riscv_use_save_libcall): Return false for interrupt handler.
10625 (riscv_first_stack_step): Add forward declaration.
10626 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
10627 for interrupt handler with large frame. Use it for saved reg list.
10628 (riscv_expand_prologue): Move flag_stack_usage_info support to
10629 eliminate duplication.
10630 (riscv_expand_epilogue): Generate mret for interrupt handler.
10631 (riscv_epilogue_uses): New.
10632 (riscv_can_use_return_insn): Return false for interrupt handler.
10633 (riscv_function_ok_for_sibcall): Likewise.
10634 (riscv_set_current_function): Add interrupt handler support.
10635 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
10636 * config/riscv/riscv.md (UNSPECV_MRET): New.
10638 (riscv_frflags, riscv_fsflags): Use tab after opcode.
10640 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
10642 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
10644 PR tree-optimization/85712
10645 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
10646 this candidate has already been replaced in-situ by a copy.
10648 2018-05-25 Jason Merrill <jason@redhat.com>
10650 PR c++/80485 - inline function non-zero address.
10651 * symtab.c (nonzero_address): Check DECL_COMDAT.
10653 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
10656 * config/alpha/alpha.md (ashlsi3): New insn pattern.
10657 (*ashlsi_se): Rename from *ashldi_se. Define as sign
10658 extension of SImode operation. Use const123_operand predicate.
10659 (*saddsi_1): Remove.
10660 (*saddl_se_1): Ditto.
10661 (*ssubsi_1): Ditto.
10662 (*ssubl_se_1): Ditto.
10663 * config/alpha/predicates.md (const123_operand): New predicate.
10664 * config/alpha/constraints.md (P): Use IN_RANGE.
10666 2018-05-25 Richard Biener <rguenther@suse.de>
10668 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
10670 (ref_maybe_used_by_stmt_p): Likewise.
10671 (stmt_may_clobber_ref_p): Likewise.
10672 (stmt_may_clobber_ref_p_1): Likewise.
10673 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
10675 (ref_maybe_used_by_stmt_p): Likewise.
10676 (stmt_may_clobber_ref_p): Likewise.
10677 (stmt_may_clobber_ref_p_1): Likewise.
10678 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
10679 the alias oracle to disambiguate DRs with stmts DR analysis
10681 (vect_analyze_data_refs): Do not give up on not analyzable
10682 DRs for BB vectorization. Remove code truncating the dataref
10685 2018-05-25 Jakub Jelinek <jakub@redhat.com>
10688 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
10689 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
10690 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
10692 2018-05-25 Richard Biener <rguenther@suse.de>
10694 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
10695 function, combining stmt data ref gathering and fatal analysis
10697 (vect_analyze_data_refs): Remove now redudnant code and simplify.
10698 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
10699 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
10700 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
10701 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
10703 2018-05-25 Bin Cheng <bin.cheng@arm.com>
10705 PR tree-optimization/85720
10706 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
10707 SCC if all partitions are builtins.
10708 (version_loop_by_alias_check): New parameter. Generate cancelable
10709 runtime alias check if all partitions are builtins.
10710 (distribute_loop): Update call to above function.
10712 2018-05-25 Bin Cheng <bin.cheng@arm.com>
10714 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
10715 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
10716 (parm_default_def_partition_arg): Ditto.
10717 (set_parm_default_def_partition): Ditto.
10718 (get_parm_default_def_partitions): Ditto and make it static.
10719 (get_undefined_value_partitions): Ditto and make it static.
10720 (remove_ssa_form): Refactor call to init_var_map here.
10721 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
10722 computation for loop region.
10723 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
10724 (register_default_def): Delete.
10725 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
10726 (parm_default_def_partition_arg): Ditto.
10727 (set_parm_default_def_partition): Ditto.
10728 (get_parm_default_def_partitions): Ditto and make it static.
10729 (get_undefined_value_partitions): Ditto and make it static.
10730 (coalesce_with_default, coalesce_with_default): Update comment.
10731 (create_coalesce_list_for_region): New func factored out from
10732 create_outofssa_var_map.
10733 (populate_coalesce_list_for_outofssa): New func factored out from
10734 create_outofssa_var_map and coalesce_ssa_name.
10735 (create_outofssa_var_map): Delete.
10736 (coalesce_ssa_name): Refactor to support live range computation.
10737 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
10738 (get_parm_default_def_partitions): Delete.
10739 (get_undefined_value_partitions): Ditto.
10740 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
10741 computation for loop region.
10742 (new_tree_live_info, loe_visit_block): Ditto.
10743 (live_worklist, set_var_live_on_entry): Ditto.
10744 (calculate_live_on_exit, verify_live_on_entry): Ditto.
10745 * tree-ssa-live.h (struct _var_map): New fields.
10746 (init_var_map): Change decl.
10747 (region_contains_p): New.
10749 2018-05-25 Bin Cheng <bin.cheng@arm.com>
10751 * tree-ssa-live.h (live_merge_and_clear): Delete.
10753 2018-05-25 Richard Biener <rguenther@suse.de>
10756 * tree-dump.c (dequeue_and_dump): Remove access to removed
10757 operand 2 of a SWITCH_EXPR.
10759 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
10761 * doc/sourcebuild.texi (vect_double_cond_arith): Include
10762 multiplication and division.
10763 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
10764 (cond_udiv@var{m}, cond_umod@var{m}): Document.
10765 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
10766 (cond_udiv_optab, cond_umod_optab): New optabs.
10767 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
10768 (IFN_COND_RDIV): New internal functions.
10769 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
10770 TRUNC_MOD_EXPR and RDIV_EXPR.
10771 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
10772 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
10774 (SVE_INT_BINARY): Include mult.
10775 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
10776 (optab, sve_int_op): Handle mult.
10777 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
10779 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
10780 for SVE_INT_BINARY_SD.
10782 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
10784 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
10785 (optab, sve_int_op): Handle div and udiv.
10786 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
10787 for SVE_INT_BINARY_SD.
10788 (*<optab><mode>3): New insn for the same.
10790 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
10792 * tree-vect-patterns.c: Include predict.h.
10793 (vect_recog_divmod_pattern): Restrict check for division support
10794 to when optimizing for size.
10796 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
10798 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
10799 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
10800 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
10801 (gimple_match_op::set_op): Likewise.
10802 (gimple_resimplify4): Declare.
10803 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
10804 (expr::gen_transform): Likewise.
10805 (decision_tree::gen): Generate a simplification routine for 4 operands.
10806 * gimple-match-head.c (gimple_simplify): Add an overload for
10807 4 operands. In the top-level function, handle up to 4 call
10808 arguments and call gimple_resimplify4.
10809 (gimple_resimplify4): New function.
10810 (build_call_internal): Pass a fourth operand.
10811 (maybe_push_to_seq): Likewise.
10812 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
10813 Fold VEC_COND_EXPRs of an operation and a default value into
10814 an IFN_COND_* function if possible.
10815 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
10817 (SVE_COND_FP_BINARY): Include them.
10818 (optab, sve_fp_op): Handle them.
10819 (SVE_INT_BINARY_REV): New code iterator.
10820 (SVE_COND_FP_BINARY_REV): New int iterator.
10821 (commutative): New int attribute.
10822 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
10824 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
10826 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
10827 (*cond_<optab><mode>): New patterns for reversed operands.
10829 2018-05-25 Richard Biener <rguenther@suse.de>
10831 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
10832 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
10833 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
10834 (STMT_VINFO_GROUPED_ACCESS): Adjust.
10835 * tree-vect-data-refs.c (everywhere): Adjust users.
10836 * tree-vect-loop.c (everywhere): Likewise.
10837 * tree-vect-slp.c (everywhere): Likewise.
10838 * tree-vect-stmts.c (everywhere): Likewise.
10839 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
10841 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10843 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
10845 (gcc_cv_as_section_exclude): ... this.
10846 Try Solaris as #exclude syntax.
10847 * configure: Regenerate.
10848 * config.in: Regenerate.
10849 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
10851 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
10852 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
10854 * varasm.c (default_elf_asm_named_section): Don't check if
10855 HAVE_GAS_SECTION_EXCLUDE is defined.
10857 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
10859 * doc/md.texi: Update the documentation of the cond_* optabs
10860 to mention the new final operand. Fix GET_MODE_NUNITS call.
10861 Describe the scalar case too.
10862 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
10863 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
10865 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
10866 (get_conditional_internal_fn): Update comment.
10867 * tree-vect-loop.c (vectorizable_reduction): Pass the original
10868 accumulator value as a final argument to conditional functions.
10869 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
10870 a define_expand and add an "else" operand. Assert for now that
10871 the else operand is equal to operand 2. Use SVE_INT_BINARY and
10872 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
10873 (*cond_<optab><mode>): New patterns.
10874 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
10875 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
10876 (UNSPEC_COND_EOR): Delete.
10877 (optab): Remove associated mappings.
10878 (SVE_INT_BINARY): New code iterator.
10879 (sve_int_op): Remove int attribute and add "minus" to the code
10881 (SVE_COND_INT_OP): Delete.
10882 (SVE_COND_FP_OP): Rename to...
10883 (SVE_COND_FP_BINARY): ...this.
10885 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
10887 * optabs.c (can_reuse_operands_p): New function.
10888 (maybe_legitimize_operands): Try to reuse the results for
10891 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
10893 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
10894 Add {q} suffix to insn mnemonic.
10896 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
10898 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
10899 (msp430_warn_func_return): New.
10901 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
10903 * fold-const.c (tree_nonzero_bits): New function.
10904 * fold-const.h (tree_nonzero_bits): Likewise.
10905 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
10906 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
10908 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
10912 * varasm.c (assemble_alias): Check ifunc_resolver only on
10915 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
10918 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
10919 when memory input operand is handled.
10921 2018-05-24 Luis Machado <luis.machado@linaro.org>
10923 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
10925 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
10927 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
10929 * match.pd: Delay FMA folds until after vectorization.
10931 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
10934 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
10935 address check not strict.
10937 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
10939 * gimple-match.h (gimple_match_op): New class.
10940 (mprts_hook): Replace parameters with a gimple_match_op *.
10941 (maybe_build_generic_op): Likewise.
10942 (gimple_simplified_result_is_gimple_val): Replace parameters with
10943 a const gimple_match_op *.
10944 (gimple_simplify): Replace code_helper * and tree * parameters with
10945 a gimple_match_op * parameter.
10946 (gimple_resimplify1): Replace code_helper *, tree and tree *
10947 parameters with a gimple_match_op * parameter.
10948 (gimple_resimplify2): Likewise.
10949 (gimple_resimplify3): Likewise.
10950 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
10951 parameters with a gimple_match_op * parameter.
10952 * gimple-match-head.c (gimple_simplify): Change prototypes of
10953 auto-generated functions to take a gimple_match_op * instead of
10954 separate code_helper * and tree * parameters. Make the same
10955 change in the top-level overload and update calls to the
10956 gimple_resimplify routines. Update calls to the auto-generated
10957 functions and to maybe_push_res_to_seq in the publicly-facing
10958 operation-specific gimple_simplify overloads.
10959 (gimple_match_op::MAX_NUM_OPS): Define.
10960 (gimple_resimplify1): Replace rcode and ops with a single res_op
10961 parameter. Update call to gimple_simplify.
10962 (gimple_resimplify2): Likewise.
10963 (gimple_resimplify3): Likewise.
10964 (mprts_hook): Replace parameters with a gimple_match_op *.
10965 (maybe_build_generic_op): Likewise.
10966 (build_call_internal): Replace type, nargs and ops with
10967 a gimple_match_op *.
10968 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
10969 with a single gimple_match_op *. Update calls to mprts_hook,
10970 build_call_internal and gimple_simplified_result_is_gimple_val.
10971 Factor out code that is common to the tree_code and combined_fn cases.
10972 * genmatch.c (expr::gen_transform): Replace tem_code and
10973 tem_ops with a gimple_match_op called tem_op. Update calls
10974 to the gimple_resimplify functions and maybe_push_res_to_seq.
10975 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
10976 res_ops. Update call to the gimple_resimplify functions.
10977 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
10978 (decision_tree::gen): Make the functions take a gimple_match_op *
10979 called res_op instead of separate res_code and res_ops parameters.
10980 Update call accordingly.
10981 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
10982 and ops with a single res_op parameter. Update calls to
10983 maybe_build_generic_op and maybe_push_res_to_seq.
10984 (fold_stmt_1): Update calls to gimple_simplify and
10985 replace_stmt_with_simplification.
10986 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
10987 and gimple_simplified_result_is_gimple_val.
10988 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
10990 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
10991 with a gimple_match_op *.
10992 (vn_nary_build_or_lookup): Likewise. Update call to
10993 vn_nary_build_or_lookup_1.
10994 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
10995 gimple_match_op *. Update calls to the gimple_resimplify routines
10996 and to gimple_simplified_result_is_gimple_val.
10997 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
10998 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
10999 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
11000 (visit_nary_op): Likewise.
11001 (visit_reference_op_load): Likewise.
11003 2018-05-23 Luis Machado <luis.machado@linaro.org>
11005 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
11006 modifier for printing the step amount.
11008 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
11011 * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
11014 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
11016 * doc/sourcebuild.texi (Endianness): New subsubsection.
11018 2018-05-23 Luis Machado <luis.machado@linaro.org>
11020 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
11021 <prefetch_dynamic_strides>: New const bool field.
11022 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
11023 prefetch_dynamic_strides.
11024 (exynosm1_prefetch_tune): Likewise.
11025 (thunderxt88_prefetch_tune): Likewise.
11026 (thunderx_prefetch_tune): Likewise.
11027 (thunderx2t99_prefetch_tune): Likewise.
11028 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
11030 (aarch64_override_options_internal): Update to set
11031 PARAM_PREFETCH_DYNAMIC_STRIDES.
11032 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
11033 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
11034 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
11035 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
11036 prefetch-dynamic-strides setting.
11038 2018-05-23 Luis Machado <luis.machado@linaro.org>
11040 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
11041 <minimum_stride>: New const int field.
11042 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
11043 minimum_stride field defaulting to -1.
11044 (exynosm1_prefetch_tune): Likewise.
11045 (thunderxt88_prefetch_tune): Likewise.
11046 (thunderx_prefetch_tune): Likewise.
11047 (thunderx2t99_prefetch_tune): Likewise.
11048 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
11049 <default_opt_level>: Set to 3.
11050 (aarch64_override_options_internal): Update to set
11051 PARAM_PREFETCH_MINIMUM_STRIDE.
11052 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
11053 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
11054 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
11055 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
11056 stride is constant and is below the minimum stride threshold.
11058 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11060 * config/arm/arm-cpus.in (mode26): Delete.
11061 (armv4): Delete mode26 reference.
11062 * config/arm/arm.c (arm_configure_build_target): Delete use of
11065 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
11067 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
11069 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
11070 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
11072 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
11073 (floatunsdidf2): Ditto.
11075 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
11077 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
11078 (fixuns_trunc<mode>si2_avx512f): Ditto.
11079 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
11080 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
11081 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
11083 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
11085 PR rtl-optimization/79985
11086 * df-scan.c (df_insn_refs_collect): Remove special case for
11087 global registers and asm statements.
11089 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
11091 * extend.texi (Global Register Variables): Rewrite the bullet list.
11092 Note that the register is available for allocation. Note that access
11093 via inline asm must use constraints. Add note about async-signal
11094 handlers. Remove paragraph about automagic register selection.
11096 2018-05-23 Richard Biener <rguenther@suse.de>
11098 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
11099 of fixed offset from memset VN.
11101 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
11103 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
11104 first_interp field.
11105 (alloc_cand_and_find_basis): Initialize first_interp field.
11106 (slsr_process_mul): Modify first_interp field.
11107 (slsr_process_add): Likewise.
11108 (slsr_process_cast): Modify first_interp field for each new
11110 (slsr_process_copy): Likewise.
11111 (dump_candidate): Dump first_interp field.
11112 (replace_mult_candidate): Process all interpretations, not just
11114 (replace_rhs_if_not_dup): Likewise.
11115 (replace_one_candidate): Likewise.
11117 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
11119 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
11121 (aarch64_needs_frame_chain): New function.
11122 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
11124 2018-05-23 Sudakshina Das <sudi.das@arm.com>
11127 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
11128 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
11129 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
11130 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
11131 as true for strict-align.
11132 (aarch64_can_inline_p): Perform checks even when callee has no
11133 attributes to check for strict alignment.
11134 * doc/extend.texi (AArch64 Function Attributes): Document
11136 * doc/invoke.texi: (AArch64 Options): Likewise.
11138 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
11140 PR tree-optimization/85853
11141 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
11142 the handling of the root of the node to...
11143 (vect_slp_analyze_node_operations_1): ...this new function,
11144 and run the whole thing with the child nodes' def types
11145 set according to their SLP node's def type.
11147 2018-05-23 Richard Biener <rguenther@suse.de>
11149 PR middle-end/85874
11150 * tree-data-ref.c (create_runtime_alias_checks): Defer
11151 and ignore overflow warnings.
11153 2018-05-23 Yury Gribov <tetra2005@gmail.com>
11155 PR tree-optimization/85822
11156 * tree-vrp.c (is_masked_range_test): Fix handling of negative
11159 2018-05-23 Richard Biener <rguenther@suse.de>
11161 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
11162 memset constants via native_interpret_expr.
11164 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
11167 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
11169 (cgraph_node::create_alias): Likewise.
11170 (cgraph_node::get_availability): Check ifunc_resolver instead
11171 of looking up ifunc attribute.
11172 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
11173 * varasm.c (do_assemble_alias): Likewise.
11174 (assemble_alias): Likewise.
11175 (default_binds_local_p_3): Likewise.
11176 * cgraph.h (cgraph_node): Add ifunc_resolver.
11177 (cgraph_node::only_called_directly_or_aliased_p): Return false
11178 for IFUNC resolver.
11179 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
11181 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
11182 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
11183 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
11184 instead of looking up ifunc attribute.
11186 2018-05-22 Luis Machado <luis.machado@linaro.org>
11188 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
11190 2018-05-22 Martin Sebor <msebor@redhat.com>
11192 PR middle-end/85359
11193 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
11194 only when expasion succeeds.
11195 (expand_builtin_strcmp): Same.
11196 (expand_builtin_strncmp): Same.
11198 2018-05-22 Martin Sebor <msebor@redhat.com>
11200 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
11202 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
11203 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11205 * config/aarch64/aarch64-ldpstp.md: Replace uses of
11206 aarch64_mem_pair_operand with memory_operand and delete operand swapping
11208 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
11209 Add check for legitimate_address.
11210 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
11211 (aarch64_swap_ldrstr_operands): New.
11212 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
11215 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
11216 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11218 * config/aarch64/aarch64.md: New patterns to generate stp
11220 (store_pair_sw, store_pair_dw): New patterns to generate stp for
11221 single words and double words.
11222 (load_pair_sw, load_pair_dw): Likewise.
11223 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
11225 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
11227 * config/aarch64/aarch64-ldpstp.md: Modify peephole
11228 for different mode ldpstp and add peephole for merged zero stores.
11229 Likewise for loads.
11230 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
11232 (aarch64_gen_store_pair): Rename calls to match new patterns.
11233 (aarch64_gen_load_pair): Rename calls to match new patterns.
11234 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
11235 (load_pair<DREG:mode><DREG2:mode>): ... This.
11236 (store_pair<mode>): Rename to...
11237 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
11238 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
11239 New mode iterators.
11240 (V_INT_EQUIV): Handle SImode.
11241 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
11244 2018-05-22 Martin Sebor <msebor@redhat.com>
11247 * calls.c (maybe_warn_nonstring_arg): Use string length to set
11248 or ajust the presumed bound on an operation to avoid unnecessary
11251 2018-05-22 Martin Sebor <msebor@redhat.com>
11253 PR tree-optimization/85826
11254 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
11255 assuming that a DECL necesarily has a constant size.
11257 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
11259 PR middle-end/85862
11260 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
11262 2018-05-22 Richard Biener <rguenther@suse.de>
11264 PR tree-optimization/85834
11265 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
11266 non-constant and non-zero memset arguments.
11268 2018-05-22 Martin Liska <mliska@suse.cz>
11271 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
11273 2018-05-22 Richard Biener <rguenther@suse.de>
11275 PR tree-optimization/85863
11276 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
11277 comparisons when vectype is specified.
11278 (vectorizable_condition): Do not specify vectype for
11279 vect_is_simple_cond when SLP vectorizing.
11281 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
11284 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
11285 define __ibm128 as long double.
11286 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
11287 as a distinct type when IEEE 128-bit support is enabled.
11288 (init_float128_ieee): Fix up conversions between IFmode and IEEE
11289 128-bit types to use the correct functions.
11290 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
11291 convert between 128-bit floating point types that have different
11292 modes but the same representation, instead of using gen_lowpart to
11294 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
11296 (IFKF_reg): New attributes to give the register constraints for
11298 (extend<mode>tf2_internal): New insns to mark an explicit
11299 conversion between 128-bit floating point types that have a
11300 different mode but share the same representation.
11302 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
11304 PR tree-optimization/85814
11305 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
11306 a null return from get_strinfo when unsharing the next
11307 strinfo in the chain.
11309 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
11312 * varasm.c (weak_finish): Clean up weak_decls.
11314 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11316 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
11317 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
11318 UNSPEC_UADALP values.
11319 * config/aarch64/iterators.md (ABAL): New int iterator.
11322 (sur): Add mappings for the above.
11323 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
11325 (aarch64_<sur>abal<mode>_4): Likewise.
11326 (aarch64_<sur>adalp<mode>_3): Likewise.
11327 (<sur>sadv16qi): New define_expand.
11329 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
11331 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
11332 (*movdf_internal): Ditto.
11333 (*rcpsf2_sse): Ditto.
11334 (*rsqrtsf2_sse): Ditto.
11335 (*sqrt<mode>2_sse): Ditto.
11337 2018-05-21 Tamar Christina <tamar.christina@arm.com>
11339 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
11341 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
11342 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
11343 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
11344 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
11346 * config/aarch64/arm_neon.h: Likewise.
11347 * config/aarch64/iterators.md (VQ_I): New.
11349 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
11351 * config.gcc: Add arc/t-multilib-linux to tmake_file for
11353 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
11356 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
11358 * config/nds32/constraints.md (S): New constraint.
11359 * config/nds32/nds32.md (call_internal): Use constraint S.
11360 (call_value_internal): Likewise.
11361 (sibcall_internal): Likewise.
11362 (sibcall_value_internal): Likewise.
11364 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
11365 Chung-Ju Wu <jasonwucj@gmail.com>
11367 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
11368 into consideration.
11370 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
11371 Chung-Ju Wu <jasonwucj@gmail.com>
11373 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
11374 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
11375 (nds32_rtx_costs_impl): Simplify.
11376 (nds32_address_cost_impl): Simplify.
11377 (nds32_init_rtx_costs): New function.
11378 (nds32_rtx_costs_speed_prefer): Likewise.
11379 (nds32_rtx_costs_size_prefer): Likewise.
11380 (nds32_address_cost_speed_prefer): Likewise.
11381 (nds32_address_cost_speed_fwprop): Likewise.
11382 (nds32_address_cost_size_prefer): Likewise.
11383 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
11384 * config/nds32/nds32.c (nds32_option_override): Use
11385 nds32_init_rtx_costs function.
11387 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
11389 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
11390 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
11391 (TARGET_PIPELINE_N8): Likewise.
11392 (TARGET_PIPELINE_N10): Likewise.
11393 (TARGET_PIPELINE_N13): Likewise.
11394 (TARGET_PIPELINE_GRAYWOLF): Likewise.
11396 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
11398 * config/nds32/nds32-fpu.md: Update copyright year.
11400 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
11402 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
11404 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
11406 * config/nds32/nds32.c
11407 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
11408 * config/nds32/nds32.opt (minline-asm-r15): New option.
11410 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
11412 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
11414 * config/nds32/nds32.md (abssi2): New pattern.
11416 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
11418 * config/i386/i386.md (rex64namesuffix): New mode attribute.
11419 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
11420 Merge insn pattern from sse_cvtsi2ss<round_name> and
11421 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
11422 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
11423 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
11424 using SWI48 mode iterator.
11425 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
11426 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
11427 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
11428 pattern from sse_cvttss2si<round_saeonly_name>
11429 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
11430 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
11431 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
11432 using SWI48 mode iterator.
11433 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
11434 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
11435 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
11436 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
11437 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
11438 using SWI48 mode iterator.
11439 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
11440 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
11441 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
11442 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
11443 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
11444 SWI48 mode iterator.
11445 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
11446 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
11447 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
11448 pattern from sse_cvttsd2si<round_saeonly_name>
11449 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
11451 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
11453 * config/nds32/nds32-md-auxiliary.c
11454 (nds32_valid_smw_lwm_base_p): Refine.
11455 (nds32_output_smw_single_word): Refine.
11456 (nds32_output_smw_double_word): New.
11457 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
11459 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
11461 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
11462 (nds32_output_stack_pop): Refine.
11463 (nds32_expand_unaligned_load): Refine.
11464 (nds32_expand_unaligned_store): Refine.
11466 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
11467 Chung-Ju Wu <jasonwucj@gmail.com>
11469 * config/nds32/constants.md: Add TP_REGNUM constant.
11470 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
11471 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
11473 * config/nds32/nds32-doubleword.md: Consider flag_pic.
11474 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
11475 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
11476 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
11477 and PIC code generation.
11478 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
11480 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
11482 * config/nds32/nds32.md: Support TLS and PIC.
11483 * config/nds32/nds32.c: Support TLS and PIC.
11484 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
11485 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
11488 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
11490 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
11491 mode with E_ prefix.
11493 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
11494 Chung-Ju Wu <jasonwucj@gmail.com>
11496 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
11497 * config/nds32/nds32-md-auxiliary.c
11498 (symbolic_reference_mentioned_p): New.
11499 (nds32_legitimize_ict_address): New.
11500 (nds32_expand_ict_move): New.
11501 (nds32_indirect_call_referenced_p): New.
11502 (nds32_symbol_binds_local_p): Delete.
11503 (nds32_long_call_p): Modify.
11504 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
11505 * config/nds32/nds32-protos.h
11506 (symbolic_reference_mentioned_p): Declare.
11507 (nds32_legitimize_ict_address): Declare.
11508 (nds32_expand_ict_move): Declare.
11509 (nds32_indirect_call_referenced_p): Declare.
11510 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
11511 (nds32_relax_group): Use nds32_ict_const_p as condition.
11512 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
11513 (nds32_asm_file_start): Output ict_model directive in asm code.
11514 (nds32_legitimate_address_p): Consider indirect call.
11515 (nds32_print_operand): Consider indirect call.
11516 (nds32_print_operand_address): Consider indirect call.
11517 (nds32_insert_attributes): Handle "indirect_call" attribute.
11518 (TARGET_LEGITIMATE_ADDRESS_P): Define.
11519 (TARGET_LEGITIMATE_CONSTANT_P): Define.
11520 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
11521 (TARGET_DELEGITIMIZE_ADDRESS): Define.
11522 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
11523 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
11524 (TARGET_ICT_MODEL_SMALL): Define.
11525 (TARGET_ICT_MODEL_LARGE): Define.
11526 * config/nds32/nds32.md (movsi): Consider ict model.
11527 (call, call_value): Consider ict model.
11528 (sibcall, sibcall_value): Consider ict model.
11529 * config/nds32/nds32.opt (mict-model): New option.
11530 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
11533 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
11534 Monk Chiang <sh.chiang04@gmail.com>
11535 Jim Wilson <jimw@sifive.com>
11537 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
11538 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
11539 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
11540 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
11541 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
11542 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
11543 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
11544 compute save_libcall_adjustment properly.
11545 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
11546 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
11547 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
11548 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
11549 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
11550 (ABI_SPEC): Handle mabi=ilp32e.
11551 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
11552 (RVE): Add RVE mask.
11553 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
11554 <-march>: Add rv32e as an example.
11556 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
11559 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
11560 (intra_create_variable_infos): Handle C++ constructors.
11562 2018-05-18 Martin Liska <mliska@suse.cz>
11564 * passes.def: Remove a redundant pass.
11566 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
11569 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
11571 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11573 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
11575 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
11577 (armv2, armv2a, armv3, armv3m): Delete architectures.
11578 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
11579 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
11580 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
11582 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
11583 (*mulsidi3adddi): Likewise.
11584 (mulsidi3): Likewise.
11585 (*mulsidi3_nov6): Likewise.
11586 (umulsidi3): Likewise.
11587 (umulsidi3_nov6): Likewise.
11588 (umaddsidi4): Likewise.
11589 (*umulsidi3adddi): Likewise.
11590 (smulsi3_highpart): Likewise.
11591 (*smulsi3_highpart_nov6): Likewise.
11592 (umulsi3_highpart): Likewise.
11593 (*umulsi3_highpart_nov6): Likewise.
11594 * config/arm/arm.h (arm_arch3m): Delete.
11595 * config/arm/arm.c (arm_arch3m): Delete.
11596 (arm_option_override_internal): Update armv3-related comment.
11597 (arm_configure_build_target): Delete use of isa_bit_mode32.
11598 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
11599 (arm_rtx_costs_internal): Delete check of arm_arch3m.
11600 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
11601 (mulsa3): Likewise.
11602 (mulusa3): Likewise.
11603 * config/arm/arm-protos.h (arm_arch3m): Delete.
11604 * config/arm/arm-tables.opt: Regenerate.
11605 * config/arm/arm-tune.md: Likewise.
11606 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
11607 deleted architectures.
11609 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11611 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
11612 (armv5t, armv5te): New features.
11613 (ARMv5, ARMv5e): Delete fgroups.
11614 (ARMv5t, ARMv5te): Adjust for above changes.
11615 (ARMv6m): Likewise.
11616 (armv5, armv5e): Delete arches.
11617 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
11619 (*call_reg_arm): Likewise.
11620 (*call_value_reg_armv5): Likewise.
11621 (*call_value_reg_arm): Likewise.
11622 (*call_symbol): Likewise.
11623 (*call_value_symbol): Likewise.
11624 (*sibcall_insn): Likewise.
11625 (*sibcall_value_insn): Likewise.
11626 (clzsi2): Likewise.
11627 (prefetch): Likewise.
11628 (define_split and define_peephole2 dependent on arm_arch5):
11630 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
11632 (TARGET_ARM_QBIT): Likewise.
11633 (TARGET_DSP_MULTIPLY): Likewise.
11634 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
11635 (arm_arch5, arm_arch5e): Delete.
11636 (arm_arch5t, arm_arch5te): Declare.
11637 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
11638 (arm_arch5t): Declare.
11639 (arm_option_reconfigure_globals): Update for the above.
11640 (arm_options_perform_arch_sanity_checks): Update comment, replace
11641 use of arm_arch5 with arm_arch5t.
11642 (use_return_insn): Likewise.
11643 (arm_emit_call_insn): Likewise.
11644 (output_return_instruction): Likewise.
11645 (arm_final_prescan_insn): Likewise.
11646 (arm_coproc_builtin_available): Likewise.
11647 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
11648 arm_arch5e with arm_arch5t and arm_arch5te.
11649 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
11650 (arm_arch5t, arm_arch5te): Declare.
11651 * config/arm/arm-tables.opt: Regenerate.
11652 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
11653 * config/arm/t-multilib: Likewise.
11654 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
11655 instead of arm_arch5.
11656 (*call_reg_thumb1): Likewise.
11657 (*call_value_reg_thumb1_v5): Likewise.
11658 (*call_value_reg_thumb1): Likewise.
11659 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
11661 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
11663 2018-05-18 Martin Liska <mliska@suse.cz>
11665 PR gcov-profile/84846
11666 * doc/gcov.texi: Document -t option of gcov tool.
11668 2018-05-18 Martin Liska <mliska@suse.cz>
11670 PR gcov-profile/84846
11671 * gcov.c (print_usage): Add new -t option.
11672 (process_args): Handle the option.
11673 (generate_results): Use stdout as output when requested by
11676 2018-05-18 Martin Liska <mliska@suse.cz>
11678 PR gcov-profile/84846
11679 * coverage.c (coverage_init): Write PWD to .gcno file.
11680 * doc/gcov.texi: Document how working directory is printed.
11681 * gcov-dump.c (dump_gcov_file): Print PWD.
11682 * gcov.c (output_intermediate_file): Likewise.
11683 (read_graph_file): Read PWD string.
11684 (output_lines): Print PWD.
11686 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11688 PR middle-end/85817
11689 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
11690 for retval and return false if all args to phi are zero.
11692 2018-05-18 Richard Biener <rguenther@suse.de>
11694 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
11696 (evrp_dom_walker::before_dom_children): Call it.
11698 2018-05-18 Richard Biener <rguenther@suse.de>
11700 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
11701 results when processing array refs with variable index.
11703 2018-05-18 Toon Moene <toon@moene.org>
11705 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
11706 directly after that of -floop-interchange. Indicate that both
11707 options are enabled by default when specifying -O3.
11709 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11711 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
11712 iterator. Delete separate integer-mode vec_set<mode> expander.
11713 (aarch64_simd_vec_setv2di): Delete.
11714 (vec_setv2di): Delete.
11715 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
11716 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
11717 the "w, r" alternative.
11719 2018-05-18 Martin Liska <mliska@suse.cz>
11721 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
11722 * tree-pass.h (make_pass_lower_switch_O0): New function.
11723 * tree-switch-conversion.c (node_has_low_bound): Remove.
11724 (node_has_high_bound): Likewise.
11725 (node_is_bounded): Likewise.
11726 (class pass_lower_switch): Make it a template type and create
11728 (pass_lower_switch::execute): Add template argument.
11729 (make_pass_lower_switch): New function.
11730 (make_pass_lower_switch_O0): New function.
11731 (do_jump_if_equal): Remove.
11732 (emit_case_nodes): Simplify to just handle all 3 cases and leave
11733 all the hard work to tree optimization passes.
11735 2018-05-18 Martin Liska <mliska@suse.cz>
11737 * dbgcnt.c (limit_low): Renamed from limit.
11738 (limit_high): New variable.
11739 (dbg_cnt_is_enabled): Check for upper limit.
11740 (dbg_cnt): Adjust dumping.
11741 (dbg_cnt_set_limit_by_index): Add new argument for high
11743 (dbg_cnt_set_limit_by_name): Likewise.
11744 (dbg_cnt_process_single_pair): Parse new format.
11745 (dbg_cnt_process_opt): Use strtok.
11746 (dbg_cnt_list_all_counters): Remove 'value' and add
11748 * doc/invoke.texi: Document changes.
11750 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
11752 * doc/sourcebuild.texi (scalar_all_fma): Document.
11753 * tree.def (FMA_EXPR): Delete.
11754 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
11755 * internal-fn.c (ternary_direct): New macro.
11756 (expand_ternary_optab_fn): Likewise.
11757 (direct_ternary_optab_supported_p): Likewise.
11758 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
11759 * builtins.c (fold_builtin_fma): Delete.
11760 (fold_builtin_3): Don't call it.
11761 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
11762 * expr.c (expand_expr_real_2): Likewise.
11763 * fold-const.c (operand_equal_p): Likewise.
11764 (fold_ternary_loc): Likewise.
11765 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
11766 * gimple.c (DEFTREECODE): Likewise.
11767 * gimplify.c (gimplify_expr): Likewise.
11768 * optabs-tree.c (optab_for_tree_code): Likewise.
11769 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
11770 * tree-eh.c (operation_could_trap_p): Likewise.
11771 (stmt_could_throw_1_p): Likewise.
11772 * tree-inline.c (estimate_operator_cost): Likewise.
11773 * tree-pretty-print.c (dump_generic_node): Likewise.
11774 (op_code_prio): Likewise.
11775 * tree-ssa-loop-im.c (stmt_cost): Likewise.
11776 * tree-ssa-operands.c (get_expr_operands): Likewise.
11777 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
11778 * fold-const-call.h (fold_fma): Delete.
11779 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
11780 CFN_FNMA and CFN_FNMS.
11781 (fold_fma): Delete.
11782 * genmatch.c (combined_fn): New enum.
11783 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
11784 (commutative_op): New function.
11785 (commutate): Use it. Handle more than 2 operands.
11786 (dt_operand::gen_gimple_expr): Use commutative_op.
11787 (parser::parse_expr): Allow :c to be used with non-binary
11788 operators if the commutative operand is known.
11789 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
11790 CFN_FMS, CFN_FNMA and CFN_FNMS.
11791 (backprop::process_assign_use): Remove FMA_EXPR handling.
11792 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
11793 (gen_hsa_fma): New function.
11794 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
11795 IFN_FNMA and IFN_FNMS.
11796 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
11797 * gimple-fold.h (follow_all_ssa_edges): Declare.
11798 * gimple-fold.c (follow_all_ssa_edges): New function.
11799 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
11800 gimple_build interface and use follow_all_ssa_edges to fold the result.
11801 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
11802 instead of checking for optabs directly.
11803 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
11804 rather than FMA_EXPRs.
11805 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
11806 call to IFN_FMA instead of an FMA_EXPR.
11808 2018-05-17 Jim Wilson <jimw@sifive.com>
11810 * expr.c (do_tablejump): When converting index to Pmode, if we have a
11811 sign extended promoted subreg, and the range does not have the sign bit
11812 set, then do a sign extend.
11814 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
11815 test, check for sign extended subreg and/or constant operands, and
11816 do a sign extend in that case.
11818 2018-05-17 Steve Ellcey <sellcey@cavium.com>
11820 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
11821 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
11823 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
11824 Change logics_shift_reg to logics_shift_imm.
11825 (thunderx2t99_fp_loadpair_basic): Delete.
11826 (thunderx2t99_fp_storepair_basic): Delete.
11827 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
11828 (thunderx2t99_asimd_polynomial): Delete.
11829 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
11830 and neon_fp_mul_d_scalar_q.
11831 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
11832 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
11833 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
11834 (thunderx2t99_asimd_lut): Add missing tbl types.
11835 (thunderx2t99_asimd_ext): Delete.
11836 (thunderx2t99_asimd_load1_1_mult): Delete.
11837 (thunderx2t99_asimd_load1_2_mult): Delete.
11838 (thunderx2t99_asimd_load1_ldp): New.
11839 (thunderx2t99_asimd_load1): New.
11840 (thunderx2t99_asimd_load2): Add missing *load2* types.
11841 (thunderx2t99_asimd_load3): New.
11842 (thunderx2t99_asimd_load4): New.
11843 (thunderx2t99_asimd_store1_1_mult): Delete.
11844 (thunderx2t99_asimd_store1_2_mult): Delete.
11845 (thunderx2t99_asimd_store2_mult): Delete.
11846 (thunderx2t99_asimd_store2_onelane): Delete.
11847 (thunderx2t99_asimd_store_stp): New.
11848 (thunderx2t99_asimd_store1): New.
11849 (thunderx2t99_asimd_store2): New.
11850 (thunderx2t99_asimd_store3): New.
11851 (thunderx2t99_asimd_store4): New.
11853 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
11855 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
11856 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
11858 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
11859 Segher Boessenkool <segher@kernel.crashing.org>
11862 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
11865 2018-05-17 Richard Biener <rguenther@suse.de>
11867 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
11868 for pruning loop and prune defs feeding only already visited PHIs.
11870 2018-05-17 Richard Biener <rguenther@suse.de>
11872 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
11874 2018-05-17 Bin Cheng <bin.cheng@arm.com>
11875 Richard Biener <rguenther@suse.de>
11877 PR tree-optimization/85793
11878 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
11879 for VMAT_ELEMENTWISE.
11881 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
11883 * internal-fn.h (lookup_internal_fn): Declare
11884 * internal-fn.c (lookup_internal_fn): New function.
11885 * gimple.c (gimple_build_call_from_tree): Handle calls to
11886 internal functions.
11887 * gimple-pretty-print.c (dump_gimple_call): Print "." before
11888 internal function names.
11889 * tree-pretty-print.c (dump_generic_node): Likewise.
11890 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
11892 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
11894 * gimple-fold.h (gimple_build): Make the function forms take
11895 combined_fn rather than built_in_function.
11896 (gimple_simplify): Likewise.
11897 * gimple-match-head.c (gimple_simplify): Likewise.
11898 * gimple-fold.c (gimple_build): Likewise.
11899 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
11900 rather than gimple_build_call_internal.
11901 (get_initial_defs_for_reduction): Likewise.
11902 (vect_create_epilog_for_reduction): Likewise.
11903 (vectorizable_live_operation): Likewise.
11905 2018-05-17 Martin Liska <mliska@suse.cz>
11907 * gimple-ssa-sprintf.c (format_directive): Do not use
11908 space in between 'G_' and '('.
11910 2018-05-17 Jakub Jelinek <jakub@redhat.com>
11913 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
11914 even if the mask is not all ones.
11917 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
11919 (ix86_gimple_fold_builtin): Likewise.
11922 * config/i386/i386.c: Include tree-vector-builder.h.
11923 (ix86_vector_shift_count): New function.
11924 (ix86_fold_builtin): Fold shift builtins by scalar count.
11925 (ix86_gimple_fold_builtin): Likewise.
11927 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
11928 _mm512_setzero): New intrinsics.
11930 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
11931 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11933 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
11934 code generation for cases where splatting a value is not useful.
11935 * simplify-rtx.c (simplify_ternary_operation): Simplify
11936 vec_merge across a vec_duplicate and a paradoxical subreg forming
11937 a vector mode to a vec_concat.
11939 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
11941 * config.gcc: Support "goldmont-plus".
11942 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
11944 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
11945 PROCESSOR_GOLDMONT_PLUS.
11946 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
11947 (processor_target_table): Add "goldmont-plus".
11948 (PTA_GOLDMONT_PLUS): Define.
11949 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
11950 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
11951 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
11952 (fold_builtin_cpu): Add "goldmont-plus".
11953 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
11954 (ix86_option_override_internal): Add "goldmont-plus".
11955 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
11956 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
11957 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
11958 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
11960 2018-05-17 Richard Biener <rguenther@suse.de>
11962 PR tree-optimization/85757
11963 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
11964 remove defs that only feed that PHI from further processing.
11966 2018-05-16 Jim Wilson <jimw@sifive.com>
11968 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
11970 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
11972 2018-05-16 Mark Wielaard <mark@klomp.org>
11974 * dwarf2out.c (count_index_strings): New function.
11975 (output_indirect_strings): Call count_index_strings and generate
11976 header for dwarf_version >= 5.
11978 2018-05-16 Mark Wielaard <mark@klomp.org>
11980 * dwarf2out.c (dwarf_FORM): New function.
11981 (set_indirect_string): Use dwarf_FORM.
11982 (reset_indirect_string): Likewise.
11983 (size_of_die): Likewise.
11984 (value_format): Likewise.
11985 (output_die): Likewise.
11986 (add_skeleton_AT_string): Likewise.
11987 (output_macinfo_op): Likewise.
11988 (index_string): Likewise.
11989 (output_index_string_offset): Likewise.
11990 (output_index_string): Likewise.
11991 (count_index_strings): Likewise.
11993 2018-05-16 Carl Love <cel@us.ibm.com>
11995 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
11996 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
11998 2018-05-16 Martin Jambor <mjambor@suse.cz>
12000 * ipa-prop.c (ipa_free_all_edge_args): Remove.
12001 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
12003 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
12005 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
12006 (fnma<mode>4): Likewise.
12007 (fms<mode>4): Likewise.
12008 (fnms<mode>4): Likewise.
12009 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
12010 (aarch64_fnma<mode>4): Likewise.
12011 (aarch64_fms<mode>4): Likewise.
12012 (aarch64_fnms<mode>4): Likewise.
12013 (aarch64_fnmadd<mode>4): Likewise.
12015 2018-05-16 Jason Merrill <jason@redhat.com>
12017 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
12019 2018-05-16 Richard Biener <rguenther@suse.de>
12021 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
12022 (dump_stmt_cost): Declare.
12023 (add_stmt_cost): Dump cost we add.
12024 (add_stmt_costs): New function.
12025 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
12026 No longer exported.
12027 (vect_analyze_stmt): Adjust prototype.
12028 (vectorizable_condition): Likewise.
12029 (vectorizable_live_operation): Likewise.
12030 (vectorizable_reduction): Likewise.
12031 (vectorizable_induction): Likewise.
12032 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
12033 cost vector to pass to vectorizable_ and record afterwards.
12034 (vect_model_reduction_cost): Take cost vector argument and adjust.
12035 (vect_model_induction_cost): Likewise.
12036 (vectorizable_reduction): Likewise.
12037 (vectorizable_induction): Likewise.
12038 (vectorizable_live_operation): Likewise.
12039 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
12040 SLP_TREE_NUMBER_OF_VEC_STMTS.
12041 (vect_analyze_slp_cost_1): Remove.
12042 (vect_analyze_slp_cost): Likewise.
12043 (vect_slp_analyze_node_operations): Take visited args and
12044 a target cost vector. Avoid processing already visited stmt sets.
12045 (vect_slp_analyze_operations): Use a local cost vector to gather
12046 costs and register those of non-discarded instances.
12047 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
12048 (vect_schedule_slp_instance): Remove copying of
12049 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
12051 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
12052 adding cost. Record cost entry location.
12053 (vect_prologue_cost_for_slp_op): Function to compute cost of
12054 a constant or invariant generated for SLP vect in the prologue,
12055 split out from vect_analyze_slp_cost_1.
12056 (vect_model_simple_cost): Make static. Adjust for SLP costing.
12057 (vect_model_promotion_demotion_cost): Likewise.
12058 (vect_model_store_cost): Likewise, make static.
12059 (vect_model_load_cost): Likewise.
12060 (vectorizable_bswap): Add cost vector arg and adjust.
12061 (vectorizable_call): Likewise.
12062 (vectorizable_simd_clone_call): Likewise.
12063 (vectorizable_conversion): Likewise.
12064 (vectorizable_assignment): Likewise.
12065 (vectorizable_shift): Likewise.
12066 (vectorizable_operation): Likewise.
12067 (vectorizable_store): Likewise.
12068 (vectorizable_load): Likewise.
12069 (vectorizable_condition): Likewise.
12070 (vectorizable_comparison): Likewise.
12071 (can_vectorize_live_stmts): Likewise.
12072 (vect_analyze_stmt): Likewise.
12073 (vect_transform_stmt): Adjust calls to vectorizable_*.
12074 * tree-vectorizer.c: Include gimple-pretty-print.h.
12075 (dump_stmt_cost): New function.
12077 2018-05-16 Richard Biener <rguenther@suse.de>
12079 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
12080 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
12081 * tree-ssa-dse.c: Include tree-ssa-loop.h.
12082 (check_name): New callback.
12083 (dse_classify_store): Track cycles via a visited bitmap of PHI
12084 defs and simplify handling of in-loop and across loop dead stores
12085 and properly fail for loop-variant refs. Handle byte-tracking with
12086 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
12089 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
12091 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
12092 (vect_get_mask_type_for_stmt): Likewise.
12093 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
12095 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
12096 to determine the statement's vector type and the vector type that
12097 should be used for calculating nunits. Deal with cases in which
12098 the type has to be deferred.
12099 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
12100 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
12101 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
12102 (vect_determine_vf_for_stmt): New functions, split out from...
12103 (vect_determine_vectorization_factor): ...here.
12104 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
12105 (vect_get_mask_type_for_stmt): New functions, split out from
12106 vect_determine_vectorization_factor.
12108 2018-05-16 Richard Biener <rguenther@suse.de>
12110 * tree-cfg.c (verify_gimple_assign_ternary): Properly
12111 verify the [VEC_]COND_EXPR embedded comparison.
12113 2018-05-15 Martin Sebor <msebor@redhat.com>
12115 PR tree-optimization/85753
12116 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
12117 RECORD_TYPE in addition to ARRAY_TYPE.
12119 2018-05-15 Martin Sebor <msebor@redhat.com>
12121 PR middle-end/85643
12122 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
12124 2018-05-15 Richard Biener <rguenther@suse.de>
12126 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
12127 add by_clobber_p one. Change algorithm to collect all defs
12128 representing uses we need to walk and try reducing them to
12129 a single one before failing.
12130 (dse_dom_walker::dse_optimize_stmt): Adjust.
12132 2018-05-13 Mark Wielaard <mark@klomp.org>
12134 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
12135 (size_of_loc_descr): Likewise.
12136 (output_loc_operands): Likewise.
12137 (output_loc_operands_raw): Likewise.
12138 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
12139 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
12140 (hash_loc_operands): Likewise.
12141 (compare_loc_operands): Likewise.
12143 2018-05-14 Mark Wielaard <mark@klomp.org>
12145 * dwarf2out.c (count_index_addrs): New function.
12146 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
12148 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12150 PR tree-optimization/83648
12151 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
12152 return value as malloc candidate.
12154 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12157 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
12158 param as true in call to suggest_attribute.
12160 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
12162 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
12163 -mreadonly-in-sdata.
12165 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12167 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
12169 (aarch64_crypto_aesd_fused): Likewise.
12171 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
12173 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
12174 (movsi_aarch64): Likewise.
12175 (load_pairsi): Likewise.
12176 (load_pairdi): Likewise.
12177 (store_pairsi): Likewise.
12178 (store_pairdi): Likewise.
12179 (load_pairsf): Likewise.
12180 (load_pairdf): Likewise.
12181 (store_pairsf): Likewise.
12182 (store_pairdf): Likewise.
12183 (zero_extend): Likewise.
12184 (trunc): Swap alternatives.
12185 (fcvt_target): Add '?' to prefer w over r.
12187 2018-05-14 Jakub Jelinek <jakub@redhat.com>
12190 * config/i386/i386.md: Disallow non-commutative arithmetics in
12191 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
12192 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
12193 in the peephole2 before it.
12195 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
12197 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
12198 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
12199 (ix86_handle_option): Handle -mcldemote.
12200 * config.gcc: New header.
12201 * config/i386/cldemoteintrin.h: New file.
12202 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
12203 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
12205 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
12206 OPTION_MASK_ISA_CLDEMOTE.
12207 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
12208 (ix86_valid_target_attribute_inner_p): Ditto.
12209 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
12210 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
12211 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
12212 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
12213 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
12215 * config/i386/i386.opt: Add -mcldemote.
12216 * config/i386/x86intrin.h: New header.
12217 * doc/invoke.texi: Add -mcldemote.
12219 2018-05-14 Richard Biener <rguenther@suse.de>
12221 * doc/match-and-simplify.texi: Adjust :s documentation.
12223 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
12225 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
12226 intended memcpy size.
12227 (REORDER_45): Likewise.
12229 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
12231 * sort.cc: New file.
12232 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
12233 * vec.c (qsort_chk): Use gcc_qsort.
12234 * Makefile.in (OBJS-libcommon): Add sort.o.
12235 (build/sort.o): New target. Use it...
12236 (BUILD_RTL): ... here, and...
12237 (build/gencfn-macros): ... here, and...
12238 (build/genmatch): ... here.
12240 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
12241 Chung-Ju Wu <jasonwucj@gmail.com>
12243 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
12244 * config/nds32/nds32-graywolf.md: New file.
12245 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
12246 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
12248 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
12249 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
12250 * config/nds32/nds32.md (pipeline_model): Add graywolf.
12251 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
12252 * config/nds32/pipelines.md: Include n15 settings.
12254 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
12255 Chung-Ju Wu <jasonwucj@gmail.com>
12257 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
12258 * config/nds32/nds32-n13.md: New file.
12259 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
12260 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
12262 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
12263 * config/nds32/nds32.md (pipeline_model): Add n13.
12264 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
12265 * config/nds32/pipelines.md: Include n13 settings.
12267 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
12268 Chung-Ju Wu <jasonwucj@gmail.com>
12270 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
12271 * config/nds32/nds32-n10.md: New file.
12272 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
12273 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
12275 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
12276 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
12277 * config/nds32/nds32.md (pipeline_model): Add n10.
12278 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
12279 * config/nds32/pipelines.md: Include n10 settings.
12281 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
12282 Kito Cheng <kito.cheng@gmail.com>
12283 Chung-Ju Wu <jasonwucj@gmail.com>
12285 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
12286 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
12287 Add enum values for DSP extension instructions.
12288 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
12290 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
12291 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
12292 New code iterators.
12293 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
12294 * config/nds32/nds32-dspext.md: New file for DSP implementation.
12295 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
12296 * config/nds32/nds32-intrinsic.md: Likewise.
12297 * config/nds32/nds32_intrinsic.h: Likewise.
12298 * config/nds32/nds32-md-auxiliary.c: Likewise.
12299 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
12300 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
12301 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
12302 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
12303 * config/nds32/nds32-protos.h: New declarations for DSP extension.
12304 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
12305 TYPE_DMAC in switch statement.
12306 * config/nds32/nds32.c: New checking and implementation for DSP
12307 extension instructions.
12308 * config/nds32/nds32.h: Likewise.
12309 * config/nds32/nds32.md: Likewise.
12310 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
12311 * config/nds32/predicates.md: Implement new predicates for DSP
12314 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
12316 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
12317 Reformat alternatives and attributes so it is easier to identify
12318 which constraints/attributes go with which instruction.
12319 (mov<mode>_hardfloat32, FMOVE64): Likewise.
12320 (mov<mode>_softfloat32, FMOVE64): Likewise.
12321 (mov<mode>_hardfloat64, FMOVE64): Likewise.
12322 (mov<mode>_softfloat64, FMOVE64): Likewise.
12324 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
12326 * doc/extend.texi (PowerPC Built-in Functions): Rename this
12328 (Basic PowerPC Built-in Functions): The new name of the
12329 subsection previously known as "PowerPC Built-in Functions".
12330 (Basic PowerPC Built-in Functions Available on all Configurations):
12332 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
12333 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
12334 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
12335 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
12337 2018-05-11 Martin Jambor <mjambor@suse.cz>
12340 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
12343 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
12346 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
12348 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
12350 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
12351 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
12352 (ix86_handle_option): Handle -mwaitpkg.
12353 * config.gcc: New header.
12354 * config/i386/cpuid.h (bit_WAITPKG): New bit.
12355 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
12356 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
12358 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
12359 OPTION_MASK_ISA_WAITPKG.
12360 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
12361 (ix86_option_override_internal): Add PTA_WAITPKG.
12362 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
12363 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
12364 IX86_BUILTIN_TPAUSE.
12365 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
12366 __builtin_ia32_umwait and __builtin_ia32_tpause.
12367 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
12368 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
12369 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
12370 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
12371 UNSPECV_TPAUSE): New.
12372 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
12373 * config/i386/i386.opt: Add -mwaitpkg.
12374 * config/i386/waitpkgintrin.h: New file.
12375 * config/i386/x86intrin.h: New header.
12376 * doc/invoke.texi: Add -mwaitpkg.
12378 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
12381 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
12383 (cortex-m0): Use armv6s-m isa.
12384 (cortex-m0plus): Likewise.
12385 (cortex-m1): Likewise.
12386 (cortex-m0.small-multiply): Likewise.
12387 (cortex-m0plus.small-multiply): Likewise.
12388 (cortex-m1.small-multiply): Likewise.
12390 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
12391 Jakub Jelinek <jakub@redhat.com>
12393 PR tree-optimization/85692
12394 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
12395 source permute as well.
12397 2018-05-11 Martin Liska <mliska@suse.cz>
12400 * doc/extend.texi: Document LLVM style format for no_sanitize
12403 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
12405 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
12406 mode_supports_vsx_dform_quad to mode_supports_dq_form.
12407 (mode_supports_vsx_dform_quad): Likewise.
12408 (mode_supports_vmx_dform): Move these functions to be next to the
12409 other mode_supports functions.
12410 (mode_supports_dq_form): Likewise.
12411 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
12412 mode_supports_dq_form.
12413 (reg_offset_addressing_ok_p): Likewise.
12414 (offsettable_ok_by_alignment): Likewise.
12415 (rs6000_legitimate_offset_address_p): Likewise.
12416 (legitimate_lo_sum_address_p): Likewise.
12417 (rs6000_legitimize_address): Likewise.
12418 (rs6000_legitimize_reload_address): Likewise.
12419 (rs6000_secondary_reload_inner): Likewise.
12420 (rs6000_preferred_reload_class): Likewise.
12421 (rs6000_output_move_128bit): Likewise.
12423 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
12425 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
12426 Generate SImode target register for null target.
12427 <case IX86_BUILTIN_XGETBV>: Ditto.
12428 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
12429 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
12431 2018-05-10 Carl Love <cel@us.ibm.com>
12433 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
12434 dcbtt and dcbtstt if operands[2] is 0.
12436 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
12439 * config/i386/sse.md (usadv64qi): New expander.
12441 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
12443 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
12444 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
12445 -maltivec=be support.
12446 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
12447 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
12448 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
12449 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
12450 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
12451 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
12452 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
12453 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
12454 altivec_vsumsws): Adjust.
12455 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
12456 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
12457 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
12459 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
12460 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
12461 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
12462 (altivec_lve<VI_char>x): Delete expand.
12463 (*altivec_lve<VI_char>x_internal): Rename to...
12464 (altivec_lve<VI_char>x): ... this.
12465 (altivec_lvxl_<mode>): Delete expand.
12466 (*altivec_lvxl_<mode>_internal): Rename to ...
12467 (altivec_lvxl_<mode>): ... this.
12468 (altivec_stvxl_<mode>): Delete expand.
12469 (*altivec_stvxl_<mode>_internal): Rename to ...
12470 (altivec_stvxl_<mode>): ... this.
12471 (altivec_stve<VI_char>x): Delete expand.
12472 (*altivec_stve<VI_char>x_internal): Rename to ...
12473 (altivec_stve<VI_char>x): ... this.
12474 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
12475 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
12476 reduc_plus_scal_<mode>): Adjust.
12477 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
12479 (rs6000_cpu_cpp_builtins): Adjust.
12480 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
12481 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
12482 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
12483 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
12484 -maltivec=be support.
12485 (rs6000_split_vec_extract_var): Adjust.
12486 (rs6000_split_v4si_init): Adjust.
12487 (swap_selector_for_mode): Delete.
12488 (altivec_expand_lvx_be, altivec_expand_stvx_be,
12489 altivec_expand_stvex_be): Delete.
12490 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
12491 -maltivec=be support.
12492 (rs6000_gimple_fold_builtin): Ditto.
12493 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
12495 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
12496 (TARGET_DIRECT_MOVE_64BIT): Adjust.
12497 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
12498 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
12499 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
12500 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
12501 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
12502 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
12503 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
12504 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
12505 anonymous split): Adjust.
12506 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
12507 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
12509 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
12511 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
12512 when --with-gxx-include-dir is also specified.
12513 * configure: Regenerate.
12515 2018-05-09 Jim Wilson <jimw@sifive.com>
12518 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
12519 * config/riscv/t-withmultilib: New.
12520 * config/riscv/withmultilib.h: New.
12521 * doc/install.texi: Document RISC-V --with-multilib-list support.
12523 2018-05-09 Richard Biener <rguenther@suse.de>
12525 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
12527 (vect_bb_vectorization_profitable_p): Adjust. Compute
12528 actual scalar cost using the cost vector and the add_stmt_cost
12531 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
12533 PR rtl-optimization/85645
12534 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
12535 in the REG_CFA_REGISTER note for LR, don't leave it empty.
12537 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
12539 PR rtl-optimization/85645
12540 * shrink-wrap.c (spread_components): Return a boolean saying if
12541 anything was changed.
12542 (try_shrink_wrapping_separate): Iterate spread_components until
12543 nothing changes anymore.
12545 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
12547 PR rtl-optimization/85645
12548 * regrename.c (build_def_use): Also kill the chains that include the
12549 destination of a REG_CFA_REGISTER note.
12551 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
12553 PR rtl-optimization/85645
12554 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
12555 insn that has a REG_CFA_REGISTER note.
12557 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
12559 * cfgexpand.c (expand_clobber): New function.
12560 (expand_gimple_stmt_1): Use it.
12561 * tree-vect-stmts.c (vect_clobber_variable): New function,
12563 (vectorizable_simd_clone_call): ...here.
12564 (vectorizable_store): Emit a clobber either side of an
12565 IFN_STORE_LANES sequence.
12566 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
12568 2018-05-09 Tom de Vries <tom@codesourcery.com>
12571 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
12572 (define_insn "trap_if_false"): Add exit after trap.
12574 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
12576 PR rtl-optimization/85638
12577 * bb-reorder.c: Include common/common-target.h.
12578 (create_forwarder_block): New function extracted from...
12579 (fix_up_crossing_landing_pad): ...here. Rename into...
12580 (dw2_fix_up_crossing_landing_pad): ...this.
12581 (sjlj_fix_up_crossing_landing_pad): New function.
12582 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
12583 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
12584 from both partitions and exit the loop after one iteration.
12586 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
12589 * doc/extend.texi (PowerPC Built-in Functions): Rename this
12591 (Basic PowerPC Built-in Functions): The new name of the
12592 subsection previously known as "PowerPC Built-in Functions".
12593 (Basic PowerPC Built-in Functions Available on all Configurations):
12595 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
12597 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
12599 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
12601 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
12604 2018-05-08 Jim Wilson <jimw@sifive.com>
12606 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
12607 (LD_EMUL_SUFFIX): New.
12608 (LINK_SPEC): Use it.
12610 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
12612 * doc/extend.texi (PowerPC Built-in Functions): Rename this
12614 (Basic PowerPC Built-in Functions): The new name of the
12615 subsection previously known as "PowerPC Built-in Functions".
12616 (Basic PowerPC Built-in Functions Available on all Configurations):
12618 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
12620 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
12622 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
12624 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
12627 2018-05-08 Jakub Jelinek <jakub@redhat.com>
12630 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
12631 after cmpelim optimization.
12633 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
12635 * config.gcc: Support "goldmont".
12636 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
12637 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
12638 PROCESSOR_GOLDMONT.
12639 * config/i386/i386.c (m_GOLDMONT): Define.
12640 (processor_target_table): Add "goldmont".
12641 (PTA_GOLDMONT): Define.
12642 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
12643 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
12644 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
12645 (fold_builtin_cpu): Add "goldmont".
12646 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
12647 (ix86_option_override_internal): Add "goldmont".
12648 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
12649 (processor_type): Add PROCESSOR_GOLDMONT.
12650 * config/i386/i386.md: Add CPU "glm".
12651 * config/i386/glm.md: New file.
12652 * config/i386/x86-tune.def: Add m_GOLDMONT.
12653 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
12655 2018-05-08 Jakub Jelinek <jakub@redhat.com>
12658 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
12660 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
12661 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
12662 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
12665 * config/i386/i386.c (ix86_fold_builtin): Handle
12666 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
12669 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
12670 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
12672 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
12675 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
12676 (check_arch): Likewise.
12677 (check_fpu): Return the result rather than printing it.
12678 (end arch): Fix operator precedence.
12679 (end cpu): Likewise.
12680 (END): Print the result from check_fpu.
12682 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
12683 Alan Hayward <alan.hayward@arm.com>
12684 David Sherwood <david.sherwood@arm.com>
12686 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
12687 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
12688 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
12689 (*fcmuo<mode>_and): New patterns.
12691 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
12693 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
12694 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
12695 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
12696 (cmp_op, sve_imm_con): New code attributes.
12697 (SVE_COND_INT_CMP, imm_con): Delete.
12698 (cmp_op): Remove above unspecs from int attribute.
12699 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
12701 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
12702 comparison-specific unspecs.
12703 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
12704 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
12705 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
12706 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
12707 (*vec_fcm<cmp_op><mode>): Rename to...
12708 (*fcm<cmp_op><mode>): ...this and adjust likewise.
12709 (*vec_fcmuo<mode>): Rename to...
12710 (*fcmuo<mode>): ...this and adjust likewise.
12711 (*pred_fcm<cmp_op><mode>): New pattern.
12712 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
12713 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
12715 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
12717 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
12718 (aarch64_emit_sve_predicated_cond): New function.
12719 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
12720 (aarch64_emit_unspec_cond_or): Replace with...
12721 (aarch64_emit_sve_or_conds): ...this new function. Use
12722 aarch64_emit_sve_ptrue_op for the individual comparisons and
12723 aarch64_emit_binop to OR them together.
12724 (aarch64_emit_inverted_unspec_cond): Replace with...
12725 (aarch64_emit_sve_inverted_cond): ...this new function. Use
12726 aarch64_emit_sve_ptrue_op for the comparison and
12727 aarch64_emit_unop to invert the result.
12728 (aarch64_expand_sve_vec_cmp_float): Update after the above
12729 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
12731 2018-05-07 Nathan Sidwell <nathan@acm.org>
12733 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
12734 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
12735 (Backwards Compatibility): Likewise.
12737 2018-05-07 Luis Machado <luis.machado@linaro.org>
12741 2018-05-07 Luis Machado <luis.machado@linaro.org>
12743 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
12744 <prefetch_dynamic_strides>: New const bool field.
12745 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
12746 prefetch_dynamic_strides.
12747 (exynosm1_prefetch_tune): Likewise.
12748 (thunderxt88_prefetch_tune): Likewise.
12749 (thunderx_prefetch_tune): Likewise.
12750 (thunderx2t99_prefetch_tune): Likewise.
12751 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
12753 (aarch64_override_options_internal): Update to set
12754 PARAM_PREFETCH_DYNAMIC_STRIDES.
12755 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
12756 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
12757 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
12758 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
12759 prefetch-dynamic-strides setting.
12761 2018-05-07 Luis Machado <luis.machado@linaro.org>
12763 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
12764 <minimum_stride>: New const int field.
12765 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
12766 minimum_stride field.
12767 (exynosm1_prefetch_tune): Likewise.
12768 (thunderxt88_prefetch_tune): Likewise.
12769 (thunderx_prefetch_tune): Likewise.
12770 (thunderx2t99_prefetch_tune): Likewise.
12771 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
12772 (aarch64_override_options_internal): Update to set
12773 PARAM_PREFETCH_MINIMUM_STRIDE.
12774 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
12775 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
12776 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
12777 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
12778 stride is constant and is below the minimum stride threshold.
12780 2018-05-07 Luis Machado <luis.machado@linaro.org>
12782 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
12785 2018-05-07 Luis Machado <luis.machado@linaro.org>
12787 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
12788 <prefetch_dynamic_strides>: New const bool field.
12789 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
12790 prefetch_dynamic_strides.
12791 (exynosm1_prefetch_tune): Likewise.
12792 (thunderxt88_prefetch_tune): Likewise.
12793 (thunderx_prefetch_tune): Likewise.
12794 (thunderx2t99_prefetch_tune): Likewise.
12795 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
12797 (aarch64_override_options_internal): Update to set
12798 PARAM_PREFETCH_DYNAMIC_STRIDES.
12799 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
12800 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
12801 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
12802 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
12803 prefetch-dynamic-strides setting.
12805 2018-05-07 Luis Machado <luis.machado@linaro.org>
12807 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
12808 <minimum_stride>: New const int field.
12809 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
12810 minimum_stride field.
12811 (exynosm1_prefetch_tune): Likewise.
12812 (thunderxt88_prefetch_tune): Likewise.
12813 (thunderx_prefetch_tune): Likewise.
12814 (thunderx2t99_prefetch_tune): Likewise.
12815 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
12816 (aarch64_override_options_internal): Update to set
12817 PARAM_PREFETCH_MINIMUM_STRIDE.
12818 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
12819 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
12820 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
12821 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
12822 stride is constant and is below the minimum stride threshold.
12824 2018-05-06 Jakub Jelinek <jakub@redhat.com>
12827 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
12828 the type is addressable. Don't force op into register if it has
12831 2018-05-05 Roland McGrath <mcgrathr@google.com>
12834 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
12835 any section for which we don't know a specific type it should have,
12836 regardless of name. Previously this was done only for the exact
12837 names ".init_array", ".fini_array", and ".preinit_array".
12838 (default_elf_asm_named_section): Add comment about
12839 relationship with default_section_type_flags and SECTION_NOTYPE.
12840 (get_section): Don't consider it a type conflict if one side has
12841 SECTION_NOTYPE and the other doesn't, as long as neither has the
12842 SECTION_BSS et al used in the default_section_type_flags logic.
12844 2018-05-05 Tom de Vries <tom@codesourcery.com>
12847 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
12848 (workaround_barsyncs): New function.
12849 (nvptx_reorg): Use workaround_barsyncs.
12850 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
12851 (define_expand "nvptx_membar_cta"): New define_expand.
12852 (define_insn "*nvptx_membar_cta"): New insn.
12854 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
12856 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
12857 To improve optimization opportunities.
12858 * builtin-types.def: The new needed builtin types for the above.
12860 2018-05-04 Richard Biener <rguenther@suse.de>
12862 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
12863 * gimple-ssa-store-merging.c
12864 (imm_store_chain_info::output_merged_store): Remove redundant create,
12865 release split_store vector contents on failure.
12866 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
12867 scalar stmt vector on cache hit.
12869 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
12871 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
12873 * config.gcc (powerpc-xilinx-eabi*): Remove.
12874 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
12876 (fusion_addis_mem_combo_load): Ditto.
12877 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
12879 (rs6000_cpu_cpp_builtins): Ditto.
12880 * config/rs6000/rs6000-linux.c
12881 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
12882 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
12883 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
12885 (rs6000_setup_reg_addr_masks): Ditto.
12886 (rs6000_init_hard_regno_mode_ok): Ditto.
12887 (rs6000_option_override_internal): Ditto.
12888 (legitimate_lo_sum_address_p): Ditto.
12889 (rs6000_legitimize_address): Ditto.
12890 (rs6000_legitimize_reload_address): Ditto.
12891 (rs6000_legitimate_address_p): Ditto.
12892 (abi_v4_pass_in_fpr): Ditto.
12893 (setup_incoming_varargs): Ditto.
12894 (rs6000_gimplify_va_arg): Ditto.
12895 (rs6000_split_multireg_move): Ditto.
12896 (rs6000_savres_strategy): Ditto.
12897 (rs6000_emit_prologue_components): Ditto.
12898 (rs6000_emit_epilogue_components): Ditto.
12899 (rs6000_emit_prologue): Ditto.
12900 (rs6000_emit_epilogue): Ditto.
12901 (rs6000_elf_file_end): Ditto.
12902 (rs6000_function_value): Ditto.
12903 (rs6000_libcall_value): Ditto.
12904 * config/rs6000/rs6000.h: Ditto.
12905 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
12906 (TARGET_MINMAX): ... this. New.
12907 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
12908 * config/rs6000/rs6000.md: Remove Xilinx FP support.
12909 (*movsi_internal1_single): Delete.
12910 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
12911 mfpu=, mxilinx-fpu): Delete.
12912 * config/rs6000/singlefp.h: Delete.
12913 * config/rs6000/sysv4.h: Remove Xilinx FP support.
12914 * config/rs6000/t-rs6000: Ditto.
12915 * config/rs6000/t-xilinx: Delete.
12916 * config/rs6000/titan.md: Adjust for fp_type removal.
12917 * config/rs6000/vsx.md: Remove Xilinx FP support.
12918 (VStype_simple): Delete.
12919 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
12920 * config/rs6000/xfpu.h: Delete.
12921 * config/rs6000/xfpu.md: Delete.
12922 * config/rs6000/xilinx.h: Delete.
12923 * config/rs6000/xilinx.opt: Delete.
12924 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
12925 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
12927 2018-05-04 Tom de Vries <tom@codesourcery.com>
12930 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
12933 2018-05-04 Richard Biener <rguenther@suse.de>
12935 PR middle-end/85627
12936 * tree-complex.c (update_complex_assignment): We are always in SSA form.
12937 (expand_complex_div_wide): Likewise.
12938 (expand_complex_operations_1): Likewise.
12939 (expand_complex_libcall): Preserve EH info of the original stmt.
12940 (tree_lower_complex): Handle removed blocks.
12941 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
12942 on complex multiplication and division libcall builtins.
12944 2018-05-04 Richard Biener <rguenther@suse.de>
12946 PR middle-end/85574
12947 * fold-const.c (negate_expr_p): Restrict negation of operand
12948 zero of a division to when we know that can happen without
12950 (fold_negate_expr_1): Likewise.
12952 2018-05-04 Jakub Jelinek <jakub@redhat.com>
12955 * real.h (real_nextafter): Declare.
12956 * real.c (real_nextafter): New function.
12957 * fold-const-call.c (fold_const_nextafter): New function.
12958 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
12959 CASE_CFN_NEXTTOWARD.
12960 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
12961 even when arg1_mode is different from arg0_mode.
12963 2018-05-03 Nathan Sidwell <nathan@acm.org>
12965 * doc/extend.texi (Deprecated Features): Remove
12966 -ffriend-injection.
12967 (Backwards Compatibility): Likewise.
12968 * doc/invoke.texi (C++ Language Options): Likewise.
12969 (C++ Dialect Options): Likewise.
12971 2018-05-03 Jakub Jelinek <jakub@redhat.com>
12974 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
12975 _mm512_mask_mullox_epi64): New intrinsics.
12977 2018-05-03 Tom de Vries <tom@codesourcery.com>
12980 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
12981 dump files): Add offload-tree.
12983 2018-05-03 Richard Biener <rguenther@suse.de>
12985 PR tree-optimization/85615
12986 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
12987 to loops not nested in BBs loop father to avoid creating multi-entry
12990 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12992 PR tree-optimization/70291
12993 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
12994 arguments. Change return type to tree. Emit libcall as a new
12995 statement rather than replacing existing one when inplace_p is true.
12996 (expand_complex_multiplication_components): New function.
12997 (expand_complex_multiplication): Expand floating-point complex
12998 multiplication using the above.
12999 (expand_complex_division): Rename inner_type parameter to type.
13000 Update expand_complex_libcall call-site.
13001 (expand_complex_operations_1): Update expand_complex_multiplication
13002 and expand_complex_division call-sites.
13004 2018-05-02 Jakub Jelinek <jakub@redhat.com>
13007 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
13008 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
13009 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
13010 the highest significant bit of the shift count mask is clear. In
13011 check whether and[sq]i3 is needed verify that all significant bits
13012 of the shift count other than the highest are set.
13014 2018-05-02 Tom de Vries <tom@codesourcery.com>
13017 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
13018 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
13019 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
13020 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
13021 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
13022 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
13023 __builtin_goacc_parlevel_size.
13025 2018-05-02 Richard Biener <rguenther@suse.de>
13027 PR tree-optimization/85597
13028 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
13029 do not use split vect_get_vec_defs call but call vect_get_slp_defs
13032 2018-05-02 Tom de Vries <tom@codesourcery.com>
13035 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
13036 dump files): Add ltrans-tree.
13038 2018-05-02 Tom de Vries <tom@codesourcery.com>
13041 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
13042 dump files): Add wpa-ipa.
13044 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
13046 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
13047 powerpc*-*-linux*paired* target.
13048 * config/rs6000/750cl.h: Delete.
13049 * config/rs6000/paired.h: Delete.
13050 * config/rs6000/paired.md: Delete.
13051 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
13053 * config/rs6000/rs6000-builtin.def: Remove paired float support.
13054 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
13055 comment. Remove paired float support.
13056 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
13057 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
13059 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
13060 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
13062 * config/rs6000/rs6000.c: Remove paired float support.
13063 (paired_expand_vector_init, paired_expand_vector_move,
13064 paired_emit_vector_compare, paired_emit_vector_cond_expr,
13065 (paired_expand_lv_builtin, paired_expand_stv_builtin,
13066 paired_expand_builtin, paired_expand_predicate_builtin,
13067 paired_init_builtins): Delete.
13068 * config/rs6000/rs6000.h: Remove paired float support.
13069 * config/rs6000/rs6000.md: Remove paired float support.
13070 (move_from_CR_ov_bit): Delete.
13071 * config/rs6000/rs6000.opt (mpaired): Delete.
13072 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
13073 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
13075 2018-05-02 Richard Biener <rguenther@suse.de>
13077 PR middle-end/85567
13078 * gimplify.c (gimplify_save_expr): When in SSA form allow
13079 SAVE_EXPRs to compute to SSA vars.
13081 2018-05-02 Jakub Jelinek <jakub@redhat.com>
13084 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
13085 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
13086 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
13087 clobber operands[2], instead use a new pseudo. Formatting fixes.
13089 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
13091 PR tree-optimization/85586
13092 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
13093 exit early for statements in the same group if the accesses are
13096 2018-05-02 Tom de Vries <tom@codesourcery.com>
13099 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
13102 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
13104 PR tree-optimization/85143
13105 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
13107 2018-05-01 Tom de Vries <tom@codesourcery.com>
13110 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
13111 not found" error message.
13113 2018-05-01 Tom de Vries <tom@codesourcery.com>
13116 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
13117 * vec.c (test_ordered_remove_if): New function.
13118 (vec_c_tests): Call test_ordered_remove_if.
13119 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
13120 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
13121 * tree-vect-patterns.c (vect_pattern_recog_1): Use
13122 VEC_ORDERED_REMOVE_IF.
13124 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13126 PR tree-optimization/82665
13127 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
13128 pointer subtraction where arguments come from a memchr call.
13130 2018-05-01 Jakub Jelinek <jakub@redhat.com>
13132 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
13133 --push-state --as-needed and --pop-state instead of --as-needed and
13134 --no-as-needed if ld supports it.
13135 * configure: Regenerated.
13138 * doc/install.texi2html: Replace _002d with - and _002a with * in
13139 generated html files using sed.
13141 2018-04-30 David Malcolm <dmalcolm@redhat.com>
13144 * gcc-rich-location.c (blank_line_before_p): New function.
13145 (use_new_line): New function.
13146 (gcc_rich_location::add_fixit_insert_formatted): New function.
13147 * gcc-rich-location.h
13148 (gcc_rich_location::add_fixit_insert_formatted): New function.
13150 2018-04-30 David Malcolm <dmalcolm@redhat.com>
13152 * selftest.c (assert_streq): Rename "expected" and "actual" to
13153 "val1" and "val2". Extend NULL-handling to cover both inputs
13154 symmetrically, while still requiring both to be non-NULL for a pass.
13155 * selftest.h (assert_streq): Rename "expected" and "actual" to
13157 (ASSERT_EQ): Likewise.
13158 (ASSERT_EQ_AT): Likewise.
13159 (ASSERT_KNOWN_EQ): Likewise.
13160 (ASSERT_KNOWN_EQ_AT): Likewise.
13161 (ASSERT_NE): Likewise.
13162 (ASSERT_MAYBE_NE): Likewise.
13163 (ASSERT_MAYBE_NE_AT): Likewise.
13164 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
13165 the assertion to pass.
13166 (ASSERT_STREQ_AT): Likewise.
13168 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
13170 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
13171 interaction with -pie.
13173 2018-04-30 David Malcolm <dmalcolm@redhat.com>
13175 * selftest.h: Fix alphabetization of per-source-file selftest
13178 2018-04-30 Jason Merrill <jason@redhat.com>
13180 PR c++/61982 - dead stores to destroyed objects.
13181 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
13184 2018-04-30 Jason Merrill <jason@redhat.com>
13186 * tree.c (build_clobber): New.
13187 * tree.h: Declare it.
13188 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
13190 2018-04-30 David Malcolm <dmalcolm@redhat.com>
13192 * diagnostic-show-locus.c (layout::layout): Update for
13193 location_get_source_line returning a char_span.
13194 (struct char_span): Move to input.h.
13195 (struct correction): Update for fields in char_span becoming
13197 (struct source_line): Update for location_get_source_line
13198 returning a char_span.
13199 (layout::print_line): Likewise.
13200 * edit-context.c (edited_file::print_content): Likewise.
13201 (edited_file::print_diff_hunk): Likewise.
13202 (edited_file::print_run_of_changed_lines): Likewise.
13203 (edited_file::get_num_lines): Likewise.
13204 (edited_line::edited_line): Likewise.
13205 * final.c (asm_show_source): Likewise.
13206 * input.c (location_get_source_line): Convert return type
13207 from const char * to char_span, losing the final "line_len"
13209 (dump_location_info): Update for the above.
13210 (get_substring_ranges_for_loc): Likewise. Use a char_span
13211 when handling the literal within the line.
13212 (test_reading_source_line): Update for location_get_source_line
13213 returning a char_span.
13214 * input.h (class char_span): Move here from
13215 diagnostic-show-locus.c, converting from a struct to a class.
13216 Make data members private.
13217 (char_span::operator bool): New.
13218 (char_span::length): New.
13219 (char_span::get_buffer): New.
13220 (char_span::operator[]): New.
13221 (char_span::subspan): Make const.
13222 (char_span::xstrdup): New.
13223 (location_get_source_line): Convert return type from const char *
13224 to char_span, losing the final "line_size" param.
13226 2018-04-30 Jan Hubicka <jh@suse.cz>
13228 * lto-wrapper.c (ltrans_priorities): New static var.
13229 (cmp_priority): New.
13230 (run_gcc): Read priorities and if doing parallel build order
13231 the Makefile by them.
13233 2018-04-30 David Malcolm <dmalcolm@redhat.com>
13235 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
13237 2018-04-30 Richard Biener <rguenther@suse.de>
13239 * tree-cfg.c (verify_address): Remove base argument, add
13240 flag whether to check TREE_ADDRESSABLE and do that.
13241 (verify_expr): Remove.
13242 (verify_types_in_gimple_reference): Add pieces from verify_expr.
13243 (verify_gimple_assign_single): Likewise.
13244 (verify_gimple_switch): Likewise.
13245 (verify_expr_location_1): Dereference tp once. Add (disabled)
13246 piece from verify_expr.
13247 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
13249 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
13251 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
13253 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
13255 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
13256 (small_data_pattern): Likewise.
13257 (arc_rewrite_small_data): Likewise.
13258 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
13259 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
13260 (get_symbol_alignment): New function.
13261 (legitimate_small_data_address_p): Likewise.
13262 (legitimate_scaled_address): Update, call
13263 legitimate_small_data_address_p.
13264 (output_sdata): New static variable.
13265 (arc_print_operand): Update how we handle small data operands.
13266 (arc_print_operand_address): Likewise.
13267 (arc_legitimate_address_p): Update, use
13268 legitimate_small_data_address_p.
13269 (arc_rewrite_small_data_p): Remove.
13270 (arc_rewrite_small_data_1): Likewise.
13271 (arc_rewrite_small_data): Likewise.
13272 (small_data_pattern): Likewise.
13273 (compact_sda_memory_operand): Update to use
13274 legitimate_small_data_address_p and get_symbol_alignment.
13275 (prepare_move_operands): Don't rewite sdata pattern.
13276 (prepare_extend_operands): Remove.
13277 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
13279 (zero_extendqisi2): Likewise.
13280 (zero_extendhisi2): Likewise.
13281 (extendqihi2): Likewise.
13282 (extendqisi2): Likewise.
13283 (extendhisi2): Likewise.
13284 (addsi3): Likewise.
13285 (subsi3): Likewise.
13286 (andsi3): Likewise.
13287 * config/arc/constraints.md (Usd): Change it to memory constraint.
13289 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
13291 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
13292 as source of std instructions.
13293 * config/arc/arc.md (movsi_insn): Update pattern predicate to
13294 allow 6-bit constants as source for store instructions.
13295 (movdi_insn): Update instruction pattern to allow 6-bit constants
13296 as source for store instructions.
13298 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
13300 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
13302 2018-04-30 Nathan Sidwell <nathan@acm.org>
13303 Sandra Loosemore <sandra@codesourcery.com>
13305 * dumpfile.c (dump_open): Allow '-' for stdout.
13306 * doc/invoke.texi (Developer Options): Document dump filename
13307 determination early. Document stdin/stdout selection.
13309 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
13311 Microblaze Target: PIC data text relative
13313 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
13314 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
13316 * config/microblaze/microblaze.h (microblaze_constant_address_p):
13317 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
13318 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
13319 New addressing mode for data-text relative position indepenedent code.
13320 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
13321 'ADDRESS_SYMBOLIC_TXT_REL'.
13322 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
13323 (microblaze_legitimate_pic_operand): Exclude function calls from
13324 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
13325 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
13327 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
13328 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
13329 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
13330 for 'address + offset'.
13331 (microblaze_expand_prologue): Add new function prologue call for
13333 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
13334 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
13335 table in case of TARGET_PIC_DATA_TEXT_REL.
13336 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
13337 * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
13338 Add new macros 'UNSPEC_TEXT',
13339 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
13340 + exclude function calls from 'UNSPEC_PLT' in case of data text
13342 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
13343 new target hook for generating address diff vector tables in case of
13345 * doc/tm.texi : Regenerate.
13346 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
13347 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
13348 of addr diff vector generation.
13349 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
13350 target hook definition.
13351 * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
13352 Add default function for generate_pic_addr_diff_vec -> flag_pic.
13353 * doc/invoke.texi (Add new pic option): Add new microblaze pic
13354 option for data text relative.
13356 2018-04-30 Richard Biener <rguenther@suse.de>
13358 * tree-chrec.h (evolution_function_is_constant_p): Remove
13360 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
13362 2018-04-30 Richard Biener <rguenther@suse.de>
13365 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
13367 2018-04-30 Richard Biener <rguenther@suse.de>
13369 PR tree-optimization/28364
13370 PR tree-optimization/85275
13371 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
13372 copying first exit test.
13374 2018-04-28 Mark Wielaard <mark@klomp.org>
13376 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
13377 dwarf_version >= 5.
13378 (dwarf_AT): Handle DW_AT_addr_base.
13379 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
13381 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
13384 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
13385 (*ashl<dwi>3_doubleword_mask_1): Ditto.
13386 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
13387 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
13389 2018-04-28 Richard Biener <rguenther@suse.de>
13391 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
13392 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
13393 to reflect use. Only add interesting stmts.
13395 2018-04-27 Martin Jambor <mjambor@suse.cz>
13398 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
13399 the jump function allows for passing through aggregate values.
13401 2018-04-27 David Malcolm <dmalcolm@redhat.com>
13403 * input.h (in_system_header_at): Convert from macro to inline
13405 (from_macro_expansion_at): Likewise.
13406 (from_macro_definition_at): Likewise.
13408 2018-04-27 Jeff Law <law@redhat.com>
13410 * config.gcc: Mark tile* targets as deprecated/obsolete.
13412 2018-04-27 Richard Biener <rguenther@suse.de>
13414 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
13417 2018-04-27 Richard Biener <rguenther@suse.de>
13419 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
13421 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
13423 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
13424 with Yd constraint. Set "preferred_for_speed" attribute from
13425 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
13426 with Yd constraint.
13427 (*movdi_internal): Ditto.
13428 (movti_interunit splitters): Remove
13429 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
13430 (movdi_interunit splitters): Ditto.
13431 * config/i386/constraints.md (Ye): Remove.
13432 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
13434 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13437 * config/aarch64/constraints.md (Usg): Limit to 31.
13438 (Usj): Limit to 63.
13440 2018-04-27 Jakub Jelinek <jakub@redhat.com>
13442 PR tree-optimization/85529
13443 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
13444 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
13445 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
13446 zero extension or masking of the MSB bit.
13447 (optimize_range_tests): Add FIRST_BB argument, pass it through
13448 to optimize_range_tests_var_bound.
13449 (maybe_optimize_range_tests, reassociate_bb): Adjust
13450 optimize_range_tests callers.
13452 2018-04-26 Richard Biener <rguenther@suse.de>
13453 Jakub Jelinek <jakub@redhat.com>
13455 * cgraph.h (symbol_table): Just declare debug method here.
13456 * symtab.c (symbol_table::debug): Define.
13458 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
13460 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
13462 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
13464 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
13465 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
13466 (*movdi_internal): Substitute Yi and Yj constraint with x
13467 and Ym and Yn constraint with y constraint. Update "isa"
13468 attribute and set "preferred_for_speed" attribute from
13469 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
13470 (*movsi_internal): Ditto.
13471 (*movdf_internal): Ditto.
13472 (*movsf_internal): Ditto.
13473 (*zero_extendsidi2): Ditto.
13474 * config/i386/sse.md (vec_set<mode>_0): Ditto.
13475 (sse2_loadld): Ditto.
13476 (*vec_extract<ssevecmodelower>_0): Ditto.
13477 (*vec_extractv4si_0_zext_sse4): Ditto.
13478 (vec_concatv2di): Ditto.
13479 (*vec_dup<mode>): Ditto.
13480 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
13481 * config/i386/constraints.md (Yi): Remove.
13486 2018-04-26 Nathan Sidwell <nathan@acm.org>
13488 * dumpfile.c (dump_open): New.
13489 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
13490 (dump_finish): Detect stdio/stderr by value not name.
13492 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
13494 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
13496 2018-04-26 Tom de Vries <tom@codesourcery.com>
13499 * config/nvptx/nvptx.c (verify_neutering_jumps)
13500 (verify_neutering_labels): New function
13501 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
13503 2018-04-26 Tom de Vries <tom@codesourcery.com>
13506 * config/nvptx/nvptx.c (needs_neutering_p): New function.
13507 (nvptx_single): Use needs_neutering_p to skip over insns that do not
13510 2018-04-26 Richard Biener <rguenther@suse.de>
13511 Tom de Vries <tom@codesourcery.com>
13514 * lto-streamer-out.c (output_function): Fixup loops if required to match
13515 discovery done in the reader.
13517 2018-04-26 Richard Biener <rguenther@suse.de>
13519 PR tree-optimization/85116
13520 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
13521 have a loop exit from the single latch predecessor. Remove
13522 case of header with just condition.
13523 (ch_base::copy_headers): Exclude infinite loops from any
13525 (pass_ch::execute): Record exits.
13527 2018-04-26 Richard Biener <rguenther@suse.de>
13529 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
13530 prologue cost vector and pass it to vect_get_load_cost.
13531 (vect_get_peeling_costs_all_drs): Likewise.
13532 (vect_peeling_hash_get_lowest_cost): Likewise.
13533 (vect_enhance_data_refs_alignment): Likewise.
13535 2018-04-26 Richard Biener <rguenther@suse.de>
13537 PR middle-end/85450
13538 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
13539 checking of integer<->pointer conversions.
13540 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
13541 sign-/zero-extending pointer types.
13542 (expand_omp_for_static_chunk): Likewise.
13544 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
13545 Jean Lee <xiaoyur347@gmail.com>
13547 * config/mips/mips.c (mips_asan_shadow_offset): New function.
13548 (TARGET_ASAN_SHADOW_OFFSET): Define.
13549 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
13550 true for -fsanitize=address.
13552 2018-04-25 Mark Wielaard <mark@klomp.org>
13554 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
13557 2018-04-25 Jakub Jelinek <jakub@redhat.com>
13559 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
13560 than "alu", remove explicit "memory" and "imm_disp" attributes.
13561 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
13563 PR middle-end/85414
13564 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
13565 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
13566 gen_lowpart_no_emit.
13568 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
13571 * config/i386/i386.c (ix86_expand_builtin): Change memory
13572 operand to XI, extend p0 to Pmode.
13573 * config/i386/i386.md: Change unspec volatile and operand
13574 1 mode to XI, change operand 0 mode to P.
13576 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
13578 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
13579 GET_MODE_MASK before any checking.
13580 (nds32_can_use_bset_p): Likewise.
13581 (nds32_can_use_btgl_p): Likewise.
13583 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
13585 * config/nds32/nds32-doubleword.md: New define_split pattern for
13586 illegal register number.
13588 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
13590 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
13592 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
13594 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
13596 2018-04-25 Richard Biener <rguenther@suse.de>
13598 * lto-streamer.h (LTO_major_version): Bump to 8.
13600 2018-04-25 Jakub Jelinek <jakub@redhat.com>
13602 * BASE-VER: Set to 9.0.0.
13604 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
13606 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
13607 in __abskf2 and __powikf2.
13609 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13612 * config/aarch64/constraints.md (Usg, Usj): New constraints.
13613 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
13614 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
13615 Use the above on operand 2. Reindent.
13616 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
13618 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
13621 * common/config/i386/i386-common.c (ix86_handle_option): Don't
13623 * config/i386/i386.opt (mcet): Removed.
13624 * doc/install.texi: Remove -mcet documentation.
13625 * doc/invoke.texi: Likewise.
13627 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
13630 * doc/install.texi: Remove -mcet from bootstrap-cet.
13632 2018-04-24 Jakub Jelinek <jakub@redhat.com>
13635 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
13636 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
13640 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
13641 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
13642 containing a CONST_VECTOR.
13644 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
13646 * doc/install.texi: Update newlib dependency for nvptx.
13648 2018-04-24 Jakub Jelinek <jakub@redhat.com>
13651 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
13652 instead of INTVAL when shifting x left.
13654 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
13656 PR tree-optimization/85478
13657 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
13658 vect_grouped_store_supported for single element vectors.
13660 2018-04-24 Richard Biener <rguenther@suse.de>
13663 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
13664 load cost increase to the case of non-constant step.
13666 2018-04-24 Jakub Jelinek <jakub@redhat.com>
13669 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
13670 destination if any_malformed_asm.
13672 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
13674 PR middle-end/85496
13675 * expr.c (store_field): In the bitfield case, if the value comes from
13676 a function call and is returned in registers by means of a PARALLEL,
13677 do not change the mode of the temporary unless BLKmode and VOIDmode.
13679 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
13681 PR rtl-optimization/85423
13682 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
13683 dependencies to debug insns when the previous insn is non-debug.
13685 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
13687 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
13688 enums into a single definition.
13689 (fls): Fix predicates and printing.
13692 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
13694 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
13695 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
13696 and short u6 immediate.
13697 (check_if_valid_sleep_operand): Remove.
13698 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
13700 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
13702 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
13703 flag_always_save_lp condition.
13704 * config/nds32/nds32.opt (malways-save-lp): New option.
13706 2018-04-22 Shiva Chen <shiva0217@gmail.com>
13708 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
13709 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
13710 * config/nds32/nds32.h
13711 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
13712 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
13714 2018-04-22 Shiva Chen <shiva0217@gmail.com>
13716 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
13717 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
13719 2018-04-22 Shiva Chen <shiva0217@gmail.com>
13720 Chung-Ju Wu <jasonwucj@gmail.com>
13722 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
13724 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
13725 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
13727 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
13729 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
13731 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
13733 * config/nds32/nds32-protos.h (nds32_data_alignment,
13734 nds32_local_alignment): Declare.
13735 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
13736 nds32_local_alignment): New functions.
13737 (TARGET_CONSTANT_ALIGNMENT): Define.
13738 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
13740 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
13742 * config/nds32/nds32.c
13743 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
13744 (TARGET_MODES_TIEABLE_P): Likewise.
13746 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
13748 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
13749 level Ofast and Og.
13751 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
13752 Chung-Ju Wu <jasonwucj@gmail.com>
13754 * config/nds32/constants.md (unspec_volatile_element): Add enum values
13755 for unaligned access.
13756 * config/nds32/nds32-intrinsic.c: Implementation of expanding
13758 * config/nds32/nds32-intrinsic.md: Likewise.
13759 * config/nds32/nds32_intrinsic.h: Likewise.
13760 * config/nds32/nds32.h (nds32_builtins): Likewise.
13761 * config/nds32/nds32.opt (munaligned-access): New option.
13762 * config/nds32/nds32.c (nds32_asm_file_start): Display
13763 flag_unaligned_access status.
13765 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
13767 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
13768 -mno-relax is present.
13769 * config/riscv/linux.h (LINK_SPEC): Ditto.
13771 2018-04-20 Martin Sebor <msebor@redhat.com>
13774 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
13776 (gimple_fold_builtin_stxcpy_chk): Same.
13777 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
13779 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
13782 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
13783 __powikf2 when long double is IEEE 128-bit.
13785 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
13787 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
13788 step to make sure stack always aligned.
13790 2018-04-20 Carl Love <cel@us.ibm.com>
13793 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
13794 size check for arg0.
13796 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
13797 Tom de Vries <tom@codesourcery.com>
13800 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
13801 Emit insns for calls too.
13802 (nvptx_find_par): Always look for worker-level predecessor insn.
13803 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
13805 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
13806 (nvptx_process_pars): Propagate frames for calls.
13808 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
13811 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
13813 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
13814 (ix86_handle_option): Don't handle OPT_mibt.
13815 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
13817 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
13819 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
13820 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
13821 (ix86_target_macros): Define __CET__ with flag_cf_protection
13822 for -fcf-protection.
13823 * config/i386/i386.c (isa2_opts): Remove -mibt.
13824 * config/i386/i386.h (TARGET_IBT): Removed.
13825 (TARGET_IBT_P): Likewise.
13826 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
13827 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
13828 * config/i386/i386.opt (mcet): Update help message.
13829 (mshstk): Likewise.
13831 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
13832 -mcet as an alias for -mshstk.
13834 2018-04-20 Richard Biener <rguenther@suse.de>
13836 PR middle-end/85475
13837 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
13838 complexity by forcing a single use of the multiply operand.
13840 2018-04-20 Martin Jambor <mjambor@suse.cz>
13843 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
13844 recursion dependency to only apply to non-clones.
13846 2018-04-20 Martin Jambor <mjambor@suse.cz>
13849 * ipa-cp.c (create_specialized_node): Check that clones of
13850 self-recursive edges exist during IPA-CP.
13852 2018-04-19 Toon Moene <toon@moene.org>
13854 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
13857 2018-04-19 Jakub Jelinek <jakub@redhat.com>
13859 PR tree-optimization/85467
13860 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
13861 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
13862 VECTOR_CST element to type.
13864 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
13867 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
13868 * config/i386/i386.md (builtin_setjmp_setup): Removed.
13869 (builtin_longjmp): Likewise.
13870 (save_stack_nonlocal): New pattern.
13871 (restore_stack_nonlocal): Likewise.
13873 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
13876 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
13877 Replace ASM_OUTPUT_LABEL with fprintf.
13879 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
13882 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
13883 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
13884 * config/i386/i386-c.c (ix86_target_macros_internal): Also
13885 define __IBT__ and __SHSTK__ for -fcf-protection.
13886 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
13888 (ix86_trampoline_init): Likewise.
13889 (x86_output_mi_thunk): Likewise.
13890 (ix86_notrack_prefixed_insn_p): Likewise.
13891 (ix86_option_override_internal): Don't disallow -fcf-protection.
13892 * config/i386/i386.md (rdssp<mode>): Also enable for
13894 (incssp<mode>): Likewise.
13895 (nop_endbr): Likewise.
13896 * config/i386/i386.opt (mcet): Change help message to built-in
13899 (mshstk): Likewise.
13900 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
13901 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
13902 enable CET built-in functions.
13904 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
13906 * common/config/i386/i386-common.c
13907 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
13908 OPTION_MASK_ISA_MOVDIRI_UNSET,
13909 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
13910 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
13911 * config.gcc (movdirintrin.h): New header.
13912 * config/i386/cpuid.h (bit_MOVDIRI,
13913 bit_MOVDIR64B): New bits.
13914 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
13916 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
13917 (VOID, PVOID, PCVOID)): New function types.
13918 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
13919 __builtin_ia32_directstoreu_u64,
13920 __builtin_ia32_movdir64b): New builtins.
13921 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
13922 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
13924 (ix86_valid_target_attribute_inner_p): Ditto.
13925 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
13926 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
13927 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
13928 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
13929 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
13930 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
13931 (movdiri<mode>, movdir64b_<mode>): New.
13932 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
13933 * config/i386/immintrin.h: Include movdirintrin.h.
13934 * config/i386/movdirintrin.h: New file.
13935 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
13937 2018-04-19 Richard Biener <rguenther@suse.de>
13939 PR middle-end/85455
13940 * cfg.c (clear_bb_flags): When loop state says we have
13941 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
13943 2018-04-19 Richard Biener <rguenther@suse.de>
13945 PR tree-optimization/84737
13946 * tree-vect-data-refs.c (vect_copy_ref_info): New function
13947 copying restrict info.
13948 (vect_setup_realignment): Use it.
13949 * tree-vectorizer.h (vect_copy_ref_info): Declare.
13950 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
13951 the first DR to all generated stores.
13952 (vectorizable_load): Likewise for loads.
13954 2018-04-19 Jakub Jelinek <jakub@redhat.com>
13956 PR tree-optimization/85446
13957 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
13958 the integral and pointer types to have the same precision.
13960 * doc/install.texi: Document --disable-cet being the default and
13963 2018-04-18 Martin Liska <mliska@suse.cz>
13965 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
13968 2018-04-18 Martin Liska <mliska@suse.cz>
13971 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
13974 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
13975 arguments if they are comparable.
13977 2018-04-18 Martin Liska <mliska@suse.cz>
13980 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
13983 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
13986 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
13989 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
13990 ENDBR after calling __morestack.
13992 2018-04-18 David Malcolm <dmalcolm@redhat.com>
13995 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
13996 by using gcc_base_ver to generate a gcc_driver_version, and use
13997 it when generating GCC_DRIVER_NAME.
13998 * configure: Regenerate.
14000 2018-04-18 Jakub Jelinek <jakub@redhat.com>
14003 * config.gcc: Obsolete powerpc*-*-*spe*.
14005 2018-04-17 Jakub Jelinek <jakub@redhat.com>
14008 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
14009 (stabstr_D): Change type of unum from unsigned int to
14010 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
14013 2018-04-17 Jim Wilson <jimw@sifive.com>
14016 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
14017 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
14018 Set arg_pointer_offset after using pretend_args_size.
14020 2018-04-17 Jakub Jelinek <jakub@redhat.com>
14022 PR rtl-optimization/85431
14023 * dse.c (record_store): Ignore zero width stores.
14026 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
14027 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
14028 __builtin_stack_restore rather than after it.
14029 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
14030 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
14031 argument instead of virtual_dynamic_stack_rtx.
14033 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
14035 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
14037 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
14038 Add note to error message to explain internal mapping of overloaded
14039 built-in function name to non-overloaded built-in function name.
14040 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
14043 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
14046 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
14047 where the inputs overlap with the output.
14049 2018-04-17 Jakub Jelinek <jakub@redhat.com>
14051 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
14052 (=v, v) alternative and explicit "memory" attribute.
14053 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
14054 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
14056 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
14057 "sselog1" type instead of "sselog".
14058 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
14059 "sselog". Remove explicit "memory" attribute.
14060 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
14061 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
14063 (vec_extract_hi_v32hi): Merge all alternatives into one, use
14064 "sselog1" type instead of "sselog". Remove explicit "memory"
14066 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
14067 use "sselog1" type instead of "sselog". Remove explicit "memory"
14069 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
14070 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
14072 (vec_extract_hi_v64qi): Merge all alternatives into one, use
14073 "sselog1" type instead of "sselog". Remove explicit "memory"
14075 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
14076 use "sselog1" type instead of "sselog". Remove explicit "memory"
14080 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
14082 PR middle-end/85414
14083 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
14086 2018-04-17 Martin Jambor <mjambor@suse.cz>
14089 * ipa-cp.c (create_specialized_node): Call
14090 expand_all_artificial_thunks if necessary.
14092 2018-04-17 Martin Liska <mliska@suse.cz>
14095 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
14096 in message, remote space in between '_G' and '('.
14098 2018-04-17 Jakub Jelinek <jakub@redhat.com>
14101 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
14102 avx512f_vmcmp<mode>3<round_saeonly_name>,
14103 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
14104 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
14105 avx512f_rndscale<mode><round_saeonly_name>,
14106 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
14107 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
14108 Use %<iptr>2 instead of %2 for -masm=intel.
14109 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
14110 avx512f_vcvttss2usi<round_saeonly_name>,
14111 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
14113 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
14114 avx512f_vcvttsd2usi<round_saeonly_name>,
14115 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
14116 Use %q1 instead of %1 for -masm=intel.
14117 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
14118 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
14119 of %3 for -masm=intel.
14120 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
14122 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
14124 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
14126 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
14127 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
14129 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
14131 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
14132 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
14133 %g1 and one with %0 and %1.
14134 (avx512er_vmrcp28<mode><round_saeonly_name>,
14135 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
14136 %1 for -masm=intel.
14137 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
14138 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
14139 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
14140 of %0 and %{%4%} for -masm=intel.
14141 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
14142 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
14143 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
14144 order of %0 and %{%5%}%{z%} for -masm=intel.
14146 2018-04-17 Jan Hubicka <jh@suse.cz>
14149 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
14151 2018-04-17 Martin Liska <mliska@suse.cz>
14154 * multiple_target.c (create_dispatcher_calls): Set apostrophes
14155 for target_clone error message. Make default implementation
14156 clone to be a local declaration.
14157 (separate_attrs): Add new argument and check for an empty
14159 (expand_target_clones): Handle it.
14160 (ipa_target_clone): Make redirection just for target_clones
14163 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
14164 Tom de Vries <tom@codesourcery.com>
14166 PR middle-end/84955
14167 * omp-expand.c (expand_oacc_for): Add dummy false branch for
14168 tiled basic blocks without omp continue statements.
14170 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
14173 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
14174 vec_extract expression as having side effects to make sure it gets
14177 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
14180 * config/i386/i386.c (get_builtin_code_for_version): Check
14183 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
14186 * config.gcc: Support "skylake".
14187 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
14189 * config/i386/i386.c (m_SKYLAKE): Define.
14190 (processor_target_table): Add "skylake".
14191 (ix86_option_override_internal): Add "skylake".
14192 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
14193 PROCESSOR_CANNONLAKE.
14194 (get_builtin_code_for_version): Fix priority for
14195 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
14196 PROCESSOR_SKYLAKE-AVX512.
14197 * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
14198 (processor_type): Add PROCESSOR_SKYLAKE.
14200 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
14201 Jason Merrill <jason@redhat.com>
14204 * convert.c (convert_to_integer_1): Use direct recursion for
14205 enumeral types and types with a precision less than the number
14206 of bits in their mode.
14208 2018-04-16 Julia Koval <julia.koval@intel.com>
14211 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
14212 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
14214 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
14217 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
14218 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
14219 and -mno-direct-move.
14221 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
14224 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
14225 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
14226 Ensure negative shifts result in {0}.
14228 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
14230 PR rtl-optimization/79916
14231 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
14232 regs (if any) to define how to gnerate SD moves when LRA is in
14235 2018-04-13 Jakub Jelinek <jakub@redhat.com>
14237 PR rtl-optimization/85393
14238 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
14239 * except.c (expand_dw2_landing_pad_for_region): Make static.
14240 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
14241 a label and unconditional jump to old_bb, rather than
14242 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
14245 PR rtl-optimization/85376
14246 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
14247 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
14248 instead of a specific value.
14250 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
14251 Bin Cheng <bin.cheng@arm.com>
14253 PR tree-optimization/82965
14254 PR tree-optimization/83991
14255 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
14256 by_profile_only parameter.
14257 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
14258 information if the loop was predicted to iterate too many times.
14259 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
14261 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
14264 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
14267 2018-04-13 Martin Liska <mliska@suse.cz>
14268 Jakub Jelinek <jakub@redhat.com>
14270 PR middle-end/81657
14271 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
14272 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
14273 * builtins.c (expand_builtin_memory_copy_args): Use
14274 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
14275 handle dest_addr == pc_rtx.
14277 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
14280 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
14281 asked to not generate direct moves.
14282 (fix_trunc<mode>si2_stfiwx): Similar.
14283 (fix_trunc<mode>si2_internal): Similar.
14285 2018-04-12 Jakub Jelinek <jakub@redhat.com>
14288 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
14289 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
14290 lookup if dest in some wider mode is known to be const0_rtx and
14291 if so, record permanent equivalence for it to be ZERO_EXTEND of
14292 the narrower mode destination.
14294 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
14296 * lto-streamer-out.c (output_function): Revert 259346.
14297 * omp-expand.c (expand_oacc_for): Likewise.
14299 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
14301 PR rtl-optimization/85354
14302 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
14303 * sel-sched.c (sel_global_init): ... here.
14305 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
14308 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
14309 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
14310 mode for PE-COFF targets.
14311 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
14312 (i386_pe_asm_lto_end): Likewise.
14313 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
14314 (TARGET_ASM_LTO_END): Likewise.
14315 * config/i386/winnt.c (saved_debug_info_level): New static variable.
14316 (i386_pe_asm_lto_start): New function.
14317 (i386_pe_asm_lto_end): Likewise.
14319 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
14320 Richard Biener <rguenther@suse.de>
14322 PR middle-end/84955
14323 * lto-streamer-out.c (output_function): Fix CFG loop state before
14325 * omp-expand.c (expand_oacc_for): Handle calls to internal
14326 functions like regular functions.
14328 2018-04-12 Richard Biener <rguenther@suse.de>
14331 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
14332 for the early LTO debug to properly generate references to it
14333 during DIE emission. Do not re-use that for the skeleton for
14335 (dwarf2out_early_finish): Likewise.
14337 2018-04-12 Jakub Jelinek <jakub@redhat.com>
14340 * config/i386/sse.md
14341 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
14342 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
14343 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
14344 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
14345 and output is a reg, avoid creating invalid lowpart subreg, but
14346 instead split into a 512-bit move. Don't split if not AVX512VL,
14347 input is xmm16+ reg and output is a mem.
14348 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
14349 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
14350 xmm16+ reg and output is a mem.
14352 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
14354 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
14355 also for flag_dwarf2_cfi_asm.
14357 2018-04-12 Jakub Jelinek <jakub@redhat.com>
14359 PR rtl-optimization/85342
14360 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
14361 a bool scalar var inside of the loop instead. Don't try to update
14362 recog_data.operand after failed apply_change_group.
14364 2018-04-12 Tom de Vries <tom@codesourcery.com>
14367 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
14368 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
14369 array with flexible array member as array without given dimension.
14370 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
14371 argument for undefined param to true.
14373 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
14376 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
14377 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
14378 from PowerPC section.
14379 * config/rs6000/sysv4.opt (mcall-): Improve help text.
14380 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
14381 help text that is too long.
14382 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
14383 help text that is too long.
14384 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
14385 help text that is too long.
14387 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
14389 * config/alpha/alpha.md (stack_probe_internal): Rename
14390 from "probe_stack". Update all callers.
14392 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
14394 PR rtl-optimization/84566
14395 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
14396 sched_macro_fuse_insns.
14398 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
14401 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
14402 (compute_block_dependences): ... from here.
14404 2018-04-11 Jakub Jelinek <jakub@redhat.com>
14406 PR tree-optimization/85331
14407 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
14408 from int to HOST_WIDE_INT.
14410 2018-04-11 Martin Jambor <mjambor@suse.cz>
14413 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
14414 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
14415 not the same as the source val.
14416 (cgraph_edge_brings_value_p): New parameter.
14417 (gather_edges_for_value): Pass destination value to
14418 cgraph_edge_brings_value_p.
14419 (perhaps_add_new_callers): Likewise.
14420 (get_info_about_necessary_edges): Likewise and exclude values brought
14421 only by self-recursive edges.
14422 (create_specialized_node): Redirect only clones of self-calling edges.
14423 (+self_recursive_pass_through_p): New function.
14424 (find_more_scalar_values_for_callers_subset): Use it.
14425 (find_aggregate_values_for_callers_subset): Likewise.
14426 (known_aggs_to_agg_replacement_list): Removed.
14427 (decide_whether_version_node): Re-calculate known constants for all
14428 remaining context clones.
14430 2018-04-11 Richard Biener <rguenther@suse.de>
14433 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
14434 from early DWARF output.
14435 (dwarf2out_early_finish): Output line info unconditionally into
14436 early DWARF and add reference to it.
14438 2018-04-11 Jakub Jelinek <jakub@redhat.com>
14441 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
14442 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
14443 other than V2DFmode using iptr mode attribute.
14444 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
14446 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
14448 PR rtl-optimization/84659
14449 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
14451 2018-04-11 Jakub Jelinek <jakub@redhat.com>
14454 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
14456 (output_loc_list): Pass address of a dummy size variable even in the
14457 locview handling loop.
14458 (index_location_lists): Add comment on why skip_loc_list_entry can't
14461 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
14464 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
14467 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
14470 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
14471 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
14472 and -mstring-compare-inline-limit.
14474 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
14477 * config/rs6000/rs6000.md (allocate_stack): Put the residual size
14478 for stack clash protection in a register whenever we need it to be in
14481 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
14483 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
14484 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
14486 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
14489 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
14491 (mlong-double-): Ditto.
14492 * config/rs6000/sysv4.opt (msdata=): Ditto.
14493 (mtls-size=): Ditto.
14495 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
14497 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
14498 erroneous entries for
14499 "vector int vec_ldl (int, long int *)", and
14500 "vector unsigned int vec_ldl (int, unsigned long int *)".
14501 Add comments and entries for
14502 "vector bool char vec_ldl (int, bool char *)",
14503 "vector bool short vec_ldl (int, bool short *)",
14504 "vector bool int vec_ldl (int, bool int *)",
14505 "vector bool long long vec_ldl (int, bool long long *)",
14506 "vector pixel vec_ldl (int, pixel *)",
14507 "vector long long vec_ldl (int, long long *)",
14508 "vector unsigned long long vec_ldl (int, unsigned long long *)".
14509 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
14510 type tree bool_long_long_type_node and correct definition of
14511 bool_V2DI_type_node to make reference to this new type tree.
14512 (rs6000_mangle_type): Replace erroneous reference to
14513 bool_long_type_node with bool_long_long_type_node.
14514 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
14515 comments to emphasize sign distinctions for char and int types and
14516 replace RS6000_BTI_bool_long constant with
14517 RS6000_BTI_bool_long_long constant. Also add comment to restrict
14518 use of RS6000_BTI_pixel.
14519 (bool_long_type_node): Remove this macro definition.
14520 (bool_long_long_type_node): New macro definition
14522 2018-04-10 Jakub Jelinek <jakub@redhat.com>
14524 PR rtl-optimization/85300
14525 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
14526 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
14527 simplify_unary_operation fails.
14529 2018-04-10 Martin Liska <mliska@suse.cz>
14531 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
14532 cgraph_edge and ipa_ref.
14534 2018-04-10 Jakub Jelinek <jakub@redhat.com>
14538 * config/i386/sse.md
14539 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
14540 computation of the VEC_MERGE selector from mask.
14541 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
14542 Fix decoding of the VEC_MERGE selector into mask.
14544 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
14546 PR tree-optimization/85286
14547 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
14549 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
14551 * final.c (final_1): Set insn_last_address as well as
14552 insn_current_address.
14554 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14557 * explow.c (emit_stack_probe): Call validize_mem on memory location
14558 before passing it to gen_probe_stack. Create address operand and
14559 legitimize it for the probe_stack_address case.
14561 2018-04-09 Jan Hubicka <jh@suse.cz>
14564 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
14565 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
14566 * tree.c (free_lang_data_in_type): Fix handling of binfos;
14568 (free_lang_data): Rebuild type inheritance graph.
14570 2018-04-09 Martin Sebor <msebor@redhat.com>
14572 * invoke.texi (-finline-small-functions): Mention other optimization
14574 (-findirect-inlining, -fpartial-inlining): Same.
14575 (-finline-functions-called-once): Same.
14576 (-freorder-blocks-and-partition): Same.
14578 2018-04-09 Jan Hubicka <jh@suse.cz>
14581 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
14582 jumps; choose last target that matches the criteria (i.e.
14583 no partition changes for non-crossing jumps).
14584 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
14585 support for redirecting crossing jumps to non-crossing.
14587 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
14589 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
14590 also for naked functions.
14592 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
14594 * config/arc/arc.md (add_shift): New pattern.
14595 (add_shift2): Likewise.
14596 (sub_shift): Likewise.
14597 (sub_shift_cmp0_noout): Likewise.
14598 (compare_si_ashiftsi): Likewise.
14599 (xbfu_cmp0_noout): New combine pattern.
14600 (xbfu_cmp0"): Likewise.
14601 (movsi_set_cc_insn): Place the predicable variant first.
14602 (commutative_binary_cmp0_noout): Remove clobber.
14603 (commutative_binary_cmp0): New pattern.
14604 (noncommutative_binary_cmp0): Likewise.
14605 (noncommutative_binary_cmp0_noout): Likewise.
14606 (noncommutative_binary_comparison_result_used): Removed.
14607 (rsub_cmp0): New pattern.
14608 (rsub_cmp0_noout): Likewise.
14609 (extzvsi): Changed, keep only meaningful variants.
14610 (SQH, SEZ): New iterators.
14611 (SQH_postfix): New mode attribute.
14612 (SEZ_prefix): New code attribute.
14613 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
14614 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
14615 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
14616 of numerical value.
14617 (noncommutative_operator): Check the availability of barrel
14620 2018-04-09 Richard Biener <rguenther@suse.de>
14622 PR tree-optimization/85284
14623 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
14624 Only use the niter constraining form of simple_iv when the exit
14625 is always executed.
14627 2018-04-09 Tom de Vries <tom@codesourcery.com>
14630 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
14631 (define_expand "*memory_barrier"): New define_expand.
14632 (define_insn "memory_barrier"): New insn.
14634 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
14636 PR rtl-optimization/80463
14637 PR rtl-optimization/83972
14638 PR rtl-optimization/83480
14640 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
14641 correct producer for the insn.
14642 (tidy_control_flow): Fixup seqnos in case of debug insns.
14644 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
14646 PR rtl-optimization/83913
14648 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
14649 different sched-times when merging exprs.
14651 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
14653 PR rtl-optimization/83962
14655 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
14656 tidy_fallthru_edge and tidy_control_flow.
14658 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
14660 PR rtl-optimization/83530
14662 * sel-sched.c (force_next_insn): New global variable.
14663 (remove_insn_for_debug): When force_next_insn is true, also leave only
14664 next insn in the ready list.
14665 (sel_sched_region): When the region wasn't scheduled, make another pass
14666 over it with force_next_insn set to 1.
14668 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
14670 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
14672 * config/nds32/constants.md (unspec_volatile_element): Add enum values
14673 for interrupt control.
14674 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
14675 functions for interrupt control.
14676 * config/nds32/nds32-intrinsic.md: Likewise.
14677 * config/nds32/nds32_intrinsic.h: Likewise.
14678 * config/nds32/nds32.h (nds32_builtins): Likewise.
14680 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
14682 * config/nds32/nds32.c (nds32_init_machine_status,
14683 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
14684 strict_aligned_p field.
14685 (nds32_expand_to_rtl_hook): New function.
14686 (TARGET_EXPAND_TO_RTL_HOOK): Define.
14687 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
14689 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
14690 Chung-Ju Wu <jasonwucj@gmail.com>
14692 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
14693 * config/nds32/nds32-n7.md: New file.
14694 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
14695 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
14697 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
14698 * config/nds32/nds32.md (pipeline_model): Add n7.
14699 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
14700 * config/nds32/pipelines.md: Include n7 settings.
14702 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
14703 Chung-Ju Wu <jasonwucj@gmail.com>
14705 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
14706 * config/nds32/nds32-e8.md: New file.
14707 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
14708 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
14710 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
14711 * config/nds32/nds32.md (pipeline_model): Add e8.
14712 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
14713 * config/nds32/pipelines.md: Include e8 settings.
14715 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
14716 Chung-Ju Wu <jasonwucj@gmail.com>
14718 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
14719 * config/nds32/nds32-n8.md: New file.
14720 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
14721 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
14723 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
14724 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
14725 * config/nds32/nds32.md (pipeline_model): Add n8.
14726 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
14727 * config/nds32/pipelines.md: Include n8 settings.
14729 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
14730 Chung-Ju Wu <jasonwucj@gmail.com>
14732 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
14733 * config/nds32/nds32-n9-2r1w.md: New file.
14734 * config/nds32/nds32-n9-3r2w.md: New file.
14735 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
14736 nds32_register_ports): New or modify for cpu n9.
14737 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
14739 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
14740 * config/nds32/nds32-utils.c: New file.
14741 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
14742 TARGET_MUL_SLOW): Define.
14743 * config/nds32/nds32.md (pipeline_model): New attribute.
14744 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
14745 New options that support cpu n9.
14746 * config/nds32/pipelines.md: Include n9 settings.
14747 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
14749 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
14751 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
14752 information if necessary.
14753 (output_cond_branch_compare_zero): Likewise.
14754 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
14755 (nds32_target_alignment): Refine for alignment.
14756 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
14757 (FUNCTION_BOUNDARY): Modify.
14758 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
14760 * config/nds32/nds32.opt (malways-align, malign-functions): New.
14762 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
14764 * config/nds32/constants.md (unspec_volatile_element): Add values for
14765 TLB operation and data prefetch.
14766 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
14767 functions for TLB operation and data prefetch.
14768 * config/nds32/nds32-intrinsic.md: Likewise.
14769 * config/nds32/nds32_intrinsic.h: Likewise.
14770 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
14771 (nds32_print_operand): Likewise.
14772 * config/nds32/nds32.h (nds32_builtins): Likewise.
14774 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
14775 Andrew Pinski <pinsika@gcc.gnu.org>
14777 PR middle-end/82976
14778 * match.pd: Use constant_boolean_node of correct type instead of
14779 boolean_true_node or boolean_false_node for simplifying
14780 pointer comparisons to zero.
14782 2018-04-07 Jakub Jelinek <jakub@redhat.com>
14784 PR tree-optimization/80021
14785 * tree.c (verify_type_variant): Make error call in verify_variant_match
14786 translatable and remove final full stop.
14788 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
14790 * config/nds32/constants.md (unspec_volatile_element): Add
14791 UNSPEC_VOLATILE_EH_RETURN.
14792 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
14793 nds32_output_stack_pop): Support dwarf exception handling process.
14794 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
14795 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
14796 exception handling process.
14797 (nds32_compute_stack_frame): Likewise.
14798 (nds32_return_addr_rtx): Likewise.
14799 (nds32_initial_elimination_offset): Likewise.
14800 (nds32_expand_prologue): Likewise.
14801 (nds32_expand_epilogue): Likewise.
14802 (nds32_dynamic_chain_address): New function.
14803 * config/nds32/nds32.h (machine_function): Add fields for dwarf
14804 exception handling.
14805 (DYNAMIC_CHAIN_ADDRESS): Define.
14806 (EH_RETURN_DATA_REGNO): Define.
14807 (EH_RETURN_STACKADJ_RTX): Define.
14808 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
14809 patterns for dwarf exception handling.
14811 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
14813 * config/nds32/nds32.h: Clean up obsolete macros.
14815 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
14817 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
14818 Add enum values for particular instructions.
14819 * config/nds32/nds32-intrinsic.c: Implementation of expanding
14820 particular intrinsic functions.
14821 * config/nds32/nds32-intrinsic.md: Likewise.
14822 * config/nds32/nds32_intrinsic.h: Likewise.
14823 * config/nds32/nds32.h (nds32_builtins): Likewise.
14824 * config/nds32/nds32.md (type): Add pbsad and pbsada.
14825 (btst, ave): New patterns for particular instructions.
14827 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
14829 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
14830 Add enum values for atomic load/store and memory sync.
14831 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
14833 * config/nds32/nds32-intrinsic.md: Likewise.
14834 * config/nds32/nds32_intrinsic.h: Likewise.
14835 * config/nds32/nds32.h (nds32_builtins): Likewise.
14837 2018-04-07 Jakub Jelinek <jakub@redhat.com>
14839 PR tree-optimization/85257
14840 * fold-const.c (native_encode_vector): If not all elts could fit
14841 and off is -1, return 0 rather than offset.
14842 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
14843 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
14844 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
14845 adjust buffer in native_interpret_expr call.
14847 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
14849 * config/nds32/constants.md (unspec_volatile_element): Add cache
14850 control enum values.
14851 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
14852 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
14853 * config/nds32/nds32.c (nds32_cctl_names): New.
14854 (nds32_print_operand): Handle cache control register names.
14855 * config/nds32/nds32.h (nds32_builtins): New enum values.
14856 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
14858 * config/nds32/nds32.md (type): Add mmu.
14859 * config/nds32/pipelines.md (simple_insn): Add mmu.
14861 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
14863 * config/nds32/nds32.md (type): Remove call.
14864 * config/nds32/pipelines.md (simple_insn): Likewise.
14866 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
14868 * config/nds32/constants.md (unspec_volatile_element): Add
14869 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
14870 UNSPEC_VOLATILE_FMFCFG.
14871 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
14872 description for fmfcfg and fmfcsr.
14873 (bdesc_1arg): Add fmtcsr.
14874 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
14875 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
14876 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
14877 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
14878 unspec_fmfcfg): New patterns.
14879 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
14880 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
14881 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
14882 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
14883 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
14884 __nds32__fmfcfg): Define.
14886 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
14888 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
14889 intrinsic register names.
14890 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
14891 intrinsic register enum values and macros.
14893 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
14895 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
14896 for load/store addressing form.
14897 (nds32_print_operand_address): Likewise.
14899 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
14902 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
14903 based on LABEL_REF. Remove useless assertion.
14904 (pic_address_needs_scratch): Fix formatting.
14905 (sparc_legitimize_pic_address): Minor tweaks.
14906 (sparc_delegitimize_address): Adjust assertion accordingly.
14907 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
14908 into symbolic_operand.
14909 (movsi_high_pic_label_ref): Likewise.
14910 (movsi_lo_sum_pic_label_ref): Likewise.
14911 (movdi_pic_label_ref): Likewise.
14912 (movdi_high_pic_label_ref): Likewise.
14913 (movdi_lo_sum_pic_label_ref): Likewise.
14915 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
14917 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
14918 custom LIB_SPEC setup.
14920 2018-04-06 Ruslan Bukin <br@bsdpad.com>
14921 Kito Cheng <kito.cheng@gmail.com>
14923 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
14924 * config/riscv/freebsd.h: New.
14926 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
14928 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
14929 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
14932 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
14933 Kito Cheng <kito.cheng@gmail.com>
14935 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
14936 nds32_output_call, nds32_symbol_binds_local_p): New functions.
14937 * config/nds32/nds32-protos.h (nds32_output_call,
14938 nds32_output_return): Declare.
14939 * config/nds32/nds32.md: Refine all the call and return patterns.
14941 2018-04-06 Jakub Jelinek <jakub@redhat.com>
14944 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
14945 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
14947 PR rtl-optimization/84872
14948 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
14949 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
14950 EDGE_CROSSING edge.
14952 2018-04-06 Tamar Christina <tamar.christina@arm.com>
14954 * expr.c (copy_blkmode_to_reg): Revert 254862.
14955 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
14957 2018-04-06 Richard Biener <rguenther@suse.de>
14959 PR middle-end/85244
14960 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
14961 after seeing a component reference with an adjacent field. Treat
14962 refs to arrays at struct end of external decls similar to
14963 refs to unconstrained commons.
14965 2018-04-06 Jakub Jelinek <jakub@redhat.com>
14968 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
14969 look through SAVE_EXPRs with non-side-effects argument. Adjust
14971 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
14974 2018-04-06 Richard Biener <rguenther@suse.de>
14976 PR middle-end/85180
14977 * alias.c (find_base_term): New wrapper around find_base_term
14978 unwinding CSELIB_VAL_PTR changes.
14979 (find_base_term): Do not restore CSELIB_VAL_PTR during the
14982 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
14984 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
14986 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
14987 constant definitions.
14988 ("nop"): lr 0,0 -> nopr r0
14989 ("nop_lr0", "nop_lr1"): New insn definitions.
14991 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
14993 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
14994 NDS32_V3PUSH_AVAILABLE_P macro.
14996 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
14997 Chung-Ju Wu <jasonwucj@gmail.com>
14999 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
15000 (nds32*-*-*): Add float and fpu_config into supported_defaults.
15001 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
15002 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
15003 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
15004 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
15005 * config/nds32/constraints.md: New constraints and checking for hard
15006 float configuration.
15007 * config/nds32/iterators.md: New mode iterator and attribute for hard
15008 float configuration.
15009 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
15011 * config/nds32/nds32-fpu.md: New file.
15012 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
15013 deal with hard float code generation.
15014 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
15016 (abi_type, float_reg_number): New enum type.
15017 * config/nds32/nds32-predicates.c: New predicates for hard float.
15018 * config/nds32/nds32-protos.h: Declare functions for hard float.
15019 * config/nds32/nds32.c: Implementation for hard float configuration.
15020 * config/nds32/nds32.h: Definitions for hard float configuration.
15021 * config/nds32/nds32.md: Include hard float machine description and
15022 modify patterns for hard float configuration.
15023 * config/nds32/nds32.opt: New options for hard float configuration.
15024 * config/nds32/predicates.md: New predicates for hard float
15027 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
15029 * common/config/nds32/nds32-common.c
15030 (nds32_option_optimization_table): Enable -mreleax-hint by default.
15032 2018-04-05 Jakub Jelinek <jakub@redhat.com>
15034 PR middle-end/85195
15035 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
15036 CONSTRUCTOR_ELT (ctor, ...)->value.
15038 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
15041 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
15043 2018-04-05 Tom de Vries <tom@codesourcery.com>
15046 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
15049 2018-04-05 Shiva Chen <shiva0217@gmail.com>
15050 Kito Cheng <kito.cheng@gmail.com>
15052 * config/nds32/constraints.md (U33): Fine-tune checking condition.
15053 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
15054 * config/nds32/nds32.h (nds32_16bit_address_type): Add
15055 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
15057 2018-04-05 Shiva Chen <shiva0217@gmail.com>
15058 Kito Cheng <kito.cheng@gmail.com>
15060 * config/nds32/constraints.md (Ufe): New memory constraint.
15061 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
15062 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
15063 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
15065 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
15066 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
15068 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
15070 * config/nds32/nds32.md: Use optimize_size in the condition for
15071 alu-shift instructions.
15073 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
15075 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
15077 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
15079 * config/nds32/nds32.md (negsi2): Refine pattern.
15081 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
15082 Chung-Ju Wu <jasonwucj@gmail.com>
15084 * config/nds32/iterators.md (shift_rotate): New code iterator.
15085 (shift): New code attribute.
15086 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
15087 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
15088 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
15089 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
15090 bit-wise operations.
15091 (andsi3, *andsi3): Ditto.
15092 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
15093 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
15094 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
15095 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
15096 nds32_ior_operand, nds32_xor_operand): New predicates.
15098 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
15100 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
15101 (addsi3, subsi3): ... this.
15103 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
15105 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
15107 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
15109 * config/nds32/nds32.md: Adjust indention.
15111 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
15113 * config/nds32/nds32.md (feature): New attribute.
15115 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
15117 * config/nds32/nds32.md (subtype): New attribute.
15119 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
15122 * config/arm/arm-builtins.c (arm_expand_builtin): Change
15123 expansion to perform a bitwise AND of the argument followed by a
15124 boolean negation of the result.
15126 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
15128 PR rtl-optimization/84878
15129 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
15130 the basic block. Assert the use reference is not artificial and that
15131 it has an associated insn.
15133 2018-04-04 Michael Matz <matz@suse.de>
15135 * builtins.c (compute_objsize): Pass correct operand
15136 to array_at_struct_end_p.
15138 2018-04-04 Richard Biener <rguenther@suse.de>
15141 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
15142 from contexts for DINFO_LEVEL_TERSE and below.
15144 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
15146 * config/nds32/nds32-doubleword.md (move_<mode>): Require
15147 resiter_operand condition.
15148 * config/nds32/nds32.md (*move<mode>): Ditto.
15150 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
15151 Monk Chiang <sh.chiang04@gmail.com>
15153 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
15155 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
15157 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
15159 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
15160 Kito Cheng <kito.cheng@gmail.com>
15162 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
15163 nds32_cond_code_str, output_cond_branch,
15164 output_cond_branch_compare_zero, nds32_expand_cbranch,
15165 nds32_expand_cstore, nds32_expand_movcc,
15166 nds32_output_cbranchsi4_equality_zero,
15167 nds32_output_cbranchsi4_equality_reg,
15168 nds32_output_cbranchsi4_equality_reg_or_const_int,
15169 nds32_output_cbranchsi4_greater_less_zero: New functions.
15170 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
15171 nds32_expand_cstore, nds32_expand_movcc,
15172 nds32_output_cbranchsi4_equality_zero,
15173 nds32_output_cbranchsi4_equality_reg,
15174 nds32_output_cbranchsi4_equality_reg_or_const_int,
15175 nds32_output_cbranchsi4_greater_less_zero): Declare.
15176 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
15177 nds32_rimm11s_operand): New predicates.
15178 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
15179 * config/nds32/nds32.md: Rewrite all the branch and conditional move
15182 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
15184 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
15185 * config/nds32/nds32.md: Ditto.
15186 * config/nds32/pipelines.md: Ditto.
15188 2018-04-04 Richard Biener <rguenther@suse.de>
15190 PR tree-optimization/85168
15191 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
15192 propagating abnormals.
15194 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
15196 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
15198 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
15199 Kito Cheng <kito.cheng@gmail.com>
15201 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
15202 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
15203 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
15204 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
15205 * config/nds32/nds32.md (sibcall_internal): New.
15206 (sibcall_register): Remove.
15207 (sibcall_immediate): Remove.
15208 (sibcall_value_internal): New.
15209 (sibcall_value_register): Remove.
15210 (sibcall_value_immediate): Remove.
15211 * config/nds32/predicates.md (nds32_general_register_operand): New.
15212 (nds32_call_address_operand): New.
15214 2018-04-03 Jakub Jelinek <jakub@redhat.com>
15216 PR rtl-optimization/85167
15217 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
15218 bb_defs if *split_p, instead preinitialize it to NULL.
15220 PR tree-optimization/85156
15221 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
15222 evaluating the argument multiple times.
15224 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
15226 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
15228 (_mm_cvtpd_ps): Likewise.
15229 (_mm_cvttpd_epi32): Likewise.
15230 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
15231 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
15232 vector, pixel, and bool following altivec.h include.
15234 2018-04-03 Martin Sebor <msebor@redhat.com>
15236 * doc/extend.texi (Common Function Attributes): Clarify.
15237 (const attribute): Likewise.
15238 (pure attribute): Likewise.
15240 2018-04-03 Jakub Jelinek <jakub@redhat.com>
15243 * config/i386/i386.c (ix86_expand_vector_set): Use
15244 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
15246 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
15248 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
15249 instructions when changing rounding bits to preserve precision bits
15250 in the x87 control word.
15252 2018-04-03 Martin Liska <mliska@suse.cz>
15254 PR tree-optimization/82491
15255 * rtl.h (strip_offset_and_add): Replace += suboffset with
15256 poly_uint64 () + suboffset.
15258 2018-03-29 Martin Liska <mliska@suse.cz>
15259 Martin Jambor <mjambor@suse.cz>
15262 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
15263 param_type is not an integral or pointer type.
15265 2018-04-03 Richard Biener <rguenther@suse.de>
15267 * sese.h (recompute_all_dominators): Remove.
15269 2018-04-02 Martin Sebor <msebor@redhat.com>
15271 * doc/invoke.texi (-Wrestrict): Fix typos.
15273 2018-04-02 Jim Wilson <jimw@sifive.com>
15275 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
15276 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
15277 (<optab>di3, <optab>si3_extend): Likewise.
15278 (<optab>si3_mask, <optab>si3_mask_1): New.
15279 (<optab>di3_mask, <optab>di3_mask_1): New.
15280 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
15281 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
15282 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
15284 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
15286 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
15289 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
15291 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
15292 (nds32_canonicalize_comparison): New function.
15294 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
15295 Kito Cheng <kito.cheng@gmail.com>
15296 Kuan-Lin Chen <kuanlinchentw@gmail.com>
15298 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
15299 * config/nds32/constants.md (unspec_volatile_element): Add
15300 UNSPEC_VOLATILE_RELAX_GROUP.
15301 * config/nds32/nds32-relax-opt.c: New file.
15302 * config/nds32/nds32-predicates.c
15303 (nds32_symbol_load_store_p): New function.
15304 * config/nds32/nds32-protos.h
15305 (nds32_symbol_load_store_p): Declare function.
15306 (make_pass_nds32_relax_opt): Declare new rtl pass function.
15307 * config/nds32/nds32.c
15308 (nds32_register_pass): New function to register pass.
15309 (nds32_register_passes): New function to register passes.
15310 * config/nds32/nds32.md (relax_group): New pattern.
15311 * config/nds32/nds32.opt (mrelax-hint): New option.
15312 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
15314 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
15316 * config/nds32/t-nds32: Modify files dependency.
15318 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
15320 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
15321 (PROFILE_HOOK): Define its implementation.
15323 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
15325 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
15326 type and 32-bit size.
15328 2018-04-01 Jakub Jelinek <jakub@redhat.com>
15330 PR middle-end/85090
15331 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
15332 (V_128_256): New mode iterator.
15333 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
15334 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
15335 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
15337 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
15338 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
15340 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
15343 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
15344 NaN inputs correctly.
15346 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
15349 * config/rs6000/vsx.md (??r): New mode attribute.
15350 (*vsx_mov<mode>_64bit): Use it.
15351 (*vsx_mov<mode>_32bit): Likewise.
15353 2018-03-30 Martin Sebor <msebor@redhat.com>
15355 PR tree-optimization/84818
15356 * builtins.c (check_access): Use warning_n.
15358 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15361 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
15363 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
15366 2018-03-30 Julia Koval <julia.koval@intel.com>
15369 * x86-tune.def (movx, partial_reg_dependency): Enable for
15372 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
15374 PR inline-asm/84985
15375 * lra-constraints.c (process_alt_operands): Move setting
15376 this_alternative_matches below.
15378 2018-03-29 Martin Liska <mliska@suse.cz>
15381 * doc/invoke.texi: Document how LTO works with debug info.
15382 Describe auto-load support of binutils. Mention 'x86-64'
15383 as valid option value of -march option.
15385 2018-03-29 Jakub Jelinek <jakub@redhat.com>
15387 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
15390 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
15391 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
15392 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
15395 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
15398 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
15399 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
15400 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
15401 for RS6000_BTM_POWERPC64.
15402 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
15403 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
15404 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
15407 (DIVDEU): Likewise.
15409 2018-03-28 Carl Love <cel@us.ibm.com>
15412 2017-09-27 Carl Love <cel@us.ibm.com>
15414 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
15415 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
15416 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
15419 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15421 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
15422 instead of __vector bool.
15423 (_mm_max_pu8): Likewise.
15424 (_mm_min_pi16): Likewise.
15426 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
15429 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
15430 (DIVWEUO): Likewise.
15431 (DIVDEO): Likewise.
15432 (DIVDEUO): Likewise.
15433 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
15434 DIVWEUO and DIVDEUO.
15435 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
15436 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
15437 (div_extend): Likewise.
15438 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
15440 (__builtin_divweuo): Likewise.
15441 (__builtin_divdeo): Likewise.
15442 (__builtin_divdeuo): Likewise.
15444 2018-03-28 Jakub Jelinek <jakub@redhat.com>
15447 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
15448 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
15450 PR tree-optimization/82004
15451 * gimple-match-head.c (optimize_pow_to_exp): New function.
15452 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
15453 Don't fold to exp if optimize_pow_to_exp is false.
15455 2018-03-28 Martin Liska <mliska@suse.cz>
15458 * calls.c (initialize_argument_information): Fix trailing space.
15459 * common.opt: Fix typo and provide better explanation for
15460 -fsanitize-coverage option.
15461 * config/i386/i386.opt: Fix typo.
15463 2018-03-28 Jakub Jelinek <jakub@redhat.com>
15464 Martin Liska <mliska@suse.cz>
15467 * gimplify.c (asan_poison_variable): Don't do the check for
15468 gimplify_omp_ctxp here.
15469 (gimplify_decl_expr): Do it here.
15470 (gimplify_target_expr): Likewise.
15472 2018-03-28 Martin Liska <mliska@suse.cz>
15475 * config/i386/i386.c (ix86_function_arg_advance): Do not call
15476 chkp_type_bounds_count if MPX is not enabled.
15478 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
15480 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
15482 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
15485 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
15486 function to create the function decl for complex long double
15487 multiply and divide for -mabi=ieeelongdouble.
15488 (init_float128_ieee): Call it.
15490 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
15493 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
15494 -fcf-protection=branch -mibt.
15495 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
15497 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
15500 * config/arm/arm.c (arm_valid_symbolic_address): Handle
15501 arm_word_relocations.
15503 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
15506 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
15507 extern array declarations.
15509 2018-03-27 Richard Biener <rguenther@suse.de>
15511 PR middle-end/84067
15512 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
15513 explicit single_use checks.
15515 2018-03-27 Richard Biener <rguenther@suse.de>
15517 PR tree-optimization/85082
15518 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
15521 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
15523 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
15524 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
15525 Turn on fasynchronous-unwind-tables and funwind-tables.
15527 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
15530 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
15531 (*bmi_blsr_<mode>_ccz): Ditto.
15533 2018-03-26 Tom de Vries <tom@codesourcery.com>
15535 PR tree-optimization/85063
15536 * omp-general.c (offloading_function_p): New function. Factor out
15538 * omp-offload.c (pass_omp_target_link::gate): ... here.
15539 * omp-general.h (offloading_function_p): Declare.
15540 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
15541 with attribute omp declare target for offloading functions.
15543 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
15545 PR tree-optimization/84005
15546 * tree-data-ref.h (get_base_for_alignment): Declare.
15547 * tree-data-ref.c (get_base_for_alignment_1): New function.
15548 (get_base_for_alignment): Likewise.
15549 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
15550 get_base_for_alignment to find a suitable base object, instead
15551 of always using drb->base_address.
15553 2018-03-23 Jakub Jelinek <jakub@redhat.com>
15555 PR inline-asm/85022
15556 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
15557 known size by default.
15559 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
15561 PR inline-asm/85030
15562 * lra-constraints.c (process_alt_operands): Don't match BLKmode
15563 and non BLKmode operands.
15565 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15568 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
15569 Clean up attributes.
15571 2018-03-23 Richard Biener <rguenther@suse.de>
15574 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
15575 we are going to emit early debug for LTO.
15577 2018-03-23 Jakub Jelinek <jakub@redhat.com>
15579 PR inline-asm/85034
15580 * function.c (match_asm_constraints_1): Don't optimize if input
15581 doesn't satisfy general_operand predicate for output's mode.
15583 PR inline-asm/85022
15584 * alias.c (write_dependence_p): Don't require for x_canonicalized
15585 non-VOIDmode if x has VOIDmode.
15588 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
15589 just don't try to optimize it rather than assert it never happens.
15591 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
15593 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
15594 macro expansions for definition of ST_INTERNAL_<mode> and
15595 LD_INTERNAL_<mode> builtins.
15596 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
15598 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
15600 (altivec_expand_st_builtin): Likewise.
15601 (altivec_expand_builtin): Remove calls to deleted functions.
15602 (rs6000_address_for_altivec): Delete this function.
15603 * config/rs6000/vector.md: Remove expands for
15604 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
15606 2018-03-22 Sudakshina Das <sudi.das@arm.com>
15609 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
15610 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
15611 re-computing once computed.
15612 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
15613 (arm_init_machine_status): Initialize
15614 machine->static_chain_stack_bytes.
15616 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
15619 * doc/extend.texi: Add four new prototypes for vec_ld.
15620 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
15621 definitions for more logical presentation.
15622 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
15623 entries for V1TI variants of __builtin_altivec_ld builtin.
15624 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
15625 handling of V1TI variant of LVX icode pattern.
15626 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
15627 (rs6000_gimple_fold_builtin): Likewise.
15628 (altivec_init_builtins): Add code to define
15629 __builtin_altivec_lvx_v1ti function.
15631 2018-03-22 Jakub Jelinek <jakub@redhat.com>
15633 PR inline-asm/84941
15634 * function.c (match_asm_constraints_1): Don't do the optimization
15635 if input isn't a REG, SUBREG, MEM or constant.
15637 2018-03-22 Tom de Vries <tom@codesourcery.com>
15639 PR tree-optimization/84956
15640 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
15641 bb_has_abnormal_pred.
15643 2018-03-22 Jakub Jelinek <jakub@redhat.com>
15646 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
15647 DECL_INITIAL (decl) to decl at the end.
15648 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
15649 adjust the comment.
15651 2018-03-21 Joseph Myers <joseph@codesourcery.com>
15653 * doc/extend.texi (__builtin_tgmath): Document when complex
15654 integer types are treated as _Complex _Float64.
15656 2018-03-21 Tom de Vries <tom@codesourcery.com>
15658 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
15660 2018-03-21 Jakub Jelinek <jakub@redhat.com>
15662 PR tree-optimization/84960
15663 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
15664 if it is ENTRY block, move them into single succ of ENTRY in that case.
15666 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
15668 PR tree-optimization/84811
15669 * poly-int.h (poly_span_traits): Remove the T3 parameter and
15670 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
15671 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
15672 (known_subrange_p): Update accordingly. Cast each value involved
15673 in the size comparison, rather than casting the result of the
15676 2018-03-21 Jakub Jelinek <jakub@redhat.com>
15678 PR tree-optimization/84982
15679 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
15680 by flipping the least significant bit rather than all bits from
15681 bitpos to bitpos + bitsize - 1.
15683 2018-03-21 Nathan Sidwell <nathan@acm.org>
15685 * doc/extend.texi (Deprecated Features): Remove mention of
15686 long-deleted deprecations.
15688 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15691 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
15692 * configure: Regenerate.
15694 2018-03-21 Tom de Vries <tom@codesourcery.com>
15696 PR tree-optimization/83126
15697 * tree-parloops.c (num_phis): New function.
15698 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
15700 2018-03-21 Nathan Sidwell <nathan@acm.org>
15702 * doc/extend.texi (Deprecated Features): Update deprecated flags,
15703 mention anon-struct/union members and trailing attributes.
15705 2018-03-21 Bin Cheng <bin.cheng@arm.com>
15707 PR tree-optimization/84969
15708 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
15709 builtin memset partitions if they set different rhs values.
15711 2018-03-21 Jakub Jelinek <jakub@redhat.com>
15713 PR rtl-optimization/84989
15714 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
15715 VEC_DUPLICATE with scalar result mode.
15717 2018-03-21 Martin Liska <mliska@suse.cz>
15720 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
15721 not intended return statement.
15723 2018-03-21 Martin Liska <mliska@suse.cz>
15726 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
15727 (chkp_find_bound_slots_1): Limit number of iterations.
15729 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
15732 * Minor grammar fixes for x86 options.
15734 2018-03-20 Jakub Jelinek <jakub@redhat.com>
15737 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
15738 holding REG_CFA_RESTORE notes, instead turn them into a USE.
15740 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
15743 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
15744 (altivec_lvx_<mode>_1op): Likewise.
15745 (altivec_stvx_<mode>_2op): Likewise.
15746 (altivec_stvx_<mode>_1op): Likewise.
15747 (altivec_lvx_<VM2:mode>): New define_expand.
15748 (altivec_stvx_<VM2:mode>): Likewise.
15749 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
15750 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
15751 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
15752 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
15753 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
15754 (rs6000_gen_lvx): Likewise.
15755 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
15756 (altivec_expand_stv_builtin): Likewise.
15757 (altivec_expand_builtin): Likewise.
15758 * config/rs6000/vector.md: Likewise.
15760 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15763 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
15766 2018-03-20 Richard Biener <rguenther@suse.de>
15769 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
15770 sign-conversions as zero, fall back to standard scalar_stmt
15773 2018-03-20 Martin Liska <mliska@suse.cz>
15776 * predict.c (rebuild_frequencies): Handle case when we have
15777 PROFILE_ABSENT, but flag_guess_branch_prob is false.
15779 2018-03-20 Jakub Jelinek <jakub@redhat.com>
15782 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
15783 flag_section_anchors.
15784 * varasm.c (use_blocks_for_decl_p): Remove hack for
15785 dw2_force_const_mem.
15788 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
15790 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
15791 be created, use lowpart_subreg of operands[0] rather than operands[0]
15793 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
15794 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
15795 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
15796 and n constraint instead of aarch64_shift_imm_di and Usd.
15797 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
15798 (*aarch64_<optab>_reg_minus<mode>3): ... this.
15800 2018-03-20 Sudakshina Das <sudi.das@arm.com>
15803 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
15804 to favor GPR over NEON registers.
15805 (<shift>di3_neon): Likewise.
15807 2018-03-20 Tom de Vries <tom@codesourcery.com>
15810 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
15811 (nvptx_process_pars): Emit bar.sync asap and alap.
15813 2018-03-20 Tom de Vries <tom@codesourcery.com>
15816 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
15817 seen_label if seen_label is already set.
15819 2018-03-20 Jakub Jelinek <jakub@redhat.com>
15822 * config/i386/i386.c (fold_builtin_cpu): For features above 31
15823 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
15824 Use 1U instead of 1. Formatting fixes.
15827 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
15828 instead of TREE_TYPE (s1) for the return value.
15830 2018-03-19 Jakub Jelinek <jakub@redhat.com>
15832 PR tree-optimization/84946
15833 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
15834 bitsize + bitsize in poly_uint64 rather than poly_int64.
15837 * dwarf2asm.c: Include fold-const.c.
15838 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
15839 of decl rather than decl itself.
15841 PR rtl-optimization/84643
15842 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
15844 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
15847 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
15848 calling assemble_variable.
15850 2018-03-19 Sudakshina Das <sudi.das@arm.com>
15853 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
15854 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
15856 2018-03-19 Jim Wilson <jimw@sifive.com>
15859 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
15860 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
15861 (riscv_first_stack_step): Likewise.
15862 (riscv_option_override): Use STACK_BOUNDARY instead of
15863 MIN_STACK_BOUNDARY.
15864 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
15865 MIN_STACK_BOUNDARY.
15866 (BIGGEST_ALIGNMENT): Set to 128.
15867 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
15868 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
15871 2018-03-19 Richard Biener <rguenther@suse.de>
15873 PR tree-optimization/84933
15874 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
15875 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
15877 2018-03-19 Richard Biener <rguenther@suse.de>
15879 PR tree-optimization/84859
15880 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
15881 (cond_if_else_store_replacement): Perform sinking operation on
15882 single-store BBs regardless of MAX_STORES_TO_SINK setting.
15883 Generalize what a BB with a single eligible store is.
15885 2018-03-19 Richard Biener <rguenther@suse.de>
15887 PR tree-optimization/84929
15888 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
15889 chrec_is_positive against non-chrec arg.
15891 2018-03-19 Tamar Christina <tamar.christina@arm.com>
15894 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
15896 2018-03-18 Martin Liska <mliska@suse.cz>
15898 PR rtl-optimization/84635
15899 * regrename.c (build_def_use): Use matches_mode only when
15902 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
15904 PR tree-optimization/84913
15905 * tree-vect-loop.c (vectorizable_reduction): Don't try to
15906 vectorize chains of COND_EXPRs.
15908 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
15910 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
15912 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
15914 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
15916 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
15918 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
15920 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
15921 Kito Cheng <kito.cheng@gmail.com>
15923 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
15924 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
15925 (nds32_adjust_reg_alloc_order): New function.
15926 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
15928 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
15930 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
15931 nds32_print_operand, nds32_print_operand_address): Use
15932 HOST_WIDE_INT_PRINT_DEC instead.
15934 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
15936 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
15938 2018-03-17 Jakub Jelinek <jakub@redhat.com>
15941 * config/i386/i386.c (initial_ix86_tune_features,
15942 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
15943 unsigned long long.
15944 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
15945 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
15946 rather than 1u << ix86_tune. Formatting fix.
15947 (ix86_option_override_internal): Change ix86_arch_mask from
15948 unsigned int to unsigned HOST_WIDE_INT, initialize to
15949 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
15950 (ix86_function_specific_restore): Likewise.
15952 2018-03-16 Jakub Jelinek <jakub@redhat.com>
15955 * postreload.c (reload_combine_recognize_pattern): Perform
15956 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
15957 truncate_int_for_mode the result for the destination's mode.
15960 * hsa-gen.c (mem_type_for_type): Fix comment typo.
15961 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
15963 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
15966 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
15969 * lra-assigns.c (lra_split_hard_reg_for): Don't use
15970 regno_allocno_class_array and sorted_pseudos.
15971 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
15972 insns where regno is used.
15974 2018-03-16 Martin Liska <mliska@suse.cz>
15977 * multiple_target.c (create_dispatcher_calls): Redirect
15978 reference in the symbol table.
15980 2018-03-16 Martin Liska <mliska@suse.cz>
15983 * multiple_target.c (create_dispatcher_calls): Redirect also
15986 2018-03-16 Jakub Jelinek <jakub@redhat.com>
15990 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
15991 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
15992 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
15994 2018-03-16 Julia Koval <julia.koval@intel.com>
15996 * doc/invoke.texi (Skylake Server): Add CLWB.
15997 Cannonlake): Remove CLWB.
15999 2018-03-16 Jakub Jelinek <jakub@redhat.com>
16001 PR tree-optimization/84841
16002 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
16004 (FLOAT_ONE_CONST_TYPE): Define.
16005 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
16006 (sort_by_operand_rank): Put entries with higher constant_type last
16007 rather than first to match comments.
16009 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
16011 * config/nios2/nios2.md (movsi_internal): Fix thinko in
16014 2018-03-15 Jakub Jelinek <jakub@redhat.com>
16017 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
16018 check and use address of target always.
16020 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
16023 * config/i386/i386.c (indirect_thunk_needed): Update comments.
16024 (indirect_thunk_bnd_needed): Likewise.
16025 (indirect_thunks_used): Likewise.
16026 (indirect_thunks_bnd_used): Likewise.
16027 (indirect_return_needed): New.
16028 (indirect_return_bnd_needed): Likewise.
16029 (output_indirect_thunk_function): Add a bool argument for
16031 (output_indirect_thunk_function): Don't generate alias for
16032 function return thunk.
16033 (ix86_code_end): Call output_indirect_thunk_function to generate
16034 function return thunks.
16035 (ix86_output_function_return): Set indirect_return_bnd_needed
16036 and indirect_return_needed instead of indirect_thunk_bnd_needed
16037 and indirect_thunk_needed.
16039 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
16041 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
16042 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
16043 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
16045 2018-03-15 David Malcolm <dmalcolm@redhat.com>
16046 Paul Hua <paul.hua.gm@gmail.com>
16049 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
16051 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
16053 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
16054 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
16055 resp. SFmode cases.
16057 2018-03-15 Tamar Christina <tamar.christina@arm.com>
16060 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
16061 instead of GET_MODE_SIZE when comparing Units.
16063 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
16066 * varasm.c (hash_section): Return an unchangeble hash value
16067 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
16068 Return !aarch64_can_use_per_function_literal_pools_p ().
16070 2018-03-15 Jakub Jelinek <jakub@redhat.com>
16073 * optabs.c (emit_conditional_move): Pass address of cmode's copy
16074 rather than address of cmode as last argument to prepare_cmp_insn.
16076 2018-03-15 Julia Koval <julia.koval@intel.com>
16078 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
16079 F_AVX512VNNI, F_AVX512BITALG): New.
16081 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
16084 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
16085 insn for floating-point loads and stores.
16087 2018-03-14 Carl Love <cel@us.ibm.com>
16089 * config/rs6000/rs6000-c.c: Add macro definitions for
16090 ALTIVEC_BUILTIN_VEC_PERMXOR.
16091 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
16092 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
16093 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
16094 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
16096 * config/doc/extend.texi: Add prototypes for vec_permxor.
16098 2018-03-14 David Malcolm <dmalcolm@redhat.com>
16101 * diagnostic-show-locus.c (class layout_point): Convert m_line
16102 from int to linenum_type.
16103 (line_span::comparator): Use linenum "compare" function when
16104 comparing line numbers.
16105 (test_line_span): New function.
16106 (layout_range::contains_point): Convert param "row" from int to
16108 (layout_range::intersects_line_p): Likewise.
16109 (layout::will_show_line_p): Likewise.
16110 (layout::print_source_line): Likewise.
16111 (layout::should_print_annotation_line_p): Likewise.
16112 (layout::print_annotation_line): Likewise.
16113 (layout::print_leading_fixits): Likewise.
16114 (layout::annotation_line_showed_range_p): Likewise.
16115 (struct line_corrections): Likewise for field m_row.
16116 (line_corrections::line_corrections): Likewise for param "row".
16117 (layout::print_trailing_fixits): Likewise.
16118 (layout::get_state_at_point): Likewise.
16119 (layout::get_x_bound_for_row): Likewise.
16120 (layout::print_line): Likewise.
16121 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
16122 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
16123 * input.c (selftest::test_linenum_comparisons): New function.
16124 (selftest::input_c_tests): Call it.
16125 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
16126 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
16127 * selftest.h (ASSERT_GT): New macro.
16128 (ASSERT_GT_AT): New macro.
16129 (ASSERT_LT): New macro.
16130 (ASSERT_LT_AT): New macro.
16132 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
16134 PR rtl-optimization/84780
16135 * combine.c (distribute_links): Don't make a link based on pc_rtx.
16137 2018-03-14 Martin Liska <mliska@suse.cz>
16139 * tree.c (record_node_allocation_statistics): Use
16140 get_stats_node_kind.
16141 (get_stats_node_kind): New function extracted from
16142 record_node_allocation_statistics.
16143 (free_node): Use get_stats_node_kind.
16145 2018-03-14 Richard Biener <rguenther@suse.de>
16147 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
16148 that the value-set of ANTIC_IN doesn't grow.
16151 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
16153 (BB_VISITED_WITH_VISITED_SUCCS): New define.
16154 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
16156 2018-03-14 Julia Koval <julia.koval@intel.com>
16158 * config.gcc (icelake-client, icelake-server): New.
16160 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
16161 (initial_ix86_arch_features): Ditto.
16162 (PTA_SKYLAKE): Add SGX.
16163 (PTA_ICELAKE): Remove.
16164 (PTA_ICELAKE_CLIENT): New.
16165 (PTA_ICELAKE_SERVER): New.
16166 (ix86_option_override_internal): Split up icelake on icelake client and
16168 (get_builtin_code_for_version): Ditto.
16169 (fold_builtin_cpu): Ditto.
16170 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
16171 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
16172 * config/i386/i386.h (processor_type): Ditto.
16173 * doc/invoke.texi: Ditto.
16175 2018-03-14 Jakub Jelinek <jakub@redhat.com>
16178 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
16179 INTEGER_CST offset, add it together with bitpos / 8 and
16180 sign extend based on POINTER_SIZE.
16184 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
16187 * config/i386/constraints.md (Yc): New register constraint.
16188 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
16189 Use Yc constraint for alternative 2 of operand 0. Remove
16190 preferred_for_speed attribute.
16192 2018-03-14 Richard Biener <rguenther@suse.de>
16194 PR tree-optimization/84830
16195 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
16196 with the old one to avoid oscillations.
16198 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
16201 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
16203 (assign_by_spills): Return a flag of reload assignment failure.
16204 Do not process the reload assignment failures. Do not spill other
16205 reload pseudos if they has the same reg class. Update n if
16207 (lra_assign): Add a return arg. Set up from the result of
16208 assign_by_spills call.
16209 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
16210 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
16211 usage_insns if it is not NULL.
16212 (spill_hard_reg_in_range): New function.
16213 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
16214 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
16215 function prototypes.
16216 (lra_assign): Change prototype.
16217 * lra.c (lra): Add code to deal with fails by splitting hard reg
16220 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
16222 * config/riscv/riscv.opt (mrelax): New option.
16223 * config/riscv/riscv.c (riscv_file_start): Emit ".option
16224 "norelax" when riscv_mrelax is disabled.
16225 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
16227 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16230 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
16231 reassociation for int modes.
16233 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
16235 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
16236 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
16238 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
16239 * config/aarch64/aarch64-sve.md
16240 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
16241 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
16242 (*extend<mode><Vwide>2): Rename to...
16243 (aarch64_sve_extend<mode><Vwide>2): ...this.
16244 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
16245 renaming the old pattern to...
16246 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
16248 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
16249 define_expand, renaming the old pattern to...
16250 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
16251 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
16252 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
16253 account when deciding which SVE instruction the optab should use.
16254 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
16256 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
16258 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
16259 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
16260 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
16261 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
16262 (tlsdesc_small_<mode>): Turn a define_expand and use
16263 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
16264 (tlsdesc_small_advsimd_<mode>): ...this.
16265 (tlsdesc_small_sve_<mode>): New pattern.
16267 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
16269 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
16270 (UNSPEC_UMUL_HIGHPART): New constants.
16271 (MUL_HIGHPART): New int iteraor.
16272 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
16273 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
16275 (*<su>mul<mode>3_highpart): New define_insn.
16277 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
16280 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
16283 2018-03-13 Martin Liska <mliska@suse.cz>
16286 * (sem_item_optimizer::sem_item_optimizer): Initialize new
16288 (sem_item_optimizer::~sem_item_optimizer): Release it.
16289 (sem_item_optimizer::merge_classes): Register variable aliases.
16290 (sem_item_optimizer::fixup_pt_set): New function.
16291 (sem_item_optimizer::fixup_points_to_sets): Likewise.
16292 * ipa-icf.h: Declare new variables and functions.
16294 2018-03-13 Jakub Jelinek <jakub@redhat.com>
16296 PR middle-end/84834
16297 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
16298 integer_pow2p@2 and test integer_pow2p in condition.
16299 (A < 0 ? C : 0): Similarly for @1.
16301 PR middle-end/84831
16302 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
16303 characters starting at p contain '\0' character, don't look beyond
16307 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
16308 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
16311 * reg-stack.c (change_stack): Change update_end var from int to
16312 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
16313 also call set_block_for_insn on the newly added insns and rescan.
16316 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
16317 on the last operand.
16320 * tree.c (stabilize_reference_1): Return save_expr (e) for
16321 STATEMENT_LIST even if it doesn't have side-effects.
16323 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
16325 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
16327 2018-03-12 Renlin Li <renlin.li@arm.com>
16329 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
16330 aarch64_output_scalar_simd_mov_immediate.
16332 2018-03-12 Martin Sebor <msebor@redhat.com>
16334 PR tree-optimization/83456
16335 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
16336 for perfectly overlapping calls to memcpy.
16337 (gimple_fold_builtin_memory_chk): Same.
16338 (gimple_fold_builtin_strcpy): Handle no-warning.
16339 (gimple_fold_builtin_stxcpy_chk): Same.
16340 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
16342 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
16344 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
16345 parameter. Use it for SFmode.
16346 (rs6000_function_arg_advance_1): Adjust.
16347 (rs6000_function_arg): Adjust.
16348 (rs6000_gimplify_va_arg): Pass false for that new parameter.
16350 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
16352 PR rtl-optimization/84169
16353 PR rtl-optimization/84780
16354 * combine.c (can_combine_p): Check for a 2-insn combination whether
16355 the destination register is used between the two insns, too.
16357 2018-03-12 Richard Biener <rguenther@suse.de>
16359 PR tree-optimization/84803
16360 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
16361 for refs DR analysis didn't process.
16363 2018-03-12 Richard Biener <rguenther@suse.de>
16365 PR tree-optimization/84777
16366 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
16367 force-vectorize loops ignore whether we are optimizing for size.
16369 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
16371 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
16372 (TARGET_MD_ASM_ADJUST): Define.
16374 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
16375 Kito Cheng <kito.cheng@gmail.com>
16376 Chung-Ju Wu <jasonwucj@gmail.com>
16378 * config/nds32/nds32.c (nds32_compute_stack_frame,
16379 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
16380 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
16381 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
16382 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
16383 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
16384 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
16385 * config/nds32/nds32.md (prologue, epilogue): Use macro
16386 NDS32_V3PUSH_AVAILABLE_P to do checking.
16388 2018-03-11 Jakub Jelinek <jakub@redhat.com>
16391 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
16392 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
16393 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
16394 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
16395 addition of most attributes on !orig_type_die or the attribute not
16396 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
16398 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
16399 Chung-Ju Wu <jasonwucj@gmail.com>
16401 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
16402 __NDS32_VH__ macro.
16403 * config/nds32/nds32.opt (mvh): New option.
16405 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
16406 Chung-Ju Wu <jasonwucj@gmail.com>
16408 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
16410 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
16411 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
16414 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
16415 Chung-Ju Wu <jasonwucj@gmail.com>
16417 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
16419 * config/nds32/nds32-multiple.md (strlensi): New pattern.
16420 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
16422 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
16423 Kito Cheng <kito.cheng@gmail.com>
16424 Chung-Ju Wu <jasonwucj@gmail.com>
16426 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
16427 UNSPEC_FFMISM and UNSPEC_FLMISM.
16428 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
16429 for ffb, ffmism and flmism.
16430 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
16431 (unspec_ffmism): Ditto.
16432 (unspec_flmism): Ditto.
16433 (nds32_expand_builtin_impl): Check if string extension is available.
16434 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
16435 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
16437 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
16440 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
16443 * lra-assigns.c (assign_by_spills): Return a flag of reload
16444 assignment failure. Do not process the reload assignment
16445 failures. Do not spill other reload pseudos if they has the same
16447 (lra_assign): Add a return arg. Set up from the result of
16448 assign_by_spills call.
16449 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
16450 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
16451 usage_insns if it is not NULL.
16452 (spill_hard_reg_in_range): New function.
16453 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
16454 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
16455 function prototypes.
16456 (lra_assign): Change prototype.
16457 * lra.c (lra): Add code to deal with fails by splitting hard reg
16460 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
16463 * config/i386/i386.opt: Replace Enforcment with Enforcement.
16465 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
16468 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
16469 (dw_val_node): Add val_symbolic_view.
16470 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
16471 (symview_upper_bound): New.
16472 (new_line_info_table): Initialize symviews_since_reset.
16473 (dwarf2out_source_line): Count symviews_since_reset and set
16474 symview_upper_bound.
16475 (dw_val_equal_p): Handle symview.
16476 (add_AT_symview): New.
16477 (print_dw_val): Handle symview.
16478 (attr_checksum, attr_checksum_ordered): Likewise.
16479 (same_dw_val_p, size_of_die): Likewise.
16480 (value_format, output_die): Likewise.
16481 (add_high_low_attributes): Use add_AT_symview for entry_view.
16482 (dwarf2out_finish): Reset symview_upper_bound, clear
16485 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
16488 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
16489 Add strict argument and use it.
16490 (rs6000_split_multireg_move): Update for new strict argument.
16491 (mem_operand_gpr): Disallow all non-offsettable addresses.
16492 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
16494 2018-03-09 Jakub Jelinek <jakub@redhat.com>
16497 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
16498 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
16499 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
16502 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
16503 decl, use remap_type if we want to use the type.
16505 2018-03-09 Martin Sebor <msebor@redhat.com>
16507 PR tree-optimization/84526
16508 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
16510 (builtin_access::generic_overlap): Be prepared to handle non-array
16513 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
16515 PR rtl-optimization/84682
16516 * lra-constraints.c (process_address_1): Check is_address flag
16517 for address constraints.
16518 (process_alt_operands): Likewise.
16519 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
16520 preprocess_constraints.
16521 * recog.h (preprocess_constraints): Add oploc parameter.
16523 * recog.c (preprocess_constraints): Test address_operand for
16524 CT_ADDRESS constraints.
16526 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
16529 * lra-assigns.c (assign_by_spills): Return a flag of reload
16530 assignment failure. Do not process the reload assignment
16531 failures. Do not spill other reload pseudos if they has the same
16533 (lra_assign): Add a return arg. Set up from the result of
16534 assign_by_spills call.
16535 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
16536 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
16537 usage_insns if it is not NULL.
16538 (spill_hard_reg_in_range): New function.
16539 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
16540 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
16541 function prototypes.
16542 (lra_assign): Change prototype.
16543 * lra.c (lra): Add code to deal with fails by splitting hard reg
16546 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16549 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
16550 Accept complain bool parameter. Only emit errors if it is true.
16551 (arm_parse_cpu_option_name): Likewise.
16552 (arm_target_thumb_only): Adjust callers of the above.
16553 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
16554 prototype to take a default true bool parameter.
16555 (arm_parse_arch_option_name): Likewise.
16557 2018-03-09 David Malcolm <dmalcolm@redhat.com>
16558 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
16562 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
16563 * configure: Regenerate.
16564 * configure.ac ("linker --version-script option"): New.
16565 ("linker soname option"): New.
16567 2018-03-09 Richard Biener <rguenther@suse.de>
16569 PR tree-optimization/84775
16570 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
16571 immediate uses of predicate stmts and mark them modified.
16574 PR tree-optimization/84178
16575 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
16577 (version_loop_for_if_conversion): Delay update_ssa call.
16578 (tree_if_conversion): Delay update_ssa until after predicate
16581 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
16584 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
16585 when the function accesses prior frames.
16587 2018-03-08 Jakub Jelinek <jakub@redhat.com>
16590 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
16591 gen_llsym, otherwise call maybe_gen_llsym.
16593 PR inline-asm/84742
16594 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
16595 has ',' character inside of it.
16597 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16600 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
16601 as clobbering CC_REGNUM.
16603 2018-03-08 Richard Biener <rguenther@suse.de>
16605 PR middle-end/84552
16606 * tree-scalar-evolution.c: Include tree-into-ssa.h.
16607 (follow_copies_to_constant): Do not follow SSA names registered
16610 2018-03-08 Richard Biener <rguenther@suse.de>
16612 PR tree-optimization/84178
16613 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
16615 (version_loop_for_if_conversion): Delay update_ssa call.
16616 (tree_if_conversion): Delay update_ssa until after predicate
16619 2018-03-08 David Malcolm <dmalcolm@redhat.com>
16621 PR tree-optimization/84178
16622 * tree-if-conv.c (release_bb_predicate): Remove the
16623 the assertion that the stmts have NULL use_ops.
16624 Discard the statements, asserting that they haven't
16625 yet been added to a BB.
16627 2018-03-08 Richard Biener <rguenther@suse.de>
16629 PR tree-optimization/84746
16630 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
16631 (phi_translate): Pass in destination ANTIC_OUT set.
16632 (phi_translate_1): Likewise. For a simplified result lookup
16633 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
16634 (phi_translate_set): Adjust.
16635 (do_pre_regular_insertion): Likewise.
16636 (do_pre_partial_partial_insertion): Likewise.
16638 2018-03-08 Martin Liska <mliska@suse.cz>
16640 PR gcov-profile/84735
16641 * doc/gcov.texi: Document usage of profile files.
16642 * gcov-io.h: Document changes in the format.
16644 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
16648 * dwarf2out.c (struct dw_line_info_table): Update comments for
16650 (FORCE_RESET_NEXT_VIEW): New.
16651 (FORCE_RESETTING_VIEW_P): New.
16652 (RESETTING_VIEW_P): Check for -1 too.
16653 (ZERO_VIEW_P): Likewise.
16654 (new_line_info_table): Force-reset next view.
16655 (dwarf2out_begin_function): Likewise.
16656 (dwarf2out_source_line): Simplify zero_view_p initialization.
16657 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
16658 view directly. Omit view when omitting .loc at line 0.
16660 2018-03-08 Jakub Jelinek <jakub@redhat.com>
16662 PR tree-optimization/84740
16663 * tree-switch-conversion.c (process_switch): Call build_constructors
16664 only if info.phi_count is non-zero.
16666 PR tree-optimization/84739
16667 * tree-tailcall.c (find_tail_calls): Check call arguments against
16668 DECL_ARGUMENTS (current_function_decl) rather than
16669 DECL_ARGUMENTS (func) when checking for tail recursion.
16671 2018-03-07 Jakub Jelinek <jakub@redhat.com>
16673 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
16674 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
16675 Volker Reichelt's entry and add entries for people that perform
16676 GCC fuzzy testing and report numerous bugs.
16678 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
16681 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
16682 readonly data in sdata, if that is disabled.
16683 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
16684 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
16685 -mreadonly-in-sdata option.
16687 2018-03-07 Martin Sebor <msebor@redhat.com>
16689 PR tree-optimization/84468
16690 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
16691 basic block when looking for nul assignment.
16693 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
16696 * except.h (output_function_exception_table): Adjust prototype.
16697 * except.c (output_function_exception_table): Remove FNNAME parameter
16698 and add SECTION parameter. Ouput one part of the table at a time.
16699 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
16700 the first part of the exception table and emit unwind directives.
16701 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
16702 (i386_pe_seh_cold_init): Likewise.
16703 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
16704 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
16705 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
16706 (ix86_output_call_insn): Emit a nop in one more case for SEH.
16707 * config/i386/winnt.c: Include except.h.
16708 (struct seh_frame_state): Add reg_offset, after_prologue and
16709 in_cold_section fields.
16710 (i386_pe_seh_end_prologue): Set seh->after_prologue.
16711 (i386_pe_seh_cold_init): New function.
16712 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
16713 to seh->in_cold_section.
16714 (seh_emit_push): Record the offset of the push.
16715 (seh_emit_save): Record the offet of the save.
16716 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
16717 Test seh->after_prologue to disregard the epilogue.
16718 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
16719 (i386_pe_end_cold_function): New function.
16721 2018-03-07 Jakub Jelinek <jakub@redhat.com>
16724 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
16725 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
16728 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
16730 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
16731 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
16733 PR middle-end/84723
16734 * multiple_target.c: Include tree-inline.h and intl.h.
16735 (expand_target_clones): Diagnose and fail if node->definition and
16736 !tree_versionable_function_p (node->decl).
16738 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
16740 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
16742 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
16743 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
16744 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
16746 2018-03-06 Jakub Jelinek <jakub@redhat.com>
16749 * combine.c (try_combine): Use reg_or_subregno instead of handling
16750 just paradoxical SUBREGs and REGs.
16752 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
16754 * config/arc/arc.c (arc_finalize_pic): Remove function.
16755 (arc_must_save_register): We use single base PIC register, remove
16756 checks to save/restore the PIC register.
16757 (arc_expand_prologue): Likewise.
16758 * config/arc/arc-protos.h (arc_set_default_type_attributes):
16760 (arc_verify_short): Likewise.
16761 (arc_attr_type): Likewise.
16762 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
16763 (walk_stores): Likewise.
16764 (arc_address_cost): Make it static.
16765 (arc_verify_short): Likewise.
16766 (branch_dest): Likewise.
16767 (arc_attr_type): Likewise.
16768 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
16769 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
16770 (arc_final_prescan_insn): Remove inserting the nops due to
16771 hardware hazards. It is done in reorg step.
16772 (insn_length_variant_t): Remove.
16773 (insn_length_parameters_t): Likewise.
16774 (arc_insn_length_parameters): Likewise.
16775 (arc_get_insn_variants): Likewise.
16776 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
16778 2018-03-06 Jakub Jelinek <jakub@redhat.com>
16780 PR inline-asm/84683
16781 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
16784 PR tree-optimization/84687
16785 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
16787 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
16789 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16791 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
16792 Rename to ppc_speculation_barrier.
16793 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
16794 __builtin_ppc_speculation_barrier.
16796 2018-03-05 Jakub Jelinek <jakub@redhat.com>
16799 * combine.c (combine_simplify_rtx): Don't try to simplify if
16800 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
16803 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
16805 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
16806 to 32 bytes when compiling for POWER9.
16808 2018-03-05 Jakub Jelinek <jakub@redhat.com>
16811 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
16812 regparm >= 3 with no arg reg available also for calls with
16813 flag_force_indirect_call. Pass decl to ix86_function_regparm.
16816 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
16818 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
16820 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
16823 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
16825 2018-03-05 Richard Biener <rguenther@suse.de>
16827 PR tree-optimization/84486
16828 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
16829 When inserting a __builtin_assume_aligned call set the LHS
16830 SSA name alignment info accordingly.
16832 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
16834 PR tree-optimization/84114
16835 * config/aarch64/aarch64.c (aarch64_reassociation_width)
16836 Avoid reassociation of FLOAT_MODE addition.
16838 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
16840 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
16841 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
16842 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
16843 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
16844 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
16845 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
16846 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
16848 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
16849 __builtin_ia32_wbinvd): New builtins.
16850 (SPECIAL_ARGS2): New.
16851 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
16852 (SPECIAL_ARGS2): New.
16853 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
16854 (ix86_valid_target_attribute_inner_p): Ditto.
16855 (ix86_init_mmx_sse_builtins): Add special_args2.
16856 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
16857 TARGET_WBNOINVD_P): New.
16858 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
16859 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
16860 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
16861 * config/i386/immintrin.h (_wbinvd): New intrinsic.
16862 * config/i386/pconfigintrin.h: New file.
16863 * config/i386/wbnoinvdintrin.h: Ditto.
16864 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
16866 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
16868 2018-03-05 Richard Biener <rguenther@suse.de>
16870 PR tree-optimization/84670
16871 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
16873 (BB_VISITED_WITH_VISITED_SUCCS): New define.
16874 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
16875 (compute_antic_aux): Only assert the number of values in ANTIC_IN
16876 doesn't grow if all successors (recursively) were visited at least
16879 2018-03-05 Richard Biener <rguenther@suse.de>
16881 PR tree-optimization/84650
16882 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
16883 if executed in the loop pipeline.
16885 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
16887 * doc/configfiles.texi (Configuration Files): Move info about
16888 conditionalizing $target-protos.h to...
16889 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
16890 differs from $target-protos.h.
16892 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
16893 Chung-Ju Wu <jasonwucj@gmail.com>
16895 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
16896 * config/nds32/nds32-multiple.md (setmemsi): Define.
16897 * config/nds32/nds32-memory-manipulation.c
16898 (nds32_gen_dup_4_byte_to_word_value): New.
16899 (emit_setmem_word_loop): New.
16900 (emit_setmem_byte_loop): New.
16901 (nds32_expand_setmem_loop): New.
16902 (nds32_expand_setmem_loop_v3m): New.
16903 (nds32_expand_setmem_unroll): New.
16904 (nds32_expand_setmem): New.
16906 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
16907 Chung-Ju Wu <jasonwucj@gmail.com>
16909 * config/nds32/nds32-memory-manipulation.c
16910 (nds32_emit_load_store): New.
16911 (nds32_emit_post_inc_load_store): New.
16912 (nds32_emit_mem_move): New.
16913 (nds32_emit_mem_move_block): New.
16914 (nds32_expand_movmemsi_loop_unknown_size): New.
16915 (nds32_expand_movmemsi_loop_known_size): New.
16916 (nds32_expand_movmemsi_loop): New.
16917 (nds32_expand_movmemsi_unroll): New.
16918 (nds32_expand_movmemqi): Rename ...
16919 (nds32_expand_movmemsi): ... to this.
16920 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
16921 (movmemsi): ... to this.
16922 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
16923 (nds32_expand_movmemsi): ... to this.
16925 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
16926 Monk Chiang <sh.chiang04@gmail.com>
16927 Chung-Ju Wu <jasonwucj@gmail.com>
16929 * config/nds32/nds32-protos.h
16930 (nds32_expand_load_multiple): New arguments.
16931 (nds32_expand_store_multiple): Ditto.
16932 (nds32_valid_multiple_load_store): Rename ...
16933 (nds32_valid_multiple_load_store_p): ... to this.
16934 * config/nds32/nds32-memory-manipulation.c
16935 (nds32_expand_load_multiple): Refine implementation.
16936 (nds32_expand_store_multiple): Ditto.
16937 * config/nds32/nds32-multiple.md
16938 (load_multiple): Update nds32_expand_load_multiple interface.
16939 (store_multiple): Update nds32_expand_store_multiple interface.
16940 * config/nds32/nds32-predicates.c
16941 (nds32_valid_multiple_load_store): Rename ...
16942 (nds32_valid_multiple_load_store_p): ... to this and refine
16944 * config/nds32/predicates.md
16945 (nds32_load_multiple_and_update_address_operation): New predicate.
16946 (nds32_store_multiple_and_update_address_operation): New predicate.
16948 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
16949 Chung-Ju Wu <jasonwucj@gmail.com>
16951 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
16952 (combo): New attribute.
16953 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
16955 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
16957 * config/nds32/nds32.opt: Change -mcmodel= default value.
16959 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
16960 Monk Chiang <sh.chiang04@gmail.com>
16961 Chung-Ju Wu <jasonwucj@gmail.com>
16963 * config/nds32/constants.md (unspec_element): New enum.
16964 * config/nds32/constraints.md (Umw): New constraint.
16965 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
16966 * config/nds32/nds32-intrinsic.md: Likewise.
16967 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
16968 (nds32_valid_smw_lwm_base_p): New.
16969 (nds32_output_smw_single_word): New.
16970 (nds32_output_lmw_single_word): New.
16971 (nds32_expand_unaligned_load): New.
16972 (nds32_expand_unaligned_store): New.
16973 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
16974 (nds32_output_smw_single_word): Declare.
16975 (nds32_output_lmw_single_word): Declare.
16976 (nds32_expand_unaligned_load): Declare.
16977 (nds32_expand_unaligned_store): Declare.
16978 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
16979 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
16980 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
16981 NDS32_BUILTIN_UASTORE_DW.
16982 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
16985 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
16986 Kito Cheng <kito.cheng@gmail.com>
16987 Chung-Ju Wu <jasonwucj@gmail.com>
16989 * config/nds32/nds32-intrinsic.c
16990 (nds32_expand_builtin_null_ftype_reg): Delete.
16991 (nds32_expand_builtin_reg_ftype_imm): Ditto.
16992 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
16993 (nds32_read_argument): New.
16994 (nds32_legitimize_target): Ditto.
16995 (nds32_legitimize_argument): Ditto.
16996 (nds32_check_constant_argument): Ditto.
16997 (nds32_expand_unop_builtin): Ditto.
16998 (nds32_expand_unopimm_builtin): Ditto.
16999 (nds32_expand_binop_builtin): Ditto.
17000 (nds32_builtin_decl_impl): Ditto.
17001 (builtin_description): Ditto.
17002 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
17003 (nds32_init_builtins_impl): Ditto.
17004 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
17005 (nds32_builtin_decl): New.
17006 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
17007 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
17009 2018-03-02 Jeff Law <law@redhat.com>
17011 * reorg.c (stop_search_p): Handle DEBUG_INSN.
17012 (redundant_insn, fill_simple_delay_slots): Likewise.
17013 (fill_slots_from_thread): Likewise.
17014 * resource.c (mark_referenced_resources): Likewise.
17015 (mark_set_resources, find_dead_or_set_registers): Likewise.
17017 2018-03-02 Jakub Jelinek <jakub@redhat.com>
17019 * substring-locations.h (format_warning_va): Formatting fix for
17020 ATTRIBUTE_GCC_DIAG.
17021 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
17023 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
17024 * substring-locations.c: Include intl.h.
17025 (format_warning_va): Turned into small wrapper around
17026 format_warning_n_va, renamed to ...
17027 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
17028 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
17030 (format_warning_at_substring_n): New function.
17031 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
17032 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
17033 format_warning_at_substring with just a shorter name instead of
17034 const function pointer.
17035 (fmtwarn_n): New function.
17036 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
17037 appropriate, get rid of all the fmtstr temporaries, move conditionals
17038 with G_() wrapped string literals directly into fmtwarn arguments,
17039 cast dir.len to (int), formatting fixes.
17041 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
17043 * doc/invoke.texi: Remove "Cilk Plus" references.
17045 2018-03-02 Jakub Jelinek <jakub@redhat.com>
17046 Richard Biener <rguenther@suse.de>
17049 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
17050 for error or warning attributes if CALL_FROM_THUNK_P is set.
17053 2018-03-02 Jakub Jelinek <jakub@redhat.com>
17056 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
17057 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
17060 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
17061 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
17063 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
17064 instead of -1U in last predictors element's probability member.
17066 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
17069 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
17070 arguments if they are comparable.
17072 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
17074 PR tree-optimization/84634
17075 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
17076 masks and masked_loop_p with a single loop_masks, making sure it's
17077 null for bb vectorization.
17079 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
17081 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
17082 (vect_analyze_data_ref_access): Use loop->safe_len rather than
17083 loop->force_vectorize to check whether there is no alias.
17085 2018-03-02 Jakub Jelinek <jakub@redhat.com>
17088 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
17090 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
17092 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
17093 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
17094 instead of a loop around prev_real_insn.
17095 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
17098 PR inline-asm/84625
17099 * config/i386/i386.c (ix86_print_operand): Use conditional
17100 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
17103 2018-03-02 Richard Biener <rguenther@suse.de>
17105 PR tree-optimization/84427
17106 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
17107 (bitmap_set_subtract_values): Rewrite to handle multiple
17110 (prune_clobbered_mems): Likewise.
17111 (phi_translate): Take edge instead of pred/phiblock.
17112 (phi_translate_1): Likewise.
17113 (phi_translate_set): Likewise. Insert all translated
17114 exprs for a value into the set, keeping possibly multiple
17115 expressions per value.
17116 (compute_antic_aux): Adjust for phi_translate changes.
17117 When intersecting union the expressions and prune those
17118 not in the final value set, keeping possibly multiple
17119 expressions per value. Do not use value-insertion
17120 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
17121 all expressions. Add verification that the value-sets
17122 only shrink during iteration.
17123 (compute_partial_antic_aux): Adjust for the phi_translate changes.
17124 (do_pre_regular_insertion): Likewise.
17125 (do_pre_partial_partial_insertion): Likewise.
17127 2018-03-02 Richard Biener <rguenther@suse.de>
17130 * config/darwin.c (saved_debug_info_level): New static global.
17131 (darwin_asm_lto_start): Disable debug info generation for LTO out.
17132 (darwin_asm_lto_end): Restore debug info generation settings.
17134 2018-03-01 Martin Liska <mliska@suse.cz>
17137 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
17138 volatile arguments.
17140 2018-03-01 Richard Biener <rguenther@suse.de>
17143 * dwarf2out.c (gen_variable_die): Properly handle late VLA
17144 type annotation with LTO when debug was disabled at compile-time.
17146 2018-03-01 Matthew Fortune <mfortune@gmail.com>
17148 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
17150 (mips_final_postscan_insn): Likewise.
17152 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
17154 PR rtl-optimization/84528
17155 * alias.c (init_alias_target): Add commentary.
17156 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
17157 a unique base value if the frame pointer is not eliminated
17158 to the stack pointer.
17160 2018-03-01 Tom de Vries <tom@codesourcery.com>
17162 PR rtl-optimization/83327
17163 * lra-int.h (hard_regs_spilled_into): Declare.
17164 * lra.c (hard_regs_spilled_into): Define.
17165 (init_reg_info): Init hard_regs_spilled_into.
17166 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
17167 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
17168 (process_bb_lives): Handle hard_regs_spilled_into.
17169 (lra_create_live_ranges_1): Before doing liveness propagation, clear
17170 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
17172 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
17174 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
17175 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
17176 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
17177 * config/rs6000/aix72.h: New file.
17179 2018-02-28 Jakub Jelinek <jakub@redhat.com>
17181 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
17182 instead of warning_at with conditional singular and plural messages
17186 * stor-layout.c (update_alignment_for_field): For
17187 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
17188 && !DECL_PACKED (field), do the alignment update, just use
17189 only desired_align instead of MAX (type_align, desired_align)
17191 (place_field): Don't do known_align < desired_align handling
17192 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
17193 is non-NULL, instead do it after rli->prev_field handling and
17194 only if not within a bitfield word. For DECL_PACKED (field)
17195 use type_align of BITS_PER_UNIT.
17197 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
17199 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
17200 superfluous parentheses and trailing spaces.
17202 2018-02-28 Richard Biener <rguenther@suse.de>
17204 PR tree-optimization/84584
17205 * graphite-scop-detection.c (scop_detection::add_scop): Discard
17206 SCoPs with fake exit edge.
17208 2018-02-28 Martin Liska <mliska@suse.cz>
17211 * timevar.c (timer::print): Fix format to properly print 100%
17214 2018-02-28 Richard Biener <rguenther@suse.de>
17216 PR middle-end/84607
17217 * genmatch.c (capture_info::walk_match): Do not mark
17218 captured expressions without operands as expr_p given
17219 they act more like predicates and should be subject to
17220 "lost tail" side-effect preserving.
17222 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
17224 PR rtl-optimization/81611
17225 * auto-inc-dec.c (attempt_change): Move dead note from
17226 mem_insn if it's the next use of regno
17227 (find_address): Take address use of reg holding
17228 non-incremented value. Add parm to limit search to the named
17230 (merge_in_block): Attempt to use a mem insn that is the next
17231 use of the original regno.
17233 2018-02-27 Martin Sebor <msebor@redhat.com>
17236 * doc/invoke.texi (-Wmissing-attributes): New option.
17237 * print-tree.c (print_node): Handle DECL_UNINLINABLE.
17239 2018-02-27 Martin Sebor <msebor@redhat.com>
17241 PR translation/84207
17242 * diagnostic-core.h (warning_n, error_n, inform_n): Change
17243 n argument to unsigned HOST_WIDE_INT.
17244 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
17245 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
17246 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
17247 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
17249 2018-02-27 Richard Biener <rguenther@suse.de>
17251 PR tree-optimization/84512
17252 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
17253 Do not use the estimate returned from record_stmt_cost for
17254 the scalar iteration cost but sum properly using add_stmt_cost.
17256 2018-02-27 Richard Biener <rguenther@suse.de>
17258 PR tree-optimization/84466
17259 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
17260 Adjust last change to less strictly validate use operands.
17262 2018-02-27 Martin Liska <mliska@suse.cz>
17264 PR gcov-profile/84548
17265 * gcov.c (process_file): Allow partial overlap and consider it
17266 also as group functions.
17267 (output_lines): Properly calculate range of lines for a group.
17269 2018-02-27 Martin Liska <mliska@suse.cz>
17271 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
17272 'ggc' suffixes. Change first column width.
17273 (timer::print): Fix formatting of the column.
17275 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
17277 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
17278 preserve inline entry blocks for the sake of debug inline
17279 entry point markers alone.
17280 (remove_unused_locals): Suggest in comments a better place to
17281 force the preservation of inline entry blocks that are
17282 otherwise unused, but do not preserve them.
17284 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
17286 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
17288 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
17291 * config/i386/constraints.md (Bs): Replace
17292 ix86_indirect_branch_register with
17293 TARGET_INDIRECT_BRANCH_REGISTER.
17295 * config/i386/i386.md (indirect_jump): Likewise.
17296 (tablejump): Likewise.
17297 (*sibcall_memory): Likewise.
17298 (*sibcall_value_memory): Likewise.
17299 Peepholes of indirect call and jump via memory: Likewise.
17300 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
17301 (*sibcall_value_GOT_32): Likewise.
17302 * config/i386/predicates.md (indirect_branch_operand): Likewise.
17303 (GOT_memory_operand): Likewise.
17304 (call_insn_operand): Likewise.
17305 (sibcall_insn_operand): Likewise.
17306 (GOT32_symbol_operand): Likewise.
17307 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
17309 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
17311 PR rtl-optimization/83496
17312 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
17313 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
17314 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
17315 redundant insn, if any.
17316 (relax_delay_slots): Likewise.
17317 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
17319 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
17321 PR tree-optimization/83965
17322 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
17323 that grouped statements are part of a reduction chain. Return
17324 true if the statement is not marked as a reduction itself but
17325 is part of a group.
17326 (vect_recog_dot_prod_pattern): Don't check whether the statement
17327 is part of a group here.
17328 (vect_recog_sad_pattern): Likewise.
17329 (vect_recog_widen_sum_pattern): Likewise.
17331 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
17334 * final.c (rest_of_clean_state): Also look for calls inside sequences.
17336 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
17339 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
17341 (ix86_output_indirect_function_return): New prototype.
17342 (ix86_split_simple_return_pop_internal): Likewise.
17343 * config/i386/i386.c (indirect_return_via_cx): New.
17344 (indirect_return_via_cx_bnd): Likewise.
17345 (indirect_thunk_name): Handle return va CX_REG.
17346 (output_indirect_thunk_function): Create alias for
17347 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
17348 (ix86_output_indirect_jmp): Remove the bool argument.
17349 (ix86_output_indirect_function_return): New function.
17350 (ix86_split_simple_return_pop_internal): Likewise.
17351 * config/i386/i386.md (*indirect_jump): Don't pass false
17352 to ix86_output_indirect_jmp.
17353 (*tablejump_1): Likewise.
17354 (simple_return_pop_internal): Change it to define_insn_and_split.
17355 Call ix86_split_simple_return_pop_internal to split it for
17356 -mfunction-return=.
17357 (simple_return_indirect_internal): Call
17358 ix86_output_indirect_function_return instead of
17359 ix86_output_indirect_jmp.
17361 2018-02-26 Jakub Jelinek <jakub@redhat.com>
17364 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
17365 memset and value initialization afterwards.
17367 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
17369 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
17371 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
17374 * common/config/aarch64/aarch64-common.c
17375 (aarch_option_optimization_table[]): Switch
17376 off fomit-frame-pointer
17378 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
17379 Chung-Ju Wu <jasonwucj@gmail.com>
17381 * config/nds32/nds32-multiple.md (load_multiple): Disallow
17383 (store_multiple): Ditto.
17385 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
17387 * config.gcc: Add --with-cpu support for nds32 target.
17388 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
17389 * config/nds32/nds32.opt: Add -mcpu= option.
17391 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
17393 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
17394 isel=yes): Warn for these deprecated options.
17396 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
17398 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
17399 ISA_2_5_MASKS_EMBEDDED.
17401 2018-02-23 Jakub Jelinek <jakub@redhat.com>
17403 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
17404 p->max as pointers rather than using iterative_hash_expr.
17406 2018-02-23 Carl Love <cel@us.ibm.com>
17408 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
17409 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
17411 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
17412 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
17413 P8V_BUILTIN_VEC_VUNSIGNED2.
17415 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
17418 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
17419 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
17421 * lra-constraints.c (curr_insn_transform): Set up
17422 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
17424 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
17425 * lra-eliminations.c (spill_pseudos): Ditto.
17426 (process_insn_for_elimination): Ditto.
17427 * lra-lives.c (reg_early_clobber_p): Use the new macros.
17428 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
17429 LRA_NON_CLOBBERED_ALT.
17431 2018-02-22 Martin Sebor <msebor@redhat.com>
17433 PR tree-optimization/84480
17434 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
17435 to maybe_diag_stxncpy_trunc. Call it.
17436 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
17437 from gimple_fold_builtin_strcpy. Print inlining stack.
17438 (handle_builtin_stxncpy): Print inlining stack.
17439 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
17441 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
17444 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
17445 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
17446 and -fcheck-pointer-bounds are used together.
17447 (indirect_thunk_prefix): New enum.
17448 (indirect_thunk_need_prefix): New function.
17449 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
17450 "_nt" instead of "_bnd" for NOTRACK prefix.
17451 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
17452 (output_indirect_thunk_function): Likewise.
17454 (ix86_code_end): Update output_indirect_thunk_function calls.
17455 (ix86_output_indirect_branch_via_reg): Replace
17456 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
17457 (ix86_output_indirect_branch_via_push): Likewise.
17458 (ix86_output_function_return): Likewise.
17459 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
17460 incompatible with -fcf-protection=branch and
17461 -fcheck-pointer-bounds.
17463 2018-02-22 Steve Ellcey <sellcey@cavium.com>
17466 * config/aarch64/aarch64.c (aarch64_print_address_internal):
17467 Change gcc_assert call to output_operand_lossage.
17469 2018-02-22 Steve Ellcey <sellcey@cavium.com>
17471 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
17473 2018-02-22 DJ Delorie <dj@redhat.com>
17474 Sebastian Perta <sebastian.perta@renesas.com>
17475 Oleg Endo <olegendo@gcc.gnu.org>
17477 * config/rx/rx.c (rx_rtx_costs): New function.
17478 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
17480 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
17482 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
17484 2018-02-22 Martin Liska <mliska@suse.cz>
17487 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
17488 Add "native" as a possible value.
17490 2018-02-22 Martin Liska <mliska@suse.cz>
17493 * config/i386/i386.c (ix86_option_override_internal):
17494 Add "native" as a possible value for -march and -mtune.
17496 2018-02-22 Jakub Jelinek <jakub@redhat.com>
17499 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
17500 to all type variants.
17502 PR tree-optimization/84503
17503 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
17504 width as info->bitpos + info->bitsize - start.
17505 (merged_store_group::merge_overlapping): Simplify width computation.
17506 (check_no_overlap): New function.
17507 (imm_store_chain_info::try_coalesce_bswap): Compute expected
17508 start + width and last_order of the group, fail if check_no_overlap
17510 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
17511 to group if check_no_overlap fails.
17513 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
17515 * config/rs6000/altivec.md: Delete contraint arguments to
17516 define_expand, define_split, and define_peephole2, and in
17517 define_insn_and_split if always unused.
17518 * config/rs6000/darwin.md: Ditto.
17519 * config/rs6000/dfp.md: Ditto.
17520 * config/rs6000/rs6000.md: Ditto.
17521 * config/rs6000/sync.md: Ditto.
17522 * config/rs6000/vector.md: Ditto.
17523 * config/rs6000/vsx.md: Ditto.
17525 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
17527 * config/rs6000/altivec.md: Write output control strings as braced
17528 blocks instead of double-quoted strings.
17529 * config/rs6000/darwin.md: Ditto.
17530 * config/rs6000/rs6000.md: Ditto.
17531 * config/rs6000/vector.md: Ditto.
17532 * config/rs6000/vsx.md: Ditto.
17534 2018-02-21 Jason Merrill <jason@redhat.com>
17536 PR c++/84314 - ICE with templates and fastcall attribute.
17537 * attribs.c (build_type_attribute_qual_variant): Remove assert.
17539 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
17541 * ipa-cp.c (determine_versionability): Fix comment typos.
17543 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
17546 * ipa-cp.c (determine_versionability): Do not version functions caling
17549 2018-02-21 Martin Liska <mliska@suse.cz>
17552 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
17553 Add "native" as a possible value.
17554 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
17555 the macro when native cpu detection is available.
17557 2018-02-21 Martin Liska <mliska@suse.cz>
17560 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
17561 Add "native" as a possible value.
17562 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
17563 when native cpu detection is available.
17565 2018-02-21 Jakub Jelinek <jakub@redhat.com>
17566 Martin Sebor <msebor@redhat.com>
17568 PR tree-optimization/84478
17569 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
17571 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
17572 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
17573 support which is conservatively correct, for 2 only stay conservative
17574 for maxlen. Formatting and comment capitalization fixes. Add STRICT
17575 argument to the 2 argument get_range_strlen, adjust 6 arg
17576 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
17578 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
17579 (gimple_fold_builtin_strlen): Pass true as last argument to
17582 2018-02-20 Martin Sebor <msebor@redhat.com>
17584 PR middle-end/84095
17585 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
17586 (builtin_memref::set_base_and_offset): Same. Handle inner references.
17587 (builtin_memref::builtin_memref): Factor out parts into
17588 set_base_and_offset and call it.
17590 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
17592 PR middle-end/84406
17593 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
17594 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
17595 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
17596 search at the associated MODE_INT.
17598 2018-02-20 Jeff Law <law@redhat.com>
17600 PR middle-end/82123
17601 PR tree-optimization/81592
17602 PR middle-end/79257
17603 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
17604 for range data rather than using global data.
17605 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
17606 range data rather than using global data.
17607 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
17608 pass it to children as needed.
17609 (struct directive::fmtresult): Similarly.
17610 (struct directive::set_width): Similarly.
17611 (struct directive::set_precision): Similarly.
17612 (format_integer, format_directive, parse_directive): Similarly.
17613 (format_none): Accept unnamed vr_values parameter.
17614 (format_percent, format_floating, format_character): Similarly.
17615 (format_string, format_plain): Similarly.
17616 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
17617 the EVRP range analyzer for range data rather than using global data.
17618 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
17619 gimple-ssa-evrp-analyze.h
17620 (class sprintf_dom_walker): Add after_dom_children member function.
17621 Add evrp_range_analyzer member.
17622 (sprintf_dom_walker::before_dom_children): Call into the EVRP
17623 range analyzer as needed.
17624 (sprintf_dom_walker::after_dom_children): New member function.
17625 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
17627 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
17628 (evrp_range_analyzer::pop_to_marker): Likewise.
17630 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
17632 PR tree-optimization/84419
17633 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
17634 with the required type if its current type is compatible but
17637 2018-02-20 Jakub Jelinek <jakub@redhat.com>
17639 PR middle-end/82004
17640 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
17641 after vectorization.
17643 2018-02-20 Martin Liska <mliska@suse.cz>
17646 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
17647 possible values if we don't have a hint.
17649 2018-02-20 Martin Liska <mliska@suse.cz>
17653 * final.c (shorten_branches): Build align_tab array with one
17655 * opts.c (finish_options): Add alignment option limit check.
17656 (MAX_CODE_ALIGN): Likewise.
17657 (MAX_CODE_ALIGN_VALUE): Likewise.
17658 * doc/invoke.texi: Document maximum allowed option value for
17659 all -falign-* options.
17661 2018-02-19 Jakub Jelinek <jakub@redhat.com>
17664 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
17665 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
17666 * var-tracking.c (emit_note_insn_var_location): Remove all references
17667 to NOTE_INSN_CALL_ARG_LOCATION.
17668 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
17669 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
17670 Use copy_rtx_if_shared.
17671 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
17672 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
17673 (dwarf2out_var_location): Remove handling of
17674 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
17676 * final.c (final_scan_insn): Remove all references to
17677 NOTE_INSN_CALL_ARG_LOCATION.
17678 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
17679 before dumping final insns.
17680 * except.c (emit_note_eh_region_end): Remove all references to
17681 NOTE_INSN_CALL_ARG_LOCATION.
17682 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
17683 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
17684 * config/arc/arc.c (hwloop_optimize): Likewise.
17685 * config/arm/arm.c (create_fix_barrier): Likewise.
17686 * config/s390/s390.c (s390_chunkify_start): Likewise.
17687 * config/sh/sh.c (find_barrier): Likewise.
17688 * config/i386/i386.c (rest_of_insert_endbranch,
17689 ix86_seh_fixup_eh_fallthru): Likewise.
17690 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
17691 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
17692 * config/frv/frv.c (frv_function_prologue): Likewise.
17693 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
17695 (note_outside_basic_block_p): Remove all references to
17696 NOTE_INSN_CALL_ARG_LOCATION.
17697 * gengtype.c (adjust_field_rtx_def): Likewise.
17698 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
17700 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
17701 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
17704 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
17707 PR tree-optimization/84452
17708 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
17709 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
17712 2018-02-19 Martin Liska <mliska@suse.cz>
17715 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
17717 2018-02-19 Martin Liska <mliska@suse.cz>
17718 Richard Sandiford <richard.sandiford@linaro.org>
17720 PR tree-optimization/82491
17721 * gimple-fold.c (get_base_constructor): Make earlier bail out
17724 2018-02-19 Carl Love <cel@us.ibm.com>
17726 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
17727 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
17729 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
17730 P8V_BUILTIN_VEC_NEG.
17732 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
17734 * config/rl78/rl78.md (movdf): New define expand.
17736 2018-02-19 Martin Liska <mliska@suse.cz>
17739 * doc/invoke.texi: Fix typo.
17740 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
17742 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
17744 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
17745 handle rs6000_single_float and rs6000_double_float specially for
17748 2018-02-16 Jeff Law <law@redhat.com>
17750 * config/rx/rx.c (add_pop_cfi_notes): New function.;
17751 (pop_regs): Use it.
17753 2018-02-16 Jakub Jelinek <jakub@redhat.com>
17756 * ipa-inline.c (inline_small_functions): Fix a typo.
17758 2018-02-16 Nathan Sidwell <nathan@acm.org>
17760 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
17762 2018-02-16 Carl Love <cel@us.ibm.com>
17764 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
17765 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
17766 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
17767 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
17768 expansion to P8V_BUILTIN_VEC_FLOAT2.
17770 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
17772 PR rtl-optimization/70023
17773 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
17774 src_regno into account.
17776 2018-02-16 Carl Love <cel@us.ibm.com>
17778 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
17779 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
17780 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
17781 * config/rs6000/rs6000.c: Remove case statements for
17782 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
17783 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
17784 and P9V_BUILTIN_VEC_VINSERT4B.
17785 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
17786 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
17787 * config/rs6000/vsx.md:
17788 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
17791 2018-02-16 Carl Love <cel@us.ibm.com>
17793 * config/rs6000/altivec.h: Add builtin names vec_extract4b
17795 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
17797 * config/rs6000/rs6000-c.c: Add the definitions for
17798 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
17799 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
17800 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
17801 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
17802 definition for insert4b and define insn *insert3b_internal.
17803 * doc/extend.texi: Add documentation for vec_extract4b.
17805 2018-02-16 Nathan Sidwell <nathan@acm.org>
17807 * doc/extend.texi (Backwards Compatibility): Mention friend
17808 injection. Note for-scope is deprecated.
17809 * doc/invoke.texi (-ffriend-injection): Deprecate.
17811 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
17813 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
17814 that moved to I2, also allow destinations that are a paradoxical
17815 subreg (instead of a normal reg).
17817 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
17820 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
17823 2018-02-16 Richard Biener <rguenther@suse.de>
17825 PR tree-optimization/84037
17826 PR tree-optimization/84016
17828 * config/i386/i386.c (ix86_builtin_vectorization_cost):
17829 Adjust vec_construct for the fact we need additional higher latency
17830 128bit inserts for AVX256 and AVX512 vector builds.
17831 (ix86_add_stmt_cost): Scale vector construction cost for
17834 2018-02-16 Richard Biener <rguenther@suse.de>
17836 PR tree-optimization/84417
17837 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
17838 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
17839 (non_rewritable_lvalue_p): Likewise, use poly-ints.
17841 2018-02-16 Martin Liska <mliska@suse.cz>
17844 * internal-fn.def (ASAN_CHECK): Set proper flags.
17845 (ASAN_MARK): Likewise.
17847 2018-02-16 Julia Koval <julia.koval@intel.com>
17849 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
17850 from PTA_CANNONLAKE.
17852 2018-02-16 Jakub Jelinek <jakub@redhat.com>
17855 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
17856 Use ++iter rather than iter++ for std::list iterators.
17857 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
17858 defer deleting them until all nodes in the forest are processed. Do
17859 free even leaf nodes. Change to_process into auto_vec.
17862 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
17863 * vec.h (vec_default_construct): Use memset instead of placement new
17864 if BROKEN_VALUE_INITIALIZATION is defined.
17865 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
17866 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
17869 PR rtl-optimization/83723
17870 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
17871 * lra.c (lra_substitute_pseudo): Likewise. If true, use
17872 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
17874 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
17876 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
17878 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
17880 PR rtl-optimization/81443
17881 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
17882 from inner REGs to paradoxical SUBREGs.
17884 2018-02-16 Richard Biener <rguenther@suse.de>
17886 PR tree-optimization/84399
17887 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
17888 For operands we can analyze at their definition make sure we can
17889 analyze them at each use as well.
17891 2018-02-16 Richard Biener <rguenther@suse.de>
17893 PR tree-optimization/84190
17894 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
17895 volatile accesses if the decl isn't volatile.
17897 2018-02-15 Jason Merrill <jason@redhat.com>
17899 PR c++/84314 - ICE with templates and fastcall attribute.
17900 * attribs.c (build_type_attribute_qual_variant): Don't clobber
17901 TYPE_CANONICAL on an existing type.
17903 2018-02-15 Jakub Jelinek <jakub@redhat.com>
17905 PR tree-optimization/84383
17906 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
17907 dstoff nor call operand_equal_p if dstbase is NULL.
17909 PR tree-optimization/84334
17910 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
17911 also a CONSTANT_CLASS_P, punt.
17913 2018-02-14 Jim Wilson <jimw@sifive.com>
17915 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
17916 first SMALL_OPERAND check. New local min_second_step. Move assert
17917 to where locals are set. Add TARGET_RVC support.
17918 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
17920 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
17922 * doc/invoke.texi: Correct -Wformat-overflow code sample.
17924 2018-02-14 Martin Sebor <msebor@redhat.com>
17926 PR tree-optimization/83698
17927 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
17928 arrays constrain the offset range to their bounds.
17929 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
17930 (builtin_access::overlap): Avoid setting the size of overlap if it's
17932 (maybe_diag_overlap): Also consider arrays when deciding what values
17933 of offsets to include in diagnostics.
17935 2018-02-14 Martin Sebor <msebor@redhat.com>
17938 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
17939 that correspond to the kind of a declaration.
17941 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
17944 * config/pa/pa.md: Load address of PIC label using the linkage table
17945 if the label is nonlocal.
17947 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
17949 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
17950 warning message if user requests -maltivec=be.
17951 * doc/invoke.texi: Document deprecation of -maltivec=be.
17953 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
17956 * config/rs6000/rs6000-c.c: Update definitions for
17957 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
17958 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
17960 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
17963 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
17964 add _get_ssp intrinsics. Remove argument from
17965 __builtin_ia32_rdssp[d|q].
17966 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
17967 * config/i386/i386-builtin.def: Remove argument from
17968 __builtin_ia32_rdssp[d|q].
17969 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
17970 ix86_expand_special_args_builtin for _rdssp[d|q].
17971 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
17972 Clear register before usage.
17973 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
17974 Add documentation for new _get_ssp and _inc_ssp intrinsics.
17976 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
17978 PR tree-optimization/84357
17979 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
17980 operand 1 of an ARRAY_REF too.
17982 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
17985 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
17986 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
17988 (set_of_reg): New struct.
17989 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
17990 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
17991 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
17993 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
17994 Split into bitclr, bitset, bitinvert patterns if appropriate.
17995 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
17996 use rx_fuse_in_memory_bitop.
17997 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
17998 to named insn, correct maximum insn length.
18000 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
18003 * machmode.def: Define a complex mode for PARTIAL_INT.
18004 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
18006 * doc/rtl.texi: Document CSPImode.
18007 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
18009 (msp430_hard_regno_nregs_with_padding): Likewise.
18011 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
18014 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
18016 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
18018 PR rtl-optimization/84169
18019 * combine.c (try_combine): New variable split_i2i3. Set it to true if
18020 we generated a parallel as new i3 and we split that to new i2 and i3
18021 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
18022 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
18023 those to i2, not i1. Partially rewrite this scan code.
18025 2018-02-13 Jakub Jelinek <jakub@redhat.com>
18028 * stor-layout.c (place_field): For variable length fields, adjust
18029 offset_align afterwards not just based on the field's alignment,
18030 but also on the size.
18032 PR middle-end/84309
18033 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
18034 of exps and logs in the use_exp2 case.
18036 2018-02-13 Jeff Law <law@redhat.com>
18038 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
18039 entry for "vector".
18041 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
18044 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
18048 * common.opt (gas-loc-support, gas-locview-support): New.
18049 (ginline-points, ginternal-reset-location-views): New.
18050 * doc/invoke.texi: Document them. Use @itemx where intended.
18051 (gvariable-location-views): Adjust.
18052 * target.def (reset_location_view): New.
18053 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
18054 (TARGET_RESET_LOCATION_VIEW): New.
18055 * doc/tm.texi: Rebuilt.
18056 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
18057 (dwarf2out_default_as_locview_support): New.
18058 (output_asm_line_debug_info): Use option variables.
18059 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
18060 (output_loc_list): Likewise.
18061 (add_high_low_attributes): Check option variables.
18062 Don't output entry view attribute in strict mode.
18063 (gen_inlined_subroutine_die): Check option variables.
18064 (dwarf2out_inline_entry): Likewise.
18065 (init_sections_and_labels): Likewise.
18066 (dwarf2out_early_finish): Likewise.
18067 (maybe_reset_location_view): New, from...
18068 (dwarf2out_var_location): ... here. Call it.
18069 * debug.h (dwarf2out_default_as_loc_support): Declare.
18070 (dwarf2out_default_as_locview_support): Declare.
18071 * hooks.c (hook_int_rtx_insn_0): New.
18072 * hooks.h (hook_int_rtx_insn_0): Declare.
18073 * toplev.c (process_options): Take -gas-loc-support and
18074 -gas-locview-support from dwarf2out. Enable
18075 -gvariable-location-views by default only with locview
18076 assembler support. Enable -ginternal-reset-location-views by
18077 default only if the target defines the corresponding hook.
18078 Enable -ginline-points by default if location views are
18079 enabled; force it disabled if statement frontiers are
18081 * tree-inline.c (expand_call_inline): Check option variables.
18082 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
18084 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
18086 PR tree-optimization/84321
18087 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
18088 handling. Also check whether the anti-range contains any values
18089 that satisfy the mask; switch to a VR_RANGE if not.
18091 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
18094 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
18096 2018-02-13 Martin Jambor <mjambor@suse.cz>
18099 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
18100 of call statements, also set location of a load to a temporary.
18102 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
18104 * config/rl78/rl78.c (add_vector_labels): New function.
18105 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
18106 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
18107 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
18108 which checks that no arguments are passed.
18109 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
18110 * doc/extend.texi: Documentation for the new attribute.
18112 2018-02-13 Andreas Schwab <schwab@suse.de>
18114 * config/riscv/linux.h (CPP_SPEC): Define.
18116 2018-02-13 Jakub Jelinek <jakub@redhat.com>
18119 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
18120 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
18121 OPTION_MASK_ISA_AES as first argument to def_builtin_const
18122 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
18123 instead of OPTION_MASK_ISA_PCLMUL as first argument to
18124 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
18125 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
18126 temporarily for AES and PCLMUL builtins.
18128 PR tree-optimization/84339
18129 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
18130 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
18133 PR middle-end/84309
18134 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
18135 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
18136 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
18138 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
18140 * omp-simd-clone.h: New file.
18141 * omp-simd-clone.c: Include omp-simd-clone.h.
18142 (expand_simd_clones): No longer static.
18143 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
18144 cgraph.h and omp-simd-clone.h.
18145 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
18146 (vect_recog_widen_shift_pattern): Formatting fix.
18147 (vect_pattern_recog_1): Don't check optab for calls.
18150 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
18151 operands[2] into a REG before using gen_lowpart on it.
18153 2018-02-12 Jeff Law <law@redhat.com>
18156 * config/sh/sh.c (find_barrier): Consider a sibling call
18159 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
18160 successfully back substituting a reg.
18162 2018-02-12 Richard Biener <rguenther@suse.de>
18164 PR tree-optimization/84037
18165 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
18166 parameter, move visited init to caller.
18167 (vect_slp_analyze_operations): Separate cost from validity
18168 check, initialize visited once for all instances.
18169 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
18171 * tree-vect-stmts.c (vect_model_simple_cost): Make early
18173 (vect_model_promotion_demotion_cost): Likewise.
18174 (vectorizable_bswap): Guard cost modeling with !slp_node
18175 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
18177 (vectorizable_call): Likewise.
18178 (vectorizable_conversion): Likewise.
18179 (vectorizable_assignment): Likewise.
18180 (vectorizable_shift): Likewise.
18181 (vectorizable_operation): Likewise.
18182 (vectorizable_store): Likewise.
18183 (vectorizable_load): Likewise.
18184 (vectorizable_condition): Likewise.
18185 (vectorizable_comparison): Likewise.
18187 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
18190 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
18191 (ASAN_MARK): Fix fnspec to account for return value, change pointer
18192 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
18194 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
18196 PR middle-end/83665
18197 * params.def (inline-min-speedup): Increase from 8 to 15.
18198 (max-inline-insns-auto): Decrease from 40 to 30.
18199 * ipa-split.c (consider_split): Add some buffer for function to
18200 be considered inlining candidate.
18201 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
18204 2018-02-12 Richard Biener <rguenther@suse.de>
18206 PR tree-optimization/84037
18207 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
18208 matched stmts if we cannot swap the non-matched ones.
18210 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
18212 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
18213 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
18214 _mm_maskz_scalef_round_ss): New intrinsics.
18215 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
18216 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
18217 __builtin_ia32_scalefss_round): Remove.
18218 (__builtin_ia32_scalefsd_mask_round,
18219 __builtin_ia32_scalefss_mask_round): New intrinsics.
18220 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
18221 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
18222 ((match_operand:VF_128 2 "<round_nimm_predicate>"
18223 "<round_constraint>")): Changed to ...
18224 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
18225 "<round_scalar_constraint>")): ... this.
18226 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
18227 %0, %1, %2<round_op3>}"): Changed to ...
18228 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
18229 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
18230 %2<round_scalar_mask_op3>}"): ... this.
18231 * config/i386/subst.md (round_scalar_nimm_predicate): New.
18233 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
18235 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
18236 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
18237 (_mm_maskz_sqrt_round_ss): New intrinsics.
18238 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
18239 (__builtin_ia32_sqrtsd_mask_round)
18240 (__builtin_ia32_sqrtss_mask_round): New builtins.
18241 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
18242 (__builtin_ia32_sqrtss_round): Remove.
18243 (__builtin_ia32_sqrtsd_mask_round)
18244 (__builtin_ia32_sqrtss_mask_round): New builtins.
18245 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
18246 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
18247 ((match_operand:VF_128 1 "vector_operand"
18248 "xBm,<round_constraint>")): Changed to ...
18249 ((match_operand:VF_128 1 "vector_operand"
18250 "xBm,<round_scalar_constraint>")): ... this.
18251 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
18252 %0, %2, %<iptr>1<round_op3>}): Changed to ...
18253 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
18254 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
18255 %<iptr>1<round_scalar_mask_op3>}): ... this.
18256 ((set_attr "prefix" "<round_prefix>")): Changed to ...
18257 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
18259 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
18262 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
18263 Cast vec_cmpeq result to correct type.
18264 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
18265 Cast vec_cmpgt result to correct type.
18267 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
18269 * final.c (final_scan_insn_1): Renamed from...
18270 (final_scan_insn): ... this. New wrapper, to recover
18271 seen from the outermost call in recursive ones.
18272 * config/sparc/sparc.c (output_return): Drop seen from call.
18273 (output_sibcall): Likewise.
18274 * config/visium/visium.c (output_branch): Likewise.
18276 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
18278 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
18281 2018-02-10 Alan Modra <amodra@gmail.com>
18284 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
18285 Specify LR as an input.
18287 2018-02-10 Jakub Jelinek <jakub@redhat.com>
18290 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
18291 remove_member_access_dummy_vars): New functions.
18292 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
18293 lower_omp_1, execute_lower_omp): Use them.
18295 PR rtl-optimization/84308
18296 * shrink-wrap.c (spread_components): Release todo vector.
18298 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
18300 PR rtl-optimization/57193
18301 * ira-color.c (struct allocno_color_data): Add member
18302 conflict_allocno_hard_prefs.
18303 (update_conflict_allocno_hard_prefs): New.
18304 (bucket_allocno_compare_func): Add a preference based on
18305 conflict_allocno_hard_prefs.
18306 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
18307 (color_allocnos): Remove a dead code. Initiate
18308 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
18310 2018-02-09 Jakub Jelinek <jakub@redhat.com>
18313 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
18314 constraint from =wa to wa. Avoid a subreg on the output operand,
18315 instead use a pseudo and subreg it in a move.
18316 (p9_xxbrd_<mode>): Changed to ...
18317 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
18318 (p9_xxbrd_v2df): New expander.
18319 (p9_xxbrw_<mode>): Changed to ...
18320 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
18321 (p9_xxbrw_v4sf): New expander.
18323 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
18325 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
18327 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
18330 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
18331 multiply in 32-bit mode.
18332 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
18333 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
18336 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
18338 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
18339 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
18340 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
18341 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
18343 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
18346 * dwarf2out.c (is_trivial_indirect_ref): New function.
18347 (dwarf2out_late_global_decl): Do not generate a location
18348 attribute for variables that have a non-trivial DECL_VALUE_EXPR
18349 and that are not defined in the current unit.
18351 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
18353 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
18354 instead of a libcall for UNORDERED.
18356 2018-02-09 Tamar Christina <tamar.christina@arm.com>
18359 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
18360 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
18362 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
18365 * config/s390/s390.c (s390_set_current_function): Invoke
18366 s390_indirect_branch_settings also if fndecl didn't change.
18368 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
18370 * config/rs6000/rs6000.md (blockage): Set length to zero.
18372 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
18374 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
18376 2018-02-09 Jakub Jelinek <jakub@redhat.com>
18379 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
18380 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
18384 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
18385 PARALLEL incoming that failed vt_get_decl_and_offset check.
18387 PR middle-end/84237
18388 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
18389 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
18391 (get_variable_section): For decls in named .bss* sections pass true as
18392 second argument to bss_initializer_p.
18394 2018-02-09 Marek Polacek <polacek@redhat.com>
18395 Jakub Jelinek <jakub@redhat.com>
18398 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
18399 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
18400 Sync some changes from cxx_fold_indirect_ref.
18402 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
18404 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
18406 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
18407 (BLOCK_INLINE_ENTRY_LABEL): New.
18408 (dwarf2out_var_location): Disregard inline entry markers.
18409 (inline_entry_data): New struct.
18410 (inline_entry_data_hasher): New hashtable type.
18411 (inline_entry_data_hasher::hash): New.
18412 (inline_entry_data_hasher::equal): New.
18413 (inline_entry_data_table): New variable.
18414 (add_high_low_attributes): Add DW_AT_entry_pc and
18415 DW_AT_GNU_entry_view attributes if a pending entry is found
18416 in inline_entry_data_table. Add old entry_pc attribute only
18417 if debug nonbinding markers are disabled.
18418 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
18419 markers are enabled.
18420 (block_within_block_p, dwarf2out_inline_entry): New.
18421 (dwarf2out_finish): Check that no entries remained in
18422 inline_entry_data_table.
18423 * final.c (reemit_insn_block_notes): Handle inline entry notes.
18424 (final_scan_insn, notice_source_line): Likewise.
18425 (rest_of_clean_state): Skip inline entry markers.
18426 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
18428 * gimple.c (gimple_build_debug_inline_entry): New.
18429 * gimple.h (enum gimple_debug_subcode): Add
18430 GIMPLE_DEBUG_INLINE_ENTRY.
18431 (gimple_build_debug_inline_entry): Declare.
18432 (gimple_debug_inline_entry_p): New.
18433 (gimple_debug_nonbind_marker_p): Adjust.
18434 * insn-notes.def (INLINE_ENTRY): New.
18435 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
18436 inline entry marker notes.
18437 (print_insn): Likewise.
18438 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
18439 (INSN_DEBUG_MARKER_KIND): Likewise.
18440 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
18441 * tree-inline.c (expand_call_inline): Build and insert
18442 debug_inline_entry stmt.
18443 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
18444 inline entry blocks early, if nonbind markers are enabled.
18445 (dump_scope_block): Dump fragment info.
18446 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
18447 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
18448 (gimple_build_debug_inline_entry): New.
18449 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
18450 Enable/disable inline entry points too.
18451 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
18452 (DEBUG_INSN): Describe inline entry markers.
18454 * common.opt (gvariable-location-views): New.
18455 (gvariable-location-views=incompat5): New.
18456 * config.in: Rebuilt.
18457 * configure: Rebuilt.
18458 * configure.ac: Test assembler for view support.
18459 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
18460 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
18461 * dwarf2out.c (var_loc_view): New typedef.
18462 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
18463 (dwarf2out_locviews_in_attribute): New.
18464 (dwarf2out_locviews_in_loclist): New.
18465 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
18466 (enum dw_line_info_opcode): Add LI_adv_address.
18467 (struct dw_line_info_table): Add view.
18468 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
18469 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
18470 (zero_view_p): New variable.
18471 (ZERO_VIEW_P): New macro.
18472 (output_asm_line_debug_info): New.
18473 (struct var_loc_node): Add view.
18474 (add_AT_view_list, AT_loc_list): New.
18475 (add_var_loc_to_decl): Add view param. Test it against last.
18476 (new_loc_list): Add view params. Record them.
18477 (AT_loc_list_ptr): Handle loc and view lists.
18478 (view_list_to_loc_list_val_node): New.
18479 (print_dw_val): Handle dw_val_class_view_list.
18480 (size_of_die): Likewise.
18481 (value_format): Likewise.
18482 (loc_list_has_views): New.
18483 (gen_llsym): Set vl_symbol too.
18484 (maybe_gen_llsym, skip_loc_list_entry): New.
18485 (dwarf2out_maybe_output_loclist_view_pair): New.
18486 (output_loc_list): Output view list or entries too.
18487 (output_view_list_offset): New.
18488 (output_die): Handle dw_val_class_view_list.
18489 (output_dwarf_version): New.
18490 (output_compilation_unit_header): Use it.
18491 (output_skeleton_debug_sections): Likewise.
18492 (output_rnglists, output_line_info): Likewise.
18493 (output_pubnames, output_aranges): Update version comments.
18494 (output_one_line_info_table): Output view numbers in asm comments.
18495 (dw_loc_list): Determine current endview, pass it to new_loc_list.
18496 Call maybe_gen_llsym.
18497 (loc_list_from_tree_1): Adjust.
18498 (add_AT_location_description): Create view list attribute if
18499 needed, check it's absent otherwise.
18500 (convert_cfa_to_fb_loc_list): Adjust.
18501 (maybe_emit_file): Call output_asm_line_debug_info for test.
18502 (dwarf2out_var_location): Reset views as needed. Precompute
18503 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
18504 attribute. Set view.
18505 (new_line_info_table): Reset next view.
18506 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
18507 (dwarf2out_source_line): Likewise. Output view resets and labels to
18508 the assembler, or select appropriate line info opcodes.
18509 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
18510 (optimize_string_length): Catch it. Adjust.
18511 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
18512 dw_val_class_view_list, and remove it if no longer needed.
18513 (hash_loc_list): Hash view numbers.
18514 (loc_list_hasher::equal): Compare them.
18515 (optimize_location_lists): Check whether a view list symbol is
18516 needed, and whether the locview attribute is present, and
18517 whether they match. Remove the locview attribute if no longer
18519 (index_location_lists): Call skip_loc_list_entry for test.
18520 (dwarf2out_finish): Call output_asm_line_debug_info for test.
18521 Use output_dwarf_version.
18522 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
18523 (struct dw_val_node): Add val_view_list.
18524 * final.c (SEEN_NEXT_VIEW): New.
18525 (set_next_view_needed): New.
18526 (clear_next_view_needed): New.
18527 (maybe_output_next_view): New.
18528 (final_start_function): Rename to...
18529 (final_start_function_1): ... this. Take pointer to FIRST,
18530 add SEEN parameter. Emit param bindings in the initial view.
18531 (final_start_function): Reintroduce SEEN-less interface.
18532 (final): Rename to...
18533 (final_1): ... this. Take SEEN parameter. Output final pending
18534 next view at the end.
18535 (final): Reintroduce seen-less interface.
18536 (final_scan_insn): Output pending next view before switching
18537 sections or ending a block. Mark the next view as needed when
18538 outputting variable locations. Notify debug backend of section
18539 changes, and of location view changes.
18540 (rest_of_handle_final): Adjust.
18541 * toplev.c (process_options): Autodetect value for debug variable
18542 location views option. Warn on incompat5 without -gdwarf-5.
18543 * doc/invoke.texi (gvariable-location-views): New.
18544 (gvariable-location-views=incompat5): New.
18545 (gno-variable-location-views): New.
18547 2018-02-08 David Malcolm <dmalcolm@redhat.com>
18549 PR tree-optimization/84136
18550 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
18551 that the result of find_edge is non-NULL.
18553 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
18556 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
18557 storing integer register in SImode. Fix cost of 256 and 512
18558 byte aligned SSE register store.
18560 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
18562 * config/i386/i386.c (ix86_multiplication_cost): Fix
18563 multiplication cost for TARGET_AVX512DQ.
18565 2018-02-08 Marek Polacek <polacek@redhat.com>
18567 PR tree-optimization/84238
18568 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
18571 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
18573 PR tree-optimization/84265
18574 * tree-vect-stmts.c (vectorizable_store): Don't treat
18575 VMAT_CONTIGUOUS accesses as grouped.
18576 (vectorizable_load): Likewise.
18578 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
18580 PR tree-optimization/81635
18581 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
18582 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
18583 (test_round_for_mask): New functions.
18584 (wide_int_cc_tests): Call test_round_for_mask.
18585 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
18586 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
18587 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
18588 range returned by get_range_info.
18590 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
18593 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
18594 * symtab.c: Include builtins.h
18595 (symtab_node::output_to_lto_symbol_table_p): Move here
18596 from lto-streamer-out.c:output_symbol_p.
18597 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
18598 (output_symbol_p): Move all logic to symtab.c
18599 (produce_symtab): Update.
18601 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
18603 * config/s390/s390-opts.h (enum indirect_branch): Define.
18604 * config/s390/s390-protos.h (s390_return_addr_from_memory)
18605 (s390_indirect_branch_via_thunk)
18606 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
18607 (enum s390_indirect_branch_type): Define.
18608 * config/s390/s390.c (struct s390_frame_layout, struct
18609 machine_function): Remove.
18610 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
18611 (indirect_branch_table_label_no, indirect_branch_table_name):
18613 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
18614 (enum s390_indirect_branch_option): Define.
18615 (s390_return_addr_from_memory): New function.
18616 (s390_handle_string_attribute): New function.
18617 (s390_attribute_table): Add new attribute handler.
18618 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
18619 (s390_indirect_branch_via_thunk): New function.
18620 (s390_indirect_branch_via_inline_thunk): New function.
18621 (s390_function_ok_for_sibcall): When jumping via thunk disallow
18622 sibling call optimization for non z10 compiles.
18623 (s390_emit_call): Force indirect branch target to be a single
18624 register. Add r1 clobber for non-z10 compiles.
18625 (s390_emit_epilogue): Emit return jump via return_use expander.
18626 (s390_reorg): Handle JUMP_INSNs as execute targets.
18627 (s390_option_override_internal): Perform validity checks for the
18628 new command line options.
18629 (s390_indirect_branch_attrvalue): New function.
18630 (s390_indirect_branch_settings): New function.
18631 (s390_set_current_function): Invoke s390_indirect_branch_settings.
18632 (s390_output_indirect_thunk_function): New function.
18633 (s390_code_end): Implement target hook.
18634 (s390_case_values_threshold): Implement target hook.
18635 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
18637 * config/s390/s390.h (struct s390_frame_layout)
18638 (struct machine_function): Move here from s390.c.
18639 (TARGET_INDIRECT_BRANCH_NOBP_RET)
18640 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
18641 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
18642 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
18643 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
18644 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
18645 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
18646 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
18647 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
18648 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
18649 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
18650 (mnemonic attribute): Add values which aren't recognized
18652 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
18653 pattern for branch conversion. Fix mnemonic attribute.
18654 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
18655 indirect branch via thunk if requested.
18656 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
18657 ("*indirect_jump"): Disable for branch conversion using out of
18659 ("indirect_jump_via_thunk<mode>_z10")
18660 ("indirect_jump_via_thunk<mode>")
18661 ("indirect_jump_via_inlinethunk<mode>_z10")
18662 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
18663 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
18664 ("casesi_jump_via_inlinethunk<mode>_z10")
18665 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
18666 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
18667 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
18668 ("*indirect2_jump"): Disable for branch conversion.
18669 ("casesi_jump"): Turn into expander and expand patterns for branch
18671 ("return_use"): New expander.
18672 ("*return"): Emit return via thunk and rename it to ...
18673 ("*return<mode>"): ... this one.
18674 * config/s390/s390.opt: Add new options and and enum for the
18677 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
18679 * lra-constraints.c (match_reload): Unconditionally use
18680 gen_lowpart_SUBREG, rather than selecting between that
18681 and equivalent gen_rtx_SUBREG code.
18683 2018-02-08 Richard Biener <rguenther@suse.de>
18685 PR tree-optimization/84233
18686 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
18687 changed flag instead of boguously re-using phi_inserted.
18689 2018-02-08 Martin Jambor <mjambor@suse.cz>
18691 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
18692 static local variables.
18694 2018-02-08 Richard Biener <rguenther@suse.de>
18696 PR tree-optimization/84278
18697 * tree-vect-stmts.c (vectorizable_store): When looking for
18698 smaller vector types to perform grouped strided loads/stores
18699 make sure the mode is supported by the target.
18700 (vectorizable_load): Likewise.
18702 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
18704 * config/aarch64/aarch64.c (aarch64_components_for_bb):
18705 Increase LDP/STP opportunities by adding adjacent callee-saves.
18707 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
18709 PR rtl-optimization/84068
18710 PR rtl-optimization/83459
18711 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
18713 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
18715 PR tree-optimization/84224
18716 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
18717 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
18718 non-zero arguments.
18720 2018-02-07 Iain Sandoe <iain@codesourcery.com>
18723 * config/rs6000/altivec.md (*restore_world): Remove LR use.
18724 * config/rs6000/predicates.md (restore_world_operation): Adjust op
18725 count, remove one USE.
18727 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
18729 * doc/install.texi (Configuration): Document the
18730 --with-long-double-format={ibm,ieee} PowerPC configuration
18734 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
18735 Convert from define_expand to be define_insn_and_split. Rework
18736 float/double/_Float128 conversions to QI/HI/SImode to work with
18737 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
18738 conversions to QI/HImode types did a store and then a load to
18739 truncate the value. For conversions to VSX registers, don't split
18740 the insn, instead emit the code directly. Use the code iterator
18741 any_fix to combine signed and unsigned conversions.
18742 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
18743 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
18744 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
18745 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
18746 (fix_<mode>di2_hw): Likewise.
18747 (fixuns_<mode>di2_hw): Likewise.
18748 (fix_<mode>si2_hw): Likewise.
18749 (fixuns_<mode>si2_hw): Likewise.
18750 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
18751 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
18752 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
18753 fix<uns>_trunc<SFDF:mode>si2_p8.
18754 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
18755 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
18756 (fix<uns>_<mode>_mem): Likewise.
18757 (fctiw<u>z_<mode>_mem): Likewise.
18758 (fix<uns>_<mode>_mem): Likewise.
18759 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
18760 the register allocator from doing a direct move to the GPRs to do
18761 a store, and instead use the ISA 3.0 store byte/half-word from
18762 vector register instruction. For IEEE 128-bit floating point,
18763 also optimize stores of 32-bit ints.
18764 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
18766 2018-02-07 Alan Hayward <alan.hayward@arm.com>
18768 * genextract.c (push_pathstr_operand): New function to support
18770 (walk_rtx): Call push_pathstr_operand.
18771 (print_path): Support [a-zA-Z].
18773 2018-02-07 Richard Biener <rguenther@suse.de>
18775 PR tree-optimization/84037
18776 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
18777 (cse_and_gimplify_to_preheader): Declare.
18778 (vect_get_place_in_interleaving_chain): Likewise.
18779 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
18781 (_loop_vec_info::~_loop_vec_info): Delete it.
18782 (cse_and_gimplify_to_preheader): New function.
18783 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
18784 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
18785 (vectorizable_load): Likewise. For grouped stores always base
18786 the IV on the first element.
18787 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
18788 condition before gimplifying.
18790 2018-02-07 Jakub Jelinek <jakub@redhat.com>
18792 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
18793 *DIV_EXPR and *MOD_EXPR.
18795 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
18798 * config/i386/i386.c (ix86_option_override_internal): Mask out
18799 the CF_SET bit when checking -fcf-protection.
18801 2018-02-07 Tom de Vries <tom@codesourcery.com>
18804 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
18807 2018-02-07 Richard Biener <rguenther@suse.de>
18809 PR tree-optimization/84204
18810 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
18813 PR tree-optimization/84205
18814 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
18815 special-case isl_ast_op_zdiv_r.
18817 PR tree-optimization/84223
18818 * graphite-scop-detection.c (gather_bbs::before_dom_children):
18819 Only add conditions from within the region.
18820 (gather_bbs::after_dom_children): Adjust.
18822 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
18825 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
18826 * config/avr/avr.md: Only post-reload split REG-REG moves if
18827 either register is GENERAL_REG_P.
18829 2018-02-07 Jakub Jelinek <jakub@redhat.com>
18831 PR tree-optimization/84235
18832 * tree-ssa-scopedtables.c
18833 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
18834 if the subtraction is performed in floating point type where NaNs are
18835 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
18836 build 1. Formatting fix.
18838 2018-02-06 Jakub Jelinek <jakub@redhat.com>
18841 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
18842 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
18843 and skip it regardless of bb boundaries. Use CALL_P macro,
18844 don't test INSN_P (insn) together with CALL_P or JUMP_P check
18845 unnecessarily, formatting fix.
18847 2018-02-06 Michael Collison <michael.collison@arm.com>
18849 * config/arm/thumb2.md:
18850 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
18851 (*thumb_mov_notscc): Ditto.
18853 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
18856 * config/rs6000/rs6000.md (su code attribute): Use "u" for
18857 unsigned_fix, not "s".
18859 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
18861 * configure.ac (gcc_fn_eh_frame_ro): New function.
18862 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
18863 correct .eh_frame permissions.
18864 * configure: Regenerate.
18866 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
18868 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
18869 irrelevant options.
18871 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18873 * config/rs6000/rs6000.c (rs6000_option_override_internal):
18874 Display warning message for -mno-speculate-indirect-jumps.
18876 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
18878 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
18880 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
18882 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
18884 PR tree-optimization/84225
18885 * tree-eh.c (find_trapping_overflow): Only call
18886 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
18888 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
18891 * config/i386/i386.c: Reimplement the check of possible options
18892 -mibt/-mshstk conbination. Change error messages.
18893 * doc/invoke.texi: Fix a typo: remove extra '='.
18895 2018-02-06 Marek Polacek <polacek@redhat.com>
18897 PR tree-optimization/84228
18898 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
18900 2018-02-06 Tamar Christina <tamar.christina@arm.com>
18903 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
18904 emitted arch directives.
18905 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
18906 __ARM_FEATURE_COPROC before changing architectures.
18908 2018-02-06 Richard Biener <rguenther@suse.de>
18910 * config/i386/i386.c (print_reg): Fix typo.
18911 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
18913 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
18915 * configure: Regenerate.
18917 2018-02-05 Martin Sebor <msebor@redhat.com>
18919 PR tree-optimization/83369
18920 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
18923 2018-02-05 Martin Liska <mliska@suse.cz>
18925 * doc/invoke.texi: Cherry-pick upstream r323995.
18927 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
18929 * ira.c (ira_init_register_move_cost): Adjust comment.
18931 2018-02-05 Martin Liska <mliska@suse.cz>
18933 PR gcov-profile/84137
18934 * doc/gcov.texi: Fix typo in documentation.
18936 2018-02-05 Martin Liska <mliska@suse.cz>
18938 PR gcov-profile/83879
18939 * doc/gcov.texi: Document necessity of --dynamic-list-data when
18940 using dlopen functionality.
18942 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
18944 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
18945 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
18946 _mm_maskz_range_ss, _mm_mask_range_round_ss,
18947 _mm_maskz_range_round_ss): New intrinsics.
18948 (__builtin_ia32_rangesd128_round)
18949 (__builtin_ia32_rangess128_round): Remove.
18950 (__builtin_ia32_rangesd128_mask_round,
18951 __builtin_ia32_rangess128_mask_round): New builtins.
18952 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
18953 __builtin_ia32_rangess128_round): Remove.
18954 (__builtin_ia32_rangesd128_mask_round,
18955 __builtin_ia32_rangess128_mask_round): New builtins.
18956 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
18957 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
18958 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
18959 "<round_saeonly_constraint>")): Changed to ...
18960 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
18961 "<round_saeonly_scalar_constraint>")): ... this.
18962 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
18963 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
18964 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
18965 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
18966 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
18968 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
18970 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
18972 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
18973 Remove all values except native, 8540 and 8548.
18975 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
18977 * config/i386/i386.c (ix86_output_function_return): Pass
18978 INVALID_REGNUM, instead of -1, as invalid register number to
18979 indirect_thunk_name and output_indirect_thunk.
18981 2018-02-02 Julia Koval <julia.koval@intel.com>
18983 * config.gcc: Add -march=icelake.
18984 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
18985 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
18986 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
18987 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
18988 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
18989 (processor_target_table): Add icelake.
18990 (ix86_option_override_internal): Handle new PTAs.
18991 (get_builtin_code_for_version): Handle icelake.
18992 (M_INTEL_COREI7_ICELAKE): New.
18993 (fold_builtin_cpu): Handle icelake.
18994 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
18995 * doc/invoke.texi: Add -march=icelake.
18997 2018-02-02 Julia Koval <julia.koval@intel.com>
18999 * config/i386/i386.c (ix86_option_override_internal): Change flags type
19000 to wide_int_bitmask.
19001 * wide-int-bitmask.h: New.
19003 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
19006 * config/i386/i386.md: Replace Pmode with word_mode in
19007 builtin_setjmp_setup and builtin_longjmp to support x32.
19009 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
19013 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
19015 (rs6000_supported_cpu_names): New static variable.
19016 (linux_cpu_translation_table): Likewise.
19017 (elf_platform) <cpu>: Define new static variable and use it.
19018 Translate kernel AT_PLATFORM name to canonical name if needed.
19019 Error if platform name is unknown.
19021 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
19024 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
19026 2018-02-01 Jeff Law <law@redhat.com>
19029 * config/i386/i386.c (release_scratch_register_on_entry): Add new
19030 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
19031 the scratch if RELEASE_VIA_POP is false.
19032 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
19033 If we have to save a temporary register, decrement SIZE appropriately.
19034 Pass new arguments to release_scratch_register_on_entry.
19035 (ix86_adjust_stack_and_probe): Likewise.
19036 (ix86_emit_probe_stack_range): Pass new arguments to
19037 release_scratch_register_on_entry.
19039 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
19041 PR rtl-optimization/84157
19042 * combine.c (change_zero_ext): Use REG_P predicate in
19043 front of HARD_REGISTER_P predicate.
19045 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
19047 * config/avr/avr.c (avr_option_override): Move disabling of
19048 -fdelete-null-pointer-checks to...
19049 * common/config/avr/avr-common.c (avr_option_optimization_table):
19052 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
19054 PR tree-optimization/81635
19055 * tree-data-ref.c (split_constant_offset_1): For types that
19056 wrap on overflow, try to use range info to prove that wrapping
19059 2018-02-01 Renlin Li <renlin.li@arm.com>
19062 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
19063 TAILCALL_ADDR_REGS.
19064 (aarch64_register_move_cost): Likewise.
19065 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
19066 TAILCALL_ADDR_REGS.
19067 (REG_CLASS_NAMES): Likewise.
19068 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
19069 TAILCALL_ADDR_REGS. Remove IP registers.
19070 * config/aarch64/aarch64.md (Ucs): Update register constraint.
19072 2018-02-01 Richard Biener <rguenther@suse.de>
19074 * domwalk.h (dom_walker::dom_walker): Add additional constructor
19075 for specifying RPO order and allow NULL for that.
19076 * domwalk.c (dom_walker::dom_walker): Likewise.
19077 (dom_walker::walk): Handle NULL RPO order.
19078 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
19080 (rewrite_update_dom_walker): Likewise.
19081 (mark_def_dom_walker): Likewise.
19083 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
19085 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
19086 (aarch64_maybe_expand_sve_subreg_move): Declare.
19087 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
19088 * config/aarch64/predicates.md (aarch64_any_register_operand): New
19090 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
19091 that are semantically a reverse operation.
19092 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
19093 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
19094 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
19096 (aarch64_can_change_mode_class): For big-endian, forbid changes
19097 between two SVE modes if they have different element sizes.
19099 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
19101 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
19102 the TImode handling for big-endian targets.
19104 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
19106 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
19107 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
19109 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
19110 Remove BSWAP handing for big-endian targets and use the form of
19111 LD1RQ appropariate for the mode.
19113 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
19115 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
19116 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
19117 duplicated element.
19119 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
19122 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
19123 check for operands that need to go through aarch64_sve_reload_be.
19125 2018-02-01 Jakub Jelinek <jakub@redhat.com>
19127 PR tree-optimization/81661
19128 PR tree-optimization/84117
19129 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
19130 * tree-eh.c: Include gimplify.h.
19131 (find_trapping_overflow, replace_trapping_overflow,
19132 rewrite_to_non_trapping_overflow): New functions.
19133 * tree-vect-loop.c: Include tree-eh.h.
19134 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
19135 * tree-data-ref.c: Include tree-eh.h.
19136 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
19138 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
19140 PR rtl-optimization/84123
19141 * combine.c (change_zero_ext): Check if hard register satisfies
19142 can_change_dest_mode before calling gen_lowpart_SUBREG.
19144 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
19147 * ira.c (ira_init_register_move_cost): Remove assert.
19149 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
19151 PR rtl-optimization/84071
19152 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
19153 * doc/tm.texi: Regenerate.
19155 2018-01-31 Richard Biener <rguenther@suse.de>
19157 PR tree-optimization/84132
19158 * tree-data-ref.c (analyze_miv_subscript): Properly
19159 check whether evolution_function_is_affine_multivariate_p
19160 before calling gcd_of_steps_may_divide_p.
19162 2018-01-31 Julia Koval <julia.koval@intel.com>
19165 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
19166 * config/i386/i386.md (rdpid_rex64) New.
19167 (rdpid): Make 32bit only.
19169 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
19172 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
19173 an IDENTIFIER_NODE for FUNCTION_TYPE's.
19175 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
19178 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
19180 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
19182 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
19184 PR rtl-optimization/84071
19185 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
19186 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
19188 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
19190 * config/arc/arc.c (arc_handle_aux_attribute): New function.
19191 (arc_attribute_table): Add 'aux' attribute.
19192 (arc_in_small_data_p): Consider aux like variables.
19193 (arc_is_aux_reg_p): New function.
19194 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
19195 (arc_get_aux_arg): New function.
19196 (prepare_move_operands): Handle aux-register access.
19197 (arc_handle_aux_attribute): New function.
19198 * doc/extend.texi (ARC Variable attributes): Add subsection.
19200 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
19202 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
19203 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
19204 (arc_attribute_table): Add 'uncached' attribute.
19205 (arc_print_operand): Print '.di' flag for uncached memory
19207 (arc_in_small_data_p): Do not consider for small data the uncached
19209 (arc_is_uncached_mem_p): New function.
19210 * config/arc/predicates.md (compact_store_memory_operand): Check
19211 for uncached memory accesses.
19212 (nonvol_nonimm_operand): Likewise.
19213 * doc/extend.texi (ARC Type Attribute): New subsection.
19215 2018-01-31 Jakub Jelinek <jakub@redhat.com>
19218 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
19219 falign-loops=): Add Optimization flag.
19221 2018-01-30 Jeff Law <law@redhat.com>
19224 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
19225 INT_REGISTERS_SAVED. Check it prior to calling
19226 get_scratch_register_on_entry.
19227 (ix86_adjust_stack_and_probe): Similarly.
19228 (ix86_emit_probe_stack_range): Similarly.
19229 (ix86_expand_prologue): Corresponding changes.
19231 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
19234 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
19235 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
19237 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
19240 * lra-constraints.c (curr_insn_transform): Process AND in the
19243 2018-01-30 Jakub Jelinek <jakub@redhat.com>
19245 PR rtl-optimization/83986
19246 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
19247 dependence against last_pending_memory_flush in addition to
19248 pending_jump_insns.
19250 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
19252 PR tree-optimization/81611
19253 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
19256 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
19259 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
19262 2018-01-30 Richard Biener <rguenther@suse.de>
19263 Jakub Jelinek <jakub@redhat.com>
19265 PR tree-optimization/84111
19266 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
19267 inner loops added during recursion, as they don't have up-to-date
19270 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
19273 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
19274 (can_inline_edge_by_limits_p): ... here.
19275 (can_early_inline_edge_p, check_callers,
19276 update_caller_keys, update_callee_keys, recursive_inlining,
19277 add_new_edges_to_heap, speculation_useful_p,
19278 inline_small_functions,
19279 inline_small_functions, flatten_function,
19280 inline_to_all_callers_1): Update.
19282 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
19284 * profile-count.c (profile_count::combine_with_ipa_count): Handle
19287 2018-01-30 Richard Biener <rguenther@suse.de>
19289 PR tree-optimization/83008
19290 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
19291 invariant and constant vector uses in stmts when they need
19292 more than one stmt.
19294 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
19297 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
19298 * configure: Regenerate.
19300 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
19302 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
19304 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
19305 Use gen_rtx_REG rather than gen_lowpart.
19307 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
19309 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
19310 rather than 0 when creating partial subregs.
19312 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
19314 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
19317 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
19320 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
19321 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
19322 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
19323 flags. This restores the settings used before the 2017-07-24.
19324 Turning off pre increment/decrement/modify allows IVOPTS to
19325 optimize DF/SF loops where the index is an int.
19327 2018-01-29 Richard Biener <rguenther@suse.de>
19328 Kelvin Nilsen <kelvin@gcc.gnu.org>
19331 * tree-vect-stmts.c (vectorizable_call): Don't call
19332 targetm.vectorize_builtin_md_vectorized_function if callee is
19335 2018-01-22 Carl Love <cel@us.ibm.com>
19337 * doc/extend.tex: Fix typo in second arg in
19338 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
19340 2018-01-29 Richard Biener <rguenther@suse.de>
19342 PR tree-optimization/84086
19343 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
19344 (flush_ssaname_freelist): When SSA names were released reset
19345 the SCEV hash table.
19347 2018-01-29 Richard Biener <rguenther@suse.de>
19349 PR tree-optimization/84057
19350 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
19351 removed paths when removing edges.
19353 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
19355 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
19356 -mfunction-return=@var{choice}.
19358 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
19360 PR diagnostic/84034
19361 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
19362 Handle CR like TAB.
19363 (layout::print_source_line): Likewise.
19364 (test_get_line_width_without_trailing_whitespace): Add test cases.
19366 2018-01-27 Jakub Jelinek <jakub@redhat.com>
19368 PR middle-end/84040
19369 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
19372 2018-01-26 Jim Wilson <jimw@sifive.com>
19374 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
19376 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
19379 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19381 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
19382 and CMP + SUB-immediate -> SUBS.
19384 2018-01-26 Martin Sebor <msebor@redhat.com>
19386 PR tree-optimization/83896
19387 * tree-ssa-strlen.c (get_string_len): Rename...
19388 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
19389 Avoid assuming length is constant.
19390 (handle_char_store): Use HOST_WIDE_INT for string length.
19392 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
19395 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
19396 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
19398 2018-01-26 Richard Biener <rguenther@suse.de>
19400 PR rtl-optimization/84003
19401 * dse.c (record_store): Only record redundant stores when
19402 the earlier store aliases at least all accesses the later one does.
19404 2018-01-26 Jakub Jelinek <jakub@redhat.com>
19406 PR rtl-optimization/83985
19407 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
19408 REG_CFA_RESTORE insns.
19409 (delete_unmarked_insns): Don't ignore separate shrink wrapping
19410 REG_CFA_RESTORE insns here.
19413 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
19414 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
19416 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
19418 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
19419 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
19420 (arc_init): Likewise.
19421 (arc_override_options): Likewise.
19422 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
19424 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
19426 * config/arc/arc.h (TARGET_DBNZ): Define.
19427 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
19428 properly set the tune attribute.
19429 (dbnz): Use TARGET_DBNZ guard.
19430 * config/arc/arc.opt (mtune): Add core3 option.
19432 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
19434 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
19435 recognize new pic like addresses.
19436 (arc_delegitimize_address): Clean up.
19438 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
19440 * config/arc/arc-arches.def: Option mrf16 valid for all
19442 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
19443 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
19444 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
19445 * config/arc/arc-tables.opt: Regenerate.
19446 * config/arc/arc.c (arc_conditional_register_usage): Handle
19447 reduced register file case.
19448 (arc_file_start): Set must have build attributes.
19449 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
19450 mrf16 option value.
19451 * config/arc/arc.opt (mrf16): Add new option.
19452 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
19453 * config/arc/genmultilib.awk: Handle new mrf16 option.
19454 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
19455 * config/arc/t-multilib: Regenerate.
19456 * doc/invoke.texi (ARC Options): Document mrf16 option.
19458 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
19460 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
19461 * config/arc/arc.c (arc_handle_secure_attribute): New function.
19462 (arc_attribute_table): Add 'secure_call' attribute.
19463 (arc_print_operand): Print secure call operand.
19464 (arc_function_ok_for_sibcall): Don't optimize tail calls when
19466 (arc_is_secure_call_p): New function. * config/arc/arc.md
19467 (call_i): Add support for sjli instruction.
19468 (call_value_i): Likewise.
19469 * config/arc/constraints.md (Csc): New constraint.
19471 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
19472 John Eric Martin <John.Martin@emmicro-us.com>
19474 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
19475 * config/arc/arc.c (_arc_jli_section): New struct.
19476 (arc_jli_section): New type.
19477 (rc_jli_sections): New static variable.
19478 (arc_handle_jli_attribute): New function.
19479 (arc_attribute_table): Add jli_always and jli_fixed attribute.
19480 (arc_file_end): New function.
19481 (TARGET_ASM_FILE_END): Define.
19482 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
19483 (arc_add_jli_section): New function.
19484 (jli_call_scan): Likewise.
19485 (arc_reorg): Call jli_call_scan.
19486 (arc_output_addsi): Remove 'S' from printing asm operand.
19487 (arc_is_jli_call_p): New function.
19488 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
19490 (movhi_insn): Likewise.
19491 (movsi_insn): Likewise.
19492 (movsi_set_cc_insn): Likewise.
19493 (loadqi_update): Likewise.
19494 (load_zeroextendqisi_update): Likewise.
19495 (load_signextendqisi_update): Likewise.
19496 (loadhi_update): Likewise.
19497 (load_zeroextendhisi_update): Likewise.
19498 (load_signextendhisi_update): Likewise.
19499 (loadsi_update): Likewise.
19500 (loadsf_update): Likewise.
19501 (movsicc_insn): Likewise.
19502 (bset_insn): Likewise.
19503 (bxor_insn): Likewise.
19504 (bclr_insn): Likewise.
19505 (bmsk_insn): Likewise.
19506 (bicsi3_insn): Likewise.
19507 (cmpsi_cc_c_insn): Likewise.
19508 (movsi_ne): Likewise.
19509 (movsi_cond_exec): Likewise.
19510 (clrsbsi2): Likewise.
19511 (norm_f): Likewise.
19521 (call_i): Remove 'S' asm letter, add jli instruction.
19522 (call_value_i): Likewise.
19523 * config/arc/arc.op (mjli-always): New option.
19524 * config/arc/constraints.md (Cji): New constraint.
19525 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
19527 (subsf3_fpx): Likewise.
19528 (mulsf3_fpx): Likewise.
19529 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
19531 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
19532 function attrbutes.
19533 * doc/invoke.texi (ARC): Document mjli-always option.
19535 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
19537 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
19538 avoid addition with 0 and use incw and decw where possible.
19540 2018-01-26 Richard Biener <rguenther@suse.de>
19542 PR tree-optimization/81082
19543 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
19544 association if it requires casting to unsigned.
19545 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
19546 from fold_plusminus_mult_expr to catch important cases late when
19547 range info is available.
19549 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
19551 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
19552 * configure.ac (hidden_linkonce): New test.
19553 * configure: Regenerate.
19554 * config.in: Regenerate.
19556 2018-01-26 Julia Koval <julia.koval@intel.com>
19558 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
19559 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
19560 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
19561 _mm_mask_bitshuffle_epi64_mask): Fix type.
19562 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
19563 USI_FTYPE_V4DI_V4DI_USI): Remove.
19564 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
19565 __builtin_ia32_vpshufbitqmb256_mask,
19566 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
19567 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
19568 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
19570 2018-01-26 Alan Modra <amodra@gmail.com>
19573 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
19574 UNSPEC_VBPERMQ. Sort other unspecs.
19576 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
19578 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
19580 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
19582 PR middle-end/83055
19583 * predict.c (drop_profile): Do not push/pop cfun; update also
19585 (handle_missing_profiles): Fix logic looking for zero profiles.
19587 2018-01-25 Jakub Jelinek <jakub@redhat.com>
19589 PR middle-end/83977
19590 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
19591 on functions with #pragma omp declare simd or functions with simd
19593 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
19594 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
19595 Remove trailing \n from warning_at calls.
19597 2018-01-25 Tom de Vries <tom@codesourcery.com>
19600 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
19601 for neutered workers.
19603 2018-01-24 Joseph Myers <joseph@codesourcery.com>
19606 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
19607 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
19609 2018-01-24 Jeff Law <law@redhat.com>
19612 * i386.c (get_probe_interval): Move to earlier point.
19613 (ix86_compute_frame_layout): If -fstack-clash-protection and
19614 the frame is larger than the probe interval, then use pushes
19615 to save registers rather than reg->mem moves.
19616 (ix86_expand_prologue): Remove conditional for int_registers_saved
19619 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
19622 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
19623 min/max for never referenced object.
19625 2018-01-24 Jakub Jelinek <jakub@redhat.com>
19627 PR middle-end/83977
19628 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
19630 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
19631 attributes from DECL_ATTRIBUTES (decl) without affecting
19632 DECL_ATTRIBUTES (current_function_decl).
19633 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
19634 functions with non-NULL DECL_ABSTRACT_ORIGIN.
19636 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
19638 PR tree-optimization/83979
19639 * fold-const.c (fold_comparison): Use constant_boolean_node
19640 instead of boolean_{true,false}_node.
19642 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
19644 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
19647 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
19649 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
19650 Simplify the clause that sets the length attribute.
19651 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
19652 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
19653 clause that sets the length attribute.
19654 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
19656 2018-01-24 Tom de Vries <tom@codesourcery.com>
19659 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
19660 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
19661 Add strict parameter.
19662 (prevent_branch_around_nothing): Insert dummy insn between branch to
19663 label and label with no ptx insn inbetween.
19664 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
19666 2018-01-24 Tom de Vries <tom@codesourcery.com>
19669 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
19670 for neutered threads in warp.
19671 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
19673 2018-01-24 Richard Biener <rguenther@suse.de>
19675 PR tree-optimization/83176
19676 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
19679 2018-01-24 Richard Biener <rguenther@suse.de>
19681 PR tree-optimization/82819
19682 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
19683 code generating pluses that are no-ops in the target precision.
19685 2018-01-24 Richard Biener <rguenther@suse.de>
19687 PR middle-end/84000
19688 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
19690 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
19692 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
19693 to merge probabilities.
19694 * predict.c (probably_never_executed): Also mark as cold functions
19695 with global 0 profile and guessed local profile.
19696 * profile-count.c (profile_probability::combine_with_count): New
19698 * profile-count.h (profile_probability::operator*,
19699 profile_probability::operator*=, profile_probability::operator/,
19700 profile_probability::operator/=): Reduce precision to adjusted
19701 and set value to guessed on contradictory divisions.
19702 (profile_probability::combine_with_freq): Remove.
19703 (profile_probability::combine_wiht_count): Declare.
19704 (profile_count::force_nonzero):: Set to adjusted.
19705 (profile_count::probability_in):: Set quality to adjusted.
19706 * tree-ssa-tail-merge.c (replace_block_by): Use
19707 combine_with_count.
19709 2018-01-23 Andrew Waterman <andrew@sifive.com>
19710 Jim Wilson <jimw@sifive.com>
19712 * config/riscv/riscv.c (riscv_stack_boundary): New.
19713 (riscv_option_override): Set riscv_stack_boundary. Handle
19714 riscv_preferred_stack_boundary_arg.
19715 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
19716 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
19717 (STACK_BOUNDARY): Set to riscv_stack_boundary.
19718 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
19719 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
19720 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
19722 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
19725 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
19726 of struct ix86_frame.
19727 (ix86_expand_epilogue): Likewise. Add a local variable for
19728 the reg_save_offset field in struct ix86_frame.
19730 2018-01-23 Bin Cheng <bin.cheng@arm.com>
19732 PR tree-optimization/82604
19733 * tree-loop-distribution.c (enum partition_kind): New enum item
19734 PKIND_PARTIAL_MEMSET.
19735 (partition_builtin_p): Support above new enum item.
19736 (generate_code_for_partition): Ditto.
19737 (compute_access_range): Differentiate cases that equality can be
19738 proven at all loops, the innermost loops or no loops.
19739 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
19740 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
19741 (finalize_partitions, distribute_loop): Don't fuse partition of
19742 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
19743 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
19744 parloop is enabled.
19746 2018-01-23 Martin Liska <mliska@suse.cz>
19748 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
19749 order to ignore the predictor.
19750 (PRED_POLYMORPHIC_CALL): Likewise.
19751 (PRED_RECURSIVE_CALL): Likewise.
19753 2018-01-23 Martin Liska <mliska@suse.cz>
19755 * tree-profile.c (tree_profiling): Print function header to
19756 aware reader which function we are working on.
19757 * value-prof.c (gimple_find_values_to_profile): Do not print
19758 not interesting value histograms.
19760 2018-01-23 Martin Liska <mliska@suse.cz>
19762 * profile-count.h (enum profile_quality): Add
19763 profile_uninitialized as the first value. Do not number values
19764 as they are zero based.
19765 (profile_count::verify): Update sanity check.
19766 (profile_probability::verify): Likewise.
19768 2018-01-23 Nathan Sidwell <nathan@acm.org>
19770 * doc/invoke.texi (ffor-scope): Deprecate.
19772 2018-01-23 David Malcolm <dmalcolm@redhat.com>
19774 PR tree-optimization/83510
19775 * domwalk.c (set_all_edges_as_executable): New function.
19776 (dom_walker::dom_walker): Convert bool param
19777 "skip_unreachable_blocks" to enum reachability. Move setup of
19778 edge flags to set_all_edges_as_executable and only do it when
19779 reachability is REACHABLE_BLOCKS.
19780 * domwalk.h (enum dom_walker::reachability): New enum.
19781 (dom_walker::dom_walker): Convert bool param
19782 "skip_unreachable_blocks" to enum reachability.
19783 (set_all_edges_as_executable): New decl.
19784 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
19785 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
19787 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
19788 but converting true to REACHABLE_BLOCKS.
19789 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
19791 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
19792 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
19793 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
19795 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
19796 if check_all_array_refs will be called.
19798 2018-01-23 David Malcolm <dmalcolm@redhat.com>
19800 * tree.c (selftest::test_location_wrappers): Add more test
19803 2018-01-23 David Malcolm <dmalcolm@redhat.com>
19805 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
19806 (selftest::test_bit_in_range): Likewise.
19808 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
19811 * doc/sourcebuild.texi (vect_float): Say that the selector
19812 only describes the situation when -funsafe-math-optimizations is on.
19813 (vect_float_strict): Document.
19815 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
19817 PR tree-optimization/83965
19818 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
19819 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
19820 instead of checking only for a reduction.
19821 (vect_recog_widen_sum_pattern): Likewise.
19823 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
19825 * predict.c (probably_never_executed): Only use precise profile info.
19826 (compute_function_frequency): Skip after inlining hack since we now
19827 have quality checking.
19829 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
19831 * profile-count.h (profile_probability::very_unlikely,
19832 profile_probability::unlikely, profile_probability::even): Set
19833 precision to guessed.
19835 2018-01-23 Richard Biener <rguenther@suse.de>
19837 PR tree-optimization/83963
19838 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
19839 Properly terminate dominator walk when crossing the exit edge not
19840 when visiting its source block.
19842 2018-01-23 Jakub Jelinek <jakub@redhat.com>
19845 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
19846 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
19848 2018-01-22 Jakub Jelinek <jakub@redhat.com>
19850 PR tree-optimization/83957
19851 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
19852 semicolon after for body surrounded by braces.
19854 PR tree-optimization/83081
19855 * profile-count.h (profile_probability::split): New method.
19856 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
19857 Use profile_probability::split.
19858 (do_compare_rtx_and_jump): Fix adjustment of probabilities
19859 when splitting a single conditional jump into 2.
19861 2018-01-22 David Malcolm <dmalcolm@redhat.com>
19863 PR tree-optimization/69452
19864 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
19867 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
19869 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
19870 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
19871 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
19873 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
19875 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
19877 * config/rl78/rl78.md (movdi): New define_expand.
19878 * config/rl78/rl78.c (rl78_split_movdi): New function.
19880 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
19883 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
19885 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
19886 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
19887 128-bit to produce an UNSPEC move to get the double word with the
19888 signbit and then a shift directly to do signbit.
19889 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
19890 implementation with a new version that just does either a direct
19891 move or a regular move. Move memory interface to separate insns.
19892 Move insns so they are next to the expander.
19893 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
19894 with signbit move. Split big and little endian case.
19895 (signbit<mode>2_dm_mem_le): Likewise.
19896 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
19897 (signbit<mode>2_dm2): Likewise.
19899 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
19901 * config/rl78/rl78.md (anddi3): New define_expand.
19903 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
19905 * config/rl78/rl78.md (umindi3): New define_expand.
19907 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
19909 * config/rl78/rl78.md (smindi3): New define_expand.
19911 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
19913 * config/rl78/rl78.md (smaxdi3): New define_expand.
19915 2018-01-22 Carl Love <cel@us.ibm.com>
19917 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
19918 LVX_V1TI): Add macro expansion.
19919 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
19920 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
19921 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
19922 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
19923 Change check to determine if the instruction is a byte reversing
19924 entry. Fix typo in comment.
19925 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
19926 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
19927 Add def_builtin calls for new builtins.
19928 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
19929 Add define_insn expansion.
19931 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
19933 * config/rl78/rl78.md (umaxdi3): New define_expand.
19935 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
19937 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
19938 for non-QImode registers.
19940 2018-01-22 Richard Biener <rguenther@suse.de>
19942 PR tree-optimization/83963
19943 * graphite-scop-detection.c (scop_detection::get_sese): Delay
19944 including the loop exit block.
19945 (scop_detection::merge_sese): Likewise.
19946 (scop_detection::add_scop): Do it here instead.
19948 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19950 * doc/sourcebuild.texi (arm_softfloat): Document.
19952 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
19955 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
19956 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
19957 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
19959 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
19960 David Edelsohn <dje.gcc@gmail.com>
19963 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
19964 Change "crset eq" to "crset 2".
19965 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
19966 (*call_indirect_aix<mode>_nospec): Likewise.
19967 (*call_value_indirect_aix<mode>_nospec): Likewise.
19968 (*call_indirect_elfv2<mode>_nospec): Likewise.
19969 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
19970 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
19971 change assembly output from . to $.
19972 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
19973 (indirect_jump<mode>_nospec): Change assembly output from . to $.
19974 (*tablejump<mode>_internal1_nospec): Likewise.
19976 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
19979 * config/sh/sh_optimize_sett_clrt.cc:
19980 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
19982 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
19984 PR tree-optimization/83940
19985 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
19986 offset_dt to vect_constant_def rather than vect_unknown_def_type.
19987 (vect_check_load_store_mask): Add a mask_dt_out parameter and
19988 use it to pass back the definition type.
19989 (vect_check_store_rhs): Likewise rhs_dt_out.
19990 (vect_build_gather_load_calls): Add a mask_dt argument and use
19991 it instead of a call to vect_is_simple_use.
19992 (vectorizable_store): Update calls to vect_check_load_store_mask
19993 and vect_check_store_rhs. Use the dt returned by the latter instead
19994 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
19995 instead of calls to vect_is_simple_use. Pass the scalar rather
19996 than the vector operand to vect_is_simple_use when handling
19997 second and subsequent copies of an rhs value.
19998 (vectorizable_load): Update calls to vect_check_load_store_mask
19999 and vect_build_gather_load_calls. Use the cached mask_dt and
20000 gs_info.offset_dt instead of calls to vect_is_simple_use.
20002 2018-01-20 Jakub Jelinek <jakub@redhat.com>
20004 PR middle-end/83945
20005 * tree-emutls.c: Include gimplify.h.
20006 (lower_emutls_2): New function.
20007 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
20008 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
20009 it before further processing.
20012 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
20013 UINTVAL (trueop1) instead of INTVAL (op1).
20015 2018-01-19 Jakub Jelinek <jakub@redhat.com>
20019 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
20020 INCOMING_FRAME_SP_OFFSET if not defined.
20021 (scan_trace): Add ENTRY argument. If true and
20022 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
20023 emit a note to adjust the CFA offset.
20024 (create_cfi_notes): Adjust scan_trace callers.
20025 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
20026 INCOMING_FRAME_SP_OFFSET in the CIE.
20027 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
20028 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
20030 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
20031 * doc/tm.texi: Regenerated.
20033 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
20035 PR rtl-optimization/83147
20036 * lra-constraints.c (remove_inheritance_pseudos): Use
20037 lra_substitute_pseudo_within_insn.
20039 2018-01-19 Tom de Vries <tom@codesourcery.com>
20040 Cesar Philippidis <cesar@codesourcery.com>
20043 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
20045 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
20048 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
20049 spaces for function labels.
20051 2018-01-19 Martin Liska <mliska@suse.cz>
20053 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
20054 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
20055 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
20056 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
20057 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
20058 (PRED_CONST_RETURN): Change from 69 to 65.
20059 (PRED_NULL_RETURN): Change from 91 to 71.
20060 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
20061 (PRED_LOOP_GUARD): Change from 66 to 73.
20063 2018-01-19 Martin Liska <mliska@suse.cz>
20065 * predict.c (predict_insn_def): Add new assert.
20066 (struct branch_predictor): Change type to signed integer.
20067 (test_prediction_value_range): Amend test to cover
20068 PROB_UNINITIALIZED.
20069 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
20070 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
20071 (PRED_LOOP_ITERATIONS_MAX): Likewise.
20072 (PRED_LOOP_IV_COMPARE): Likewise.
20073 * predict.h (PROB_UNINITIALIZED): Define new constant.
20075 2018-01-19 Martin Liska <mliska@suse.cz>
20077 * predict.c (dump_prediction): Add new format for
20078 analyze_brprob.py script which is enabled with -details
20080 * profile-count.h (precise_p): New function.
20082 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
20084 PR tree-optimization/83922
20085 * tree-vect-loop.c (vect_verify_full_masking): Return false if
20086 there are no statements that need masking.
20087 (vect_active_double_reduction_p): New function.
20088 (vect_analyze_loop_operations): Use it when handling phis that
20089 are not in the loop header.
20091 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
20093 PR tree-optimization/83914
20094 * tree-vect-loop.c (vectorizable_induction): Don't convert
20095 init_expr or apply the peeling adjustment for inductions
20096 that are nested within the vectorized loop.
20098 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20100 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
20103 2018-01-18 Jakub Jelinek <jakub@redhat.com>
20107 * function.h (gimplify_parameters): Add gimple_seq * argument.
20108 * function.c: Include gimple.h and options.h.
20109 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
20110 for the added local temporaries if needed.
20111 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
20112 if there are any parameter cleanups, wrap whole body into a
20113 try/finally with the cleanups.
20115 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
20118 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
20119 Use GET_MODE_CLASS for scalar floating point.
20121 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
20125 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
20126 Fix call of call_cgraph_insertion_hooks.
20128 2018-01-18 Martin Sebor <msebor@redhat.com>
20130 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
20132 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
20135 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
20138 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
20141 * common.opt: (-ffile-prefix-map): New option.
20142 * opts.c (common_handle_option): Defer it.
20143 * opts-global.c (handle_common_deferred_options): Handle it.
20144 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
20145 * file-prefix-map.h: New file.
20146 (remap_debug_filename, add_debug_prefix_map): ...here.
20147 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
20148 * final.c (debug_prefix_map, add_debug_prefix_map
20149 remap_debug_filename): Move to...
20150 * file-prefix-map.c: New file.
20151 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
20152 generalize, get rid of alloca(), use strrchr() instead of strchr().
20153 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
20154 Implement in terms of add_prefix_map().
20155 (remap_macro_filename, remap_debug_filename): Implement in term of
20157 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
20158 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
20159 * dbxout.c: Include file-prefix-map.h.
20160 * varasm.c: Likewise.
20161 * vmsdbgout.c: Likewise.
20162 * xcoffout.c: Likewise.
20163 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
20164 * doc/cppopts.texi (-fmacro-prefix-map): Document.
20165 * doc/invoke.texi (-ffile-prefix-map): Document.
20166 (-fdebug-prefix-map): Update description.
20168 2018-01-18 Martin Liska <mliska@suse.cz>
20170 * config/i386/i386.c (indirect_thunk_name): Document that also
20172 (output_indirect_thunk): Document why both instructions
20173 (pause and lfence) are generated.
20175 2018-01-18 Richard Biener <rguenther@suse.de>
20177 PR tree-optimization/83887
20178 * graphite-scop-detection.c
20179 (scop_detection::get_nearest_dom_with_single_entry): Remove.
20180 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
20181 (scop_detection::merge_sese): Re-implement with a flood-fill
20182 algorithm that properly finds a SESE region if it exists.
20184 2018-01-18 Jakub Jelinek <jakub@redhat.com>
20187 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
20188 pointer_diff optimizations use view_convert instead of convert.
20190 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
20192 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
20193 Generate different code for -mno-speculate-indirect-jumps.
20194 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
20195 (*call_indirect_aix<mode>): Disable for
20196 -mno-speculate-indirect-jumps.
20197 (*call_indirect_aix<mode>_nospec): New define_insn.
20198 (*call_value_indirect_aix<mode>): Disable for
20199 -mno-speculate-indirect-jumps.
20200 (*call_value_indirect_aix<mode>_nospec): New define_insn.
20201 (*sibcall_nonlocal_sysv<mode>): Generate different code for
20202 -mno-speculate-indirect-jumps.
20203 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
20205 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
20207 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
20208 long double type, set the flags for noting the default long double
20209 type, even if we don't pass or return a long double type.
20211 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
20214 * ipa-inline.c (flatten_function): Do not overwrite final inlining
20217 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
20219 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
20220 support for merge[hl].
20221 (fold_mergehl_helper): New helper function.
20222 (tree-vector-builder.h): New #include for tree_vector_builder usage.
20223 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
20224 (altivec_vmrglw_direct): Add xxmrglw insn.
20226 2018-01-17 Andrew Waterman <andrew@sifive.com>
20228 * config/riscv/riscv.c (riscv_conditional_register_usage): If
20229 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
20231 2018-01-17 David Malcolm <dmalcolm@redhat.com>
20234 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
20235 call the lto_location_cache before reading the
20236 DECL_SOURCE_LOCATION of the types.
20238 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
20239 Richard Sandiford <richard.sandiford@linaro.org>
20241 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
20242 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
20243 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
20244 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
20245 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
20247 * config/aarch64/constraints.md (aarch64_movti_operand):
20249 * config/aarch64/predicates.md (Uti): Add new constraint.
20251 2018-01-17 Carl Love <cel@us.ibm.com>
20253 * config/rs6000/vsx.md (define_expand xl_len_r,
20254 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
20255 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
20257 (define_expand, define_insn): Move the shift left from the
20258 define_insn to the define_expand for lxvl and stxvl instructions.
20259 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
20260 and XL_LEN_R definitions to PURE.
20262 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
20264 * config/i386/i386.c (indirect_thunk_name): Declare regno
20265 as unsigned int. Compare regno with INVALID_REGNUM.
20266 (output_indirect_thunk): Ditto.
20267 (output_indirect_thunk_function): Ditto.
20268 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
20269 in the call to output_indirect_thunk_function.
20271 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
20273 PR middle-end/83884
20274 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
20275 rather than the size of inner_type to determine the stack slot size
20276 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
20278 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
20281 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
20284 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
20286 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
20287 endian Linux systems to optionally enable multilibs for selecting
20288 the long double type if the user configured an explicit type.
20289 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
20290 have no long double multilibs if not defined.
20291 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
20292 warn if the user used -mabi={ieee,ibm}longdouble and we built
20293 multilibs for long double.
20294 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
20295 appropriate multilib option.
20296 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
20298 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
20299 for building long double multilibs.
20300 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
20302 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
20304 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
20307 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
20309 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
20312 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
20315 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
20318 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
20320 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
20321 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
20323 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
20325 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
20326 different rtl trees depending on TARGET_64BIT.
20327 (rs6000_gen_lvx): Likewise.
20329 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
20331 * config/visium/visium.md (nop): Tweak comment.
20332 (hazard_nop): Likewise.
20334 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
20336 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
20337 -mspeculate-indirect-jumps.
20338 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
20339 for -mno-speculate-indirect-jumps.
20340 (*call_indirect_elfv2<mode>_nospec): New define_insn.
20341 (*call_value_indirect_elfv2<mode>): Disable for
20342 -mno-speculate-indirect-jumps.
20343 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
20344 (indirect_jump): Emit different RTL for
20345 -mno-speculate-indirect-jumps.
20346 (*indirect_jump<mode>): Disable for
20347 -mno-speculate-indirect-jumps.
20348 (*indirect_jump<mode>_nospec): New define_insn.
20349 (tablejump): Emit different RTL for
20350 -mno-speculate-indirect-jumps.
20351 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
20352 (tablejumpsi_nospec): New define_expand.
20353 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
20354 (tablejumpdi_nospec): New define_expand.
20355 (*tablejump<mode>_internal1): Disable for
20356 -mno-speculate-indirect-jumps.
20357 (*tablejump<mode>_internal1_nospec): New define_insn.
20358 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
20361 2018-01-16 Artyom Skrobov tyomitch@gmail.com
20363 * caller-save.c (insert_save): Drop unnecessary parameter. All
20366 2018-01-16 Jakub Jelinek <jakub@redhat.com>
20367 Richard Biener <rguenth@suse.de>
20370 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
20371 return early, inline manually is_gimple_sizepos. Make sure if we
20372 call gimplify_expr we don't end up with a gimple constant.
20373 * tree.c (variably_modified_type_p): Don't return true for
20374 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
20375 * gimplify.h (is_gimple_sizepos): Remove.
20377 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
20379 PR tree-optimization/83857
20380 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
20381 vectorizable_live_operation for pure SLP statements.
20382 (vectorizable_live_operation): Handle PHIs.
20384 2018-01-16 Richard Biener <rguenther@suse.de>
20386 PR tree-optimization/83867
20387 * tree-vect-stmts.c (vect_transform_stmt): Precompute
20388 nested_in_vect_loop_p since the scalar stmt may get invalidated.
20390 2018-01-16 Jakub Jelinek <jakub@redhat.com>
20393 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
20394 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
20395 If off is not INTEGER_CST, issue a may not be aligned warning
20396 rather than isn't aligned. Use isn%'t rather than isn't.
20397 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
20399 <case MULT_EXPR>: Improve the case when bottom and one of the
20400 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
20401 operand, in that case check if the other operand is multiple of
20402 bottom divided by the INTEGER_CST operand.
20404 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
20407 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
20408 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
20409 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
20410 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
20411 * config/pa/pa.c (pa_function_arg_advance): Likewise.
20412 (pa_function_arg, pa_arg_partial_bytes): Likewise.
20413 (pa_function_arg_size): New function.
20415 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
20417 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
20418 in a separate statement.
20420 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
20422 PR tree-optimization/83847
20423 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
20424 group gathers and scatters.
20426 2018-01-16 Jakub Jelinek <jakub@redhat.com>
20428 PR rtl-optimization/86620
20429 * params.def (max-sched-ready-insns): Bump minimum value to 1.
20431 PR rtl-optimization/83213
20432 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
20433 to last if both are JUMP_INSNs.
20435 PR tree-optimization/83843
20436 * gimple-ssa-store-merging.c
20437 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
20438 store_immediate_info for bswap/nop orig_stores.
20440 2018-01-15 Andrew Waterman <andrew@sifive.com>
20442 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
20444 <UDIV>: Increase cost if !TARGET_DIV.
20446 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
20448 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
20449 (define_attr "cr_logical_3op"): New.
20450 (cceq_ior_compare): Adjust.
20451 (cceq_ior_compare_complement): Adjust.
20452 (*cceq_rev_compare): Adjust.
20453 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
20454 (is_cracked_insn): Adjust.
20455 (insn_must_be_first_in_group): Adjust.
20456 * config/rs6000/40x.md: Adjust.
20457 * config/rs6000/440.md: Adjust.
20458 * config/rs6000/476.md: Adjust.
20459 * config/rs6000/601.md: Adjust.
20460 * config/rs6000/603.md: Adjust.
20461 * config/rs6000/6xx.md: Adjust.
20462 * config/rs6000/7450.md: Adjust.
20463 * config/rs6000/7xx.md: Adjust.
20464 * config/rs6000/8540.md: Adjust.
20465 * config/rs6000/cell.md: Adjust.
20466 * config/rs6000/e300c2c3.md: Adjust.
20467 * config/rs6000/e500mc.md: Adjust.
20468 * config/rs6000/e500mc64.md: Adjust.
20469 * config/rs6000/e5500.md: Adjust.
20470 * config/rs6000/e6500.md: Adjust.
20471 * config/rs6000/mpc.md: Adjust.
20472 * config/rs6000/power4.md: Adjust.
20473 * config/rs6000/power5.md: Adjust.
20474 * config/rs6000/power6.md: Adjust.
20475 * config/rs6000/power7.md: Adjust.
20476 * config/rs6000/power8.md: Adjust.
20477 * config/rs6000/power9.md: Adjust.
20478 * config/rs6000/rs64.md: Adjust.
20479 * config/rs6000/titan.md: Adjust.
20481 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
20483 * config/i386/predicates.md (indirect_branch_operand): Rewrite
20484 ix86_indirect_branch_register logic.
20486 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
20488 * config/i386/constraints.md (Bs): Update
20489 ix86_indirect_branch_register check. Don't check
20490 ix86_indirect_branch_register with GOT_memory_operand.
20492 * config/i386/predicates.md (GOT_memory_operand): Don't check
20493 ix86_indirect_branch_register here.
20494 (GOT32_symbol_operand): Likewise.
20496 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
20498 * config/i386/predicates.md (constant_call_address_operand):
20499 Rewrite ix86_indirect_branch_register logic.
20500 (sibcall_insn_operand): Likewise.
20502 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
20504 * config/i386/constraints.md (Bs): Replace
20505 ix86_indirect_branch_thunk_register with
20506 ix86_indirect_branch_register.
20508 * config/i386/i386.md (indirect_jump): Likewise.
20509 (tablejump): Likewise.
20510 (*sibcall_memory): Likewise.
20511 (*sibcall_value_memory): Likewise.
20512 Peepholes of indirect call and jump via memory: Likewise.
20513 * config/i386/i386.opt: Likewise.
20514 * config/i386/predicates.md (indirect_branch_operand): Likewise.
20515 (GOT_memory_operand): Likewise.
20516 (call_insn_operand): Likewise.
20517 (sibcall_insn_operand): Likewise.
20518 (GOT32_symbol_operand): Likewise.
20520 2018-01-15 Jakub Jelinek <jakub@redhat.com>
20522 PR middle-end/83837
20523 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
20524 type rather than type addr's type points to.
20525 (expand_omp_atomic_mutex): Likewise.
20526 (expand_omp_atomic): Likewise.
20528 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
20531 * config/i386/i386.c (output_indirect_thunk_function): Use
20532 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
20533 for __x86_return_thunk.
20535 2018-01-15 Richard Biener <rguenther@suse.de>
20537 PR middle-end/83850
20538 * expmed.c (extract_bit_field_1): Fix typo.
20540 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20543 * config/arm/iterators.md (VF): New mode iterator.
20544 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
20545 Remove integer-related logic from pattern.
20546 (neon_vabd<mode>_3): Likewise.
20548 2018-01-15 Jakub Jelinek <jakub@redhat.com>
20550 PR middle-end/82694
20551 * common.opt (fstrict-overflow): No longer an alias.
20552 (fwrapv-pointer): New option.
20553 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
20554 also for pointer types based on flag_wrapv_pointer.
20555 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
20556 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
20557 opts->x_flag_wrapv got set.
20558 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
20559 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
20560 POINTER_TYPE_OVERFLOW_UNDEFINED.
20561 * match.pd: Likewise in address comparison pattern.
20562 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
20564 2018-01-15 Richard Biener <rguenther@suse.de>
20567 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
20568 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
20569 Reset type names to their identifier if their TYPE_DECL doesn't
20570 have linkage (and thus is used for ODR and devirt).
20571 (save_debug_info_for_decl): Remove.
20572 (save_debug_info_for_type): Likewise.
20573 (add_tree_to_fld_list): Adjust.
20574 * tree-pretty-print.c (dump_generic_node): Make dumping of
20575 type names more robust.
20577 2018-01-15 Richard Biener <rguenther@suse.de>
20579 * BASE-VER: Bump to 8.0.1.
20581 2018-01-14 Martin Sebor <msebor@redhat.com>
20584 * builtins.c (check_access): Avoid warning when the no-warning bit
20587 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
20589 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
20590 * ira-color (allocno_hard_regs_compare): Likewise.
20592 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
20595 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
20596 Use .pushsection/.popsection.
20598 2018-01-14 Martin Sebor <msebor@redhat.com>
20601 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
20603 2018-01-14 Jakub Jelinek <jakub@redhat.com>
20605 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
20606 entry from extra_headers.
20607 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
20608 extra_headers, make the list bitwise identical to the i?86-*-* one.
20610 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
20612 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
20613 -mcmodel=large with -mindirect-branch=thunk,
20614 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
20615 -mfunction-return=thunk-extern.
20616 * doc/invoke.texi: Document -mcmodel=large is incompatible with
20617 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
20618 -mfunction-return=thunk and -mfunction-return=thunk-extern.
20620 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
20622 * config/i386/i386.c (print_reg): Print the name of the full
20623 integer register without '%'.
20624 (ix86_print_operand): Handle 'V'.
20625 * doc/extend.texi: Document 'V' modifier.
20627 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
20629 * config/i386/constraints.md (Bs): Disallow memory operand for
20630 -mindirect-branch-register.
20632 * config/i386/predicates.md (indirect_branch_operand): Likewise.
20633 (GOT_memory_operand): Likewise.
20634 (call_insn_operand): Likewise.
20635 (sibcall_insn_operand): Likewise.
20636 (GOT32_symbol_operand): Likewise.
20637 * config/i386/i386.md (indirect_jump): Call convert_memory_address
20638 for -mindirect-branch-register.
20639 (tablejump): Likewise.
20640 (*sibcall_memory): Likewise.
20641 (*sibcall_value_memory): Likewise.
20642 Disallow peepholes of indirect call and jump via memory for
20643 -mindirect-branch-register.
20644 (*call_pop): Replace m with Bw.
20645 (*call_value_pop): Likewise.
20646 (*sibcall_pop_memory): Replace m with Bs.
20647 * config/i386/i386.opt (mindirect-branch-register): New option.
20648 * doc/invoke.texi: Document -mindirect-branch-register option.
20650 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
20652 * config/i386/i386-protos.h (ix86_output_function_return): New.
20653 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
20654 set function_return_type.
20655 (indirect_thunk_name): Add ret_p to indicate thunk for function
20657 (output_indirect_thunk_function): Pass false to
20658 indirect_thunk_name.
20659 (ix86_output_indirect_branch_via_reg): Likewise.
20660 (ix86_output_indirect_branch_via_push): Likewise.
20661 (output_indirect_thunk_function): Create alias for function
20662 return thunk if regno < 0.
20663 (ix86_output_function_return): New function.
20664 (ix86_handle_fndecl_attribute): Handle function_return.
20665 (ix86_attribute_table): Add function_return.
20666 * config/i386/i386.h (machine_function): Add
20667 function_return_type.
20668 * config/i386/i386.md (simple_return_internal): Use
20669 ix86_output_function_return.
20670 (simple_return_internal_long): Likewise.
20671 * config/i386/i386.opt (mfunction-return=): New option.
20672 (indirect_branch): Mention -mfunction-return=.
20673 * doc/extend.texi: Document function_return function attribute.
20674 * doc/invoke.texi: Document -mfunction-return= option.
20676 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
20678 * config/i386/i386-opts.h (indirect_branch): New.
20679 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
20680 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
20681 with local indirect jump when converting indirect call and jump.
20682 (ix86_set_indirect_branch_type): New.
20683 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
20684 (indirectlabelno): New.
20685 (indirect_thunk_needed): Likewise.
20686 (indirect_thunk_bnd_needed): Likewise.
20687 (indirect_thunks_used): Likewise.
20688 (indirect_thunks_bnd_used): Likewise.
20689 (INDIRECT_LABEL): Likewise.
20690 (indirect_thunk_name): Likewise.
20691 (output_indirect_thunk): Likewise.
20692 (output_indirect_thunk_function): Likewise.
20693 (ix86_output_indirect_branch_via_reg): Likewise.
20694 (ix86_output_indirect_branch_via_push): Likewise.
20695 (ix86_output_indirect_branch): Likewise.
20696 (ix86_output_indirect_jmp): Likewise.
20697 (ix86_code_end): Call output_indirect_thunk_function if needed.
20698 (ix86_output_call_insn): Call ix86_output_indirect_branch if
20700 (ix86_handle_fndecl_attribute): Handle indirect_branch.
20701 (ix86_attribute_table): Add indirect_branch.
20702 * config/i386/i386.h (machine_function): Add indirect_branch_type
20703 and has_local_indirect_jump.
20704 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
20706 (tablejump): Likewise.
20707 (*indirect_jump): Use ix86_output_indirect_jmp.
20708 (*tablejump_1): Likewise.
20709 (simple_return_indirect_internal): Likewise.
20710 * config/i386/i386.opt (mindirect-branch=): New option.
20711 (indirect_branch): New.
20714 (thunk-inline): Likewise.
20715 (thunk-extern): Likewise.
20716 * doc/extend.texi: Document indirect_branch function attribute.
20717 * doc/invoke.texi: Document -mindirect-branch= option.
20719 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
20722 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
20724 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
20726 * ipa-inline.c (want_inline_small_function_p): Return false if
20727 inlining has already failed with CIF_FINAL_ERROR.
20728 (update_caller_keys): Call want_inline_small_function_p before
20730 (update_callee_keys): Likewise.
20732 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
20734 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
20736 (rs6000_quadword_masked_address_p): Likewise.
20737 (quad_aligned_load_p): Likewise.
20738 (quad_aligned_store_p): Likewise.
20739 (const_load_sequence_p): Add comment to describe the outer-most loop.
20740 (mimic_memory_attributes_and_flags): New function.
20741 (rs6000_gen_stvx): Likewise.
20742 (replace_swapped_aligned_store): Likewise.
20743 (rs6000_gen_lvx): Likewise.
20744 (replace_swapped_aligned_load): Likewise.
20745 (replace_swapped_load_constant): Capitalize argument name in
20746 comment describing this function.
20747 (rs6000_analyze_swaps): Add a third pass to search for vector loads
20748 and stores that access quad-word aligned addresses and replace
20749 with stvx or lvx instructions when appropriate.
20750 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
20751 New function prototype.
20752 (rs6000_quadword_masked_address_p): Likewise.
20753 (rs6000_gen_lvx): Likewise.
20754 (rs6000_gen_stvx): Likewise.
20755 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
20756 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
20757 when memory address is aligned.
20758 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
20759 this split to select lvx instruction when memory address is aligned.
20760 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
20761 instruction when memory address is aligned.
20762 (*vsx_le_perm_load_v16qi): Likewise.
20763 (four unnamed splitters): Modify to select the stvx instruction
20764 when memory is aligned.
20766 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
20768 * predict.c (determine_unlikely_bbs): Handle correctly BBs
20769 which appears in the queue multiple times.
20771 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20772 Alan Hayward <alan.hayward@arm.com>
20773 David Sherwood <david.sherwood@arm.com>
20775 * tree-vectorizer.h (vec_lower_bound): New structure.
20776 (_loop_vec_info): Add check_nonzero and lower_bounds.
20777 (LOOP_VINFO_CHECK_NONZERO): New macro.
20778 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
20779 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
20780 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
20781 fields. Make seg_len the distance travelled, not including the
20783 (dr_direction_indicator): Declare.
20784 (dr_zero_step_indicator): Likewise.
20785 (dr_known_forward_stride_p): Likewise.
20786 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
20788 (runtime_alias_check_p): Allow runtime alias checks with
20790 (operator ==): Compare access_size and align.
20791 (prune_runtime_alias_test_list): Rework for new distinction between
20792 the access_size and seg_len.
20793 (create_intersect_range_checks_index): Likewise. Cope with polynomial
20795 (get_segment_min_max): New function.
20796 (create_intersect_range_checks): Use it.
20797 (dr_step_indicator): New function.
20798 (dr_direction_indicator): Likewise.
20799 (dr_zero_step_indicator): Likewise.
20800 (dr_known_forward_stride_p): Likewise.
20801 * tree-loop-distribution.c (data_ref_segment_size): Return
20802 DR_STEP * (niters - 1).
20803 (compute_alias_check_pairs): Update call to the dr_with_seg_len
20805 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
20806 (vect_preserves_scalar_order_p): New function, split out from...
20807 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
20808 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
20809 (vect_vfa_access_size): New function.
20810 (vect_vfa_align): Likewise.
20811 (vect_compile_time_alias): Take access_size_a and access_b arguments.
20812 (dump_lower_bound): New function.
20813 (vect_check_lower_bound): Likewise.
20814 (vect_small_gap_p): Likewise.
20815 (vectorizable_with_step_bound_p): Likewise.
20816 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
20817 depencies if the vectorization factor is 1. Convert the checks
20818 for nonzero steps into checks on the bounds of DR_STEP. Try using
20819 a bunds check for variable steps if the minimum required step is
20820 relatively small. Update calls to the dr_with_seg_len
20821 constructor and to vect_compile_time_alias.
20822 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
20824 (vect_loop_versioning): Call it.
20825 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
20827 (vect_estimate_min_profitable_iters): Account for any bounds checks.
20829 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20830 Alan Hayward <alan.hayward@arm.com>
20831 David Sherwood <david.sherwood@arm.com>
20833 * doc/sourcebuild.texi (vect_scatter_store): Document.
20834 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
20836 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
20838 * genopinit.c (main): Add supports_vec_scatter_store and
20839 supports_vec_scatter_store_cached to target_optabs.
20840 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
20841 IFN_MASK_SCATTER_STORE.
20842 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
20844 * internal-fn.h (internal_store_fn_p): Declare.
20845 (internal_fn_stored_value_index): Likewise.
20846 * internal-fn.c (scatter_store_direct): New macro.
20847 (expand_scatter_store_optab_fn): New function.
20848 (direct_scatter_store_optab_supported_p): New macro.
20849 (internal_store_fn_p): New function.
20850 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
20851 IFN_MASK_SCATTER_STORE.
20852 (internal_fn_mask_index): Likewise.
20853 (internal_fn_stored_value_index): New function.
20854 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
20855 for scatter stores.
20856 * optabs-query.h (supports_vec_scatter_store_p): Declare.
20857 * optabs-query.c (supports_vec_scatter_store_p): New function.
20858 * tree-vectorizer.h (vect_get_store_rhs): Declare.
20859 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
20860 true for scatter stores.
20861 (vect_gather_scatter_fn_p): Handle scatter stores too.
20862 (vect_check_gather_scatter): Consider using scatter stores if
20863 supports_vec_scatter_store_p.
20864 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
20865 scatter stores too.
20866 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
20867 internal_fn_stored_value_index.
20868 (check_load_store_masking): Handle scatter stores too.
20869 (vect_get_store_rhs): Make public.
20870 (vectorizable_call): Use internal_store_fn_p.
20871 (vectorizable_store): Handle scatter store internal functions.
20872 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
20873 when deciding whether the end of the group has been reached.
20874 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
20875 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
20876 (mask_scatter_store<mode>): New insns.
20878 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20879 Alan Hayward <alan.hayward@arm.com>
20880 David Sherwood <david.sherwood@arm.com>
20882 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
20883 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
20884 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
20886 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
20887 Use vect_truncate_gather_scatter_offset if we can't treat the
20888 operation as a normal gather load or scatter store.
20889 (get_group_load_store_type): Take the gather_scatter_info
20890 as argument. Try using a gather load or scatter store for
20891 single-element groups.
20892 (get_load_store_type): Update calls to get_group_load_store_type
20893 and vect_use_strided_gather_scatters_p.
20895 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20896 Alan Hayward <alan.hayward@arm.com>
20897 David Sherwood <david.sherwood@arm.com>
20899 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
20900 optional tree argument.
20901 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
20903 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
20904 but continue to use the current value as a fallback.
20905 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
20906 to compare the updates.
20907 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
20908 (get_load_store_type): Use it when handling a strided access.
20909 (vect_get_strided_load_store_ops): New function.
20910 (vect_get_data_ptr_increment): Likewise.
20911 (vectorizable_load): Handle strided gather loads. Always pass
20912 a step to vect_create_data_ref_ptr and bump_vector_ptr.
20914 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20915 Alan Hayward <alan.hayward@arm.com>
20916 David Sherwood <david.sherwood@arm.com>
20918 * doc/md.texi (gather_load@var{m}): Document.
20919 (mask_gather_load@var{m}): Likewise.
20920 * genopinit.c (main): Add supports_vec_gather_load and
20921 supports_vec_gather_load_cached to target_optabs.
20922 * optabs-tree.c (init_tree_optimization_optabs): Use
20923 ggc_cleared_alloc to allocate target_optabs.
20924 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
20925 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
20927 * internal-fn.h (internal_load_fn_p): Declare.
20928 (internal_gather_scatter_fn_p): Likewise.
20929 (internal_fn_mask_index): Likewise.
20930 (internal_gather_scatter_fn_supported_p): Likewise.
20931 * internal-fn.c (gather_load_direct): New macro.
20932 (expand_gather_load_optab_fn): New function.
20933 (direct_gather_load_optab_supported_p): New macro.
20934 (direct_internal_fn_optab): New function.
20935 (internal_load_fn_p): Likewise.
20936 (internal_gather_scatter_fn_p): Likewise.
20937 (internal_fn_mask_index): Likewise.
20938 (internal_gather_scatter_fn_supported_p): Likewise.
20939 * optabs-query.c (supports_at_least_one_mode_p): New function.
20940 (supports_vec_gather_load_p): Likewise.
20941 * optabs-query.h (supports_vec_gather_load_p): Declare.
20942 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
20943 and memory_type field.
20944 (NUM_PATTERNS): Bump to 15.
20945 * tree-vect-data-refs.c: Include internal-fn.h.
20946 (vect_gather_scatter_fn_p): New function.
20947 (vect_describe_gather_scatter_call): Likewise.
20948 (vect_check_gather_scatter): Try using internal functions for
20949 gather loads. Recognize existing calls to a gather load function.
20950 (vect_analyze_data_refs): Consider using gather loads if
20951 supports_vec_gather_load_p.
20952 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
20953 (vect_get_gather_scatter_offset_type): Likewise.
20954 (vect_convert_mask_for_vectype): Likewise.
20955 (vect_add_conversion_to_patterm): Likewise.
20956 (vect_try_gather_scatter_pattern): Likewise.
20957 (vect_recog_gather_scatter_pattern): New pattern recognizer.
20958 (vect_vect_recog_func_ptrs): Add it.
20959 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
20960 internal_fn_mask_index and internal_gather_scatter_fn_p.
20961 (check_load_store_masking): Take the gather_scatter_info as an
20962 argument and handle gather loads.
20963 (vect_get_gather_scatter_ops): New function.
20964 (vectorizable_call): Check internal_load_fn_p.
20965 (vectorizable_load): Likewise. Handle gather load internal
20967 (vectorizable_store): Update call to check_load_store_masking.
20968 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
20969 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
20970 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
20971 (aarch64_gather_scale_operand_d): New predicates.
20972 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
20973 (mask_gather_load<mode>): New insns.
20975 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20976 Alan Hayward <alan.hayward@arm.com>
20977 David Sherwood <david.sherwood@arm.com>
20979 * optabs.def (fold_left_plus_optab): New optab.
20980 * doc/md.texi (fold_left_plus_@var{m}): Document.
20981 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
20982 * internal-fn.c (fold_left_direct): Define.
20983 (expand_fold_left_optab_fn): Likewise.
20984 (direct_fold_left_optab_supported_p): Likewise.
20985 * fold-const-call.c (fold_const_fold_left): New function.
20986 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
20987 * tree-parloops.c (valid_reduction_p): New function.
20988 (gather_scalar_reductions): Use it.
20989 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
20990 (vect_finish_replace_stmt): Declare.
20991 * tree-vect-loop.c (fold_left_reduction_fn): New function.
20992 (needs_fold_left_reduction_p): New function, split out from...
20993 (vect_is_simple_reduction): ...here. Accept reductions that
20994 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
20995 (vect_force_simple_reduction): Also store the reduction type in
20996 the assignment's STMT_VINFO_REDUC_TYPE.
20997 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
20998 (merge_with_identity): New function.
20999 (vect_expand_fold_left): Likewise.
21000 (vectorize_fold_left_reduction): Likewise.
21001 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
21002 scalar phi in place for it. Check for target support and reject
21003 cases that would reassociate the operation. Defer the transform
21004 phase to vectorize_fold_left_reduction.
21005 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
21006 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
21007 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
21009 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21011 * tree-if-conv.c (predicate_mem_writes): Remove redundant
21012 call to ifc_temp_var.
21014 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21015 Alan Hayward <alan.hayward@arm.com>
21016 David Sherwood <david.sherwood@arm.com>
21018 * target.def (legitimize_address_displacement): Take the original
21019 offset as a poly_int.
21020 * targhooks.h (default_legitimize_address_displacement): Update
21022 * targhooks.c (default_legitimize_address_displacement): Likewise.
21023 * doc/tm.texi: Regenerate.
21024 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
21025 as an argument, moving assert of ad->disp == ad->disp_term to...
21026 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
21027 Try calling targetm.legitimize_address_displacement before expanding
21028 the address rather than afterwards, and adjust for the new interface.
21029 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
21030 Match the new hook interface. Handle SVE addresses.
21031 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
21032 new hook interface.
21034 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21036 * Makefile.in (OBJS): Add early-remat.o.
21037 * target.def (select_early_remat_modes): New hook.
21038 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
21039 * doc/tm.texi: Regenerate.
21040 * targhooks.h (default_select_early_remat_modes): Declare.
21041 * targhooks.c (default_select_early_remat_modes): New function.
21042 * timevar.def (TV_EARLY_REMAT): New timevar.
21043 * passes.def (pass_early_remat): New pass.
21044 * tree-pass.h (make_pass_early_remat): Declare.
21045 * early-remat.c: New file.
21046 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
21048 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
21050 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21051 Alan Hayward <alan.hayward@arm.com>
21052 David Sherwood <david.sherwood@arm.com>
21054 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
21055 vfm1 with a bound_epilog parameter.
21056 (vect_do_peeling): Update calls accordingly, and move the prologue
21057 call earlier in the function. Treat the base bound_epilog as 0 for
21058 fully-masked loops and retain vf - 1 for other loops. Add 1 to
21059 this base when peeling for gaps.
21060 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
21061 with fully-masked loops.
21062 (vect_estimate_min_profitable_iters): Handle the single peeled
21063 iteration in that case.
21065 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21066 Alan Hayward <alan.hayward@arm.com>
21067 David Sherwood <david.sherwood@arm.com>
21069 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
21070 single-element interleaving even if the size is not a power of 2.
21071 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
21072 accesses for single-element interleaving if the group size is
21075 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21076 Alan Hayward <alan.hayward@arm.com>
21077 David Sherwood <david.sherwood@arm.com>
21079 * doc/md.texi (fold_extract_last_@var{m}): Document.
21080 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
21081 * optabs.def (fold_extract_last_optab): New optab.
21082 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
21083 * internal-fn.c (fold_extract_direct): New macro.
21084 (expand_fold_extract_optab_fn): Likewise.
21085 (direct_fold_extract_optab_supported_p): Likewise.
21086 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
21087 * tree-vect-loop.c (vect_model_reduction_cost): Handle
21088 EXTRACT_LAST_REDUCTION.
21089 (get_initial_def_for_reduction): Do not create an initial vector
21090 for EXTRACT_LAST_REDUCTION reductions.
21091 (vectorizable_reduction): Leave the scalar phi in place for
21092 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
21093 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
21094 epilogue code for EXTRACT_LAST_REDUCTION and defer the
21095 transform phase to vectorizable_condition.
21096 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
21098 (vect_finish_stmt_generation): ...here.
21099 (vect_finish_replace_stmt): New function.
21100 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
21101 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
21103 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
21105 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21106 Alan Hayward <alan.hayward@arm.com>
21107 David Sherwood <david.sherwood@arm.com>
21109 * doc/md.texi (extract_last_@var{m}): Document.
21110 * optabs.def (extract_last_optab): New optab.
21111 * internal-fn.def (EXTRACT_LAST): New internal function.
21112 * internal-fn.c (cond_unary_direct): New macro.
21113 (expand_cond_unary_optab_fn): Likewise.
21114 (direct_cond_unary_optab_supported_p): Likewise.
21115 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
21116 loops using EXTRACT_LAST.
21117 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
21118 (extract_last_<mode>): ...this optab.
21119 (vec_extract<mode><Vel>): Update accordingly.
21121 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21122 Alan Hayward <alan.hayward@arm.com>
21123 David Sherwood <david.sherwood@arm.com>
21125 * target.def (empty_mask_is_expensive): New hook.
21126 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
21127 * doc/tm.texi: Regenerate.
21128 * targhooks.h (default_empty_mask_is_expensive): Declare.
21129 * targhooks.c (default_empty_mask_is_expensive): New function.
21130 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
21131 if the target says that empty masks are expensive.
21132 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
21134 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
21136 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21137 Alan Hayward <alan.hayward@arm.com>
21138 David Sherwood <david.sherwood@arm.com>
21140 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
21141 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
21142 (vect_use_loop_mask_for_alignment_p): New function.
21143 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
21144 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
21145 niters_skip argument. Make sure that the first niters_skip elements
21146 of the first iteration are inactive.
21147 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
21148 Update call to vect_set_loop_masks_directly.
21149 (get_misalign_in_elems): New function, split out from...
21150 (vect_gen_prolog_loop_niters): ...here.
21151 (vect_update_init_of_dr): Take a code argument that specifies whether
21152 the adjustment should be added or subtracted.
21153 (vect_update_init_of_drs): Likewise.
21154 (vect_prepare_for_masked_peels): New function.
21155 (vect_do_peeling): Skip prologue peeling if we're using a mask
21156 instead. Update call to vect_update_inits_of_drs.
21157 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
21159 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
21160 alignment. Do not include the number of peeled iterations in
21161 the minimum threshold in that case.
21162 (vectorizable_induction): Adjust the start value down by
21163 LOOP_VINFO_MASK_SKIP_NITERS iterations.
21164 (vect_transform_loop): Call vect_prepare_for_masked_peels.
21165 Take the number of skipped iterations into account when calculating
21167 * tree-vect-stmts.c (vect_gen_while_not): New function.
21169 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21170 Alan Hayward <alan.hayward@arm.com>
21171 David Sherwood <david.sherwood@arm.com>
21173 * doc/sourcebuild.texi (vect_fully_masked): Document.
21174 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
21175 default value to 0.
21176 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
21178 (vect_analyze_loop_2): ...here. Don't check the vectorization
21179 factor against the number of loop iterations if the loop is
21182 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21183 Alan Hayward <alan.hayward@arm.com>
21184 David Sherwood <david.sherwood@arm.com>
21186 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
21187 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
21188 (dump_groups): Update accordingly.
21189 (iv_use::mem_type): New member variable.
21190 (address_p): New function.
21191 (record_use): Add a mem_type argument and initialize the new
21193 (record_group_use): Add a mem_type argument. Use address_p.
21194 Remove obsolete null checks of base_object. Update call to record_use.
21195 (find_interesting_uses_op): Update call to record_group_use.
21196 (find_interesting_uses_cond): Likewise.
21197 (find_interesting_uses_address): Likewise.
21198 (get_mem_type_for_internal_fn): New function.
21199 (find_address_like_use): Likewise.
21200 (find_interesting_uses_stmt): Try find_address_like_use before
21201 calling find_interesting_uses_op.
21202 (addr_offset_valid_p): Use the iv mem_type field as the type
21203 of the addressed memory.
21204 (add_autoinc_candidates): Likewise.
21205 (get_address_cost): Likewise.
21206 (split_small_address_groups_p): Use address_p.
21207 (split_address_groups): Likewise.
21208 (add_iv_candidate_for_use): Likewise.
21209 (autoinc_possible_for_pair): Likewise.
21210 (rewrite_groups): Likewise.
21211 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
21212 (determine_group_iv_cost): Update after split of USE_ADDRESS.
21213 (get_alias_ptr_type_for_ptr_address): New function.
21214 (rewrite_use_address): Rewrite address uses in calls that were
21215 identified by find_address_like_use.
21217 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21218 Alan Hayward <alan.hayward@arm.com>
21219 David Sherwood <david.sherwood@arm.com>
21221 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
21223 * gimple-expr.h (is_gimple_addressable: Likewise.
21224 * gimple-expr.c (is_gimple_address): Likewise.
21225 * internal-fn.c (expand_call_mem_ref): New function.
21226 (expand_mask_load_optab_fn): Use it.
21227 (expand_mask_store_optab_fn): Likewise.
21229 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21230 Alan Hayward <alan.hayward@arm.com>
21231 David Sherwood <david.sherwood@arm.com>
21233 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
21234 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
21235 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
21236 (cond_umax@var{mode}): Document.
21237 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
21238 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
21239 (cond_umin_optab, cond_umax_optab): New optabs.
21240 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
21241 (COND_IOR, COND_XOR): New internal functions.
21242 * internal-fn.h (get_conditional_internal_fn): Declare.
21243 * internal-fn.c (cond_binary_direct): New macro.
21244 (expand_cond_binary_optab_fn): Likewise.
21245 (direct_cond_binary_optab_supported_p): Likewise.
21246 (get_conditional_internal_fn): New function.
21247 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
21248 Cope with reduction statements that are vectorized as calls rather
21250 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
21251 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
21252 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
21253 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
21254 (UNSPEC_COND_EOR): New unspecs.
21255 (optab): Add mappings for them.
21256 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
21257 (sve_int_op, sve_fp_op): New int attributes.
21259 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21260 Alan Hayward <alan.hayward@arm.com>
21261 David Sherwood <david.sherwood@arm.com>
21263 * optabs.def (while_ult_optab): New optab.
21264 * doc/md.texi (while_ult@var{m}@var{n}): Document.
21265 * internal-fn.def (WHILE_ULT): New internal function.
21266 * internal-fn.h (direct_internal_fn_supported_p): New override
21267 that takes two types as argument.
21268 * internal-fn.c (while_direct): New macro.
21269 (expand_while_optab_fn): New function.
21270 (convert_optab_supported_p): Likewise.
21271 (direct_while_optab_supported_p): New macro.
21272 * wide-int.h (wi::udiv_ceil): New function.
21273 * tree-vectorizer.h (rgroup_masks): New structure.
21274 (vec_loop_masks): New typedef.
21275 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
21276 and fully_masked_p.
21277 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
21278 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
21279 (vect_max_vf): New function.
21280 (slpeel_make_loop_iterate_ntimes): Delete.
21281 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
21282 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
21283 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
21284 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
21285 internal-fn.h, stor-layout.h and optabs-query.h.
21286 (vect_set_loop_mask): New function.
21287 (add_preheader_seq): Likewise.
21288 (add_header_seq): Likewise.
21289 (interleave_supported_p): Likewise.
21290 (vect_maybe_permute_loop_masks): Likewise.
21291 (vect_set_loop_masks_directly): Likewise.
21292 (vect_set_loop_condition_masked): Likewise.
21293 (vect_set_loop_condition_unmasked): New function, split out from
21294 slpeel_make_loop_iterate_ntimes.
21295 (slpeel_make_loop_iterate_ntimes): Rename to..
21296 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
21297 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
21298 (vect_do_peeling): Update call accordingly.
21299 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
21301 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
21302 mask_compare_type, can_fully_mask_p and fully_masked_p.
21303 (release_vec_loop_masks): New function.
21304 (_loop_vec_info): Use it to free the loop masks.
21305 (can_produce_all_loop_masks_p): New function.
21306 (vect_get_max_nscalars_per_iter): Likewise.
21307 (vect_verify_full_masking): Likewise.
21308 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
21309 retries, and free the mask rgroups before retrying. Check loop-wide
21310 reasons for disallowing fully-masked loops. Make the final decision
21311 about whether use a fully-masked loop or not.
21312 (vect_estimate_min_profitable_iters): Do not assume that peeling
21313 for the number of iterations will be needed for fully-masked loops.
21314 (vectorizable_reduction): Disable fully-masked loops.
21315 (vectorizable_live_operation): Likewise.
21316 (vect_halve_mask_nunits): New function.
21317 (vect_double_mask_nunits): Likewise.
21318 (vect_record_loop_mask): Likewise.
21319 (vect_get_loop_mask): Likewise.
21320 (vect_transform_loop): Handle the case in which the final loop
21321 iteration might handle a partial vector. Call vect_set_loop_condition
21322 instead of slpeel_make_loop_iterate_ntimes.
21323 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
21324 (check_load_store_masking): New function.
21325 (prepare_load_store_mask): Likewise.
21326 (vectorizable_store): Handle fully-masked loops.
21327 (vectorizable_load): Likewise.
21328 (supportable_widening_operation): Use vect_halve_mask_nunits for
21330 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
21331 (vect_gen_while): New function.
21332 * config/aarch64/aarch64.md (umax<mode>3): New expander.
21333 (aarch64_uqdec<mode>): New insn.
21335 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21336 Alan Hayward <alan.hayward@arm.com>
21337 David Sherwood <david.sherwood@arm.com>
21339 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
21340 (reduc_xor_scal_optab): New optabs.
21341 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
21342 (reduc_xor_scal_@var{m}): Document.
21343 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
21344 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
21345 internal functions.
21346 * fold-const-call.c (fold_const_call): Handle them.
21347 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
21348 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
21349 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
21350 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
21351 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
21352 (UNSPEC_XORV): New unspecs.
21353 (optab): Add entries for them.
21354 (BITWISEV): New int iterator.
21355 (bit_reduc_op): New int attributes.
21357 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21358 Alan Hayward <alan.hayward@arm.com>
21359 David Sherwood <david.sherwood@arm.com>
21361 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
21362 * internal-fn.def (VEC_SHL_INSERT): New internal function.
21363 * optabs.def (vec_shl_insert_optab): New optab.
21364 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
21365 (duplicate_and_interleave): Likewise.
21366 * tree-vect-loop.c: Include internal-fn.h.
21367 (neutral_op_for_slp_reduction): New function, split out from
21368 get_initial_defs_for_reduction.
21369 (get_initial_def_for_reduction): Handle option 2 for variable-length
21370 vectors by loading the neutral value into a vector and then shifting
21371 the initial value into element 0.
21372 (get_initial_defs_for_reduction): Replace the code argument with
21373 the neutral value calculated by neutral_op_for_slp_reduction.
21374 Use gimple_build_vector for constant-length vectors.
21375 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
21376 but the first group_size elements have a neutral value.
21377 Use duplicate_and_interleave otherwise.
21378 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
21379 Update call to get_initial_defs_for_reduction. Handle SLP
21380 reductions for variable-length vectors by creating one vector
21381 result for each scalar result, with the elements associated
21382 with other scalar results stubbed out with the neutral value.
21383 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
21384 Require IFN_VEC_SHL_INSERT for double reductions on
21385 variable-length vectors, or SLP reductions that have
21386 a neutral value. Require can_duplicate_and_interleave_p
21387 support for variable-length unchained SLP reductions if there
21388 is no neutral value, such as for MIN/MAX reductions. Also require
21389 the number of vector elements to be a multiple of the number of
21390 SLP statements when doing variable-length unchained SLP reductions.
21391 Update call to vect_create_epilog_for_reduction.
21392 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
21393 and remove initial values.
21394 (duplicate_and_interleave): Make public.
21395 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
21396 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
21398 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21399 Alan Hayward <alan.hayward@arm.com>
21400 David Sherwood <david.sherwood@arm.com>
21402 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
21403 (can_duplicate_and_interleave_p): New function.
21404 (vect_get_and_check_slp_defs): Take the vector of statements
21405 rather than just the current one. Remove excess parentheses.
21406 Restriction rejectinon of vect_constant_def and vect_external_def
21407 for variable-length vectors to boolean types, or types for which
21408 can_duplicate_and_interleave_p is false.
21409 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
21410 (duplicate_and_interleave): New function.
21411 (vect_get_constant_vectors): Use gimple_build_vector for
21412 constant-length vectors and suitable variable-length constant
21413 vectors. Use duplicate_and_interleave for other variable-length
21414 vectors. Don't defer the update when inserting new statements.
21416 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21417 Alan Hayward <alan.hayward@arm.com>
21418 David Sherwood <david.sherwood@arm.com>
21420 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
21421 min_profitable_iters doesn't go negative.
21423 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21424 Alan Hayward <alan.hayward@arm.com>
21425 David Sherwood <david.sherwood@arm.com>
21427 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
21428 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
21429 * optabs.def (vec_mask_load_lanes_optab): New optab.
21430 (vec_mask_store_lanes_optab): Likewise.
21431 * internal-fn.def (MASK_LOAD_LANES): New internal function.
21432 (MASK_STORE_LANES): Likewise.
21433 * internal-fn.c (mask_load_lanes_direct): New macro.
21434 (mask_store_lanes_direct): Likewise.
21435 (expand_mask_load_optab_fn): Handle masked operations.
21436 (expand_mask_load_lanes_optab_fn): New macro.
21437 (expand_mask_store_optab_fn): Handle masked operations.
21438 (expand_mask_store_lanes_optab_fn): New macro.
21439 (direct_mask_load_lanes_optab_supported_p): Likewise.
21440 (direct_mask_store_lanes_optab_supported_p): Likewise.
21441 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
21443 (vect_load_lanes_supported): Likewise.
21444 * tree-vect-data-refs.c (strip_conversion): New function.
21445 (can_group_stmts_p): Likewise.
21446 (vect_analyze_data_ref_accesses): Use it instead of checking
21447 for a pair of assignments.
21448 (vect_store_lanes_supported): Take a masked_p parameter.
21449 (vect_load_lanes_supported): Likewise.
21450 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
21451 vect_store_lanes_supported and vect_load_lanes_supported.
21452 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
21453 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
21454 parameter. Don't allow gaps for masked accesses.
21455 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
21456 and vect_load_lanes_supported.
21457 (get_load_store_type): Take a masked_p parameter and update
21458 call to get_group_load_store_type.
21459 (vectorizable_store): Update call to get_load_store_type.
21460 Handle IFN_MASK_STORE_LANES.
21461 (vectorizable_load): Update call to get_load_store_type.
21462 Handle IFN_MASK_LOAD_LANES.
21464 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21465 Alan Hayward <alan.hayward@arm.com>
21466 David Sherwood <david.sherwood@arm.com>
21468 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
21470 * config/aarch64/aarch64-protos.h
21471 (aarch64_sve_struct_memory_operand_p): Declare.
21472 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
21473 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
21474 (VPRED, vpred): Handle SVE structure modes.
21475 * config/aarch64/constraints.md (Utx): New constraint.
21476 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
21477 (aarch64_sve_struct_nonimmediate_operand): New predicates.
21478 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
21479 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
21480 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
21481 structure modes. Split into pieces after RA.
21482 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
21483 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
21485 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
21486 SVE structure modes.
21487 (aarch64_classify_address): Likewise.
21488 (sizetochar): Move earlier in file.
21489 (aarch64_print_operand): Handle SVE register lists.
21490 (aarch64_array_mode): New function.
21491 (aarch64_sve_struct_memory_operand_p): Likewise.
21492 (TARGET_ARRAY_MODE): Redefine.
21494 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21495 Alan Hayward <alan.hayward@arm.com>
21496 David Sherwood <david.sherwood@arm.com>
21498 * target.def (array_mode): New target hook.
21499 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
21500 * doc/tm.texi: Regenerate.
21501 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
21502 * hooks.c (hook_optmode_mode_uhwi_none): New function.
21503 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
21504 targetm.array_mode.
21505 * stor-layout.c (mode_for_array): Likewise. Support polynomial
21508 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21509 Alan Hayward <alan.hayward@arm.com>
21510 David Sherwood <david.sherwood@arm.com>
21512 * fold-const.c (fold_binary_loc): Check the argument types
21513 rather than the result type when testing for a vector operation.
21515 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21517 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
21518 * doc/tm.texi: Regenerate.
21520 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21521 Alan Hayward <alan.hayward@arm.com>
21522 David Sherwood <david.sherwood@arm.com>
21524 * doc/invoke.texi (-msve-vector-bits=): Document new option.
21525 (sve): Document new AArch64 extension.
21526 * doc/md.texi (w): Extend the description of the AArch64
21527 constraint to include SVE vectors.
21528 (Upl, Upa): Document new AArch64 predicate constraints.
21529 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
21531 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
21532 (msve-vector-bits=): New option.
21533 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
21534 SVE when these are disabled.
21535 (sve): New extension.
21536 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
21537 modes. Adjust their number of units based on aarch64_sve_vg.
21538 (MAX_BITSIZE_MODE_ANY_MODE): Define.
21539 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
21540 aarch64_addr_query_type.
21541 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
21542 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
21543 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
21544 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
21545 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
21546 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
21547 (aarch64_simd_imm_zero_p): Delete.
21548 (aarch64_check_zero_based_sve_index_immediate): Declare.
21549 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
21550 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
21551 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
21552 (aarch64_sve_float_mul_immediate_p): Likewise.
21553 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
21554 rather than an rtx.
21555 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
21556 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
21557 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
21558 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
21559 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
21560 (aarch64_regmode_natural_size): Likewise.
21561 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
21562 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
21564 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
21565 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
21566 for VG and the SVE predicate registers.
21567 (V_ALIASES): Add a "z"-prefixed alias.
21568 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
21569 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
21570 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
21571 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
21572 (REG_CLASS_NAMES): Add entries for them.
21573 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
21574 and the predicate registers.
21575 (aarch64_sve_vg): Declare.
21576 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
21577 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
21578 (REGMODE_NATURAL_SIZE): Define.
21579 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
21581 * config/aarch64/aarch64.c: Include cfgrtl.h.
21582 (simd_immediate_info): Add a constructor for series vectors,
21583 and an associated step field.
21584 (aarch64_sve_vg): New variable.
21585 (aarch64_dbx_register_number): Handle VG and the predicate registers.
21586 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
21587 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
21588 (VEC_ANY_DATA, VEC_STRUCT): New constants.
21589 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
21590 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
21591 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
21592 (aarch64_get_mask_mode): New functions.
21593 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
21594 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
21595 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
21596 predicate modes and predicate registers. Explicitly restrict
21597 GPRs to modes of 16 bytes or smaller. Only allow FP registers
21598 to store a vector mode if it is recognized by
21599 aarch64_classify_vector_mode.
21600 (aarch64_regmode_natural_size): New function.
21601 (aarch64_hard_regno_caller_save_mode): Return the original mode
21603 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
21604 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
21605 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
21606 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
21608 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
21609 does not overlap dest if the function is frame-related. Handle
21611 (aarch64_split_add_offset): New function.
21612 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
21613 them aarch64_add_offset.
21614 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
21615 and update call to aarch64_sub_sp.
21616 (aarch64_add_cfa_expression): New function.
21617 (aarch64_expand_prologue): Pass extra temporary registers to the
21618 functions above. Handle the case in which we need to emit new
21619 DW_CFA_expressions for registers that were originally saved
21620 relative to the stack pointer, but now have to be expressed
21621 relative to the frame pointer.
21622 (aarch64_output_mi_thunk): Pass extra temporary registers to the
21624 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
21625 IP0 and IP1 values for SVE frames.
21626 (aarch64_expand_vec_series): New function.
21627 (aarch64_expand_sve_widened_duplicate): Likewise.
21628 (aarch64_expand_sve_const_vector): Likewise.
21629 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
21630 Handle SVE constants. Use emit_move_insn to move a force_const_mem
21631 into the register, rather than emitting a SET directly.
21632 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
21633 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
21634 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
21635 (offset_9bit_signed_scaled_p): New functions.
21636 (aarch64_replicate_bitmask_imm): New function.
21637 (aarch64_bitmask_imm): Use it.
21638 (aarch64_cannot_force_const_mem): Reject expressions involving
21639 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
21640 (aarch64_classify_index): Handle SVE indices, by requiring
21641 a plain register index with a scale that matches the element size.
21642 (aarch64_classify_address): Handle SVE addresses. Assert that
21643 the mode of the address is VOIDmode or an integer mode.
21644 Update call to aarch64_classify_symbol.
21645 (aarch64_classify_symbolic_expression): Update call to
21646 aarch64_classify_symbol.
21647 (aarch64_const_vec_all_in_range_p): New function.
21648 (aarch64_print_vector_float_operand): Likewise.
21649 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
21650 "vN" for FP registers with SVE modes. Handle (const ...) vectors
21651 and the FP immediates 1.0 and 0.5.
21652 (aarch64_print_address_internal): Handle SVE addresses.
21653 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
21654 (aarch64_regno_regclass): Handle predicate registers.
21655 (aarch64_secondary_reload): Handle big-endian reloads of SVE
21657 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
21658 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
21659 (aarch64_convert_sve_vector_bits): New function.
21660 (aarch64_override_options): Use it to handle -msve-vector-bits=.
21661 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
21662 rather than an rtx.
21663 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
21664 Handle SVE vector and predicate modes. Accept VL-based constants
21665 that need only one temporary register, and VL offsets that require
21666 no temporary registers.
21667 (aarch64_conditional_register_usage): Mark the predicate registers
21668 as fixed if SVE isn't available.
21669 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
21670 Return true for SVE vector and predicate modes.
21671 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
21672 rather than an unsigned int. Handle SVE modes.
21673 (aarch64_preferred_simd_mode): Update call accordingly. Handle
21675 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
21677 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
21678 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
21679 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
21680 (aarch64_sve_float_mul_immediate_p): New functions.
21681 (aarch64_sve_valid_immediate): New function.
21682 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
21683 Explicitly reject structure modes. Check for INDEX constants.
21684 Handle PTRUE and PFALSE constants.
21685 (aarch64_check_zero_based_sve_index_immediate): New function.
21686 (aarch64_simd_imm_zero_p): Delete.
21687 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
21688 vector modes. Accept constants in the range of CNT[BHWD].
21689 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
21690 ask for an Advanced SIMD mode.
21691 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
21692 (aarch64_simd_vector_alignment): Handle SVE predicates.
21693 (aarch64_vectorize_preferred_vector_alignment): New function.
21694 (aarch64_simd_vector_alignment_reachable): Use it instead of
21696 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
21697 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
21699 (MAX_VECT_LEN): Delete.
21700 (expand_vec_perm_d): Add a vec_flags field.
21701 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
21702 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
21703 (aarch64_evpc_ext): Don't apply a big-endian lane correction
21705 (aarch64_evpc_rev): Rename to...
21706 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
21707 (aarch64_evpc_rev_global): New function.
21708 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
21709 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
21711 (aarch64_evpc_sve_tbl): New function.
21712 (aarch64_expand_vec_perm_const_1): Update after rename of
21713 aarch64_evpc_rev. Handle SVE permutes too, trying
21714 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
21715 than aarch64_evpc_tbl.
21716 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
21717 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
21718 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
21719 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
21720 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
21721 (aarch64_expand_sve_vcond): New functions.
21722 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
21723 of aarch64_vector_mode_p.
21724 (aarch64_dwarf_poly_indeterminate_value): New function.
21725 (aarch64_compute_pressure_classes): Likewise.
21726 (aarch64_can_change_mode_class): Likewise.
21727 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
21728 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
21729 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
21730 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
21731 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
21732 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
21733 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
21734 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
21736 (Dn, Dl, Dr): Accept const as well as const_vector.
21737 (Dz): Likewise. Compare against CONST0_RTX.
21738 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
21739 of "vector" where appropriate.
21740 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
21741 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
21742 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
21743 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
21744 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
21745 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
21746 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
21747 (v_int_equiv): Extend to SVE modes.
21748 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
21750 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
21751 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
21752 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
21753 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
21754 (SVE_COND_FP_CMP): New int iterators.
21755 (perm_hilo): Handle the new unpack unspecs.
21756 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
21758 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
21759 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
21760 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
21761 (aarch64_equality_operator, aarch64_constant_vector_operand)
21762 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
21763 (aarch64_sve_nonimmediate_operand): Likewise.
21764 (aarch64_sve_general_operand): Likewise.
21765 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
21766 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
21767 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
21768 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
21769 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
21770 (aarch64_sve_float_arith_immediate): Likewise.
21771 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
21772 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
21773 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
21774 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
21775 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
21776 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
21777 (aarch64_sve_float_arith_operand): Likewise.
21778 (aarch64_sve_float_arith_with_sub_operand): Likewise.
21779 (aarch64_sve_float_mul_operand): Likewise.
21780 (aarch64_sve_vec_perm_operand): Likewise.
21781 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
21782 (aarch64_mov_operand): Accept const_poly_int and const_vector.
21783 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
21784 as well as const_vector.
21785 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
21786 in file. Use CONST0_RTX and CONSTM1_RTX.
21787 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
21788 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
21789 Use aarch64_simd_imm_zero.
21790 * config/aarch64/aarch64-sve.md: New file.
21791 * config/aarch64/aarch64.md: Include it.
21792 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
21793 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
21794 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
21795 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
21796 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
21797 (sve): New attribute.
21798 (enabled): Disable instructions with the sve attribute unless
21800 (movqi, movhi): Pass CONST_POLY_INT operaneds through
21801 aarch64_expand_mov_immediate.
21802 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
21803 CNT[BHSD] immediates.
21804 (movti): Split CONST_POLY_INT moves into two halves.
21805 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
21806 Split additions that need a temporary here if the destination
21807 is the stack pointer.
21808 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
21809 (*add<mode>3_poly_1): New instruction.
21810 (set_clobber_cc): New expander.
21812 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21814 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
21815 parameter and use it instead of GET_MODE_SIZE (innermode). Use
21816 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
21817 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
21818 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
21819 Change innermode from fixed_mode_size to machine_mode.
21820 (simplify_subreg): Update call accordingly. Handle a constant-sized
21821 subreg of a variable-length CONST_VECTOR.
21823 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21824 Alan Hayward <alan.hayward@arm.com>
21825 David Sherwood <david.sherwood@arm.com>
21827 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
21828 (add_offset_to_base): New function, split out from...
21829 (create_mem_ref): ...here. When handling a scale other than 1,
21830 check first whether the address is valid without the offset.
21831 Add it into the base if so, leaving the index and scale as-is.
21833 2018-01-12 Jakub Jelinek <jakub@redhat.com>
21836 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
21837 fold_for_warn before checking if arg2 is INTEGER_CST.
21839 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
21841 * config/rs6000/predicates.md (load_multiple_operation): Delete.
21842 (store_multiple_operation): Delete.
21843 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
21844 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
21845 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
21846 guarded by TARGET_STRING.
21847 (rs6000_output_load_multiple): Delete.
21848 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
21849 OPTION_MASK_STRING / TARGET_STRING handling.
21850 (print_operand) <'N', 'O'>: Add comment that these are unused now.
21851 (const rs6000_opt_masks) <"string">: Change mask to 0.
21852 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
21853 (MASK_STRING): Delete.
21854 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
21856 (load_multiple): Delete.
21863 (store_multiple): Delete.
21870 (movmemsi_8reg): Delete.
21871 (corresponding unnamed define_insn): Delete.
21872 (movmemsi_6reg): Delete.
21873 (corresponding unnamed define_insn): Delete.
21874 (movmemsi_4reg): Delete.
21875 (corresponding unnamed define_insn): Delete.
21876 (movmemsi_2reg): Delete.
21877 (corresponding unnamed define_insn): Delete.
21878 (movmemsi_1reg): Delete.
21879 (corresponding unnamed define_insn): Delete.
21880 * config/rs6000/rs6000.opt (mno-string): New.
21881 (mstring): Replace by deprecation warning stub.
21882 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
21884 2018-01-12 Jakub Jelinek <jakub@redhat.com>
21886 * regrename.c (regrename_do_replace): If replacing the same
21887 reg multiple times, try to reuse last created gen_raw_REG.
21890 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
21891 main to workaround a bug in GDB.
21893 2018-01-12 Tom de Vries <tom@codesourcery.com>
21896 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
21898 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
21900 PR rtl-optimization/80481
21901 * ira-color.c (get_cap_member): New function.
21902 (allocnos_conflict_by_live_ranges_p): Use it.
21903 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
21904 (setup_slot_coalesced_allocno_live_ranges): Ditto.
21906 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
21909 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
21910 (*saddl_se_1): Ditto.
21911 (*ssubsi_1): Ditto.
21912 (*ssubl_se_1): Ditto.
21914 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
21916 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
21917 rather than wi::to_widest for DR_INITs.
21918 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
21919 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
21920 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
21922 (vect_analyze_group_access_1): Note that here.
21924 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
21926 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
21927 polynomial type sizes.
21929 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
21931 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
21932 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
21933 (gimple_add_tmp_var): Likewise.
21935 2018-01-12 Martin Liska <mliska@suse.cz>
21937 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
21938 (gimple_alloc_sizes): Likewise.
21939 (dump_gimple_statistics): Use PRIu64 in printf format.
21940 * gimple.h: Change uint64_t to int.
21942 2018-01-12 Martin Liska <mliska@suse.cz>
21944 * tree-core.h: Use uint64_t instead of int.
21945 * tree.c (tree_node_counts): Likewise.
21946 (tree_node_sizes): Likewise.
21947 (dump_tree_statistics): Use PRIu64 in printf format.
21949 2018-01-12 Martin Liska <mliska@suse.cz>
21951 * Makefile.in: As qsort_chk is implemented in vec.c, add
21952 vec.o to linkage of gencfn-macros.
21953 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
21954 passing the info to record_node_allocation_statistics.
21955 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
21957 * ggc-common.c (struct ggc_usage): Add operator== and use
21958 it in operator< and compare function.
21959 * mem-stats.h (struct mem_usage): Likewise.
21960 * vec.c (struct vec_usage): Remove operator< and compare
21961 function. Can be simply inherited.
21963 2018-01-12 Martin Jambor <mjambor@suse.cz>
21966 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
21967 * tree-ssa-math-opts.c: Include domwalk.h.
21968 (convert_mult_to_fma_1): New function.
21969 (fma_transformation_info): New type.
21970 (fma_deferring_state): Likewise.
21971 (cancel_fma_deferring): New function.
21972 (result_of_phi): Likewise.
21973 (last_fma_candidate_feeds_initial_phi): Likewise.
21974 (convert_mult_to_fma): Added deferring logic, split actual
21975 transformation to convert_mult_to_fma_1.
21976 (math_opts_dom_walker): New type.
21977 (math_opts_dom_walker::after_dom_children): New method, body moved
21978 here from pass_optimize_widening_mul::execute, added deferring logic
21980 (pass_optimize_widening_mul::execute): Moved most of code to
21981 math_opts_dom_walker::after_dom_children.
21982 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
21983 * config/i386/i386.c (ix86_option_override_internal): Added
21984 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
21986 2018-01-12 Richard Biener <rguenther@suse.de>
21989 * dwarf2out.c (gen_variable_die): Do not reset old_die for
21990 inline instance vars.
21992 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
21995 * config/rx/rx.c (rx_is_restricted_memory_address):
21996 Handle SUBREG case.
21998 2018-01-12 Richard Biener <rguenther@suse.de>
22000 PR tree-optimization/80846
22001 * target.def (split_reduction): New target hook.
22002 * targhooks.c (default_split_reduction): New function.
22003 * targhooks.h (default_split_reduction): Declare.
22004 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
22005 target requests first reduce vectors by combining low and high
22007 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
22008 (get_vectype_for_scalar_type_and_size): Export.
22009 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
22010 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
22011 * doc/tm.texi: Regenerate.
22012 * config/i386/i386.c (ix86_split_reduction): Implement
22013 TARGET_VECTORIZE_SPLIT_REDUCTION.
22015 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
22018 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
22019 in PIC mode except for TARGET_VXWORKS_RTP.
22020 * config/sparc/sparc.c: Include cfgrtl.h.
22021 (TARGET_INIT_PIC_REG): Define.
22022 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
22023 (sparc_pic_register_p): New predicate.
22024 (sparc_legitimate_address_p): Use it.
22025 (sparc_legitimize_pic_address): Likewise.
22026 (sparc_delegitimize_address): Likewise.
22027 (sparc_mode_dependent_address_p): Likewise.
22028 (gen_load_pcrel_sym): Remove 4th parameter.
22029 (load_got_register): Adjust call to above. Remove obsolete stuff.
22030 (sparc_expand_prologue): Do not call load_got_register here.
22031 (sparc_flat_expand_prologue): Likewise.
22032 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
22033 (sparc_use_pseudo_pic_reg): New function.
22034 (sparc_init_pic_reg): Likewise.
22035 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
22036 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
22038 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
22040 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
22041 Add item for branch_cost.
22043 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
22045 PR rtl-optimization/83565
22046 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
22047 not extend the result to a larger mode for rotate operations.
22048 (num_sign_bit_copies1): Likewise.
22050 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
22053 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
22055 Use values-Xc.o for -pedantic.
22056 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
22058 2018-01-12 Martin Liska <mliska@suse.cz>
22061 * ipa-devirt.c (final_warning_record::grow_type_warnings):
22063 (possible_polymorphic_call_targets): Use it.
22064 (ipa_devirt): Likewise.
22066 2018-01-12 Martin Liska <mliska@suse.cz>
22068 * profile-count.h (enum profile_quality): Use 0 as invalid
22069 enum value of profile_quality.
22071 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
22073 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
22074 -mext-string options.
22076 2018-01-12 Richard Biener <rguenther@suse.de>
22078 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
22079 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
22080 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
22082 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
22084 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
22086 * configure.ac (--with-long-double-format): Add support for the
22087 configuration option to change the default long double format on
22089 * config.gcc (powerpc*-linux*-*): Likewise.
22090 * configure: Regenerate.
22091 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
22092 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
22093 used without modification.
22095 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
22097 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
22098 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
22099 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
22100 MISC_BUILTIN_SPEC_BARRIER.
22101 (rs6000_init_builtins): Likewise.
22102 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
22104 (speculation_barrier): New define_insn.
22105 * doc/extend.texi: Document __builtin_speculation_barrier.
22107 2018-01-11 Jakub Jelinek <jakub@redhat.com>
22110 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
22111 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
22112 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
22114 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
22115 integral modes instead of "ss" and "sd".
22116 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
22117 vectors with 32-bit and 64-bit elements.
22118 (vecdupssescalarmodesuffix): New mode attribute.
22119 (vec_dup<mode>): Use it.
22121 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
22124 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
22125 frame if argument is passed on stack.
22127 2018-01-11 Jakub Jelinek <jakub@redhat.com>
22130 * ree.c (combine_reaching_defs): Optimize also
22131 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
22132 reg2=any_extend(exp); reg1=reg2;, formatting fix.
22134 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
22136 PR middle-end/83189
22137 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
22139 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
22141 PR middle-end/83718
22142 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
22143 after they are computed.
22145 2018-01-11 Bin Cheng <bin.cheng@arm.com>
22147 PR tree-optimization/83695
22148 * gimple-loop-linterchange.cc
22149 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
22150 reset cached scev information after interchange.
22151 (pass_linterchange::execute): Remove call to scev_reset_htab.
22153 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22155 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
22156 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
22157 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
22158 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
22159 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
22160 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
22161 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
22162 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
22163 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
22164 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
22165 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
22166 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
22167 (V_lane_reg): Likewise.
22168 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
22170 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
22171 (vfmal_lane_low<mode>_intrinsic,
22172 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
22173 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
22174 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
22175 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
22176 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
22177 vfmsl_lane_high<mode>_intrinsic): New define_insns.
22179 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22181 * config/arm/arm-cpus.in (fp16fml): New feature.
22182 (ALL_SIMD): Add fp16fml.
22183 (armv8.2-a): Add fp16fml as an option.
22184 (armv8.3-a): Likewise.
22185 (armv8.4-a): Add fp16fml as part of fp16.
22186 * config/arm/arm.h (TARGET_FP16FML): Define.
22187 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
22189 * config/arm/arm-modes.def (V2HF): Define.
22190 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
22191 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
22192 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
22193 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
22194 vfmsl_low, vfmsl_high): New set of builtins.
22195 * config/arm/iterators.md (PLUSMINUS): New code iterator.
22196 (vfml_op): New code attribute.
22197 (VFMLHALVES): New int iterator.
22198 (VFML, VFMLSEL): New mode attributes.
22199 (V_reg): Define mapping for V2HF.
22200 (V_hi, V_lo): New mode attributes.
22201 (VF_constraint): Likewise.
22202 (vfml_half, vfml_half_selector): New int attributes.
22203 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
22205 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
22206 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
22208 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
22209 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
22210 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
22211 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
22213 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
22214 Document new effective target and option set.
22216 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22218 * config/arm/arm-cpus.in (armv8_4): New feature.
22219 (ARMv8_4a): New fgroup.
22220 (armv8.4-a): New arch.
22221 * config/arm/arm-tables.opt: Regenerate.
22222 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
22223 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
22224 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
22225 Add matching rules for -march=armv8.4-a and extensions.
22226 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
22228 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
22231 * config/rx/rx.md (BW): New mode attribute.
22232 (sync_lock_test_and_setsi): Add mode suffix to insn output.
22234 2018-01-11 Richard Biener <rguenther@suse.de>
22236 PR tree-optimization/83435
22237 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
22238 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
22239 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
22241 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
22242 Alan Hayward <alan.hayward@arm.com>
22243 David Sherwood <david.sherwood@arm.com>
22245 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
22247 (aarch64_classify_address): Initialize it. Track polynomial offsets.
22248 (aarch64_print_address_internal): Use it to check for a zero offset.
22250 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
22251 Alan Hayward <alan.hayward@arm.com>
22252 David Sherwood <david.sherwood@arm.com>
22254 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
22255 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
22256 Return a poly_int64 rather than a HOST_WIDE_INT.
22257 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
22258 rather than a HOST_WIDE_INT.
22259 * config/aarch64/aarch64.h (aarch64_frame): Protect with
22260 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
22261 hard_fp_offset, frame_size, initial_adjust, callee_offset and
22262 final_offset from HOST_WIDE_INT to poly_int64.
22263 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
22264 to_constant when getting the number of units in an Advanced SIMD
22266 (aarch64_builtin_vectorized_function): Check for a constant number
22268 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
22270 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
22271 attribute instead of GET_MODE_NUNITS.
22272 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
22273 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
22274 GET_MODE_SIZE for fixed-size registers.
22275 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
22276 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
22277 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
22278 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
22279 (aarch64_print_operand, aarch64_print_address_internal)
22280 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
22281 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
22282 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
22283 Handle polynomial GET_MODE_SIZE.
22284 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
22285 wider than SImode without modification.
22286 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
22287 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
22288 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
22289 passing and returning SVE modes.
22290 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
22291 rather than GEN_INT.
22292 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
22293 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
22294 (aarch64_allocate_and_probe_stack_space): Likewise.
22295 (aarch64_layout_frame): Cope with polynomial offsets.
22296 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
22297 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
22298 polynomial offsets.
22299 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
22300 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
22301 poly_int64 rather than a HOST_WIDE_INT.
22302 (aarch64_get_separate_components, aarch64_process_components)
22303 (aarch64_expand_prologue, aarch64_expand_epilogue)
22304 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
22305 (aarch64_anchor_offset): New function, split out from...
22306 (aarch64_legitimize_address): ...here.
22307 (aarch64_builtin_vectorization_cost): Handle polynomial
22308 TYPE_VECTOR_SUBPARTS.
22309 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
22311 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
22312 number of elements from the PARALLEL rather than the mode.
22313 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
22314 rather than GET_MODE_BITSIZE.
22315 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
22316 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
22317 (aarch64_expand_vec_perm_const_1): Handle polynomial
22318 d->perm.length () and d->perm elements.
22319 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
22320 Apply to_constant to d->perm elements.
22321 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
22322 polynomial CONST_VECTOR_NUNITS.
22323 (aarch64_move_pointer): Take amount as a poly_int64 rather
22325 (aarch64_progress_pointer): Avoid temporary variable.
22326 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
22327 the mode attribute instead of GET_MODE.
22329 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
22330 Alan Hayward <alan.hayward@arm.com>
22331 David Sherwood <david.sherwood@arm.com>
22333 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
22334 x exists before using it.
22335 (aarch64_add_constant_internal): Rename to...
22336 (aarch64_add_offset_1): ...this. Replace regnum with separate
22337 src and dest rtxes. Handle the case in which they're different,
22338 including when the offset is zero. Replace scratchreg with an rtx.
22339 Use 2 additions if there is no spare register into which we can
22340 move a 16-bit constant.
22341 (aarch64_add_constant): Delete.
22342 (aarch64_add_offset): Replace reg with separate src and dest
22343 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
22344 Use aarch64_add_offset_1.
22345 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
22346 an rtx rather than an int. Take the delta as a poly_int64
22347 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
22348 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
22349 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
22350 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
22351 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
22352 and aarch64_add_sp.
22353 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
22354 aarch64_add_constant.
22356 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
22358 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
22359 Use scalar_float_mode.
22361 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
22363 * config/aarch64/aarch64-simd.md
22364 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
22365 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
22366 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
22367 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
22368 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
22369 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
22370 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
22371 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
22372 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
22373 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
22375 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
22378 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
22379 targ_options->x_arm_arch_string is non NULL.
22381 2018-01-11 Tamar Christina <tamar.christina@arm.com>
22383 * config/aarch64/aarch64.h
22384 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
22386 2018-01-11 Sudakshina Das <sudi.das@arm.com>
22389 * expmed.c (emit_store_flag_force): Swap if const op0
22390 and change VOIDmode to mode of op0.
22392 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
22394 PR rtl-optimization/83761
22395 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
22396 than bytes to mode_for_size.
22398 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
22400 PR middle-end/83189
22401 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
22402 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
22405 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
22407 PR middle-end/83575
22408 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
22409 when in layout mode.
22410 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
22411 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
22414 2018-01-10 Michael Collison <michael.collison@arm.com>
22416 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
22417 * config/aarch64/aarch64-option-extension.def: Add
22418 AARCH64_OPT_EXTENSION of 'fp16fml'.
22419 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
22420 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
22421 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
22422 * config/aarch64/constraints.md (Ui7): New constraint.
22423 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
22424 (VFMLA_SEL_W): Ditto.
22427 (VFMLA16_LOW): New int iterator.
22428 (VFMLA16_HIGH): Ditto.
22429 (UNSPEC_FMLAL): New unspec.
22430 (UNSPEC_FMLSL): Ditto.
22431 (UNSPEC_FMLAL2): Ditto.
22432 (UNSPEC_FMLSL2): Ditto.
22433 (f16mac): New code attribute.
22434 * config/aarch64/aarch64-simd-builtins.def
22435 (aarch64_fmlal_lowv2sf): Ditto.
22436 (aarch64_fmlsl_lowv2sf): Ditto.
22437 (aarch64_fmlalq_lowv4sf): Ditto.
22438 (aarch64_fmlslq_lowv4sf): Ditto.
22439 (aarch64_fmlal_highv2sf): Ditto.
22440 (aarch64_fmlsl_highv2sf): Ditto.
22441 (aarch64_fmlalq_highv4sf): Ditto.
22442 (aarch64_fmlslq_highv4sf): Ditto.
22443 (aarch64_fmlal_lane_lowv2sf): Ditto.
22444 (aarch64_fmlsl_lane_lowv2sf): Ditto.
22445 (aarch64_fmlal_laneq_lowv2sf): Ditto.
22446 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
22447 (aarch64_fmlalq_lane_lowv4sf): Ditto.
22448 (aarch64_fmlsl_lane_lowv4sf): Ditto.
22449 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
22450 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
22451 (aarch64_fmlal_lane_highv2sf): Ditto.
22452 (aarch64_fmlsl_lane_highv2sf): Ditto.
22453 (aarch64_fmlal_laneq_highv2sf): Ditto.
22454 (aarch64_fmlsl_laneq_highv2sf): Ditto.
22455 (aarch64_fmlalq_lane_highv4sf): Ditto.
22456 (aarch64_fmlsl_lane_highv4sf): Ditto.
22457 (aarch64_fmlalq_laneq_highv4sf): Ditto.
22458 (aarch64_fmlsl_laneq_highv4sf): Ditto.
22459 * config/aarch64/aarch64-simd.md:
22460 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
22461 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
22462 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
22463 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
22464 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
22465 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
22466 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
22467 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
22468 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
22469 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
22470 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
22471 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
22472 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
22473 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
22474 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
22475 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
22476 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
22477 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
22478 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
22479 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
22480 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
22481 (vfmlsl_low_u32): Ditto.
22482 (vfmlalq_low_u32): Ditto.
22483 (vfmlslq_low_u32): Ditto.
22484 (vfmlal_high_u32): Ditto.
22485 (vfmlsl_high_u32): Ditto.
22486 (vfmlalq_high_u32): Ditto.
22487 (vfmlslq_high_u32): Ditto.
22488 (vfmlal_lane_low_u32): Ditto.
22489 (vfmlsl_lane_low_u32): Ditto.
22490 (vfmlal_laneq_low_u32): Ditto.
22491 (vfmlsl_laneq_low_u32): Ditto.
22492 (vfmlalq_lane_low_u32): Ditto.
22493 (vfmlslq_lane_low_u32): Ditto.
22494 (vfmlalq_laneq_low_u32): Ditto.
22495 (vfmlslq_laneq_low_u32): Ditto.
22496 (vfmlal_lane_high_u32): Ditto.
22497 (vfmlsl_lane_high_u32): Ditto.
22498 (vfmlal_laneq_high_u32): Ditto.
22499 (vfmlsl_laneq_high_u32): Ditto.
22500 (vfmlalq_lane_high_u32): Ditto.
22501 (vfmlslq_lane_high_u32): Ditto.
22502 (vfmlalq_laneq_high_u32): Ditto.
22503 (vfmlslq_laneq_high_u32): Ditto.
22504 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
22505 (AARCH64_FL_FOR_ARCH8_4): New.
22506 (AARCH64_ISA_F16FML): New ISA flag.
22507 (TARGET_F16FML): New feature flag for fp16fml.
22508 (doc/invoke.texi): Document new fp16fml option.
22510 2018-01-10 Michael Collison <michael.collison@arm.com>
22512 * config/aarch64/aarch64-builtins.c:
22513 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
22514 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
22515 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
22516 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
22517 (AARCH64_ISA_SHA3): New ISA flag.
22518 (TARGET_SHA3): New feature flag for sha3.
22519 * config/aarch64/iterators.md (sha512_op): New int attribute.
22520 (CRYPTO_SHA512): New int iterator.
22521 (UNSPEC_SHA512H): New unspec.
22522 (UNSPEC_SHA512H2): Ditto.
22523 (UNSPEC_SHA512SU0): Ditto.
22524 (UNSPEC_SHA512SU1): Ditto.
22525 * config/aarch64/aarch64-simd-builtins.def
22526 (aarch64_crypto_sha512hqv2di): New builtin.
22527 (aarch64_crypto_sha512h2qv2di): Ditto.
22528 (aarch64_crypto_sha512su0qv2di): Ditto.
22529 (aarch64_crypto_sha512su1qv2di): Ditto.
22530 (aarch64_eor3qv8hi): Ditto.
22531 (aarch64_rax1qv2di): Ditto.
22532 (aarch64_xarqv2di): Ditto.
22533 (aarch64_bcaxqv8hi): Ditto.
22534 * config/aarch64/aarch64-simd.md:
22535 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
22536 (aarch64_crypto_sha512su0qv2di): Ditto.
22537 (aarch64_crypto_sha512su1qv2di): Ditto.
22538 (aarch64_eor3qv8hi): Ditto.
22539 (aarch64_rax1qv2di): Ditto.
22540 (aarch64_xarqv2di): Ditto.
22541 (aarch64_bcaxqv8hi): Ditto.
22542 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
22543 (vsha512h2q_u64): Ditto.
22544 (vsha512su0q_u64): Ditto.
22545 (vsha512su1q_u64): Ditto.
22546 (veor3q_u16): Ditto.
22547 (vrax1q_u64): Ditto.
22548 (vxarq_u64): Ditto.
22549 (vbcaxq_u16): Ditto.
22550 * config/arm/types.md (crypto_sha512): New type attribute.
22551 (crypto_sha3): Ditto.
22552 (doc/invoke.texi): Document new sha3 option.
22554 2018-01-10 Michael Collison <michael.collison@arm.com>
22556 * config/aarch64/aarch64-builtins.c:
22557 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
22558 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
22559 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
22560 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
22561 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
22562 (AARCH64_ISA_SM4): New ISA flag.
22563 (TARGET_SM4): New feature flag for sm4.
22564 * config/aarch64/aarch64-simd-builtins.def
22565 (aarch64_sm3ss1qv4si): Ditto.
22566 (aarch64_sm3tt1aq4si): Ditto.
22567 (aarch64_sm3tt1bq4si): Ditto.
22568 (aarch64_sm3tt2aq4si): Ditto.
22569 (aarch64_sm3tt2bq4si): Ditto.
22570 (aarch64_sm3partw1qv4si): Ditto.
22571 (aarch64_sm3partw2qv4si): Ditto.
22572 (aarch64_sm4eqv4si): Ditto.
22573 (aarch64_sm4ekeyqv4si): Ditto.
22574 * config/aarch64/aarch64-simd.md:
22575 (aarch64_sm3ss1qv4si): Ditto.
22576 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
22577 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
22578 (aarch64_sm4eqv4si): Ditto.
22579 (aarch64_sm4ekeyqv4si): Ditto.
22580 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
22581 (sm3part_op): Ditto.
22582 (CRYPTO_SM3TT): Ditto.
22583 (CRYPTO_SM3PART): Ditto.
22584 (UNSPEC_SM3SS1): New unspec.
22585 (UNSPEC_SM3TT1A): Ditto.
22586 (UNSPEC_SM3TT1B): Ditto.
22587 (UNSPEC_SM3TT2A): Ditto.
22588 (UNSPEC_SM3TT2B): Ditto.
22589 (UNSPEC_SM3PARTW1): Ditto.
22590 (UNSPEC_SM3PARTW2): Ditto.
22591 (UNSPEC_SM4E): Ditto.
22592 (UNSPEC_SM4EKEY): Ditto.
22593 * config/aarch64/constraints.md (Ui2): New constraint.
22594 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
22595 * config/arm/types.md (crypto_sm3): New type attribute.
22596 (crypto_sm4): Ditto.
22597 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
22598 (vsm3tt1aq_u32): Ditto.
22599 (vsm3tt1bq_u32): Ditto.
22600 (vsm3tt2aq_u32): Ditto.
22601 (vsm3tt2bq_u32): Ditto.
22602 (vsm3partw1q_u32): Ditto.
22603 (vsm3partw2q_u32): Ditto.
22604 (vsm4eq_u32): Ditto.
22605 (vsm4ekeyq_u32): Ditto.
22606 (doc/invoke.texi): Document new sm4 option.
22608 2018-01-10 Michael Collison <michael.collison@arm.com>
22610 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
22611 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
22612 (AARCH64_FL_FOR_ARCH8_4): New.
22613 (AARCH64_FL_V8_4): New flag.
22614 (doc/invoke.texi): Document new armv8.4-a option.
22616 2018-01-10 Michael Collison <michael.collison@arm.com>
22618 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
22619 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
22620 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
22621 * config/aarch64/aarch64-option-extension.def: Add
22622 AARCH64_OPT_EXTENSION of 'sha2'.
22623 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
22624 (crypto): Disable sha2 and aes if crypto disabled.
22625 (crypto): Enable aes and sha2 if enabled.
22626 (simd): Disable sha2 and aes if simd disabled.
22627 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
22629 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
22630 (TARGET_SHA2): New feature flag for sha2.
22631 (TARGET_AES): New feature flag for aes.
22632 * config/aarch64/aarch64-simd.md:
22633 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
22634 conditional on TARGET_AES.
22635 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
22636 (aarch64_crypto_sha1hsi): Make pattern conditional
22638 (aarch64_crypto_sha1hv4si): Ditto.
22639 (aarch64_be_crypto_sha1hv4si): Ditto.
22640 (aarch64_crypto_sha1su1v4si): Ditto.
22641 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
22642 (aarch64_crypto_sha1su0v4si): Ditto.
22643 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
22644 (aarch64_crypto_sha256su0v4si): Ditto.
22645 (aarch64_crypto_sha256su1v4si): Ditto.
22646 (doc/invoke.texi): Document new aes and sha2 options.
22648 2018-01-10 Martin Sebor <msebor@redhat.com>
22650 PR tree-optimization/83781
22651 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
22654 2018-01-11 Martin Sebor <msebor@gmail.com>
22655 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
22657 PR tree-optimization/83501
22658 PR tree-optimization/81703
22660 * tree-ssa-strlen.c (get_string_cst): Rename...
22661 (get_string_len): ...to this. Handle global constants.
22662 (handle_char_store): Adjust.
22664 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
22665 Jim Wilson <jimw@sifive.com>
22667 * config/riscv/riscv-protos.h (riscv_output_return): New.
22668 * config/riscv/riscv.c (struct machine_function): New naked_p field.
22669 (riscv_attribute_table, riscv_output_return),
22670 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
22671 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
22672 (riscv_compute_frame_info): Only compute frame->mask if not a naked
22674 (riscv_expand_prologue): Add early return for naked function.
22675 (riscv_expand_epilogue): Likewise.
22676 (riscv_function_ok_for_sibcall): Return false for naked function.
22677 (riscv_set_current_function): New.
22678 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
22679 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
22680 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
22681 * doc/extend.texi (RISC-V Function Attributes): New.
22683 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
22685 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
22686 check for 128-bit long double before checking TCmode.
22687 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
22688 128-bit long doubles before checking TFmode or TCmode.
22689 (FLOAT128_IBM_P): Likewise.
22691 2018-01-10 Martin Sebor <msebor@redhat.com>
22693 PR tree-optimization/83671
22694 * builtins.c (c_strlen): Unconditionally return zero for the empty
22696 Use -Warray-bounds for warnings.
22697 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
22698 for non-constant array indices with COMPONENT_REF, arrays of
22699 arrays, and pointers to arrays.
22700 (gimple_fold_builtin_strlen): Determine and set length range for
22701 non-constant character arrays.
22703 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
22705 PR middle-end/81897
22706 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
22709 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
22711 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
22713 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
22716 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
22717 VECTOR_MEM_ALTIVEC_OR_VSX_P.
22718 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
22719 indexed_or_indirect_operand predicate.
22720 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
22721 (*vsx_le_perm_load_v8hi): Likewise.
22722 (*vsx_le_perm_load_v16qi): Likewise.
22723 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
22724 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
22725 (*vsx_le_perm_store_v8hi): Likewise.
22726 (*vsx_le_perm_store_v16qi): Likewise.
22727 (eight unnamed splitters): Likewise.
22729 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
22731 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
22732 * config/rs6000/emmintrin.h: Likewise.
22733 * config/rs6000/mmintrin.h: Likewise.
22734 * config/rs6000/xmmintrin.h: Likewise.
22736 2018-01-10 David Malcolm <dmalcolm@redhat.com>
22739 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
22741 * tree.c (tree_nop_conversion): Return true for location wrapper
22743 (maybe_wrap_with_location): New function.
22744 (selftest::check_strip_nops): New function.
22745 (selftest::test_location_wrappers): New function.
22746 (selftest::tree_c_tests): Call it.
22747 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
22748 (maybe_wrap_with_location): New decl.
22749 (EXPR_LOCATION_WRAPPER_P): New macro.
22750 (location_wrapper_p): New inline function.
22751 (tree_strip_any_location_wrapper): New inline function.
22753 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
22756 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
22757 stack_realign_offset for the largest alignment of stack slot
22759 (ix86_find_max_used_stack_alignment): New function.
22760 (ix86_finalize_stack_frame_flags): Use it. Set
22761 max_used_stack_alignment if we don't realign stack.
22762 * config/i386/i386.h (machine_function): Add
22763 max_used_stack_alignment.
22765 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
22767 * config/arm/arm.opt (-mbranch-cost): New option.
22768 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
22771 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
22774 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
22775 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
22777 2018-01-10 Richard Biener <rguenther@suse.de>
22780 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
22781 early out so it also covers the case where we have a non-NULL
22784 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
22786 PR tree-optimization/83753
22787 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
22788 for non-strided grouped accesses if the number of elements is 1.
22790 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
22793 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
22794 * i386.h (TARGET_USE_GATHER): Define.
22795 * x86-tune.def (X86_TUNE_USE_GATHER): New.
22797 2018-01-10 Martin Liska <mliska@suse.cz>
22800 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
22801 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
22803 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
22804 CLEANUP_NO_PARTITIONING is not set.
22806 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
22808 * doc/rtl.texi: Remove documentation of (const ...) wrappers
22809 for vectors, as a partial revert of r254296.
22810 * rtl.h (const_vec_p): Delete.
22811 (const_vec_duplicate_p): Don't test for vector CONSTs.
22812 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
22813 * expmed.c (make_tree): Likewise.
22816 * common.md (E, F): Use CONSTANT_P instead of checking for
22818 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
22819 checking for CONST_VECTOR.
22821 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
22823 PR middle-end/83575
22824 * predict.c (force_edge_cold): Handle in more sane way edges
22825 with no prediction.
22827 2018-01-09 Carl Love <cel@us.ibm.com>
22829 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
22831 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
22832 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
22833 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
22834 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
22835 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
22836 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
22837 * config/rs6000/rs6000-protos.h: Add extern defition for
22838 rs6000_generate_float2_double_code.
22839 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
22841 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
22842 (float2_v2df): Add define_expand.
22844 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
22847 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
22848 op_mode in the force_to_mode call.
22850 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
22852 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
22853 instead of checking each element individually.
22854 (aarch64_evpc_uzp): Likewise.
22855 (aarch64_evpc_zip): Likewise.
22856 (aarch64_evpc_ext): Likewise.
22857 (aarch64_evpc_rev): Likewise.
22858 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
22859 instead of checking each element individually. Return true without
22861 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
22862 whether all selected elements come from the same input, instead of
22863 checking each element individually. Remove calls to gen_rtx_REG,
22864 start_sequence and end_sequence and instead assert that no rtl is
22867 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
22869 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
22870 order of HIGH and CONST checks.
22872 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
22874 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
22875 if the destination isn't an SSA_NAME.
22877 2018-01-09 Richard Biener <rguenther@suse.de>
22879 PR tree-optimization/83668
22880 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
22882 (canonicalize_loop_form): ... here, renamed from ...
22883 (canonicalize_loop_closed_ssa_form): ... this and amended to
22884 swap successor edges for loop exit blocks to make us use
22885 the RPO order we need for initial schedule generation.
22887 2018-01-09 Joseph Myers <joseph@codesourcery.com>
22889 PR tree-optimization/64811
22890 * match.pd: When optimizing comparisons with Inf, avoid
22891 introducing or losing exceptions from comparisons with NaN.
22893 2018-01-09 Martin Liska <mliska@suse.cz>
22896 * asan.c (shadow_mem_size): Add gcc_assert.
22898 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
22900 Don't save registers in main().
22903 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
22904 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
22905 * config/avr/avr.c (avr_set_current_function): Don't error if
22906 naked, OS_task or OS_main are specified at the same time.
22907 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
22909 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
22911 * common/config/avr/avr-common.c (avr_option_optimization_table):
22912 Switch on -mmain-is-OS_task for optimizing compilations.
22914 2018-01-09 Richard Biener <rguenther@suse.de>
22916 PR tree-optimization/83572
22917 * graphite.c: Include cfganal.h.
22918 (graphite_transform_loops): Connect infinite loops to exit
22919 and remove fake edges at the end.
22921 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
22923 * ipa-inline.c (edge_badness): Revert accidental checkin.
22925 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
22928 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
22929 symbols; not inline clones.
22931 2018-01-09 Jakub Jelinek <jakub@redhat.com>
22934 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
22935 hard registers. Formatting fixes.
22937 PR preprocessor/83722
22938 * gcc.c (try_generate_repro): Pass
22939 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
22940 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
22943 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
22944 Kito Cheng <kito.cheng@gmail.com>
22946 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
22947 (riscv_leaf_function_p): Delete.
22948 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
22950 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
22952 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
22954 (do_ifelse): New function.
22955 (do_isel): New function.
22956 (do_sub3): New function.
22957 (do_add3): New function.
22958 (do_load_mask_compare): New function.
22959 (do_overlap_load_compare): New function.
22960 (expand_compare_loop): New function.
22961 (expand_block_compare): Call expand_compare_loop() when appropriate.
22962 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
22963 option description.
22964 (-mblock-compare-inline-loop-limit): New option.
22966 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
22969 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
22970 Reverse order of second and third operands in first alternative.
22971 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
22972 of first and second elements in UNSPEC_VPERMR vector.
22973 (altivec_expand_vec_perm_le): Likewise.
22975 2018-01-08 Jeff Law <law@redhat.com>
22977 PR rtl-optimizatin/81308
22978 * tree-switch-conversion.c (cfg_altered): New file scoped static.
22979 (process_switch): If group_case_labels makes a change, then set
22981 (pass_convert_switch::execute): If a switch is converted, then
22982 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
22984 PR rtl-optimization/81308
22985 * recog.c (split_all_insns): Conditionally cleanup the CFG after
22988 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
22990 PR target/83663 - Revert r255946
22991 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
22992 generation for cases where splatting a value is not useful.
22993 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
22994 across a vec_duplicate and a paradoxical subreg forming a vector
22995 mode to a vec_concat.
22997 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22999 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
23000 -march=armv8.3-a variants.
23001 * config/arm/t-multilib: Likewise.
23002 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
23004 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
23006 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
23008 (cceq_ior_compare_complement): Give it a name so I can use it, and
23009 change boolean_or_operator predicate to boolean_operator so it can
23010 be used to generate a crand.
23011 (eqne): New code iterator.
23012 (bd/bd_neg): New code_attrs.
23013 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
23014 a single define_insn.
23015 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
23016 decrement (bdnzt/bdnzf/bdzt/bdzf).
23017 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
23018 with the new names of the branch decrement patterns, and added the
23019 names of the branch decrement conditional patterns.
23021 2018-01-08 Richard Biener <rguenther@suse.de>
23023 PR tree-optimization/83563
23024 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
23027 2018-01-08 Richard Biener <rguenther@suse.de>
23029 PR middle-end/83713
23030 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
23032 2018-01-08 Richard Biener <rguenther@suse.de>
23034 PR tree-optimization/83685
23035 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
23036 references to abnormals.
23038 2018-01-08 Richard Biener <rguenther@suse.de>
23041 * dwarf2out.c (output_indirect_strings): Handle empty
23042 skeleton_debug_str_hash.
23043 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
23045 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
23047 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
23048 (emit_store_direct): Likewise.
23049 (arc_trampoline_adjust_address): Likewise.
23050 (arc_asm_trampoline_template): New function.
23051 (arc_initialize_trampoline): Use asm_trampoline_template.
23052 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
23053 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
23054 * config/arc/arc.md (flush_icache): Delete pattern.
23056 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
23058 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
23059 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
23062 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
23065 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
23066 by not USED_FOR_TARGET.
23067 (make_pass_resolve_sw_modes): Likewise.
23069 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
23071 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
23074 2018-01-08 Richard Biener <rguenther@suse.de>
23076 PR middle-end/83580
23077 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
23079 2018-01-08 Richard Biener <rguenther@suse.de>
23081 PR middle-end/83517
23082 * match.pd ((t * 2) / 2) -> t): Add missing :c.
23084 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
23086 PR middle-end/81897
23087 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
23088 basic blocks with a small number of successors.
23089 (convert_control_dep_chain_into_preds): Improve handling of
23091 (dump_predicates): Split apart into...
23092 (dump_pred_chain): ...here...
23093 (dump_pred_info): ...and here.
23094 (can_one_predicate_be_invalidated_p): Add debugging printfs.
23095 (can_chain_union_be_invalidated_p): Improve check for invalidation
23097 (uninit_uses_cannot_happen): Avoid unnecessary if
23098 convert_control_dep_chain_into_preds yielded nothing.
23100 2018-01-06 Martin Sebor <msebor@redhat.com>
23102 PR tree-optimization/83640
23103 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
23104 subtracting negative offset from size.
23105 (builtin_access::overlap): Adjust offset bounds of the access to fall
23106 within the size of the object if possible.
23108 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
23110 PR rtl-optimization/83699
23111 * expmed.c (extract_bit_field_1): Restrict the vector usage of
23112 extract_bit_field_as_subreg to cases in which the extracted
23113 value is also a vector.
23115 * lra-constraints.c (process_alt_operands): Test for the equivalence
23116 substitutions when detecting a possible reload cycle.
23118 2018-01-06 Jakub Jelinek <jakub@redhat.com>
23121 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
23122 by default if flag_selective_schedling{,2}. Formatting fixes.
23124 PR rtl-optimization/83682
23125 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
23126 if it has non-VECTOR_MODE element mode.
23127 (vec_duplicate_p): Likewise.
23129 PR middle-end/83694
23130 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
23131 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
23133 2018-01-05 Jakub Jelinek <jakub@redhat.com>
23136 * config/i386/i386-builtin.def
23137 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
23138 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
23139 Require also OPTION_MASK_ISA_AVX512F in addition to
23140 OPTION_MASK_ISA_GFNI.
23141 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
23142 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
23143 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
23144 to OPTION_MASK_ISA_GFNI.
23145 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
23146 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
23147 OPTION_MASK_ISA_AVX512BW.
23148 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
23149 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
23150 addition to OPTION_MASK_ISA_GFNI.
23151 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
23152 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
23153 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
23154 to OPTION_MASK_ISA_GFNI.
23155 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
23156 a requirement for all ISAs rather than any of them with a few
23158 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
23160 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
23161 bitmasks to be enabled with 3 exceptions, instead of requiring any
23162 enabled ISA with lots of exceptions.
23163 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
23164 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
23165 Change avx512bw in isa attribute to avx512f.
23166 * config/i386/sgxintrin.h: Add license boilerplate.
23167 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
23168 to __AVX512F__ and __AVX512VL to __AVX512VL__.
23169 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
23170 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
23172 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
23173 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
23174 temporarily sse2 rather than sse if not enabled already.
23177 * config/i386/sse.md (VI248_VLBW): Rename to ...
23178 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
23179 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
23180 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
23181 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
23182 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
23183 mode iterator instead of VI248_VLBW.
23185 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
23187 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
23188 (record_modified): Skip clobbers; add debug output.
23189 (param_change_prob): Use sreal frequencies.
23191 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
23193 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
23194 punt for user-aligned variables.
23196 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
23198 * tree-chrec.c (chrec_contains_symbols): Return true for
23201 2018-01-05 Sudakshina Das <sudi.das@arm.com>
23204 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
23205 of (x|y) == x for BICS pattern.
23207 2018-01-05 Jakub Jelinek <jakub@redhat.com>
23209 PR tree-optimization/83605
23210 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
23211 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
23214 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
23216 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
23217 * config/epiphany/rtems.h: New file.
23219 2018-01-04 Jakub Jelinek <jakub@redhat.com>
23220 Uros Bizjak <ubizjak@gmail.com>
23223 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
23224 QIreg_operand instead of register_operand predicate.
23225 * config/i386/i386.c (ix86_rop_should_change_byte_p,
23226 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
23227 comments instead of -fmitigate[-_]rop.
23229 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
23232 * cgraphunit.c (symbol_table::compile): Switch to text_section
23233 before calling assembly_start debug hook.
23234 * run-rtl-passes.c (run_rtl_passes): Likewise.
23237 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
23239 * tree-vrp.c (extract_range_from_binary_expr_1): Check
23240 range_int_cst_p rather than !symbolic_range_p before calling
23241 extract_range_from_multiplicative_op_1.
23243 2018-01-04 Jeff Law <law@redhat.com>
23245 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
23246 redundant test in assertion.
23248 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
23250 * doc/rtl.texi: Document machine_mode wrapper classes.
23252 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
23254 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
23255 using tree_to_uhwi.
23257 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
23259 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
23260 the VEC_PERM_EXPR fold to fail.
23262 2018-01-04 Jakub Jelinek <jakub@redhat.com>
23265 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
23266 to switched_sections.
23268 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
23271 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
23272 test for d.testing.
23274 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
23277 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
23278 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
23280 2018-01-04 Jakub Jelinek <jakub@redhat.com>
23283 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
23284 is BLKmode and bitpos not zero or mode change is needed.
23286 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
23289 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
23292 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
23295 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
23296 instead of MULT rtx. Update all corresponding splitters.
23297 (*saddl_se): Ditto.
23298 (*ssub<modesuffix>): Ditto.
23299 (*ssubl_se): Ditto.
23300 (*cmp_sadd_di): Update split patterns.
23301 (*cmp_sadd_si): Ditto.
23302 (*cmp_sadd_sidi): Ditto.
23303 (*cmp_ssub_di): Ditto.
23304 (*cmp_ssub_si): Ditto.
23305 (*cmp_ssub_sidi): Ditto.
23306 * config/alpha/predicates.md (const23_operand): New predicate.
23307 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
23308 Look for ASHIFT, not MULT inner operand.
23309 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
23311 2018-01-04 Martin Liska <mliska@suse.cz>
23313 PR gcov-profile/83669
23314 * gcov.c (output_intermediate_file): Add version to intermediate
23316 * doc/gcov.texi: Document new field 'version' in intermediate
23317 file format. Fix location of '-k' option of gcov command.
23319 2018-01-04 Martin Liska <mliska@suse.cz>
23322 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
23324 2018-01-04 Jakub Jelinek <jakub@redhat.com>
23326 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
23328 2018-01-03 Martin Sebor <msebor@redhat.com>
23330 PR tree-optimization/83655
23331 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
23332 checking calls with invalid arguments.
23334 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23336 * tree-vect-stmts.c (vect_get_store_rhs): New function.
23337 (vectorizable_mask_load_store): Delete.
23338 (vectorizable_call): Return false for masked loads and stores.
23339 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
23340 instead of gimple_assign_rhs1.
23341 (vectorizable_load): Handle IFN_MASK_LOAD.
23342 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
23344 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23346 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
23348 (vectorizable_mask_load_store): ...here.
23349 (vectorizable_load): ...and here.
23351 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23353 * tree-vect-stmts.c (vect_build_all_ones_mask)
23354 (vect_build_zero_merge_argument): New functions, split out from...
23355 (vectorizable_load): ...here.
23357 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23359 * tree-vect-stmts.c (vect_check_store_rhs): New function,
23361 (vectorizable_mask_load_store): ...here.
23362 (vectorizable_store): ...and here.
23364 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23366 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
23368 (vectorizable_mask_load_store): ...here.
23370 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23372 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
23373 (vect_model_store_cost): Take a vec_load_store_type instead of a
23375 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
23376 (vect_model_store_cost): Take a vec_load_store_type instead of a
23378 (vectorizable_mask_load_store): Update accordingly.
23379 (vectorizable_store): Likewise.
23380 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
23382 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23384 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
23385 IFN_MASK_LOAD calls here rather than...
23386 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
23388 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23389 Alan Hayward <alan.hayward@arm.com>
23390 David Sherwood <david.sherwood@arm.com>
23392 * expmed.c (extract_bit_field_1): For vector extracts,
23393 fall back to extract_bit_field_as_subreg if vec_extract
23396 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23397 Alan Hayward <alan.hayward@arm.com>
23398 David Sherwood <david.sherwood@arm.com>
23400 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
23401 they are variable or constant sized.
23402 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
23403 slots for constant-sized data.
23405 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23406 Alan Hayward <alan.hayward@arm.com>
23407 David Sherwood <david.sherwood@arm.com>
23409 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
23410 handling COND_EXPRs with boolean comparisons, try to find a better
23411 basis for the mask type than the boolean itself.
23413 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23415 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
23416 is calculated and how it can be overridden.
23417 * genmodes.c (max_bitsize_mode_any_mode): New variable.
23418 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
23420 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
23423 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23424 Alan Hayward <alan.hayward@arm.com>
23425 David Sherwood <david.sherwood@arm.com>
23427 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
23428 Remove the mode argument.
23429 (aarch64_simd_valid_immediate): Remove the mode and inverse
23431 * config/aarch64/iterators.md (bitsize): New iterator.
23432 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
23433 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
23434 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
23435 aarch64_simd_valid_immediate.
23436 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
23437 (aarch64_reg_or_bic_imm): Likewise.
23438 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
23439 with an insn_type enum and msl with a modifier_type enum.
23440 Replace element_width with a scalar_mode. Change the shift
23441 to unsigned int. Add constructors for scalar_float_mode and
23442 scalar_int_mode elements.
23443 (aarch64_vect_float_const_representable_p): Delete.
23444 (aarch64_can_const_movi_rtx_p)
23445 (aarch64_simd_scalar_immediate_valid_for_move)
23446 (aarch64_simd_make_constant): Update call to
23447 aarch64_simd_valid_immediate.
23448 (aarch64_advsimd_valid_immediate_hs): New function.
23449 (aarch64_advsimd_valid_immediate): Likewise.
23450 (aarch64_simd_valid_immediate): Remove mode and inverse
23451 arguments. Rewrite to use the above. Use const_vec_duplicate_p
23452 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
23453 and aarch64_float_const_representable_p on the result.
23454 (aarch64_output_simd_mov_immediate): Remove mode argument.
23455 Update call to aarch64_simd_valid_immediate and use of
23456 simd_immediate_info.
23457 (aarch64_output_scalar_simd_mov_immediate): Update call
23460 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23461 Alan Hayward <alan.hayward@arm.com>
23462 David Sherwood <david.sherwood@arm.com>
23464 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
23465 (mode_nunits): Likewise CONST_MODE_NUNITS.
23466 * machmode.def (ADJUST_NUNITS): Document.
23467 * genmodes.c (mode_data::need_nunits_adj): New field.
23468 (blank_mode): Update accordingly.
23469 (adj_nunits): New variable.
23470 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
23472 (emit_mode_size_inline): Set need_bytesize_adj for all modes
23473 listed in adj_nunits.
23474 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
23475 listed in adj_nunits. Don't emit case statements for such modes.
23476 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
23477 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
23478 nothing if adj_nunits is nonnull.
23479 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
23480 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
23481 (emit_mode_fbit): Update use of print_maybe_const_decl.
23482 (emit_move_size): Likewise. Treat the array as non-const
23484 (emit_mode_adjustments): Handle adj_nunits.
23486 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23488 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
23489 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
23490 (VECTOR_MODES): Use it.
23491 (make_vector_modes): Take the prefix as an argument.
23493 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23494 Alan Hayward <alan.hayward@arm.com>
23495 David Sherwood <david.sherwood@arm.com>
23497 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
23498 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
23499 for MODE_VECTOR_BOOL.
23500 * machmode.def (VECTOR_BOOL_MODE): Document.
23501 * genmodes.c (VECTOR_BOOL_MODE): New macro.
23502 (make_vector_bool_mode): New function.
23503 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
23505 * lto-streamer-in.c (lto_input_mode_table): Likewise.
23506 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
23508 * stor-layout.c (int_mode_for_mode): Likewise.
23509 * tree.c (build_vector_type_for_mode): Likewise.
23510 * varasm.c (output_constant_pool_2): Likewise.
23511 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
23512 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
23513 for MODE_VECTOR_BOOL.
23514 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
23515 of mode class checks.
23516 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
23517 instead of a list of mode class checks.
23518 (expand_vector_scalar_condition): Likewise.
23519 (type_for_widest_vector_mode): Handle BImode as an inner mode.
23521 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23522 Alan Hayward <alan.hayward@arm.com>
23523 David Sherwood <david.sherwood@arm.com>
23525 * machmode.h (mode_size): Change from unsigned short to
23527 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
23528 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
23529 or if measurement_type is not polynomial.
23530 (fixed_size_mode::includes_p): Check for constant-sized modes.
23531 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
23532 return a poly_uint16 rather than an unsigned short.
23533 (emit_mode_size): Change the type of mode_size from unsigned short
23534 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
23535 (emit_mode_adjustments): Cope with polynomial vector sizes.
23536 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
23538 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
23540 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
23541 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
23542 * caller-save.c (setup_save_areas): Likewise.
23543 (replace_reg_with_saved_mem): Likewise.
23544 * calls.c (emit_library_call_value_1): Likewise.
23545 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
23546 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
23547 (gen_lowpart_for_combine): Likewise.
23548 * convert.c (convert_to_integer_1): Likewise.
23549 * cse.c (equiv_constant, cse_insn): Likewise.
23550 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
23551 (cselib_subst_to_values): Likewise.
23552 * dce.c (word_dce_process_block): Likewise.
23553 * df-problems.c (df_word_lr_mark_ref): Likewise.
23554 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
23555 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
23556 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
23557 (rtl_for_decl_location): Likewise.
23558 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
23559 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
23560 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
23561 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
23562 (expand_expr_real_1): Likewise.
23563 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
23564 (pad_below): Likewise.
23565 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
23566 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
23567 * ira.c (get_subreg_tracking_sizes): Likewise.
23568 * ira-build.c (ira_create_allocno_objects): Likewise.
23569 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
23570 (ira_sort_regnos_for_alter_reg): Likewise.
23571 * ira-costs.c (record_operand_costs): Likewise.
23572 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
23573 (resolve_simple_move): Likewise.
23574 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
23575 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
23576 (lra_constraints): Likewise.
23577 (CONST_POOL_OK_P): Reject variable-sized modes.
23578 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
23579 (add_pseudo_to_slot, lra_spill): Likewise.
23580 * omp-low.c (omp_clause_aligned_alignment): Likewise.
23581 * optabs-query.c (get_best_extraction_insn): Likewise.
23582 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
23583 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
23584 (expand_mult_highpart, valid_multiword_target_p): Likewise.
23585 * recog.c (offsettable_address_addr_space_p): Likewise.
23586 * regcprop.c (maybe_mode_change): Likewise.
23587 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
23588 * regrename.c (build_def_use): Likewise.
23589 * regstat.c (dump_reg_info): Likewise.
23590 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
23591 (find_reloads, find_reloads_subreg_address): Likewise.
23592 * reload1.c (eliminate_regs_1): Likewise.
23593 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
23594 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
23595 (simplify_binary_operation_1, simplify_subreg): Likewise.
23596 * targhooks.c (default_function_arg_padding): Likewise.
23597 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
23598 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
23599 (verify_gimple_assign_ternary): Likewise.
23600 * tree-inline.c (estimate_move_cost): Likewise.
23601 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
23602 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
23603 (get_address_cost_ainc): Likewise.
23604 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
23605 (vect_supportable_dr_alignment): Likewise.
23606 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
23607 (vectorizable_reduction): Likewise.
23608 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
23609 (vectorizable_operation, vectorizable_load): Likewise.
23610 * tree.c (build_same_sized_truth_vector_type): Likewise.
23611 * valtrack.c (cleanup_auto_inc_dec): Likewise.
23612 * var-tracking.c (emit_note_insn_var_location): Likewise.
23613 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
23614 (ADDR_VEC_ALIGN): Likewise.
23616 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23617 Alan Hayward <alan.hayward@arm.com>
23618 David Sherwood <david.sherwood@arm.com>
23620 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
23622 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
23623 or if measurement_type is polynomial.
23624 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
23625 * combine.c (make_extraction): Likewise.
23626 * dse.c (find_shift_sequence): Likewise.
23627 * dwarf2out.c (mem_loc_descriptor): Likewise.
23628 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
23629 (extract_bit_field, extract_low_bits): Likewise.
23630 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
23631 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
23632 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
23633 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
23634 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
23635 * reload.c (find_reloads): Likewise.
23636 * reload1.c (alter_reg): Likewise.
23637 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
23638 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
23639 * tree-if-conv.c (predicate_mem_writes): Likewise.
23640 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
23641 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
23642 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
23643 * valtrack.c (dead_debug_insert_temp): Likewise.
23644 * varasm.c (mergeable_constant_section): Likewise.
23645 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
23647 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23648 Alan Hayward <alan.hayward@arm.com>
23649 David Sherwood <david.sherwood@arm.com>
23651 * expr.c (expand_assignment): Cope with polynomial mode sizes
23652 when assigning to a CONCAT.
23654 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23655 Alan Hayward <alan.hayward@arm.com>
23656 David Sherwood <david.sherwood@arm.com>
23658 * machmode.h (mode_precision): Change from unsigned short to
23660 (mode_to_precision): Return a poly_uint16 rather than an unsigned
23662 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
23663 or if measurement_type is not polynomial.
23664 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
23665 in which the mode is already known to be a scalar_int_mode.
23666 * genmodes.c (emit_mode_precision): Change the type of mode_precision
23667 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
23669 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
23670 for GET_MODE_PRECISION.
23671 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
23672 for GET_MODE_PRECISION.
23673 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
23675 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
23676 (expand_field_assignment, make_extraction): Likewise.
23677 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
23678 (get_last_value): Likewise.
23679 * convert.c (convert_to_integer_1): Likewise.
23680 * cse.c (cse_insn): Likewise.
23681 * expr.c (expand_expr_real_1): Likewise.
23682 * lra-constraints.c (simplify_operand_subreg): Likewise.
23683 * optabs-query.c (can_atomic_load_p): Likewise.
23684 * optabs.c (expand_atomic_load): Likewise.
23685 (expand_atomic_store): Likewise.
23686 * ree.c (combine_reaching_defs): Likewise.
23687 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
23688 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
23689 * tree.h (type_has_mode_precision_p): Likewise.
23690 * ubsan.c (instrument_si_overflow): Likewise.
23692 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23693 Alan Hayward <alan.hayward@arm.com>
23694 David Sherwood <david.sherwood@arm.com>
23696 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
23697 polynomial numbers of units.
23698 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
23699 (valid_vector_subparts_p): New function.
23700 (build_vector_type): Remove temporary shim and take the number
23701 of units as a poly_uint64 rather than an int.
23702 (build_opaque_vector_type): Take the number of units as a
23703 poly_uint64 rather than an int.
23704 * tree.c (build_vector_from_ctor): Handle polynomial
23705 TYPE_VECTOR_SUBPARTS.
23706 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
23707 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
23708 (build_vector_from_val): If the number of units is variable,
23709 use build_vec_duplicate_cst for constant operands and
23710 VEC_DUPLICATE_EXPR otherwise.
23711 (make_vector_type): Remove temporary is_constant ().
23712 (build_vector_type, build_opaque_vector_type): Take the number of
23713 units as a poly_uint64 rather than an int.
23714 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
23716 * cfgexpand.c (expand_debug_expr): Likewise.
23717 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
23718 (store_constructor, expand_expr_real_1): Likewise.
23719 (const_scalar_mask_from_tree): Likewise.
23720 * fold-const-call.c (fold_const_reduction): Likewise.
23721 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
23722 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
23723 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
23724 (fold_relational_const): Likewise.
23725 (native_interpret_vector): Likewise. Change the size from an
23726 int to an unsigned int.
23727 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
23728 TYPE_VECTOR_SUBPARTS.
23729 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
23730 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
23731 duplicating a non-constant operand into a variable-length vector.
23732 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
23733 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
23734 * ipa-icf.c (sem_variable::equals): Likewise.
23735 * match.pd: Likewise.
23736 * omp-simd-clone.c (simd_clone_subparts): Likewise.
23737 * print-tree.c (print_node): Likewise.
23738 * stor-layout.c (layout_type): Likewise.
23739 * targhooks.c (default_builtin_vectorization_cost): Likewise.
23740 * tree-cfg.c (verify_gimple_comparison): Likewise.
23741 (verify_gimple_assign_binary): Likewise.
23742 (verify_gimple_assign_ternary): Likewise.
23743 (verify_gimple_assign_single): Likewise.
23744 * tree-pretty-print.c (dump_generic_node): Likewise.
23745 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
23746 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
23747 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
23748 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
23749 (vect_shift_permute_load_chain): Likewise.
23750 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
23751 (expand_vector_condition, optimize_vector_constructor): Likewise.
23752 (lower_vec_perm, get_compute_type): Likewise.
23753 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
23754 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
23755 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
23756 (vect_recog_mask_conversion_pattern): Likewise.
23757 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
23758 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
23759 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
23760 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
23761 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
23762 (vectorizable_shift, vectorizable_operation, vectorizable_store)
23763 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
23764 (supportable_widening_operation): Likewise.
23765 (supportable_narrowing_operation): Likewise.
23766 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
23768 * varasm.c (output_constant): Likewise.
23770 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23771 Alan Hayward <alan.hayward@arm.com>
23772 David Sherwood <david.sherwood@arm.com>
23774 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
23775 so that both the length == 3 and length != 3 cases set up their
23776 own permute vectors. Add comments explaining why we know the
23777 number of elements is constant.
23778 (vect_permute_load_chain): Likewise.
23780 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23781 Alan Hayward <alan.hayward@arm.com>
23782 David Sherwood <david.sherwood@arm.com>
23784 * machmode.h (mode_nunits): Change from unsigned char to
23786 (ONLY_FIXED_SIZE_MODES): New macro.
23787 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
23788 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
23789 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
23791 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
23792 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
23793 or if measurement_type is not polynomial.
23794 * genmodes.c (ZERO_COEFFS): New macro.
23795 (emit_mode_nunits_inline): Make mode_nunits_inline return a
23797 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
23798 Use ZERO_COEFFS when emitting initializers.
23799 * data-streamer.h (bp_pack_poly_value): New function.
23800 (bp_unpack_poly_value): Likewise.
23801 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
23802 for GET_MODE_NUNITS.
23803 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
23804 for GET_MODE_NUNITS.
23805 * tree.c (make_vector_type): Remove temporary shim and make
23806 the real function take the number of units as a poly_uint64
23807 rather than an int.
23808 (build_vector_type_for_mode): Handle polynomial nunits.
23809 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
23810 * emit-rtl.c (const_vec_series_p_1): Likewise.
23811 (gen_rtx_CONST_VECTOR): Likewise.
23812 * fold-const.c (test_vec_duplicate_folding): Likewise.
23813 * genrecog.c (validate_pattern): Likewise.
23814 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
23815 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
23816 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
23817 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
23818 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
23819 * rtlanal.c (subreg_get_info): Likewise.
23820 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
23821 (vect_grouped_load_supported): Likewise.
23822 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
23823 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
23824 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
23825 (simplify_const_unary_operation, simplify_binary_operation_1)
23826 (simplify_const_binary_operation, simplify_ternary_operation)
23827 (test_vector_ops_duplicate, test_vector_ops): Likewise.
23828 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
23829 instead of CONST_VECTOR_NUNITS.
23830 * varasm.c (output_constant_pool_2): Likewise.
23831 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
23832 explicit-encoded elements in the XVEC for variable-length vectors.
23834 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23836 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
23838 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23839 Alan Hayward <alan.hayward@arm.com>
23840 David Sherwood <david.sherwood@arm.com>
23842 * coretypes.h (fixed_size_mode): Declare.
23843 (fixed_size_mode_pod): New typedef.
23844 * builtins.h (target_builtins::x_apply_args_mode)
23845 (target_builtins::x_apply_result_mode): Change type to
23846 fixed_size_mode_pod.
23847 * builtins.c (apply_args_size, apply_result_size, result_vector)
23848 (expand_builtin_apply_args_1, expand_builtin_apply)
23849 (expand_builtin_return): Update accordingly.
23851 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23853 * cse.c (hash_rtx_cb): Hash only the encoded elements.
23854 * cselib.c (cselib_hash_rtx): Likewise.
23855 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
23856 CONST_VECTOR encoding.
23858 2018-01-03 Jakub Jelinek <jakub@redhat.com>
23859 Jeff Law <law@redhat.com>
23862 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
23863 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
23864 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
23865 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
23868 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
23869 explicitly probe *sp in a noreturn function if there were any callee
23870 register saves or frame pointer is needed.
23872 2018-01-03 Jakub Jelinek <jakub@redhat.com>
23875 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
23876 BLKmode for ternary, binary or unary expressions.
23879 * var-tracking.c (delete_vta_debug_insn): New inline function.
23880 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
23881 insns from get_insns () to NULL instead of each bb separately.
23882 Use delete_vta_debug_insn. No longer static.
23883 (vt_debug_insns_local, variable_tracking_main_1): Adjust
23884 delete_vta_debug_insns callers.
23885 * rtl.h (delete_vta_debug_insns): Declare.
23886 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
23887 instead of variable_tracking_main.
23889 2018-01-03 Martin Sebor <msebor@redhat.com>
23891 PR tree-optimization/83603
23892 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
23893 arguments past the endof the argument list in functions declared
23894 without a prototype.
23895 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
23896 Avoid checking when arguments are null.
23898 2018-01-03 Martin Sebor <msebor@redhat.com>
23901 * doc/extend.texi (attribute const): Fix a typo.
23902 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
23903 issuing -Wsuggest-attribute for void functions.
23905 2018-01-03 Martin Sebor <msebor@redhat.com>
23907 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
23908 offset_int::from instead of wide_int::to_shwi.
23909 (maybe_diag_overlap): Remove assertion.
23910 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
23911 * gimple-ssa-sprintf.c (format_directive): Same.
23912 (parse_directive): Same.
23913 (sprintf_dom_walker::compute_format_length): Same.
23914 (try_substitute_return_value): Same.
23916 2018-01-03 Jeff Law <law@redhat.com>
23918 PR middle-end/83654
23919 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
23920 non-constant residual for zero at runtime and avoid probing in
23921 that case. Reorganize code for trailing problem to mirror handling
23924 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
23926 PR tree-optimization/83501
23927 * tree-ssa-strlen.c (get_string_cst): New.
23928 (handle_char_store): Call get_string_cst.
23930 2018-01-03 Martin Liska <mliska@suse.cz>
23932 PR tree-optimization/83593
23933 * tree-ssa-strlen.c: Include tree-cfg.h.
23934 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
23935 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
23936 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
23938 (strlen_dom_walker::before_dom_children): Call
23939 gimple_purge_dead_eh_edges. Dump tranformation with details
23941 (strlen_dom_walker::before_dom_children): Update call by adding
23942 new argument cleanup_eh.
23943 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
23945 2018-01-03 Martin Liska <mliska@suse.cz>
23948 * cif-code.def (VARIADIC_THUNK): New enum value.
23949 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
23952 2018-01-03 Jan Beulich <jbeulich@suse.com>
23954 * sse.md (mov<mode>_internal): Tighten condition for when to use
23955 vmovdqu<ssescalarsize> for TI and OI modes.
23957 2018-01-03 Jakub Jelinek <jakub@redhat.com>
23959 Update copyright years.
23961 2018-01-03 Martin Liska <mliska@suse.cz>
23964 * ipa-visibility.c (function_and_variable_visibility): Skip
23965 functions with noipa attribure.
23967 2018-01-03 Jakub Jelinek <jakub@redhat.com>
23969 * gcc.c (process_command): Update copyright notice dates.
23970 * gcov-dump.c (print_version): Ditto.
23971 * gcov.c (print_version): Ditto.
23972 * gcov-tool.c (print_version): Ditto.
23973 * gengtype.c (create_file): Ditto.
23974 * doc/cpp.texi: Bump @copying's copyright year.
23975 * doc/cppinternals.texi: Ditto.
23976 * doc/gcc.texi: Ditto.
23977 * doc/gccint.texi: Ditto.
23978 * doc/gcov.texi: Ditto.
23979 * doc/install.texi: Ditto.
23980 * doc/invoke.texi: Ditto.
23982 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23984 * vector-builder.h (vector_builder::m_full_nelts): Change from
23985 unsigned int to poly_uint64.
23986 (vector_builder::full_nelts): Update prototype accordingly.
23987 (vector_builder::new_vector): Likewise.
23988 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
23989 (vector_builder::operator ==): Likewise.
23990 (vector_builder::finalize): Likewise.
23991 * int-vector-builder.h (int_vector_builder::int_vector_builder):
23992 Take the number of elements as a poly_uint64 rather than an
23994 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
23995 from unsigned int to poly_uint64.
23996 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
23997 (vec_perm_indices::new_vector): Likewise.
23998 (vec_perm_indices::length): Likewise.
23999 (vec_perm_indices::nelts_per_input): Likewise.
24000 (vec_perm_indices::input_nelts): Likewise.
24001 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
24002 number of elements per input as a poly_uint64 rather than an
24003 unsigned int. Use the original encoding for variable-length
24004 vectors, rather than clamping each individual element.
24005 For the second and subsequent elements in each pattern,
24006 clamp the step and base before clamping their sum.
24007 (vec_perm_indices::series_p): Handle polynomial element counts.
24008 (vec_perm_indices::all_in_range_p): Likewise.
24009 (vec_perm_indices_to_tree): Likewise.
24010 (vec_perm_indices_to_rtx): Likewise.
24011 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
24012 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
24013 (tree_vector_builder::new_binary_operation): Handle polynomial
24014 element counts. Return false if we need to know the number
24015 of elements at compile time.
24016 * fold-const.c (fold_vec_perm): Punt if the number of elements
24017 isn't known at compile time.
24019 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24021 * vec-perm-indices.h (vec_perm_builder): Change element type
24022 from HOST_WIDE_INT to poly_int64.
24023 (vec_perm_indices::element_type): Update accordingly.
24024 (vec_perm_indices::clamp): Handle polynomial element_types.
24025 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
24026 (vec_perm_indices::all_in_range_p): Likewise.
24027 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
24029 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
24030 polynomial vec_perm_indices element types.
24031 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
24032 * fold-const.c (fold_vec_perm): Likewise.
24033 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
24034 * tree-vect-generic.c (lower_vec_perm): Likewise.
24035 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
24036 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
24037 element type to HOST_WIDE_INT.
24039 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24040 Alan Hayward <alan.hayward@arm.com>
24041 David Sherwood <david.sherwood@arm.com>
24043 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
24044 rather than an int. Use plus_constant.
24045 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
24046 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
24048 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24049 Alan Hayward <alan.hayward@arm.com>
24050 David Sherwood <david.sherwood@arm.com>
24052 * calls.c (emit_call_1, expand_call): Change struct_value_size from
24053 a HOST_WIDE_INT to a poly_int64.
24055 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24056 Alan Hayward <alan.hayward@arm.com>
24057 David Sherwood <david.sherwood@arm.com>
24059 * calls.c (load_register_parameters): Cope with polynomial
24060 mode sizes. Require a constant size for BLKmode parameters
24061 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
24062 forces a parameter to be padded at the lsb end in order to
24063 fill a complete number of words, require the parameter size
24064 to be ordered wrt UNITS_PER_WORD.
24066 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24067 Alan Hayward <alan.hayward@arm.com>
24068 David Sherwood <david.sherwood@arm.com>
24070 * reload1.c (spill_stack_slot_width): Change element type
24071 from unsigned int to poly_uint64_pod.
24072 (alter_reg): Treat mode sizes as polynomial.
24074 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24075 Alan Hayward <alan.hayward@arm.com>
24076 David Sherwood <david.sherwood@arm.com>
24078 * reload.c (complex_word_subreg_p): New function.
24079 (reload_inner_reg_of_subreg, push_reload): Use it.
24081 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24082 Alan Hayward <alan.hayward@arm.com>
24083 David Sherwood <david.sherwood@arm.com>
24085 * lra-constraints.c (process_alt_operands): Reject matched
24086 operands whose sizes aren't ordered.
24087 (match_reload): Refer to this check here.
24089 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24090 Alan Hayward <alan.hayward@arm.com>
24091 David Sherwood <david.sherwood@arm.com>
24093 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
24094 that the mode size is in the set {1, 2, 4, 8, 16}.
24096 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24097 Alan Hayward <alan.hayward@arm.com>
24098 David Sherwood <david.sherwood@arm.com>
24100 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
24101 Use plus_constant instead of gen_rtx_PLUS.
24103 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24104 Alan Hayward <alan.hayward@arm.com>
24105 David Sherwood <david.sherwood@arm.com>
24107 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
24108 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
24109 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
24110 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
24111 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
24112 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
24113 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
24114 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
24115 * config/i386/i386.c (ix86_push_rounding): ...this new function.
24116 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
24118 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
24119 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
24120 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
24121 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
24122 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
24123 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
24124 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
24125 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
24126 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
24127 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
24129 * expr.c (emit_move_resolve_push): Treat the input and result
24130 of PUSH_ROUNDING as a poly_int64.
24131 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
24132 (emit_push_insn): Likewise.
24133 * lra-eliminations.c (mark_not_eliminable): Likewise.
24134 * recog.c (push_operand): Likewise.
24135 * reload1.c (elimination_effects): Likewise.
24136 * rtlanal.c (nonzero_bits1): Likewise.
24137 * calls.c (store_one_arg): Likewise. Require the padding to be
24138 known at compile time.
24140 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24141 Alan Hayward <alan.hayward@arm.com>
24142 David Sherwood <david.sherwood@arm.com>
24144 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
24145 Use plus_constant instead of gen_rtx_PLUS.
24147 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24148 Alan Hayward <alan.hayward@arm.com>
24149 David Sherwood <david.sherwood@arm.com>
24151 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
24152 rather than an int.
24154 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24155 Alan Hayward <alan.hayward@arm.com>
24156 David Sherwood <david.sherwood@arm.com>
24158 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
24159 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
24160 via stack temporaries. Treat the mode size as polynomial too.
24162 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24163 Alan Hayward <alan.hayward@arm.com>
24164 David Sherwood <david.sherwood@arm.com>
24166 * expr.c (expand_expr_real_2): When handling conversions involving
24167 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
24168 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
24169 as a poly_uint64 too.
24171 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24172 Alan Hayward <alan.hayward@arm.com>
24173 David Sherwood <david.sherwood@arm.com>
24175 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
24177 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24178 Alan Hayward <alan.hayward@arm.com>
24179 David Sherwood <david.sherwood@arm.com>
24181 * combine.c (can_change_dest_mode): Handle polynomial
24182 REGMODE_NATURAL_SIZE.
24183 * expmed.c (store_bit_field_1): Likewise.
24184 * expr.c (store_constructor): Likewise.
24185 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
24186 and polynomial REGMODE_NATURAL_SIZE.
24187 (gen_lowpart_common): Likewise.
24188 * reginfo.c (record_subregs_of_mode): Likewise.
24189 * rtlanal.c (read_modify_subreg_p): Likewise.
24191 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24192 Alan Hayward <alan.hayward@arm.com>
24193 David Sherwood <david.sherwood@arm.com>
24195 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
24196 numbers of elements.
24198 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24199 Alan Hayward <alan.hayward@arm.com>
24200 David Sherwood <david.sherwood@arm.com>
24202 * match.pd: Cope with polynomial numbers of vector elements.
24204 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24205 Alan Hayward <alan.hayward@arm.com>
24206 David Sherwood <david.sherwood@arm.com>
24208 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
24209 in a POINTER_PLUS_EXPR.
24211 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24212 Alan Hayward <alan.hayward@arm.com>
24213 David Sherwood <david.sherwood@arm.com>
24215 * omp-simd-clone.c (simd_clone_subparts): New function.
24216 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
24217 (ipa_simd_modify_function_body): Likewise.
24219 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24220 Alan Hayward <alan.hayward@arm.com>
24221 David Sherwood <david.sherwood@arm.com>
24223 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
24224 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
24225 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
24226 (expand_vector_condition, vector_element): Likewise.
24227 (subparts_gt): New function.
24228 (get_compute_type): Use subparts_gt.
24229 (count_type_subparts): Delete.
24230 (expand_vector_operations_1): Use subparts_gt instead of
24231 count_type_subparts.
24233 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24234 Alan Hayward <alan.hayward@arm.com>
24235 David Sherwood <david.sherwood@arm.com>
24237 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
24238 (vect_compile_time_alias): ...this new function. Do the calculation
24239 on poly_ints rather than trees.
24240 (vect_prune_runtime_alias_test_list): Update call accordingly.
24242 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24243 Alan Hayward <alan.hayward@arm.com>
24244 David Sherwood <david.sherwood@arm.com>
24246 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
24248 (vect_schedule_slp_instance): Likewise.
24250 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24251 Alan Hayward <alan.hayward@arm.com>
24252 David Sherwood <david.sherwood@arm.com>
24254 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
24255 constant and extern definitions for variable-length vectors.
24256 (vect_get_constant_vectors): Note that the number of units
24257 is known to be constant.
24259 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24260 Alan Hayward <alan.hayward@arm.com>
24261 David Sherwood <david.sherwood@arm.com>
24263 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
24264 of units as polynomial. Choose between WIDE and NARROW based
24267 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24268 Alan Hayward <alan.hayward@arm.com>
24269 David Sherwood <david.sherwood@arm.com>
24271 * tree-vect-stmts.c (simd_clone_subparts): New function.
24272 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
24274 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24275 Alan Hayward <alan.hayward@arm.com>
24276 David Sherwood <david.sherwood@arm.com>
24278 * tree-vect-stmts.c (vectorizable_call): Treat the number of
24279 vectors as polynomial. Use build_index_vector for
24280 IFN_GOMP_SIMD_LANE.
24282 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24283 Alan Hayward <alan.hayward@arm.com>
24284 David Sherwood <david.sherwood@arm.com>
24286 * tree-vect-stmts.c (get_load_store_type): Treat the number of
24287 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
24288 for variable-length vectors.
24289 (vectorizable_mask_load_store): Treat the number of units as
24290 polynomial, asserting that it is constant if the condition has
24291 already been enforced.
24292 (vectorizable_store, vectorizable_load): Likewise.
24294 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24295 Alan Hayward <alan.hayward@arm.com>
24296 David Sherwood <david.sherwood@arm.com>
24298 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
24299 of units as polynomial. Punt if we can't tell at compile time
24300 which vector contains the final result.
24302 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24303 Alan Hayward <alan.hayward@arm.com>
24304 David Sherwood <david.sherwood@arm.com>
24306 * tree-vect-loop.c (vectorizable_induction): Treat the number
24307 of units as polynomial. Punt on SLP inductions. Use an integer
24308 VEC_SERIES_EXPR for variable-length integer reductions. Use a
24309 cast of such a series for variable-length floating-point
24312 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24313 Alan Hayward <alan.hayward@arm.com>
24314 David Sherwood <david.sherwood@arm.com>
24316 * tree.h (build_index_vector): Declare.
24317 * tree.c (build_index_vector): New function.
24318 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
24319 of units as polynomial, forcibly converting it to a constant if
24320 vectorizable_reduction has already enforced the condition.
24321 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
24322 to create a {1,2,3,...} vector.
24323 (vectorizable_reduction): Treat the number of units as polynomial.
24324 Choose vectype_in based on the largest scalar element size rather
24325 than the smallest number of units. Enforce the restrictions
24328 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24329 Alan Hayward <alan.hayward@arm.com>
24330 David Sherwood <david.sherwood@arm.com>
24332 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
24333 number of units as polynomial.
24335 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24336 Alan Hayward <alan.hayward@arm.com>
24337 David Sherwood <david.sherwood@arm.com>
24339 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
24340 * target.def (autovectorize_vector_sizes): Return the vector sizes
24341 by pointer, using vector_sizes rather than a bitmask.
24342 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
24343 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
24344 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
24346 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
24347 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
24348 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
24349 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
24350 * omp-general.c (omp_max_vf): Likewise.
24351 * omp-low.c (omp_clause_aligned_alignment): Likewise.
24352 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
24353 * tree-vect-loop.c (vect_analyze_loop): Likewise.
24354 * tree-vect-slp.c (vect_slp_bb): Likewise.
24355 * doc/tm.texi: Regenerate.
24356 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
24358 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
24359 the vector size as a poly_uint64 rather than an unsigned int.
24360 (current_vector_size): Change from an unsigned int to a poly_uint64.
24361 (get_vectype_for_scalar_type): Update accordingly.
24362 * tree.h (build_truth_vector_type): Take the size and number of
24363 units as a poly_uint64 rather than an unsigned int.
24364 (build_vector_type): Add a temporary overload that takes
24365 the number of units as a poly_uint64 rather than an unsigned int.
24366 * tree.c (make_vector_type): Likewise.
24367 (build_truth_vector_type): Take the number of units as a poly_uint64
24368 rather than an unsigned int.
24370 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24371 Alan Hayward <alan.hayward@arm.com>
24372 David Sherwood <david.sherwood@arm.com>
24374 * target.def (get_mask_mode): Take the number of units and length
24375 as poly_uint64s rather than unsigned ints.
24376 * targhooks.h (default_get_mask_mode): Update accordingly.
24377 * targhooks.c (default_get_mask_mode): Likewise.
24378 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
24379 * doc/tm.texi: Regenerate.
24381 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24382 Alan Hayward <alan.hayward@arm.com>
24383 David Sherwood <david.sherwood@arm.com>
24385 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
24386 * omp-general.c (omp_max_vf): Likewise.
24387 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
24388 (expand_omp_simd): Handle polynomial safelen.
24389 * omp-low.c (omplow_simd_context): Add a default constructor.
24390 (omplow_simd_context::max_vf): Change from int to poly_uint64.
24391 (lower_rec_simd_input_clauses): Update accordingly.
24392 (lower_rec_input_clauses): Likewise.
24394 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24395 Alan Hayward <alan.hayward@arm.com>
24396 David Sherwood <david.sherwood@arm.com>
24398 * tree-vectorizer.h (vect_nunits_for_cost): New function.
24399 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
24400 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
24401 (vect_analyze_slp_cost): Likewise.
24402 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
24403 (vect_model_load_cost): Likewise.
24405 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24406 Alan Hayward <alan.hayward@arm.com>
24407 David Sherwood <david.sherwood@arm.com>
24409 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
24410 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
24411 from an unsigned int * to a poly_uint64_pod *.
24412 (calculate_unrolling_factor): New function.
24413 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
24415 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24416 Alan Hayward <alan.hayward@arm.com>
24417 David Sherwood <david.sherwood@arm.com>
24419 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
24420 from an unsigned int to a poly_uint64.
24421 (_loop_vec_info::slp_unrolling_factor): Likewise.
24422 (_loop_vec_info::vectorization_factor): Change from an int
24424 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
24425 (vect_get_num_vectors): New function.
24426 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
24427 (vect_get_num_copies): Use vect_get_num_vectors.
24428 (vect_analyze_data_ref_dependences): Change max_vf from an int *
24429 to an unsigned int *.
24430 (vect_analyze_data_refs): Change min_vf from an int * to a
24432 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
24433 than an unsigned HOST_WIDE_INT.
24434 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
24435 (vect_analyze_data_ref_dependence): Change max_vf from an int *
24436 to an unsigned int *.
24437 (vect_analyze_data_ref_dependences): Likewise.
24438 (vect_compute_data_ref_alignment): Handle polynomial vf.
24439 (vect_enhance_data_refs_alignment): Likewise.
24440 (vect_prune_runtime_alias_test_list): Likewise.
24441 (vect_shift_permute_load_chain): Likewise.
24442 (vect_supportable_dr_alignment): Likewise.
24443 (dependence_distance_ge_vf): Take the vectorization factor as a
24444 poly_uint64 rather than an unsigned HOST_WIDE_INT.
24445 (vect_analyze_data_refs): Change min_vf from an int * to a
24447 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
24448 vfm1 as a poly_uint64 rather than an int. Make the same change
24449 for the returned bound_scalar.
24450 (vect_gen_vector_loop_niters): Handle polynomial vf.
24451 (vect_do_peeling): Likewise. Update call to
24452 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
24453 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
24455 * tree-vect-loop.c (vect_determine_vectorization_factor)
24456 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
24457 (vect_get_known_peeling_cost): Likewise.
24458 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
24459 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
24460 (vect_transform_loop): Likewise. Use the lowest possible VF when
24461 updating the upper bounds of the loop.
24462 (vect_min_worthwhile_factor): Make static. Return an unsigned int
24463 rather than an int.
24464 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
24465 polynomial unroll factors.
24466 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
24467 (vect_make_slp_decision): Likewise.
24468 (vect_supported_load_permutation_p): Likewise, and polynomial
24470 (vect_analyze_slp_cost): Handle polynomial vf.
24471 (vect_slp_analyze_node_operations): Likewise.
24472 (vect_slp_analyze_bb_1): Likewise.
24473 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
24474 than an unsigned HOST_WIDE_INT.
24475 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
24476 (vectorizable_load): Handle polynomial vf.
24477 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
24479 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
24481 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24482 Alan Hayward <alan.hayward@arm.com>
24483 David Sherwood <david.sherwood@arm.com>
24485 * match.pd: Handle bit operations involving three constants
24486 and try to fold one pair.
24488 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24490 * tree-vect-loop-manip.c: Include gimple-fold.h.
24491 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
24492 niters_maybe_zero parameters. Handle other cases besides a step of 1.
24493 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
24494 Add a path that uses a step of VF instead of 1, but disable it
24496 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
24497 and niters_no_overflow parameters. Update calls to
24498 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
24499 Create a new SSA name if the latter choses to use a ste other
24500 than zero, and return it via niters_vector_mult_vf_var.
24501 * tree-vect-loop.c (vect_transform_loop): Update calls to
24502 vect_do_peeling, vect_gen_vector_loop_niters and
24503 slpeel_make_loop_iterate_ntimes.
24504 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
24505 (vect_gen_vector_loop_niters): Update declarations after above changes.
24507 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
24509 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
24510 128-bit round to integer instructions.
24511 (ceil<mode>2): Likewise.
24512 (btrunc<mode>2): Likewise.
24513 (round<mode>2): Likewise.
24515 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
24517 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
24518 unaligned VSX load/store on P8/P9.
24519 (expand_block_clear): Allow the use of unaligned VSX
24520 load/store on P8/P9.
24522 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
24524 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
24526 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
24527 swap associated with both a load and a store.
24529 2018-01-02 Andrew Waterman <andrew@sifive.com>
24531 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
24532 * config/riscv/riscv.md (clear_cache): Use it.
24534 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
24536 * web.c: Remove out-of-date comment.
24538 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24540 * expr.c (fixup_args_size_notes): Check that any existing
24541 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
24542 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
24543 (emit_single_push_insn): ...here.
24545 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24547 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
24548 (const_vector_encoded_nelts): New function.
24549 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
24550 (const_vector_int_elt, const_vector_elt): Declare.
24551 * emit-rtl.c (const_vector_int_elt_1): New function.
24552 (const_vector_elt): Likewise.
24553 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
24554 of CONST_VECTOR_ELT.
24556 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24558 * expr.c: Include rtx-vector-builder.h.
24559 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
24560 directly on the tree encoding.
24561 (const_vector_from_tree): Likewise.
24562 * optabs.c: Include rtx-vector-builder.h.
24563 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
24564 sequence of "u" values.
24565 * vec-perm-indices.c: Include rtx-vector-builder.h.
24566 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
24567 directly on the vec_perm_indices encoding.
24569 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24571 * doc/rtl.texi (const_vector): Describe new encoding scheme.
24572 * Makefile.in (OBJS): Add rtx-vector-builder.o.
24573 * rtx-vector-builder.h: New file.
24574 * rtx-vector-builder.c: Likewise.
24575 * rtl.h (rtx_def::u2): Add a const_vector field.
24576 (CONST_VECTOR_NPATTERNS): New macro.
24577 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
24578 (CONST_VECTOR_DUPLICATE_P): Likewise.
24579 (CONST_VECTOR_STEPPED_P): Likewise.
24580 (CONST_VECTOR_ENCODED_ELT): Likewise.
24581 (const_vec_duplicate_p): Check for a duplicated vector encoding.
24582 (unwrap_const_vec_duplicate): Likewise.
24583 (const_vec_series_p): Check for a non-duplicated vector encoding.
24584 Say that the function only returns true for integer vectors.
24585 * emit-rtl.c: Include rtx-vector-builder.h.
24586 (gen_const_vec_duplicate_1): Delete.
24587 (gen_const_vector): Call gen_const_vec_duplicate instead of
24588 gen_const_vec_duplicate_1.
24589 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
24590 (gen_const_vec_duplicate): Use rtx_vector_builder.
24591 (gen_const_vec_series): Likewise.
24592 (gen_rtx_CONST_VECTOR): Likewise.
24593 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
24594 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
24595 Build a new vector rather than modifying a CONST_VECTOR in-place.
24596 (handle_special_swappables): Update call accordingly.
24597 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
24598 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
24599 Build a new vector rather than modifying a CONST_VECTOR in-place.
24600 (handle_special_swappables): Update call accordingly.
24602 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24604 * simplify-rtx.c (simplify_const_binary_operation): Use
24605 CONST_VECTOR_ELT instead of XVECEXP.
24607 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24609 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
24610 the selector elements to be different from the data elements
24611 if the selector is a VECTOR_CST.
24612 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
24613 ssizetype for the selector.
24615 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24617 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
24618 before testing each element individually.
24619 * tree-vect-generic.c (lower_vec_perm): Likewise.
24621 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24623 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
24624 * selftest-run-tests.c (selftest::run_tests): Call it.
24625 * vector-builder.h (vector_builder::operator ==): New function.
24626 (vector_builder::operator !=): Likewise.
24627 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
24628 (vec_perm_indices::all_from_input_p): New function.
24629 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
24630 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
24631 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
24632 instead of reading the VECTOR_CST directly. Detect whether both
24633 vector inputs are the same before constructing the vec_perm_indices,
24634 and update the number of inputs argument accordingly. Use the
24635 utility functions added above. Only construct sel2 if we need to.
24637 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24639 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
24640 the broadcast of the low byte.
24641 (expand_mult_highpart): Use an explicit encoding for the permutes.
24642 * optabs-query.c (can_mult_highpart_p): Likewise.
24643 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
24644 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
24645 (vectorizable_bswap): Likewise.
24646 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
24647 explicit encoding for the power-of-2 permutes.
24648 (vect_permute_store_chain): Likewise.
24649 (vect_grouped_load_supported): Likewise.
24650 (vect_permute_load_chain): Likewise.
24652 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24654 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
24655 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
24656 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
24657 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
24658 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
24659 (vect_gen_perm_mask_any): Likewise.
24661 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24663 * int-vector-builder.h: New file.
24664 * vec-perm-indices.h: Include int-vector-builder.h.
24665 (vec_perm_indices): Redefine as an int_vector_builder.
24666 (auto_vec_perm_indices): Delete.
24667 (vec_perm_builder): Redefine as a stand-alone class.
24668 (vec_perm_indices::vec_perm_indices): New function.
24669 (vec_perm_indices::clamp): Likewise.
24670 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
24671 (vec_perm_indices::new_vector): New function.
24672 (vec_perm_indices::new_expanded_vector): Update for new
24673 vec_perm_indices class.
24674 (vec_perm_indices::rotate_inputs): New function.
24675 (vec_perm_indices::all_in_range_p): Operate directly on the
24676 encoded form, without computing elided elements.
24677 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
24678 encoding. Update for new vec_perm_indices class.
24679 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
24680 the given vec_perm_builder.
24681 (expand_vec_perm_var): Update vec_perm_builder constructor.
24682 (expand_mult_highpart): Use vec_perm_builder instead of
24683 auto_vec_perm_indices.
24684 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
24685 vec_perm_indices instead of auto_vec_perm_indices. Use a single
24686 or double series encoding as appropriate.
24687 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
24688 vec_perm_indices instead of auto_vec_perm_indices.
24689 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
24690 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
24691 (vect_permute_store_chain): Likewise.
24692 (vect_grouped_load_supported): Likewise.
24693 (vect_permute_load_chain): Likewise.
24694 (vect_shift_permute_load_chain): Likewise.
24695 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
24696 (vect_transform_slp_perm_load): Likewise.
24697 (vect_schedule_slp_instance): Likewise.
24698 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
24699 (vectorizable_mask_load_store): Likewise.
24700 (vectorizable_bswap): Likewise.
24701 (vectorizable_store): Likewise.
24702 (vectorizable_load): Likewise.
24703 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
24704 vec_perm_indices instead of auto_vec_perm_indices. Use
24705 tree_to_vec_perm_builder to read the vector from a tree.
24706 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
24707 vec_perm_builder instead of a vec_perm_indices.
24708 (have_whole_vector_shift): Use vec_perm_builder and
24709 vec_perm_indices instead of auto_vec_perm_indices. Leave the
24710 truncation to calc_vec_perm_mask_for_shift.
24711 (vect_create_epilog_for_reduction): Likewise.
24712 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
24713 from auto_vec_perm_indices to vec_perm_indices.
24714 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
24715 instead of changing individual elements.
24716 (aarch64_vectorize_vec_perm_const): Use new_vector to install
24717 the vector in d.perm.
24718 * config/arm/arm.c (expand_vec_perm_d::perm): Change
24719 from auto_vec_perm_indices to vec_perm_indices.
24720 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
24721 instead of changing individual elements.
24722 (arm_vectorize_vec_perm_const): Use new_vector to install
24723 the vector in d.perm.
24724 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
24725 Update vec_perm_builder constructor.
24726 (rs6000_expand_interleave): Likewise.
24727 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
24728 (rs6000_expand_interleave): Likewise.
24730 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24732 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
24733 to qimode could truncate the indices.
24734 * optabs.c (expand_vec_perm_var): Likewise.
24736 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24738 * Makefile.in (OBJS): Add vec-perm-indices.o.
24739 * vec-perm-indices.h: New file.
24740 * vec-perm-indices.c: Likewise.
24741 * target.h (vec_perm_indices): Replace with a forward class
24743 (auto_vec_perm_indices): Move to vec-perm-indices.h.
24744 * optabs.h: Include vec-perm-indices.h.
24745 (expand_vec_perm): Delete.
24746 (selector_fits_mode_p, expand_vec_perm_var): Declare.
24747 (expand_vec_perm_const): Declare.
24748 * target.def (vec_perm_const_ok): Replace with...
24749 (vec_perm_const): ...this new hook.
24750 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
24751 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
24752 * doc/tm.texi: Regenerate.
24753 * optabs.def (vec_perm_const): Delete.
24754 * doc/md.texi (vec_perm_const): Likewise.
24755 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
24756 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
24757 expand_vec_perm for constant permutation vectors. Assert that
24758 the mode of variable permutation vectors is the integer equivalent
24759 of the mode that is being permuted.
24760 * optabs-query.h (selector_fits_mode_p): Declare.
24761 * optabs-query.c: Include vec-perm-indices.h.
24762 (selector_fits_mode_p): New function.
24763 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
24764 is defined, instead of checking whether the vec_perm_const_optab
24765 exists. Use targetm.vectorize.vec_perm_const instead of
24766 targetm.vectorize.vec_perm_const_ok. Check whether the indices
24767 fit in the vector mode before using a variable permute.
24768 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
24769 vec_perm_indices instead of an rtx.
24770 (expand_vec_perm): Replace with...
24771 (expand_vec_perm_const): ...this new function. Take the selector
24772 as a vec_perm_indices rather than an rtx. Also take the mode of
24773 the selector. Update call to shift_amt_for_vec_perm_mask.
24774 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
24775 Use vec_perm_indices::new_expanded_vector to expand the original
24776 selector into bytes. Check whether the indices fit in the vector
24777 mode before using a variable permute.
24778 (expand_vec_perm_var): Make global.
24779 (expand_mult_highpart): Use expand_vec_perm_const.
24780 * fold-const.c: Includes vec-perm-indices.h.
24781 * tree-ssa-forwprop.c: Likewise.
24782 * tree-vect-data-refs.c: Likewise.
24783 * tree-vect-generic.c: Likewise.
24784 * tree-vect-loop.c: Likewise.
24785 * tree-vect-slp.c: Likewise.
24786 * tree-vect-stmts.c: Likewise.
24787 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
24789 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
24790 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
24791 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
24792 (aarch64_vectorize_vec_perm_const): ...this new function.
24793 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
24794 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
24795 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
24796 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
24797 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
24798 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
24799 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
24801 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
24802 check for NEON modes.
24803 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
24804 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
24805 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
24806 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
24808 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
24809 the old VEC_PERM_CONST conditions.
24810 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
24811 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
24812 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
24813 (ia64_vectorize_vec_perm_const_ok): Merge into...
24814 (ia64_vectorize_vec_perm_const): ...this new function.
24815 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
24816 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
24817 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
24818 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
24819 * config/mips/mips.c (mips_expand_vec_perm_const)
24820 (mips_vectorize_vec_perm_const_ok): Merge into...
24821 (mips_vectorize_vec_perm_const): ...this new function.
24822 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
24823 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
24824 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
24825 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
24826 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
24827 (rs6000_expand_vec_perm_const): Delete.
24828 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
24830 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
24831 (altivec_expand_vec_perm_const_le): Take each operand individually.
24832 Operate on constant selectors rather than rtxes.
24833 (altivec_expand_vec_perm_const): Likewise. Update call to
24834 altivec_expand_vec_perm_const_le.
24835 (rs6000_expand_vec_perm_const): Delete.
24836 (rs6000_vectorize_vec_perm_const_ok): Delete.
24837 (rs6000_vectorize_vec_perm_const): New function.
24838 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
24839 an element count and rtx array.
24840 (rs6000_expand_extract_even): Update call accordingly.
24841 (rs6000_expand_interleave): Likewise.
24842 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
24843 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
24844 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
24845 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
24846 (rs6000_expand_vec_perm_const): Delete.
24847 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
24848 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
24849 (altivec_expand_vec_perm_const_le): Take each operand individually.
24850 Operate on constant selectors rather than rtxes.
24851 (altivec_expand_vec_perm_const): Likewise. Update call to
24852 altivec_expand_vec_perm_const_le.
24853 (rs6000_expand_vec_perm_const): Delete.
24854 (rs6000_vectorize_vec_perm_const_ok): Delete.
24855 (rs6000_vectorize_vec_perm_const): New function. Remove stray
24856 reference to the SPE evmerge intructions.
24857 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
24858 an element count and rtx array.
24859 (rs6000_expand_extract_even): Update call accordingly.
24860 (rs6000_expand_interleave): Likewise.
24861 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
24862 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
24864 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
24866 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24868 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
24869 vector mode and that that mode matches the mode of the data
24871 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
24872 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
24873 directly using expand_vec_perm_1 when forcing selectors into
24875 (expand_vec_perm_var): New function, split out from expand_vec_perm.
24877 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24879 * optabs-query.h (can_vec_perm_p): Delete.
24880 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
24881 * optabs-query.c (can_vec_perm_p): Split into...
24882 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
24883 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
24884 particular selector is valid.
24885 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
24886 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
24887 (vect_grouped_load_supported): Likewise.
24888 (vect_shift_permute_load_chain): Likewise.
24889 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
24890 (vect_transform_slp_perm_load): Likewise.
24891 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
24892 (vectorizable_bswap): Likewise.
24893 (vect_gen_perm_mask_checked): Likewise.
24894 * fold-const.c (fold_ternary_loc): Likewise. Don't take
24895 implementations of variable permutation vectors into account
24896 when deciding which selector to use.
24897 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
24898 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
24899 with a false third argument.
24900 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
24901 to test whether the constant selector is valid and can_vec_perm_var_p
24902 to test whether a variable selector is valid.
24904 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24906 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
24907 * optabs-query.c (can_vec_perm_p): Likewise.
24908 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
24909 instead of vec_perm_indices.
24910 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
24911 (vect_gen_perm_mask_checked): Likewise,
24912 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
24913 (vect_gen_perm_mask_checked): Likewise,
24915 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24917 * optabs-query.h (qimode_for_vec_perm): Declare.
24918 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
24919 (qimode_for_vec_perm): ...this new function.
24920 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
24922 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
24924 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
24925 does not have a conditional at the top.
24927 2018-01-02 Richard Biener <rguenther@suse.de>
24929 * ipa-inline.c (big_speedup_p): Fix expression.
24931 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
24934 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
24937 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
24941 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
24942 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
24943 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
24944 cond_taken_branch_cost 3->4.
24946 2018-01-01 Jakub Jelinek <jakub@redhat.com>
24948 PR tree-optimization/83581
24949 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
24950 TODO_cleanup_cfg if any changes have been made.
24952 PR middle-end/83608
24953 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
24954 convert_modes if target mode has the right side, but different mode
24957 PR middle-end/83609
24958 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
24959 last argument when extracting from CONCAT. If either from_real or
24960 from_imag is NULL, use expansion through memory. If result is not
24961 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
24962 the parts directly to inner mode, if even that fails, use expansion
24965 PR middle-end/83623
24966 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
24967 check for bswap in mode rather than HImode and use that in expand_unop
24970 Copyright (C) 2018 Free Software Foundation, Inc.
24972 Copying and distribution of this file, with or without modification,
24973 are permitted in any medium without royalty provided the copyright
24974 notice and this notice are preserved.