1 2018-09-05 Richard Biener <rguenther@suse.de>
4 * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
7 2018-09-05 Siddhesh Poyarekar <siddhesh@sourceware.org>
8 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
10 * config/aarch64/falkor-tag-collision-avoidance.c: New file.
11 * config.gcc (extra_objs): Build it.
12 * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
14 * config/aarch64/aarch64-passes.def
15 (pass_tag_collision_avoidance): New pass.
16 * config/aarch64/aarch64.c (qdf24xx_tunings): Add
17 AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
18 (aarch64_classify_address): Remove static qualifier.
19 (aarch64_address_info, aarch64_address_type): Move to...
20 * config/aarch64/aarch64-protos.h: ... here.
21 (make_pass_tag_collision_avoidance): New function.
22 * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
25 2018-09-05 Martin Liska <mliska@suse.cz>
27 * doc/gcov.texi: Update documentation of humar
29 * gcov.c (format_count): Print one decimal place, it provides
30 more fine number of situations like '1G' vs. '1.4G'.
32 2018-09-05 Martin Liska <mliska@suse.cz>
35 * config/rs6000/rs6000.opt: Mark the option as Deprecated.
36 * optc-gen.awk: Allow 'Var' for Deprecated options in order
37 to generate a MASK value.
39 2018-09-04 H.J. Lu <hongjiu.lu@intel.com>
42 * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
43 if frame pointer isn't used.
44 (compute_frame_pointer_to_fb_displacement): Likewise.
46 2018-09-04 Jakub Jelinek <jakub@redhat.com>
49 * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
50 OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
51 OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
52 (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
53 and OPTION_MASK_ISA_XSAVEC_UNSET.
55 2018-09-04 Max Filippov <jcmvbkbc@gmail.com>
57 * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
58 XOR operations in NAND case.
60 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
62 * wide-int-range.cc (wide_int_range_convert): New.
63 * wide-int-range.h (wide_int_range_convert): New.
64 * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
65 code into wide_int_range_convert.
66 (extract_range_into_wide_ints): Do not munge anti range constants
67 into the entire domain. Just return the range back.
69 2018-09-04 Martin Liska <mliska@suse.cz>
71 * genmatch.c (output_line_directive): Add new argument
73 (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
75 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
77 * doc/invoke.texi (Option Summary): Add whitespace.
79 * doc/invoke.texi (Option Summary): Add -Waligned-new.
81 2018-09-04 Richard Biener <rguenther@suse.de>
83 PR tree-optimization/87211
84 * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
85 backedge value we're supposed to treat as VARYING also number
86 the PHI to VARYING in case it got a different value-number already.
88 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
90 * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
91 (extract_range_from_binary_expr_1): Do not call
92 vrp_can_optimize_bit_op.
93 * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
95 (wide_int_range_get_mask_and_bounds): New.
96 (wide_int_range_optimize_bit_op): New.
97 (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
98 (wide_int_range_bit_and): Same.
99 * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
100 (wide_int_range_optimize_bit_op): New.
101 (wide_int_range_get_mask_and_bounds): New.
103 2018-09-04 Richard Biener <rguenther@suse.de>
105 PR tree-optimization/87176
106 * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
107 variable. When value-numbering a virtual PHI node make sure
108 to not value-number to the backedge value.
110 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
112 * doc/extend.texi (Long Long, Hex Floats): Document support for
113 long long and hex floats in more recent versions of ISO C++.
115 2018-09-03 Richard Biener <rguenther@suse.de>
117 PR tree-optimization/87177
118 * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
121 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
123 * bb-reorder.c (edge_order): Convert to C-qsort-style
124 tri-state comparator.
125 (reorder_basic_blocks_simple): Change std::stable_sort to gcc_stablesort.
127 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
129 * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
130 tri-state comparator.
131 (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
133 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
135 * sort.cc (struct sort_ctx): New field 'nlim'. Use it...
136 (mergesort): ... here as maximum count for using netsort.
137 (gcc_qsort): Set nlim to 3 if stable sort is requested.
138 (gcc_stablesort): New.
139 * system.h (gcc_stablesort): Declare.
141 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
143 * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
144 * system.h (qsort): Always redirect to gcc_qsort. Update comment.
145 * vec.c (qsort_chk): Do not call gcc_qsort. Update comment.
147 2018-09-03 Segher Boessenkool <segher@kernel.crashing.org>
149 * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
150 lxsdx and stxsdx alternatives.
151 (*mov<mode>_hardfloat64): Ditto.
152 * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
154 2018-09-03 Richard Biener <rguenther@suse.de>
156 PR tree-optimization/87200
157 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
160 2018-09-03 Martin Liska <mliska@suse.cz>
162 PR tree-optimization/87201
163 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
164 Fix parenthesis in an expression.
166 2018-09-03 Richard Biener <rguenther@suse.de>
168 PR tree-optimization/87197
169 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
170 visited. CSE the VN_INFO hashtable lookup.
172 PR tree-optimization/87169
173 * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
174 iterating make sure there's no extra backedges from irreducible
175 regions feeding the header. Mark the destination block
178 2018-09-03 Martin Liska <mliska@suse.cz>
181 * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
182 * common/common-targhooks.c (default_get_valid_option_values):
184 * common/common-targhooks.h (default_get_valid_option_values):
186 * common/config/i386/i386-common.c: Move processor_target_table
188 (ix86_get_valid_option_values): New function.
189 (TARGET_GET_VALID_OPTION_VALUES): New macro.
190 * config/i386/i386.c (struct ptt): Move to i386-common.c.
191 (PTA_*): Move all defined masks into i386-common.c.
192 (ix86_function_specific_restore): Use new processor_cost_table.
193 * config/i386/i386.h (struct ptt): Moved from i386.c.
194 (struct pta): Likewise.
195 * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
196 * doc/tm.texi.in: Likewise.
197 * opt-suggestions.c (option_proposer::suggest_option):
198 Pass prefix to build_option_suggestions.
199 (option_proposer::get_completions): Likewise.
200 (option_proposer::build_option_suggestions): Use the new target
202 * opts.c (struct option_help_tuple): New struct.
203 (print_filtered_help): Use the new target hook.
205 2018-09-03 Martin Liska <mliska@suse.cz>
208 * predict.c (set_even_probabilities): Add likely_edges
209 argument and handle cases where we have precisely one
211 (combine_predictions_for_bb): Catch also likely_edges.
212 (tree_predict_by_opcode): Handle gswitch statements.
213 * tree-cfg.h (find_case_label_for_value): New declaration.
214 (find_taken_edge_switch_expr): Likewise.
215 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
216 Find pivot in decision tree based on probabily, not by number of
219 2018-09-02 Gerald Pfeifer <gerald@pfeifer.com>
221 * doc/standards.texi (Standards): Update Objective-C reference.
223 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
225 * doc/install.texi (Prerequisites): Update link for MPC.
227 2018-09-01 Michael Matz <matz@suse.de>
229 PR tree-optimization/87074
230 * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
231 PHIs for outer-loop uses.
233 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
235 * doc/generic.texi (OpenMP): Adjust link to openmp.org.
236 * doc/invoke.texi (C Dialect Options): Ditto.
238 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
240 * doc/install.texi (Prerequisites): Adjust link mpfr.org.
242 2018-08-31 Richard Biener <rguenther@suse.de>
244 PR tree-optimization/87168
245 * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
246 (rpo_elim::eliminate_avail): When OP was not visited it must
249 2018-08-31 David Malcolm <dmalcolm@redhat.com>
251 * tree-vrp.c (copy_value_range): Convert param "from" from
252 "value_range *" to "const value_range *".
253 (range_is_null): Likewise for param "vr".
254 (range_int_cst_p): Likewise.
255 (range_int_cst_singleton_p): Likewise.
256 (symbolic_range_p): Likewise.
257 (value_ranges_intersect_p): Likewise for both params.
258 (value_range_nonnegative_p): Likewise for param "vr".
259 (value_range_constant_singleton): Likewise.
260 (vrp_set_zero_nonzero_bits): Likewise for param "ar".
261 (extract_range_into_wide_ints): Likewise for param "vr".
262 (extract_range_from_multiplicative_op): Likewise for params "vr0"
264 (vrp_can_optimize_bit_op): Likewise.
265 (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
267 (extract_range_from_unary_expr): Likewise.
268 (debug_value_range): Likewise for param "vr".
269 (value_range::dump): Add "const" qualifier.
270 (vrp_prop::check_array_ref): Convert local "vr" from
271 "value_range *" to "const value_range *".
272 (vrp_prop::check_mem_ref): Likewise.
273 (vrp_prop::visit_stmt): Likewise for local "old_vr".
274 (vrp_intersect_ranges_1): Likewise for param "vr_1".
275 (vrp_intersect_ranges): Likewise.
276 (simplify_stmt_for_jump_threading): Likewise for local "vr".
277 (vrp_prop::vrp_finalize): Likewise.
278 * tree-vrp.h (value_range::dump): Add "const" qualifier.
279 (vrp_intersect_ranges): Add "const" qualifier to params as above.
280 (extract_range_from_unary_expr): Likewise.
281 (value_range_constant_singleton): Likewise.
282 (symbolic_range_p): Likewise.
283 (copy_value_range): Likewise.
284 (extract_range_from_binary_expr_1): Likewise.
285 (range_int_cst_p): Likewise.
286 (vrp_set_zero_nonzero_bits): Likewise.
287 (range_int_cst_singleton_p): Likewise.
289 2018-08-31 Vlad Lazar <vlad.lazar@arm.com>
291 * config/aarch64/arm_neon.h (vabsd_s64): New.
292 (vnegd_s64): Likewise.
294 2018-08-31 Martin Jambor <mjambor@suse.cz>
296 * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
298 2018-08-31 Martin Liska <mliska@suse.cz>
300 * ipa-icf.c (sem_item::add_type): Use
301 sem_item::m_type_hash_cache.
302 * ipa-icf.h: Move the cache from sem_item_optimizer
305 2018-08-31 Nathan Sidwell <nathan@acm.org>
307 * doc/extend.texi (Backwards Compatibility): Remove implicit
308 extern C leeway of () being (...).
310 2018-08-31 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
312 * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
314 2018-08-31 Segher Boessenkool <segher@kernel.crashing.org>
318 * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
320 2018-08-31 Jakub Jelinek <jakub@redhat.com>
323 * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
324 gen_int_mode. Formatting fixes.
326 2018-08-30 Sandra Loosemore <sandra@codesourcery.com>
328 * target.def (custom_function_descriptors): Improve documentation.
329 * doc/tm.texi.in (Trampolines): Expand discussion of function
330 descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
331 beginning of the section.
332 * doc/tm.texi: Regenerated.
334 2018-08-30 Jose E. Marchesi <jose.marchesi@oracle.com>
336 * cfg.h (class auto_edge_flag): Spell out the template-id of the
337 base class in the initializer list. This is a workaround for
338 building with older GCC.
339 (class auto_bb_flag): Likewise.
341 2018-08-30 Aaron Sawdey <acsawdey@linux.ibm.com>
343 * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
344 (altivec_vcmpequ<VI_char>_p): Remove star.
345 * config/rs6000/rs6000-string.c (do_load_for_compare): Support
347 (expand_strncmp_vec_sequence): New function.
348 (emit_final_str_compare_vec): New function.
349 (expand_strn_compare): Add support for vector strncmp.
350 * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
351 length specification to bytes.
352 * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
353 (vcmpnezb_p): New pattern.
354 * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
355 for option -mstring-compare-inline-limit.
357 2018-08-30 Thiago Macieira <thiago.macieira@intel.com>
359 * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
360 (PTA_SKYLAKE): Add PTA_AES.
361 (PTA_GOLDMONT): Likewise.
363 2018-08-29 Jan Hubicka <jh@suse.cz>
366 * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
367 * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
369 2018-08-29 Jan Hubicka <jh@suse.cz>
371 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
373 (hash_tree): Do not visit TYPE_STUB_DECL.
374 * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
375 stream TYPE_STUB_DECL.
376 * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
377 * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
378 after free_lang_data.
379 (type_in_anonymous_namespace_p): Likewise.
381 2018-08-29 Jan Hubicka <jh@suse.cz>
383 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
384 comment that it has to be even number.
385 (class sreal): Change m_sig type to int32_t.
386 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
387 int64_t for temporary calculations.
388 (sreal_verify_basics): Drop one bit from minimum and maximum.
390 2018-08-30 Richard Biener <rguenther@suse.de>
392 PR tree-optimization/87147
393 * tree-ssa-sccvn.c (SSA_VISITED): New function.
394 (visit_phi): When the degenerate result is from the backedge and
395 we didn't visit its definition yet drop to VARYING.
396 (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
398 2018-08-29 Jan Hubicka <jh@suse.cz>
400 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
402 (hash_tree): Likewise.
404 2018-08-29 Jan Hubicka <jh@suse.cz>
406 * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
407 and TYPE_NEXT_REF_TO.
409 2018-08-29 Jan Hubicka <jh@suse.cz>
411 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
412 comment that it has to be even number.
413 (class sreal): Change m_sig type to int32_t.
414 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
415 int64_t for temporary calculations.
416 (sreal_verify_basics): Drop one bit from minimum and maximum.
418 2018-08-30 Tamar Christina <tamar.christina@arm.com>
420 * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
422 2018-08-30 Vlad Lazar <vlad.lazar@arm.com>
425 * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
426 if to_add is negative.
428 2018-08-29 Bernd Edlinger <bernd.edlinger@hotmail.de>
431 * builtins.c (c_strlen): Improve range checks.
433 2018-08-29 Martin Sebor <msebor@redhat.com>
434 Jeff Law <law@redhat.com>
436 PR tree-optimization/86714
437 PR tree-optimization/86711
438 * builtins.c (c_strlen): Add arguments to call to string_constant.
439 * expr.c (string_constant): Add argument. Detect missing nul
440 terminator and outermost declaration it's missing in.
441 * expr.h (string_constant): Add argument.
442 * fold-const.c (read_from_constant_string): Add arguments to call to
444 (c_getstr): Likewise.
445 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
447 * tree-ssa-strlen.c (get_stridx): Likewise.
449 2018-08-29 Jan Hubicka <jh@suse.cz>
451 * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
452 Do not stream DECL_VINDEX.
453 * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
454 * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
455 (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
457 2018-08-29 Richard Biener <rguenther@suse.de>
459 * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
460 virtual operands that are not default defs to honor region
462 (rpo_vn_valueize): Remove ineffective code here.
464 2018-08-29 Richard Biener <rguenther@suse.de>
466 PR tree-optimization/87132
467 * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
468 when skipping defs reachable over backedges.
470 2018-08-29 Richard Biener <rguenther@suse.de>
472 * tree-core.h: Document use of deprecated_flag in SSA_NAME.
473 * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
474 * tree-into-ssa.c (pass_build_ssa::execute): Initialize
475 function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
476 * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
477 (vn_reference_lookup_3): Remove use of const_parms.
478 (free_rpo_vn): Do not free const_parms.
479 (do_rpo_vn): Do not call init_const_parms.
480 * tree-ssa-alias.c (refs_may_alias_p_1): Honor
481 SSA_NAME_POINTS_TO_READONLY_MEMORY.
482 (call_may_clobber_ref_p_1): Likewise.
484 2018-08-29 Alexander Monakov <amonakov@ispras.ru>
487 * invoke.texi (Optimization Options): List -ftree-scev-cprop.
489 (-ftree-scev-cprop): Document.
491 2018-08-29 Jan Hubicka <jh@suse.cz>
493 * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
495 (sreal constructor): Update.
496 * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
497 sreal:operator/): Update.
499 2018-08-29 Martin Liska <mliska@suse.cz>
501 * tree-switch-conversion.c (switch_conversion::expand):
502 Strenghten assumption about gswitch statements.
504 2018-08-29 Richard Biener <rguenther@suse.de>
506 PR tree-optimization/87117
507 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
508 re-value-number released SSA VDEFs.
510 2018-08-29 Richard Biener <rguenther@suse.de>
512 PR tree-optimization/87126
513 * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
515 2018-08-28 Jim Wilson <jimw@sifive.com>
517 * config/riscv/pic.md: Rewrite.
518 * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
520 * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
521 (SOFTF, default_load, softload, softstore): New.
523 2018-08-28 Jeff Law <law@redhat.com>
525 * fold-const.c (fold_binary_loc): Remove recently added assert.
527 2018-08-28 Joern Rennecke <joern.rennecke@riscy-ip.com>
529 * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
530 to OP parmeter of generated function.
532 2018-08-28 MCC CS <deswurstes@users.noreply.github.com>
534 PR tree-optimization/87009
535 * match.pd: Add boolean optimizations.
537 2018-08-28 Martin Sebor <msebor@redhat.com>
540 * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
541 * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
542 (pass_walloca::gate): Use it.
543 (alloca_call_type): Same.
544 (pass_walloca::execute): Same.
545 * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
547 2018-08-28 David Malcolm <dmalcolm@redhat.com>
549 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
550 GCC_VERSION for usage of "__gcc_dump_printf__" format from
553 2018-08-28 Richard Biener <rguenther@suse.de>
555 PR tree-optimization/87124
556 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
557 constants before looking up avail.
559 2018-08-28 Jakub Jelinek <jakub@redhat.com>
562 * calls.c (maybe_warn_nonstring_arg): Punt early if
563 warn_stringop_overflow is zero. Don't call get_range_strlen
564 on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
565 Swap comparison operands to have constants on rhs. Only use
566 lenrng[1] if non-NULL and INTEGER_CST. Don't uselessly
569 2018-08-28 Richard Sandiford <richard.sandiford@arm.com>
571 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
572 use of tree_to_shwi. Remove duplicated test for the size being
573 a whole number of bytes.
575 2018-08-28 Richard Biener <rguenther@suse.de>
577 PR tree-optimization/87117
578 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
579 Handle removed stmt without LHS (GIMPLE_NOP).
581 2018-08-28 Richard Biener <rguenther@suse.de>
583 PR tree-optimization/87117
584 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
585 void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
587 2018-08-28 Richard Biener <rguenther@suse.de>
589 PR tree-optimization/87117
590 * tree-ssa-pre.c (compute_avail): Do not make expressions
591 with predicated values available.
592 (get_expr_value_id): Assert we do not run into predicated value
595 2018-08-28 Richard Biener <rguenther@suse.de>
597 PR tree-optimization/87117
598 * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
599 get virtual operands.
600 (get_expr_operands): Handle STRING_CST like other decls.
602 2018-08-28 Martin Liska <mliska@suse.cz>
604 * tree.h: Update documentation of fndecl_built_in_p
608 2018-08-27 Jeff Law <law@redhat.com>
609 PR tree-optimization/87110
610 * tree-ssa-dse.c (compute_trims): Handle non-constant
613 2018-08-27 Martin Sebor <msebor@redhat.com>
615 PR tree-optimization/86914
616 * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
618 2018-08-27 Martin Sebor <msebor@redhat.com>
620 PR tree-optimization/87112
621 * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
622 the type of the bound argument.
624 2018-08-27 Jeff Law <law@redhat.com>
626 * tree-ssa-dse.c (compute_trims): Handle case where the reference's
627 type does not have a TYPE_SIZE_UNIT.
629 2018-08-27 Steve Ellcey <sellcey@cavium.com>
631 * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
632 with include of backend.h.
634 2018-08-27 Richard Biener <rguenther@suse.de>
636 PR tree-optimization/86927
637 * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
638 use const cond reduction code.
640 2018-08-27 Alexander Monakov <amonakov@ispras.ru>
642 PR tree-optimization/85758
643 * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
645 2018-08-27 David Malcolm <dmalcolm@redhat.com>
648 * diagnostic-show-locus.c (class layout_range): Update for
649 conversion of show_caret_p to a tri-state.
650 (layout_range::layout_range): Likewise.
651 (make_range): Likewise.
652 (layout::maybe_add_location_range): Likewise.
653 (layout::should_print_annotation_line_p): Don't show annotation
654 lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
655 (layout::get_state_at_point): Update for conversion of
656 show_caret_p to a tri-state. Bail out early for
657 SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
658 underlining or source colorization.
659 (gcc_rich_location::add_location_if_nearby): Update for conversion
660 of show_caret_p to a tri-state.
661 (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
662 (selftest::test_one_liner_fixit_replace_equal_secondary_range):
664 (selftest::test_one_liner_labels): Likewise.
665 * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
666 conversion of show_caret_p to a tri-state.
667 * pretty-print.c (text_info::set_location): Likewise.
668 * pretty-print.h (text_info::set_location): Likewise.
669 * substring-locations.c (format_warning_n_va): Likewise.
670 * tree-diagnostic.c (default_tree_printer): Likewise.
671 * tree-pretty-print.c (newline_and_indent): Likewise.
673 2018-08-27 David Malcolm <dmalcolm@redhat.com>
676 * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
677 line above for line-insertion fix-it hints.
678 (selftest::test_fixit_insert_containing_newline): Update the
679 expected results, and add a test with line-numbering enabled.
681 2018-08-27 Martin Liska <mliska@suse.cz>
684 * sanopt.c (sanitize_rewrite_addressable_params): Ignore
685 params with DECL_HAS_VALUE_EXPR_P.
687 2018-08-27 Martin Liska <mliska@suse.cz>
689 * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
690 selected expansion strategy.
692 2018-08-27 Martin Liska <mliska@suse.cz>
694 * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
695 * builtins.c (is_builtin_fn): Likewise.
696 * attribs.c (diag_attr_exclusions): Use new function
697 fndecl_built_in_p and remove check for FUNCTION_DECL if
699 (builtin_mathfn_code): Likewise.
700 (fold_builtin_expect): Likewise.
701 (fold_call_expr): Likewise.
702 (fold_builtin_call_array): Likewise.
703 (fold_call_stmt): Likewise.
704 (set_builtin_user_assembler_name): Likewise.
705 (is_simple_builtin): Likewise.
706 * calls.c (gimple_alloca_call_p): Likewise.
707 (maybe_warn_nonstring_arg): Likewise.
708 * cfgexpand.c (expand_call_stmt): Likewise.
709 * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
710 (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
711 (cgraph_node::verify_node): Likewise.
712 * cgraphclones.c (build_function_decl_skip_args): Likewise.
713 (cgraph_node::create_clone): Likewise.
714 * config/arm/arm.c (arm_insert_attributes): Likewise.
715 * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
716 * dse.c (scan_insn): Likewise.
717 * expr.c (expand_expr_real_1): Likewise.
718 * fold-const.c (operand_equal_p): Likewise.
719 (fold_binary_loc): Likewise.
720 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
721 * gimple-low.c (lower_stmt): Likewise.
722 * gimple-pretty-print.c (dump_gimple_call): Likewise.
723 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Likewise.
724 * gimple.c (gimple_build_call_from_tree): Likewise.
725 (gimple_call_builtin_p): Likewise.
726 (gimple_call_combined_fn): Likewise.
727 * gimplify.c (gimplify_call_expr): Likewise.
728 (gimple_boolify): Likewise.
729 (gimplify_modify_expr): Likewise.
730 (gimplify_addr_expr): Likewise.
731 * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
732 * ipa-cp.c (determine_versionability): Likewise.
733 * ipa-fnsummary.c (compute_fn_summary): Likewise.
734 * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
735 * ipa-split.c (visit_bb): Likewise.
736 (split_function): Likewise.
737 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
738 * lto-cgraph.c (input_node): Likewise.
739 * lto-streamer-out.c (write_symbol): Likewise.
740 * omp-low.c (setjmp_or_longjmp_p): Likewise.
741 (lower_omp_1): Likewise.
742 * predict.c (strip_predict_hints): Likewise.
743 * print-tree.c (print_node): Likewise.
744 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
745 * trans-mem.c (is_tm_irrevocable): Likewise.
746 (is_tm_load): Likewise.
747 (is_tm_simple_load): Likewise.
748 (is_tm_store): Likewise.
749 (is_tm_simple_store): Likewise.
750 (is_tm_abort): Likewise.
751 (tm_region_init_1): Likewise.
752 * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
753 * tree-cfg.c (verify_gimple_call): Likewise.
754 (move_stmt_r): Likewise.
755 (stmt_can_terminate_bb_p): Likewise.
756 * tree-eh.c (lower_eh_constructs_2): Likewise.
757 * tree-if-conv.c (if_convertible_stmt_p): Likewise.
758 * tree-inline.c (remap_gimple_stmt): Likewise.
760 (estimate_num_insns): Likewise.
761 (fold_marked_statements): Likewise.
762 * tree-sra.c (scan_function): Likewise.
763 * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
764 (optimize_stack_restore): Likewise.
765 (pass_fold_builtins::execute): Likewise.
766 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
767 (mark_all_reaching_defs_necessary_1): Likewise.
768 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
769 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
770 (pass_forwprop::execute): Likewise.
771 * tree-ssa-loop-im.c (stmt_cost): Likewise.
772 * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
773 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
774 * tree-ssa-strlen.c (get_string_length): Likewise.
775 * tree-ssa-structalias.c (handle_lhs_call): Likewise.
776 (find_func_aliases_for_call): Likewise.
777 * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
778 * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
779 * tree-tailcall.c (find_tail_calls): Likewise.
780 * tree.c (need_assembler_name_p): Likewise.
781 (free_lang_data_in_decl): Likewise.
782 (get_call_combined_fn): Likewise.
783 * ubsan.c (is_ubsan_builtin_p): Likewise.
784 * varasm.c (incorporeal_function_p): Likewise.
785 * tree.h (DECL_BUILT_IN): Remove and replace with
787 (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
788 (fndecl_built_in_p): New.
790 2018-08-27 Martin Liska <mliska@suse.cz>
792 PR tree-optimization/86847
793 * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
794 Dump also subtree probability.
795 (switch_decision_tree::do_jump_if_equal): New function.
796 (switch_decision_tree::emit_case_nodes): Handle special
797 situations in balanced tree that can be emitted much simpler.
798 Fix calculation of probabilities that happen in tree expansion.
799 * tree-switch-conversion.h (struct cluster): Add
801 (struct simple_cluster): Likewise.
802 (struct case_tree_node): Add new function has_child.
803 (do_jump_if_equal): New.
805 2018-08-27 Martin Liska <mliska@suse.cz>
807 * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
808 Add new argument to bit_test_cluster constructor.
809 (bit_test_cluster::emit): Set bits really number of values
811 (bit_test_cluster::hoist_edge_and_branch_if_true): Add
812 probability argument.
813 * tree-switch-conversion.h (struct bit_test_cluster):
814 Add m_handles_entire_switch.
816 2018-08-27 Martin Liska <mliska@suse.cz>
818 PR tree-optimization/86702
819 * tree-switch-conversion.c (jump_table_cluster::emit):
820 Make probabilities even for values in jump table
821 according to number of cases handled.
822 (switch_decision_tree::compute_cases_per_edge): Pass
823 argument to reset_out_edges_aux function.
824 (switch_decision_tree::analyze_switch_statement): Likewise.
825 * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
828 2018-08-27 Martin Liska <mliska@suse.cz>
830 * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
831 cfun argument explicitly.
832 * gimple-pretty-print.c (dump_gimple_switch): Likewise.
833 * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
834 function gimple_switch_default_bb.
835 (convert_switch_statements):
837 * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
838 * stmt.c (label_to_block_fn): Use label_to_block and pass
839 cfun argument explicitly and use gimple_switch_label_bb.
840 (expand_case): Likewise.
841 * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
842 cfun argument explicitly. Likewise.
843 (make_edges_bb): Likewise.
844 (make_cond_expr_edges): Likewise.
845 (get_cases_for_edge): Likewise.
846 (make_gimple_switch_edges): Likewise.
847 (label_to_block_fn): Likewise.
848 (label_to_block): Likewise.
849 (make_goto_expr_edges): Likewise.
850 (make_gimple_asm_edges): Likewise.
851 (main_block_label): Likewise.
852 (group_case_labels_stmt): Likewise.
853 (find_taken_edge_computed_goto): Likewise.
854 (find_taken_edge_switch_expr): Likewise.
855 (gimple_verify_flow_info): Likewise.
856 (gimple_redirect_edge_and_branch): Likewise.
857 (gimple_switch_label_bb): New function.
858 (gimple_switch_default_bb): Likewise.
859 (gimple_switch_edge): Likewise.
860 (gimple_switch_default_edge): Likewise.
861 * tree-cfg.h (label_to_block_fn): Remove and replace ...
862 (label_to_block): ... with this.
863 (gimple_switch_label_bb): New.
864 (gimple_switch_default_bb): Likewise.
865 (gimple_switch_edge): Likewise.
866 (gimple_switch_default_edge): Likewise.
867 * tree-cfgcleanup.c (convert_single_case_switch): Use
868 new gimple functions and pass new argument to label_to_block.
869 (cleanup_control_flow_bb):
870 * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
871 cfun argument explicitly.
872 (make_eh_edges): Likewise.
873 (redirect_eh_dispatch_edge): Likewise.
874 (lower_resx): Likewise.
875 (lower_eh_dispatch): Likewise.
876 (maybe_remove_unreachable_handlers): Likewise.
877 (unsplit_eh): Likewise.
878 (cleanup_empty_eh): Likewise.
879 (verify_eh_edges): Likewise.
880 (verify_eh_dispatch_edge): Likewise.
881 * tree-ssa-dom.c (record_edge_info): Likewise.
882 * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
883 * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
884 (thread_through_normal_block): Likewise.
885 * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
886 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
887 * tree-switch-conversion.c (switch_conversion::collect): Use new
889 (switch_conversion::check_final_bb): Likewise.
890 (switch_conversion::gather_default_values): Pass new argument
892 (switch_conversion::build_constructors): Likewise.
893 (switch_decision_tree::compute_cases_per_edge): Use new
894 gimple_switch_edge function.
895 (switch_decision_tree::analyze_switch_statement): Pass new argument
897 (switch_decision_tree::try_switch_expansion): Use
898 gimple_switch_default_edge.
899 * tree-vrp.c (find_switch_asserts): Pass new argument
901 * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
902 (vr_values::simplify_switch_using_ranges): Likewise.
904 2018-08-27 Richard Biener <rguenther@suse.de>
906 * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
907 * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
909 * tree-ssa-sccvn.h (struct vn_pval): New structure.
910 (struct vn_nary_op_s): Add unwind_to member. Add
911 predicated_values flag and put result into a union together
912 with a linked list of vn_pval.
913 (struct vn_ssa_aux): Add name member to make maintaining
914 a map of SSA name to vn_ssa_aux possible. Remove no longer
915 needed info, dfsnum, low, visited, on_sccstack, use_processed
916 and range_info_anti_range_p members.
917 (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
918 (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
920 (vn_valueize): New global.
921 (vn_context_bb): Likewise.
922 (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
923 VN_INFO_PTR_INFO): Remove.
924 * tree-ssa-sccvn.c: ... (rewrite)
925 (pass_fre::execute): For -O2+ initialize loops and run
926 RPO VN in optimistic mode (iterating). For -O1 and -Og
927 run RPO VN in non-optimistic mode.
928 * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
929 (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
930 * doc/invoke.texi (sccvn-max-scc-size): Remove.
931 (rpo-vn-max-loop-depth): Document.
932 * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
933 when valuezing the VUSE signals we walked out of the region.
934 * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
935 (phi_translate): Set VN context block to use for availability
937 (compute_avail): Likewise.
938 (pre_valueize): New function.
939 (pass_pre::execute): Adjust to the RPO VN API.
941 * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
942 (propagate_constants_for_unrolling): Remove.
943 (tree_unroll_loops_completely): Perform value-numbering
944 on the unrolled bodies loop parent.
946 2018-08-27 Richard Biener <rguenther@suse.de>
948 * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
949 for partial antic compute.
951 2018-08-27 Jakub Jelinek <jakub@redhat.com>
953 PR rtl-optimization/87065
954 * combine.c (simplify_if_then_else): Formatting fix.
955 (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
957 (known_cond): Don't return const_true_rtx for vector modes. Use
958 CONST0_RTX instead of const0_rtx. Formatting fixes.
960 2018-08-27 Martin Liska <mliska@suse.cz>
962 PR gcov-profile/87069
963 * gcov.c (process_file): Record files already processed
964 and warn about a file being processed multiple times.
966 2018-08-27 Martin Liska <mliska@suse.cz>
969 * config/aarch64/aarch64.c (aarch64_override_options_internal):
970 Set default values for x_aarch64_*_string strings.
971 * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
972 prefix. For -mabi do not print '=ABI' in help and use
973 <option_value> format for -msve-vector-bits and -moverride
976 2018-08-26 Jeff Law <law@redhat.com>
978 * config/mips/frame-header-opt.c: Include "backend.h" rather than
981 2018-08-26 Marek Polacek <polacek@redhat.com>
983 PR c++/87029, Implement -Wredundant-move.
984 * doc/invoke.texi: Document -Wredundant-move.
986 2018-08-25 Martin Sebor <msebor@redhat.com>
988 PR tree-optimization/87059
989 * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
990 to the same type as the other.
991 * fold-const.c (fold_binary_loc): Assert expectation.
993 2018-08-25 Iain Sandoe <iain@sandoe.co.uk>
995 * config/darwin.c (machopic_legitimize_pic_address): Clean up
996 extraneous parentheses, dead code section and formatting.
998 2018-08-24 David Malcolm <dmalcolm@redhat.com>
1001 * diagnostic-show-locus.c (layout::layout): Ensure the margin is
1002 wide enough for jumps in the line-numbering to be visible.
1003 (layout::print_gap_in_line_numbering): New member function.
1004 (layout::calculate_line_spans): When using line numbering, merge
1005 line spans that are only 1 line apart.
1006 (diagnostic_show_locus): When printing line numbers, show gaps in
1007 line numbering directly, rather than printing headers.
1008 (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
1009 line-numbering with multiple line spans.
1010 (selftest::test_fixit_insert_containing_newline_2): Add test of
1011 line-numbering, in which the spans are close enough to be merged.
1013 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
1015 * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
1016 to range_includes_zero_p. Do not special case VR_ANTI_RANGE.
1017 * tree-vrp.c (range_is_nonnull): Remove.
1018 (range_includes_zero_p): Accept value_range instead of min/max.
1019 (extract_range_from_binary_expr_1): Do not early bail on
1021 Use range_includes_zero_p instead of range_is_nonnull.
1022 (extract_range_from_unary_expr): Use range_includes_zero_p instead
1023 of range_is_nonnull.
1024 (vrp_meet_1): Pass value_range to range_includes_zero_p. Do not
1025 special case VR_ANTI_RANGE.
1026 (vrp_finalize): Same.
1027 * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
1029 (range_is_nonnull): Remove.
1030 * vr-values.c (vrp_stmt_computes_nonzero): Use
1031 range_includes_zero_p instead of range_is_nonnull.
1032 (extract_range_basic): Pass value_range to range_includes_zero_p
1033 instead of range_is_nonnull.
1035 2018-08-24 Uros Bizjak <ubizjak@gmail.com>
1037 * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
1038 * emit-rtl.h (rtl_data): Remove return_bnd.
1039 * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
1040 * function.c (diddle_return_value): Do not handle crtl->return_bnd.
1041 * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
1042 (POINTER_BOUNDS_MODE): Remove definition.
1043 (make_pointer_bounds_mode): Remove.
1044 (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
1045 * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
1046 (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
1047 * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
1048 * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
1049 * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
1050 * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
1052 * config/i386/i386-modes.def (BND32, BND64): Remove.
1053 * config/i386/i386.c (dbx_register_map): Remove bound registers.
1054 (dbx64_register_map): Ditto.
1055 (svr4_dbx_register_map): Ditto.
1056 (indirect_thunk_bnd_needed): Remove.
1057 (indirect_thunks_bnd_used): Ditto.
1058 (indirect_return_bnd_needed): Ditto.
1059 (indirect_return_via_cx_bnd): Ditto.
1060 (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
1061 (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
1062 (output_indirect_thunk): Ditto. Remove need_prefix argument.
1063 (output_indirect_thunk_function): Remove handling of
1064 indirect_return_bnd_needed, indirect_return_via_cx_bnd,
1065 indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
1066 (ix86_save_reg): Remove handling of crtl->return_bnd.
1067 (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
1068 (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
1069 and UNSPEC_BNDLX_ADDR.
1070 (ix86_output_indirect_branch_via_reg): Remove handling of
1071 indirect_thunk_prefix_bnd.
1072 (ix86_output_indirect_branch_via_push): Ditto.
1073 (ix86_output_function_return): Ditto.
1074 (ix86_output_indirect_function_return): Ditto.
1075 (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
1076 * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
1077 (CALL_USED_REGISTERS): Ditto.
1078 (REG_ALLOC_ORDER): Update for removal of bound registers.
1079 (HI_REGISTER_NAMES): Ditto.
1080 * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
1081 (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
1082 (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
1083 (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
1084 (FIRST_PSEUDO_REG): Update.
1085 (BND): Remove mode iterator.
1086 * config/i386/predicates.md (bnd_mem_operator): Remove.
1088 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
1090 * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
1093 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
1095 * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
1096 the case in which the permute needs only a single element and
1097 repeats for every vector of the result. Extend that case to
1098 handle variable-length vectors.
1099 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
1101 2018-08-24 H.J. Lu <hongjiu.lu@intel.com>
1104 * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
1105 on DW_MACINFO_start_file for -gsplit-dwarf -g3.
1107 2018-08-24 Richard Biener <rguenther@suse.de>
1109 * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
1110 bb_flags_allocated members.
1111 (auto_flag): New RAII class for allocating flags.
1112 (auto_edge_flag): New RAII class for allocating edge flags.
1113 (auto_bb_flag): New RAII class for allocating bb flags.
1114 * cfgloop.c (verify_loop_structure): Allocate temporary edge
1116 * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
1117 in favor of temporarily allocated BB flag.
1118 * hsa-brig.c: Re-order includes.
1119 * hsa-dump.c: Likewise.
1120 * hsa-regalloc.c: Likewise.
1121 * print-rtl.c: Likewise.
1122 * profile-count.c: Likewise.
1124 2018-08-24 Segher Boessenkool <segher@kernel.crashing.org>
1127 * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
1130 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
1133 * wide-int-range.cc (wide_int_range_div): Do not ignore result
1134 from wide_int_range_multiplicative_op.
1136 2018-08-23 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1138 * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
1141 2018-08-23 Giuliano Belinassi <giuliano.belinassi@usp.br>
1143 * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
1146 2018-08-23 Alexander Monakov <amonakov@ispras.ru>
1148 * tree-scalar-evolution.c (final_value_replacement_loop): Dump
1149 full GENERIC expression used for replacement.
1151 2018-08-23 Aldy Hernandez <aldyh@redhat.com>
1153 * tree-vrp.c (abs_extent_range): Remove.
1154 (extract_range_into_wide_ints): Pass wide ints by reference.
1155 (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
1156 Pass wide ints by reference in all calls to
1157 extract_range_into_wide_ints.
1158 * wide-int-range.cc (wide_int_range_div): New.
1159 * wide-int-range.h (wide_int_range_div): New.
1160 (wide_int_range_includes_zero_p): New.
1161 (wide_int_range_zero_p): New.
1163 2018-08-23 Matthew Malcomson <matthew.malcomson@arm.com>
1165 * config/aarch64/aarch64.md (arches): New enum.
1166 (arch): New enum attr.
1167 (arch_enabled): New attr.
1168 (enabled): Now uses arch_enabled only.
1169 (simd, sve, fp16): Removed attribute.
1170 (fp): Attr now defined in terms of 'arch'.
1171 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
1172 *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
1173 <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
1174 <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
1175 attributes into 'arch'.
1176 (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
1177 subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
1178 *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
1179 *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
1180 *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
1181 'simd' attribute into 'arch'.
1182 (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
1183 store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
1184 Convert use of 'fp' attribute to 'arch'.
1185 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
1186 move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
1188 (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
1189 (different modes) Merge 'fp' and 'simd' into 'arch'.
1190 (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
1193 2018-08-23 Segher Boessenkool <segher@kernel.crashing.org>
1195 PR rtl-optimization/87026
1196 * expmed.c (canonicalize_comparison): If we can no longer create
1197 pseudoregisters, don't.
1199 2018-08-23 Richard Earnshaw <rearnsha@arm.com>
1202 * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
1204 * config/arm/arm.c (speculation_barrier_libfunc): New static
1206 (arm_init_libfuncs): Initialize it.
1207 (arm_emit_speculation_barrier): New function.
1208 * config/arm/arm.md (speculation_barrier): Call
1209 arm_emit_speculation_barrier for architectures that do not have
1211 (speculation_barrier_insn): Only match on Armv7 or later.
1213 2018-08-23 Richard Biener <rguenther@suse.de>
1216 * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
1219 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
1221 * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
1222 of single-vector TBLs.
1223 (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
1226 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
1229 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
1230 aarch64_evpc_tbl guard.
1232 2018-08-22 Bernd Edlinger <bernd.edlinger@hotmail.de>
1234 * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
1237 2018-08-22 Martin Sebor <msebor@redhat.com>
1240 * tree-pretty-print.c (pretty_print_string): Add argument.
1241 (dump_generic_node): Call to pretty_print_string with string size.
1243 2018-08-22 Segher Boessenkool <segher@kernel.crashing.org>
1245 PR rtl-optimization/86771
1246 * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
1247 of two SETs into those two SETs, one to be placed at i2, if that SETs
1248 destination is modified between i2 and i3.
1250 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
1252 PR tree-optimization/86725
1253 * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
1255 (vect_analyze_scalar_cycles_1): Check it.
1257 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
1259 PR tree-optimization/86725
1260 * tree-vect-loop.c (vect_is_simple_reduction): When treating
1261 an outer loop phi as a double reduction, make sure that the
1262 single user of the phi result is an inner loop phi.
1264 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
1266 * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
1267 grouped stores with gaps to a strided group.
1269 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
1271 * tree-vect-stmts.c (get_group_load_store_type)
1272 (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
1273 first statement in a group.
1275 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
1277 * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
1278 the sequence used in gcc/gcc.c.
1280 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
1283 * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
1286 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
1288 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
1289 Darwin10-specific unwinder-shim.
1290 * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
1291 * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
1292 New to cater for Darwin10 Rosetta.
1294 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
1296 * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
1297 specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
1299 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
1304 * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
1305 (dwarf2out_switch_text_section): Generate a local label for the
1306 second function sub-section and apply it as the second FDE start
1308 * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
1309 second sub-section start.
1311 2018-08-22 Richard Biener <rguenther@suse.de>
1313 PR tree-optimization/86988
1314 * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
1316 2018-08-22 Richard Biener <rguenther@suse.de>
1318 PR tree-optimization/86945
1319 * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
1321 2018-08-22 Alexandre Oliva <oliva@adacore.com>
1323 * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
1324 a comment about how uses of r2 for .sdata2 come about.
1326 2018-08-22 Alexandre Oliva <aoliva@redhat.com>
1328 * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
1330 2018-08-21 Marek Polacek <polacek@redhat.com>
1332 PR c++/86981, Implement -Wpessimizing-move.
1333 * doc/invoke.texi: Document -Wpessimizing-move.
1335 2018-08-21 Jan Hubicka <jh@suse.cz>
1337 * tree.c (find_decls_types_r): Do not check for redundant typedefs.
1338 * tree.h (is_redundant_typedef): Remove.
1339 * dwarf2out.c (is_redundant_typedef): Turn into static function.
1341 2018-08-21 Jan Hubicka <jh@suse.cz>
1343 * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
1346 2018-08-21 Tamar Christina <tamar.christina@arm.com>
1348 * expmed.c (extract_low_bits): Reject invalid subregs early.
1350 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
1353 * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
1356 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
1358 * config/vxworks.h: Guard vxworks_asm_out_constructor and
1359 vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
1360 * config/vxworks.c: Likewise.
1362 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
1364 * config/vxworks.c: Set targetm.have_ctors_dtors
1365 if HAVE_INITFINI_ARRAY_SUPPORT.
1366 * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
1367 if HAVE_INITFINI_ARRAY_SUPPORT.
1369 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
1371 * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
1372 default search path for VxWorks < 7.
1374 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
1376 * gimple-ssa-sprintf.c (decl_constant_value): Remove.
1377 (get_format_string): Refer to c_getstr.
1379 2018-08-21 Tom de Vries <tdevries@suse.de>
1381 * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
1382 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
1383 (debuginfo_early_stop): Declare.
1384 * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
1385 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
1386 (debuginfo_early_stop): New function.
1387 (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
1388 and debuginfo_early_stop.
1389 * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
1390 * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
1391 (general_init): Call debuginfo_early_init.
1392 (finalize): Call debuginfo_fini.
1393 (do_compile): Call debuginfo_init.
1394 * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
1396 (@item -fdump-debug, @item -fdump-earlydebug): Add.
1398 2018-08-21 Tom de Vries <tdevries@suse.de>
1400 * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
1401 flag_dump_noaddr and flag_dump_unnumbered.
1403 2018-08-21 Aldy Hernandez <aldyh@redhat.com>
1405 * wide-int-range.cc (wide_int_range_abs): New.
1406 (wide_int_range_order_set): Rename from wide_int_range_min_max.
1407 * wide-int-range.h (wide_int_range_abs): New.
1408 (wide_int_range_min_max): New.
1409 * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
1410 case to call wide_int_range_abs.
1411 Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
1412 (extract_range_from_abs_expr): Delete.
1414 2018-08-20 Michael Meissner <meissner@linux.ibm.com>
1417 * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
1418 from 'Y' to 'YZ' to enable the LWAX instruction to be generated
1421 2018-08-20 Nathan Sidwell <nathan@acm.org>
1422 Jeff Law <law@redhat.com>
1424 * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
1425 * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
1427 2018-08-20 David Malcolm <dmalcolm@redhat.com>
1430 * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
1431 (decl_attributes): Likewise.
1432 * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
1434 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
1435 * diagnostic-core.h (class auto_diagnostic_group): New class.
1436 * diagnostic.c (diagnostic_initialize): Initialize the new fields.
1437 (diagnostic_report_diagnostic): Handle the first diagnostics within
1439 (emit_diagnostic): Add auto_diagnostic_group instance.
1441 (inform_n): Likewise.
1442 (warning): Likewise.
1443 (warning_at): Likewise.
1444 (warning_n): Likewise.
1445 (pedwarn): Likewise.
1446 (permerror): Likewise.
1448 (error_n): Likewise.
1449 (error_at): Likewise.
1451 (fatal_error): Likewise.
1452 (internal_error): Likewise.
1453 (internal_error_no_backtrace): Likewise.
1454 (auto_diagnostic_group::auto_diagnostic_group): New ctor.
1455 (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
1456 * diagnostic.h (struct diagnostic_context): Add fields
1457 "diagnostic_group_nesting_depth",
1458 "diagnostic_group_emission_count", "begin_group_cb",
1460 * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
1461 Add auto_diagnostic_group instance(s).
1462 (find_explicit_erroneous_behavior): Likewise.
1463 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
1464 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
1465 * gimplify.c (warn_implicit_fallthrough_r): Likewise.
1466 (gimplify_va_arg_expr): Likewise.
1467 * hsa-gen.c (HSA_SORRY_ATV): Likewise.
1468 (HSA_SORRY_AT): Likewise.
1469 * ipa-devirt.c (compare_virtual_tables): Likewise.
1470 (warn_odr): Likewise.
1471 * multiple_target.c (expand_target_clones): Likewise.
1472 * opts-common.c (cmdline_handle_error): Likewise.
1473 * reginfo.c (globalize_reg): Likewise.
1474 * substring-locations.c (format_warning_n_va): Likewise.
1475 * tree-inline.c (expand_call_inline): Likewise.
1476 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
1477 * tree-ssa-loop-niter.c
1478 (do_warn_aggressive_loop_optimizations): Likewise.
1479 * tree-ssa-uninit.c (warn_uninit): Likewise.
1480 * tree.c (warn_deprecated_use): Likewise.
1482 2018-08-20 H.J. Lu <hongjiu.lu@intel.com>
1485 * config/i386/i386.md (eh_return): Always update EH return
1486 address in word_mode.
1488 2018-08-20 Chung-Lin Tang <cltang@codesourcery.com>
1490 * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
1491 TARGET_SPLIT_COMPLEX_ARG is defined.
1493 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
1495 * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
1497 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
1500 * expr.c (expand_assignment): Assert that bitpos is positive.
1501 (store_field): Likewise
1502 (expand_expr_real_1): Make sure that bitpos is positive.
1503 * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
1506 2018-08-20 Nathan Sidwell <nathan@acm.org>
1508 * Makefile.in (CPP_ID_DATA_H): Delete.
1509 (CPP_INTERNAL_H): Don't add it.
1510 (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
1511 * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
1513 2018-08-20 Richard Biener <rguenther@suse.de>
1515 PR tree-optimization/78655
1516 * tree-vrp.c (extract_range_from_binary_expr_1): Make
1517 pointer + offset nonnull if either operand is nonnull work.
1519 2018-08-20 Tom de Vries <tdevries@suse.de>
1521 * dwarf2out.c (add_scalar_info): Don't add reference to existing die
1522 unless the referenced die describes the added property using
1523 DW_AT_location or DW_AT_const_value. Fall back to exprloc case.
1524 Otherwise, add a DW_AT_location to the referenced die.
1526 2018-08-19 Uros Bizjak <ubizjak@gmail.com>
1529 * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
1530 register_operand when calling ix86_set_reg_reg_cost.
1531 [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
1532 Set *total to 0 for operands that satisfy x86_64_immediate_operand
1533 predicate and to 1 otherwise.
1535 2018-08-18 Iain Sandoe <iain@sandoe.co.uk>
1537 * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
1538 emit a diagnostic that it is not supported and reset the option.
1539 * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
1540 supported and consume the option. (ASM_FINAL_SPEC): New.
1542 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
1544 * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
1547 2018-08-17 Sandra Loosemore <sandra@codesourcery.com>
1549 C-SKY port: Documentation
1551 * doc/extend.texi (C-SKY Function Attributes): New section.
1552 * doc/invoke.texi (Option Summary): Add C-SKY options.
1553 (C-SKY Options): New section.
1554 * doc/md.texi (Machine Constraints): Document C-SKY constraints.
1556 2018-08-17 Jojo <jijie_rong@c-sky.com>
1557 Huibin Wang <huibin_wang@c-sky.com>
1558 Sandra Loosemore <sandra@codesourcery.com>
1559 Chung-Lin Tang <cltang@codesourcery.com>
1561 C-SKY port: Backend implementation
1563 * config/csky/*: New.
1564 * common/config/csky/*: New.
1566 2018-08-17 Jojo <jijie_rong@c-sky.com>
1567 Huibin Wang <huibin_wang@c-sky.com>
1568 Sandra Loosemore <sandra@codesourcery.com>
1569 Chung-Lin Tang <cltang@codesourcery.com>
1570 Andrew Jenner <andrew@codesourcery.com>
1572 C-SKY port: Configury
1574 * config.gcc (csky-*-*): New.
1575 * configure.ac: Add csky to targets for dwarf2 debug_line support.
1576 * configure: Regenerated.
1578 2018-08-17 David Malcolm <dmalcolm@redhat.com>
1580 * dump-context.h: Include "dumpfile.h".
1581 (dump_context::dump_printf_va): Convert final param from va_list
1582 to va_list *. Convert from ATTRIBUTE_PRINTF to
1583 ATTRIBUTE_GCC_DUMP_PRINTF.
1584 (dump_context::dump_printf_loc_va): Likewise.
1585 * dumpfile.c: Include "stringpool.h".
1586 (make_item_for_dump_printf_va): Delete.
1587 (make_item_for_dump_printf): Delete.
1588 (class dump_pretty_printer): New class.
1589 (dump_pretty_printer::dump_pretty_printer): New ctor.
1590 (dump_pretty_printer::emit_items): New member function.
1591 (dump_pretty_printer::emit_any_pending_textual_chunks): New member
1593 (dump_pretty_printer::emit_item): New member function.
1594 (dump_pretty_printer::stash_item): New member function.
1595 (dump_pretty_printer::format_decoder_cb): New member function.
1596 (dump_pretty_printer::decode_format): New member function.
1597 (dump_context::dump_printf_va): Reimplement in terms of
1598 dump_pretty_printer.
1599 (dump_context::dump_printf_loc_va): Convert final param from va_list
1601 (dump_context::begin_scope): Reimplement call to
1602 make_item_for_dump_printf.
1603 (dump_printf): Update for change to dump_printf_va.
1604 (dump_printf_loc): Likewise.
1605 (selftest::test_capture_of_dump_calls): Convert "stmt" from
1606 greturn * to gimple *. Add a test_decl. Add tests of dump_printf
1607 with %T, %E, and %G.
1608 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
1609 (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
1610 ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
1611 (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
1612 ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
1613 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
1614 use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
1615 within a dump_printf_loc call to "%wu".
1616 (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
1617 converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd". Add a
1618 missing space after "=".
1619 * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
1620 call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
1621 * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
1622 convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
1623 * tree-vectorizer.c (try_vectorize_loop_1): Likewise. Remove
1624 duplicate "vectorized" from message.
1626 2018-08-17 Szabolcs Nagy <szabolcs.nagy@arm.com>
1628 * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
1629 polyNxK_t element's TYPE_STRING_FLAG.
1631 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
1633 * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
1634 (they were unnamed before). Fix comments.
1636 2018-08-17 Nathan Sidwell <nathan@acm.org>
1638 * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
1640 2018-08-17 Richard Biener <rguenther@suse.de>
1642 PR tree-optimization/86841
1643 * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
1645 2018-08-17 Martin Liska <mliska@suse.cz>
1647 * common.opt: Remove Warn, Init and Report for options with
1648 Ignore/Deprecated flag. Warning is done automatically for
1650 * config/i386/i386.opt: Likewise.
1651 * config/ia64/ia64.opt: Likewise.
1652 * config/rs6000/rs6000.opt: Likewise.
1653 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
1654 Remove usage of flag_check_pointer_bounds.
1655 * lto-wrapper.c (merge_and_complain): Do not handle
1656 OPT_fcheck_pointer_bounds.
1657 (append_compiler_options): Likewise.
1658 * opt-functions.awk: Do not handle Deprecated.
1659 * optc-gen.awk: Check that Var, Report and Init are not
1660 used for an option with Ignore/Deprecated flag.
1661 * opts-common.c (decode_cmdline_option): Do not report
1663 (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
1665 * opts.h (struct cl_option): Remove cl_deprecated flag.
1666 (CL_ERR_DEPRECATED): Remove error enum value.
1668 2018-08-17 Richard Biener <rguenther@suse.de>
1671 * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
1672 across a va-arg-pack using call adjust its return value accordingly.
1674 2018-08-16 Martin Sebor <msebor@redhat.com>
1676 PR tree-optimization/86853
1677 * gimple-ssa-sprintf.c (struct format_result): Rename member.
1678 (struct fmtresult): Add member and initialize it in ctors.
1679 (format_character): Handle %C. Extend range to NUL. Set MAYFAIL.
1680 (format_string): Handle %S the same as %ls. Set MAYFAIL.
1681 (format_directive): Set POSUNDER4K when MAYFAIL is set.
1682 (parse_directive): Handle %C same as %c.
1683 (sprintf_dom_walker::compute_format_length): Adjust.
1684 (is_call_safe): Adjust.
1686 2018-08-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
1688 * builtins.c (c_strlen): Add new parameter eltsize. Use it
1689 for determining how to count the elements.
1690 * builtins.h (c_strlen): Adjust prototype.
1691 * expr.c (string_constant): Add new parameter mem_size.
1692 Set *mem_size appropriately.
1693 * expr.h (string_constant): Adjust protoype.
1694 * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
1695 * gimple-fold.h (get_range_strlen): Adjust prototype.
1696 * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
1697 (format_string): Call get_string_length with eltsize.
1699 2018-08-16 David Malcolm <dmalcolm@redhat.com>
1701 * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
1702 to emit the span, rather than setting it as the prefix.
1704 2018-08-16 David Malcolm <dmalcolm@redhat.com>
1706 * diagnostic-show-locus.c (layout::start_annotation_line): Add
1707 "margin_char" parameter, defaulting to space. Use it in place
1708 of pp_space for the initial part of the margin.
1709 (layout::print_leading_fixits): Use '+' when filling the margin
1710 of line-insertion fix-it hints.
1712 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
1714 * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
1717 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
1719 * config/rs6000/altivec.md: Don't set length attribute to the default
1721 * config/rs6000/darwin.md: Ditto.
1722 * config/rs6000/dfp.md: Ditto.
1723 * config/rs6000/htm.md: Ditto.
1724 * config/rs6000/rs6000.md: Ditto.
1725 * config/rs6000/sync.md: Ditto.
1726 * config/rs6000/vsx.md: Ditto.
1728 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
1730 * config/rs6000/altivec.md: Don't set length attribute to the default
1731 value, for branch instructions.
1732 * config/rs6000/darwin.md: Ditto.
1733 * config/rs6000/rs6000.md: Ditto.
1735 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
1737 * config/rs6000/rs6000.md (length): Always define as const_int 4.
1738 (unnamed conditional branch define_insn): Set length to 4 or 8
1739 depending on offset.
1740 (<bd>_<mode>): Similar, for alternative 0.
1741 (<bd>tf_<mode>): Ditto.
1743 2018-08-16 Tamar Christina <tamar.christina@arm.com>
1745 * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
1747 2018-08-16 Matthew Malcomson <matthew.malcomson@arm.com>
1749 * doc/rtl.texi: Replace old RTX class names with new names.
1752 2018-08-16 Vlad Lazar <vlad.lazar@arm.com>
1754 * expmed.h (canonicalize_comparison): New declaration.
1755 * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
1756 * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
1757 * optabs.c (prepare_cmp_insn): Likewise.
1758 * rtl.h (unsigned_condition_p): New function which checks if a
1759 comparison operator is unsigned.
1761 2018-08-16 Nathan Sidwell <nathan@acm.org>
1763 * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
1764 * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
1766 2018-08-16 Tamar Christina <tamar.christina@arm.com>
1769 * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
1770 * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
1771 (mov<mov>): ..this and enable unconditionally.
1773 2018-08-16 Tamar Christina <tamar.christina@arm.com>
1775 * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
1777 2018-08-16 Sam Tebbs <sam.tebbs@arm.com>
1779 * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
1780 (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
1781 "Common" with "Target".
1783 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
1785 * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
1786 * doc/invoke.texi (mmitigate-rop): Remove.
1787 * config/i386/i386.c: Do not include "regrename.h".
1788 (ix86_rop_should_change_byte_p, reg_encoded_number)
1789 (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
1791 (ix86_reorg): Remove call to ix86_mitigate_rop.
1792 * config/i386/i386.md (attr "modrm_class"): Remove.
1793 (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
1794 (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
1795 (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
1797 2018-08-15 Will Schmidt <will_schmidt@vnet.ibm.com>
1799 * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
1800 allow folding of mergeh() and mergel() for the float and double types.
1801 (fold_mergehl_helper): Rework to handle building a permute tree
1804 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
1806 * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
1809 2018-08-15 David Malcolm <dmalcolm@redhat.com>
1811 * common.opt (fdiagnostics-show-labels): New option.
1812 * diagnostic-show-locus.c (class layout_range): Add field
1814 (class layout): Add field "m_show_labels_p".
1815 (layout_range::layout_range): Add param "label" and use it to
1817 (make_range): Pass in NULL for new "label" param of layout_range's
1819 (layout::layout): Initialize m_show_labels_p.
1820 (layout::maybe_add_location_range): Pass in loc_range->m_label
1821 when constructing layout_range instances.
1822 (struct line_label): New struct.
1823 (layout::print_any_labels): New member function.
1824 (layout::print_line): Call it if label-printing is enabled.
1825 (selftest::test_one_liner_labels): New test.
1826 (selftest::test_diagnostic_show_locus_one_liner): Call it.
1827 * diagnostic.c (diagnostic_initialize): Initialize
1828 context->show_labels_p.
1829 * diagnostic.h (struct diagnostic_context): Add field
1831 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
1832 -fno-diagnostics-show-labels.
1833 * dwarf2out.c (gen_producer_string): Add
1834 OPT_fdiagnostics_show_labels to the ignored options.
1835 * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
1837 (gcc_rich_location::maybe_add_expr): Likewise.
1838 * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
1839 label" param, defaulting to NULL.
1840 (gcc_rich_location::add_expr): Add "label" param.
1841 (gcc_rich_location::maybe_add_expr): Likewise.
1842 (class text_range_label): New class.
1843 (class range_label_for_type_mismatch): New class.
1844 * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
1845 of format_warning_va.
1846 (fmtwarn_n): Likewise for new params of format_warning_n_va.
1847 * lto-wrapper.c (merge_and_complain): Add
1848 OPT_fdiagnostics_show_labels to the "pick one setting" options.
1849 (append_compiler_options): Likewise to the dropped options.
1850 (append_diag_options): Likewise to the passed-on options.
1851 * opts.c (common_handle_option): Handle the new option.
1852 * selftest-diagnostic.c
1853 (test_diagnostic_context::test_diagnostic_context): Enable
1855 * substring-locations.c: Include "gcc-rich-location.h".
1856 (format_warning_n_va): Add "fmt_label" and "param_label" params
1857 and use them as appropriate.
1858 (format_warning_va): Add "fmt_label" and "param_label" params,
1859 passing them on to format_warning_n_va.
1860 (format_warning_at_substring): Likewise.
1861 (format_warning_at_substring_n): Likewise.
1862 * substring-locations.h (format_warning_va): Add "fmt_label" and
1863 "param_label" params.
1864 (format_warning_n_va): Likewise.
1865 (format_warning_at_substring): Likewise.
1866 (format_warning_at_substring_n): Likewise.
1867 * toplev.c (general_init): Initialize global_dc->show_labels_p.
1869 2018-08-15 Qing Zhao <qing.zhao@oracle.com>
1872 * builtins.c (expand_builtin_memcmp): Do not expand the call
1873 when overflow is detected.
1875 2018-08-15 Martin Sebor <msebor@redhat.com>
1877 PR tree-optimization/71625
1878 * config/aarch64/aarch64-builtins.c
1879 (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
1881 2018-08-15 Ilya Leoshkevich <iii@linux.ibm.com>
1883 * config/s390/s390.c (s390_reorg): Remove loop.
1885 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
1888 (darwin_function_switched_text_sections): Delete.
1889 * gcc/config/darwin.h
1890 (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
1892 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
1895 * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
1896 DEBUG_RNGLISTS_SECTION) new macros. (DEBUG_PUBNAMES_SECTION,
1897 DEBUG_PUBTYPES_SECTION) update to include GNU variant.
1899 2018-08-15 Martin Liska <mliska@suse.cz>
1901 PR tree-optimization/86925
1902 * predict.c (expr_expected_value_1): When taking
1903 later predictor, assign also probability.
1904 Use fold_build2_initializer_loc in order to fold
1905 the expression in -frounding-math.
1907 2018-08-14 Allan Sandfeld Jensen <allan.jensen@qt.io>
1909 * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
1911 (expand_vec_perm_1): Try the new method.
1913 2018-08-14 Ilya Leoshkevich <iii@linux.ibm.com>
1916 * lra-lives.c (remove_some_program_points_and_update_live_ranges):
1917 Check whether lra_live_max_point is 0 before dividing.
1919 2018-08-14 Martin Sebor <msebor@redhat.com>
1921 PR tree-optimization/86650
1922 * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
1923 (vrp_prop::check_mem_ref): Same.
1925 2018-08-13 Liu Hao <lh_mouse@126.com>
1927 * pretty-print.c (eat_esc_sequence): Swap the foreground and
1928 background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
1929 and clear it thereafter, as it only works for DBCS.
1931 2018-08-13 Liu Hao <lh_mouse@126.com>
1933 * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
1934 handle returned by _get_osf_handle().
1936 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
1938 * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
1939 for folding vec_perm.
1941 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
1943 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
1944 Add support for gimple-folding of vec_pack() and vec_unpack()
1947 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
1949 * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
1950 vec_xst variants to the list.
1951 (rs6000_gimple_fold_builtin): Add support for folding unaligned
1952 vector loads and stores.
1954 2018-08-13 David Edelsohn <dje.gcc@gmail.com>
1956 * config.gcc (rs6000-ibm-aix4.x): Delete.
1957 (rs6000-ibm-aix5.1): Delete.
1958 (rs6000-ibm-aix5.2): Delete.
1959 (rs6000-ibm-aix5.3): Delete.
1960 * config/rs6000/aix43.h: Delete.
1961 * config/rs6000/aix51.h: Delete.
1962 * config/rs6000/aix52.h: Delete.
1963 * config/rs6000/t-aix43: Delete.
1965 2018-08-13 Ilya Leoshkevich <iii@linux.ibm.com>
1967 * config/s390/s390.c (s390_decompose_constant_pool_ref):
1969 (s390_decompose_address): Factor out constant pool ref
1972 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
1974 * config/nds32/nds32-predicates.c
1975 (nds32_can_use_bclr_p): Change return type as bool.
1976 (nds32_can_use_bset_p): Ditto.
1977 (nds32_can_use_btgl_p): Ditto.
1978 (nds32_can_use_bitci_p): Ditto.
1979 * config/nds32/nds32-protos.h
1980 (nds32_can_use_bclr_p): Change declaration.
1981 (nds32_can_use_bset_p): Ditto.
1982 (nds32_can_use_btgl_p): Ditto.
1983 (nds32_can_use_bitci_p): Ditto.
1985 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
1987 * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
1988 Support -msched-prolog-epilog option.
1989 * config/nds32/nds32.opt (msched-prolog-epilog): New option.
1991 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
1993 * common/config/nds32/nds32-common.c
1994 (nds32_option_optimization_table): Enalbe -malways-align.
1996 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
1998 * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
2000 * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
2001 OPT_misr_secure_ case.
2002 * config/nds32/nds32-isr.c: Implementation of backward compatibility.
2003 * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
2004 * config/nds32/nds32.c (nds32_attribute_table): Add critical and
2006 * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
2007 (nds32_isr_info): New field security_level.
2008 (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
2009 * config/nds32/nds32.md (return_internal): Consider critical attribute.
2010 * config/nds32/nds32.opt (misr-secure): New option.
2011 * config/nds32/nds32_init.inc: New file.
2012 * config/nds32/nds32_isr.h: New file.
2014 2018-08-11 John David Anglin <danglin@gcc.gnu.org>
2016 * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
2017 Update comment for atomic instructions.
2018 (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
2019 atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
2021 (atomic_loaddi): Revise fence expansion to only emit fence prior to
2022 load for __ATOMIC_SEQ_CST model.
2023 (atomic_loaddi_1): Remove float register target.
2024 (atomic_storedi): Handle CONST_INT values.
2025 (atomic_storedi_1): Remove float register source. Add special case
2027 (memory_barrier): New expander and insn.
2029 2018-08-11 Jakub Jelinek <jakub@redhat.com>
2031 PR tree-optimization/86835
2032 * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
2033 new_stmt after def_gsi, make sure to insert new_square_stmt after
2034 that stmt, not 2 stmts before it.
2036 2018-08-10 Alexander Monakov <amonakov@ispras.ru>
2039 * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
2042 2018-08-10 Martin Liska <mliska@suse.cz>
2045 * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
2047 * builtins.c (expand_builtin_expect_with_probability):
2049 (expand_builtin_expect_with_probability): New function.
2050 (build_builtin_expect_predicate): Add new argumnet probability
2051 for BUILT_IN_EXPECT_WITH_PROBABILITY.
2052 (fold_builtin_expect):
2055 * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
2056 * builtins.h (fold_builtin_expect): Set new argument.
2057 * doc/extend.texi: Document __builtin_expect_with_probability.
2058 * doc/invoke.texi: Likewise.
2059 * gimple-fold.c (gimple_fold_call): Pass new argument.
2060 * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
2061 also BUILT_IN_EXPECT_WITH_PROBABILITY.
2062 * predict.c (get_predictor_value): New function.
2063 (expr_expected_value): Add new argument probability. Assume
2064 that predictor and probability are always non-null.
2065 (expr_expected_value_1): Likewise. For __builtin_expect and
2066 __builtin_expect_with_probability set probability. Handle
2067 combination in binary expressions.
2068 (tree_predict_by_opcode): Simplify code by simply calling
2069 get_predictor_value.
2070 (pass_strip_predict_hints::execute): Add handling of
2071 BUILT_IN_EXPECT_WITH_PROBABILITY.
2072 * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
2074 * tree.h (DECL_BUILT_IN_P): New function.
2076 2018-08-10 Martin Liska <mliska@suse.cz>
2078 PR tree-optimization/85799
2079 * passes.def: Add argument for pass_strip_predict_hints.
2080 * predict.c (class pass_strip_predict_hints): Add new argument
2082 (strip_predictor_early): New function.
2083 (pass_strip_predict_hints::execute): Call the function to
2085 (strip_predict_hints): New function.
2086 * predict.def: Fix comment.
2088 2018-08-10 Thomas Preud'homme <thomas.preudhomme@linaro.org>
2090 * Makefile.in: Clarify which tm.texi to copy over to assert the
2091 right to grant a GFDL license for all.
2093 2018-08-09 Jeff Law <law@redhat.com>
2095 * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
2098 2018-08-09 Andreas Schwab <schwab@linux-m68k.org>
2100 * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
2103 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
2105 * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
2106 reductions for variable-length vectors.
2108 2018-08-09 David Malcolm <dmalcolm@redhat.com>
2111 * common.opt (fdiagnostics-show-line-numbers): New option.
2112 * diagnostic-show-locus.c (class layout): Add fields
2113 "m_show_line_numbers_p" and "m_linenum_width";
2114 (num_digits): New function.
2115 (test_num_digits): New function.
2116 (layout::layout): Initialize new fields. Update m_x_offset
2117 logic to handle any left margin.
2118 (layout::print_source_line): Print line number when requested.
2119 (layout::start_annotation_line): New member function.
2120 (layout::print_annotation_line): Call it.
2121 (layout::print_leading_fixits): Likewise.
2122 (layout::print_trailing_fixits): Likewise. Update calls to
2123 move_to_column for new parameter.
2124 (layout::get_x_bound_for_row): Add "add_left_margin" param and use
2125 it to potentially call start_annotation_line.
2126 (layout::show_ruler): Call start_annotation_line.
2127 (selftest::test_line_numbers_multiline_range): New selftest.
2128 (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
2129 and selftest::test_line_numbers_multiline_range.
2130 * diagnostic.c (diagnostic_initialize): Initialize
2131 show_line_numbers_p.
2132 * diagnostic.h (struct diagnostic_context): Add field
2133 "show_line_numbers_p".
2134 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
2135 -fno-diagnostics-show-line-numbers.
2136 * dwarf2out.c (gen_producer_string): Add
2137 OPT_fdiagnostics_show_line_numbers to the ignored options.
2138 * lto-wrapper.c (merge_and_complain): Likewise to the "pick
2139 one setting" options.
2140 (append_compiler_options): Likewise to the dropped options.
2141 (append_diag_options): Likewise to the passed-on options.
2142 * opts.c (common_handle_option): Handle the new option.
2143 * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
2145 2018-08-09 Kelvin Nilsen <kelvin@gcc.gnu.org>
2147 * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
2148 ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub. Add
2149 third argument of type "const signed char" to descriptions of
2150 __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
2151 __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
2152 __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
2153 __builtin_bcdsub_ov functions.
2155 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
2157 PR tree-optimization/86858
2158 * tree-vect-loop.c (vect_is_simple_reduction): Restore
2159 flow_bb_inside_loop_p calls.
2161 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
2163 PR tree-optimization/86871
2164 * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
2165 instead of gimple_assign_lhs.
2167 2018-08-09 Richard Earnshaw <rearnsha@arm.com>
2170 * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
2171 register constraint to operand 0.
2172 (add<mode>3_carryinC): Likewise.
2173 (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
2175 2018-08-09 Martin Liska <mliska@suse.cz>
2178 * common.opt: Remove extra line.
2180 2018-08-09 Martin Liska <mliska@suse.cz>
2182 * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
2183 at the end of a line, make first letter capital and end up
2184 a sentence with a dot.
2185 (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
2186 (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
2187 (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
2188 (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
2189 (PARAM_MAX_ISL_OPERATIONS): Likewise.
2190 (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
2191 (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
2192 (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
2193 (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
2194 (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
2195 (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
2196 (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
2197 (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
2198 (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
2199 (PARAM_TREE_REASSOC_WIDTH): Likewise.
2200 (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
2201 (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
2202 (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
2204 2018-08-09 Andreas Krebbel <krebbel@linux.ibm.com>
2207 * config/s390/s390.c (s390_option_override_internal): Reduce the
2208 stack-clash-protection-probe-interval param if it would be too big
2211 2018-08-08 Andreas Schwab <schwab@linux-m68k.org>
2214 * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
2215 * config/m68k/m68k.c (handle_move_double): Don't call
2216 m68k_final_prescan_insn.
2217 (m68k_adjust_decorated_operand): Renamed from
2218 m68k_final_prescan_insn, remove first and third operand and
2220 (print_operand): Call it.
2221 (print_operand_address): Call it.
2223 2018-08-08 Nathan Sidwell <nathan@acm.org>
2225 * diagnostic.c (diagnostic_report_current_module): Use
2226 linemap_included_from & linemap_included_from_linemap.
2228 2018-08-08 Hongbo Zhang <hongbo.zhang@linaro.org>
2230 * config/aarch64/aarch64-cores.def: Add phecda core.
2231 * config/aarch64/aarch64-tune.md: Regenerate.
2232 * doc/invoke.texi: Add phecda core.
2234 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
2237 * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
2239 * config/s390/s390.md ("movti"): Add more alternatives for
2240 constant to GPR copies.
2242 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
2244 * config/s390/s390.c: Fix whitespace damage throughout the file.
2245 * config/s390/s390.h: Likewise.
2246 * config/s390/tpf.h: Likewise.
2248 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
2250 * config/s390/s390.c (s390_loadrelative_operand_p):
2251 Remove TARGET_CPU_ZARCH usages.
2252 (s390_rtx_costs): Likewise.
2253 (s390_legitimate_constant_p): Likewise.
2254 (s390_cannot_force_const_mem): Likewise.
2255 (legitimate_reload_constant_p): Likewise.
2256 (s390_preferred_reload_class): Likewise.
2257 (legitimize_pic_address): Likewise.
2258 (legitimize_tls_address): Likewise.
2259 (s390_split_branches): Removed.
2260 (s390_add_execute): Removed.
2261 (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
2262 (s390_mainpool_start): Likewise.
2263 (s390_mainpool_finish): Likewise.
2264 (s390_mainpool_cancel): Removed.
2265 (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
2266 (s390_chunkify_cancel): Likewise.
2267 (s390_return_addr_rtx): Likewise.
2268 (s390_register_info): Remove split_branches_pending_p uages.
2269 (s390_optimize_register_info): Likewise.
2270 (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
2271 split_branches_pending_p usages.
2272 (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
2273 (s390_load_got): Likewise.
2274 (s390_expand_split_stack_prologue): Likewise.
2275 (output_asm_nops): Likewise.
2276 (s390_function_profiler): Likewise.
2277 (s390_emit_call): Likewise.
2278 (s390_conditional_register_usage): Likewise.
2279 (s390_optimize_prologue): Likewise.
2280 (s390_reorg): Remove TARGET_CPU_ZARCH and
2281 split_branches_pending_p usages.
2282 (s390_option_override_internal): Remove TARGET_CPU_ZARCH
2284 (s390_output_indirect_thunk_function): Likewise.
2285 * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
2286 (TARGET_CPU_ZARCH_P): Removed.
2287 (struct machine_function): Remove split_branches_pending_p.
2288 * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
2290 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
2292 * common/config/s390/s390-common.c (processor_flags_table):
2294 * config.gcc: Remove with_arch/with_tune support.
2295 * config/s390/2064.md: Remove cpu attribute comparisons.
2296 * config/s390/driver-native.c (s390_host_detect_local_cpu):
2298 * config/s390/linux.h (ASM_SPEC):
2299 Remove -march support.
2300 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
2301 Use a table to get an arch level.
2302 * config/s390/s390-opts.h (enum processor_type):
2304 * config/s390/s390.c
2305 (processor_table): Remove entries, add arch_level values.
2306 (s390_issue_rate): Remove cases.
2307 (s390_option_override): Adjust
2308 s390_option_override_internal() call.
2309 (s390_option_override_internal): Remove deprecation warning.
2310 (s390_valid_target_attribute_tree): Adjust
2311 s390_option_override_internal() call.
2312 * config/s390/s390.h (struct s390_processor):
2313 Share with s390-c.c, add arch_level field.
2314 * config/s390/s390.md:
2315 Remove occurrences in cpu attribute.
2316 * config/s390/s390.opt: Remove -march/-mtune support.
2317 * config/s390/tpf.h (ASM_SPEC): Remove -march support.
2318 * doc/invoke.texi: Remove deprecation warning.
2320 2018-08-08 Luis Machado <luis.machado@linaro.org>
2322 * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
2324 (qdf24xx_tunings): Set vector cost structure to
2325 qdf24xx_vector_cost.
2327 * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
2328 <register_sextend>: Set to 3.
2330 2018-08-07 Richard Sandiford <richard.sandiford@arm.com>
2333 * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
2334 * config/aarch64/aarch64-simd.md
2335 (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
2336 (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
2337 (aarch64_frecpx<mode>): ...this new pattern.
2338 * config/aarch64/aarch64-simd-builtins.def: Remove comment
2339 about aarch64_frecp<FRECP:frecp_suffix><mode>.
2341 2018-08-07 Martin Liska <mliska@suse.cz>
2344 * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
2345 BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
2346 * predict.def (PRED_MALLOC_NONNULL): New predictor.
2347 * doc/extend.texi: Document that malloc attribute adds
2350 2018-08-06 John David Anglin <danglin@gcc.gnu.org>
2353 * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2354 Define to speculation_safe_value_not_needed.
2356 2018-08-06 Jeff Law <law@redhat.com>
2358 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
2359 the vr_values instance to cprop_into_stmt.
2360 (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
2361 (cprop_operand): Also query EVRP to determine if OP is a constant.
2363 2018-08-06 Nathan Sidwell <nathan@acm.org>
2365 * diagnostic.c (diagnostic_report_current_module): Reroll
2366 included-at loop. Translate text.
2368 2018-08-06 David Malcolm <dmalcolm@redhat.com>
2370 * function-tests.c (selftest::test_expansion_to_rtl): Call
2371 free_after_compilation.
2373 2018-08-06 Alan Hayward <alan.hayward@arm.com>
2375 * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
2377 2018-08-06 Andreas Krebbel <krebbel@linux.ibm.com>
2379 * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
2380 loops with memory block operations from getting unrolled.
2382 2018-08-06 Ulrich Weigand <uweigand@de.ibm.com>
2385 * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2386 Define to speculation_safe_value_not_needed.
2388 2018-08-06 Jeff Law <law@redhat.com>
2390 * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
2393 2018-08-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2396 * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
2397 with all enabled __intN types.
2399 * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
2401 2018-08-06 Alan Hayward <alan.hayward@arm.com>
2403 * alias.c (record_set): Check for clobber high.
2404 * cfgexpand.c (expand_gimple_stmt): Likewise.
2405 * combine-stack-adj.c (single_set_for_csa): Likewise.
2406 * combine.c (find_single_use_1): Likewise.
2407 (set_nonzero_bits_and_sign_copies): Likewise.
2408 (get_combine_src_dest): Likewise.
2409 (is_parallel_of_n_reg_sets): Likewise.
2410 (try_combine): Likewise.
2411 (record_dead_and_set_regs_1): Likewise.
2412 (reg_dead_at_p_1): Likewise.
2413 (reg_dead_at_p): Likewise.
2414 * dce.c (deletable_insn_p): Likewise.
2415 (mark_nonreg_stores_1): Likewise.
2416 (mark_nonreg_stores_2): Likewise.
2417 * df-scan.c (df_find_hard_reg_defs): Likewise.
2418 (df_uses_record): Likewise.
2419 (df_get_call_refs): Likewise.
2420 * dwarf2out.c (mem_loc_descriptor): Likewise.
2421 * haifa-sched.c (haifa_classify_rtx): Likewise.
2422 * ira-build.c (create_insn_allocnos): Likewise.
2423 * ira-costs.c (scan_one_insn): Likewise.
2424 * ira.c (equiv_init_movable_p): Likewise.
2425 (rtx_moveable_p): Likewise.
2426 (interesting_dest_for_shprep): Likewise.
2427 * jump.c (mark_jump_label_1): Likewise.
2428 * postreload-gcse.c (record_opr_changes): Likewise.
2429 * postreload.c (reload_cse_simplify): Likewise.
2430 (struct reg_use): Add source expr.
2431 (reload_combine): Check for clobber high.
2432 (reload_combine_note_use): Likewise.
2433 (reload_cse_move2add): Likewise.
2434 (move2add_note_store): Likewise.
2435 * print-rtl.c (print_pattern): Likewise.
2436 * recog.c (decode_asm_operands): Likewise.
2437 (store_data_bypass_p): Likewise.
2438 (if_test_bypass_p): Likewise.
2439 * regcprop.c (kill_clobbered_value): Likewise.
2440 (kill_set_value): Likewise.
2441 * reginfo.c (reg_scan_mark_refs): Likewise.
2442 * reload1.c (maybe_fix_stack_asms): Likewise.
2443 (eliminate_regs_1): Likewise.
2444 (elimination_effects): Likewise.
2445 (mark_not_eliminable): Likewise.
2446 (scan_paradoxical_subregs): Likewise.
2447 (forget_old_reloads_1): Likewise.
2448 * reorg.c (find_end_label): Likewise.
2449 (try_merge_delay_insns): Likewise.
2450 (redundant_insn): Likewise.
2451 (own_thread_p): Likewise.
2452 (fill_simple_delay_slots): Likewise.
2453 (fill_slots_from_thread): Likewise.
2454 (dbr_schedule): Likewise.
2455 * resource.c (update_live_status): Likewise.
2456 (mark_referenced_resources): Likewise.
2457 (mark_set_resources): Likewise.
2458 * rtl.c (copy_rtx): Likewise.
2459 * rtlanal.c (reg_referenced_p): Likewise.
2460 (single_set_2): Likewise.
2461 (noop_move_p): Likewise.
2462 (note_stores): Likewise.
2463 * sched-deps.c (sched_analyze_reg): Likewise.
2464 (sched_analyze_insn): Likewise.
2466 2018-08-06 Alan Hayward <alan.hayward@arm.com>
2468 * cse.c (invalidate_reg): New function extracted from...
2469 (invalidate): ...here.
2470 (canonicalize_insn): Check for clobber high.
2471 (invalidate_from_clobbers): invalidate clobber highs.
2472 (invalidate_from_sets_and_clobbers): Likewise.
2473 (count_reg_usage): Check for clobber high.
2474 (insn_live_p): Likewise.
2475 * cselib.c (cselib_expand_value_rtx_1):Likewise.
2476 (cselib_invalidate_regno): Check for clobber in setter.
2477 (cselib_invalidate_rtx): Pass through setter.
2478 (cselib_invalidate_rtx_note_stores):
2479 (cselib_process_insn): Check for clobber high.
2480 * cselib.h (cselib_invalidate_rtx): Add operand.
2482 2018-08-06 Alan Hayward <alan.hayward@arm.com>
2484 * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
2485 (mark_not_eliminable): Likewise.
2486 * lra-int.h (struct lra_insn_reg): Add clobber high marker.
2487 * lra-lives.c (process_bb_lives): Check for clobber high.
2488 * lra.c (new_insn_reg): Remember clobber highs.
2489 (collect_non_operand_hard_regs): Check for clobber high.
2490 (lra_set_insn_recog_data): Likewise.
2491 (add_regs_to_insn_regno_info): Likewise.
2492 (lra_update_insn_regno_info): Likewise.
2494 2018-08-06 Alan Hayward <alan.hayward@arm.com>
2496 * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
2497 * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
2499 2018-08-06 Alan Hayward <alan.hayward@arm.com>
2501 * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
2502 (copy_insn_1): Likewise.
2503 (gen_hard_reg_clobber_high): New gen function.
2504 * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
2505 * genemit.c (gen_exp): Likewise.
2506 (gen_emit_seq): Pass through info.
2507 (gen_insn): Check for CLOBBER_HIGH.
2508 (gen_expand): Pass through info.
2509 (gen_split): Likewise.
2510 (output_add_clobbers): Likewise.
2511 * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
2512 (remove_clobbers): Likewise.
2513 * rtl.h (gen_hard_reg_clobber_high): New declaration.
2515 2018-08-06 Alan Hayward <alan.hayward@arm.com>
2517 * doc/rtl.texi (clobber_high): Add.
2518 (parallel): Add in clobber high
2519 * rtl.c (rtl_check_failed_code3): Add function.
2520 * rtl.def (CLOBBER_HIGH): Add expression.
2521 * rtl.h (RTL_CHECKC3): Add macro.
2522 (rtl_check_failed_code3): Add declaration.
2523 (XC3EXP): Add macro.
2525 2018-08-05 H.J. Lu <hongjiu.lu@intel.com>
2528 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
2529 cfun->machine->max_used_stack_alignment if needed.
2531 2018-08-04 Martin Sebor <msebor@redhat.com>
2533 PR tree-optimization/86571
2534 * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
2537 2018-08-03 Sandra Loosemore <sandra@codesourcery.com>
2540 * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2543 2018-08-03 Jeff Law <law@redhat.com>
2546 * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2547 Define to speculation_safe_value_not_needed.
2549 2018-08-03 David Malcolm <dmalcolm@redhat.com>
2551 * doc/gcov.texi (-x): Remove duplicate "to".
2552 * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
2553 (-Wif-not-aligned): Remove duplicate "is".
2554 (-flto): Remove duplicate "the".
2555 (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
2556 duplicate "v5.00.b".
2557 (MSP430 Options): Remove duplicate "and" from the description
2558 of "-mgprel-sec=regexp".
2559 (x86 Options): Remove duplicate copies of "vmldLog102" and
2560 vmlsLog104 from description of "-mveclibabi=type".
2562 2018-08-03 Richard Sandiford <richard.sandiford@arm.com>
2564 * internal-fn.h (first_commutative_argument): Declare.
2565 * internal-fn.c (first_commutative_argument): New function.
2566 * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
2567 restrictions for pattern statements. Use first_commutative_argument
2568 to look for commutative operands in calls to internal functions.
2570 2018-08-03 Aldy Hernandez <aldyh@redhat.com>
2572 * Makefile.in (wide-int-range.o): New.
2573 * tree-vrp.c: Move all the wide_int_* functions to...
2574 * wide-int-range.cc: ...here.
2575 * tree-vrp.h: Move all the wide_int_* prototypes to...
2576 * wide-int-range.h: ...here.
2578 2018-08-03 Tom de Vries <tdevries@suse.de>
2580 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
2582 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
2583 * except.c (output_function_exception_table): Do early exit if
2584 targetm_common.except_unwind_info (&global_options) == UI_NONE.
2586 2018-08-03 Martin Liska <mliska@suse.cz>
2588 * predict.c (dump_prediction): Change to 2 digits
2591 2018-08-03 Siddhesh Poyarekar <siddhesh@sourceware.org>
2593 * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
2595 (falkor_am_1_gtov_gtov): ... a new insn reservation.
2597 2018-07-19 Ilya Leoshkevich <iii@linux.ibm.com>
2599 * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
2600 * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
2601 * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
2602 * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
2603 * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
2605 2018-08-02 David Malcolm <dmalcolm@redhat.com>
2607 * diagnostic-show-locus.c (diagnostic_show_locus): Use
2608 pp_take_prefix when saving the existing prefix.
2609 * diagnostic.c (diagnostic_append_note): Likewise.
2610 * langhooks.c (lhd_print_error_function): Likewise.
2611 * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
2612 param's type. Free the existing prefix.
2613 (pp_take_prefix): New function.
2614 (pretty_printer::pretty_printer): Drop the prefix parameter.
2615 Rename the length parameter to match the comment.
2616 (pretty_printer::~pretty_printer): Free the prefix.
2617 * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
2619 (struct pretty_printer): Drop the "const" from "prefix" field's
2620 type and clarify memory management.
2621 (pp_set_prefix): Drop the "const" from the 2nd param.
2622 (pp_take_prefix): New decl.
2624 2018-08-02 Aaron Sawdey <acsawdey@linux.ibm.com>
2626 * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
2627 for word_mode_ok here instead of passing as argument.
2628 (expand_block_compare): Change select_block_compare_mode() call.
2629 (expand_strncmp_gpr_sequence): New function.
2630 (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
2632 2018-08-02 Jeff Law <law@redhat.com>
2635 * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2636 Define to speculation_safe_value_not_needed.
2639 * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2640 Define to speculation_safe_value_not_needed.
2642 2018-08-02 Tom de Vries <tdevries@suse.de>
2645 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
2646 function. Return UI_TARGET unconditionally.
2647 (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
2648 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
2650 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
2652 * genemit.c (print_overload_test): Fix typo.
2654 2018-08-02 Richard Biener <rguenther@suse.de>
2656 PR tree-optimization/86816
2657 * tree-ssa-tail-merge.c (tail_merge_valueize): New function
2658 which checks for value availability before querying it.
2659 (gvn_uses_equal): Use it.
2660 (same_succ_hash): Likewise.
2661 (gimple_equal_p): Likewise.
2663 2018-08-02 Nick Clifton <nickc@redhat.com>
2666 * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2667 Define to speculation_safe_value_not_needed.
2670 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2671 Define to speculation_safe_value_not_needed.
2674 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2675 Define to speculation_safe_value_not_needed.
2678 * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2679 Define to speculation_safe_value_not_needed.
2682 * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2683 Define to speculation_safe_value_not_needed.
2686 * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2687 Define to speculation_safe_value_not_needed.
2690 * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2691 Define to speculation_safe_value_not_needed.
2694 * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2695 Define to speculation_safe_value_not_needed.
2698 * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
2699 speculation_safe_value_not_needed.
2702 * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
2703 to speculation_safe_value_not_needed.
2705 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
2707 * doc/md.texi: Expand the documentation of instruction names
2708 to mention port-local uses. Document '@' in pattern names.
2709 * read-md.h (overloaded_instance, overloaded_name): New structs.
2711 (md_reader::handle_overloaded_name): New member function.
2712 (md_reader::get_overloads): Likewise.
2713 (md_reader::m_first_overload): New member variable.
2714 (md_reader::m_next_overload_ptr): Likewise.
2715 (md_reader::m_overloads_htab): Likewise.
2716 * read-md.c (md_reader::md_reader): Initialize m_first_overload,
2717 m_next_overload_ptr and m_overloads_htab.
2718 * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
2719 (get_mode_token, get_code_token, get_int_token): New functions.
2720 (map_attr_string): Add an optional argument that passes back
2721 the associated iterator.
2722 (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
2723 (md_reader::handle_overloaded_name, add_overload_instance): New
2725 (apply_iterators): Handle '@' names. Report an error if '@'
2726 is used without iterators.
2727 (initialize_iterators): Initialize the new iterator_group fields.
2728 * genopinit.c (handle_overloaded_code_for)
2729 (handle_overloaded_gen): New functions.
2730 (main): Use them to print declarations of maybe_code_for_* and
2731 maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
2732 * genemit.c (print_overload_arguments, print_overload_test)
2733 (handle_overloaded_code_for, handle_overloaded_gen): New functions.
2734 (main): Use it to print definitions of maybe_code_for_* and
2735 maybe_gen_* functions.
2736 * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
2737 gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
2738 instead of explicit mode checks.
2739 (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
2740 (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
2741 (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
2742 (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
2743 (aarch64_expand_compare_and_swap): Likewise
2744 gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
2745 (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
2746 (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
2747 (aarch64_constant_pool_reload_icode): Delete.
2748 (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
2749 instead of aarch64_constant_pool_reload_icode. Use
2750 code_for_aarch64_reload_mov instead of explicit mode checks.
2751 (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
2752 (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
2753 get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
2754 (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
2755 (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
2756 get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
2757 (aarch64_atomic_load_op_code): Delete.
2758 (aarch64_emit_atomic_load_op): Likewise.
2759 (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
2760 aarch64_atomic_load_op_code. Use gen_aarch64_atomic_load
2761 instead of aarch64_emit_atomic_load_op.
2762 * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
2763 (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
2764 (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
2765 (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
2766 character before the pattern name.
2767 * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
2768 (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
2769 (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
2770 (aarch64_frecps<mode>): Likewise.
2771 * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
2772 (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
2773 (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
2774 (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
2775 (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
2777 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
2779 * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
2780 Allow HFmode constants if TARGET_FP_F16INST.
2782 2018-08-02 Jackson Woodruff <jackson.woodruff@arm.com>
2785 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
2786 No longer check last store for clobber of address register.
2788 2018-08-02 Martin Liska <mliska@suse.cz>
2790 PR gcov-profile/86817
2791 * gcov.c (process_all_functions): New function.
2793 (process_file): Move functions processing to
2794 process_all_functions.
2796 2018-08-02 David Malcolm <dmalcolm@redhat.com>
2798 * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
2799 "const" to the "gimple *" and "rtx_insn *" parameters.
2800 * dumpfile.h (dump_user_location_t::dump_user_location_t):
2802 (dump_location_t::dump_location_t): Likewise.
2804 2018-08-01 Martin Sebor <msebor@redhat.com>
2806 PR tree-optimization/86650
2807 * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
2808 rather than a "gcall *". Directly pass the data of interest
2809 to percent_K_format, rather than building a temporary CALL_EXPR
2811 * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
2812 (gimple_fold_builtin_strncat): Adjust.
2813 * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
2814 gcall* argument with gimple*.
2815 * gimple-ssa-warn-restrict.c (check_call): Same.
2816 (wrestrict_dom_walker::before_dom_children): Same.
2817 (builtin_access::builtin_access): Same.
2818 (check_bounds_or_overlap): Same
2819 (maybe_diag_overlap): Same.
2820 (maybe_diag_offset_bounds): Same.
2821 * tree-diagnostic.c (default_tree_printer): Move usage of
2822 EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
2824 * tree-pretty-print.c (percent_K_format): Add argument.
2825 * tree-pretty-print.h: Add argument.
2826 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
2827 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
2828 (maybe_diag_stxncpy_trunc): Same.
2829 (handle_builtin_stxncpy): Same.
2830 (handle_builtin_strcat): Same.
2832 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
2834 * match.pd: Optimise pointer range checks.
2836 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
2838 PR tree-optimization/86758
2839 * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
2840 to remove pattern statements.
2842 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
2844 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
2845 result of dfs_enumerate_from when constructing stmt_vec_infos,
2846 instead of additionally calling get_loop_body.
2848 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
2850 * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
2852 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
2853 When creating an iv, assert that the step is not known to be zero.
2854 (vect_setup_realignment): Update call accordingly.
2855 * tree-vect-stmts.c (vectorizable_store): Likewise.
2856 (vectorizable_load): Likewise. Handle VMAT_INVARIANT separately.
2858 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
2860 * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
2861 * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
2862 (vectorizable_reduction): Likewise.
2863 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
2864 (vect_detect_hybrid_slp_stmts): Likewise.
2865 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
2867 2018-08-01 Aldy Hernandez <aldyh@redhat.com>
2869 * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
2870 (wide_int_set_zero_nonzero_bits): ...this.
2871 (zero_nonzero_bits_from_vr): Rename to...
2872 (vrp_set_zero_nonzero_bits): ...this.
2873 (extract_range_from_multiplicative_op_1): Abstract wide int
2875 (wide_int_range_multiplicative_op): ...here.
2876 (extract_range_from_binary_expr_1): Extract wide int binary
2877 operations into their own functions.
2878 (wide_int_range_lshift): New.
2879 (wide_int_range_can_optimize_bit_op): New.
2880 (wide_int_range_shift_undefined_p): New.
2881 (wide_int_range_bit_xor): New.
2882 (wide_int_range_bit_ior): New.
2883 (wide_int_range_bit_and): New.
2884 (wide_int_range_trunc_mod): New.
2885 (extract_range_into_wide_ints): New.
2886 (vrp_shift_undefined_p): New.
2887 (extract_range_from_multiplicative_op): New.
2888 (vrp_can_optimize_bit_op): New.
2889 * tree-vrp.h (value_range::dump): New.
2890 (wide_int_range_multiplicative_op): New.
2891 (wide_int_range_lshift):New.
2892 (wide_int_range_shift_undefined_p): New.
2893 (wide_int_range_bit_xor): New.
2894 (wide_int_range_bit_ior): New.
2895 (wide_int_range_bit_and): New.
2896 (wide_int_range_trunc_mod): New.
2897 (zero_nonzero_bits_from_bounds): Rename to...
2898 (wide_int_set_zero_nonzero_bits): ...this.
2899 (zero_nonzero_bits_from_vr): Rename to...
2900 (vrp_set_zero_nonzero_bits): ...this.
2901 (range_easy_mask_min_max): Rename to...
2902 (wide_int_range_can_optimize_bit_op): this.
2903 * vr-values.c (simplify_bit_ops_using_ranges): Rename
2904 zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
2906 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
2908 * tree-vectorizer.h (vect_orig_stmt): New function.
2909 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
2910 * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
2911 (vect_create_epilog_for_reduction): Likewise.
2912 (vectorizable_live_operation): Likewise.
2913 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
2914 (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
2915 * tree-vect-stmts.c (vectorizable_call): Likewise.
2916 (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
2918 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
2920 * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
2922 * tree-vect-stmts.c (vect_transform_stmt): Likewise.
2923 * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
2924 (vect_transform_loop): Likewise.
2925 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
2927 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
2929 * tree-vectorizer.h (vect_schedule_slp): Return void.
2930 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
2931 (vect_schedule_slp): Likewise.
2933 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
2935 * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
2937 (vect_transform_loop): Update calls accordingly. Schedule SLP
2938 instances before the main loop, if any exist.
2940 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
2942 PR tree-optimization/86749
2943 * tree-vect-patterns.c (vect_determine_min_output_precision_1):
2944 If the lhs is used in a COND_EXPR, check that it is being used
2945 as the "then" or "else" value.
2947 2018-08-01 Tom de Vries <tdevries@suse.de>
2950 * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
2951 speculation_safe_value_not_needed.
2953 2018-08-01 Richard Biener <rguenther@suse.de>
2955 * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
2958 2018-08-01 Martin Liska <mliska@suse.cz>
2960 * value-prof.c (gimple_divmod_fixed_value_transform): Unify
2961 format how successful transformation is dumped.
2962 (gimple_mod_pow2_value_transform): Likewise.
2963 (gimple_mod_subtract_transform): Likewise.
2964 (gimple_stringops_transform): Likewise.
2966 2018-08-01 Martin Liska <mliska@suse.cz>
2969 * value-prof.c (interesting_stringop_to_profile_p):
2970 Simplify the code and add BUILT_IN_MEMMOVE.
2971 (gimple_stringops_transform): Likewise.
2973 2018-08-01 Sam Tebbs <sam.tebbs@arm.com>
2975 * config/aarch64/aarch64-simd.md
2976 (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
2977 (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
2978 use GPI iterator instead of SI mode.
2980 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
2982 * config/rs6000/rs6000.md (speculation_barrier): Renamed from
2983 rs6000_speculation_barrier.
2984 * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
2985 new barrier pattern name.
2987 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
2989 * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
2990 (speculation_barrier): New insn.
2992 2018-08-01 Richard Biener <rguenther@suse.de>
2995 * graphite.h: Include isl/id.h and isl/space.h to allow build
2998 2018-08-01 Jan Willem Jagersma <jwjagersma@gmail.com>
3001 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
3002 mode for COFF targets.
3003 * defaults.h (TARGET_COFF): Define.
3004 * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
3005 TARGET_COFF): Define.
3006 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
3007 * config/i386/djgpp.c (saved_debug_info_level): New static variable.
3008 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
3010 2018-07-31 Alexandre Oliva <oliva@adacore.com>
3012 * gimple-streamer-in.c (input_bb): Restore BB discriminator.
3013 * gimple-streamer-out.c (output_bb): Save it.
3014 * lto-streamer-in.c (input_struct_function_base): Restore
3015 instance discriminator if available. Create map on demand.
3016 * lto-streamer-out.c (output_struct_function_base): Save it if
3018 * final.c (decl_to_instance_map): Document LTO strategy.
3020 2018-07-31 Alexandre Oliva <oliva@adacore.com>
3021 Olivier Hainque <hainque@adacore.com>
3023 * debug.h (decl_to_instance_map_t): New type.
3024 (decl_to_instance_map): Declare.
3025 (maybe_create_decl_to_instance_map): New inline function.
3026 * final.c (bb_discriminator, last_bb_discriminator): New statics,
3027 to track basic block discriminators.
3028 (final_start_function_1): Initialize them.
3029 (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
3031 (decl_to_instance_map): New variable.
3032 (map_decl_to_instance, maybe_set_discriminator): New functions.
3033 (notice_source_line): Set discriminator.
3035 2018-07-31 Ian Lance Taylor <iant@golang.org>
3037 * targhooks.c (default_have_speculation_safe_value): Add
3040 2018-07-31 David Malcolm <dmalcolm@redhat.com>
3042 * dump-context.h: Include "pretty-print.h".
3043 (dump_context::refresh_dumps_are_enabled): New decl.
3044 (dump_context::emit_item): New decl.
3045 (class dump_context): Add fields "m_test_pp" and
3047 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
3048 (temp_dump_context::get_dumped_text): New decl.
3049 (class temp_dump_context): Add field "m_pp".
3050 * dumpfile.c (refresh_dumps_are_enabled): Convert to...
3051 (dump_context::refresh_dumps_are_enabled): ...and add a test for
3053 (set_dump_file): Update for above change.
3054 (set_alt_dump_file): Likewise.
3055 (dump_loc): New overload, taking a pretty_printer *.
3056 (dump_context::dump_loc): Call end_any_optinfo. Dump the location
3057 to any test pretty-printer.
3058 (make_item_for_dump_gimple_stmt): New function, adapted from
3059 optinfo::add_gimple_stmt.
3060 (dump_context::dump_gimple_stmt): Call it, and use the result,
3061 eliminating the direct usage of dump_file and alt_dump_file in
3062 favor of indirectly using them via emit_item.
3063 (make_item_for_dump_gimple_expr): New function, adapted from
3064 optinfo::add_gimple_expr.
3065 (dump_context::dump_gimple_expr): Call it, and use the result,
3066 eliminating the direct usage of dump_file and alt_dump_file in
3067 favor of indirectly using them via emit_item.
3068 (make_item_for_dump_generic_expr): New function, adapted from
3070 (dump_context::dump_generic_expr): Call it, and use the result,
3071 eliminating the direct usage of dump_file and alt_dump_file in
3072 favor of indirectly using them via emit_item.
3073 (make_item_for_dump_printf_va): New function, adapted from
3074 optinfo::add_printf_va.
3075 (make_item_for_dump_printf): New function.
3076 (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
3077 and use the result, eliminating the direct usage of dump_file and
3078 alt_dump_file in favor of indirectly using them via emit_item.
3079 (make_item_for_dump_dec): New function.
3080 (dump_context::dump_dec): Call it, and use the result,
3081 eliminating the direct usage of dump_file and alt_dump_file in
3082 favor of indirectly using them via emit_item.
3083 (make_item_for_dump_symtab_node): New function, adapted from
3084 optinfo::add_symtab_node.
3085 (dump_context::dump_symtab_node): Call it, and use the result,
3086 eliminating the direct usage of dump_file and alt_dump_file in
3087 favor of indirectly using them via emit_item.
3088 (dump_context::begin_scope): Reimplement, avoiding direct usage
3089 of dump_file and alt_dump_file in favor of indirectly using them
3091 (dump_context::emit_item): New member function.
3092 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
3093 Set up test pretty-printer on the underlying context. Call
3094 refresh_dumps_are_enabled.
3095 (temp_dump_context::~temp_dump_context): Call
3096 refresh_dumps_are_enabled.
3097 (temp_dump_context::get_dumped_text): New member function.
3098 (selftest::verify_dumped_text): New function.
3099 (ASSERT_DUMPED_TEXT_EQ): New macro.
3100 (selftest::test_capture_of_dump_calls): Run all tests twice, with
3101 and then without optinfo enabled. Add uses of
3102 ASSERT_DUMPED_TEXT_EQ to all tests. Add test of nested scopes.
3103 * dumpfile.h: Update comment for the dump_* API.
3104 * optinfo-emit-json.cc
3105 (selftest::test_building_json_from_dump_calls): Update for new
3106 param for temp_dump_context ctor.
3107 * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
3108 and "m_owned" field.
3109 (optinfo_item::~optinfo_item): Likewise.
3110 (optinfo::add_item): New member function.
3111 (optinfo::emit): Update comment.
3112 (optinfo::add_string): Delete.
3113 (optinfo::add_printf): Delete.
3114 (optinfo::add_printf_va): Delete.
3115 (optinfo::add_gimple_stmt): Delete.
3116 (optinfo::add_gimple_expr): Delete.
3117 (optinfo::add_tree): Delete.
3118 (optinfo::add_symtab_node): Delete.
3119 (optinfo::add_dec): Delete.
3120 * optinfo.h (class dump_context): New forward decl.
3121 (optinfo::add_item): New decl.
3122 (optinfo::add_string): Delete.
3123 (optinfo::add_printf): Delete.
3124 (optinfo::add_printf_va): Delete.
3125 (optinfo::add_gimple_stmt): Delete.
3126 (optinfo::add_gimple_expr): Delete.
3127 (optinfo::add_tree): Delete.
3128 (optinfo::add_symtab_node): Delete.
3129 (optinfo::add_dec): Delete.
3130 (optinfo::add_poly_int): Delete.
3131 (optinfo_item::optinfo_item): Remove "owned" param.
3132 (class optinfo_item): Remove field "m_owned".
3134 2018-07-31 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3137 * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
3138 MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
3139 requested variable alignment.
3140 (expand_one_ssa_partition): Likewise.
3141 (expand_one_var): Likewise.
3143 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3145 * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
3146 to speculation_safe_value_not_needed.
3148 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3150 * targhooks.h (speculation_safe_value_not_needed): New prototype.
3151 * targhooks.c (speculation_safe_value_not_needed): New function.
3152 * target.def (have_speculation_safe_value): Update documentation.
3153 * doc/tm.texi: Regenerated.
3155 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3157 * config/aarch64/iterators.md (ALLI_TI): New iterator.
3158 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
3160 (despeculate_copy<ALLI:mode>_insn): New insn.
3161 (despeculate_copyti_insn): New insn.
3162 (despeculate_simple<ALLI:mode>): New insn
3163 (despeculate_simpleti): New insn.
3164 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
3166 (TARGET_SPECULATION_SAFE_VALUE): Redefine to
3167 aarch64_speculation_safe_value.
3168 (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
3170 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3172 * config/aarch64/aarch64-speculation.cc: New file.
3173 * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
3174 before pass_reorder_blocks.
3175 * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
3177 * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
3178 X14 and X15 when tracking speculation.
3179 * config/aarch64/aarch64.md (register name constants): Add
3180 SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
3181 (unspec): Add UNSPEC_SPECULATION_TRACKER.
3182 (speculation_barrier): New insn attribute.
3183 (cmp<mode>): Allow SP in comparisons.
3184 (speculation_tracker): New insn.
3185 (speculation_barrier): Add speculation_barrier attribute.
3186 * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
3187 * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
3188 * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
3190 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3192 * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
3193 aarch64_track_speculation is true.
3194 (tb<optab><mode>1): Likewise.
3195 * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
3196 generate CB[N]Z when tracking speculation.
3197 (aarch64_split_compare_and_swap): Likewise.
3198 (aarch64_split_atomic_op): Likewise.
3200 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3202 * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
3204 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3206 * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
3207 (speculation_barrier): New insn.
3209 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3211 * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
3212 * config/arm/arm.md (speculation_barrier): New expand.
3213 (speculation_barrier_insn): New pattern.
3215 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3217 * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
3218 (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
3219 (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
3220 * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
3222 * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
3223 (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
3224 (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
3225 (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
3226 (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
3227 (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
3228 (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
3229 * builtins.c (expand_speculation_safe_value): New function.
3230 (expand_builtin): Call it.
3231 * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
3232 * doc/extend.texi: Document __builtin_speculation_safe_value.
3233 * doc/md.texi: Document "speculation_barrier" pattern.
3234 * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
3235 TARGET_HAVE_SPECULATION_SAFE_VALUE.
3236 * doc/tm.texi: Regenerated.
3237 * target.def (have_speculation_safe_value, speculation_safe_value): New
3239 * targhooks.c (default_have_speculation_safe_value): New function.
3240 (default_speculation_safe_value): New function.
3241 * targhooks.h (default_have_speculation_safe_value): Add prototype.
3242 (default_speculation_safe_value): Add prototype.
3244 2018-07-31 David Malcolm <dmalcolm@redhat.com>
3246 * dump-context.h (dump_context::dump_loc): New decl.
3247 * dumpfile.c (dump_context::dump_loc): New member function.
3248 (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
3249 and dump_gimple_stmt.
3250 (dump_context::dump_gimple_expr_loc): Likewise, using
3252 (dump_context::dump_generic_expr_loc): Likewise, using
3254 (dump_context::dump_printf_loc_va): Likewise, using
3256 (dump_context::begin_scope): Explicitly using the global function
3257 "dump_loc", rather than the member function.
3259 2018-07-31 Martin Sebor <msebor@redhat.com>
3261 PR tree-optimization/86741
3262 * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
3264 2018-07-31 Andreas Krebbel <krebbel@linux.ibm.com>
3266 * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
3267 depend on whether prefetch instructions will be emitted or not.
3268 Use TARGET_SETMEM_PFD for checking whether prefetch instructions
3269 will be emitted or not.
3270 * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
3271 (TARGET_SETMEM_PFD): New macros.
3273 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3275 * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
3276 (NULL_STMT_VEC_INFO): Delete.
3277 (stmt_vec_info::operator*): Likewise.
3278 (stmt_vec_info::operator gimple *): Likewise.
3279 * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
3280 of NULL_STMT_VEC_INFO.
3281 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
3282 (vect_reassociating_reduction_p): Likewise.
3283 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
3284 (vectorizable_store): Likewise.
3285 * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
3286 (vec_info::free_stmt_vec_infos): Likewise.
3288 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3290 * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
3291 * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
3293 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3295 * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
3296 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
3297 (vec_info::free_stmt_vec_info): New private member functions.
3298 (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
3299 (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
3300 * tree-parloops.c (gather_scalar_reductions): Remove calls to
3301 set_stmt_vec_info_vec and free_stmt_vec_infos.
3302 * tree-vect-loop.c (_loop_vec_info): Remove call to
3303 set_stmt_vec_info_vec.
3304 * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
3305 (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
3306 * tree-vectorizer.c (vec_info::new_stmt_vec_info)
3307 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
3308 (vec_info::free_stmt_vec_info): ...these new functions. Remove
3309 assignments in {vec_info::,}new_stmt_vec_info that are redundant
3310 with the clearing in the xcalloc.
3311 (stmt_vec_info_vec): Delete.
3312 (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
3313 (vectorize_loops): Likewise.
3314 (vec_info::~vec_info): Remove argument from call to
3315 free_stmt_vec_infos.
3316 (vec_info::add_stmt): Remove vinfo argument from call to
3319 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3321 * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
3322 rather than a gimple stmt.
3323 * tree-vect-stmts.c (free_stmt_vec_info): Likewise. Don't free
3324 information for pattern statements when passed the original
3325 statement; instead wait to be passed the pattern statement itself.
3326 Don't call set_vinfo_for_stmt here.
3327 (free_stmt_vec_infos): Update call to free_stmt_vec_info.
3328 * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
3329 stmt_vec_infos here.
3330 * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
3331 * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
3332 stmt_vec_infos entry.
3334 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3336 * tree-vectorizer.h (vec_info::replace_stmt): Declare.
3337 * tree-vectorizer.c (vec_info::replace_stmt): New function.
3338 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
3339 * tree-vect-stmts.c (vectorizable_call): Likewise.
3340 (vectorizable_simd_clone_call): Likewise.
3342 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3344 * tree-vectorizer.h (vec_info::remove_stmt): Declare.
3345 * tree-vectorizer.c (vec_info::remove_stmt): New function.
3346 * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
3347 * tree-vect-loop.c (vect_transform_loop): Likewise.
3348 * tree-vect-slp.c (vect_schedule_slp): Likewise.
3349 * tree-vect-stmts.c (vect_remove_stores): Likewise.
3351 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3353 * tree-vectorizer.h (vec_info::lookup_dr): New member function.
3354 (vect_dr_stmt): Delete.
3355 * tree-vectorizer.c (vec_info::lookup_dr): New function.
3356 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
3358 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
3359 (vect_analyze_data_ref_dependence, vect_record_base_alignments)
3360 (vect_verify_datarefs_alignment, vect_peeling_supportable)
3361 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
3362 (vect_analyze_data_refs): Likewise.
3363 (vect_slp_analyze_data_ref_dependence): Likewise. Take a vec_info
3365 (vect_find_same_alignment_drs): Likewise.
3366 (vect_slp_analyze_node_dependences): Update calls accordingly.
3367 (vect_analyze_data_refs_alignment): Likewise. Use vec_info::lookup_dr
3368 instead of DR_VECT_AUX.
3369 (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
3370 of a vector data references. Use vec_info::lookup_dr instead of
3372 (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
3373 (vect_enhance_data_refs_alignment): Likewise. Use vec_info::lookup_dr
3374 instead of DR_VECT_AUX.
3376 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3378 * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
3380 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
3382 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
3383 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
3384 (vect_gen_prolog_loop_niters): Likewise.
3386 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3388 * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
3389 (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
3390 (vect_known_alignment_in_bytes, vect_dr_behavior)
3391 (vect_get_scalar_dr_size): Take references as dr_vec_infos
3392 instead of data_references. Update calls to other routines for
3393 which the same change has been made.
3394 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
3395 dr_vec_infos instead of stmt_vec_infos.
3396 (vect_analyze_data_ref_dependence): Update call accordingly.
3397 (vect_slp_analyze_data_ref_dependence)
3398 (vect_record_base_alignments): Use DR_VECT_AUX.
3399 (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
3400 (vect_update_misalignment_for_peel, verify_data_ref_alignment)
3401 (vector_alignment_reachable_p, vect_get_data_access_cost)
3402 (vect_peeling_supportable, vect_analyze_group_access_1)
3403 (vect_analyze_group_access, vect_analyze_data_ref_access)
3404 (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
3405 (vect_compile_time_alias, vect_small_gap_p)
3406 (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
3407 (vect_supportable_dr_alignment): Take references as dr_vec_infos
3408 instead of data_references. Update calls to other routines for
3409 which the same change has been made.
3410 (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
3411 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
3412 (vect_slp_analyze_and_verify_node_alignment)
3413 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
3414 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
3415 (vect_setup_realignment): Use dr_vec_infos. Update calls after
3417 (_vect_peel_info::dr): Replace with...
3418 (_vect_peel_info::dr_info): ...this new field.
3419 (vect_peeling_hash_get_most_frequent)
3420 (vect_peeling_hash_choose_best_peeling): Update accordingly.
3421 (vect_peeling_hash_get_lowest_cost):
3422 (vect_enhance_data_refs_alignment): Likewise. Update calls to other
3423 routines for which the same change has been made.
3424 (vect_peeling_hash_insert): Likewise. Take a dr_vec_info instead of a
3426 * tree-vect-loop-manip.c (get_misalign_in_elems)
3427 (vect_gen_prolog_loop_niters): Use dr_vec_infos. Update calls after
3429 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
3430 * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
3431 (vect_truncate_gather_scatter_offset, compare_step_with_zero)
3432 (get_group_load_store_type, get_negative_load_store_type)
3433 (vect_get_data_ptr_increment, vectorizable_store)
3434 (vectorizable_load): Likewise.
3435 (ensure_base_align): Take a dr_vec_info instead of a data_reference.
3436 Update calls to other routines for which the same change has been made.
3438 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3440 * tree-vectorizer.h (vec_info::move_dr): New member function.
3441 (dataref_aux): Rename to...
3442 (dr_vec_info): ...this and add "dr" and "stmt" fields.
3443 (_stmt_vec_info::dr_aux): Update accordingly.
3444 (_stmt_vec_info::data_ref_info): Delete.
3445 (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
3446 (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
3447 (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
3448 (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
3450 (STMT_VINFO_DATA_REF): Likewise. Turn into an lvalue.
3451 (STMT_VINFO_DR_INFO): New macro.
3452 (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
3453 (set_dr_misalignment): Update after rename of dataref_aux.
3454 (vect_dr_stmt): Move earlier in file. Return dr_aux.stmt.
3455 * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
3456 initialization of STMT_VINFO_DATA_REF.
3457 * tree-vectorizer.c (vec_info::move_dr): New function.
3458 * tree-vect-patterns.c (vect_recog_bool_pattern)
3459 (vect_recog_mask_conversion_pattern)
3460 (vect_recog_gather_scatter_pattern): Use it.
3461 * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
3462 the "dr" and "stmt" fields of dr_vec_info instead of
3463 STMT_VINFO_DATA_REF.
3465 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3467 * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
3468 (is_pattern_stmt_p): Use it.
3469 * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
3470 on pattern statements.
3472 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3474 * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
3475 original stmt as a stmt_vec_info rather than a gimple stmt.
3476 (vect_pattern_recog_1): Take the statement directly as a
3477 stmt_vec_info, rather than via a gimple_stmt_iterator.
3478 Update call to vect_mark_pattern_stmts.
3479 (vect_pattern_recog): Update calls accordingly.
3481 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3483 * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
3484 (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
3485 a vect_def_type for the first argument.
3486 * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
3487 (vect_get_vec_def_for_stmt_copy): Likewise. Return the original
3488 operand if it isn't defined by a vectorized statement.
3489 (vect_build_gather_load_calls): Remove the mask_dt argument and
3490 update calls to vect_get_vec_def_for_stmt_copy.
3491 (vectorizable_bswap): Likewise the dt argument.
3492 (vectorizable_call): Update calls to vectorizable_bswap and
3493 vect_get_vec_def_for_stmt_copy.
3494 (vectorizable_simd_clone_call, vectorizable_assignment)
3495 (vectorizable_shift, vectorizable_operation, vectorizable_condition)
3496 (vectorizable_comparison): Update calls to
3497 vect_get_vec_def_for_stmt_copy.
3498 (vectorizable_store): Likewise. Remove now-unnecessary calls to
3500 (vect_get_loop_based_defs): Remove dt argument and update call
3501 to vect_get_vec_def_for_stmt_copy.
3502 (vectorizable_conversion): Update calls to vect_get_loop_based_defs
3503 and vect_get_vec_def_for_stmt_copy.
3504 (vectorizable_load): Update calls to vect_build_gather_load_calls
3505 and vect_get_vec_def_for_stmt_copy.
3506 * tree-vect-loop.c (vect_create_epilog_for_reduction)
3507 (vectorizable_reduction, vectorizable_live_operation): Update calls
3508 to vect_get_vec_def_for_stmt_copy.
3510 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3512 * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
3513 and gimple stmt arguments with a stmt_vec_info.
3514 (vect_record_base_alignments): Update calls accordingly.
3515 * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
3516 and gimple stmt arguments with a stmt_vec_info.
3517 (vect_build_slp_tree_1): Remove vinfo argument and update call
3518 to vect_record_max_nunits.
3519 (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
3520 and vect_record_max_nunits.
3522 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3524 * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
3525 file and take a stmt_vec_info instead of a gimple stmt.
3526 (supportable_widening_operation, vect_finish_replace_stmt)
3527 (vect_finish_stmt_generation, vect_get_store_rhs)
3528 (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
3529 (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
3530 (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
3531 (vect_get_smallest_scalar_type, vect_check_gather_scatter)
3532 (vect_create_data_ref_ptr, bump_vector_ptr)
3533 (vect_permute_store_chain, vect_setup_realignment)
3534 (vect_transform_grouped_load, vect_record_grouped_load_vectors)
3535 (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
3536 (vectorizable_reduction, vectorizable_induction)
3537 (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
3538 (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
3539 than gimple stmts as arguments.
3540 * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
3541 (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
3542 (can_group_stmts_p, vect_check_gather_scatter)
3543 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
3544 (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
3545 (vect_permute_load_chain, vect_shift_permute_load_chain)
3546 (vect_transform_grouped_load)
3547 (vect_record_grouped_load_vectors): Likewise.
3548 * tree-vect-loop.c (vect_fixup_reduc_chain)
3549 (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
3550 (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
3551 (vectorizable_reduction, vectorizable_induction)
3552 (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
3553 * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
3554 (vect_get_load_store_mask): Likewise.
3555 * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
3556 (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
3557 * tree-vect-stmts.c (vect_mark_relevant)
3558 (is_simple_and_all_uses_invariant)
3559 (exist_non_indexing_operands_for_use_p, process_use)
3560 (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
3561 (vect_get_vec_def_for_operand, vect_get_vec_defs)
3562 (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
3563 (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
3564 (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
3565 (get_negative_load_store_type, get_load_store_type)
3566 (vect_check_load_store_mask, vect_check_store_rhs)
3567 (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
3568 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
3569 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
3570 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
3571 (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
3572 (vectorizable_load, vectorizable_condition, vectorizable_comparison)
3573 (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
3574 (supportable_widening_operation): Likewise.
3576 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3578 * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
3579 a stmt_vec_info instead of a gcall.
3580 (vect_check_gather_scatter): Update call accordingly.
3581 * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
3583 (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
3584 (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
3585 * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
3586 instead of a gimple stmt.
3587 (vect_transform_loop): Update calls accordingly.
3588 * tree-vect-slp.c (vect_split_slp_store_group): Take and return
3589 stmt_vec_infos instead of gimple stmts.
3590 (vect_analyze_slp_instance): Update use accordingly.
3591 * tree-vect-stmts.c (read_vector_array, write_vector_array)
3592 (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
3593 (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
3594 (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
3595 (vect_gen_widened_results_half, vect_get_loop_based_defs)
3596 (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
3597 Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
3598 down to subroutines.
3600 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3602 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
3603 of the worklist from a vector of gimple stmts to a vector of
3605 * tree-vect-stmts.c (vect_mark_relevant, process_use)
3606 (vect_mark_stmts_to_be_vectorized): Likewise
3608 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3610 * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
3611 statement before passing it to vect_analyze_stmt.
3612 (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
3613 the chain of phi vector definitions. Track the exit phi via its
3615 (vectorizable_reduction): Set cond_stmt_vinfo directly from the
3616 STMT_VINFO_REDUC_DEF.
3617 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
3618 stmt_vec_infos to handle the statement chains.
3619 (vect_get_slp_defs): Record the first statement in the node
3620 using a stmt_vec_info.
3621 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
3622 statements here and pass their stmt_vec_info down to subroutines.
3623 (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
3624 down to vect_finish_stmt_generation.
3625 (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
3626 (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
3627 stmt_vec_infos to subroutines.
3628 (vect_remove_stores): Use stmt_vec_infos to handle the statement
3631 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3633 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
3634 (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
3635 (vect_permute_store_chain, vect_setup_realignment)
3636 (vect_permute_load_chain, vect_shift_permute_load_chain)
3637 (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
3638 stmts internally, and when passing values to other vectorizer routines.
3639 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
3640 * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
3641 (vect_analyze_loop_operations, get_initial_def_for_reduction)
3642 (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
3643 (vectorizable_reduction, vectorizable_induction)
3644 (vectorizable_live_operation, vect_transform_loop_stmt)
3645 (vect_transform_loop): Likewise.
3646 * tree-vect-patterns.c (vect_reassociating_reduction_p)
3647 (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
3648 (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
3649 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
3650 (vect_slp_analyze_node_operations_1): Likewise.
3651 * tree-vect-stmts.c (vect_mark_relevant, process_use)
3652 (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
3653 (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
3654 (vect_finish_stmt_generation_1, get_group_load_store_type)
3655 (get_load_store_type, vect_build_gather_load_calls)
3656 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
3657 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
3658 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
3659 (vectorizable_store, vectorizable_load, vectorizable_condition)
3660 (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
3661 (supportable_widening_operation): Likewise.
3662 (vect_get_vector_types_for_stmt): Likewise.
3663 * tree-vectorizer.h (vect_dr_behavior): Likewise.
3665 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3667 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
3668 (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
3669 (vect_permute_store_chain, vect_permute_load_chain)
3670 (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
3671 repeated stmt_vec_info lookups.
3672 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
3673 (vect_update_ivs_after_vectorizer): Likewise.
3674 * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
3675 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
3676 * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
3677 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
3678 (vect_bb_slp_scalar_cost): Likewise.
3679 * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
3681 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3683 * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
3684 gcall rather than the generic gimple stmt to gimple_call_internal_fn.
3685 (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
3686 to get gassigns and gcalls, rather than operating on generc gimple
3688 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
3689 (vect_mark_stmts_to_be_vectorized, vectorizable_store)
3690 (vectorizable_load, vect_analyze_stmt): Likewise.
3691 * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
3693 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3695 * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
3696 return stmt_vec_infos rather than gimple stmts. Do not accept
3698 (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
3700 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
3701 Update use of get_later_stmt.
3702 (vect_get_constant_vectors): Update call accordingly.
3703 (vect_schedule_slp_instance): Likewise
3704 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
3705 (vect_slp_analyze_instance_dependence): Likewise.
3706 (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
3708 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3710 * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
3711 (stmt_info_for_cost::stmt_info): ...this new field.
3712 (add_stmt_costs): Update accordingly.
3713 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
3714 (vect_get_known_peeling_cost): Likewise.
3715 (vect_estimate_min_profitable_iters): Likewise.
3716 * tree-vect-stmts.c (record_stmt_cost): Likewise.
3718 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3720 * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
3721 from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
3722 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
3724 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
3726 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3728 * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
3729 a gimple stmt to a stmt_vec_info.
3730 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
3732 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3734 * tree-vectorizer.h (vec_info::grouped_stores): Change from
3735 an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
3736 (_loop_vec_info::reduction_chains): Likewise.
3737 * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
3739 * tree-vect-slp.c (vect_analyze_slp): Likewise.
3741 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3743 * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
3744 a gimple stmt to a stmt_vec_info.
3745 (_stmt_vec_info::next_element): Likewise.
3746 * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
3747 (vect_slp_analyze_and_verify_node_alignment)
3748 (vect_analyze_group_access_1, vect_analyze_group_access)
3749 (vect_small_gap_p, vect_prune_runtime_alias_test_list)
3750 (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
3751 (vect_supportable_dr_alignment): Update accordingly.
3752 * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
3753 (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
3754 (vect_is_simple_reduction, vectorizable_reduction): Likewise.
3755 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
3756 * tree-vect-slp.c (vect_build_slp_tree_1)
3757 (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
3758 (vect_split_slp_store_group, vect_analyze_slp_instance)
3759 (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
3760 * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
3761 (get_group_load_store_type, get_load_store_type)
3762 (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
3763 (vect_transform_stmt, vect_remove_stores): Likewise.
3765 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3767 * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
3769 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
3770 (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
3771 (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
3772 (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
3773 (vector_alignment_reachable_p, vect_get_data_access_cost)
3774 (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
3775 (vect_peeling_supportable, vect_enhance_data_refs_alignment)
3776 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
3777 (vect_analyze_group_access_1, vect_analyze_group_access)
3778 (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
3779 (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
3780 (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
3781 result of vect_dr_stmt and use the stmt_vec_info instead of
3782 the associated gimple stmt.
3783 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
3784 (vect_gen_prolog_loop_niters): Likewise.
3785 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
3787 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3789 * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
3790 to a vec<stmt_vec_info>.
3791 * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
3792 (vect_create_new_slp_node): Take a vec<gimple *> instead of a
3794 (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
3795 to a vec<stmt_vec_info>.
3796 (bst_traits::value_type, bst_traits::value_type): Likewise.
3797 (bst_traits::hash): Update accordingly.
3798 (vect_get_and_check_slp_defs): Change the stmts parameter from
3799 a vec<gimple *> to a vec<stmt_vec_info>.
3800 (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
3801 (vect_build_slp_tree): Likewise.
3802 (vect_build_slp_tree_2): Likewise. Update uses of
3803 SLP_TREE_SCALAR_STMTS.
3804 (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
3805 (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
3806 (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
3807 (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
3808 (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
3809 (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
3810 (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
3811 (vect_get_constant_vectors, vect_get_slp_defs)
3812 (vect_transform_slp_perm_load, vect_schedule_slp_instance)
3813 (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
3814 (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
3815 instead of gimple stmts.
3816 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
3817 the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
3818 (vect_slp_analyze_instance_dependence): Update uses of
3819 SLP_TREE_SCALAR_STMTS.
3820 (vect_slp_analyze_and_verify_node_alignment): Likewise.
3821 (vect_slp_analyze_and_verify_instance_alignment): Likewise.
3822 * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
3823 (get_initial_defs_for_reduction): Likewise.
3824 (vect_create_epilog_for_reduction): Likewise.
3825 (vectorize_fold_left_reduction): Likewise.
3826 * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
3827 (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
3828 (can_vectorize_live_stmts): Likewise.
3830 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3832 * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
3833 auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
3834 (vect_force_simple_reduction): Take and return stmt_vec_infos rather
3836 * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
3838 (gather_scalar_reductions): Update after above interface changes.
3839 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
3840 (vect_is_simple_reduction): Take and return stmt_vec_infos rather
3842 (vect_force_simple_reduction): Likewise.
3843 * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
3844 LOOP_VINFO_REDUCTIONS.
3845 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
3847 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3849 * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
3850 a gimple stmt to a stmt_vec_info.
3851 * tree-vect-loop.c (vect_active_double_reduction_p)
3852 (vect_force_simple_reduction, vectorizable_reduction): Update
3855 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3857 * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
3858 vec<gimple *> to a vec<stmt_vec_info>.
3859 * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
3860 the reduction_phis argument from a vec<gimple *> to a
3862 (vectorizable_reduction): Likewise the phis local variable that
3863 is passed to vect_create_epilog_for_reduction. Update for new type
3864 of SLP_TREE_VEC_STMTS.
3865 (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
3866 (vectorizable_live_operation): Likewise.
3867 * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
3868 (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
3870 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3872 * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
3873 a gimple stmt to a stmt_vec_info.
3874 (vectorizable_condition, vectorizable_live_operation)
3875 (vectorizable_reduction, vectorizable_induction): Pass back the
3876 vectorized statement as a stmt_vec_info.
3877 * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
3878 use of STMT_VINFO_VEC_STMT.
3879 * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
3880 accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
3881 as stmt_vec_infos rather than gimple stmts.
3882 (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
3884 (vectorizable_live_operation): Likewise.
3885 (vectorizable_reduction, vectorizable_induction): Likewise,
3886 updating use of STMT_VINFO_VEC_STMT.
3887 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
3888 of STMT_VINFO_VEC_STMT.
3889 (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
3890 (vectorizable_simd_clone_call, vectorizable_conversion)
3891 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
3892 (vectorizable_store, vectorizable_load, vectorizable_condition)
3893 (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
3894 from a gimple stmt to a stmt_vec_info.
3895 (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT. Pass a
3896 pointer to a stmt_vec_info to the vectorizable_* routines.
3898 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3900 * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
3901 a gimple stmt to a stmt_vec_info.
3902 (is_pattern_stmt_p): Update accordingly.
3903 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
3904 (vect_record_grouped_load_vectors): Likewise.
3905 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
3906 (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
3907 (vect_model_reduction_cost): Likewise.
3908 (vect_create_epilog_for_reduction): Likewise.
3909 (vectorizable_reduction, vectorizable_induction): Likewise.
3910 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
3911 Return the stmt_vec_info for the pattern statement.
3912 (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
3913 (vect_split_statement, vect_mark_pattern_stmts): Likewise.
3914 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
3915 (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
3916 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
3917 (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
3918 (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
3919 (free_stmt_vec_info, vect_is_simple_use): Likewise.
3921 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3923 * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
3924 (vect_finish_stmt_generation): Likewise.
3925 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
3926 (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
3927 (vect_build_gather_load_calls): Use the return value of the above
3928 functions instead of a separate call to vinfo_for_stmt. Use narrow
3929 scopes for the input gimple stmt and wider scopes for the associated
3930 stmt_vec_info. Use vec_info::lookup_def when setting these
3931 stmt_vec_infos from an SSA_NAME definition.
3932 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
3933 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
3934 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
3935 (vectorizable_store, vectorizable_load, vectorizable_condition)
3936 (vectorizable_comparison): Likewise.
3937 * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
3938 (vectorizable_reduction): Likewise.
3940 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3942 * tree-vectorizer.h (vect_is_simple_use): Add an optional
3943 stmt_vec_info * parameter before the optional gimple **.
3944 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
3945 (process_use, vect_get_vec_def_for_operand_1): Update callers.
3946 (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
3947 * tree-vect-loop.c (vectorizable_reduction): Likewise.
3948 (vectorizable_live_operation): Likewise.
3949 * tree-vect-patterns.c (type_conversion_p): Likewise.
3950 (vect_look_through_possible_promotion): Likewise.
3951 (vect_recog_rotate_pattern): Likewise.
3952 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
3954 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3956 * tree-vectorizer.h (stmt_vec_info): Temporarily change from
3957 a typedef to a wrapper class.
3958 (NULL_STMT_VEC_INFO): New macro.
3959 (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
3960 (stmt_vec_info::operator*): New function.
3961 (stmt_vec_info::operator gimple *): Likewise.
3962 (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
3963 (add_stmt_costs): Likewise.
3964 * tree-vect-loop-manip.c (iv_phi_p): Likewise.
3965 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
3966 (vect_get_known_peeling_cost): Likewise.
3967 (vect_estimate_min_profitable_iters): Likewise.
3968 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
3969 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
3970 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
3971 (vectorizable_store, free_stmt_vec_infos): Likewise.
3972 (new_stmt_vec_info): Change return type of xcalloc to
3975 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3977 * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
3978 * tree-vectorizer.c (vec_info::lookup_single_use): New function.
3979 * tree-vect-loop.c (vectorizable_reduction): Use it instead of
3980 a single_imm_use-based sequence.
3981 * tree-vect-stmts.c (supportable_widening_operation): Likewise.
3983 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3985 * tree-vectorizer.h (vec_info::lookup_def): Declare.
3986 * tree-vectorizer.c (vec_info::lookup_def): New function.
3987 * tree-vect-patterns.c (vect_get_internal_def): Use it.
3988 (vect_widened_op_tree): Likewise.
3989 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
3990 * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
3991 (vectorizable_reduction): Likewise.
3992 (vect_valid_reduction_input_p): Take a stmt_vec_info instead
3994 (vect_is_slp_reduction): Update calls accordingly. Use
3995 vec_info::lookup_def.
3996 (vect_is_simple_reduction): Likewise
3997 * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
3999 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4001 * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
4002 * tree-vectorizer.c (vec_info::lookup_stmt): New function.
4003 * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
4005 (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
4006 (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
4007 (vect_update_vf_for_slp, vect_analyze_loop_operations)
4008 (vect_is_slp_reduction, vectorizable_induction)
4009 (vect_transform_loop_stmt, vect_transform_loop): Likewise.
4010 * tree-vect-patterns.c (vect_init_pattern_stmt):
4011 (vect_determine_min_output_precision_1, vect_determine_precisions)
4012 (vect_pattern_recog): Likewise.
4013 * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
4014 * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
4015 * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
4016 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
4017 (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
4018 (vect_detect_hybrid_slp): Likewise. Change the walk_stmt_info
4019 info field from a loop to a loop_vec_info.
4021 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4023 * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
4024 (vec_info::add_stmt): Declare.
4025 * tree-vectorizer.c (vec_info::add_stmt): New function.
4026 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
4027 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
4028 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
4029 (vectorizable_induction): Likewise.
4030 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
4031 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
4032 (vectorizable_simd_clone_call, vectorizable_store): Likewise.
4033 (vectorizable_load): Likewise.
4034 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
4035 (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
4036 (vect_recog_gather_scatter_pattern): Likewise.
4037 (append_pattern_def_seq): Likewise. Remove a check that is
4038 performed by add_stmt itself.
4040 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4042 * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
4043 which make_ssa_name was called with new_stmt before new_stmt
4046 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4048 * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
4050 (vect_is_slp_reduction): ...here...
4051 (vect_is_simple_reduction): ...and here. Remove repetition of tests
4052 that are already known to be false.
4054 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4056 * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
4057 * tree-vect-slp.c (vect_free_slp_tree): Likewise. Don't update
4058 STMT_VINFO_NUM_SLP_USES when it's true.
4059 (vect_free_slp_instance): Add a final_p parameter and pass it to
4061 (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
4062 (vect_analyze_slp_instance): Likewise.
4063 (vect_slp_analyze_operations): Likewise.
4064 (vect_slp_analyze_bb_1): Likewise.
4065 * tree-vectorizer.c (vec_info): Likewise.
4066 * tree-vect-loop.c (vect_transform_loop): Likewise.
4068 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4070 * tree-vect-loop.c (vectorizable_reduction): Assert that the
4071 function is not called for second and subsequent members of
4074 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4076 * tree-vect-loop.c (get_initial_def_for_reduction): Move special
4077 cases for nested loops from here to ...
4078 (vect_create_epilog_for_reduction): ...here. Only call
4079 vect_is_simple_use for inner-loop reductions.
4081 2018-07-31 Martin Liska <mliska@suse.cz>
4083 PR gcov-profile/85338
4084 PR gcov-profile/85350
4085 PR gcov-profile/85372
4086 * profile.c (struct location_triplet): New.
4087 (struct location_triplet_hash): Likewise.
4088 (output_location): Do not output a BB that
4089 is already recorded for a line.
4090 (branch_prob): Use streamed_locations.
4092 2018-07-31 Martin Liska <mliska@suse.cz>
4094 PR gcov-profile/85370
4095 * coverage.c (coverage_begin_function): Do not mark target
4096 clones as artificial functions.
4098 2018-07-31 Martin Liska <mliska@suse.cz>
4100 PR gcov-profile/83813
4101 PR gcov-profile/84758
4102 PR gcov-profile/85217
4103 PR gcov-profile/85332
4104 * profile.c (branch_prob): Do not record GOTO expressions
4105 for GIMPLE statements which locations are already streamed.
4107 2018-07-31 Olivier Hainque <hainque@adacore.com>
4109 * gcc.c (handle_spec_function): Accept a soft_matched_part
4110 argument, as do_spec_1. Pass it down to ...
4111 (eval_spec_function): Accept a soft_matched_part argument,
4112 and pass it down to ...
4113 (do_spec_2): Accept a soft_matched_part argument, and pass
4114 it down to do_spec_1.
4115 (do_spec_1): Pass soft_matched_part to handle_spec_function.
4116 (handle_braces): Update call to handle_spec_function.
4117 (driver::set_up_specs): Update calls to do_spec_2.
4118 (compare_debug_dump_opt_spec_function): Likewise.
4119 (compare_debug_self_opt_spec_function): Likewise.
4121 2018-07-31 Olivier Hainque <hainque@adacore.com>
4123 * common.opt (nolibc): New option.
4124 * doc/invoke.texi (Link Options): Document it.
4125 * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
4126 * config/alpha/linux.h: Likewise.
4127 * config/arc/elf.h: Likewise.
4128 * config/arm/uclinux-elf.h: Likewise.
4129 * config/arm/unknown-elf.h: Likewise.
4130 * config/avr/avrlibc.h: Likewise.
4131 * config/bfin/bfin.h: Likewise.
4132 * config/bfin/linux.h: Likewise.
4133 * config/bfin/uclinux.h: Likewise.
4134 * config/darwin.h: Likewise.
4135 * config/darwin10.h: Likewise.
4136 * config/darwin12.h: Likewise.
4137 * config/gnu-user.h: Likewise.
4138 * config/lm32/uclinux-elf.h: Likewise.
4139 * config/pa/pa-hpux11.h: Likewise.
4140 * config/pa/pa64-hpux.h: Likewise.
4141 * config/sparc/sparc.h: Likewise.
4143 2018-07-31 Olivier Hainque <hainque@adacore.com>
4145 * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
4146 undefined variables.
4148 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
4151 * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
4154 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
4156 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
4157 terminated string literal.
4159 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
4161 PR rtl-optimization/85160
4162 * combine.c (is_just_move): New function.
4163 (try_combine): Allow combining two instructions into two if neither of
4164 the original instructions was a move.
4166 2018-07-30 Alexander Monakov <amonakov@ispras.ru>
4169 * doc/extend.texi (Global Register Variables): Discourage use of type
4171 (Local Register Variables): Likewise.
4173 2018-07-30 Richard Sandiford <richard.sandiford@arm.com>
4175 PR tree-optimization/86506
4176 * hwint.h (ceil_log2): Resync with hwint.c implementation.
4178 2018-07-30 Ilya Leoshkevich <iii@linux.ibm.com>
4181 * lra-constraints.c (spill_hard_reg_in_range): When selecting the
4182 hard_regno, make sure no insn between `from` and `to` clobbers it.
4184 2018-07-30 Cesar Philippidis <cesar@codesourcery.com>
4185 Tom de Vries <tdevries@suse.de>
4187 * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
4188 (PTX_DEFAULT_RUNTIME_DIM): ... this.
4189 (nvptx_goacc_validate_dims): Set default worker and gang dims to
4190 PTX_DEFAULT_RUNTIME_DIM.
4191 (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
4193 2018-07-29 John David Anglin <danglin@gcc.gnu.org>
4195 * config/pa/pa.c (pa_output_addr_vec): Align address table.
4196 * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
4197 * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
4199 2018-07-27 Michael Meissner <meissner@linux.ibm.com>
4201 * config/rs6000/constraints.md (wG constraint): Delete, no longer
4203 * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
4204 predicate to reflect toc fusion has been deleted.
4205 (toc_fusion_mem_raw): Delete, no longer used.
4206 (toc_fusion_mem_wrapped): Likewise.
4207 * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
4209 * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
4210 Delete, no longer used.
4211 * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
4212 meant to be used for toc fusion.
4213 (rs6000_debug_print_mode): Delete toc fusion debugging.
4214 (rs6000_debug_reg_global): Likewise.
4215 (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
4216 fusion and secondary reload support that were never used.
4217 (rs6000_option_override_internal): Delete TOC fusion, that was only
4218 partially defined, and it did not work unless you also used the
4220 (rs6000_legitimate_address_p): Delete TOC fusion support.
4221 (rs6000_opt_masks): Likewise.
4222 (fusion_wrap_memory_address): Delete function, no longer used.
4223 (fusion_split_address); Delete TOC fusion support.
4224 * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
4225 longer used with toc fusion being deleted.
4226 (TARGET_TOC_FUSION_FP): Likewise.
4227 * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
4229 (toc fusion spliter): Delete TOC fusion support.
4230 (toc_fusionload_<mode>): Likewise.
4231 (toc_fusionload_di): Likewise.
4232 (fusion_gpr_load_<mode>): Delete generator function, this insn no
4233 longer needs to be named. Rename predicate to delete TOC fusion.
4234 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
4235 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
4236 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
4237 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
4238 (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
4240 2018-07-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
4242 * doc/extend.texi (Basic PowerPC Built-in Functions Available on
4243 ISA 2.05): Replace __uint128_t with __uint128 and __int128_t with
4244 __int128 in built-in function prototypes.
4245 (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
4246 (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
4248 2018-07-27 Martin Sebor <msebor@redhat.com>
4250 PR tree-optimization/86696
4251 * tree-ssa-strlen.c (get_min_string_length): Handle all integer
4252 types, including enums.
4253 (handle_char_store): Be prepared for the above function to fail.
4255 2018-07-26 Qing Zhao <qing.zhao@oracle.com>
4257 * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
4258 when optimization level is lower than 2 or optimize for size.
4260 2018-07-26 Martin Sebor <msebor@redhat.com>
4262 PR tree-optimization/86043
4263 PR tree-optimization/86042
4264 * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
4265 (get_string_cst_length): Rename...
4266 (get_min_string_length): ...to this. Add argument.
4267 (handle_char_store): Extend to handle multi-character stores by
4269 * tree.c (initializer_zerop): Use new argument. Handle MEM_REF.
4270 * tree.h (initializer_zerop): Add argument.
4272 2018-07-26 Jakub Jelinek <jakub@redhat.com>
4275 * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
4276 declare target to variables if they have always,{to,from,tofrom} map
4279 2018-07-26 Martin Liska <mliska@suse.cz>
4282 * lto-wrapper.c: Add linker_output as prefix
4283 for ltrans_output_file.
4285 2018-07-26 Segher Boessenkool <segher@kernel.crashing.org>
4287 PR rtl-optimization/85805
4288 * combine.c (reg_nonzero_bits_for_combine): Only use the last set
4289 value for hard registers if that was written in the same mode.
4291 2018-07-26 Martin Liska <mliska@suse.cz>
4293 PR gcov-profile/86536
4294 * gcov.c (format_gcov): Use printf format %.*f directly
4295 and do not handle special values.
4297 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
4299 * common/config/arc/arc-common.c (arc_option_optimization_table):
4300 Update default optimizations for size.
4302 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
4304 * config/arc/arc.md (movsf_insn): Add short instruction selection.
4305 * config/arc/constraints.md (CfZ): New constraint.
4306 * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
4307 (subsf3_fpu): Likewise.
4308 (cmpsf_fpu): Likewise.
4309 (cmpsf_fpu_uneq): Likewise.
4311 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
4313 * config/arc/arc.c (compact_memory_operand_p): Check for uncached
4315 (arc_is_uncached_mem_p): uncached applies to both the variable and
4318 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
4320 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
4323 2018-07-25 David Malcolm <dmalcolm@redhat.com>
4325 * optinfo-emit-json.cc (class optrecord_json_writer): Convert
4326 field "m_scopes" from vec to auto_vec.
4328 2018-07-25 Martin Liska <mliska@suse.cz>
4330 * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
4333 2018-07-25 Richard Biener <rguenther@suse.de>
4336 * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
4337 special wrt context_die late.
4338 (gen_subprogram_die): Re-use DIEs in local scope.
4340 2018-07-25 Richard Sandiford <richard.sandiford@arm.com>
4342 PR tree-optimization/86644
4343 * hwint.c (ceil_log2): Fix comment. Return 0 for 0.
4345 2018-07-25 Martin Liska <mliska@suse.cz>
4348 * dumpfile.c: And excluded values with TDF_ALL_VALUES.
4349 * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
4351 2018-07-25 Martin Liska <mliska@suse.cz>
4354 * params.def: Explain ASan abbreviation and provide
4355 a documentation link.
4357 2018-07-24 Martin Sebor <msebor@redhat.com>
4359 PR tree-optimization/86622
4360 PR tree-optimization/86532
4361 * builtins.h (string_length): Declare.
4362 * builtins.c (c_strlen): Correct handling of non-constant offsets.
4363 (check_access): Be prepared for non-constant length ranges.
4364 (string_length): Make extern.
4365 * expr.c (string_constant): Only handle the minor non-constant
4366 array index. Use string_constant to compute the length of
4367 a generic string constant.
4369 2018-07-24 Richard Sandiford <richard.sandiford@arm.com>
4371 PR tree-optimization/86618
4372 * tree-vect-stmts.c (vectorizable_call): Don't take the address
4373 of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
4375 2018-07-24 David Malcolm <dmalcolm@redhat.com>
4377 PR tree-optimization/86636
4378 * json.cc (json::object::set): Fix comment. Add assertions.
4379 (json::array::append): Move here from json.h. Add comment and an
4381 (json::string::string): Likewise.
4382 * json.h (json::array::append): Move to json.cc.
4383 (json::string::string): Likewise.
4384 * optinfo-emit-json.cc
4385 (optrecord_json_writer::impl_location_to_json): Assert that we
4386 aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
4387 wrapper around it. Expand the location once, rather than three
4389 (optrecord_json_writer::inlining_chain_to_json): Fix the check for
4390 UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
4392 (optrecord_json_writer::optinfo_to_json): Likewise, in four
4393 places. Fix some overlong lines.
4395 2018-07-24 Matthew Malcomson <matthew.malcomson@arm.com>
4397 * config/aarch64/aarch64-simd.md
4398 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
4399 (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
4400 (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
4401 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
4402 (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
4403 (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
4404 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
4405 (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
4406 (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
4408 2018-07-24 Jakub Jelinek <jakub@redhat.com>
4411 * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
4412 and size > HOST_BITS_PER_WIDE_INT. For size > HOST_BITS_PER_WIDE_INT
4413 and abs_d == d, do the power of two handling if profitable.
4415 2018-07-24 Richard Biener <rguenther@suse.de>
4417 * match.pd: Add BIT_FIELD_REF canonicalizations.
4419 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
4422 * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
4424 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
4426 * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
4427 terminated STRING_CST object.
4429 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
4431 hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
4433 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
4435 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
4436 * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
4437 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
4438 the elements into a register.
4439 (rs6000_split_v4si_init_di_reg): Delete.
4440 (rs6000_split_v4si_init): Delete.
4441 * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
4442 (vsx_init_v4si): Rewrite as a define_expand.
4444 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
4446 * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
4447 zero_extend argument from memory): New.
4449 2018-07-22 Martin Sebor <msebor@redhat.com>
4452 * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
4453 diagnosing calls with unknown arguments unless -Walloca-larger-than
4454 is restricted to less than PTRDIFF_MAX bytes.
4456 2018-07-22 Gerald Pfeifer <gerald@pfeifer.com>
4458 * doc/gcov.texi (Invoking Gcov): Editorial changes.
4460 2018-07-20 David Malcolm <dmalcolm@redhat.com>
4462 * pretty-print.c (text_info::set_location): Remove redundant
4463 "line_table" parameter from call to rich_location::set_range.
4465 2018-07-20 Martin Sebor <msebor@redhat.com>
4468 * builtins.c (expand_builtin_alloca): Adjust.
4469 * calls.c (alloc_max_size): Simplify.
4470 * cgraphunit.c (cgraph_node::expand): Adjust.
4471 * common.opt (larger_than_size, warn_frame_larger_than): Remove
4473 (frame_larger_than_size): Same.
4474 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
4475 to take a HOST_WIDE_INT argument and accept a byte-size suffix.
4477 * doc/invoke.texi (GCC Command Options): Document option arguments.
4478 Explain byte-size arguments and suffixes.
4479 (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
4480 (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
4481 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
4482 * doc/options.texi (UInteger): Expand.
4483 (Host_Wide_Int, ByteSize): Document new properties.
4484 * final.c (final_start_function_1): Include sizes in an error message.
4485 * function.c (frame_offset_overflow): Same.
4486 * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
4487 (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
4488 Diagnose unbounded alloca calls only for limits of less than
4490 (alloca_call_type): Adjust. Diagnose possibly out-of-bounds alloca
4491 calls and VLA size only for limits of less than PTRDIFF_MAX. Same
4493 (pass_walloca::execute): Adjust. Diagnose alloca calls in loops
4494 only for limits of less than PTRDIFF_MAX.
4495 * langhooks-def.h (lhd_handle_option): Change function argument
4497 * langhooks.c (lhd_handle_option): Same.
4498 * langhooks.h (handle_option): Same.
4499 * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
4501 (var_type, var_type_struct): Same.
4502 (var_set): Handle ByteSize flag.
4503 * optc-gen.awk: Add comments to output to ease debugging. Make
4504 use of HOST_WIDE_INT where appropriate.
4505 * opts-gen-save.awk: Use %lx to format unsigned long.
4506 * opth-gen.awk: Change function argument to HOST_WIDE_INT.
4507 * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
4508 arguments. Parse bytes-size suffixes.
4509 (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
4510 (enum_value_to_arg): Same.
4511 (decode_cmdline_option): Handle cl_host_wide_int. Adjust.
4512 (handle_option): Adjust.
4513 (generate_option): Change function argument to HOST_WIDE_INT.
4514 (cmdline_handle_error): Adjust.
4515 (read_cmdline_option): Change function argument to HOST_WIDE_INT.
4516 (set_option): Change function argument to HOST_WIDE_INT.
4517 (option_enabled): Handle cl_host_wide_int.
4518 (get_option_state): Handle CLVC_SIZE.
4519 (control_warning_option): Same.
4520 * opts.c (common_handle_option): Change function argument to
4521 HOST_WIDE_INT. Remove handling of OPT_Walloca_larger_than_ and
4522 OPT_Wvla_larger_than_.
4523 * opts.h (enum cl_var_type): Add an enumerator.
4524 * stor-layout.c (layout_decl): Print a more meaningful warning.
4525 * toplev.c (output_stack_usage): Adjust.
4527 2018-07-20 Qing Zhao <qing.zhao@oracle.com>
4529 * builtins.c (expand_builtin_memcmp): Delete the last parameter for
4530 call to inline_expand_builtin_string_cmp.
4531 (expand_builtin_strcmp): Likewise.
4532 (expand_builtin_strncmp): Likewise.
4533 (inline_string_cmp): Delete the last parameter, change char_type_node
4534 to unsigned_char_type_node for strcmp/strncmp, add conversions to the
4536 (inline_expand_builtin_string_cmp): Delete the last parameter, give up
4537 the inlining expansion on target where the type of the call has same or
4538 narrower precision than unsigned char.
4540 2018-07-20 David Malcolm <dmalcolm@redhat.com>
4542 * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
4543 (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
4544 * common.opt (fsave-optimization-record): New option.
4545 * coretypes.h (struct kv_pair): Move here from dumpfile.c.
4546 * doc/invoke.texi (-fsave-optimization-record): New option.
4547 * dumpfile.c: Include "optinfo-emit-json.h".
4548 (struct kv_pair): Move to coretypes.h.
4549 (optgroup_options): Make non-static.
4550 (dump_context::end_scope): Call
4551 optimization_records_maybe_pop_dump_scope.
4552 * dumpfile.h (optgroup_options): New decl.
4553 * json.cc: New file.
4555 * optinfo-emit-json.cc: New file.
4556 * optinfo-emit-json.h: New file.
4557 * optinfo.cc: Include "optinfo-emit-json.h".
4558 (optinfo::emit): Call optimization_records_maybe_record_optinfo.
4559 (optinfo_enabled_p): Check optimization_records_enabled_p.
4560 (optinfo_wants_inlining_info_p): Likewise.
4561 * optinfo.h: Update comment.
4562 * profile-count.c (profile_quality_as_string): New function.
4563 * profile-count.h (profile_quality_as_string): New decl.
4564 (profile_count::quality): New accessor.
4565 * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
4566 and optinfo_emit_json_cc_tests.
4567 * selftest.h (selftest::json_cc_tests): New decl.
4568 (selftest::optinfo_emit_json_cc_tests): New decl.
4569 * toplev.c: Include "optinfo-emit-json.h".
4570 (compile_file): Call optimization_records_finish.
4571 (do_compile): Call optimization_records_start.
4572 * tree-ssa-live.c: Include optinfo.h.
4573 (remove_unused_scope_block_p): Retain inlining information if
4574 optinfo_wants_inlining_info_p returns true.
4576 2018-07-20 Richard Biener <rguenther@suse.de>
4579 * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
4580 to cover -flto-partition=none.
4582 2018-07-20 Martin Liska <mliska@suse.cz>
4584 * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
4585 (get_decl_source_range): Remove unused function.
4587 2018-07-20 Richard Biener <rguenther@suse.de>
4589 * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
4590 (struct vn_phi_s): Likewise.
4591 (struct vn_reference_s): Likewise.
4592 * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
4593 for searching the slot of an entry known to be in the hash itself.
4594 (vn_phi_hasher::equal): Likewise.
4595 (vn_reference_hasher::equal): Likewise.
4596 (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
4598 (optimistic_info, current_info): Remove, keeping only valid_info.
4599 (vn_reference_lookup_1): Remove fallback lookup.
4600 (vn_reference_lookup_2): Likewise.
4601 (vn_nary_op_lookup_1): Likewise.
4602 (vn_phi_lookup): Likewise.
4603 (vn_nary_build_or_lookup_1): Make sure to not chain the built
4605 (vn_reference_insert): Adjust, chain the inserted hash element
4606 at last_inserted_ref.
4607 (vn_reference_insert_pieces): Likewise.
4608 (visit_reference_op_call): Likewise.
4609 (vn_nary_op_insert_into): Chain the inserted hash element at
4611 (vn_nary_op_insert_pieces): Adjust.
4612 (vn_nary_op_insert): Likewise.
4613 (vn_nary_op_insert_stmt): Likewise.
4614 (vn_phi_insert): Adjust, chain the inserted hash element at
4616 (process_scc): Remove clearing and copying the optimistic
4617 table. Instead remove elements inserted during an optimistic
4618 iteration from the single table we maintain.
4619 (init_scc_vn): Adjust.
4620 (free_scc_vn): Likewise.
4621 (sccvn_dom_walker::record_cond): Likewise.
4622 (sccvn_dom_walker::after_dom_children): Likewise.
4624 2018-07-19 Martin Sebor <msebor@redhat.com>
4626 PR tree-optimization/84047
4627 PR tree-optimization/83776
4628 * tree-vrp.c (vrp_prop::check_mem_ref): New function.
4629 (check_array_bounds): Call it.
4631 2018-07-19 Martin Sebor <msebor@redhat.com>
4633 * align.h (align_flags): Use member initialization.
4635 2018-07-19 David Malcolm <dmalcolm@redhat.com>
4637 * Makefile.in (OBJS): Add optinfo.o.
4638 * coretypes.h (class symtab_node): New forward decl.
4639 (struct cgraph_node): New forward decl.
4640 (class varpool_node): New forward decl.
4641 * dump-context.h: New file.
4642 * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
4644 (refresh_dumps_are_enabled): Use optinfo_enabled_p.
4645 (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
4646 (set_alt_dump_file): Likewise.
4647 (dump_context::~dump_context): New dtor.
4648 (dump_gimple_stmt): Move implementation to...
4649 (dump_context::dump_gimple_stmt): ...this new member function.
4650 Add the stmt to any pending optinfo, creating one if need be.
4651 (dump_gimple_stmt_loc): Move implementation to...
4652 (dump_context::dump_gimple_stmt_loc): ...this new member function.
4653 Start a new optinfo and add the stmt to it.
4654 (dump_gimple_expr): Move implementation to...
4655 (dump_context::dump_gimple_expr): ...this new member function.
4656 Add the stmt to any pending optinfo, creating one if need be.
4657 (dump_gimple_expr_loc): Move implementation to...
4658 (dump_context::dump_gimple_expr_loc): ...this new member function.
4659 Start a new optinfo and add the stmt to it.
4660 (dump_generic_expr): Move implementation to...
4661 (dump_context::dump_generic_expr): ...this new member function.
4662 Add the tree to any pending optinfo, creating one if need be.
4663 (dump_generic_expr_loc): Move implementation to...
4664 (dump_context::dump_generic_expr_loc): ...this new member
4665 function. Add the tree to any pending optinfo, creating one if
4667 (dump_printf): Move implementation to...
4668 (dump_context::dump_printf_va): ...this new member function. Add
4669 the text to any pending optinfo, creating one if need be.
4670 (dump_printf_loc): Move implementation to...
4671 (dump_context::dump_printf_loc_va): ...this new member function.
4672 Start a new optinfo and add the stmt to it.
4673 (dump_dec): Move implementation to...
4674 (dump_context::dump_dec): ...this new member function. Add the
4675 value to any pending optinfo, creating one if need be.
4676 (dump_context::dump_symtab_node): New member function.
4677 (dump_context::get_scope_depth): New member function.
4678 (dump_context::begin_scope): New member function.
4679 (dump_context::end_scope): New member function.
4680 (dump_context::ensure_pending_optinfo): New member function.
4681 (dump_context::begin_next_optinfo): New member function.
4682 (dump_context::end_any_optinfo): New member function.
4683 (dump_context::s_current): New global.
4684 (dump_context::s_default): New global.
4685 (dump_scope_depth): Delete global.
4686 (dumpfile_ensure_any_optinfo_are_flushed): New function.
4687 (dump_symtab_node): New function.
4688 (get_dump_scope_depth): Reimplement in terms of dump_context.
4689 (dump_begin_scope): Likewise.
4690 (dump_end_scope): Likewise.
4691 (selftest::temp_dump_context::temp_dump_context): New ctor.
4692 (selftest::temp_dump_context::~temp_dump_context): New dtor.
4693 (selftest::verify_item): New function.
4694 (ASSERT_IS_TEXT): New macro.
4695 (ASSERT_IS_TREE): New macro.
4696 (ASSERT_IS_GIMPLE): New macro.
4697 (selftest::test_capture_of_dump_calls): New test.
4698 (selftest::dumpfile_c_tests): Call it.
4699 * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
4700 (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
4701 (dump_gimple_stmt, dump_dec): Gather these related decls and add a
4702 descriptive comment.
4703 (dump_function, print_combine_total_stats, enable_rtl_dump_file)
4704 (dump_node, dump_bb): Move these unrelated decls.
4705 (class dump_manager): Add leading comment.
4706 * optinfo.cc: New file.
4707 * optinfo.h: New file.
4709 2018-07-19 Michael Collison <michael.collison@arm.com>
4710 Richard Henderson <rth@redhat.com>
4712 * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
4713 (subti): Handle op1 zero.
4714 (subvti4, usub4ti4): New.
4715 (*sub<GPI>3_compare1_imm): New.
4716 (sub<GPI>3_carryinCV): New.
4717 (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
4718 (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
4720 2018-07-19 Michael Collison <michael.collison@arm.com>
4721 Richard Henderson <rth@redhat.com>
4723 * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
4724 (addti3): Create simpler code if low part is already known to be 0.
4725 (addvti4, uaddvti4): New.
4726 (*add<GPI>3_compareC_cconly_imm): New.
4727 (*add<GPI>3_compareC_cconly): New.
4728 (*add<GPI>3_compareC_imm): New.
4729 (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
4730 handle constants within this pattern..
4731 (*add<GPI>3_compareV_cconly_imm): New.
4732 (*add<GPI>3_compareV_cconly): New.
4733 (*add<GPI>3_compareV_imm): New.
4734 (add<GPI>3_compareV): New.
4735 (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
4736 (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
4737 (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
4738 ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
4739 with 'comparison' operator.
4740 (*add<GPI>3_compareV_cconly_imm): Ditto.
4741 (*add<GPI>3_compareV_cconly): Ditto.
4742 (*add<GPI>3_compareV_imm): Ditto.
4743 (add<GPI>3_compareV): Ditto.
4744 (add<mode>3_carryinC): Ditto.
4745 (*add<mode>3_carryinC_zero): Ditto.
4746 (*add<mode>3_carryinC): Ditto.
4747 (add<mode>3_carryinV): Ditto.
4748 (*add<mode>3_carryinV_zero): Ditto.
4749 (*add<mode>3_carryinV): Ditto.
4751 2018-07-19 Michael Collison <michael.collison@arm.com>
4752 Richard Henderson <rth@redhat.com>
4754 * config/aarch64/aarch64-modes.def (CC_V): New.
4755 * config/aarch64/aarch64-protos.h
4756 (aarch64_addti_scratch_regs): Declare
4757 (aarch64_subvti_scratch_regs): Declare.
4758 (aarch64_expand_subvti): Declare.
4759 (aarch64_gen_unlikely_cbranch): Declare
4760 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
4761 for signed overflow using CC_Vmode.
4762 (aarch64_get_condition_code_1): Handle CC_Vmode.
4763 (aarch64_gen_unlikely_cbranch): New function.
4764 (aarch64_addti_scratch_regs): New function.
4765 (aarch64_subvti_scratch_regs): New function.
4766 (aarch64_expand_subvti): New function.
4768 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
4770 * config/aarch64/aarch64-option-extensions.def: New entry for profile
4772 * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
4773 * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
4776 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
4779 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
4780 address check not strict.
4782 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
4784 * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
4786 (store_pair_lanes<mode>): Likewise.
4787 * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
4788 enum value 'ADDR_QUERY_LDP_STP_N'.
4789 * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
4790 (aarch64_print_address_internal): Add declaration.
4791 (aarch64_print_ldpstp_address): Remove.
4792 (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
4793 (aarch64_print_operand): Change printing of 'y'.
4794 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
4795 new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
4796 'true' rather than '1'.
4797 * config/aarch64/constraints.md (Uml): Likewise.
4798 (Uml): Rename to Umn.
4801 2018-07-19 Richard Biener <rguenther@suse.de>
4803 * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
4805 * tree-ssa-sccvn.c: Remove alloc-pool.h use.
4806 (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
4807 (vn_reference_hasher): Likewise.
4808 (struct vn_tables_s): Remove obstack and alloc-pool members.
4809 (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
4810 (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
4811 (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
4812 (vn_reference_insert_pieces): Likewise.
4813 (alloc_vn_nary_op_noinit): Adjust.
4814 (vn_nary_op_insert_stmt): Allocate phiargs in-place.
4815 (vn_phi_eq): Adjust.
4816 (shared_lookup_phiargs): Remove.
4817 (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
4818 (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
4819 (visit_reference_op_call): Likewise.
4820 (copy_nary, copy_phi, copy_reference): Remove.
4821 (process_scc): Rewind the obstack when iterating. Do not
4822 copy the elements to valid_info but just move them from one
4823 hashtable to the other.
4824 (allocate_vn_table): Adjust.
4825 (free_vn_table): Likewise.
4826 (init_scc_vn): Likewise.
4827 (free_scc_vn): Likewise.
4829 2018-07-19 H.J. Lu <hongjiu.lu@intel.com>
4832 * config/i386/i386.c (rest_of_insert_endbranch): Lookup
4833 indirect_return as function type attribute.
4834 (ix86_attribute_table): Change indirect_return to function
4836 * doc/extend.texi: Update indirect_return attribute.
4838 2018-07-19 Aldy Hernandez <aldyh@redhat.com>
4840 * wide-int.h (widest2_int): New.
4841 * gimple-fold.c (arith_overflowed_p): Use it.
4842 * tree.h (widest2_int_cst): New.
4843 * tree-vrp.c (wide_int_binop_overflow): Rename from
4844 vrp_int_const_binop.
4845 Rewrite to work on trees.
4846 (extract_range_from_multiplicative_op_1): Abstract code to...
4847 (wide_int_range_min_max): ...here.
4848 (wide_int_range_cross_product): ...and here.
4849 (extract_range_from_binary_expr_1): Abstract overflow code to...
4850 (wide_int_range_mult_wrapping): ...here.
4851 * tree-vrp.h (wide_int_range_cross_product): New.
4852 (wide_int_range_mult_wrapping): New.
4854 2018-07-19 Andrew Senkevich <andrew.senkevich@intel.com>
4855 Julia Koval <julia.koval@intel.com>
4857 * config/i386/x86-tune-costs.h (skylake_memcpy,
4858 skylake_memset): Replace rep_prefix with unrolling for size 512.
4860 2018-07-18 Kugan Vivekanandarajah <kuganv@linaro.org>
4863 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
4864 comparision with EQ_EXPR in last stmt.
4866 2018-07-18 Kelvin Nilsen <kelvin@gcc.gnu.org>
4868 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
4869 this subsection to "PowerPC AltiVec/VSX Built-in Functions".
4870 (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
4871 previously known as "PowerPC AltiVec Built-in Functions". Move
4872 some material to new subsubsections "PowerPC AltiVec Built-in
4873 Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
4875 (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
4876 (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
4877 (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
4878 (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
4880 2018-07-18 Richard Biener <rguenther@suse.de>
4882 PR tree-optimization/86557
4883 * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
4886 2018-07-18 Ilya Leoshkevich <iii@linux.ibm.com>
4888 * config/s390/s390.c (s390_function_profiler): Generate CFI.
4890 2018-07-17 Jeff Law <law@redhat.com>
4892 * config/arm/arm.c (get_label_padding): Update for recent
4893 changes to label_to_alignment.
4895 PR tree-optimization/86010
4896 * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
4898 * config/mips/mips.c (vr4130_align_insns): Update for recent
4899 changes to label_to_alignment.
4901 * config/frv/frv.c (frv_label_align): Update for recent changes
4902 to label_to_alignment.
4904 * config/nios2/nios2.c (nios2_label_align): Update for recent
4905 changes which dropped ALIGN_LABELS_LOG.
4907 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
4909 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
4910 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
4912 2018-07-17 Claudiu Zissulescu <claziss@synopsys.com>
4914 * config/arc/arc.c (arc_label_align): Use align_labels instead of
4915 deprecated align_labels_log.
4917 2018-07-17 Richard Biener <rguenther@suse.de>
4920 * dwarf2out.c (init_sections_and_labels): Always generate
4921 a debug_line_str_section for early LTO debug.
4922 (dwarf2out_finish): Reset debug_line_str_hash output early.
4923 Bump counter for extra dwarf5 .debug_loc labels to not conflict
4925 (dwarf2out_early_finish): Output debug_line_str.
4927 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
4929 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
4930 index register on z196 or later.
4932 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
4934 * config/s390/s390.c (s390_default_align): Set default function
4936 (s390_override_options_after_change): Call s390_default align.
4937 (s390_option_override_internal): Call s390_default_align.
4938 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
4940 2018-07-17 Jakub Jelinek <jakub@redhat.com>
4943 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
4944 to _looptemp_ clauses, other than the first two.
4946 2018-07-17 Martin Liska <mliska@suse.cz>
4948 * opts.c: Do not enable OPT_falign_* for -Os.
4950 2018-07-17 Martin Liska <mliska@suse.cz>
4952 * align.h (MAX_CODE_ALIGN): New.
4953 (MAX_CODE_ALIGN_VALUE): New.
4954 * common/config/i386/i386-common.c (ix86_handle_option):
4955 (MAX_CODE_ALIGN): Moved to align.h.
4956 * final.c (MAX_CODE_ALIGN): Likewise.
4957 * opts.c (parse_and_check_align_values):
4958 (MAX_CODE_ALIGN): Likewise.
4959 (MAX_CODE_ALIGN_VALUE): Likewise.
4961 2018-07-17 Martin Liska <mliska@suse.cz>
4963 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
4964 in order to fulfil coding style.
4965 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
4966 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4967 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
4968 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
4969 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
4970 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
4971 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
4972 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
4973 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4975 2018-07-17 Martin Liska <mliska@suse.cz>
4977 * align.h: New file.
4978 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
4980 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
4981 align_flags of label_to_alignment.
4982 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
4984 * config/m68k/m68k.c: Do not use removed align_labels_value and
4986 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
4987 (LOOP_ALIGN): Likewise.
4988 (LABEL_ALIGN): Likewise.
4989 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
4990 Remove not used macro.
4991 (rs6000_loop_align): Change return type to align_flags.
4992 (rs6000_loop_align_max_skip): Remove.
4993 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
4994 Change return type to align_flags.
4995 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
4996 Remove not used macro.
4997 (rs6000_loop_align): Change return type to align_flags.
4998 (rs6000_loop_align_max_skip): Remove.
4999 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
5000 * config/rx/rx-protos.h (rx_align_for_label): Make it
5002 * config/rx/rx.c (rx_align_for_label): Change return type
5004 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
5006 into align_flags class.
5007 (LABEL_ALIGN): Likewise.
5008 (LOOP_ALIGN): Likewise.
5009 * config/s390/s390.c (s390_label_align): Use align_flags
5011 (s390_asm_output_function_label): Likewise.
5012 * config/sh/sh.c (sh_override_options_after_change):
5013 Use align_flags class directly without macros.
5014 (find_barrier): Likewise.
5015 (barrier_align): Likewise.
5016 (sh_loop_align): Likewise.
5017 * config/spu/spu.c (spu_option_override):
5018 Use align_flags_tuple::get_value instead of removed macros.
5019 (spu_sched_init): Likewise.
5020 * config/spu/spu.h (GTY): Likewise.
5021 * config/visium/visium.c (visium_option_override):
5022 Set "8" as default secondary alignment.
5023 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
5024 in order to guarantee secondary alignment of 8.
5025 * coretypes.h: Include align.h header file.
5026 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
5027 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
5028 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
5029 * doc/tm.texi.in: Likewise.
5030 * final.c (struct label_alignment): Remove not used structure.
5031 (LABEL_ALIGN): Change type to align_flags.
5032 (LOOP_ALIGN): Likewise.
5033 (JUMP_ALIGN): Likewise.
5034 (default_loop_align_max_skip): Remove.
5035 (default_label_align_max_skip): Likewise.
5036 (default_jump_align_max_skip): Likewise.
5037 (default_label_align_after_barrier_max_skip):
5038 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
5039 (LABEL_TO_MAX_SKIP): Remove.
5040 (label_to_alignment): Return align_flags type instead of integer.
5041 (label_to_max_skip): Remove.
5042 (align_fuzz): Use align_flags type.
5043 (compute_alignments): Use align_flags type and use align_flags::max
5044 to combine multiple alignments.
5045 (grow_label_align): Grow vec instead of C array.
5046 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
5047 (shorten_branches): Use align_flags type and use align_flags::max
5048 to combine multiple alignments.
5049 (final_scan_insn_1): Remove usage of secondary alignment that comes
5050 from label alignment, but instead use proper secondary alignment
5051 which is computed in grow_label_align.
5052 * flags.h (struct align_flags_tuple): Move to align.h.
5053 (struct align_flags): Likewise.
5054 (state_align_loops): Rename to align_loops.
5055 (state_align_jumps): Rename to align_jumps.
5056 (state_align_labels): Rename to align_labels.
5057 (state_align_functions): Rename to align_functions.
5058 (align_loops_log): Remove.
5059 (align_jumps_log): Remove.
5060 (align_labels_log): Remove.
5061 (align_functions_log): Remove.
5062 (align_loops_max_skip): Remove.
5063 (align_jumps_max_skip): Remove.
5064 (align_labels_max_skip): Remove.
5065 (align_functions_max_skip): Remove.
5066 (align_loops_value): Remove.
5067 (align_jumps_value): Remove.
5068 (align_labels_value): Remove.
5069 (align_functions_value): Remove.
5070 * output.h (label_to_alignment): Change return type to align_flags.
5071 (label_to_max_skip): Remove.
5072 * target.def: Remove loop_align_max_skip, label_align_max_skip,
5073 jump_align_max_skip macros.
5074 * targhooks.h (default_loop_align_max_skip): Remove.
5075 (default_label_align_max_skip): Likewise.
5076 (default_jump_align_max_skip): Likewise.
5077 (default_label_align_after_barrier_max_skip): Remove.
5078 * toplev.c (read_log_maxskip): Use ::normalize function.
5079 (parse_N_M): Remove not used argument and also call ::normalize.
5080 (parse_alignment_opts): Do not pass unused arguments.
5081 * varasm.c (assemble_start_function): Use directly align_functions
5082 instead of removed macros.
5083 * system.h: Do not poison removed macros.
5085 2018-07-17 Jakub Jelinek <jakub@redhat.com>
5088 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
5089 and cond temporaries don't have reference type if iterator has
5090 pointer type. For init use &for_pre_body instead of pre_p if
5091 for_pre_body is non-empty.
5093 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
5095 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
5096 double-double modes to SFmode directly directly.
5097 (trunc<mode>sf2_fprs): Delete.
5099 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
5101 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
5102 for conversions between IFmode and the decimal floating point modes.
5103 (init_float128_ieee): Use the correct names for conversions between
5104 KFmode and the decimal floating point modes.
5106 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
5108 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
5109 for the conversions between TDmode and IFmode.
5110 (init_float128_ieee): Use more correct names for the conversions
5111 between TDmode and KFmode.
5113 2018-07-16 Jakub Jelinek <jakub@redhat.com>
5115 PR tree-optimization/86526
5116 * builtins.c (expand_builtin_memcmp): Formatting fixes.
5117 (inline_expand_builtin_string_cmp): Likewise.
5118 (inline_string_cmp): Likewise. Use c_readstr instead of
5119 builtin_memcpy_read_str. Add unit_mode temporary.
5121 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
5124 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
5125 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
5127 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
5129 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
5130 Alphabetize prototypes of built-in functions, separating out
5131 built-in functions that are listed in this section but should be
5132 described elsewhere.
5134 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
5137 * expmed.c (emit_store_flag): Do not emit setcc followed by a
5138 conditional move when trapping comparison was split to a
5139 non-trapping one (and vice versa).
5141 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
5143 * config/s390/s390.c (s390_function_profiler): Generate nops
5144 instead of profiler call sequences.
5145 * config/s390/s390.opt: Add the new option.
5147 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
5149 * config/s390/s390.c (s390_function_profiler): Generate
5150 __mcount_loc section.
5151 * config/s390/s390.opt: Add the new option.
5153 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
5155 * common.opt: Add the new warning.
5156 * config/s390/s390.c (s390_function_profiler): Emit "brasl
5157 %r0,__fentry__" when -mfentry is specified.
5158 (s390_option_override_internal): Disallow -mfentry for 31-bit
5160 * config/s390/s390.opt: Add the new option.
5162 2018-07-16 Richard Biener <rguenther@suse.de>
5165 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
5166 for function-local FUNCTION_DECL and RESULT_DECL immediately.
5168 2018-07-16 Martin Liska <mliska@suse.cz>
5171 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
5174 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
5176 * config/arc/arcHS.md: Update ARCHS scheduling rules.
5178 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
5180 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
5182 * config/arc/arc-cpus.def (hs4x): New cpu.
5184 * config/arc/arc-tables.opt: Regenerate.
5185 * config/arc/arc.c (arc_sched_issue_rate): New function.
5186 (TARGET_SCHED_ISSUE_RATE): Define.
5187 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
5188 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
5190 (attr tune): Add ARCHS4x tune values.
5191 (attr tune_dspmpy): Define.
5192 (*tst): Correct instruction type.
5193 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
5194 * config/arc/arcHS4x.md: New file.
5195 * config/arc/fpu.md: Update instruction type attributes.
5196 * config/arc/t-multilib: Regenerate.
5198 2018-07-16 Tom de Vries <tdevries@suse.de>
5201 * var-tracking.c (vt_initialize): Fix pre_dec handling.
5203 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5205 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
5208 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
5210 PR tree-optimization/86514
5211 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
5212 conversion to a boolean type from a type with greater precision.
5214 2018-07-16 Tom de Vries <tdevries@suse.de>
5216 * var-tracking.c (vt_initialize): Print adjusted insn slim if
5217 dump_flags request TDF_SLIM.
5219 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
5221 * fold-const.c (int_const_binop_1): Abstract...
5222 (wide_int_binop): ...wide int code here.
5223 (poly_int_binop): ...poly int code here.
5224 Abstract the rest of int_const_binop_1 into int_const_binop.
5225 * fold-const.h (wide_int_binop): New.
5226 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
5227 Remove useless PLUS/MINUS_EXPR case.
5228 (zero_nonzero_bits_from_vr): Move wide int code...
5229 (zero_nonzero_bits_from_bounds): ...here.
5230 (extract_range_from_binary_expr_1): Move mask optimization code...
5231 (range_easy_mask_min_max): ...here.
5232 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
5233 (range_easy_mask_min_max): New.
5235 2018-07-15 Jeff Law <law@redhat.com>
5238 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
5241 2018-07-14 Jim Wilson <jimw@sifive.com>
5243 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
5245 2018-07-14 Paul Koning <ni1d@arrl.net>
5247 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
5249 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
5251 * lto-streamer-out.c (copy_function_or_variable): Dump info about
5254 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
5255 Steve Munroe <munroesj52@gmail.com>
5257 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
5258 (_mm_andnot_si128): Likewise.
5259 (_mm_or_si128): Likewise.
5260 (_mm_xor_si128): Likewise.
5262 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
5265 * builtins.c (expand_builtin_memcmp): Inline the calls first
5266 when result_eq is false.
5267 (expand_builtin_strcmp): Inline the calls first.
5268 (expand_builtin_strncmp): Likewise.
5269 (inline_string_cmp): New routine. Expand a string compare
5270 call by using a sequence of char comparison.
5271 (inline_expand_builtin_string_cmp): New routine. Inline expansion
5272 a call to str(n)cmp/memcmp.
5273 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
5275 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
5277 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
5279 * config/arm/driver-arm.c: Include arm-native.h.
5280 (host_detect_local_cpu): Use auto-generated data tables.
5281 (vendors, arm_cpu_table): Delete. Move part information to ...
5282 * config/arm/arm-cpus.in: ... here.
5283 * config/arm/parsecpu.awk (gen_native): New function.
5284 (vendor, part): New CPU fields.
5285 (END): Add support for building the native CPU detection tables.
5286 * config/arm/t-arm (arm-native.h): Add build rule.
5287 (driver-arm.o): Add dependency on arm-native.h.
5289 2018-07-13 Richard Biener <rguenther@suse.de>
5292 * match.pd (addr1 - addr2): Allow either of the operand to
5295 2018-07-13 Tom de Vries <tdevries@suse.de>
5297 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
5300 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
5302 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
5303 arrays instead of numbered variables.
5305 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
5307 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
5308 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
5310 2018-07-13 Richard Biener <rguenther@suse.de>
5313 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
5314 instead of get_context_die.
5316 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
5317 Richard Biener <rguenther@suse.de>
5320 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
5321 that the loop latch destination where phi is defined.
5323 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
5325 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
5326 (riscv_expand_epilogue): Add assertion to check interrupt mode.
5327 (riscv_set_current_function): Extract getting interrupt type to new
5329 (riscv_get_interrupt_type): New function.
5330 (riscv_merge_decl_attributes): New function, checking interrupt type is
5332 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
5334 2018-07-12 Paul Koning <ni1d@arrl.net>
5336 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
5339 2018-07-12 Paul Koning <ni1d@arrl.net>
5341 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
5342 zero reference, add doloop_end instead.
5343 * doc/md.texi (decrement_and_branch_until_zero): Remove.
5344 (Looping patterns): Remove decrement_and_branch_until_zero. Add
5345 detail for doloop_end.
5347 2018-07-12 Martin Sebor <msebor@redhat.com>
5350 * attribs.c (decl_attributes): Reject conflicting attributes before
5351 calling attribute handlers.
5353 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
5355 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
5357 (gcc::dump_manager::get_dump_file_name): likewise.
5358 (dump_begin): Likewise.
5359 * dumpfile.h (dump_begin): Update prototype.
5360 (gcc::dump_manager::get_dump_file_name,
5361 gcc::dump_manager::get_dump_file_name): Update prototype.
5363 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
5365 * internal-fn.h (vectorizable_internal_fn_p): New function.
5366 * tree-vect-slp.c (compatible_calls_p): Likewise.
5367 (vect_build_slp_tree_1): Remove nops argument. Handle calls
5368 to internal functions.
5369 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
5371 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
5373 * fold-const.h (inverse_conditions_p): Declare.
5374 * fold-const.c (inverse_conditions_p): New function.
5375 * match.pd: Use inverse_conditions_p. Add folds of view_converts
5376 that test the inverse condition of a conditional internal function.
5377 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
5378 * internal-fn.c (internal_fn_mask_index): Handle conditional
5380 (vectorized_internal_fn_supported_p): New function.
5381 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
5382 (any_pred_load_store): Replace with...
5383 (need_to_predicate): ...this new variable.
5384 (redundant_ssa_names): New variable.
5385 (ifcvt_can_use_mask_load_store): Move initial checks to...
5386 (ifcvt_can_predicate): ...this new function. Handle tree codes
5387 for which a conditional internal function exists.
5388 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
5389 instead of ifcvt_can_use_mask_load_store. Update after variable
5391 (predicate_load_or_store): New function, split out from
5392 predicate_mem_writes.
5393 (check_redundant_cond_expr): New function.
5394 (value_available_p): Likewise.
5395 (predicate_rhs_code): Likewise.
5396 (predicate_mem_writes): Rename to...
5397 (predicate_statements): ...this. Use predicate_load_or_store
5398 and predicate_rhs_code.
5399 (combine_blocks, tree_if_conversion): Update after above name changes.
5400 (ifcvt_local_dce): Handle redundant_ssa_names.
5401 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
5402 general conditional functions.
5403 * tree-vect-stmts.c (vectorizable_call): Likewise.
5405 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
5406 Alan Hayward <alan.hayward@arm.com>
5407 David Sherwood <david.sherwood@arm.com>
5409 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
5410 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
5411 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
5412 plus and minus and convert them into IFN_COND_FMA-based sequences.
5413 (convert_mult_to_fma): Handle conditional plus and minus.
5415 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
5417 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
5418 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
5419 (cond_fnms_optab): New optabs.
5420 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
5422 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
5423 * internal-fn.h (get_conditional_internal_fn): Declare.
5424 (get_unconditional_internal_fn): Likewise.
5425 * internal-fn.c (cond_ternary_direct): New macro.
5426 (expand_cond_ternary_optab_fn): Likewise.
5427 (direct_cond_ternary_optab_supported_p): Likewise.
5428 (FOR_EACH_COND_FN_PAIR): Likewise.
5429 (get_conditional_internal_fn): New function.
5430 (get_unconditional_internal_fn): Likewise.
5431 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
5432 (gimple_match_op::gimple_match_op): Add a new overload for 5
5434 (gimple_match_op::set_op): Likewise.
5435 (gimple_resimplify5): Declare.
5436 * genmatch.c (decision_tree::gen): Generate simplifications for
5438 * gimple-match-head.c (gimple_simplify): Define an overload for
5439 5 operands. Handle calls with 5 arguments in the top-level overload.
5440 (convert_conditional_op): Handle conversions from unconditional
5441 internal functions to conditional ones.
5442 (gimple_resimplify5): New function.
5443 (build_call_internal): Pass a fifth operand.
5444 (maybe_push_res_to_seq): Likewise.
5445 (try_conditional_simplification): Try converting conditional
5446 internal functions to unconditional internal functions.
5447 Handle 3-operand unconditional forms.
5448 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
5449 Define ternary equivalents of the current rules for binary conditional
5451 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
5453 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
5454 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
5455 (optab): Handle them.
5456 (SVE_COND_FP_TERNARY): New int iterator.
5457 (sve_fmla_op, sve_fmad_op): New int attributes.
5458 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
5459 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
5460 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
5462 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
5464 * target.def (preferred_else_value): New target hook.
5465 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
5466 * doc/tm.texi: Regenerate.
5467 * targhooks.h (default_preferred_else_value): Declare.
5468 * targhooks.c (default_preferred_else_value): New function.
5469 * internal-fn.h (conditional_internal_fn_code): Declare.
5470 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
5471 (get_conditional_internal_fn): Use it.
5472 (conditional_internal_fn_code): New function.
5473 * gimple-match.h (gimple_match_cond): New struct.
5474 (gimple_match_op): Add a cond member function.
5475 (gimple_match_op::gimple_match_op): Update all forms to take a
5477 * genmatch.c (expr::gen_transform): Use the same condition as res_op
5478 for the suboperation, but don't specify a particular else_value.
5479 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
5480 (visit_nary_op, visit_reference_op_load): Pass
5481 gimple_match_cond::UNCOND to the gimple_match_op constructor.
5482 * gimple-match-head.c: Include tree-eh.h
5483 (convert_conditional_op): New function.
5484 (maybe_resimplify_conditional_op): Likewise.
5485 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
5486 (gimple_resimplify2): Likewise.
5487 (gimple_resimplify3): Likewise.
5488 (gimple_resimplify4): Likewise.
5489 (maybe_push_res_to_seq): Return null for conditional operations.
5490 (try_conditional_simplification): New function.
5491 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
5493 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
5495 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
5497 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
5499 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
5501 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
5503 (hash_tree): Do not hash DECL_FCONTEXT
5504 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
5505 Do not stream DECL_FCONTEXT.
5506 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
5507 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
5509 2018-07-12 Richard Biener <rguenther@suse.de>
5512 * dwarf2out.c (gen_block_die): Only output blocks when they have
5513 at least one !DECL_IGNORED_P variable.
5515 2018-07-12 Richard Biener <rguenther@suse.de>
5518 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
5519 Remove -mieee-fp handling.
5521 2018-07-12 Richard Biener <rguenther@suse.de>
5523 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
5524 left-over from last patch.
5526 2018-07-12 Jakub Jelinek <jakub@redhat.com>
5528 PR tree-optimization/86492
5529 * gimple-ssa-store-merging.c
5530 (imm_store_chain_info::coalesce_immediate_stores): Call
5531 check_no_overlap even for the merge_overlapping case. Formatting fix.
5533 2018-07-12 Richard Biener <rguenther@suse.de>
5536 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
5537 move possibly trapping operations into the conditional.
5539 2018-07-12 Richard Biener <rguenther@suse.de>
5541 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
5542 (vn_lookup_simplify_result): Remove recursion limit applied
5544 (vn_nary_build_or_lookup_1): Adjust.
5545 (try_to_simplify): Likewise.
5546 * gimple-match-head.c (gimple_resimplify1): Instead apply one
5548 (gimple_resimplify2): Likewise.
5549 (gimple_resimplify3): Likewise.
5550 (gimple_resimplify4): Likewise.
5552 2018-07-11 Jakub Jelinek <jakub@redhat.com>
5554 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
5555 Use __mmask64 type instead of __mmask8 for __M argument.
5556 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
5557 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
5559 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
5560 __mmask16 for __M argument.
5561 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
5562 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
5563 to __mmask16 instead of __mmask8.
5564 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
5565 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
5566 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
5567 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
5568 instead of __mmask16 for __U argument.
5569 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
5570 __mmask16 instead of __mmask8 for __U argument.
5571 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
5573 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
5575 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
5577 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
5579 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
5581 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
5583 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
5585 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
5587 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
5589 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
5591 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
5592 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
5593 return type as well as __M argument type and all casts from __mmask8
5595 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
5596 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
5597 return type as well as __M argument type and all casts from __mmask8
5599 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
5600 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
5601 return type as well as __M argument type and all casts from __mmask8
5603 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
5604 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
5605 return type as well as __M argument type and all casts from __mmask8
5607 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
5608 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
5611 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
5613 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
5614 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
5617 2018-07-11 Paul Koning <ni1d@arrl.net>
5619 * doc/md.texi (define_subst): Document how multiple occurrences of
5620 the same argument in the replacement pattern are handled.
5622 2018-07-11 Paul Koning <ni1d@arrl.net>
5624 * doc/extend.texi (Common Variable Attributes): Move "mode" into
5626 (Common Type Attributes): Add "mode" attribute.
5628 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
5630 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
5631 stream DECL_ORIGINAL_TYPE.
5632 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
5633 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
5634 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
5635 Do not walk original type.
5636 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
5638 (write_ts_decl_non_common_tree_pointers): Do not stream
5640 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
5641 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
5643 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
5645 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
5646 thread twice from the same starting edge.
5648 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
5650 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
5651 * gimple.c (gimple_call_nonnull_result_p): ...here...
5652 (gimple_call_nonnull_arg): ...and here.
5653 * gimple.h (gimple_call_nonnull_result_p): New.
5654 (gimple_call_nonnull_arg): New.
5656 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
5658 * config/arm/arm-cpus.in: Move information from fpu field of each
5659 cpu definition to the isa field.
5660 * config/arm/parsecpu.awk (fpu): Delete match rule.
5661 (gen_comm_data): Don't add bits from the CPU's FPU entry.
5663 2018-07-11 Richard Biener <rguenther@suse.de>
5666 * dwarf2out.c (init_sections_and_labels): Use
5667 output_asm_line_debug_info consistently.
5668 (dwarf2out_early_finish): Likewise.
5669 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
5672 2018-07-11 Richard Biener <rguenther@suse.de>
5674 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
5675 Rework father_bb setting in a way to avoid propagating constants
5676 multiple times on a loop body.
5678 2018-07-10 Mark Wielaard <mark@klomp.org>
5681 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
5683 2018-07-10 Richard Biener <rguenther@suse.de>
5685 * hash-map.h (hash_map::iterator::operator*): Return
5686 references to key and value.
5688 2018-07-10 Jakub Jelinek <jakub@redhat.com>
5691 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
5692 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
5693 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
5694 (gimplify_omp_for): For composite loops, move outer
5695 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
5696 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
5697 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
5698 TREE_LIST for both the original class iterator and the "last" helper
5699 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
5700 loop, remember has_decl_expr from outer composite loops for the
5701 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
5703 2018-07-09 Martin Sebor <msebor@redhat.com>
5707 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
5708 accessing implicitly initialized array elements.
5709 * expr.c (string_constant): Handle string initializers of
5710 character arrays within aggregates.
5711 * gimple-fold.c (fold_array_ctor_reference): Add argument.
5712 Store element offset. As a special case, handle zero size.
5713 (fold_nonarray_ctor_reference): Same.
5714 (fold_ctor_reference): Add argument. Store subobject offset.
5715 * gimple-fold.h (fold_ctor_reference): Add argument.
5717 2018-07-09 Paul Koning <ni1d@arrl.net>
5719 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
5720 (pdp11_insn_cost): New function.
5721 (pdp11_md_asm_adjust): New function.
5722 (TARGET_INVALID_WITHIN_DOLOOP): Define.
5723 (pdp11_rtx_costs): Update to match machine better.
5724 (output_addr_const_pdp11): Correct format mismatch warnings.
5725 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
5726 * config/pdp11/pdp11.md: General change to add base_cost and/or
5727 length attributes for use by new pdp11_insn_cost function.
5728 (MIN_BRANCH): Correct definition.
5730 (doloop_end): Use standard pattern name for looping pattern.
5731 (doloop_end_nocc): New.
5732 (movsf): Add another constraint alternative.
5733 (zero_extendqihi2): Add constraint alternatives for not in place
5735 (zero_extendhisi2): Remove.
5736 (shift patterns): Add CC handling variants.
5740 (define_peephole2): New peephole to recognize mov that sets CC for
5743 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5745 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
5746 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
5749 2018-07-09 Jakub Jelinek <jakub@redhat.com>
5752 * real.c (real_nextafter): Return true if result is denormal.
5754 2018-07-09 Martin Liska <mliska@suse.cz>
5756 * common.opt: Add back wrongly removed attribute.
5758 2018-07-09 Richard Biener <rguenther@suse.de>
5761 * dwarf2out.c (gen_block_die): For an early generated DIE
5762 always output high/low PC attributes.
5764 2018-07-09 Tom de Vries <tdevries@suse.de>
5766 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
5767 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
5768 onto VAR_DECL with abstract origin.
5770 2018-07-07 Jim Wilson <jimw@sifive.com>
5772 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
5774 2018-07-07 Tom de Vries <tdevries@suse.de>
5776 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
5778 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
5780 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
5782 (combine_bound): Use wide-int overflow calculation instead of
5784 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
5786 * fold-const.c (int_const_binop_2): Same.
5787 (extract_muldiv_1): Same.
5788 (fold_div_compare): Same.
5789 (fold_abs_const): Same.
5791 * poly-int.h (add): Same.
5795 * predict.c (predict_iv_comparison): Same.
5796 * profile-count.c (slow_safe_scale_64bit): Same.
5797 * simplify-rtx.c (simplify_const_binary_operation): Same.
5798 * tree-chrec.c (tree_fold_binomial): Same.
5799 * tree-data-ref.c (split_constant_offset_1): Same.
5800 * tree-if-conv.c (idx_within_array_bound): Same.
5801 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
5802 * tree-ssa-phiopt.c (minmax_replacement): Same.
5803 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
5804 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
5805 * vr-values.c (vr_values::adjust_range_with_scev): Same.
5806 * wide-int.cc (wi::add_large): Same.
5807 (wi::mul_internal): Same.
5808 (wi::sub_large): Same.
5809 (wi::divmod_internal): Same.
5810 * wide-int.h: Change overflow type to overflow_type for neg, add,
5811 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
5812 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
5813 mul_internal, divmod_internal.
5814 (overflow_type): New enum.
5815 (accumulate_overflow): New.
5817 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
5819 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
5820 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
5822 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
5824 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
5825 argument is checked for zero before entering loop, avoid checking again.
5827 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
5829 * gimplify.h (generic_expr_could_trap_p): Set as global function.
5830 * gimplify.h (generic_expr_could_trap_p): Likwise.
5831 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
5833 2018-07-06 Jakub Jelinek <jakub@redhat.com>
5835 PR tree-optimization/86401
5836 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
5837 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
5838 (fold_bit_and_mask): ... here. New helper function for match.pd.
5839 * fold-const.h (fold_bit_and_mask): Declare.
5840 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
5842 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
5845 * target.def (translate_mode_attribute): New hook.
5846 * targhooks.h (default_translate_mode_attribute): Declare.
5847 * targhooks.c (default_translate_mode_attribute): New function.
5848 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
5849 * doc/tm.texi: Regenerate.
5850 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
5851 (rs6000_translate_mode_attribute): New function.
5853 2018-07-06 Paul Koning <ni1d@arrl.net>
5855 * doc/md.texi (define_split): Document DONE and FAIL.
5856 (define_peephole2): Ditto.
5858 2018-07-05 Jeff Law <law@redhat.com>
5860 PR tree-optimization/86010
5861 * tree-ssa-dse.c (compute_trims): More aggressively trim at
5862 both the head and tail of mem* and str* calls.
5864 2018-07-05 Jim Wilson <jimw@sifive.com>
5866 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
5868 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
5870 * config/aarch64/aarch64-simd.md: correct flags text for
5871 MIN_EXPR replacement.
5873 2018-07-05 James Clarke <jrtc27@jrtc27.com>
5875 * configure: Regenerated.
5877 2018-07-05 Carl Love <cel@us.ibm.com>
5879 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
5880 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
5881 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
5882 VSX_BUILTIN_DOUBLEL_V4SF.
5884 2018-07-05 Martin Sebor <msebor@redhat.com>
5887 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
5888 than its domain to compute its the upper bound of a char array.
5890 2018-07-05 Nathan Sidwell <nathan@acm.org>
5892 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
5893 * doc/cpp.texi: Update comment.
5894 * doc/tm.texi: Rebuilt.
5895 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
5896 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
5897 * doc/extend.texi (Backwards Compatibility): Clarify it is system
5898 headers affected by extern "C".
5899 * system.h: Poison NO_IMPLICIT_EXTERN_C.
5900 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
5901 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
5902 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
5903 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
5904 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
5905 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
5906 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
5907 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
5908 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
5909 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
5910 config/sparc/sp64-elf.h, config/spu/spu.h,
5911 config/stormy16/stormy16.h, config/v850/v850.h,
5912 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
5913 define NO_IMPLICIT_EXTERN_C.
5914 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
5916 2018-07-05 Tamar Christina <tamar.christina@arm.com>
5919 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
5920 instead of GET_MODE_SIZE when comparing Units.
5922 2018-07-05 Tamar Christina <tamar.christina@arm.com>
5925 * rtlanal.c (set_noop_p): Constrain on mode change,
5926 include hard-reg-set.h
5928 2018-07-05 Tamar Christina <tamar.christina@arm.com>
5930 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
5932 2018-07-05 Jakub Jelinek <jakub@redhat.com>
5935 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
5938 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
5940 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
5942 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
5945 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
5947 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
5949 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
5952 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
5954 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
5956 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
5957 neg_*_op* variables.
5959 2018-07-04 Martin Liska <mliska@suse.cz>
5961 * tree-switch-conversion.c: Define
5962 max_ratio_for_speed and max_ratio_for_size constants.
5964 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
5965 Martin Liska <mliska@suse.cz>
5970 * common.opt: Rename align options with 'str_' prefix.
5971 * common/config/i386/i386-common.c (set_malign_value): New
5973 (ix86_handle_option): Use it to set -falign-* options/
5974 * config/aarch64/aarch64-protos.h (struct tune_params): Change
5975 type from int to string.
5976 * config/aarch64/aarch64.c: Update default values from int
5978 * config/alpha/alpha.c (alpha_override_options_after_change):
5980 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
5981 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
5982 max skip conditionally.
5983 * config/i386/freebsd.h (SUBALIGN_LOG): New.
5984 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
5985 max skip conditionally.
5986 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
5987 max skip conditionally.
5988 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
5989 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
5990 max skip conditionally.
5991 * config/i386/i386.c (struct ptt): Change type from int to
5993 (ix86_default_align): Set default values.
5994 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
5995 max skip conditionally.
5996 * config/i386/iamcu.h (SUBALIGN_LOG): New.
5997 (ASM_OUTPUT_MAX_SKIP_ALIGN):
5998 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
5999 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
6000 max skip conditionally.
6001 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
6002 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
6003 * config/i386/x86-64.h (SUBALIGN_LOG): New.
6004 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
6005 max skip conditionally.
6006 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
6007 * config/ia64/ia64.c (ia64_option_override): Set default values
6008 for alignment options.
6009 * config/m68k/m68k.c: Handle new str_align_* options.
6010 * config/mips/mips.c (mips_set_compression_mode): Change
6012 (mips_option_override): Set default values for options.
6013 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
6015 * config/rs6000/rs6000.c (rs6000_option_override_internal):
6017 * config/rx/rx.c (rx_option_override): Likewise.
6018 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
6019 (LABEL_ALIGN): Use align_labels_log.
6020 (LOOP_ALIGN): Use align_loops_align.
6021 * config/s390/s390.c (s390_asm_output_function_label): Use new
6023 * config/sh/sh.c (sh_override_options_after_change):
6024 Change type of constants.
6025 * config/spu/spu.c (spu_sched_init): Likewise.
6026 * config/sparc/sparc.c (sparc_option_override): Set default
6028 * config/visium/visium.c (visium_option_override): Likewise.
6029 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
6030 emit p2align format with last argument if it's not needed.
6031 * doc/invoke.texi: Document extended format of -falign-*.
6032 * final.c: Use align_labels alignment.
6033 * flags.h (struct target_flag_state): Change type to use
6035 (struct align_flags_tuple): New.
6036 (struct align_flags): Likewise.
6037 (align_loops_log): Redefine macro to use new types.
6038 (align_loops_max_skip): Redefine macro to use new types.
6039 (align_jumps_log): Redefine macro to use new types.
6040 (align_jumps_max_skip): Redefine macro to use new types.
6041 (align_labels_log): Redefine macro to use new types.
6042 (align_labels_max_skip): Redefine macro to use new types.
6043 (align_functions_log): Redefine macro to use new types.
6044 (align_loops): Redefine macro to use new types.
6045 (align_jumps): Redefine macro to use new types.
6046 (align_labels): Redefine macro to use new types.
6047 (align_functions): Redefine macro to use new types.
6048 (align_functions_max_skip): Redefine macro to use new types.
6049 (align_loops_value): New macro.
6050 (align_jumps_value): New macro.
6051 (align_labels_value): New macro.
6052 (align_functions_value): New macro.
6053 * function.c (invoke_set_current_function_hook): Propagate
6054 alignment values from flags to global variables default in
6056 * ipa-icf.c (sem_function::equals_wpa): Use
6057 cl_optimization_option_eq instead of memcmp.
6058 * lto-streamer.h (cl_optimization_stream_out): Support streaming
6060 (cl_optimization_stream_in): Likewise.
6061 * optc-save-gen.awk: Support strings in cl_optimization.
6062 * opth-gen.awk: Likewise.
6063 * opts.c (finish_options): Remove error checking of invalid
6065 (MAX_CODE_ALIGN): Remove.
6066 (MAX_CODE_ALIGN_VALUE): Likewise.
6067 (parse_and_check_align_values): New function.
6068 (check_alignment_argument): Likewise.
6069 (common_handle_option): Use check_alignment_argument.
6070 * opts.h (parse_and_check_align_values): Declare.
6071 * toplev.c (init_alignments): Remove.
6072 (read_log_maxskip): New.
6073 (parse_N_M): Likewise.
6074 (parse_alignment_opts): Likewise.
6075 (backend_init_target): Remove usage of init_alignments.
6076 * toplev.h (parse_alignment_opts): Declare.
6077 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
6079 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
6080 * tree.c (cl_option_hasher::equal): New.
6081 * varasm.c: Use new global macros.
6083 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
6085 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
6086 Use a simpler align directive also if MAXSKIP = ALIGN-1.
6087 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6088 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6089 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6090 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
6091 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
6092 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
6093 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
6094 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6095 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6096 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6097 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6099 2018-07-04 Martin Liska <mliska@suse.cz>
6100 Jonathan Wakely <jwakely@redhat.com>
6102 * coverage.c: Use correct type.
6103 * doc/invoke.texi: Language correction.
6105 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
6108 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
6109 ENDBRANCH for non-tail call which may return via indirect branch.
6110 * doc/extend.texi: Document indirect_return attribute.
6112 2018-07-03 Martin Sebor <msebor@redhat.com>
6114 PR tree-optimization/86274
6115 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
6117 (format_floating): Correct handling of infinities and NaNs.
6119 2018-07-03 Martin Sebor <msebor@redhat.com>
6121 * print-tree.c (print_real_cst): New function.
6122 (print_node_brief): Call it.
6123 (print_node): Ditto.
6125 2018-07-03 Jeff Law <law@redhat.com>
6127 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
6128 into a single pattern.
6130 * config/h8300/h8300.md (ors code_iterator): New.
6131 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
6132 a single pattern and single splitter.
6133 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
6134 (iorqi3_1, xorqi3_1): Likewise.
6135 (iorqi3, xorqi3 expanders): Similarly.
6137 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
6138 (movmd_internal) into a single pattern using the P mode iterator.
6139 (movmd splitters): Similarly.
6140 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
6141 (movsd splitters): Similarly.
6143 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
6144 ADDB, ADDW and ADDL into a single ADD attribute which selects the
6145 right table based on the size of the operand.
6146 * config/h8300/h8300.md (length_table): Corresponding changes. All
6147 references to "addb", "addw" and "addl" changed to "add".
6148 (btst patterns): Merge two variants into a single pattern.
6149 (tstqi, tsthi): Likewise.
6150 (addhi3_incdec, addsi3_incdec): Likewise.
6151 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
6152 (mulhi3, mulsi3): Likewise.
6153 (udivhi3, udivsi3): Likewise.
6154 (divhi3, divsi3): Likewise.
6155 (andorqi3, andorhi3, andorsi3): Likewise.
6157 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
6160 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
6161 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
6163 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
6165 PR tree-optimization/85694
6166 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
6167 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
6169 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
6170 (<u>avg<mode>3_ceil): New patterns.
6172 2018-07-03 David Malcolm <dmalcolm@redhat.com>
6174 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
6175 scan-tree-dump directive.
6176 * gcc.dg/vect/slp-perm-2.c: Likewise.
6177 * gcc.dg/vect/slp-perm-3.c: Likewise.
6178 * gcc.dg/vect/slp-perm-5.c: Likewise.
6179 * gcc.dg/vect/slp-perm-6.c: Likewise.
6180 * gcc.dg/vect/slp-perm-7.c: Likewise.
6181 * gcc.dg/vect/slp-perm-8.c: Likewise.
6183 2018-07-03 Marek Polacek <polacek@redhat.com>
6186 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
6188 2018-07-03 Richard Biener <rguenther@suse.de>
6191 * tree-ssa-structalias.c (find_func_clobbers): Properly
6192 handle indirect calls.
6194 2018-07-03 Jeff Law <law@redhat.com>
6196 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
6197 (shifts): New code iterator.
6198 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
6199 expander. Fix HImode handling on H8/SX.
6200 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
6201 (subqi3, subhi3, subsi3 expanders): Likewise.
6202 (andqi3, andhi3, andsi3 expanders): Likewise.
6203 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
6204 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
6205 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
6206 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
6207 (zero_extendqihi2, zero_extendqisi2): Likewise.
6208 (extendqihi2, extendqisi2): Likewise.
6209 (rotlqi3, rotlhi3, rotlsi3): Likewise.
6210 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
6211 (rotlqi3_1, rotlhi3_1): Likewise.
6212 (logicalhi3_sn, logicalsi3_sn): Likewise.
6213 (logicalhi3, logicalsi3): Likewise.
6215 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
6217 * tree-vect-patterns.c (vect_recog_rotate_pattern)
6218 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
6219 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
6220 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
6221 type to append_pattern_def_seq instead of creating a stmt_vec_info
6223 (build_mask_conversion): Likewise. Remove vinfo argument.
6224 (vect_add_conversion_to_patterm): Likewise, renaming to...
6225 (vect_add_conversion_to_pattern): ...this.
6226 (vect_recog_mask_conversion_pattern): Update call to
6227 build_mask_conversion. Pass the vector type to
6228 append_pattern_def_seq here too.
6229 (vect_recog_gather_scatter_pattern): Update call to
6230 vect_add_conversion_to_pattern.
6232 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
6234 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
6235 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
6236 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
6237 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
6238 STMT_VINFO_PATTERN_DEF_SEQ to null here.
6239 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
6240 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
6241 append_pattern_def_seq instead of new_pattern_def_seq.
6242 (vect_recog_divmod_pattern): Do both of the above.
6243 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
6246 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
6248 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
6249 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
6250 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
6251 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
6252 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
6253 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
6254 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
6255 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
6256 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
6257 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
6258 parameter with a single stmt_vec_info.
6259 (vect_recog_func_ptr): Likewise.
6260 (vect_recog_gather_scatter_pattern): Likewise, folding in...
6261 (vect_try_gather_scatter_pattern): ...this.
6262 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
6263 the stmt_vec_info of the statement to be matched. Don't clear
6264 STMT_VINFO_RELATED_STMT.
6265 (vect_pattern_recog): Update call accordingly.
6267 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
6269 PR tree-optimization/85694
6270 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
6271 (uavgM3_ceil): Document new optabs.
6272 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
6273 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
6275 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
6276 (savg_ceil_optab): New optabs.
6277 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
6278 (vect_vect_recog_func_ptrs): Add it.
6279 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
6280 constant directly from the associated lhs.
6282 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
6284 * tree-vect-patterns.c (vect_split_statement): New function.
6285 (vect_convert_input): Use it to try to split an existing cast.
6287 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
6289 * poly-int.h (print_hex): New function.
6290 * dumpfile.h (dump_dec, dump_hex): Declare.
6291 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
6292 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
6293 min_input_precision, operation_precision and operation_sign.
6294 * tree-vect-patterns.c (vect_get_range_info): New function.
6295 (vect_same_loop_or_bb_p, vect_single_imm_use)
6296 (vect_operation_fits_smaller_type): Delete.
6297 (vect_look_through_possible_promotion): Add an optional
6298 single_use_p parameter.
6299 (vect_recog_over_widening_pattern): Rewrite to use new
6300 stmt_vec_info infomration. Handle one operation at a time.
6301 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
6302 (vect_truncatable_operation_p, vect_set_operation_type)
6303 (vect_set_min_input_precision): New functions.
6304 (vect_determine_min_output_precision_1): Likewise.
6305 (vect_determine_min_output_precision): Likewise.
6306 (vect_determine_precisions_from_range): Likewise.
6307 (vect_determine_precisions_from_users): Likewise.
6308 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
6309 (vect_vect_recog_func_ptrs): Put over_widening first.
6311 (vect_pattern_recog): Call vect_determine_precisions.
6313 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
6315 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
6316 statements that have been replaced by further pattern statements.
6317 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
6319 2018-07-03 Richard Biener <rguenther@suse.de>
6321 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
6322 always set *dt. Dump vectype in vectype overload.
6323 * dumpfile.h (dump_gimple_expr): New function.
6324 (dump_gimple_expr_loc): Likewise.
6325 * dumpfile.c (dump_gimple_expr): New function.
6326 (dump_gimple_expr_loc): Likewise.
6328 2018-07-02 Jeff Law <law@redhat.com>
6330 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
6331 the H8/300, H8/300H and H8/S variants into a single pattern.
6332 (movhi_h8300, movqi_h8300hs): Similarly.
6333 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
6334 (QHI mode iterator): New.
6336 * config/h8300/h8300.md: Remove trailing whitespace.
6338 2018-07-02 Jim Wilson <jimw@sifive.com>
6340 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
6341 instead of emit_insn for interrupt returns.
6342 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
6343 (riscv_sret, riscv_uret): Likewise.
6345 2018-07-02 David Malcolm <dmalcolm@redhat.com>
6347 * pretty-print.c (selftest::test_pp_format): Move save and restore
6348 of quotes to class auto_fix_quotes, and add an instance.
6349 * selftest.c: Include "intl.h".
6350 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
6351 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
6352 * selftest.h (selftest::auto_fix_quotes): New class.
6354 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
6356 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
6357 (aarch64_sve_prepare_conditional_op): Remove.
6358 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
6359 Allow aarch64_simd_reg_or_zero as select operand; remove
6360 the aarch64_sve_prepare_conditional_op call.
6361 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
6362 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
6363 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
6364 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
6365 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
6366 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
6367 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
6368 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
6369 and a splitters to match all of the *_any patterns.
6370 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
6372 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
6373 (SVE_COND_FP_BINARY_REV): Remove.
6374 (sve_int_op_rev, sve_fp_op_rev): New.
6375 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
6376 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
6377 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
6378 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
6379 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
6380 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
6381 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
6382 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
6383 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
6385 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
6386 Remove match_dup 1 from the inner unspec.
6387 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
6389 * config/aarch64/aarch64.md (movprfx): New attr.
6390 (length): Default movprfx to 8.
6391 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
6392 (*madd<SVE_I>, *msub<SVE_I): Likewise.
6393 (*<su>mul<SVE_I>3_highpart): Likewise.
6394 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
6395 (*v<ASHIFT><SVE_I>3): Likewise.
6396 (*<su><MAXMIN><SVE_I>3): Likewise.
6397 (*<su><MAXMIN><SVE_F>3): Likewise.
6398 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
6399 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
6400 (*div<SVE_F>4): Likewise.
6402 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
6404 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
6407 2018-07-02 Richard Biener <rguenther@suse.de>
6409 PR tree-optimization/86363
6410 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
6411 memset argument refers to a non-variable address.
6413 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
6415 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
6416 {PLUS,MINUS}_EXPR code to...
6417 (adjust_symbolic_bound): ...here,
6418 (combine_bound): ...here,
6419 (set_value_range_with_overflow): ...and here.
6421 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
6423 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
6425 (extract_range_from_abs_expr): ...here.
6427 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
6429 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
6430 -fno-omit-frame-pointer when not optimizing.
6432 2018-07-02 Martin Liska <mliska@suse.cz>
6435 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
6436 (propagate_nothrow): Likewise.
6438 2018-07-02 Martin Liska <mliska@suse.cz>
6441 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
6443 2018-07-02 David Malcolm <dmalcolm@redhat.com>
6445 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
6446 function in r262149, changing "loc" param from source_location to
6447 const dump_location_t &.
6448 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
6449 declaration, as above.
6451 2018-07-01 Paul Koning <ni1d@arrl.net>
6453 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
6454 -munit-asm, -mgnu-asm, -mdec-asm.
6455 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
6456 (pdp11_output_labelref): New.
6457 (pdp11_output_def): New.
6458 (pdp11_output_addr_vec_elt): New.
6459 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
6460 %# and %@ format codes.
6461 (pdp11_option_override): New.
6462 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
6463 (pdp11_output_ident): New.
6464 (pdp11_asm_named_section): New.
6465 (pdp11_asm_init_sections): New.
6466 (pdp11_file_start): New.
6467 (pdp11_file_end): New.
6468 (output_ascii): Use .ascii/.asciz for -mdec-asm.
6469 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
6470 %o, like %c but octal.
6471 (pdp11_option_override): New.
6472 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
6474 (DATA_SECTION_ASM_OP): Ditto.
6475 (READONLY_DATA_SECTION_ASM_OP): New.
6476 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
6477 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
6478 (ASM_OUTPUT_LABELREF): Ditto.
6479 (ASM_OUTPUT_DEF): Ditto.
6480 (ASM_OUTPUT_EXTERNAL): New.
6481 (ASM_OUTPUT_SOURCE_FILENAME): New.
6482 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
6483 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
6484 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
6485 %# and %@ format codes.
6486 * config/pdp11/pdp11.opt (mgnu-asm): New.
6487 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
6488 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
6489 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
6491 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
6493 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
6494 dereferencing path[] beyond its length.
6496 (debug_all_paths): New.
6497 (rewire_first_differing_edge): New.
6498 (adjust_paths_after_duplication): New.
6499 (duplicate_thread_path): Call adjust_paths_after_duplication.
6501 (thread_through_all_blocks): Add new argument to
6502 duplicate_thread_path.
6504 2018-06-30 Jim Wilson <jimw@sifive.com>
6506 * config/riscv/predicates.md (p2m1_shift_operand): New.
6507 (high_mask_shift_operand): New.
6508 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
6509 pattern using p2m1_shift_operand.
6510 (lshsi3_zero_extend_3+2): New combiner pattern using
6511 high_mask_shift_operand.
6513 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
6515 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
6517 (vect_recog_rotate_pattern): ...here.
6518 (vect_convert_input): Try to insert casts of invariants in the
6520 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
6521 preheader to be empty.
6523 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
6525 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
6526 vector type. If given, install it in the new statement's
6528 (vect_element_precision): New function.
6529 (vect_unpromoted_value): New struct.
6530 (vect_unpromoted_value::vect_unpromoted_value): New function.
6531 (vect_unpromoted_value::set_op): Likewise.
6532 (vect_look_through_possible_promotion): Likewise.
6533 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
6534 (vect_widened_op_tree, vect_convert_input): Likewise.
6535 (vect_convert_inputs, vect_convert_output): Likewise.
6536 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
6537 to handle the optional cast of the multiplication result and
6538 vect_widened_op_tree to detect the widened multiplication itself.
6539 Do not require the input and output of promotion casts to have
6540 the same sign, but base the signedness of the operation on the
6541 input rather than the result. If the pattern includes two
6542 promotions, check that those promotions have the same sign.
6543 Do not restrict the MULT_EXPR handling to a double-width result;
6544 handle quadruple-width results and wider. Use vect_convert_inputs
6545 to convert the inputs to the common type.
6546 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
6547 to handle the optional cast of the ABS result. Also allow a sign
6548 change or a sign extension between the ABS and MINUS.
6549 Use vect_widened_op_tree to detect the widened subtraction and use
6550 vect_convert_inputs to convert the inputs to the common type.
6551 (vect_handle_widen_op_by_const): Delete.
6552 (vect_recog_widen_op_pattern): New function.
6553 (vect_recog_widen_mult_pattern): Use it.
6554 (vect_recog_widen_shift_pattern): Likewise.
6555 (vect_recog_widen_sum_pattern): Use
6556 vect_look_through_possible_promotion to handle the promoted
6559 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
6561 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
6562 the containing gimple_seq *.
6563 * gimple-iterator.h (gsi_for_stmt): Declare it.
6564 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
6565 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
6566 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
6567 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
6568 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
6570 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
6572 (vect_mark_pattern_stmts): ...here. Handle cases in which the
6573 statement being replaced is part of an existing pattern
6574 definition sequence, inserting the new pattern statements before
6576 (vect_pattern_recog_1): Don't return a bool. If the statement
6577 is already part of a pattern, instead apply pattern matching
6578 to the pattern definition statements. Don't clear the
6579 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
6580 (vect_pattern_recog): Don't break after the first match;
6581 continue processing the pattern definition statements instead.
6582 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
6584 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
6586 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
6587 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
6588 (vect_recog_widen_sum_pattern): Use it.
6590 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
6592 * tree-vect-loop.c (vectorizable_reduction): Assert that the
6593 phi is not a pattern statement and has not been replaced by
6594 a pattern statement.
6595 * tree-vect-patterns.c (type_conversion_p): Don't check
6596 STMT_VINFO_IN_PATTERN_P.
6597 (vect_recog_vector_vector_shift_pattern): Likewise.
6598 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
6599 the pattern statement rather than the original statement; check
6600 directly for a WIDEN_MULT_EXPR here.
6601 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
6602 vect_is_simple_use to return the pattern statement rather
6603 than the original statement; use is_pattern_stmt_p to check
6604 for such a pattern statement.
6605 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
6606 to return the pattern statement rather than the original statement;
6607 don't do the same transformation here.
6608 (vect_is_simple_use): If the defining statement has been replaced
6609 by a pattern statement, return the pattern statement instead.
6610 Remove the corresponding (local) transformation from the vectype
6613 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
6615 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
6616 end and default to null.
6617 * tree-vect-loop.c (vect_create_epilog_for_reduction)
6618 (vectorizable_reduction): Update calls accordingly, dropping the
6619 gimple ** argument if the passed-back statement isn't needed.
6620 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
6621 (vect_recog_rotate_pattern): Likewise.
6622 (vect_recog_mask_conversion_pattern): Likewise.
6623 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
6624 (vect_mask_constant_operand_p): Likewise.
6625 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
6626 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
6627 (get_group_load_store_type, get_load_store_type): Likewise.
6628 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
6629 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
6630 (vectorizable_conversion, vectorizable_assignment): Likewise.
6631 (vectorizable_shift, vectorizable_operation): Likewise.
6632 (vectorizable_store, vect_is_simple_cond): Likewise.
6633 (vectorizable_condition, vectorizable_comparison): Likewise.
6634 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
6635 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
6636 and move it to the end. Cope with null def_stmt_outs.
6638 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
6640 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
6642 2018-06-29 Jeff Law <law@redhat.com>
6644 * config/v850/v850.c (v850_legitimate_address_p): Handle large
6645 displacements for TARGET_V850E2V3 and newer.
6646 (TARGET_LRA_P): Remove. Defaults to LRA now.
6647 * config/v850/v850.md (sign23byte_load): Remove.
6648 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
6649 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
6651 2018-06-29 Martin Liska <mliska@suse.cz>
6654 * coverage.c (coverage_init): Mangle full path name.
6655 * doc/invoke.texi: Document the change.
6656 * gcov-io.c (mangle_path): New.
6657 * gcov-io.h (mangle_path): Likewise.
6658 * gcov.c (mangle_name): Use mangle_path for path mangling.
6660 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6662 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
6663 if starting source register is not even.
6665 2018-06-29 Martin Liska <mliska@suse.cz>
6667 PR tree-optimization/86263
6668 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
6669 Make edge redirection.
6671 2018-06-29 David Malcolm <dmalcolm@redhat.com>
6673 * dumpfile.c (dump_loc): Add indentation based on scope depth.
6674 (dump_scope_depth): New variable.
6675 (get_dump_scope_depth): New function.
6676 (dump_begin_scope): New function.
6677 (dump_end_scope): New function.
6678 * dumpfile.h (get_dump_scope_depth): New declaration.
6679 (dump_begin_scope): New declaration.
6680 (dump_end_scope): New declaration.
6681 (class auto_dump_scope): New class.
6682 (AUTO_DUMP_SCOPE): New macro.
6683 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
6686 2018-06-29 Richard Biener <rguenther@suse.de>
6688 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
6689 compute_all_dependences succeeds.
6690 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
6691 exceed --param loop-max-datarefs-for-datadeps.
6693 2018-06-29 Jakub Jelinek <jakub@redhat.com>
6695 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
6697 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
6700 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
6701 alternative 0 in preferred_for_speed attribute.
6703 2018-06-28 Paul Koning <ni1d@arrl.net>
6705 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
6706 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
6707 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
6708 * config/pdp11/pdp11.md: Correct "length" attribute calculation
6709 for shift insn patterns.
6711 2018-06-28 David Malcolm <dmalcolm@redhat.com>
6713 * cgraph.c (cgraph_node::get_body): Replace assignments to
6714 "dump_file" with calls to set_dump_file.
6715 * dumpfile.c (alt_dump_file): Make static, and group with...
6716 (alt_flags): ...this definition.
6717 (dumps_are_enabled): New variable.
6718 (refresh_dumps_are_enabled): New function.
6719 (set_dump_file): New function.
6720 (set_alt_dump_file): New function.
6721 (gcc::dump_manager::dump_start): Replace assignments to
6722 "dump_file" and "alt_dump_file" with calls to set_dump_file and
6724 (gcc::dump_manager::dump_finish): Likewise.
6725 * dumpfile.h (alt_dump_file): Delete decl.
6726 (dumps_are_enabled): New variable decl.
6727 (set_dump_file): New function decl.
6728 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
6730 * tree-nested.c (lower_nested_functions): Replace assignments to
6731 "dump_file" with calls to set_dump_file.
6733 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
6735 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
6736 goto_locus of each outgoing edge of each basic block.
6738 2018-06-28 Richard Biener <rguenther@suse.de>
6740 * dwarf2out.c (decl_scope_table): Remove.
6741 (push_decl_scope): Likewise.
6742 (pop_decl_scope): Likewise.
6743 (gen_type_die_for_member): Do not call push/pop_decl_scope.
6744 (gen_struct_or_union_type_die): Likewise.
6745 (gen_tagged_type_die): Likewise.
6746 (dwarf2out_init): Do not initialize decl_scope_table.
6747 (dwarf2out_c_finalize): Do not free it.
6749 2018-06-28 Richard Biener <rguenther@suse.de>
6751 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
6752 deciding whether to not re-use a DIE.
6754 2018-06-28 Richard Biener <rguenther@suse.de>
6756 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
6757 DW_AT_abstract_origin attribute.
6759 2018-06-28 Martin Liska <mliska@suse.cz>
6761 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
6762 Use newly introduced constants.
6763 * tree-switch-conversion.h (struct jump_table_cluster):
6764 Define max_ratio_for_size and max_ratio_for_speed.
6766 2018-06-28 Martin Liska <mliska@suse.cz>
6768 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
6769 Add new checking assert to catch invalid state.
6770 (jump_table_cluster::can_be_handled): Handle single case
6772 (jump_table_cluster::is_beneficial): Bail out for such case.
6773 (bit_test_cluster::find_bit_tests):
6774 Add new checking assert to catch invalid state.
6775 (bit_test_cluster::can_be_handled): Handle single case
6777 (bit_test_cluster::is_beneficial): Bail out for such case.
6778 (switch_decision_tree::analyze_switch_statement):
6781 2018-06-28 Martin Liska <mliska@suse.cz>
6783 * common.opt: Introduce -completion option.
6784 * gcc.c (driver_handle_option): Handle it.
6785 (driver::main): Print completions if completion
6787 * opt-suggestions.c (option_proposer::get_completions):
6789 (option_proposer::suggest_completion): Likewise.
6790 (option_proposer::find_param_completions): Likewise.
6791 (verify_autocompletions): Likewise.
6792 (test_completion_valid_options): Likewise.
6793 (test_completion_valid_params): Likewise.
6794 (in_completion_p): Likewise.
6795 (empty_completion_p): Likewise.
6796 (test_completion_partial_match): Likewise.
6797 (test_completion_garbage): Likewise.
6798 (opt_proposer_c_tests): Likewise.
6799 * opt-suggestions.h: Declare new functions.
6800 * opts.c (common_handle_option): Handle OPT__completion_.
6801 * selftest-run-tests.c (selftest::run_tests): Add
6802 opt_proposer_c_tests.
6803 * selftest.c (assert_str_startswith): New.
6804 * selftest.h (assert_str_startswith): Likewise.
6805 (opt_proposer_c_tests): New.
6806 (ASSERT_STR_STARTSWITH): Likewise.
6808 2018-06-28 Martin Liska <mliska@suse.cz>
6810 * Makefile.in: Add opt-suggestions.o.
6811 * gcc-main.c: Include opt-suggestions.h.
6812 * gcc.c (driver::driver): Likewise.
6813 (driver::~driver): Remove m_option_suggestions.
6814 (driver::build_option_suggestions): Moved to option_proposer.
6815 (driver::suggest_option): Likewise.
6816 (driver::handle_unrecognized_options): Use option_proposer.
6817 * gcc.h (class driver): Add new memver m_option_proposer.
6818 * opt-suggestions.c: New file.
6819 * opt-suggestions.h: New file.
6821 2018-06-28 Martin Liska <mliska@suse.cz>
6823 * vec.h (class auto_string_vec): New (moved from auto_argvec).
6824 (auto_string_vec::~auto_string_vec): Likewise.
6826 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
6828 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
6829 prevent_decl_creation_for_types fields up and add reset_location field.
6830 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
6831 statement if id->reset_location is true.
6832 (copy_edges_for_bb): Do not set goto_locus on the new edges if
6833 id->reset_location is true.
6834 (copy_phis_for_bb): Force input_location on the arguments if
6835 id->reset_location is true.
6836 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
6837 is set on the function to be inlined.
6839 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
6841 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
6843 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
6845 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
6846 registers for Pmode.
6847 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
6848 hard registers for the clobbered pseudo.
6850 2018-06-27 Paul Koning <ni1d@arrl.net>
6852 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
6853 mutually exclusive options.
6854 * config/pdp11/constraints.md (h): New constraint.
6855 (O): Update definition to match shift code generation.
6856 (D): New constraint.
6857 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
6859 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
6861 (output_jump): Change arguments.
6862 (pdp11_fixed_cc_regs): New function.
6863 (pdp11_cc_mode): Ditto.
6864 (pdp11_expand_shift): Ditto.
6865 (pdp11_assemble_shift): Ditto.
6866 (pdp11_small_shift): Ditto.
6867 (pdp11_branch_cost): Remove.
6868 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
6870 (pdp11_register_move_cost): Update for CC registers.
6871 (pdp11_rtx_costs): Add case for LSHIFTRT.
6872 (pdp11_output_jump): Add CCNZ mode conditional branches.
6873 (notice_update_cc_on_set): Remove.
6874 (pdp11_cc_mode): New function.
6875 (simple_memory_operand): Correct pre/post decrement case.
6876 (no_side_effect_operand): New function.
6877 (pdp11_regno_reg_class): Add CC_REGS class.
6878 (pdp11_fixed_cc_regs): New function.
6879 (pdp11_small_shift): New function.
6880 (pdp11_expand_shift): New function to expand shift insns.
6881 (pdp11_assemble_shift): New function to output shifts.
6882 (pdp11_branch_cost): Remove.
6883 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
6884 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
6885 (WCHAR_TYPE): Ditto.
6886 (PTRDIFF_TYPE): Ditto.
6887 (ADJUST_INSN_LENGTH): New macro.
6888 (FIXED_REGISTERS): Add CC registers.
6889 (CALL_USED_REGISTERS): Ditto.
6891 (REG_CLASS_NAMES): Ditto.
6892 (REG_CLASS_CONTENTS): Ditto.
6893 (SELECT_CC_MODE): Use new function.
6894 (TARGET_FLAGS_REGNUM): New macro.
6895 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
6896 (cc0_reg_rtx): Remove.
6897 (CC_STATUS_MDEP): Remove.
6898 (CC_STATUS_MDEFP_INIT): Remove.
6899 (CC_IN_FPU): Remove.
6900 (NOTICE_UPDATE_CC): Remove.
6901 (REGISTER_NAMES): Add CC registers.
6902 (BRANCH_COST): Change to constant 1.
6903 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
6905 * config/pdp11/pdp11.opt (mbcopy): Remove.
6906 (mbcopy-builtin): Remove.
6907 (mbranch-cheap): Remove.
6908 (mbranch-expensive): Remove.
6909 * config/pdp11/predicates.md (expand_shift_operand): Update to
6910 match shift code generation.
6911 (ccnz_operator): New predicate.
6912 * doc/invoke.texi (PDP-11 Options): Remove deleted options
6913 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
6914 Remove non-existent option -mabshi, -mno-abshi. Document mutually
6916 * doc/md.texi (PDP-11): Document new D and h constraints. Update
6917 description of O constraint.
6919 2018-06-27 Jeff Law <law@redhat.com>
6920 Austin Law <austinklaw@gmail.com>
6922 * config/v850/v850.md (addsi3_set_flags): New pattern.
6923 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
6924 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
6925 (zero_extendhisi2_v850_set_flags): Likewise.
6926 (zero_extendqisi2_v850_set_flags): Likewise.
6927 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
6928 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
6929 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
6931 * config/v850/v850-protos.h (notice_update_cc): Remove.
6932 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
6933 (v850_print_operand): Handle 'D' and "d".
6934 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
6935 Add handling of arithmetic/logical operations compared against zero.
6936 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
6937 Do not look at v850_compare_op, instead get mode from last argument.
6938 (v850_gen_compare): Remove
6939 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
6940 after reload for prologue insns.
6941 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
6943 (construct_save_jarl): Likewise.
6944 (TARGET_FLAGS_REGNUM): Define.
6945 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
6946 (NOTICE_UPDATE_CC): Remove.
6947 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
6948 than cc0. Conditionalize on reload_completed.
6949 (cmpsi_insn, setfcc_insn): Likewise.
6950 (tst1 splitter): Turn into define_and_split which sets the flags
6952 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
6953 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
6954 (cstoresf4, cstoredf4): Clobber the flags.
6955 (cmpsi, cmpsf, cmpdf): Remove expanders.
6956 (setf_insn): Remove pattern.
6957 (addsi3): Turn into define_and_split which clobbers the flags after
6958 reload and a suitable pattern (addsi3_clobber_flags) for use after
6960 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
6961 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
6962 (ashrsi3, ashrsi3_v850e2): Likewise.
6963 (bins): Clobber the flags.
6964 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
6965 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
6966 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
6967 (call_value_internal_short, call_value_internal_long): Likewise.
6968 (callt_save_interrupt, callt_return_interrupt): Likewise.
6969 (save_interrupt, return_interrupt): Likewise.
6970 (callt_save_all_interrupt, save_all_interrupt): Likewise.
6971 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
6972 (restore_all_interrupt, _restore_all_interrupt): Likewise.
6973 (All FP comparisons): Only allow after reload has completed.
6975 (divh, divhu): Tweak output template.
6976 (branch_z_normal, branch_z_invert): Remove
6977 (branch_nz_normal, branch_nz_invert): Likewise.
6978 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
6980 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
6981 * config/v850/v850.c (notice_update_cc): Remove.
6982 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
6983 (CC_NO_CARRY): Likewise.
6984 (NOTICE_UPDATE_CC): Define to nothing.
6985 * config/v850/v850.md: Remove block comment on cc0 handling
6986 Remove "cc" attribute from all patterns. Remove cc_status handling
6987 from all patterns. Minor formatting fixes.
6989 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6991 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
6992 (cortex-a76.cortex-a55): Likewise.
6993 * config/aarch64/aarch64-tune.md: Regenerate.
6994 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
6995 cortex-a76.cortex-a55.
6997 2018-06-27 Jeff Law <law@redhat.com>
6999 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
7000 (MULTILIB_DIRNAMES): Similarly.
7002 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
7004 * gimple.h (gimple_return_retbnd): Delete.
7005 (gimple_return_set_retbnd): Likewise.
7006 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
7007 gimple_return_set_retbnd.
7008 * gimple-pretty-print.c (dump_gimple_return): Remove call to
7009 gimple_return_retbnd and adjust.
7010 * tree-inline.h (struct copy_body_data): Remove retbnd field.
7011 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
7012 Explicitly return NULL in a couple more cases. Move assertion
7013 on debug statements and remove unreachable code.
7014 (reset_debug_binding): Do not test id->retbnd.
7015 (expand_call_inline): Do not set it.
7017 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
7019 * configure.ac: Add --disable-gcov option.
7020 * configure: Regenerate.
7021 * Makefile.in: Honour @enable_gcov@.
7022 * doc/install.texi: Document --disable-gcov.
7024 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7026 * config/arm/arm-cpus.in (cortex-a76): New entry.
7027 (cortex-a76.cortex-a55): Likewise.
7028 * config/arm/arm-tables.opt: Regenerate.
7029 * config/arm/arm-tune.md: Likewise.
7030 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
7031 * doc/invoke.texi (ARM Options): Document cortex-a76 and
7032 cortex-a76.cortex-a55.
7034 2018-06-27 Tamar Christina <tamar.christina@arm.com>
7037 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
7039 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
7041 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
7043 (EPILOGUE_USES): Likewise.
7045 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
7047 * tree-inline.c (remap_location): New function extracted from...
7048 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
7049 (copy_phis_for_bb): ...here. Call remap_location.
7050 (copy_cfg_body): Adjust call to copy_edges_for_bb.
7052 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
7054 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
7055 unaligned vsx for 16B memset.
7057 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
7060 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
7061 ieee128_float_type_node to long_double_type_node unless
7062 TARGET_LONG_DOUBLE_128 is set.
7064 2018-06-26 David Malcolm <dmalcolm@redhat.com>
7066 * cfgloop.c (get_loop_location): Convert return type from
7067 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
7068 by implicit construction from rtx_insn *, and using
7069 dump_user_location_t::from_function_decl for the fallback case.
7070 * cfgloop.h (get_loop_location): Convert return type from
7071 location_t to dump_user_location_t.
7072 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
7073 dump_printf_loc to pass in a dump_location_t rather than a
7074 location_t, via the gimple stmt.
7075 * coverage.c (get_coverage_counts): Update calls to
7076 dump_printf_loc to pass in dump_location_t rather than a
7078 * doc/optinfo.texi (Dump types): Convert example of
7079 dump_printf_loc from taking "locus" to taking "insn". Update
7080 description of the "_loc" calls to cover dump_location_t.
7081 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
7083 (dump_user_location_t::dump_user_location_t): New constructors,
7084 from gimple *stmt and rtx_insn *.
7085 (dump_user_location_t::from_function_decl): New function.
7086 (dump_loc): Make static.
7087 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
7088 const dump_location_t &.
7089 (dump_generic_expr_loc): Delete.
7090 (dump_printf_loc): Convert param "loc" from location_t to
7091 const dump_location_t &.
7092 (selftest::test_impl_location): New function.
7093 (selftest::dumpfile_c_tests): New function.
7094 * dumpfile.h: Include "profile-count.h".
7095 (class dump_user_location_t): New class.
7096 (struct dump_impl_location_t): New struct.
7097 (class dump_location_t): New class.
7098 (dump_printf_loc): Convert 2nd param from source_location to
7099 const dump_location_t &.
7100 (dump_generic_expr_loc): Delete.
7101 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
7102 const dump_location_t &.
7103 * gimple-fold.c (fold_gimple_assign): Update call to
7104 dump_printf_loc to pass in a dump_location_t rather than a
7105 location_t, via the gimple stmt.
7106 (gimple_fold_call): Likewise.
7107 * gimple-loop-interchange.cc
7108 (loop_cand::analyze_iloop_reduction_var): Update for change to
7109 check_reduction_path.
7110 (tree_loop_interchange::interchange): Update for change to
7112 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
7113 change in return-type of find_loop_location.
7114 (graphite_regenerate_ast_isl): Likewise.
7115 * graphite-optimize-isl.c (optimize_isl): Likewise.
7116 * graphite.c (graphite_transform_loops): Likewise.
7117 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
7118 pass in a dump_location_t rather than a location_t, via the
7120 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
7121 * ipa.c (walk_polymorphic_call_targets): Likewise.
7122 * loop-unroll.c (report_unroll): Convert "locus" param from
7123 location_t to dump_location_t.
7124 (decide_unrolling): Update for change to get_loop_location's
7126 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
7127 location_t to dump_user_location_t.
7128 (grid_find_single_omp_among_assignments_1): Updates calls to
7129 dump_printf_loc to pass in a dump_location_t rather than a
7130 location_t, via the gimple stmt.
7131 (grid_parallel_clauses_gridifiable): Convert "tloc" from
7132 location_t to dump_location_t. Updates calls to dump_printf_loc
7133 to pass in a dump_location_t rather than a location_t, via the
7135 (grid_inner_loop_gridifiable_p): Likewise.
7136 (grid_dist_follows_simple_pattern): Likewise.
7137 (grid_gfor_follows_tiling_pattern): Likewise.
7138 (grid_target_follows_gridifiable_pattern): Likewise.
7139 (grid_attempt_target_gridification): Convert initialization
7140 of local "grid" from memset to zero-initialization; FIXME: does
7141 this require C++11? Update call to dump_printf_loc to pass in a
7142 optinfo_location rather than a location_t, via the gimple stmt.
7143 * profile.c (read_profile_edge_counts): Updates call to
7144 dump_printf_loc to pass in a dump_location_t rather than a
7146 (compute_branch_probabilities): Likewise.
7147 * selftest-run-tests.c (selftest::run_tests): Call
7149 * selftest.h (dumpfile_c_tests): New decl.
7150 * tree-loop-distribution.c (pass_loop_distribution::execute):
7151 Update for change in return type of find_loop_location.
7152 * tree-parloops.c (parallelize_loops): Likewise.
7153 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
7154 "locus" from location_t to dump_user_location_t.
7155 (canonicalize_loop_induction_variables): Likewise.
7156 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
7157 for change in return type of find_loop_location.
7158 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
7159 to dump_printf_loc to pass in a dump_location_t rather than a
7160 location_t, via the stmt.
7161 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
7163 * tree-vect-loop-manip.c (find_loop_location): Convert return
7164 type from source_location to dump_user_location_t.
7165 (vect_do_peeling): Update for above change.
7166 (vect_loop_versioning): Update for change in type of
7168 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
7169 from location_t to dump_user_location_t.
7170 (vect_estimate_min_profitable_iters): Update for change in type
7172 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
7173 location_t to dump_location_t.
7174 (vect_slp_bb): Update for change in type of vect_location.
7175 * tree-vectorizer.c (vect_location): Convert from source_location
7176 to dump_user_location_t.
7177 (try_vectorize_loop_1): Update for change in vect_location's type.
7178 (vectorize_loops): Likewise.
7179 (increase_alignment): Likewise.
7180 * tree-vectorizer.h (vect_location): Convert from source_location
7181 to dump_user_location_t.
7182 (find_loop_location): Convert return type from source_location to
7183 dump_user_location_t.
7184 (check_reduction_path): Convert 1st param from location_t to
7185 dump_user_location_t.
7186 * value-prof.c (check_counter): Update call to dump_printf_loc to
7187 pass in a dump_user_location_t rather than a location_t; update
7188 call to error_at for change in type of "locus".
7189 (check_ic_target): Update call to dump_printf_loc to
7190 pass in a dump_user_location_t rather than a location_t, via the
7193 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
7195 * config/s390/s390.h (enum processor_flags): Do not use
7196 default tune parameter when -march was specified.
7198 2018-06-26 Jakub Jelinek <jakub@redhat.com>
7201 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
7202 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
7205 2018-06-26 Richard Biener <rguenther@suse.de>
7207 PR tree-optimization/86287
7209 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
7210 (vect_analyze_loop): Initialize n_stmts.
7212 2018-06-26 Richard Biener <rguenther@suse.de>
7215 * fold-const.c (fold_convertible_p): Pointer extension
7218 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
7221 * dwarf2out.c (loc_list_has_views): Adjust comments.
7222 (dw_loc_list): Split single cross-partition range with
7225 2018-06-25 Jeff Law <law@redhat.com>
7227 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
7228 on -mbig-switch by default.
7230 * config/v850/predicates.md (const_float_1_operand): Fix match_code
7232 (const_float_0_operand): Remove unused predicate.
7233 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
7234 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
7235 (recipsf2): New expander. Original pattern now called
7237 (recipdf2, recipdf2_insn): Similarly.
7238 (rsqrtsf2, rsqrtsf2_insn): Similarly
7239 (rsqrtdf2, rsqrtdf2_insn): Similarly
7241 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
7243 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
7244 Simplify logic for FreeBSD (twice).
7246 2018-06-25 Martin Sebor <msebor@redhat.com>
7248 PR tree-optimization/86204
7249 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
7250 a strnlen result if it's less than the length of the string.
7252 2018-06-25 Martin Sebor <msebor@redhat.com>
7254 PR tree-optimization/85700
7255 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
7256 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
7257 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
7259 2018-06-25 Martin Sebor <msebor@redhat.com>
7261 * doc/extend.texi (Zero-length arrays): Update and clarify.
7263 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
7265 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
7266 added IEEE/IBM long double multilib support on PowerPC little
7267 endian Linux systems.
7268 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
7269 (MULTILIB_DEFAULTS): Likewise.
7270 * config/rs6000/rs6000.c (rs6000_option_override_internal):
7272 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
7273 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
7274 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
7276 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
7279 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
7280 (REORDER_45): Likewise.
7282 2018-06-25 Jeff Law <law@redhat.com>
7284 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
7285 dividend to 32 bits. Adjust length.
7286 (udivmodhi4): Cleanup output template. Fix length.
7288 2018-06-25 Carl Love <cel@us.ibm.com>
7290 * config/rs6000/vsx.md: Change word selector to prefered location.
7292 2018-06-25 Richard Biener <rguenther@suse.de>
7294 PR tree-optimization/86304
7295 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
7296 epilogue-if-converted loops as well.
7298 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
7300 * lto-section-out.c (lto_begin_section): Do not print section
7301 name for noaddr and unnumbered dumps.
7303 2018-06-25 Richard Biener <rguenther@suse.de>
7305 * tree-vectorizer.h (struct vec_info_shared): New structure
7306 with parts split out from struct vec_info and loop_nest from
7307 struct _loop_vec_info.
7308 (struct vec_info): Adjust accordingly.
7309 (struct _loop_vec_info): Likewise.
7310 (LOOP_VINFO_LOOP_NEST): Adjust.
7311 (LOOP_VINFO_DATAREFS): Likewise.
7312 (LOOP_VINFO_DDRS): Likewise.
7313 (struct _bb_vec_info): Likewise.
7314 (BB_VINFO_DATAREFS): Likewise.
7315 (BB_VINFO_DDRS): Likewise.
7316 (struct _stmt_vec_info): Add dr_aux member.
7317 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
7318 (DR_MISALIGNMENT_UNINITIALIZED): New.
7319 (set_dr_misalignment): Adjust.
7320 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
7321 (vect_analyze_loop): Adjust prototype.
7322 (vect_analyze_loop_form): Likewise.
7323 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
7324 Compute dependences lazily.
7325 (vect_record_base_alignments): Use shared datarefs/ddrs.
7326 (vect_verify_datarefs_alignment): Likewise.
7327 (vect_analyze_data_refs_alignment): Likewise.
7328 (vect_analyze_data_ref_accesses): Likewise.
7329 (vect_analyze_data_refs): Likewise.
7330 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
7331 constructor parameter for shared part.
7332 (vect_analyze_loop_form): Pass in shared part and adjust.
7333 (vect_analyze_loop_2): Pass in storage for the number of
7334 stmts. Move loop nest finding to the caller. Compute
7336 (vect_analyze_loop): Pass in shared part.
7337 (vect_transform_loop): Verify shared datarefs are unchanged.
7338 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
7339 constructor parameter for shared part.
7340 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
7341 (vect_slp_bb): Verify shared datarefs are unchanged before
7343 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
7345 (new_stmt_vec_info): Initialize DR_AUX misalignment to
7346 DR_MISALIGNMENT_UNINITIALIZED.
7347 * tree-vectorizer.c (vec_info::vec_info): Add constructor
7348 parameter for shared part.
7349 (vec_info::~vec_info): Adjust.
7350 (vec_info_shared::vec_info_shared): New.
7351 (vec_info_shared::~vec_info_shared): Likewise.
7352 (vec_info_shared::save_datarefs): Likewise.
7353 (vec_info_shared::check_datarefs): Likewise.
7354 (try_vectorize_loop_1): Construct shared part live for analyses
7355 of a single loop for multiple vector sizes.
7356 * tree-parloops.c (gather_scalar_reductions): Adjust.
7358 2018-06-25 Richard Biener <rguenther@suse.de>
7360 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
7361 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
7362 (vect_analyze_data_refs): Remove similar code from here and
7363 simplify accordingly.
7365 2018-06-25 Richard Biener <rguenther@suse.de>
7367 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
7368 for reverse storage order accesses rather than asserting
7369 they cannot happen here.
7371 2018-06-25 Tom de Vries <tdevries@suse.de>
7374 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
7375 Use data16 instead of .byte for insn prefix.
7377 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
7380 * parser.c (make_char_string_pack): Pass this literal chars
7381 through cpp_interpret_string.
7382 (cp_parser_userdef_numeric_literal): Check the result of
7383 make_char_string_pack.
7385 2018-06-24 Maya Rashish <coypu@sdf.org>
7387 * ginclude/stddef.h: Simplify conditions around avoiding
7388 re-definition of __size_t.
7390 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
7392 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
7393 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
7395 2018-06-22 Maya Rashish <coypu@sdf.org>
7397 * doc/invoke.texi (mno-fancy-math-387): Update for changes
7398 made to OpenBSD and NetBSD through the years.
7400 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
7402 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
7403 behavior of vec_pack (vector double, vector double) to match
7404 behavior of vec_float2 (vector double, vector double).
7406 2018-06-22 Olivier Hainque <hainque@adacore.com>
7408 * gimplify.c (gimplify_function_tree): Prevent creation
7409 of a trampoline for the address of the current function
7410 passed to entry/exit instrumentation hooks.
7412 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
7415 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
7418 2018-06-22 Martin Liska <mliska@suse.cz>
7420 PR tree-optimization/86263
7421 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
7422 Bail out if is_enabled is false.
7423 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
7425 (jump_table_cluster::is_enabled): New function.
7427 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
7429 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
7430 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
7431 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
7432 (lto_input_ts_binfo_tree_pointers): Likewise.
7433 * tree-streamer-out.c (streamer_write_tree_bitfields,
7434 write_ts_binfo_tree_pointers): Likewise.
7435 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
7437 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
7439 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
7441 2018-06-22 Martin Liska <mliska@suse.cz>
7443 * symbol-summary.h (get): Make it pure and inline move
7444 functionality from ::get function.
7445 (get): Remove and inline into ::get and ::get_create.
7446 (get_create): Move code from ::get function.
7448 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7451 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
7452 -x assembler-with-cpp.
7454 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7456 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
7457 _FILE_OFFSET_BITS=64 for C++.
7459 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
7461 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
7462 conversion insn that shows up when pr85657-3.c is compiled using
7463 IEEE 128-bit long double.
7464 (neg<mode>2_internal): Use the correct mode to check whether the
7465 mode is IBM extended.
7466 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
7467 multiply and divide external functions from being created more
7470 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
7472 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
7474 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
7475 the edge can be forwarded.
7476 (cfg_layout_merge_blocks): Likewise.
7478 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
7480 * except.c (finish_eh_generation): Commit edge insertions only after
7481 the EH edges have been redirected from post-landing to landing pads.
7483 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
7485 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
7486 create_tmp_var_for to create the FRAME decl.
7487 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
7489 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
7491 * tree-inline.c (copy_edges_for_bb): Minor tweak.
7492 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
7493 debug statement when resetting its value.
7494 (expand_call_inline): Copy the locus of the call onto the assignment
7495 of the return value, if any. Use local variable in more cases.
7497 2018-06-21 Martin Liska <mliska@suse.cz>
7499 * ipa-pure-const.c (propagate_nothrow): Use
7500 funct_state_summaries->get.
7501 (dump_malloc_lattice): Likewise.
7502 (propagate_malloc): Likewise.
7504 2018-06-21 Richard Biener <rguenther@suse.de>
7506 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
7507 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
7508 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
7510 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
7511 BLOCK_ABSTRACT_ORIGIN unconditionally.
7513 2018-06-21 David Malcolm <dmalcolm@redhat.com>
7515 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
7517 * ipa-reference.c (ipa_reference_c_finalize): Delete
7518 ipa_ref_opt_sum_summaries and set it to NULL.
7520 2018-06-21 Tom de Vries <tdevries@suse.de>
7522 PR tree-optimization/85859
7523 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
7524 test with comment from bb_no_side_effects_p.
7526 2018-06-21 Richard Biener <rguenther@suse.de>
7528 PR tree-optimization/86232
7529 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
7530 max for constant niter.
7532 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
7534 * config/aarch64/aarch64-simd.md
7535 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
7537 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
7539 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
7540 Make opernads of the unspec commutative.
7542 2018-06-21 Richard Biener <rguenther@suse.de>
7544 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
7545 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
7546 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
7547 (vect_analyze_data_ref_dependence): Re-order checks to deal with
7549 (vect_record_base_alignments): Do not record base alignment
7550 for gathers or scatters.
7551 (vect_compute_data_ref_alignment): Drop return value that is always
7552 true. Bail out early for gathers or scatters.
7553 (vect_enhance_data_refs_alignment): Bail out early for gathers
7555 (vect_find_same_alignment_drs): Likewise.
7556 (vect_analyze_data_refs_alignment): Remove dead code.
7557 (vect_slp_analyze_and_verify_node_alignment): Likewise.
7558 (vect_analyze_data_refs): For possible gathers or scatters do
7559 not create an alternate DR, just check their possible validity
7560 and mark them. Adjust DECL_NONALIASED handling to not rely
7562 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
7563 update inits of gathers or scatters.
7564 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
7565 Also copy gather/scatter flag to pattern vinfo.
7567 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
7569 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
7570 behavior of vec_packsu (vector unsigned long long, vector unsigned
7571 long long) to match behavior of vec_packs with same signature.
7573 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
7574 Thomas Schwinge <thomas@codesourcery.com>
7575 Cesar Philippidis <cesar@codesourcery.com>
7577 * gimplify.c (gimplify_scan_omp_clauses): Add support for
7578 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
7579 (gimplify_adjust_omp_clauses): Likewise.
7580 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
7581 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
7582 (gimplify_omp_target_update): Update handling of acc update and
7584 * omp-low.c (install_var_field): Remove unused parameter
7585 base_pointers_restrict.
7586 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
7587 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
7589 (omp_target_base_pointers_restrict_p): Delete.
7590 (scan_omp_target): Update call to scan_sharing_clauses.
7591 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
7593 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
7594 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
7595 (convert_local_omp_clauses): Likewise.
7596 * tree-pretty-print.c (dump_omp_clause): Likewise.
7597 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
7599 (omp_clause_code_name): Likewise.
7601 2018-06-20 Jakub Jelinek <jakub@redhat.com>
7604 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
7607 PR tree-optimization/86231
7608 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
7609 anti-range don't overwrite *vr0min before using it to compute *vr0max.
7611 2018-06-20 Tom de Vries <tdevries@suse.de>
7613 PR tree-optimization/86097
7614 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
7615 iv type if signedness of iv type is not the same as that of *nit.
7617 2018-06-20 Jakub Jelinek <jakub@redhat.com>
7619 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
7620 EDGE_EH edges, verify they are all EDGE_EH.
7622 2018-06-20 Maya Rashish <coypu@sdf.org>
7624 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
7626 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7628 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
7629 * config/aarch64/aarch64.c (xgene1_tunings): Add
7630 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
7631 (aarch64_mode_valid_for_sched_fusion_p):
7632 Allow 16-byte modes.
7633 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
7634 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
7636 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
7638 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
7639 * config/aarch64/iterators.md (VQ2): New mode iterator.
7641 2018-06-20 Martin Liska <mliska@suse.cz>
7643 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
7644 Change default ratio from 10 to 8.
7646 2018-06-20 Martin Liska <mliska@suse.cz>
7648 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
7650 (bit_test_cluster::find_bit_tests): Likewise.
7651 (switch_decision_tree::analyze_switch_statement): Find clusters.
7652 * tree-switch-conversion.h (struct jump_table_cluster): Document
7655 2018-06-20 Martin Liska <mliska@suse.cz>
7657 * tree-switch-conversion.c (switch_conversion::collect):
7658 Record m_uniq property.
7659 (switch_conversion::expand): Bail out for special conditions.
7660 (group_cluster::~group_cluster): New.
7661 (group_cluster::group_cluster): Likewise.
7662 (group_cluster::dump): Likewise.
7663 (jump_table_cluster::emit): New.
7664 (switch_decision_tree::fix_phi_operands_for_edges): New.
7665 (struct case_node): Remove struct.
7666 (jump_table_cluster::can_be_handled): New.
7667 (case_values_threshold): Moved to header.
7668 (reset_out_edges_aux): Likewise.
7669 (jump_table_cluster::is_beneficial): New.
7670 (bit_test_cluster::can_be_handled): Likewise.
7671 (add_case_node): Remove.
7672 (bit_test_cluster::is_beneficial): New.
7673 (case_bit_test::cmp): New.
7674 (bit_test_cluster::emit): New.
7675 (expand_switch_as_decision_tree_p): Remove.
7676 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
7677 (fix_phi_operands_for_edge): Likewise.
7678 (switch_decision_tree::analyze_switch_statement): New.
7679 (compute_cases_per_edge): Move ...
7680 (switch_decision_tree::compute_cases_per_edge): ... here.
7681 (try_switch_expansion): Likewise.
7682 (switch_decision_tree::try_switch_expansion): Likewise.
7683 (record_phi_operand_mapping): Likewise.
7684 (switch_decision_tree::record_phi_operand_mapping): Likewise.
7685 (emit_case_decision_tree): Likewise.
7686 (switch_decision_tree::emit): Likewise.
7687 (balance_case_nodes): Likewise.
7688 (switch_decision_tree::balance_case_nodes): Likewise.
7689 (dump_case_nodes): Likewise.
7690 (switch_decision_tree::dump_case_nodes): Likewise.
7691 (emit_jump): Likewise.
7692 (switch_decision_tree::emit_jump): Likewise.
7693 (emit_cmp_and_jump_insns): Likewise.
7694 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
7695 (emit_case_nodes): Likewise.
7696 (switch_decision_tree::emit_case_nodes): Likewise.
7697 (conditional_probability): Remove.
7698 * tree-switch-conversion.h (enum cluster_type): New.
7700 (struct cluster): Likewise.
7701 (cluster::cluster): Likewise.
7702 (struct simple_cluster): Likewise.
7703 (simple_cluster::simple_cluster): Likewise.
7704 (struct group_cluster): Likewise.
7705 (struct jump_table_cluster): Likewise.
7706 (struct bit_test_cluster): Likewise.
7707 (struct min_cluster_item): Likewise.
7708 (struct case_tree_node): Likewise.
7709 (case_tree_node::case_tree_node): Likewise.
7710 (jump_table_cluster::case_values_threshold): Likewise.
7711 (struct case_bit_test): Likewise.
7712 (struct switch_decision_tree): Likewise.
7713 (struct switch_conversion): Likewise.
7714 (switch_decision_tree::reset_out_edges_aux): Likewise.
7716 2018-06-20 Martin Liska <mliska@suse.cz>
7718 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
7719 (hoist_edge_and_branch_if_true): Likewise.
7720 (expand_switch_using_bit_tests_p): Likewise.
7721 (struct case_bit_test): Likewise.
7722 (case_bit_test_cmp): Likewise.
7723 (emit_case_bit_tests): Likewise.
7724 (switch_conversion::switch_conversion): New class.
7725 (struct switch_conv_info): Remove old struct.
7726 (collect_switch_conv_info): More to ...
7727 (switch_conversion::collect): ... this.
7728 (check_range): Likewise.
7729 (switch_conversion::check_range): Likewise.
7730 (check_all_empty_except_final): Likewise.
7731 (switch_conversion::check_all_empty_except_final): Likewise.
7732 (check_final_bb): Likewise.
7733 (switch_conversion::check_final_bb): Likewise.
7734 (create_temp_arrays): Likewise.
7735 (switch_conversion::create_temp_arrays): Likewise.
7736 (free_temp_arrays): Likewise.
7737 (gather_default_values): Likewise.
7738 (switch_conversion::gather_default_values): Likewise.
7739 (build_constructors): Likewise.
7740 (switch_conversion::build_constructors): Likewise.
7741 (constructor_contains_same_values_p): Likewise.
7742 (switch_conversion::contains_same_values_p): Likewise.
7743 (array_value_type): Likewise.
7744 (switch_conversion::array_value_type): Likewise.
7745 (build_one_array): Likewise.
7746 (switch_conversion::build_one_array): Likewise.
7747 (build_arrays): Likewise.
7748 (switch_conversion::build_arrays): Likewise.
7749 (gen_def_assigns): Likewise.
7750 (switch_conversion::gen_def_assigns): Likewise.
7751 (prune_bbs): Likewise.
7752 (switch_conversion::prune_bbs): Likewise.
7753 (fix_phi_nodes): Likewise.
7754 (switch_conversion::fix_phi_nodes): Likewise.
7755 (gen_inbound_check): Likewise.
7756 (switch_conversion::gen_inbound_check): Likewise.
7757 (process_switch): Use the newly created class.
7758 (switch_conversion::expand): New.
7759 (switch_conversion::~switch_conversion): New.
7760 * tree-switch-conversion.h: New file.
7762 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
7764 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
7765 tree-vect-patterns.c.
7766 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
7767 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
7768 (vect_recog_sad_pattern): Likewise.
7769 (vect_recog_widen_sum_pattern): Likewise.
7770 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
7771 (vect_recog_widen_shift_pattern): Remove the type_in argument.
7772 (vect_recog_rotate_pattern): Likewise.
7773 (vect_recog_mult_pattern): Likewise.
7774 (vect_recog_vector_vector_shift_pattern): Likewise.
7775 (vect_recog_divmod_pattern): Likewise.
7776 (vect_recog_mixed_size_cond_pattern): Likewise.
7777 (vect_recog_bool_pattern): Likewise.
7778 (vect_recog_mask_conversion_pattern): Likewise.
7779 (vect_try_gather_scatter_pattern): Likewise.
7780 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
7781 (vect_recog_over_widening_pattern): Likewise.
7782 (vect_recog_gather_scatter_pattern): Likewise.
7783 (vect_recog_func_ptr): Move from tree-vectorizer.h
7784 (vect_vect_recog_func_ptrs): Move further down the file.
7785 (vect_recog_func): Likewise. Remove the third argument.
7786 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
7787 (vect_pattern_recog_1): Expect the pattern function to do any
7788 necessary target tests. Also expect it to provide a vector type.
7789 Remove the type_in handling.
7791 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
7793 * tree-vect-patterns.c (vect_pattern_detected): New function.
7794 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
7795 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
7796 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
7797 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
7798 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
7799 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
7800 (vect_recog_mask_conversion_pattern)
7801 (vect_try_gather_scatter_pattern): Likewise.
7803 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
7805 * tree-vect-patterns.c (vect_get_internal_def): New function.
7806 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
7807 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
7808 (search_type_for_mask_1): Use it.
7810 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
7812 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
7813 redundant WIDEN_SUM_EXPR handling.
7814 (vect_recog_sad_pattern): Likewise.
7816 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
7818 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
7819 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
7820 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
7821 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
7822 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
7824 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
7826 * tree-vect-stmts.c (vectorizable_call): Make sure that we
7827 use the stmt_vec_info of the original bb statement for the
7828 new zero assignment, even if the call is part of a pattern.
7830 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
7832 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
7833 that the sequence is attached to the original statement rather
7834 than the pattern statement.
7835 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
7836 PATTERN_DEF_SEQ from the original statement rather than
7837 the main pattern statement.
7838 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
7839 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
7840 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
7842 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
7844 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
7845 definition statements before the early exit for statements that aren't
7847 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
7849 (vect_transform_loop): ...here. Process pattern definition
7850 statements without first checking whether the main pattern
7851 statement is live or relevant.
7853 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
7855 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
7856 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
7858 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
7860 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
7861 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
7862 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
7863 (expand_block_compare): Change select_block_compare_mode call.
7864 (expand_strncmp_align_check): Use new functions, fix comment.
7865 (emit_final_str_compare_gpr): New function.
7866 (expand_strn_compare): Refactor and clean up code.
7867 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
7869 2018-06-19 Tony Reix <tony.reix@atos.com>
7870 Damien Bergamini <damien.bergamini@atos.com>
7871 David Edelsohn <dje.gcc@gmail.com>
7873 * collect2.c (static_obj): New variable.
7874 (static_libs): New variable.
7875 (is_in_list): Uncomment declaration.
7876 (main): Track AIX libraries linked statically.
7877 (is_in_list): Uncomment definition.
7878 (scan_prog_file): Don't add AIX shared libraries initializer
7879 to constructor list if linking statically.
7881 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
7883 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
7885 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
7887 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
7889 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
7892 2018-06-19 Martin Liska <mliska@suse.cz>
7894 * config/i386/i386.c (ix86_can_inline_p): Do not use
7895 ipa_fn_summaries::get_create.
7896 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
7898 (devirtualization_time_bonus): Likewise.
7899 (ipcp_propagate_stage): Likewise.
7900 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
7901 (edge_set_predicate): Likewise.
7902 (evaluate_conditions_for_known_args): Likewise.
7903 (evaluate_properties_for_edge): Likewise.
7904 (ipa_call_summary::reset): Tranform to ...
7905 (ipa_call_summary::~ipa_call_summary): ... this.
7906 (ipa_fn_summary::reset): Transform to ...
7907 (ipa_fn_summary::~ipa_fn_summary): ... this.
7908 (ipa_fn_summary_t::remove): Rename to ...
7909 (ipa_fn_summary_t::remove_callees): ... this.
7910 (ipa_fn_summary_t::duplicate): Use placement new
7911 instead of memory copy.
7912 (ipa_call_summary_t::duplicate): Likewise.
7913 (ipa_call_summary_t::remove): Remove.
7914 (dump_ipa_call_summary): Change get_create to get.
7915 (ipa_dump_fn_summary): Dump only when summary exists.
7916 (analyze_function_body): Use symbol_summary::get instead
7918 (compute_fn_summary): Likewise.
7919 (estimate_edge_devirt_benefit): Likewise.
7920 (estimate_edge_size_and_time): Likewise.
7921 (inline_update_callee_summaries): Likewise.
7922 (remap_edge_change_prob): Likewise.
7923 (remap_edge_summaries): Likewise.
7924 (ipa_merge_fn_summary_after_inlining): Likewise.
7925 (write_ipa_call_summary): Likewise.
7926 (ipa_fn_summary_write): Likewise.
7927 (ipa_free_fn_summary): Likewise.
7928 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
7929 (struct ipa_call_summary): Likewise.
7930 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
7932 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
7933 (estimate_size_after_inlining): Likewise.
7934 (estimate_growth): Likewise.
7935 (growth_likely_positive): Likewise.
7936 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
7937 (inline_call): Likewise.
7938 * ipa-inline.c (caller_growth_limits): Likewise.
7939 (can_inline_edge_p): Likewise.
7940 (can_inline_edge_by_limits_p): Likewise.
7941 (compute_uninlined_call_time): Likewise.
7942 (compute_inlined_call_time): Likewise.
7943 (want_inline_small_function_p): Likewise.
7944 (edge_badness): Likewise.
7945 (update_caller_keys): Likewise.
7946 (update_callee_keys): Likewise.
7947 (inline_small_functions): Likewise.
7948 (inline_to_all_callers_1): Likewise.
7949 (dump_overall_stats): Likewise.
7950 (early_inline_small_functions): Likewise.
7951 (early_inliner): Likewise.
7952 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
7953 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
7954 * ipa-pure-const.c (malloc_candidate_p): Likewise.
7955 * ipa-split.c (execute_split_functions): Likewise.
7956 * symbol-summary.h: Likewise.
7957 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
7959 2018-06-19 Richard Biener <rguenther@suse.de>
7961 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
7962 (vectorize_loops): ... here. Fix dbgcnt handling.
7963 (try_vectorize_loop): Wrap try_vectorize_loop_1.
7965 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
7968 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
7969 ieee128 argument takes up only one (vector) register, not two (floating
7972 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
7974 * gimplify.c (gimplify_init_constructor): Really never clear for an
7975 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
7977 2018-06-19 Richard Biener <rguenther@suse.de>
7979 PR tree-optimization/86179
7980 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
7981 after failed recognition.
7983 2018-06-18 Martin Sebor <msebor@redhat.com>
7986 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
7987 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
7988 Handle integer subtraction.
7989 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
7990 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
7992 2018-06-18 David Malcolm <dmalcolm@redhat.com>
7994 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
7995 param from rtx to rtx_insn *.
7996 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
7998 (frv_ifcvt_modify_insn): Likwise.
7999 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
8000 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
8001 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
8002 as_a <rtx_insn *> cast to local "unprotected_region" once
8003 it's been established that it's not NULL or pc_rtx.
8004 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
8005 param "sethi" from rtx to rtx_insn *.
8006 (nds32_group_float_insns): Likewise for param "insn".
8007 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
8009 (vax_output_int_subtract): Likewise.
8010 * config/vax/vax.c (vax_output_int_add): Likewise for param
8012 (vax_output_int_subtract): Likewise.
8013 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
8014 (emit_pattern_after): Likewise for param "after".
8015 (emit_insn_after): Likewise.
8016 (emit_jump_insn_after): Likewise.
8017 (emit_call_insn_after): Likewise.
8018 (emit_debug_insn_after): Likewise.
8019 (emit_pattern_before): Likewise for param "before".
8020 (emit_insn_before): Likewise.
8021 (emit_jump_insn_before): Likewise.
8022 * final.c (get_insn_template): Likewise for param "insn", removing
8024 * output.h (get_insn_template): Likewise for 2nd param.
8025 * rtl.h (emit_insn_before): Likewise.
8026 (emit_jump_insn_before): Likewise.
8027 (emit_debug_insn_before_noloc): Likewise.
8028 (emit_insn_after): Likewise.
8029 (emit_jump_insn_after): Likewise.
8030 (emit_call_insn_after): Likewise.
8031 (emit_debug_insn_after): Likewise.
8032 (set_insn_deleted): Likewise for param.
8034 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
8037 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
8038 floating point modes, so that IFmode is numerically greater than
8039 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
8040 to declare the ordering. This prevents IFmode from being
8041 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
8042 machine. Include rs6000-modes.h to share the fractional values
8043 between genmodes* and the rest of the compiler.
8047 * config/rs6000/rs6000-modes.h: New file.
8048 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
8049 meaning of rs6000_long_double_size so that 126..128 selects an
8050 appropriate 128-bit floating point type.
8051 (rs6000_option_override_internal): Likewise.
8052 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
8053 (TARGET_LONG_DOUBLE_128): Change the meaning of
8054 rs6000_long_double_size so that 126..128 selects an appropriate
8055 128-bit floating point type.
8056 (LONG_DOUBLE_TYPE_SIZE): Update comment.
8057 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
8058 source and destination to match the standard usage.
8059 (truncifkf2): Likewise.
8060 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
8061 ISA 2.07 to use an explicit clobber, instead of passing in a
8063 (copysign<mode>3_soft): Likewise.
8065 2018-06-18 David Malcolm <dmalcolm@redhat.com>
8067 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
8068 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
8069 (vect_slp_analyze_instance_dependence): Likewise.
8070 (vect_enhance_data_refs_alignment): Likewise.
8071 (vect_analyze_data_refs_alignment): Likewise.
8072 (vect_slp_analyze_and_verify_instance_alignment
8073 (vect_analyze_data_ref_accesses): Likewise.
8074 (vect_prune_runtime_alias_test_list): Likewise.
8075 (vect_analyze_data_refs): Likewise.
8076 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
8077 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
8078 (vect_analyze_scalar_cycles_1): Likewise.
8079 (vect_get_loop_niters): Likewise.
8080 (vect_analyze_loop_form_1): Likewise.
8081 (vect_update_vf_for_slp): Likewise.
8082 (vect_analyze_loop_operations): Likewise.
8083 (vect_analyze_loop): Likewise.
8084 (vectorizable_induction): Likewise.
8085 (vect_transform_loop): Likewise.
8086 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
8087 * tree-vect-slp.c (vect_analyze_slp): Likewise.
8088 (vect_make_slp_decision): Likewise.
8089 (vect_detect_hybrid_slp): Likewise.
8090 (vect_slp_analyze_operations): Likewise.
8091 (vect_slp_bb): Likewise.
8092 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
8093 (vectorizable_bswap): Likewise.
8094 (vectorizable_call): Likewise.
8095 (vectorizable_simd_clone_call): Likewise.
8096 (vectorizable_conversion): Likewise.
8097 (vectorizable_assignment): Likewise.
8098 (vectorizable_shift): Likewise.
8099 (vectorizable_operation): Likewise.
8100 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
8102 2018-06-18 Martin Sebor <msebor@redhat.com>
8104 PR tree-optimization/81384
8105 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
8106 * builtins.c (expand_builtin_strnlen): New function.
8107 (expand_builtin): Call it.
8108 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
8109 * builtins.def (BUILT_IN_STRNLEN): New.
8110 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
8111 Warn for bounds in excess of maximum object size.
8112 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
8113 single-value ranges. Handle strnlen.
8114 (handle_builtin_strlen): Handle strnlen.
8115 (strlen_check_and_optimize_stmt): Same.
8116 * doc/extend.texi (Other Builtins): Document strnlen.
8118 2018-06-18 Maya Rashish <coypu@sdf.org>
8120 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
8121 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
8122 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
8124 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
8126 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
8128 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8130 * tree.c (escaped_string::escape): Replace cast to char * by
8131 const_cast<char *> (unescaped).
8133 2018-06-18 Nick Clifton <nickc@redhat.com>
8136 * tree.c (escaped_string): New class. Converts an unescaped
8137 string into its escaped equivalent.
8138 (warn_deprecated_use): Use the new class to convert the
8139 deprecation message, if present.
8140 (test_escaped_strings): New self test.
8141 (test_c_tests): Add test_escaped_strings.
8142 * doc/extend.texi (deprecated): Add a note that the
8143 deprecation message is affected by the -fmessage-length
8144 option, and that control characters will be escaped.
8145 (#pragma GCC error): Document this pragma.
8146 (#pragma GCC warning): Likewise.
8147 * doc/invoke.texi (-fmessage-length): Document this option's
8148 effect on the #warning and #error preprocessor directives and
8149 the deprecated attribute.
8151 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
8153 * tree.c (decl_value_expr_lookup): Revert latest change.
8154 (decl_value_expr_insert): Likewise.
8156 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
8158 * gimplify.c (nonlocal_vlas): Delete.
8159 (nonlocal_vla_vars): Likewise.
8160 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
8162 (gimplify_body): Do not create and destroy nonlocal_vlas.
8163 * tree-nested.c: Include diagnostic.h.
8164 (use_pointer_in_frame): Tweak.
8165 (lookup_field_for_decl): Add assertion and declare the transformation.
8166 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
8167 internal error when the reference is in a wrong context. Do not
8168 create a debug decl by default.
8169 (note_nonlocal_block_vlas): Delete.
8170 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
8171 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
8172 create a debug decl by default.
8173 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
8174 call is in a wrong context.
8175 (fixup_vla_decls): New function.
8176 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
8177 debug variables were created.
8178 * tree.c (decl_value_expr_lookup): Add checking assertion.
8179 (decl_value_expr_insert): Likewise.
8181 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
8184 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
8185 * tree-scalar-evolution.c (interpret_expr): Likewise.
8186 (expression_expensive_p): Likewise.
8187 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
8188 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
8189 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
8190 (ssa_defined_by_minus_one_stmt_p): New.
8192 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
8195 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
8196 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
8197 * dojump.c (do_jump): Likewise.
8198 * expr.c (expand_expr_real_2): Check operand type's sign.
8199 * fold-const.c (const_unop): Handle ABSU_EXPR.
8200 (fold_abs_const): Likewise.
8201 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
8202 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
8203 (strip_sign_op_1): Likesise.
8204 * match.pd: Add new pattern to generate ABSU_EXPR.
8205 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
8206 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
8207 * tree-eh.c (operation_could_trap_helper_p): Likewise.
8208 * tree-inline.c (estimate_operator_cost): Likewise.
8209 * tree-pretty-print.c (dump_generic_node): Likewise.
8210 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
8211 * tree.def (ABSU_EXPR): New.
8213 2018-06-16 Jakub Jelinek <jakub@redhat.com>
8216 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
8217 documented as preserved for backward compatibility only.
8218 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
8220 PR rtl-optimization/86108
8221 * bb-reorder.c (create_forwarder_block): Renamed to ...
8222 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
8223 jump from new landing pad to the second part.
8224 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
8227 2018-06-15 Jakub Jelinek <jakub@redhat.com>
8230 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
8231 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
8232 Only call store_expr for halves if the mode is the same.
8235 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
8236 Fix up comment formatting.
8238 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
8240 * typed-splay-tree.h (typed_splay_tree::remove): New function.
8241 (typed_splay_tree::closure,
8242 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
8243 (typed_splay_tree::typed_splay_tree,
8244 typed_splay_tree::operator =): Declared private.
8245 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
8246 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
8247 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
8248 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
8249 typed_splay_tree::splay_tree_splay,
8250 typed_splay_tree::splay_tree_foreach_helper,
8251 typed_splay_tree::splay_tree_insert,
8252 typed_splay_tree::splay_tree_remove,
8253 typed_splay_tree::splay_tree_lookup,
8254 typed_splay_tree::splay_tree_predecessor,
8255 typed_splay_tree::splay_tree_successor,
8256 typed_splay_tree::splay_tree_min,
8257 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
8258 (typed_splay_tree::root, typed_splay_tree::comp,
8259 typed_splay_tree::delete_key,
8260 typed_splay_tree::delete_value): New data members.
8261 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
8262 typed_splay_tree::remove.
8264 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
8266 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
8267 -mginv and -mno-ginv to the assembler.
8268 * config/mips/mips.opt (-mcrc): New option.
8270 * doc/invoke.text (-mcrc): Document.
8273 2018-06-15 Nick Clifton <nickc@redhat.com>
8276 * tree.c (escaped_string): New class. Converts an unescaped
8277 string into its escaped equivalent.
8278 (warn_deprecated_use): Use the new class to convert the
8279 deprecation message, if present.
8280 (test_escaped_strings): New self test.
8281 (test_c_tests): Add test_escaped_strings.
8282 * doc/extend.texi (deprecated): Add a note that the
8283 deprecation message is affected by the -fmessage-length
8284 option, and that control characters will be escaped.
8285 (#pragma GCC error): Document this pragma.
8286 (#pragma GCC warning): Likewise.
8287 * doc/invoke.texi (-fmessage-length): Document this option's
8288 effect on the #warning and #error preprocessor directives and
8289 the deprecated attribute.
8291 2018-06-15 Richard Biener <rguenther@suse.de>
8293 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
8294 here, also noting vector size used.
8295 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
8296 size used in MSG_OPTIMIZED_LOCATIONS dump.
8297 (pass_slp_vectorize::execute): Adjust.
8299 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
8302 * config/arc/arc.c (arc_return_address_register): Fix
8305 2018-06-15 Richard Biener <rguenther@suse.de>
8308 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
8309 leave useless conversion stripping to force_gimple_operand_gsi.
8310 (gimplify_build2): Likewise.
8311 (gimplify_build1): Likewise.
8313 2018-06-15 Richard Biener <rguenther@suse.de>
8316 * tree-cfg.c (move_stmt_op): unshare invariant addresses
8317 before adjusting their block.
8319 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
8321 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
8322 multilibs for *-*-rtems*.
8323 * config/riscv/t-rtems: New file.
8325 2018-06-14 Jakub Jelinek <jakub@redhat.com>
8328 * match.pd ((A +- CST1) +- CST2): Punt if last resort
8329 unsigned_type_for returns NULL.
8332 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
8333 subregs of multi-word pseudos unless the float mode has word size.
8335 2018-06-14 Richard Biener <rguenther@suse.de>
8338 * tree-vect-generic.c (build_word_mode_vector_type): Remove
8339 duplicate and harmful type_hash_canon.
8340 * tree.c (type_hash_canon): Assert we didn't find ourselves.
8342 2018-06-14 Richard Biener <rguenther@suse.de>
8345 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
8348 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
8350 * config/rtems.h (STDINT_LONG32): Define.
8352 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
8353 Prachi Godbole <prachi.godbole@imgtec.com>
8355 * config/mips/mips-cpus.def: Define P6600.
8356 * config/mips/mips-tables.opt: Regenerate.
8357 * config/mips/mips.c (mips_ucbranch_type): New enum.
8358 (mips_rtx_cost_data): Add support for P6600.
8359 (mips_issue_rate): Likewise.
8360 (mips_multipass_dfa_lookahead): Likewise.
8361 (mips_avoid_hazard): Likewise.
8362 (mips_reorg_process_insns): Likewise.
8363 (mips_classify_branch_p6600): New function.
8364 * config/mips/mips.h (TUNE_P6600): New define.
8365 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
8366 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
8367 * config/mips/mips.md: Include p6600.md.
8368 (processor): Add p6600.
8369 * config/mips/p6600.md: New file.
8370 * doc/invoke.texi: Add p6600 to supported architectures.
8372 2018-06-13 Martin Sebor <msebor@redhat.com>
8374 PR tree-optimization/86114
8375 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
8377 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
8379 2018-06-13 Richard Biener <rguenther@suse.de>
8381 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
8382 Properly set vector type of the intermediate stmt.
8383 * tree-vect-stmts.c (vectorizable_operation): The destination
8384 var always has vectype_out type.
8386 2018-06-13 Jeff Law <law@redhat.com>
8388 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
8389 integer 0 for argument to print_rtl_with_bb.
8390 (rl78_reorg): Likewise.
8392 2018-06-13 David Malcolm <dmalcolm@redhat.com>
8394 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
8395 from rtx to rtx_insn *.
8396 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
8398 (add_sched_insns_for_speculation): Likewise for local "target",
8399 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
8400 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
8401 from rtx_insn ** to rtx_code_label **.
8402 (reorg_emit_nops): Likewise.
8403 (c6x_reorg): Likewise for local "call_labels".
8404 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
8406 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
8407 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
8408 the loops over LABEL_REFS.
8409 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
8411 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
8412 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
8413 (split_branches): Strengthen local "olabel" from rtx to
8414 rtx_insn *, adding a safe_as_a cast.
8415 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
8417 (add_insn_after): Likewise for first two params.
8418 (add_insn_before): Likewise.
8419 (remove_insn): Likewise for param.
8420 (emit_pattern_before_noloc): Likewise for second and third params.
8421 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
8422 (emit_call_insn_before_noloc): Likewise.
8423 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
8425 (emit_barrier_before): Likewise.
8426 (emit_label_before): Strengthen "label" param from "rtx" to
8427 "rtx_code_label *". Strengthen "before" param from "rtx" to
8429 (emit_insn_after_1): Strengthen "after" param from "rtx" to
8431 (emit_pattern_after_noloc): Likewise.
8432 (emit_insn_after_noloc): Likewise.
8433 (emit_jump_insn_after_noloc): Likewise.
8434 (emit_call_insn_after_noloc): Likewise.
8435 (emit_debug_insn_after_noloc): Likewise.
8436 (emit_barrier_after): Likewise.
8437 (emit_label_after): Likewise for both params.
8438 (emit_pattern_after_setloc): Likewise for "after" param. Convert
8439 "loc" param from "int" to "location_t".
8440 (emit_insn_after_setloc): Likewise.
8441 (emit_jump_insn_after_setloc): Likewise.
8442 (emit_call_insn_after_setloc): Likewise.
8443 (emit_debug_insn_after_setloc): Likewise.
8444 (emit_pattern_before_setloc): Likewise for "before" param. Convert
8445 "loc" param from "int" to "location_t".
8446 (emit_pattern_before): Convert NULL_RTX to NULL.
8447 (emit_insn_before_setloc): Convert "loc" param from "int" to
8449 (emit_jump_insn_before_setloc): Likewise.
8450 (emit_call_insn_before_setloc): Likewise.
8451 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
8452 rtx_insn *. Convert "loc" param from "int" to "location_t".
8453 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
8454 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
8455 Convert 3rd param from "int" to "location_t".
8456 (emit_barrier_before, emit_barrier_after, next_real_insn):
8457 Strengthen param from rtx to rtx_insn *.
8458 (emit_label_before): Strengthen 1st param from "rtx" to
8459 "rtx_code_label *". Strengthen 2nd param from "rtx" to
8461 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
8462 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
8463 Strengthen 2nd param from "rtx" to "rtx_insn *".
8464 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
8465 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
8466 Likewise. Convert 3rd param from "int" to "location_t".
8467 (emit_label_after): Strengthen 1st param from "rtx" to
8469 (next_real_insn, remove_insn): Strengthen param from "rtx" to
8471 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
8472 from "rtx" to "rtx_insn *".
8474 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
8476 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
8477 bodies streamed in with -Q.
8478 * dumpfile.c (dump_files): Add lto-stream-out dump file.
8479 * dumpfile.h (tree_dump_index): Add lto_stream_out.
8480 * gimple-streamer-out.c: Include gimple-pretty-print.h
8481 (output_bb): Dump stmts streamed.
8482 * lto-section-out.c: Include print-tree.h
8483 (lto_begin_section): Dump sections created.
8484 (lto_output_decl_index): Dump decl encoded.
8485 * lto-streamer-out.c: Include print-tree.h
8486 (create_output_block): Dump output block created.
8487 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
8488 (output_function): Dump function output.
8489 (output_constructor): Dump constructor streamed.
8490 (write_global_stream): Output indexes encoded.
8491 (produce_asm_for_decls): Dump streams encoded.
8492 * lto-streamer.c (streamer_dump_file): New global var.
8493 * lto-streamer.h (streamer_dump_file): Declare.
8494 * passes.c (ipa_write_summaries): Initialize streamer dump.
8495 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
8498 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
8501 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
8502 offsets for register save directives. Emit a second batch of save
8503 directives, if need be, when the function accesses prior frames.
8505 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
8507 * config/arc/fpu.md (fmasf4): Force operand to register.
8508 (fnmasf4): Likewise.
8510 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
8512 * config/arc/arc-protos.h (arc_pad_return): Remove.
8513 * config/arc/arc.c (machine_function): Remove force_short_suffix
8515 (arc_print_operand): Adjust printing of '&'.
8516 (arc_verify_short): Remove conditional printing of short suffix.
8517 (arc_final_prescan_insn): Remove reference to size_reason.
8518 (pad_return): New function.
8519 (arc_reorg): Call pad_return.
8520 (arc_pad_return): Remove.
8521 (arc_init_machine_status): Remove reference to force_short_suffix.
8522 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
8523 (attr length): When attribute iscompact is true force to 2
8524 regardless; in the case of maybe check if we want to force the
8525 instruction to have 4 bytes length.
8526 (nopv): Change it to generate 4 byte long nop as well.
8527 (blockage): New pattern.
8528 (simple_return): Remove call to arc_pad_return.
8529 (p_return_i): Likewise.
8531 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
8533 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
8535 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
8537 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
8540 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
8542 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
8543 for ARC700 and ARCv2.
8545 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
8548 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
8549 operands[2] instead of operands[1].
8552 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
8554 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
8555 case, check whether the outer register overlaps an unallocatable
8556 register, not just whether it fits the required class.
8558 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
8560 * poly-int.h (can_div_trunc_p): Add new overload in which all values
8562 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
8563 (memrefs_conflict_p): Likewise.
8564 (init_alias_analysis): Likewise.
8565 * cfgexpand.c (expand_debug_expr): Likewise.
8566 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
8567 * cse.c (fold_rtx): Likewise.
8568 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
8569 * expr.c (emit_block_move_hints): Likewise.
8570 (clear_storage_hints, push_block, emit_push_insn): Likewise.
8571 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
8572 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
8573 (emit_group_store): Likewise.
8574 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
8575 to read the PRE/POST_MODIFY increment.
8576 * calls.c (store_one_arg): Use strip_offset.
8577 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
8579 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
8581 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
8582 (simplify_binary_operation_1): Extend CONST_INT handling to
8584 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
8585 than a HOST_WIDE_INT.
8586 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
8588 (adjust_mems, add_stores): Update accodingly.
8589 (vt_canonicalize_addr): Track polynomial offsets.
8590 (emit_note_insn_var_location): Likewise.
8591 (vt_add_function_parameter): Likewise.
8592 (vt_initialize): Likewise.
8594 2018-06-12 Jeff Law <law@redhat.com>
8596 * config.gcc (alpha*-*-freebsd*): Remove.
8597 * config/alpha/freebsd.h: Remove.
8599 2018-06-12 David Malcolm <dmalcolm@redhat.com>
8602 * spellcheck-tree.c (levenshtein_distance): Rename to...
8603 (get_edit_distance): ...this, and update for underlying renaming.
8604 * spellcheck-tree.h (levenshtein_distance): Rename to...
8605 (get_edit_distance): ...this.
8606 * spellcheck.c (levenshtein_distance): Rename to...
8607 (get_edit_distance): ...this. Convert from Levenshtein distance
8608 to Damerau-Levenshtein distance by supporting transpositions of
8609 adjacent characters. Rename "v1" to "v_next" and "v0" to
8611 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
8612 (selftest::test_edit_distance_unit_test_oneway): ...this, and
8613 update for underlying renaming.
8614 (selftest::levenshtein_distance_unit_test): Rename to...
8615 (selftest::test_get_edit_distance_unit): ...this, and update for
8616 underlying renaming.
8617 (selftest::test_find_closest_string): Add example from PR 69968
8618 where transposition helps
8619 (selftest::test_metric_conditions): Update for renaming.
8620 (selftest::test_metric_conditions): Likewise.
8621 (selftest::spellcheck_c_tests): Likewise.
8622 * spellcheck.h (levenshtein_distance): Rename both overloads to...
8623 (get_edit_distance): ...this.
8624 (best_match::consider): Update for renaming.
8626 2018-06-12 Martin Sebor <msebor@redhat.com>
8628 PR tree-optimization/85259
8629 * builtins.c (compute_objsize): Handle constant offsets.
8630 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
8631 true iff a warning has been issued.
8632 * gimple.h (gimple_nonartificial_location): New function.
8633 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
8634 gimple_nonartificial_location and handle -Wno-system-headers.
8635 (handle_builtin_stxncpy): Same.
8637 2018-06-12 Martin Sebor <msebor@redhat.com>
8640 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
8642 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
8644 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
8645 BUILTIN_VEC_XST entries for pointer to double and long long.
8647 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
8650 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
8652 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
8655 2018-06-12 Martin Liska <mliska@suse.cz>
8657 * doc/options.texi: Document IntegerRange.
8659 2018-06-12 Martin Liska <mliska@suse.cz>
8661 * config/i386/i386.opt: Make MPX-related options as Deprecated.
8662 * opt-functions.awk: Handle Deprecated flag.
8663 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
8665 (read_cmdline_option): Report warning for a deprecated option.
8666 * opts.h (struct cl_option): Add new field cl_deprecated.
8667 (CL_ERR_DEPRECATED): New.
8669 2018-06-12 Martin Liska <mliska@suse.cz>
8671 * doc/options.texi: Document Deprecated option flag.
8673 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
8675 * config/arc/arc-arch.h (arc_extras): New enum.
8676 (arc_cpu_t):Add field extra.
8677 (arc_cpu_types): Consider the extras.
8678 * config/arc/arc-cpus.def: Add extras info.
8679 * config/arc/arc-opts.h (processor_type): Consider extra field.
8680 * config/arc/arc.c (arc_override_options): Handle extra field.
8682 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
8684 * config/arc/arc-arch.h: Update ARC_OPTX macro.
8685 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
8687 * config/arc/arc.c (arc_init): Update pic warning.
8688 (irq_range): Update irq range parsing warnings.
8689 (arc_override_options): Update various warning messages.
8690 (arc_handle_aux_attribute): Likewise.
8692 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
8694 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
8696 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
8698 * doc/sourcebuild.texi: Document usage of line number 0 in verify
8699 compiler messages directives.
8701 2018-06-12 Matthew Fortune <mfortune@gmail.com>
8703 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
8704 * config/mips/mips-tables.opt: Regenerate.
8705 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
8707 * doc/invoke.texi: Document -march=i6500.
8709 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
8711 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
8712 (i6400_gpmul): Add cpu_unit.
8713 (i6400_gpdiv): Likewise.
8714 (i6400_msa_add_d): Update reservations.
8715 (i6400_msa_int_add) Likewise.
8716 (i6400_msa_short_logic3) Likewise.
8717 (i6400_msa_short_logic2) Likewise.
8718 (i6400_msa_short_logic) Likewise.
8719 (i6400_msa_move) Likewise.
8720 (i6400_msa_cmp) Likewise.
8721 (i6400_msa_short_float2) Likewise.
8722 (i6400_msa_div_d) Likewise.
8723 (i6400_msa_long_logic1) Likewise.
8724 (i6400_msa_long_logic2) Likewise.
8725 (i6400_msa_mult) Likewise.
8726 (i6400_msa_long_float2) Likewise.
8727 (i6400_msa_long_float4) Likewise.
8728 (i6400_msa_long_float5) Likewise.
8729 (i6400_msa_long_float8) Likewise.
8730 (i6400_fpu_fadd): Include frint type.
8731 (i6400_fpu_store): New define_insn_reservation.
8732 (i6400_fpu_load): Likewise.
8733 (i6400_fpu_move): Likewise.
8734 (i6400_fpu_fcmp): Likewise.
8735 (i6400_fpu_fmadd): Likewise.
8736 (i6400_int_mult): Include imul3nc type and update reservation.
8737 (i6400_int_div): Include idiv3 type and update reservation.
8738 (i6400_int_load): Update to check type not move_type.
8739 (i6400_int_store): Likewise.
8740 (i6400_int_prefetch): Set zero latency.
8742 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
8744 * gcc.c: Document new %@{...} sequence.
8745 (LINK_COMMAND_SPEC): Use it for the -L switches.
8746 (cpp_unique_options): Use it for the -I switches.
8747 (at_file_argbuf): New global variable.
8748 (in_at_file): Likewise.
8749 (alloc_args): Create at_file_argbuf.
8750 (clear_args): Truncate at_file_argbuf.
8751 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
8752 (open_at_file): New function.
8753 (close_at_file): Likewise.
8754 (create_at_file): Delete.
8755 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
8758 (validate_switches_from_spec): Deal with %@{...} sequence.
8759 (validate_switches): Likewise.
8760 (driver::finalize): Call clear_args.
8762 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
8764 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
8766 2018-06-11 Martin Sebor <msebor@redhat.com>
8768 * doc/invoke.texi (-Wall): List -Wc++17-compat.
8769 (Wno-class-memaccess): Add @opindex.
8770 (Wno-templates, Wno-multiple-inheritance): Same.
8771 (Wno-virtual-inheritance, Wno-namespaces): Same.
8772 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
8773 (Wno-format-overflow, Wno-format-truncation): Same.
8774 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
8775 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
8776 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
8777 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
8778 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
8779 (Wno-misspelled-isr): Same.
8781 2018-06-11 Martin Sebor <msebor@redhat.com>
8783 * PR tree-optimization/86083
8784 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
8786 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
8788 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
8790 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
8793 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
8794 on the correct operand.
8795 (*movdi_internal64): Ditto.
8797 2018-06-11 Martin Liska <mliska@suse.cz>
8799 PR tree-optimization/86089
8800 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
8802 2018-06-11 Julia Koval <julia.koval@intel.com>
8804 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
8805 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
8806 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
8808 2018-06-11 Olivier Hainque <hainque@adacore.com>
8810 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
8811 for Ada with strict dwarf2.
8813 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
8816 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
8819 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
8821 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
8823 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
8825 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
8828 2018-06-08 Martin Liska <mliska@suse.cz>
8830 * tree-cfg.h (debug_function): Fix argument type to match
8833 2018-06-08 Martin Liska <mliska@suse.cz>
8835 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
8836 Remove usage of MPX-related (and removed) fields.
8837 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
8839 2018-06-08 David Malcolm <dmalcolm@redhat.com>
8841 * cfg.c (debug): Use TDF_NONE rather than 0.
8842 * cfghooks.c (debug): Likewise.
8843 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
8844 (struct dump_option_value_info): Convert to...
8845 (struct kv_pair): ...this template type.
8846 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
8848 (optinfo_verbosity_options): Likewise.
8849 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
8851 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
8852 than int for "optgroup_flags" param.
8853 (dump_generic_expr_loc): Use dump_flags_t rather than int for
8855 (dump_dec): Likewise.
8856 (dump_finish): Use TDF_NONE rather than 0.
8857 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
8858 rather than int for "optgroup_flags" param. Use TDF_NONE rather
8859 than 0. Update for change to option_ptr.
8860 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
8861 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
8862 0. Update for changes to optinfo_verbosity_options and
8864 (opt_info_switch_p): Convert optgroup_flags from int to
8866 (dump_basic_block): Use dump_flags_t rather than int
8867 for "dump_kind" param.
8868 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
8869 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
8870 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
8871 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
8872 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
8873 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
8874 TDF_NONE): Convert from macros to...
8875 (enum dump_flag): ...this new enum.
8876 (dump_flags_t): Update to use enum.
8877 (operator|, operator&, operator~, operator|=, operator&=):
8878 Implement for dump_flags_t.
8879 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
8880 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
8881 Convert from macros to...
8882 (enum optgroup_flag): ...this new enum.
8883 (optgroup_flags_t): New typedef.
8884 (operator|, operator|=): Implement for optgroup_flags_t.
8885 (struct dump_file_info): Convert field "alt_flags" to
8886 dump_flags_t. Convert field "optgroup_flags" to
8888 (dump_basic_block): Use dump_flags_t rather than int for param.
8889 (dump_generic_expr_loc): Likewise.
8890 (dump_dec): Likewise.
8891 (dump_register): Convert param "optgroup_flags" to
8893 (opt_info_enable_passes): Likewise.
8894 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
8896 * gimple-pretty-print.c (debug): Likewise.
8897 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
8898 (merged_store_group::apply_stores): Likewise.
8899 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
8900 * gimple.c (verify_gimple_pp): Likewise.
8901 * graphite-poly.c (print_pbb_body): Likewise.
8902 * passes.c (pass_manager::register_one_dump_file): Convert
8903 local "optgroup_flags" to optgroup_flags_t.
8904 * print-tree.c (print_node): Use TDF_NONE rather than 0.
8906 (debug_body): Likewise.
8907 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
8908 to optgroup_flags_t.
8909 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
8911 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
8912 (convert_mult_to_fma): Likewise.
8913 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
8914 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
8915 * tree-vect-data-refs.c (dump_lower_bound): Convert param
8916 "dump_kind" to dump_flags_t.
8918 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
8920 * config/rs6000/rs6000.c (min, max): Delete.
8922 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
8924 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
8925 -mabi=spe and -mabi=no-spe.
8927 2018-06-08 Martin Liska <mliska@suse.cz>
8929 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
8930 where we expect an existing summary.
8932 2018-06-08 Martin Liska <mliska@suse.cz>
8934 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
8935 * ipa-inline.h (estimate_edge_growth): Likewise.
8937 2018-06-08 Martin Liska <mliska@suse.cz>
8939 * cgraph.c (function_version_hasher::hash): Use
8940 cgraph_node::get_uid ().
8941 (function_version_hasher::equal):
8942 * cgraph.h (cgraph_node::get_uid): New method.
8943 * ipa-inline.c (update_caller_keys): Use
8944 cgraph_node::get_uid ().
8945 (update_callee_keys): Likewise.
8946 * ipa-utils.c (searchc): Likewise.
8947 (ipa_reduced_postorder): Likewise.
8948 * lto-cgraph.c (input_node): Likewise.
8949 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
8950 * symbol-summary.h (symtab_insertion): Likewise.
8951 (symtab_removal): Likewise.
8952 (symtab_duplication): Likewise.
8953 * tree-pretty-print.c (dump_function_header): Likewise.
8954 * tree-sra.c (convert_callers_for_node): Likewise.
8956 2018-06-08 Martin Liska <mliska@suse.cz>
8958 * cgraph.c (symbol_table::create_edge): Always assign a new
8960 (symbol_table::free_edge): Do not recycle numbers.
8961 * cgraph.h (cgraph_edge::get): New method.
8962 * symbol-summary.h (symtab_removal): Use it.
8963 (symtab_duplication): Likewise.
8964 (call_summary::hashable_uid): Remove.
8966 2018-06-08 Martin Liska <mliska@suse.cz>
8968 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
8969 (initialize_growth_caches): Remove.
8970 (free_growth_caches): Likewise.
8971 (do_estimate_edge_time): Use edge_growth_cache.
8972 (do_estimate_edge_size): Likewise.
8973 (do_estimate_edge_hints): Likewise.
8974 * ipa-inline.c (reset_edge_caches): Likewise.
8975 (recursive_inlining): Likewise.
8976 (inline_small_functions): Likewise.
8977 * ipa-inline.h (initialize_growth_caches): Remove.
8978 (estimate_edge_size): Likewise.
8979 (estimate_edge_time): Likewise.
8980 (estimate_edge_hints): Likewise.
8981 (reset_edge_growth_cache): Likewise.
8982 * symbol-summary.h (call_summary::remove): New method.
8984 2018-06-08 Martin Liska <mliska@suse.cz>
8986 * ipa-cp.c (class edge_clone_summary): New summary.
8987 (grow_edge_clone_vectors): Remove.
8988 (ipcp_edge_duplication_hook): Remove.
8989 (class edge_clone_summary_t): New call_summary class.
8990 (ipcp_edge_removal_hook): Remove.
8991 (edge_clone_summary_t::duplicate): New function.
8992 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
8993 (create_specialized_node): Likewise.
8994 (ipcp_driver): Initialize edge_clone_summaries and do not
8997 2018-06-08 Martin Liska <mliska@suse.cz>
8999 * symbol-summary.h (get): New function.
9000 (call_summary::m_initialize_when_cloning): New class member.
9002 2018-06-08 Martin Liska <mliska@suse.cz>
9004 * cgraph.c (cgraph_node::remove): Do not recycle uid.
9005 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
9006 (symbol_table::allocate_cgraph_symbol): Do not set uid.
9007 * passes.c (uid_hash_t): Record removed_nodes by their uids.
9008 (remove_cgraph_node_from_order): Use the removed_nodes set.
9009 (do_per_function_toporder): Likwise.
9010 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
9011 instead of summary_uid.
9012 (symtab_removal): Likewise.
9013 (symtab_duplication): Likewise.
9015 2018-06-08 Martin Liska <mliska@suse.cz>
9017 * ipa-cp.c (ipcp_store_bits_results): Use
9018 ipcp_transformation_sum.
9019 (ipcp_store_vr_results): Likewise.
9020 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
9022 (ipcp_transformation_initialize): ... this.
9023 (ipa_set_node_agg_value_chain):
9024 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
9025 (write_ipcp_transformation_info): Likewise.
9026 (read_ipcp_transformation_info): Likewise.
9027 (ipcp_update_bits): Likewise.
9028 (ipcp_update_vr): Likewise.
9029 (ipcp_transform_function): Likewise.
9030 * ipa-prop.h: Rename ipcp_transformation_summary to
9031 ipcp_transformation.
9032 (class ipcp_transformation_t): New function summary.
9033 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
9034 (ipa_get_agg_replacements_for_node): Likewise.
9036 2018-06-08 Martin Liska <mliska@suse.cz>
9038 * ipa-pure-const.c (struct funct_state_d): Do it class instead
9040 (class funct_state_summary_t): New function_summary class.
9041 (has_function_state): Remove.
9042 (get_function_state): Likewise.
9043 (set_function_state): Likewise.
9044 (add_new_function): Likewise.
9045 (funct_state_summary_t::insert): New function.
9046 (duplicate_node_data): Remove.
9047 (remove_node_data): Remove.
9048 (funct_state_summary_t::duplicate): New function.
9049 (register_hooks): Create new funct_state_summaries.
9050 (pure_const_generate_summary): Use it.
9051 (pure_const_write_summary): Likewise.
9052 (pure_const_read_summary): Likewise.
9053 (propagate_pure_const): Likewise.
9054 (propagate_nothrow): Likewise.
9055 (dump_malloc_lattice): Likewise.
9056 (propagate_malloc): Likewise.
9057 (execute): Do not register hooks, just remove summary
9059 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
9062 2018-06-08 Martin Liska <mliska@suse.cz>
9064 * ipa-reference.c (remove_node_data): Remove.
9065 (duplicate_node_data): Likewise.
9066 (class ipa_ref_var_info_summary_t): New class.
9067 (class ipa_ref_opt_summary_t): Likewise.
9068 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
9069 (get_reference_optimization_summary): Use
9070 ipa_ref_opt_sum_summaries.
9071 (set_reference_vars_info): Remove.
9072 (set_reference_optimization_summary): Likewise.
9073 (ipa_init): Create summaries.
9074 (init_function_info): Use function summary.
9075 (ipa_ref_opt_summary_t::duplicate): New function.
9076 (ipa_ref_opt_summary_t::remove): New function.
9077 (get_read_write_all_from_node): Fix GNU coding style.
9078 (propagate): Use function summary.
9079 (write_node_summary_p): Fix GNU coding style.
9080 (stream_out_bitmap): Likewise.
9081 (ipa_reference_read_optimization_summary): Use function summary.
9082 (ipa_reference_c_finalize): Do not release hooks.
9084 2018-06-08 Martin Liska <mliska@suse.cz>
9086 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
9087 (analyze_function_body): Extract multiple calls of get_create.
9088 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
9089 * ipa-inline.c (recursive_inlining): Use ::get method.
9090 * ipa-inline.h (estimate_edge_growth): Likewise.
9092 2018-06-08 Martin Liska <mliska@suse.cz>
9094 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
9096 (hsa_function_summary::hsa_function_summary): Use the new enum
9098 (hsa_gpu_implementation_p): Use hsa_summaries::get.
9099 * hsa-gen.c (hsa_get_host_function): Likewise.
9100 (get_brig_function_name): Likewise.
9101 * ipa-hsa.c (process_hsa_functions): Likewise.
9102 (ipa_hsa_write_summary): Likewise.
9103 * symbol-summary.h (symtab_duplication): Use ::get function/
9104 (get): New function.
9106 2018-06-08 Martin Liska <mliska@suse.cz>
9108 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
9110 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
9111 (hsa_register_kernel): Likewise.
9112 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
9113 * hsa-gen.c (hsa_get_host_function): Likewise.
9114 (get_brig_function_name): Likewise.
9115 (generate_hsa): Likewise.
9116 (pass_gen_hsail::execute): Likewise.
9117 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
9118 (devirtualization_time_bonus): Likewise.
9119 (ipcp_propagate_stage): Likewise.
9120 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
9121 (edge_set_predicate): Likewise.
9122 (evaluate_conditions_for_known_args): Likewise.
9123 (evaluate_properties_for_edge): Likewise.
9124 (ipa_fn_summary::reset): Likewise.
9125 (ipa_fn_summary_t::duplicate): Likewise.
9126 (dump_ipa_call_summary): Likewise.
9127 (ipa_dump_fn_summary): Likewise.
9128 (analyze_function_body): Likewise.
9129 (compute_fn_summary): Likewise.
9130 (estimate_edge_devirt_benefit): Likewise.
9131 (estimate_edge_size_and_time): Likewise.
9132 (estimate_calls_size_and_time): Likewise.
9133 (estimate_node_size_and_time): Likewise.
9134 (inline_update_callee_summaries): Likewise.
9135 (remap_edge_change_prob): Likewise.
9136 (remap_edge_summaries): Likewise.
9137 (ipa_merge_fn_summary_after_inlining): Likewise.
9138 (ipa_update_overall_fn_summary): Likewise.
9139 (read_ipa_call_summary): Likewise.
9140 (inline_read_section): Likewise.
9141 (write_ipa_call_summary): Likewise.
9142 (ipa_fn_summary_write): Likewise.
9143 (ipa_free_fn_summary): Likewise.
9144 * ipa-hsa.c (process_hsa_functions): Likewise.
9145 (ipa_hsa_write_summary): Likewise.
9146 (ipa_hsa_read_section): Likewise.
9147 * ipa-icf.c (sem_function::merge): Likewise.
9148 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
9149 (do_estimate_edge_time): Likewise.
9150 (estimate_size_after_inlining): Likewise.
9151 (estimate_growth): Likewise.
9152 (growth_likely_positive): Likewise.
9153 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
9154 (inline_call): Likewise.
9155 * ipa-inline.c (caller_growth_limits): Likewise.
9156 (can_inline_edge_p): Likewise.
9157 (can_inline_edge_by_limits_p): Likewise.
9158 (compute_uninlined_call_time): Likewise.
9159 (compute_inlined_call_time): Likewise.
9160 (want_inline_small_function_p): Likewise.
9161 (edge_badness): Likewise.
9162 (update_caller_keys): Likewise.
9163 (update_callee_keys): Likewise.
9164 (recursive_inlining): Likewise.
9165 (inline_small_functions): Likewise.
9166 (inline_to_all_callers_1): Likewise.
9167 (dump_overall_stats): Likewise.
9168 (early_inline_small_functions): Likewise.
9169 (early_inliner): Likewise.
9170 * ipa-inline.h (estimate_edge_growth): Likewise.
9171 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
9172 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
9173 * ipa-prop.h (IPA_NODE_REF): Likewise.
9174 (IPA_EDGE_REF): Likewise.
9175 * ipa-pure-const.c (malloc_candidate_p): Likewise.
9176 (propagate_malloc): Likewise.
9177 * ipa-split.c (execute_split_functions): Likewise.
9178 * symbol-summary.h: Rename get to get_create.
9180 (get_create): Likewise.
9181 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
9183 2018-06-08 Martin Liska <mliska@suse.cz>
9185 * symbol-summary.h (release): Move definition out of class
9187 (symtab_removal): Likewise.
9188 (symtab_duplication): Likewise.
9190 2018-06-08 Martin Liska <mliska@suse.cz>
9192 * symbol-summary.h (function_summary): Move constructor
9193 implementation out of class declaration.
9194 (release): Likewise.
9195 (symtab_insertion): Likewise.
9196 (symtab_removal): Likewise.
9197 (symtab_duplication): Likewise.
9200 2018-06-08 Martin Liska <mliska@suse.cz>
9202 * Makefile.in: Remove support for MPX (macros, related functions,
9203 fields in cgraph_node, ...).
9204 * builtin-types.def (BT_BND): Likewise.
9205 (BT_FN_BND_CONST_PTR): Likewise.
9206 (BT_FN_CONST_PTR_BND): Likewise.
9207 (BT_FN_VOID_PTR_BND): Likewise.
9208 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
9209 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
9210 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
9211 (expand_builtin_mempcpy_with_bounds): Likewise.
9212 (expand_builtin_memset_with_bounds): Likewise.
9213 (expand_builtin_memset_args): Likewise.
9214 (std_expand_builtin_va_start): Likewise.
9215 (expand_builtin): Likewise.
9216 (expand_builtin_with_bounds): Likewise.
9217 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
9218 (DEF_LIB_BUILTIN_CHKP): Likewise.
9219 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
9220 (DEF_CHKP_BUILTIN): Likewise.
9221 (BUILT_IN_MEMCPY): Likewise.
9222 (BUILT_IN_MEMMOVE): Likewise.
9223 (BUILT_IN_MEMPCPY): Likewise.
9224 (BUILT_IN_MEMSET): Likewise.
9225 (BUILT_IN_STPCPY): Likewise.
9226 (BUILT_IN_STRCAT): Likewise.
9227 (BUILT_IN_STRCHR): Likewise.
9228 (BUILT_IN_STRCPY): Likewise.
9229 (BUILT_IN_STRLEN): Likewise.
9230 (BUILT_IN_MEMCPY_CHK): Likewise.
9231 (BUILT_IN_MEMMOVE_CHK): Likewise.
9232 (BUILT_IN_MEMPCPY_CHK): Likewise.
9233 (BUILT_IN_MEMSET_CHK): Likewise.
9234 (BUILT_IN_STPCPY_CHK): Likewise.
9235 (BUILT_IN_STRCAT_CHK): Likewise.
9236 (BUILT_IN_STRCPY_CHK): Likewise.
9237 * calls.c (store_bounds): Likewise.
9238 (emit_call_1): Likewise.
9239 (special_function_p): Likewise.
9240 (maybe_warn_nonstring_arg): Likewise.
9241 (initialize_argument_information): Likewise.
9242 (finalize_must_preallocate): Likewise.
9243 (compute_argument_addresses): Likewise.
9244 (expand_call): Likewise.
9245 * cfgexpand.c (expand_call_stmt): Likewise.
9246 (expand_return): Likewise.
9247 (expand_gimple_stmt_1): Likewise.
9248 (pass_expand::execute): Likewise.
9249 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
9250 (cgraph_node::remove): Likewise.
9251 (cgraph_node::dump): Likewise.
9252 (cgraph_node::verify_node): Likewise.
9253 * cgraph.h (chkp_function_instrumented_p): Likewise.
9254 (symtab_node::get_alias_target): Likewise.
9255 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
9256 (cgraph_local_p): Likewise.
9257 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
9258 (cgraph_edge::rebuild_references): Likewise.
9259 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
9260 (walk_polymorphic_call_targets): Likewise.
9261 (cgraph_node::expand_thunk): Likewise.
9262 (symbol_table::output_weakrefs): Likewise.
9263 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
9264 (ix86_handle_option): Likewise.
9265 * config/i386/constraints.md: Likewise.
9266 * config/i386/i386-builtin-types.def (BND): Likewise.
9270 * config/i386/i386-builtin.def (BDESC_END): Likewise.
9271 (BDESC_FIRST): Likewise.
9273 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
9274 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
9275 * config/i386/i386.c (enum reg_class): Likewise.
9276 (ix86_target_string): Likewise.
9277 (ix86_option_override_internal): Likewise.
9278 (ix86_conditional_register_usage): Likewise.
9279 (ix86_valid_target_attribute_inner_p): Likewise.
9280 (ix86_set_indirect_branch_type): Likewise.
9281 (ix86_set_current_function): Likewise.
9282 (ix86_function_arg_regno_p): Likewise.
9283 (init_cumulative_args): Likewise.
9284 (ix86_function_arg_advance): Likewise.
9285 (ix86_function_arg): Likewise.
9286 (ix86_pass_by_reference): Likewise.
9287 (ix86_function_value_regno_p): Likewise.
9288 (ix86_function_value_1): Likewise.
9289 (ix86_function_value_bounds): Likewise.
9290 (ix86_return_in_memory): Likewise.
9291 (ix86_setup_incoming_vararg_bounds): Likewise.
9292 (ix86_va_start): Likewise.
9293 (indirect_thunk_need_prefix): Likewise.
9294 (print_reg): Likewise.
9295 (ix86_print_operand): Likewise.
9296 (ix86_expand_call): Likewise.
9297 (ix86_output_function_return): Likewise.
9298 (reg_encoded_number): Likewise.
9299 (BDESC_VERIFYS): Likewise.
9300 (ix86_init_mpx_builtins): Likewise.
9301 (ix86_init_builtins): Likewise.
9302 (ix86_emit_cmove): Likewise.
9303 (ix86_emit_move_max): Likewise.
9304 (ix86_expand_builtin): Likewise.
9305 (ix86_builtin_mpx_function): Likewise.
9306 (ix86_get_arg_address_for_bt): Likewise.
9307 (ix86_load_bounds): Likewise.
9308 (ix86_store_bounds): Likewise.
9309 (ix86_load_returned_bounds): Likewise.
9310 (ix86_store_returned_bounds): Likewise.
9311 (ix86_class_likely_spilled_p): Likewise.
9312 (ix86_hard_regno_mode_ok): Likewise.
9313 (x86_order_regs_for_local_alloc): Likewise.
9314 (ix86_mitigate_rop): Likewise.
9315 (ix86_bnd_prefixed_insn_p): Likewise.
9316 (ix86_mpx_bound_mode): Likewise.
9317 (ix86_make_bounds_constant): Likewise.
9318 (ix86_initialize_bounds): Likewise.
9319 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
9320 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
9321 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
9322 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
9323 (TARGET_CHKP_BOUND_MODE): Likewise.
9324 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
9325 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
9326 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
9327 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
9328 * config/i386/i386.h (TARGET_MPX): Likewise.
9329 (TARGET_MPX_P): Likewise.
9330 (VALID_BND_REG_MODE): Likewise.
9331 (FIRST_BND_REG): Likewise.
9332 (LAST_BND_REG): Likewise.
9333 (enum reg_class): Likewise.
9334 (BND_REG_P): Likewise.
9335 (BND_REGNO_P): Likewise.
9336 (BNDmode): Likewise.
9337 (ADJUST_INSN_LENGTH): Likewise.
9338 * config/i386/i386.md: Likewise.
9339 * config/i386/i386.opt: Likewise.
9340 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
9341 (defined): Likewise.
9342 (LINK_MPX): Likewise.
9343 (MPX_SPEC): Likewise.
9344 (LIBMPX_SPEC): Likewise.
9345 (LIBMPXWRAPPERS_SPEC): Likewise.
9346 (CHKP_SPEC): Likewise.
9347 * config/i386/predicates.md: Likewise.
9348 * dbxout.c (dbxout_type): Likewise.
9349 * doc/extend.texi: Likewise.
9350 * doc/invoke.texi: Likewise.
9351 * doc/md.texi: Likewise.
9352 * doc/tm.texi: Likewise.
9353 * doc/tm.texi.in: Likewise.
9354 * dwarf2out.c (is_base_type): Likewise.
9355 (gen_formal_types_die): Likewise.
9356 (gen_subprogram_die): Likewise.
9357 (gen_type_die_with_usage): Likewise.
9358 (gen_decl_die): Likewise.
9359 (dwarf2out_late_global_decl): Likewise.
9360 * expr.c (expand_assignment): Likewise.
9361 (emit_storent_insn): Likewise.
9362 (store_expr_with_bounds): Likewise.
9363 (store_expr): Likewise.
9364 (expand_expr_real_1): Likewise.
9365 * expr.h (store_expr_with_bounds): Likewise.
9366 * function.c (use_register_for_decl): Likewise.
9367 (struct bounds_parm_data): Likewise.
9368 (assign_parms_augmented_arg_list): Likewise.
9369 (assign_parm_find_entry_rtl): Likewise.
9370 (assign_parm_is_stack_parm): Likewise.
9371 (assign_parm_load_bounds): Likewise.
9372 (assign_bounds): Likewise.
9373 (assign_parms): Likewise.
9374 (expand_function_start): Likewise.
9375 * gcc.c (CHKP_SPEC): Likewise.
9376 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
9377 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
9378 (wrestrict_dom_walker::check_call): Likewise.
9379 * gimple.c (gimple_build_call_from_tree): Likewise.
9380 * gimple.h (enum gf_mask): Likewise.
9381 (gimple_call_with_bounds_p): Likewise.
9382 (gimple_call_set_with_bounds): Likewise.
9383 * gimplify.c (gimplify_init_constructor): Likewise.
9384 * ipa-cp.c (initialize_node_lattices): Likewise.
9385 (propagate_constants_across_call): Likewise.
9386 (find_more_scalar_values_for_callers_subset): Likewise.
9387 * ipa-hsa.c (process_hsa_functions): Likewise.
9388 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
9389 * ipa-icf.c (sem_function::merge): Likewise.
9390 * ipa-inline.c (early_inliner): Likewise.
9391 * ipa-pure-const.c (warn_function_noreturn): Likewise.
9392 (warn_function_cold): Likewise.
9393 (propagate_pure_const): Likewise.
9394 * ipa-ref.h (enum GTY): Likewise.
9395 * ipa-split.c (find_retbnd): Likewise.
9396 (consider_split): Likewise.
9397 (split_function): Likewise.
9398 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
9399 * ipa.c (walk_polymorphic_call_targets): Likewise.
9400 (symbol_table::remove_unreachable_nodes): Likewise.
9401 (process_references): Likewise.
9402 (cgraph_build_static_cdtor_1): Likewise.
9403 * lto-cgraph.c (lto_output_node): Likewise.
9404 (output_refs): Likewise.
9405 (compute_ltrans_boundary): Likewise.
9406 (input_overwrite_node): Likewise.
9407 (input_node): Likewise.
9408 (input_cgraph_1): Likewise.
9409 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
9410 * passes.c (pass_manager::execute_early_local_passes): Likewise.
9411 (class pass_chkp_instrumentation_passes): Likewise.
9412 (make_pass_chkp_instrumentation_passes): Likewise.
9413 * passes.def: Likewise.
9414 * rtl.h (struct GTY): Likewise.
9415 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
9416 * stor-layout.c (layout_type): Likewise.
9417 * symtab.c: Likewise.
9418 * target.def: Likewise.
9419 * targhooks.c (default_chkp_bound_type): Likewise.
9420 (default_chkp_bound_mode): Likewise.
9421 (default_builtin_chkp_function): Likewise.
9422 (default_chkp_function_value_bounds): Likewise.
9423 (default_chkp_make_bounds_constant): Likewise.
9424 (default_chkp_initialize_bounds): Likewise.
9425 * targhooks.h (default_chkp_bound_type): Likewise.
9426 (default_chkp_bound_mode): Likewise.
9427 (default_builtin_chkp_function): Likewise.
9428 (default_chkp_function_value_bounds): Likewise.
9429 (default_chkp_make_bounds_constant): Likewise.
9430 (default_chkp_initialize_bounds): Likewise.
9431 * toplev.c (compile_file): Likewise.
9432 (process_options): Likewise.
9433 * tree-core.h (DEF_BUILTIN): Likewise.
9434 (DEF_BUILTIN_CHKP): Likewise.
9435 * tree-inline.c (declare_return_variable): Likewise.
9436 (remap_gimple_stmt): Likewise.
9437 (copy_bb): Likewise.
9438 (initialize_inlined_parameters): Likewise.
9439 (expand_call_inline): Likewise.
9440 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
9441 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
9442 (make_pass_ipa_chkp_produce_thunks): Likewise.
9443 (make_pass_chkp): Likewise.
9444 (make_pass_chkp_opt): Likewise.
9445 (make_pass_chkp_instrumentation_passes): Likewise.
9446 * tree-pretty-print.c (dump_generic_node): Likewise.
9447 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
9448 * tree-ssa-dce.c (propagate_necessity): Likewise.
9449 (eliminate_unnecessary_stmts): Likewise.
9450 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
9451 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
9452 * tree-ssa-sccvn.h: Likewise.
9453 * tree-ssa-strlen.c (get_string_length): Likewise.
9454 (valid_builtin_call): Likewise.
9455 (adjust_last_stmt): Likewise.
9456 (handle_builtin_strchr): Likewise.
9457 (handle_builtin_strcpy): Likewise.
9458 (handle_builtin_stxncpy): Likewise.
9459 (handle_builtin_memcpy): Likewise.
9460 (handle_builtin_strcat): Likewise.
9461 (strlen_check_and_optimize_stmt): Likewise.
9462 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
9463 * tree-streamer-in.c: Likewise.
9464 * tree-streamer.c (record_common_node): Likewise.
9465 * tree.c (tree_code_size): Likewise.
9466 (wide_int_to_tree_1): Likewise.
9467 (type_contains_placeholder_1): Likewise.
9468 (build_common_tree_nodes): Likewise.
9469 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
9470 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
9471 (POINTER_BOUNDS_P): Likewise.
9472 (BOUNDED_TYPE_P): Likewise.
9473 (BOUNDED_P): Likewise.
9474 (CALL_WITH_BOUNDS_P): Likewise.
9475 (pointer_bounds_type_node): Likewise.
9476 * value-prof.c (gimple_ic): Likewise.
9477 * var-tracking.c (vt_add_function_parameters): Likewise.
9478 * varasm.c (make_decl_rtl): Likewise.
9479 (assemble_start_function): Likewise.
9480 (output_constant): Likewise.
9481 (maybe_assemble_visibility): Likewise.
9482 * varpool.c (ctor_for_folding): Likewise.
9483 * chkp-builtins.def: Remove.
9484 * ipa-chkp.c: Remove.
9485 * ipa-chkp.h: Remove.
9486 * rtl-chkp.c: Remove.
9487 * rtl-chkp.h: Remove.
9488 * tree-chkp-opt.c: Remove.
9489 * tree-chkp.c: Remove.
9490 * tree-chkp.h: Remove.
9492 2018-06-07 Carl Love <cel@us.ibm.com>
9494 * config/rs6000/vsx.md (vextract_fp_from_shorth,
9495 vextract_fp_from_shortl): Add BE support.
9497 2018-06-07 Paul Koning <ni1d@arrl.net>
9499 * compare-elim.c (try_merge_compare): Don't merge compare if
9500 address contains a side effect.
9501 (try_eliminate_compare): Likewise.
9503 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
9505 * config.gcc: Support "tremont".
9506 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
9507 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
9509 * config/i386/i386.c (m_TREMONT): Define.
9510 (processor_target_table): Add "tremont".
9511 (PTA_TREMONT): Define.
9512 (ix86_lea_outperforms): Add TARGET_TREMONT.
9513 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
9514 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
9515 and M_INTEL_GOLDMONT_PLUS.
9516 (fold_builtin_cpu): Add "tremont".
9517 (ix86_add_stmt_cost): Add TARGET_TREMONT.
9518 (ix86_option_override_internal): Add "tremont".
9519 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
9520 (processor_type): Add PROCESSOR_TREMONT.
9521 * config/i386/x86-tune.def: Add m_TREMONT.
9522 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
9524 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
9526 * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
9527 symbol defined for msp430i* devices to be lower case.
9529 2018-06-07 Richard Biener <rguenther@suse.de>
9531 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
9532 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
9533 Properly wrap signed arithmetic if overflow wraps.
9535 2018-06-07 Jakub Jelinek <jakub@redhat.com>
9537 PR tree-optimization/69615
9538 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
9539 of a cast from a same precision integral SSA_NAME in a bb dominated
9540 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
9541 cast to utype if rhs2 has already a compatible type.
9543 2018-06-07 Richard Biener <rguenther@suse.de>
9545 PR tree-optimization/85935
9546 * graphite-scop-detection.c (find_params_in_bb): Analyze
9547 condition operands with respect to the correct loop. Assert
9548 the analysis doesn't fail.
9550 2018-06-04 Carl Love <cel@us.ibm.com>
9552 * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
9553 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
9554 as it is slightly cheaper.
9555 (first_match_or_eos_index_<mode>):
9556 Calculate index using natural element order.
9557 (first_match_index_<mode>):
9558 Calculate index using natural element order.
9559 (first_match_or_eos_index_<mode>):
9560 Calculate index using natural order.
9561 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
9562 for BE and LE modes.
9563 * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
9564 P9V_BUILTIN_VCLZLSBB_V16QI.
9565 * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
9568 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
9570 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
9571 indentation and line wrap for many prototypes. Add missing
9572 @smallexample directives around block of prototypes for vec_xl and
9575 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
9577 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
9578 track if we pass or return IEEE 128-bit floating point.
9579 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
9580 C++ mangling that is compatible with GCC 8.1.
9581 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
9582 (init_cumulative_args): Note if we pass or return IEEE 128-bit
9583 floating point types.
9584 (rs6000_function_arg_advance_1): Likewise.
9585 (rs6000_mangle_type): Optionally generate mangled names that match
9586 what GCC 8.1 generated for IEEE 128-bit floating point types.
9587 (rs6000_globalize_decl_name): If we have an external function that
9588 passes or returns IEEE 128-bit types, generate a weak reference
9589 from the mangled name used in GCC 8.1 to the current mangled
9591 (rs6000_init_builtins): Make __ibm128 use the long double type if
9592 long double is IBM extended double. Make __float128 use the long
9593 double type if long double is IEEE 128-bit.
9596 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
9597 macro for __ibm128 built-in functions.
9598 (PACK_IF): Add __ibm128 pack/unpack functions.
9599 (UNPACK_IF): Likewise.
9600 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
9601 enable long double built-in functions if long double is IEEE
9602 128-bit floating point.
9603 (rs6000_invalid_builtin): Update long double built-in function
9605 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
9606 functions, adjust the built-in function to use the long double
9607 built-in function if __ibm128 and long double are the same type.
9608 * doc/extend.texi (PowerPC builtins): Update documention for
9609 __builtin_{,un}pack_longdouble. Add documentation for
9610 __builtin_{,un}pack_ibm128.
9612 2018-06-06 Jim Wilson <jimw@sifive.com>
9614 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
9615 (struct machine_function): New field interrupt_mode.
9616 (riscv_handle_type_attribute): New function. Add forward declaration.
9617 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
9618 (riscv_expand_epilogue): Check interrupt_mode field.
9619 (riscv_set_current_function): Check interrupt attribute args and
9620 set interrupt_mode field.
9621 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
9622 (riscv_sret, riscv_uret): New.
9623 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
9624 new arguments to interrupt attribute.
9626 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
9629 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
9630 Don't handle -mcpu=power8 if -mpower9-vector is also used.
9632 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
9634 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
9635 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
9636 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
9637 several redundant entries.
9639 2018-06-06 David Malcolm <dmalcolm@redhat.com>
9641 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
9642 type from "rtx" to "rtx_insn *".
9643 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
9644 for local "call_insn", removing cast.
9645 (ix86_expand_call): Likewise, introducing a "call_insn" local.
9647 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
9649 PR tree-optimization/86066
9650 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
9651 for BIT_INSERT_EXPR stores.
9653 2018-06-06 Richard Biener <rguenther@suse.de>
9655 PR tree-optimization/86062
9656 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
9657 component refs ontop
9658 of to be offsetted base.
9660 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
9662 * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
9663 to be static and remove check on interrupt attribute name.
9665 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
9667 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
9668 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
9670 2018-06-05 Steve Ellcey <sellcey@cavium.com>
9673 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
9675 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
9676 Remove second argument, change how error is called.
9677 (aarch64_layout_arg): Remove second argument from
9678 aarch64_err_no_fpadvsimd call.
9679 (aarch64_init_cumulative_args): Ditto.
9680 (aarch64_gimplify_va_arg_expr): Ditto.
9681 * config/aarch64/aarch64.md (mov<mode>): Ditto.
9683 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
9685 * config/i386/i386.md (simple_return_indirect_internal): New expander.
9686 (*simple_return_indirect_internal<mode>): Rename from
9687 simple_return_indirect_internal. Use W mode iterator.
9688 (rstorssp): New expander.
9689 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
9690 (clrssbsy): New expander.
9691 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
9693 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
9695 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
9697 (cmse_check_pointed_object): Likewise.
9699 2018-06-05 Martin Liska <mliska@suse.cz>
9701 PR gcov-profile/47618
9702 * doc/invoke.texi: Document how -fprofile-dir format
9705 2018-06-05 Richard Biener <rguenther@suse.de>
9707 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
9708 removal pretend DOM info isn't available so we do not update
9709 it and only remove edges, not dominated blocks. Actually free
9710 DOM info in case we removed something. Remove unreachable blocks.
9711 (mfb_keep_latches): Work with either DOM info or marked backedges.
9712 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
9713 first. Mark backedges if DOM info isn't available.
9714 (Re-)compute DOM info after cleanup_control_flow_pre.
9716 2018-06-05 Richard Biener <rguenther@suse.de>
9718 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
9719 (locus_discrim_hasher::hash): Adjust.
9720 (locus_discrim_hasher::equal): Likewise.
9721 (next_discriminator_for_locus): Work on line directly.
9722 (same_line_p): Pass in expanded locus1 as well.
9723 (assign_discriminators): Avoid redundant location expansions.
9725 2018-06-05 Richard Biener <rguenther@suse.de>
9727 PR tree-optimization/86046
9728 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
9729 if required after clearing TREE_ADDRESSABLE.
9731 2018-06-05 Richard Biener <rguenther@suse.de>
9733 PR tree-optimization/86047
9734 * tree-ssa-loop.c (for_each_index): Glob handling of all
9735 decls and constants and really handle all of them.
9737 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9740 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
9741 qualifier_void_pointer and qualifier_const_void_pointer.
9742 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
9743 (arm_init_builtins): Handle the above.
9744 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
9745 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
9746 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
9749 2018-06-05 Martin Liska <mliska@suse.cz>
9751 * auto-profile.c (read_autofdo_file): Do not use
9752 gcov_ctr_summary struct.
9753 (afdo_callsite_hot_enough_for_early_inline): Likewise.
9754 * coverage.c (struct counts_entry): Likewise.
9755 (read_counts_file): Read just single summary entry.
9756 (get_coverage_counts): Use gcov_summary struct.
9757 * coverage.h (get_coverage_counts): Likewise.
9758 * gcov-dump.c (dump_working_sets): Likewise.
9759 (tag_summary): Dump just single summary.
9760 * gcov-io.c (gcov_write_summary): Write just histogram
9762 (gcov_read_summary): Read just single summary.
9763 (compute_working_sets): Use gcov_summary struct.
9764 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
9765 of GCOV_COUNTERS_SUMMABLE.
9766 (GCOV_COUNTERS_SUMMABLE): Remove.
9767 (GCOV_FIRST_VALUE_COUNTER): Replace with
9768 GCOV_COUNTER_V_INTERVAL.
9769 (struct gcov_ctr_summary): Remove.
9770 (struct gcov_summary): Directly use fields of former
9772 (compute_working_sets): Use gcov_summary struct.
9773 * gcov.c (read_count_file): Do not use ctrs fields.
9774 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
9776 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
9778 * profile.c: Likewise.
9779 * profile.h: Likewise.
9781 2018-06-05 Martin Liska <mliska@suse.cz>
9783 PR gcov-profile/84846
9784 * gcov.c (output_lines): Print working directory only
9785 in intermediate format.
9787 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
9789 * config/s390/s390-builtin-types.def: Add void function type.
9790 * config/s390/s390-builtins.def: Use the function type for the
9793 2018-06-04 Jim Wilson <jimw@sifive.com>
9795 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
9797 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
9798 and maybe_eh_return. Change regno to unsigned int. Use new args to
9799 handle EH_RETURN_DATA_REGNO registers properly.
9800 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
9801 (riscv_expand_epilogue): Update comment. Change argument name and
9802 type. Update code to use new name and type. Pass new args to
9803 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
9805 * config/riscv/riscv.md (NORMAL_RETURN): New.
9806 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
9807 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
9808 (eh_return): Call gen_eh_return_internal and emit barrier.
9809 (eh_return_internal): Call riscv_expand_epilogue.
9811 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
9813 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
9814 bit_insertion field and declare can_be_merged_into method.
9815 (merged_store_group::can_be_merged_into): New method.
9816 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
9817 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
9818 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
9820 2018-06-04 Richard Biener <rguenther@suse.de>
9822 PR tree-optimization/85955
9823 * builtins.c (fold_builtin_sincos): Convert pointers to
9824 destination to appropriate type before dereferencing.
9826 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
9828 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
9830 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
9832 * expr.c (expand_expr_real_1): Force the operand into memory if
9833 its TYPE_MODE is BLKmode and if there is no integer mode for
9834 the number of bits being extracted.
9836 2018-06-04 Jakub Jelinek <jakub@redhat.com>
9840 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
9841 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
9843 2018-06-04 Richard Biener <rguenther@suse.de>
9845 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
9846 (cleanup_tree_cfg_noloop): ... single caller. Do
9847 start_recording_case_labels later.
9849 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
9851 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
9852 to _IMMINTRIN_H_INCLUDED.
9853 * config/i386/pconfigintrin.h: Ditto.
9854 * config/i386/waitpkgintrin.h: Ditto.
9855 * config/i386/immintrin.h: Add includes for sgxintrin.h,
9856 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
9857 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
9858 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
9859 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
9860 waitpkgintrin.h and cldemoteintrin.h.
9862 2018-06-04 Richard Biener <rguenther@suse.de>
9864 PR tree-optimization/86038
9865 * tracer.c (find_best_successor): Check probability for
9866 being initialized, bail out if not.
9868 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
9871 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
9872 of bits to ignore when comparing architectures.
9874 2018-06-04 Jakub Jelinek <jakub@redhat.com>
9876 PR tree-optimization/69615
9877 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
9878 maximum or minimum of the type, try to merge it also as if
9879 range1 is + [-, x - 1] or + [x + 1, -].
9882 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
9884 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
9886 PR tree-optimization/86034
9887 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
9888 the unsigned bitfield type in a bit insertion sequence if it does not
9889 have a larger precision than the bitfield size.
9890 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
9892 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
9894 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
9896 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
9898 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
9899 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
9900 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
9901 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
9903 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
9905 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
9906 Disable -fdelete-null-pointer-checks for ELF toolchain.
9908 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
9909 Kito Cheng <kito.cheng@gmail.com>
9911 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
9912 (nds32le-*-*, nds32be-*-*): Integrate checking process.
9913 (nds32*-*-*): Add glibc and uclibc conditions.
9914 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
9915 (TARGET_EXCEPT_UNWIND_INFO): Define.
9916 * config/nds32/elf.h: New file.
9917 * config/nds32/linux.h: New file.
9918 * config/nds32/nds32-elf.opt: New file.
9919 * config/nds32/nds32-linux.opt: New file.
9920 * config/nds32/nds32-fp-as-gp.c
9921 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
9922 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
9924 (nds32_asm_file_end): Ditto.
9925 (nds32_print_operand): Ditto.
9926 (nds32_insert_attributes): Ditto.
9927 (nds32_init_libfuncs): New function.
9928 (TARGET_HAVE_TLS): Define.
9929 (TARGET_INIT_LIBFUNCS): Define.
9930 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
9932 (TARGET_ELF): Apply different mcmodel setting.
9933 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
9934 been migrated into elf.h and linux.h files.
9935 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
9936 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
9937 (mcmodel): The content has been migrated into nds32-elf.opt and
9938 nds32-linux.opt files.
9939 * config/nds32/t-elf: New file.
9940 * config/nds32/t-linux: New file.
9942 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
9943 Shiva Chen <shiva0217@gmail.com>
9945 * config/nds32/constants.md (unspec_volatile_element): Add
9946 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
9947 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
9949 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
9950 (make_pass_nds32_fp_as_gp): Declare.
9951 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
9953 (nds32_asm_function_end_prologue): Remove unused asm output.
9954 (nds32_asm_function_begin_epilogue): Remove unused asm output.
9955 (nds32_asm_file_start): Output necessary fp_as_gp information.
9956 (nds32_option_override): Adjust register usage.
9957 (nds32_expand_prologue): Consider fp_as_gp situation.
9958 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
9959 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
9962 (simple_return): Ditto.
9963 (omit_fp_begin): Output special directive for fp_as_gp.
9964 (omit_fp_end): Output special directive for fp_as_gp.
9965 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
9966 mforbid-fp-as-gp): New options.
9968 2018-06-01 Mark Wielaard <mark@klomp.org>
9970 * dwarf2out.c (dwarf2out_finish): Remove generation of
9971 DW_AT_loclists_base.
9973 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
9975 * gimple-ssa-store-merging.c: Include gimple-fold.h.
9976 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
9977 (struct merged_store_group): Add bit_insertion field.
9978 (dump_char_array): Use standard hexadecimal format.
9979 (merged_store_group::merged_store_group): Set bit_insertion to false.
9980 (merged_store_group::apply_stores): Use optimal buffer size. Deal
9981 with BIT_INSERT_EXPR stores. Move up code updating the mask and
9982 also print the mask in the dump file.
9983 (pass_store_merging::gate): Minor tweak.
9984 (imm_store_chain_info::coalesce_immediate): Fix wrong association
9985 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
9986 stores with INTEGER_CST stores.
9987 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
9988 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
9989 and use it throughout. Generate bit insertion sequences if need be.
9990 (pass_store_merging::process_store): Remove redundant condition.
9991 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
9993 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
9995 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
9996 the 128-bit floating point types. Fix function comment.
9998 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10000 * config/aarch64/aarch64-simd.md
10001 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
10003 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
10006 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
10008 PR tree-optimization/85989
10009 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
10011 (backprop::intersect_uses): Check it when deciding whether this
10012 is a backedge reference.
10013 (backprop::process_block): Add each phi to m_visited_phis
10014 after visiting it, then clear it at the end.
10016 2018-06-01 Richard Biener <rguenther@suse.de>
10018 * tree-vectorizer.h (vect_dr_stmt): New function.
10019 (vect_get_load_cost): Adjust.
10020 (vect_get_store_cost): Likewise.
10021 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
10022 Use vect_dr_stmt instead of DR_SMTT.
10023 (vect_record_base_alignments): Likewise.
10024 (vect_calculate_target_alignment): Likewise.
10025 (vect_compute_data_ref_alignment): Likewise and make static.
10026 (vect_update_misalignment_for_peel): Likewise.
10027 (vect_verify_datarefs_alignment): Likewise.
10028 (vector_alignment_reachable_p): Likewise.
10029 (vect_get_data_access_cost): Likewise. Pass down
10030 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
10031 (vect_get_peeling_costs_all_drs): Likewise.
10032 (vect_peeling_hash_get_lowest_cost): Likewise.
10033 (vect_enhance_data_refs_alignment): Likewise.
10034 (vect_find_same_alignment_drs): Likewise.
10035 (vect_analyze_data_refs_alignment): Likewise.
10036 (vect_analyze_group_access_1): Likewise.
10037 (vect_analyze_group_access): Likewise.
10038 (vect_analyze_data_ref_access): Likewise.
10039 (vect_analyze_data_ref_accesses): Likewise.
10040 (vect_vfa_segment_size): Likewise.
10041 (vect_small_gap_p): Likewise.
10042 (vectorizable_with_step_bound_p): Likewise.
10043 (vect_prune_runtime_alias_test_list): Likewise.
10044 (vect_analyze_data_refs): Likewise.
10045 (vect_supportable_dr_alignment): Likewise.
10046 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
10047 (vect_gen_prolog_loop_niters): Likewise.
10048 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
10049 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
10051 (vect_recog_mask_conversion_pattern): Likewise.
10052 (vect_try_gather_scatter_pattern): Likewise.
10053 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
10054 to vect_get_store_cost.
10055 (vect_get_store_cost): Get stmt_info instead of DR.
10056 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
10057 (vect_get_load_cost): Get stmt_info instead of DR.
10059 2018-06-01 Richard Biener <rguenther@suse.de>
10061 PR middle-end/86017
10062 * gimple-fold.c (var_decl_component_p): Also allow offsetted
10063 vars wrapped in MEM_REFs.
10065 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
10067 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
10068 Fix subreg tests so that we only return a choice between
10069 GENERAL_REGS and FP_REGS if the original classes included both.
10071 2018-06-01 Richard Biener <rguenther@suse.de>
10074 * tree-ssa-structalias.c (get_function_part_constraint):
10075 Handle NULL fi->decl.
10076 (find_func_aliases_for_call): Properly handle indirect
10077 fi from direct call.
10078 (find_func_clobbers): Likewise.
10079 (ipa_pta_execute): Likewise.
10080 (create_variable_info_for): For functions that are ifunc_resolver
10081 resolve to a varinfo that contains the result of the resolver call.
10082 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
10085 2018-05-31 Michael Collison <michael.collison@arm.com>
10087 * config/aarch64/aarch64.md:
10088 (*fix_to_zero_extenddfdi2): New pattern.
10089 * gcc.target/aarch64/fix_extend1.c: New testcase.
10091 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
10093 PR middle-end/78809
10094 PR middle-end/83026
10095 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
10096 and BUILT_IN_STRNCMP_EQ.
10097 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
10098 BUILT_IN_STRNCMP_EQ.
10099 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
10100 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
10101 (gimple_fold_builtin): Likewise.
10102 * tree-ssa-strlen.c (compute_string_length): New function.
10103 (determine_min_obsize): New function.
10104 (handle_builtin_string_cmp): New function to handle calls to
10105 string compare functions.
10106 (strlen_optimize_stmt): Add handling to builtin string compare
10108 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
10109 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
10110 * tree.c (build_common_builtin_nodes): Add new defines of
10111 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
10113 2018-05-31 Jakub Jelinek <jakub@redhat.com>
10116 * bb-reorder.c (pass_partition_blocks::gate): Return false for
10117 functions with naked attribute.
10119 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
10121 * config/i386/sse.md (avx_vec_concat<mode>):
10122 Substitute concat_tg_mode mode attribute with xtg_mode.
10123 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
10124 (concat_tg_mode): Remove mode attribute.
10126 2018-05-31 Martin Sebor <msebor@redhat.com>
10129 * calls.c (alloc_max_size): Correct a logic error/typo.
10130 Treat excessive arguments as infinite. Warn for invalid arguments.
10131 * doc/invoke.texi (-Walloc-size-larger-than): Update.
10133 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
10136 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
10137 and movx for Haswell.
10139 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
10140 Cesar Philippidis <cesar@codesourcery.com>
10142 PR middle-end/85879
10143 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
10144 when emitting error on private/firstprivate reductions.
10145 * omp-low.c (lower_omp_target): Avoid reference-type processing
10146 on pointers for firstprivate clause.
10148 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
10150 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
10153 * config/aarch64/aarch64-simd.md
10154 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
10155 (aarch64_ld1_x3_<mode>): Likewise
10156 (aarch64_st1x2<VALLDIF:mode>): Likewise
10157 (aarch64_st1_x2_<mode>): Likewise
10158 (aarch64_st1x3<VALLDIF:mode>): Likewise
10159 (aarch64_st1_x3_<mode>): Likewise
10160 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
10161 (vld1_s8_x3): Likewise.
10162 (vld1_u16_x3): Likewise.
10163 (vld1_s16_x3): Likewise.
10164 (vld1_u32_x3): Likewise.
10165 (vld1_s32_x3): Likewise.
10166 (vld1_u64_x3): Likewise.
10167 (vld1_s64_x3): Likewise.
10168 (vld1_f16_x3): Likewise.
10169 (vld1_f32_x3): Likewise.
10170 (vld1_f64_x3): Likewise.
10171 (vld1_p8_x3): Likewise.
10172 (vld1_p16_x3): Likewise.
10173 (vld1_p64_x3): Likewise.
10174 (vld1q_u8_x3): Likewise.
10175 (vld1q_s8_x3): Likewise.
10176 (vld1q_u16_x3): Likewise.
10177 (vld1q_s16_x3): Likewise.
10178 (vld1q_u32_x3): Likewise.
10179 (vld1q_s32_x3): Likewise.
10180 (vld1q_u64_x3): Likewise.
10181 (vld1q_s64_x3): Likewise.
10182 (vld1q_f16_x3): Likewise.
10183 (vld1q_f32_x3): Likewise.
10184 (vld1q_f64_x3): Likewise.
10185 (vld1q_p8_x3): Likewise.
10186 (vld1q_p16_x3): Likewise.
10187 (vld1q_p64_x3): Likewise.
10188 (vst1_s64_x2): Likewise.
10189 (vst1_u64_x2): Likewise.
10190 (vst1_f64_x2): Likewise.
10191 (vst1_s8_x2): Likewise.
10192 (vst1_p8_x2): Likewise.
10193 (vst1_s16_x2): Likewise.
10194 (vst1_p16_x2): Likewise.
10195 (vst1_s32_x2): Likewise.
10196 (vst1_u8_x2): Likewise.
10197 (vst1_u16_x2): Likewise.
10198 (vst1_u32_x2): Likewise.
10199 (vst1_f16_x2): Likewise.
10200 (vst1_f32_x2): Likewise.
10201 (vst1_p64_x2): Likewise.
10202 (vst1q_s8_x2): Likewise.
10203 (vst1q_p8_x2): Likewise.
10204 (vst1q_s16_x2): Likewise.
10205 (vst1q_p16_x2): Likewise.
10206 (vst1q_s32_x2): Likewise.
10207 (vst1q_s64_x2): Likewise.
10208 (vst1q_u8_x2): Likewise.
10209 (vst1q_u16_x2): Likewise.
10210 (vst1q_u32_x2): Likewise.
10211 (vst1q_u64_x2): Likewise.
10212 (vst1q_f16_x2): Likewise.
10213 (vst1q_f32_x2): Likewise.
10214 (vst1q_f64_x2): Likewise.
10215 (vst1q_p64_x2): Likewise.
10216 (vst1_s64_x3): Likewise.
10217 (vst1_u64_x3): Likewise.
10218 (vst1_f64_x3): Likewise.
10219 (vst1_s8_x3): Likewise.
10220 (vst1_p8_x3): Likewise.
10221 (vst1_s16_x3): Likewise.
10222 (vst1_p16_x3): Likewise.
10223 (vst1_s32_x3): Likewise.
10224 (vst1_u8_x3): Likewise.
10225 (vst1_u16_x3): Likewise.
10226 (vst1_u32_x3): Likewise.
10227 (vst1_f16_x3): Likewise.
10228 (vst1_f32_x3): Likewise.
10229 (vst1_p64_x3): Likewise.
10230 (vst1q_s8_x3): Likewise.
10231 (vst1q_p8_x3): Likewise.
10232 (vst1q_s16_x3): Likewise.
10233 (vst1q_p16_x3): Likewise.
10234 (vst1q_s32_x3): Likewise.
10235 (vst1q_s64_x3): Likewise.
10236 (vst1q_u8_x3): Likewise.
10237 (vst1q_u16_x3): Likewise.
10238 (vst1q_u32_x3): Likewise.
10239 (vst1q_u64_x3): Likewise.
10240 (vst1q_f16_x3): Likewise.
10241 (vst1q_f32_x3): Likewise.
10242 (vst1q_f64_x3): Likewise.
10243 (vst1q_p64_x3): Likewise.
10245 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
10247 * config/msp430/msp430.c (msp430_output_labelref): Prepend
10248 user_label_prefix to name.
10250 * tree-core.h: Update comment about the format of NAME string
10251 passed to handler in attribute_spec.
10253 * config/msp430/msp430.md: Remove erroneous subreg expression from
10254 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
10255 zero_extend{q,h}isi2.
10257 2018-05-30 Borislav Petkov <bp@suse.de>
10259 * doc/extend.texi: Document some architecture specific
10260 constraints and sort entries.
10262 2018-05-30 Martin Sebor <msebor@redhat.com>
10264 PR middle-end/85369
10265 * builtins.c (expand_builtin_stpcpy_1): New function.
10266 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
10267 only if the former succeeds.
10269 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
10271 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
10274 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
10276 * doc/invoke.texi (-flinker-output): Document
10278 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
10280 * passes.c (ipa_write_summaries): Only modify statements if body
10282 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
10283 incrementally linking.
10284 (ipa_passes): Likewise.
10285 * lto-cgraph.c (lto_output_node): When incrementally linking do not
10286 pass down resolution info.
10287 * common.opt (flag_incremental_link): Update info.
10288 * gcc.c (plugin specs): Turn flinker-output=* to
10289 -plugin-opt=-linker-output-known
10290 * toplev.c (compile_file): Also cut compilation when doing incremental
10292 * flag-types. (enum lto_partition_model): Add
10293 LTO_LINKER_OUTPUT_NOLTOREL.
10294 (invoke.texi): Add -flinker-output docs.
10295 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
10296 link same way as WPA; do not stream in dead initializers.
10298 * dwarf2out.c (dwarf2out_die_ref_for_decl,
10299 darf2out_register_external_decl): Support incremental link.
10301 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
10303 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
10305 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
10307 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
10308 it down to simple_object_copy_lto_debug_sections.
10309 (run_gcc): Determine incremental LTO link time and configure
10310 lto1 into non-wpa mode, disable renaming of debug sections.
10312 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
10314 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
10315 descriptions of various incorrectly documented functions.
10317 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
10320 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
10321 address check not strict.
10323 2018-05-30 Richard Biener <rguenther@suse.de>
10325 PR tree-optimization/85964
10326 * tracer.c (better_p): Drop initialized count check, we only
10327 call the function with initialized counts now.
10328 (find_best_successor): Do find a best edge if one
10329 has uninitialized count.
10330 (find_best_predecessor): Likewise. Do BB frequency check only
10331 if count is initialized.
10333 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
10335 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
10336 (aarch64_ldrstr_offset_compare): New.
10337 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
10338 load/store orderings.
10339 (aarch64_gen_adjusted_ldpstp): Likewise.
10341 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
10343 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
10344 Check for subset of GENERAL_REGS and FP_REGS.
10345 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
10348 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
10350 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
10351 and wi::to_poly_offset. Add the current offset and then check
10352 whether the sum fits, rather than using an unchecked addition of
10353 a checked term. Check for a shwi rather than a uhwi.
10354 * expr.c (get_bit_range): Use tree_to_poly_uint64.
10355 (store_constructor): Use poly_int_tree_p.
10356 (expand_expr_real_1): Likewise.
10357 * function.c (assign_temp): Likewise.
10358 * fold-const.c (const_binop): Use poly_int_tree_p and
10359 wi::to_poly_offset.
10360 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
10362 * ipa-icf-gimple.c (func_checker::compare_operand): Use
10363 to_poly_offset for MEM offsets.
10364 * ipa-icf.c (sem_variable::equals): Likewise.
10365 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
10366 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
10367 wi::to_poly_offset for BIT_FIELD_REF offsets.
10368 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
10369 wi::to_poly_offset.
10370 * var-tracking.c (emit_note_insn_var_location): Use
10371 tree_to_poly_uint64.
10373 2018-05-29 Jim Wilson <jimw@sifive.com>
10375 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
10377 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
10380 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
10381 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
10383 (sse4_1_round<mode>2): Use nonimmediate_operand
10384 for operand 1 predicate.
10386 2018-05-29 Martin Sebor <msebor@redhat.com>
10387 Richard Biener <rguenther@suse.de>
10390 * calls.c (get_size_range): Call determine_value_range instead
10391 of get_value_range..
10392 * tree-vrp.h (determine_value_range): Declared new function.
10393 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
10395 2018-05-29 Richard Biener <rguenther@suse.de>
10397 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
10398 sure to use non-pattern stmts for get_earlier_stmt arguments.
10399 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
10400 called on pattern stmts.
10401 (get_later_stmt): Likewise.
10403 2018-05-29 Martin Liska <mliska@suse.cz>
10405 PR gcov-profile/85759
10406 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
10409 2018-05-29 Jakub Jelinek <jakub@redhat.com>
10411 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
10413 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
10417 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
10418 VEC_PACK_FLOAT_EXPR): New tree codes.
10419 * tree-pretty-print.c (op_code_prio): Handle
10420 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
10421 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
10422 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
10423 * tree-inline.c (estimate_operator_cost): Likewise.
10424 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
10425 * fold-const.c (const_binop): Likewise.
10426 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
10427 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
10428 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
10429 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
10430 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
10431 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
10432 * expr.c (expand_expr_real_2): Likewise.
10433 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
10434 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
10435 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
10437 * optabs.c (expand_widen_pattern_expr): For
10438 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
10439 sign from result type rather than operand's type.
10440 (expand_binop_directly): For vec_packu_float_optab and
10441 vec_packs_float_optab allow result type to be different from operand's
10443 * optabs-tree.c (optab_for_tree_code): Handle
10444 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
10445 VEC_PACK_FLOAT_EXPR. Formatting fixes.
10446 * tree-vect-generic.c (expand_vector_operations_1): Handle
10447 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
10448 VEC_PACK_FLOAT_EXPR.
10449 * tree-vect-stmts.c (supportable_widening_operation): Handle
10451 (supportable_narrowing_operation): Handle FLOAT_EXPR.
10452 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
10453 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
10454 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
10455 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
10457 (vec_pack<floatprefix>_float_<mode>): New expander.
10458 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
10460 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
10461 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
10462 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
10463 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
10464 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
10466 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
10467 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
10468 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
10469 VEC_PACK_FLOAT_EXPR): Document.
10471 2018-05-29 Richard Biener <rguenther@suse.de>
10473 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
10475 (stmt_vec_info_vec): Make pointer.
10476 (init_stmt_vec_info_vec): Remove.
10477 (free_stmt_vec_info_vec): Likewise.
10478 (set_stmt_vec_info_vec): New function.
10479 (free_stmt_vec_infos): Likewise.
10480 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
10481 (set_vinfo_for_stmt): Likewise.
10482 (get_earlier_stmt): Likewise.
10483 (get_later_stmt): Likewise.
10484 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
10485 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
10486 (vec_info::~vec_info): Free stmt_vec_infos.
10487 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
10488 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
10489 (pass_slp_vectorize::execute): Likewise.
10490 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
10491 (free_stmt_vec_info_vec): Likewise.
10492 (set_stmt_vec_info_vec): New function.
10493 (free_stmt_vec_infos): Likewise.
10494 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
10495 the global stmt_vec_info_vec.
10496 * tree-parloops.c (gather_scalar_reductions): Use
10497 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
10500 2018-05-29 Richard Biener <rguenther@suse.de>
10502 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
10504 2018-05-29 Martin Liska <mliska@suse.cz>
10505 David Malcolm <dmalcolm@redhat.com>
10507 * vec.c (test_reverse): New.
10508 (vec_c_tests): Add new test.
10509 * vec.h (vl_ptr>::reverse): New function.
10511 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
10513 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
10515 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
10518 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
10520 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
10522 2018-05-28 Richard Biener <rguenther@suse.de>
10524 PR tree-optimization/85933
10525 * tree-vect-data-refs.c (vect_record_base_alignments): Only
10526 look at stmts marked as vectorizable.
10528 2018-05-28 Richard Biener <rguenther@suse.de>
10530 PR tree-optimization/85934
10531 * tree-vect-generic.c (expand_vector_operations_1): Hoist
10532 vector boolean check before scalar optimization.
10534 2018-05-28 Jakub Jelinek <jakub@redhat.com>
10536 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
10539 2018-05-28 Mark Wielaard <mark@klomp.org>
10541 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
10542 if it is an expression containing a minus sign.
10544 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
10546 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
10548 2018-05-27 Paul Koning <ni1d@arrl.net>
10550 * config/pdp11/pdp11.md (truncsihi2): Remove.
10552 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
10553 Chung-Ju Wu <jasonwucj@gmail.com>
10555 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
10557 (unaligned_store_dw): Ditto.
10558 * config/nds32/nds32-memory-manipulation.c
10559 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
10560 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
10561 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
10562 (emit_setmem_word_loop): Rename to ...
10563 (emit_setmem_doubleword_loop): ... this.
10564 (nds32_gen_dup_4_byte_to_word_value): New function.
10565 (nds32_gen_dup_8_byte_to_double_word_value): New function.
10566 (nds32_expand_setmem_loop): Refine implementation.
10567 (nds32_expand_setmem_loop_v3m): Ditto.
10568 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
10571 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
10573 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
10575 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
10577 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
10578 (nds32_init_machine_status): Initialize machine->attr_naked_p and
10579 machine->attr_no_prologue_p.
10580 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
10581 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
10582 (nds32_expand_epilogue): Consider attr_naked_p.
10583 (nds32_expand_epilogue_v3pop): Likewise.
10584 (nds32_can_use_return_insn): Likewise.
10585 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
10586 attr_no_prologue_p fields.
10587 * config/nds32/nds32.opt (mret-in-naked-func): New option.
10589 2018-05-27 Jakub Jelinek <jakub@redhat.com>
10592 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
10594 * config/i386/sse.md
10595 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
10597 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
10599 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
10601 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
10603 (*<floatsuffix>floatv2div2sf2): Rename to ...
10604 (*float<floatunssuffix>v2div2sf2): ... this.
10605 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
10606 (float<floatunssuffix>v2div2sf2_mask): ... this.
10607 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
10608 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
10609 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
10611 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
10613 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
10615 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
10617 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
10619 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
10621 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
10622 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
10623 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
10624 gen_ufix_truncv8dfv8si2.
10625 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
10626 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
10627 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
10628 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
10629 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
10630 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
10631 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
10632 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
10634 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
10638 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
10640 2018-05-25 Jim Wilson <jimw@sifive.com>
10642 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
10643 * config/riscv/riscv.c (struct machine_function): Add
10644 interrupt_handler_p and attribute_checked_p fields.
10645 (riscv_attribute_table): Add interrupt.
10646 (riscv_interrupt_type_p): New.
10647 (riscv_save_reg_p): Save extra regs for interrupt handler.
10648 (riscv_use_save_libcall): Return false for interrupt handler.
10649 (riscv_first_stack_step): Add forward declaration.
10650 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
10651 for interrupt handler with large frame. Use it for saved reg list.
10652 (riscv_expand_prologue): Move flag_stack_usage_info support to
10653 eliminate duplication.
10654 (riscv_expand_epilogue): Generate mret for interrupt handler.
10655 (riscv_epilogue_uses): New.
10656 (riscv_can_use_return_insn): Return false for interrupt handler.
10657 (riscv_function_ok_for_sibcall): Likewise.
10658 (riscv_set_current_function): Add interrupt handler support.
10659 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
10660 * config/riscv/riscv.md (UNSPECV_MRET): New.
10662 (riscv_frflags, riscv_fsflags): Use tab after opcode.
10664 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
10666 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
10668 PR tree-optimization/85712
10669 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
10670 this candidate has already been replaced in-situ by a copy.
10672 2018-05-25 Jason Merrill <jason@redhat.com>
10674 PR c++/80485 - inline function non-zero address.
10675 * symtab.c (nonzero_address): Check DECL_COMDAT.
10677 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
10680 * config/alpha/alpha.md (ashlsi3): New insn pattern.
10681 (*ashlsi_se): Rename from *ashldi_se. Define as sign
10682 extension of SImode operation. Use const123_operand predicate.
10683 (*saddsi_1): Remove.
10684 (*saddl_se_1): Ditto.
10685 (*ssubsi_1): Ditto.
10686 (*ssubl_se_1): Ditto.
10687 * config/alpha/predicates.md (const123_operand): New predicate.
10688 * config/alpha/constraints.md (P): Use IN_RANGE.
10690 2018-05-25 Richard Biener <rguenther@suse.de>
10692 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
10694 (ref_maybe_used_by_stmt_p): Likewise.
10695 (stmt_may_clobber_ref_p): Likewise.
10696 (stmt_may_clobber_ref_p_1): Likewise.
10697 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
10699 (ref_maybe_used_by_stmt_p): Likewise.
10700 (stmt_may_clobber_ref_p): Likewise.
10701 (stmt_may_clobber_ref_p_1): Likewise.
10702 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
10703 the alias oracle to disambiguate DRs with stmts DR analysis
10705 (vect_analyze_data_refs): Do not give up on not analyzable
10706 DRs for BB vectorization. Remove code truncating the dataref
10709 2018-05-25 Jakub Jelinek <jakub@redhat.com>
10712 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
10713 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
10714 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
10716 2018-05-25 Richard Biener <rguenther@suse.de>
10718 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
10719 function, combining stmt data ref gathering and fatal analysis
10721 (vect_analyze_data_refs): Remove now redudnant code and simplify.
10722 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
10723 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
10724 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
10725 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
10727 2018-05-25 Bin Cheng <bin.cheng@arm.com>
10729 PR tree-optimization/85720
10730 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
10731 SCC if all partitions are builtins.
10732 (version_loop_by_alias_check): New parameter. Generate cancelable
10733 runtime alias check if all partitions are builtins.
10734 (distribute_loop): Update call to above function.
10736 2018-05-25 Bin Cheng <bin.cheng@arm.com>
10738 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
10739 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
10740 (parm_default_def_partition_arg): Ditto.
10741 (set_parm_default_def_partition): Ditto.
10742 (get_parm_default_def_partitions): Ditto and make it static.
10743 (get_undefined_value_partitions): Ditto and make it static.
10744 (remove_ssa_form): Refactor call to init_var_map here.
10745 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
10746 computation for loop region.
10747 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
10748 (register_default_def): Delete.
10749 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
10750 (parm_default_def_partition_arg): Ditto.
10751 (set_parm_default_def_partition): Ditto.
10752 (get_parm_default_def_partitions): Ditto and make it static.
10753 (get_undefined_value_partitions): Ditto and make it static.
10754 (coalesce_with_default, coalesce_with_default): Update comment.
10755 (create_coalesce_list_for_region): New func factored out from
10756 create_outofssa_var_map.
10757 (populate_coalesce_list_for_outofssa): New func factored out from
10758 create_outofssa_var_map and coalesce_ssa_name.
10759 (create_outofssa_var_map): Delete.
10760 (coalesce_ssa_name): Refactor to support live range computation.
10761 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
10762 (get_parm_default_def_partitions): Delete.
10763 (get_undefined_value_partitions): Ditto.
10764 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
10765 computation for loop region.
10766 (new_tree_live_info, loe_visit_block): Ditto.
10767 (live_worklist, set_var_live_on_entry): Ditto.
10768 (calculate_live_on_exit, verify_live_on_entry): Ditto.
10769 * tree-ssa-live.h (struct _var_map): New fields.
10770 (init_var_map): Change decl.
10771 (region_contains_p): New.
10773 2018-05-25 Bin Cheng <bin.cheng@arm.com>
10775 * tree-ssa-live.h (live_merge_and_clear): Delete.
10777 2018-05-25 Richard Biener <rguenther@suse.de>
10780 * tree-dump.c (dequeue_and_dump): Remove access to removed
10781 operand 2 of a SWITCH_EXPR.
10783 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
10785 * doc/sourcebuild.texi (vect_double_cond_arith): Include
10786 multiplication and division.
10787 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
10788 (cond_udiv@var{m}, cond_umod@var{m}): Document.
10789 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
10790 (cond_udiv_optab, cond_umod_optab): New optabs.
10791 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
10792 (IFN_COND_RDIV): New internal functions.
10793 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
10794 TRUNC_MOD_EXPR and RDIV_EXPR.
10795 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
10796 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
10798 (SVE_INT_BINARY): Include mult.
10799 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
10800 (optab, sve_int_op): Handle mult.
10801 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
10803 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
10804 for SVE_INT_BINARY_SD.
10806 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
10808 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
10809 (optab, sve_int_op): Handle div and udiv.
10810 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
10811 for SVE_INT_BINARY_SD.
10812 (*<optab><mode>3): New insn for the same.
10814 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
10816 * tree-vect-patterns.c: Include predict.h.
10817 (vect_recog_divmod_pattern): Restrict check for division support
10818 to when optimizing for size.
10820 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
10822 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
10823 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
10824 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
10825 (gimple_match_op::set_op): Likewise.
10826 (gimple_resimplify4): Declare.
10827 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
10828 (expr::gen_transform): Likewise.
10829 (decision_tree::gen): Generate a simplification routine for 4 operands.
10830 * gimple-match-head.c (gimple_simplify): Add an overload for
10831 4 operands. In the top-level function, handle up to 4 call
10832 arguments and call gimple_resimplify4.
10833 (gimple_resimplify4): New function.
10834 (build_call_internal): Pass a fourth operand.
10835 (maybe_push_to_seq): Likewise.
10836 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
10837 Fold VEC_COND_EXPRs of an operation and a default value into
10838 an IFN_COND_* function if possible.
10839 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
10841 (SVE_COND_FP_BINARY): Include them.
10842 (optab, sve_fp_op): Handle them.
10843 (SVE_INT_BINARY_REV): New code iterator.
10844 (SVE_COND_FP_BINARY_REV): New int iterator.
10845 (commutative): New int attribute.
10846 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
10848 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
10850 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
10851 (*cond_<optab><mode>): New patterns for reversed operands.
10853 2018-05-25 Richard Biener <rguenther@suse.de>
10855 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
10856 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
10857 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
10858 (STMT_VINFO_GROUPED_ACCESS): Adjust.
10859 * tree-vect-data-refs.c (everywhere): Adjust users.
10860 * tree-vect-loop.c (everywhere): Likewise.
10861 * tree-vect-slp.c (everywhere): Likewise.
10862 * tree-vect-stmts.c (everywhere): Likewise.
10863 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
10865 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10867 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
10869 (gcc_cv_as_section_exclude): ... this.
10870 Try Solaris as #exclude syntax.
10871 * configure: Regenerate.
10872 * config.in: Regenerate.
10873 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
10875 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
10876 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
10878 * varasm.c (default_elf_asm_named_section): Don't check if
10879 HAVE_GAS_SECTION_EXCLUDE is defined.
10881 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
10883 * doc/md.texi: Update the documentation of the cond_* optabs
10884 to mention the new final operand. Fix GET_MODE_NUNITS call.
10885 Describe the scalar case too.
10886 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
10887 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
10889 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
10890 (get_conditional_internal_fn): Update comment.
10891 * tree-vect-loop.c (vectorizable_reduction): Pass the original
10892 accumulator value as a final argument to conditional functions.
10893 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
10894 a define_expand and add an "else" operand. Assert for now that
10895 the else operand is equal to operand 2. Use SVE_INT_BINARY and
10896 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
10897 (*cond_<optab><mode>): New patterns.
10898 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
10899 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
10900 (UNSPEC_COND_EOR): Delete.
10901 (optab): Remove associated mappings.
10902 (SVE_INT_BINARY): New code iterator.
10903 (sve_int_op): Remove int attribute and add "minus" to the code
10905 (SVE_COND_INT_OP): Delete.
10906 (SVE_COND_FP_OP): Rename to...
10907 (SVE_COND_FP_BINARY): ...this.
10909 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
10911 * optabs.c (can_reuse_operands_p): New function.
10912 (maybe_legitimize_operands): Try to reuse the results for
10915 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
10917 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
10918 Add {q} suffix to insn mnemonic.
10920 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
10922 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
10923 (msp430_warn_func_return): New.
10925 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
10927 * fold-const.c (tree_nonzero_bits): New function.
10928 * fold-const.h (tree_nonzero_bits): Likewise.
10929 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
10930 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
10932 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
10936 * varasm.c (assemble_alias): Check ifunc_resolver only on
10939 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
10942 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
10943 when memory input operand is handled.
10945 2018-05-24 Luis Machado <luis.machado@linaro.org>
10947 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
10949 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
10951 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
10953 * match.pd: Delay FMA folds until after vectorization.
10955 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
10958 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
10959 address check not strict.
10961 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
10963 * gimple-match.h (gimple_match_op): New class.
10964 (mprts_hook): Replace parameters with a gimple_match_op *.
10965 (maybe_build_generic_op): Likewise.
10966 (gimple_simplified_result_is_gimple_val): Replace parameters with
10967 a const gimple_match_op *.
10968 (gimple_simplify): Replace code_helper * and tree * parameters with
10969 a gimple_match_op * parameter.
10970 (gimple_resimplify1): Replace code_helper *, tree and tree *
10971 parameters with a gimple_match_op * parameter.
10972 (gimple_resimplify2): Likewise.
10973 (gimple_resimplify3): Likewise.
10974 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
10975 parameters with a gimple_match_op * parameter.
10976 * gimple-match-head.c (gimple_simplify): Change prototypes of
10977 auto-generated functions to take a gimple_match_op * instead of
10978 separate code_helper * and tree * parameters. Make the same
10979 change in the top-level overload and update calls to the
10980 gimple_resimplify routines. Update calls to the auto-generated
10981 functions and to maybe_push_res_to_seq in the publicly-facing
10982 operation-specific gimple_simplify overloads.
10983 (gimple_match_op::MAX_NUM_OPS): Define.
10984 (gimple_resimplify1): Replace rcode and ops with a single res_op
10985 parameter. Update call to gimple_simplify.
10986 (gimple_resimplify2): Likewise.
10987 (gimple_resimplify3): Likewise.
10988 (mprts_hook): Replace parameters with a gimple_match_op *.
10989 (maybe_build_generic_op): Likewise.
10990 (build_call_internal): Replace type, nargs and ops with
10991 a gimple_match_op *.
10992 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
10993 with a single gimple_match_op *. Update calls to mprts_hook,
10994 build_call_internal and gimple_simplified_result_is_gimple_val.
10995 Factor out code that is common to the tree_code and combined_fn cases.
10996 * genmatch.c (expr::gen_transform): Replace tem_code and
10997 tem_ops with a gimple_match_op called tem_op. Update calls
10998 to the gimple_resimplify functions and maybe_push_res_to_seq.
10999 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
11000 res_ops. Update call to the gimple_resimplify functions.
11001 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
11002 (decision_tree::gen): Make the functions take a gimple_match_op *
11003 called res_op instead of separate res_code and res_ops parameters.
11004 Update call accordingly.
11005 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
11006 and ops with a single res_op parameter. Update calls to
11007 maybe_build_generic_op and maybe_push_res_to_seq.
11008 (fold_stmt_1): Update calls to gimple_simplify and
11009 replace_stmt_with_simplification.
11010 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
11011 and gimple_simplified_result_is_gimple_val.
11012 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
11014 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
11015 with a gimple_match_op *.
11016 (vn_nary_build_or_lookup): Likewise. Update call to
11017 vn_nary_build_or_lookup_1.
11018 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
11019 gimple_match_op *. Update calls to the gimple_resimplify routines
11020 and to gimple_simplified_result_is_gimple_val.
11021 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
11022 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
11023 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
11024 (visit_nary_op): Likewise.
11025 (visit_reference_op_load): Likewise.
11027 2018-05-23 Luis Machado <luis.machado@linaro.org>
11029 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
11030 modifier for printing the step amount.
11032 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
11035 * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
11038 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
11040 * doc/sourcebuild.texi (Endianness): New subsubsection.
11042 2018-05-23 Luis Machado <luis.machado@linaro.org>
11044 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
11045 <prefetch_dynamic_strides>: New const bool field.
11046 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
11047 prefetch_dynamic_strides.
11048 (exynosm1_prefetch_tune): Likewise.
11049 (thunderxt88_prefetch_tune): Likewise.
11050 (thunderx_prefetch_tune): Likewise.
11051 (thunderx2t99_prefetch_tune): Likewise.
11052 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
11054 (aarch64_override_options_internal): Update to set
11055 PARAM_PREFETCH_DYNAMIC_STRIDES.
11056 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
11057 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
11058 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
11059 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
11060 prefetch-dynamic-strides setting.
11062 2018-05-23 Luis Machado <luis.machado@linaro.org>
11064 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
11065 <minimum_stride>: New const int field.
11066 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
11067 minimum_stride field defaulting to -1.
11068 (exynosm1_prefetch_tune): Likewise.
11069 (thunderxt88_prefetch_tune): Likewise.
11070 (thunderx_prefetch_tune): Likewise.
11071 (thunderx2t99_prefetch_tune): Likewise.
11072 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
11073 <default_opt_level>: Set to 3.
11074 (aarch64_override_options_internal): Update to set
11075 PARAM_PREFETCH_MINIMUM_STRIDE.
11076 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
11077 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
11078 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
11079 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
11080 stride is constant and is below the minimum stride threshold.
11082 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11084 * config/arm/arm-cpus.in (mode26): Delete.
11085 (armv4): Delete mode26 reference.
11086 * config/arm/arm.c (arm_configure_build_target): Delete use of
11089 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
11091 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
11093 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
11094 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
11096 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
11097 (floatunsdidf2): Ditto.
11099 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
11101 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
11102 (fixuns_trunc<mode>si2_avx512f): Ditto.
11103 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
11104 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
11105 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
11107 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
11109 PR rtl-optimization/79985
11110 * df-scan.c (df_insn_refs_collect): Remove special case for
11111 global registers and asm statements.
11113 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
11115 * extend.texi (Global Register Variables): Rewrite the bullet list.
11116 Note that the register is available for allocation. Note that access
11117 via inline asm must use constraints. Add note about async-signal
11118 handlers. Remove paragraph about automagic register selection.
11120 2018-05-23 Richard Biener <rguenther@suse.de>
11122 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
11123 of fixed offset from memset VN.
11125 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
11127 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
11128 first_interp field.
11129 (alloc_cand_and_find_basis): Initialize first_interp field.
11130 (slsr_process_mul): Modify first_interp field.
11131 (slsr_process_add): Likewise.
11132 (slsr_process_cast): Modify first_interp field for each new
11134 (slsr_process_copy): Likewise.
11135 (dump_candidate): Dump first_interp field.
11136 (replace_mult_candidate): Process all interpretations, not just
11138 (replace_rhs_if_not_dup): Likewise.
11139 (replace_one_candidate): Likewise.
11141 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
11143 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
11145 (aarch64_needs_frame_chain): New function.
11146 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
11148 2018-05-23 Sudakshina Das <sudi.das@arm.com>
11151 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
11152 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
11153 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
11154 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
11155 as true for strict-align.
11156 (aarch64_can_inline_p): Perform checks even when callee has no
11157 attributes to check for strict alignment.
11158 * doc/extend.texi (AArch64 Function Attributes): Document
11160 * doc/invoke.texi: (AArch64 Options): Likewise.
11162 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
11164 PR tree-optimization/85853
11165 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
11166 the handling of the root of the node to...
11167 (vect_slp_analyze_node_operations_1): ...this new function,
11168 and run the whole thing with the child nodes' def types
11169 set according to their SLP node's def type.
11171 2018-05-23 Richard Biener <rguenther@suse.de>
11173 PR middle-end/85874
11174 * tree-data-ref.c (create_runtime_alias_checks): Defer
11175 and ignore overflow warnings.
11177 2018-05-23 Yury Gribov <tetra2005@gmail.com>
11179 PR tree-optimization/85822
11180 * tree-vrp.c (is_masked_range_test): Fix handling of negative
11183 2018-05-23 Richard Biener <rguenther@suse.de>
11185 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
11186 memset constants via native_interpret_expr.
11188 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
11191 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
11193 (cgraph_node::create_alias): Likewise.
11194 (cgraph_node::get_availability): Check ifunc_resolver instead
11195 of looking up ifunc attribute.
11196 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
11197 * varasm.c (do_assemble_alias): Likewise.
11198 (assemble_alias): Likewise.
11199 (default_binds_local_p_3): Likewise.
11200 * cgraph.h (cgraph_node): Add ifunc_resolver.
11201 (cgraph_node::only_called_directly_or_aliased_p): Return false
11202 for IFUNC resolver.
11203 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
11205 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
11206 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
11207 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
11208 instead of looking up ifunc attribute.
11210 2018-05-22 Luis Machado <luis.machado@linaro.org>
11212 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
11214 2018-05-22 Martin Sebor <msebor@redhat.com>
11216 PR middle-end/85359
11217 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
11218 only when expasion succeeds.
11219 (expand_builtin_strcmp): Same.
11220 (expand_builtin_strncmp): Same.
11222 2018-05-22 Martin Sebor <msebor@redhat.com>
11224 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
11226 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
11227 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11229 * config/aarch64/aarch64-ldpstp.md: Replace uses of
11230 aarch64_mem_pair_operand with memory_operand and delete operand swapping
11232 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
11233 Add check for legitimate_address.
11234 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
11235 (aarch64_swap_ldrstr_operands): New.
11236 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
11239 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
11240 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11242 * config/aarch64/aarch64.md: New patterns to generate stp
11244 (store_pair_sw, store_pair_dw): New patterns to generate stp for
11245 single words and double words.
11246 (load_pair_sw, load_pair_dw): Likewise.
11247 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
11249 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
11251 * config/aarch64/aarch64-ldpstp.md: Modify peephole
11252 for different mode ldpstp and add peephole for merged zero stores.
11253 Likewise for loads.
11254 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
11256 (aarch64_gen_store_pair): Rename calls to match new patterns.
11257 (aarch64_gen_load_pair): Rename calls to match new patterns.
11258 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
11259 (load_pair<DREG:mode><DREG2:mode>): ... This.
11260 (store_pair<mode>): Rename to...
11261 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
11262 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
11263 New mode iterators.
11264 (V_INT_EQUIV): Handle SImode.
11265 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
11268 2018-05-22 Martin Sebor <msebor@redhat.com>
11271 * calls.c (maybe_warn_nonstring_arg): Use string length to set
11272 or ajust the presumed bound on an operation to avoid unnecessary
11275 2018-05-22 Martin Sebor <msebor@redhat.com>
11277 PR tree-optimization/85826
11278 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
11279 assuming that a DECL necesarily has a constant size.
11281 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
11283 PR middle-end/85862
11284 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
11286 2018-05-22 Richard Biener <rguenther@suse.de>
11288 PR tree-optimization/85834
11289 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
11290 non-constant and non-zero memset arguments.
11292 2018-05-22 Martin Liska <mliska@suse.cz>
11295 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
11297 2018-05-22 Richard Biener <rguenther@suse.de>
11299 PR tree-optimization/85863
11300 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
11301 comparisons when vectype is specified.
11302 (vectorizable_condition): Do not specify vectype for
11303 vect_is_simple_cond when SLP vectorizing.
11305 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
11308 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
11309 define __ibm128 as long double.
11310 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
11311 as a distinct type when IEEE 128-bit support is enabled.
11312 (init_float128_ieee): Fix up conversions between IFmode and IEEE
11313 128-bit types to use the correct functions.
11314 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
11315 convert between 128-bit floating point types that have different
11316 modes but the same representation, instead of using gen_lowpart to
11318 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
11320 (IFKF_reg): New attributes to give the register constraints for
11322 (extend<mode>tf2_internal): New insns to mark an explicit
11323 conversion between 128-bit floating point types that have a
11324 different mode but share the same representation.
11326 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
11328 PR tree-optimization/85814
11329 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
11330 a null return from get_strinfo when unsharing the next
11331 strinfo in the chain.
11333 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
11336 * varasm.c (weak_finish): Clean up weak_decls.
11338 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11340 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
11341 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
11342 UNSPEC_UADALP values.
11343 * config/aarch64/iterators.md (ABAL): New int iterator.
11346 (sur): Add mappings for the above.
11347 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
11349 (aarch64_<sur>abal<mode>_4): Likewise.
11350 (aarch64_<sur>adalp<mode>_3): Likewise.
11351 (<sur>sadv16qi): New define_expand.
11353 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
11355 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
11356 (*movdf_internal): Ditto.
11357 (*rcpsf2_sse): Ditto.
11358 (*rsqrtsf2_sse): Ditto.
11359 (*sqrt<mode>2_sse): Ditto.
11361 2018-05-21 Tamar Christina <tamar.christina@arm.com>
11363 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
11365 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
11366 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
11367 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
11368 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
11370 * config/aarch64/arm_neon.h: Likewise.
11371 * config/aarch64/iterators.md (VQ_I): New.
11373 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
11375 * config.gcc: Add arc/t-multilib-linux to tmake_file for
11377 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
11380 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
11382 * config/nds32/constraints.md (S): New constraint.
11383 * config/nds32/nds32.md (call_internal): Use constraint S.
11384 (call_value_internal): Likewise.
11385 (sibcall_internal): Likewise.
11386 (sibcall_value_internal): Likewise.
11388 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
11389 Chung-Ju Wu <jasonwucj@gmail.com>
11391 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
11392 into consideration.
11394 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
11395 Chung-Ju Wu <jasonwucj@gmail.com>
11397 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
11398 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
11399 (nds32_rtx_costs_impl): Simplify.
11400 (nds32_address_cost_impl): Simplify.
11401 (nds32_init_rtx_costs): New function.
11402 (nds32_rtx_costs_speed_prefer): Likewise.
11403 (nds32_rtx_costs_size_prefer): Likewise.
11404 (nds32_address_cost_speed_prefer): Likewise.
11405 (nds32_address_cost_speed_fwprop): Likewise.
11406 (nds32_address_cost_size_prefer): Likewise.
11407 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
11408 * config/nds32/nds32.c (nds32_option_override): Use
11409 nds32_init_rtx_costs function.
11411 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
11413 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
11414 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
11415 (TARGET_PIPELINE_N8): Likewise.
11416 (TARGET_PIPELINE_N10): Likewise.
11417 (TARGET_PIPELINE_N13): Likewise.
11418 (TARGET_PIPELINE_GRAYWOLF): Likewise.
11420 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
11422 * config/nds32/nds32-fpu.md: Update copyright year.
11424 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
11426 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
11428 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
11430 * config/nds32/nds32.c
11431 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
11432 * config/nds32/nds32.opt (minline-asm-r15): New option.
11434 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
11436 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
11438 * config/nds32/nds32.md (abssi2): New pattern.
11440 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
11442 * config/i386/i386.md (rex64namesuffix): New mode attribute.
11443 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
11444 Merge insn pattern from sse_cvtsi2ss<round_name> and
11445 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
11446 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
11447 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
11448 using SWI48 mode iterator.
11449 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
11450 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
11451 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
11452 pattern from sse_cvttss2si<round_saeonly_name>
11453 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
11454 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
11455 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
11456 using SWI48 mode iterator.
11457 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
11458 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
11459 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
11460 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
11461 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
11462 using SWI48 mode iterator.
11463 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
11464 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
11465 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
11466 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
11467 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
11468 SWI48 mode iterator.
11469 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
11470 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
11471 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
11472 pattern from sse_cvttsd2si<round_saeonly_name>
11473 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
11475 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
11477 * config/nds32/nds32-md-auxiliary.c
11478 (nds32_valid_smw_lwm_base_p): Refine.
11479 (nds32_output_smw_single_word): Refine.
11480 (nds32_output_smw_double_word): New.
11481 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
11483 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
11485 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
11486 (nds32_output_stack_pop): Refine.
11487 (nds32_expand_unaligned_load): Refine.
11488 (nds32_expand_unaligned_store): Refine.
11490 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
11491 Chung-Ju Wu <jasonwucj@gmail.com>
11493 * config/nds32/constants.md: Add TP_REGNUM constant.
11494 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
11495 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
11497 * config/nds32/nds32-doubleword.md: Consider flag_pic.
11498 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
11499 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
11500 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
11501 and PIC code generation.
11502 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
11504 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
11506 * config/nds32/nds32.md: Support TLS and PIC.
11507 * config/nds32/nds32.c: Support TLS and PIC.
11508 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
11509 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
11512 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
11514 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
11515 mode with E_ prefix.
11517 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
11518 Chung-Ju Wu <jasonwucj@gmail.com>
11520 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
11521 * config/nds32/nds32-md-auxiliary.c
11522 (symbolic_reference_mentioned_p): New.
11523 (nds32_legitimize_ict_address): New.
11524 (nds32_expand_ict_move): New.
11525 (nds32_indirect_call_referenced_p): New.
11526 (nds32_symbol_binds_local_p): Delete.
11527 (nds32_long_call_p): Modify.
11528 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
11529 * config/nds32/nds32-protos.h
11530 (symbolic_reference_mentioned_p): Declare.
11531 (nds32_legitimize_ict_address): Declare.
11532 (nds32_expand_ict_move): Declare.
11533 (nds32_indirect_call_referenced_p): Declare.
11534 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
11535 (nds32_relax_group): Use nds32_ict_const_p as condition.
11536 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
11537 (nds32_asm_file_start): Output ict_model directive in asm code.
11538 (nds32_legitimate_address_p): Consider indirect call.
11539 (nds32_print_operand): Consider indirect call.
11540 (nds32_print_operand_address): Consider indirect call.
11541 (nds32_insert_attributes): Handle "indirect_call" attribute.
11542 (TARGET_LEGITIMATE_ADDRESS_P): Define.
11543 (TARGET_LEGITIMATE_CONSTANT_P): Define.
11544 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
11545 (TARGET_DELEGITIMIZE_ADDRESS): Define.
11546 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
11547 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
11548 (TARGET_ICT_MODEL_SMALL): Define.
11549 (TARGET_ICT_MODEL_LARGE): Define.
11550 * config/nds32/nds32.md (movsi): Consider ict model.
11551 (call, call_value): Consider ict model.
11552 (sibcall, sibcall_value): Consider ict model.
11553 * config/nds32/nds32.opt (mict-model): New option.
11554 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
11557 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
11558 Monk Chiang <sh.chiang04@gmail.com>
11559 Jim Wilson <jimw@sifive.com>
11561 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
11562 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
11563 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
11564 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
11565 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
11566 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
11567 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
11568 compute save_libcall_adjustment properly.
11569 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
11570 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
11571 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
11572 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
11573 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
11574 (ABI_SPEC): Handle mabi=ilp32e.
11575 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
11576 (RVE): Add RVE mask.
11577 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
11578 <-march>: Add rv32e as an example.
11580 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
11583 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
11584 (intra_create_variable_infos): Handle C++ constructors.
11586 2018-05-18 Martin Liska <mliska@suse.cz>
11588 * passes.def: Remove a redundant pass.
11590 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
11593 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
11595 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11597 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
11599 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
11601 (armv2, armv2a, armv3, armv3m): Delete architectures.
11602 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
11603 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
11604 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
11606 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
11607 (*mulsidi3adddi): Likewise.
11608 (mulsidi3): Likewise.
11609 (*mulsidi3_nov6): Likewise.
11610 (umulsidi3): Likewise.
11611 (umulsidi3_nov6): Likewise.
11612 (umaddsidi4): Likewise.
11613 (*umulsidi3adddi): Likewise.
11614 (smulsi3_highpart): Likewise.
11615 (*smulsi3_highpart_nov6): Likewise.
11616 (umulsi3_highpart): Likewise.
11617 (*umulsi3_highpart_nov6): Likewise.
11618 * config/arm/arm.h (arm_arch3m): Delete.
11619 * config/arm/arm.c (arm_arch3m): Delete.
11620 (arm_option_override_internal): Update armv3-related comment.
11621 (arm_configure_build_target): Delete use of isa_bit_mode32.
11622 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
11623 (arm_rtx_costs_internal): Delete check of arm_arch3m.
11624 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
11625 (mulsa3): Likewise.
11626 (mulusa3): Likewise.
11627 * config/arm/arm-protos.h (arm_arch3m): Delete.
11628 * config/arm/arm-tables.opt: Regenerate.
11629 * config/arm/arm-tune.md: Likewise.
11630 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
11631 deleted architectures.
11633 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11635 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
11636 (armv5t, armv5te): New features.
11637 (ARMv5, ARMv5e): Delete fgroups.
11638 (ARMv5t, ARMv5te): Adjust for above changes.
11639 (ARMv6m): Likewise.
11640 (armv5, armv5e): Delete arches.
11641 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
11643 (*call_reg_arm): Likewise.
11644 (*call_value_reg_armv5): Likewise.
11645 (*call_value_reg_arm): Likewise.
11646 (*call_symbol): Likewise.
11647 (*call_value_symbol): Likewise.
11648 (*sibcall_insn): Likewise.
11649 (*sibcall_value_insn): Likewise.
11650 (clzsi2): Likewise.
11651 (prefetch): Likewise.
11652 (define_split and define_peephole2 dependent on arm_arch5):
11654 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
11656 (TARGET_ARM_QBIT): Likewise.
11657 (TARGET_DSP_MULTIPLY): Likewise.
11658 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
11659 (arm_arch5, arm_arch5e): Delete.
11660 (arm_arch5t, arm_arch5te): Declare.
11661 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
11662 (arm_arch5t): Declare.
11663 (arm_option_reconfigure_globals): Update for the above.
11664 (arm_options_perform_arch_sanity_checks): Update comment, replace
11665 use of arm_arch5 with arm_arch5t.
11666 (use_return_insn): Likewise.
11667 (arm_emit_call_insn): Likewise.
11668 (output_return_instruction): Likewise.
11669 (arm_final_prescan_insn): Likewise.
11670 (arm_coproc_builtin_available): Likewise.
11671 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
11672 arm_arch5e with arm_arch5t and arm_arch5te.
11673 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
11674 (arm_arch5t, arm_arch5te): Declare.
11675 * config/arm/arm-tables.opt: Regenerate.
11676 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
11677 * config/arm/t-multilib: Likewise.
11678 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
11679 instead of arm_arch5.
11680 (*call_reg_thumb1): Likewise.
11681 (*call_value_reg_thumb1_v5): Likewise.
11682 (*call_value_reg_thumb1): Likewise.
11683 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
11685 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
11687 2018-05-18 Martin Liska <mliska@suse.cz>
11689 PR gcov-profile/84846
11690 * doc/gcov.texi: Document -t option of gcov tool.
11692 2018-05-18 Martin Liska <mliska@suse.cz>
11694 PR gcov-profile/84846
11695 * gcov.c (print_usage): Add new -t option.
11696 (process_args): Handle the option.
11697 (generate_results): Use stdout as output when requested by
11700 2018-05-18 Martin Liska <mliska@suse.cz>
11702 PR gcov-profile/84846
11703 * coverage.c (coverage_init): Write PWD to .gcno file.
11704 * doc/gcov.texi: Document how working directory is printed.
11705 * gcov-dump.c (dump_gcov_file): Print PWD.
11706 * gcov.c (output_intermediate_file): Likewise.
11707 (read_graph_file): Read PWD string.
11708 (output_lines): Print PWD.
11710 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11712 PR middle-end/85817
11713 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
11714 for retval and return false if all args to phi are zero.
11716 2018-05-18 Richard Biener <rguenther@suse.de>
11718 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
11720 (evrp_dom_walker::before_dom_children): Call it.
11722 2018-05-18 Richard Biener <rguenther@suse.de>
11724 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
11725 results when processing array refs with variable index.
11727 2018-05-18 Toon Moene <toon@moene.org>
11729 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
11730 directly after that of -floop-interchange. Indicate that both
11731 options are enabled by default when specifying -O3.
11733 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11735 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
11736 iterator. Delete separate integer-mode vec_set<mode> expander.
11737 (aarch64_simd_vec_setv2di): Delete.
11738 (vec_setv2di): Delete.
11739 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
11740 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
11741 the "w, r" alternative.
11743 2018-05-18 Martin Liska <mliska@suse.cz>
11745 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
11746 * tree-pass.h (make_pass_lower_switch_O0): New function.
11747 * tree-switch-conversion.c (node_has_low_bound): Remove.
11748 (node_has_high_bound): Likewise.
11749 (node_is_bounded): Likewise.
11750 (class pass_lower_switch): Make it a template type and create
11752 (pass_lower_switch::execute): Add template argument.
11753 (make_pass_lower_switch): New function.
11754 (make_pass_lower_switch_O0): New function.
11755 (do_jump_if_equal): Remove.
11756 (emit_case_nodes): Simplify to just handle all 3 cases and leave
11757 all the hard work to tree optimization passes.
11759 2018-05-18 Martin Liska <mliska@suse.cz>
11761 * dbgcnt.c (limit_low): Renamed from limit.
11762 (limit_high): New variable.
11763 (dbg_cnt_is_enabled): Check for upper limit.
11764 (dbg_cnt): Adjust dumping.
11765 (dbg_cnt_set_limit_by_index): Add new argument for high
11767 (dbg_cnt_set_limit_by_name): Likewise.
11768 (dbg_cnt_process_single_pair): Parse new format.
11769 (dbg_cnt_process_opt): Use strtok.
11770 (dbg_cnt_list_all_counters): Remove 'value' and add
11772 * doc/invoke.texi: Document changes.
11774 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
11776 * doc/sourcebuild.texi (scalar_all_fma): Document.
11777 * tree.def (FMA_EXPR): Delete.
11778 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
11779 * internal-fn.c (ternary_direct): New macro.
11780 (expand_ternary_optab_fn): Likewise.
11781 (direct_ternary_optab_supported_p): Likewise.
11782 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
11783 * builtins.c (fold_builtin_fma): Delete.
11784 (fold_builtin_3): Don't call it.
11785 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
11786 * expr.c (expand_expr_real_2): Likewise.
11787 * fold-const.c (operand_equal_p): Likewise.
11788 (fold_ternary_loc): Likewise.
11789 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
11790 * gimple.c (DEFTREECODE): Likewise.
11791 * gimplify.c (gimplify_expr): Likewise.
11792 * optabs-tree.c (optab_for_tree_code): Likewise.
11793 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
11794 * tree-eh.c (operation_could_trap_p): Likewise.
11795 (stmt_could_throw_1_p): Likewise.
11796 * tree-inline.c (estimate_operator_cost): Likewise.
11797 * tree-pretty-print.c (dump_generic_node): Likewise.
11798 (op_code_prio): Likewise.
11799 * tree-ssa-loop-im.c (stmt_cost): Likewise.
11800 * tree-ssa-operands.c (get_expr_operands): Likewise.
11801 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
11802 * fold-const-call.h (fold_fma): Delete.
11803 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
11804 CFN_FNMA and CFN_FNMS.
11805 (fold_fma): Delete.
11806 * genmatch.c (combined_fn): New enum.
11807 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
11808 (commutative_op): New function.
11809 (commutate): Use it. Handle more than 2 operands.
11810 (dt_operand::gen_gimple_expr): Use commutative_op.
11811 (parser::parse_expr): Allow :c to be used with non-binary
11812 operators if the commutative operand is known.
11813 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
11814 CFN_FMS, CFN_FNMA and CFN_FNMS.
11815 (backprop::process_assign_use): Remove FMA_EXPR handling.
11816 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
11817 (gen_hsa_fma): New function.
11818 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
11819 IFN_FNMA and IFN_FNMS.
11820 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
11821 * gimple-fold.h (follow_all_ssa_edges): Declare.
11822 * gimple-fold.c (follow_all_ssa_edges): New function.
11823 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
11824 gimple_build interface and use follow_all_ssa_edges to fold the result.
11825 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
11826 instead of checking for optabs directly.
11827 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
11828 rather than FMA_EXPRs.
11829 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
11830 call to IFN_FMA instead of an FMA_EXPR.
11832 2018-05-17 Jim Wilson <jimw@sifive.com>
11834 * expr.c (do_tablejump): When converting index to Pmode, if we have a
11835 sign extended promoted subreg, and the range does not have the sign bit
11836 set, then do a sign extend.
11838 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
11839 test, check for sign extended subreg and/or constant operands, and
11840 do a sign extend in that case.
11842 2018-05-17 Steve Ellcey <sellcey@cavium.com>
11844 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
11845 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
11847 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
11848 Change logics_shift_reg to logics_shift_imm.
11849 (thunderx2t99_fp_loadpair_basic): Delete.
11850 (thunderx2t99_fp_storepair_basic): Delete.
11851 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
11852 (thunderx2t99_asimd_polynomial): Delete.
11853 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
11854 and neon_fp_mul_d_scalar_q.
11855 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
11856 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
11857 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
11858 (thunderx2t99_asimd_lut): Add missing tbl types.
11859 (thunderx2t99_asimd_ext): Delete.
11860 (thunderx2t99_asimd_load1_1_mult): Delete.
11861 (thunderx2t99_asimd_load1_2_mult): Delete.
11862 (thunderx2t99_asimd_load1_ldp): New.
11863 (thunderx2t99_asimd_load1): New.
11864 (thunderx2t99_asimd_load2): Add missing *load2* types.
11865 (thunderx2t99_asimd_load3): New.
11866 (thunderx2t99_asimd_load4): New.
11867 (thunderx2t99_asimd_store1_1_mult): Delete.
11868 (thunderx2t99_asimd_store1_2_mult): Delete.
11869 (thunderx2t99_asimd_store2_mult): Delete.
11870 (thunderx2t99_asimd_store2_onelane): Delete.
11871 (thunderx2t99_asimd_store_stp): New.
11872 (thunderx2t99_asimd_store1): New.
11873 (thunderx2t99_asimd_store2): New.
11874 (thunderx2t99_asimd_store3): New.
11875 (thunderx2t99_asimd_store4): New.
11877 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
11879 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
11880 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
11882 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
11883 Segher Boessenkool <segher@kernel.crashing.org>
11886 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
11889 2018-05-17 Richard Biener <rguenther@suse.de>
11891 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
11892 for pruning loop and prune defs feeding only already visited PHIs.
11894 2018-05-17 Richard Biener <rguenther@suse.de>
11896 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
11898 2018-05-17 Bin Cheng <bin.cheng@arm.com>
11899 Richard Biener <rguenther@suse.de>
11901 PR tree-optimization/85793
11902 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
11903 for VMAT_ELEMENTWISE.
11905 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
11907 * internal-fn.h (lookup_internal_fn): Declare
11908 * internal-fn.c (lookup_internal_fn): New function.
11909 * gimple.c (gimple_build_call_from_tree): Handle calls to
11910 internal functions.
11911 * gimple-pretty-print.c (dump_gimple_call): Print "." before
11912 internal function names.
11913 * tree-pretty-print.c (dump_generic_node): Likewise.
11914 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
11916 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
11918 * gimple-fold.h (gimple_build): Make the function forms take
11919 combined_fn rather than built_in_function.
11920 (gimple_simplify): Likewise.
11921 * gimple-match-head.c (gimple_simplify): Likewise.
11922 * gimple-fold.c (gimple_build): Likewise.
11923 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
11924 rather than gimple_build_call_internal.
11925 (get_initial_defs_for_reduction): Likewise.
11926 (vect_create_epilog_for_reduction): Likewise.
11927 (vectorizable_live_operation): Likewise.
11929 2018-05-17 Martin Liska <mliska@suse.cz>
11931 * gimple-ssa-sprintf.c (format_directive): Do not use
11932 space in between 'G_' and '('.
11934 2018-05-17 Jakub Jelinek <jakub@redhat.com>
11937 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
11938 even if the mask is not all ones.
11941 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
11943 (ix86_gimple_fold_builtin): Likewise.
11946 * config/i386/i386.c: Include tree-vector-builder.h.
11947 (ix86_vector_shift_count): New function.
11948 (ix86_fold_builtin): Fold shift builtins by scalar count.
11949 (ix86_gimple_fold_builtin): Likewise.
11951 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
11952 _mm512_setzero): New intrinsics.
11954 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
11955 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11957 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
11958 code generation for cases where splatting a value is not useful.
11959 * simplify-rtx.c (simplify_ternary_operation): Simplify
11960 vec_merge across a vec_duplicate and a paradoxical subreg forming
11961 a vector mode to a vec_concat.
11963 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
11965 * config.gcc: Support "goldmont-plus".
11966 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
11968 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
11969 PROCESSOR_GOLDMONT_PLUS.
11970 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
11971 (processor_target_table): Add "goldmont-plus".
11972 (PTA_GOLDMONT_PLUS): Define.
11973 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
11974 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
11975 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
11976 (fold_builtin_cpu): Add "goldmont-plus".
11977 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
11978 (ix86_option_override_internal): Add "goldmont-plus".
11979 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
11980 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
11981 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
11982 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
11984 2018-05-17 Richard Biener <rguenther@suse.de>
11986 PR tree-optimization/85757
11987 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
11988 remove defs that only feed that PHI from further processing.
11990 2018-05-16 Jim Wilson <jimw@sifive.com>
11992 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
11994 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
11996 2018-05-16 Mark Wielaard <mark@klomp.org>
11998 * dwarf2out.c (count_index_strings): New function.
11999 (output_indirect_strings): Call count_index_strings and generate
12000 header for dwarf_version >= 5.
12002 2018-05-16 Mark Wielaard <mark@klomp.org>
12004 * dwarf2out.c (dwarf_FORM): New function.
12005 (set_indirect_string): Use dwarf_FORM.
12006 (reset_indirect_string): Likewise.
12007 (size_of_die): Likewise.
12008 (value_format): Likewise.
12009 (output_die): Likewise.
12010 (add_skeleton_AT_string): Likewise.
12011 (output_macinfo_op): Likewise.
12012 (index_string): Likewise.
12013 (output_index_string_offset): Likewise.
12014 (output_index_string): Likewise.
12015 (count_index_strings): Likewise.
12017 2018-05-16 Carl Love <cel@us.ibm.com>
12019 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
12020 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
12022 2018-05-16 Martin Jambor <mjambor@suse.cz>
12024 * ipa-prop.c (ipa_free_all_edge_args): Remove.
12025 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
12027 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
12029 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
12030 (fnma<mode>4): Likewise.
12031 (fms<mode>4): Likewise.
12032 (fnms<mode>4): Likewise.
12033 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
12034 (aarch64_fnma<mode>4): Likewise.
12035 (aarch64_fms<mode>4): Likewise.
12036 (aarch64_fnms<mode>4): Likewise.
12037 (aarch64_fnmadd<mode>4): Likewise.
12039 2018-05-16 Jason Merrill <jason@redhat.com>
12041 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
12043 2018-05-16 Richard Biener <rguenther@suse.de>
12045 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
12046 (dump_stmt_cost): Declare.
12047 (add_stmt_cost): Dump cost we add.
12048 (add_stmt_costs): New function.
12049 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
12050 No longer exported.
12051 (vect_analyze_stmt): Adjust prototype.
12052 (vectorizable_condition): Likewise.
12053 (vectorizable_live_operation): Likewise.
12054 (vectorizable_reduction): Likewise.
12055 (vectorizable_induction): Likewise.
12056 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
12057 cost vector to pass to vectorizable_ and record afterwards.
12058 (vect_model_reduction_cost): Take cost vector argument and adjust.
12059 (vect_model_induction_cost): Likewise.
12060 (vectorizable_reduction): Likewise.
12061 (vectorizable_induction): Likewise.
12062 (vectorizable_live_operation): Likewise.
12063 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
12064 SLP_TREE_NUMBER_OF_VEC_STMTS.
12065 (vect_analyze_slp_cost_1): Remove.
12066 (vect_analyze_slp_cost): Likewise.
12067 (vect_slp_analyze_node_operations): Take visited args and
12068 a target cost vector. Avoid processing already visited stmt sets.
12069 (vect_slp_analyze_operations): Use a local cost vector to gather
12070 costs and register those of non-discarded instances.
12071 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
12072 (vect_schedule_slp_instance): Remove copying of
12073 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
12075 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
12076 adding cost. Record cost entry location.
12077 (vect_prologue_cost_for_slp_op): Function to compute cost of
12078 a constant or invariant generated for SLP vect in the prologue,
12079 split out from vect_analyze_slp_cost_1.
12080 (vect_model_simple_cost): Make static. Adjust for SLP costing.
12081 (vect_model_promotion_demotion_cost): Likewise.
12082 (vect_model_store_cost): Likewise, make static.
12083 (vect_model_load_cost): Likewise.
12084 (vectorizable_bswap): Add cost vector arg and adjust.
12085 (vectorizable_call): Likewise.
12086 (vectorizable_simd_clone_call): Likewise.
12087 (vectorizable_conversion): Likewise.
12088 (vectorizable_assignment): Likewise.
12089 (vectorizable_shift): Likewise.
12090 (vectorizable_operation): Likewise.
12091 (vectorizable_store): Likewise.
12092 (vectorizable_load): Likewise.
12093 (vectorizable_condition): Likewise.
12094 (vectorizable_comparison): Likewise.
12095 (can_vectorize_live_stmts): Likewise.
12096 (vect_analyze_stmt): Likewise.
12097 (vect_transform_stmt): Adjust calls to vectorizable_*.
12098 * tree-vectorizer.c: Include gimple-pretty-print.h.
12099 (dump_stmt_cost): New function.
12101 2018-05-16 Richard Biener <rguenther@suse.de>
12103 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
12104 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
12105 * tree-ssa-dse.c: Include tree-ssa-loop.h.
12106 (check_name): New callback.
12107 (dse_classify_store): Track cycles via a visited bitmap of PHI
12108 defs and simplify handling of in-loop and across loop dead stores
12109 and properly fail for loop-variant refs. Handle byte-tracking with
12110 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
12113 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
12115 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
12116 (vect_get_mask_type_for_stmt): Likewise.
12117 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
12119 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
12120 to determine the statement's vector type and the vector type that
12121 should be used for calculating nunits. Deal with cases in which
12122 the type has to be deferred.
12123 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
12124 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
12125 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
12126 (vect_determine_vf_for_stmt): New functions, split out from...
12127 (vect_determine_vectorization_factor): ...here.
12128 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
12129 (vect_get_mask_type_for_stmt): New functions, split out from
12130 vect_determine_vectorization_factor.
12132 2018-05-16 Richard Biener <rguenther@suse.de>
12134 * tree-cfg.c (verify_gimple_assign_ternary): Properly
12135 verify the [VEC_]COND_EXPR embedded comparison.
12137 2018-05-15 Martin Sebor <msebor@redhat.com>
12139 PR tree-optimization/85753
12140 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
12141 RECORD_TYPE in addition to ARRAY_TYPE.
12143 2018-05-15 Martin Sebor <msebor@redhat.com>
12145 PR middle-end/85643
12146 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
12148 2018-05-15 Richard Biener <rguenther@suse.de>
12150 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
12151 add by_clobber_p one. Change algorithm to collect all defs
12152 representing uses we need to walk and try reducing them to
12153 a single one before failing.
12154 (dse_dom_walker::dse_optimize_stmt): Adjust.
12156 2018-05-13 Mark Wielaard <mark@klomp.org>
12158 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
12159 (size_of_loc_descr): Likewise.
12160 (output_loc_operands): Likewise.
12161 (output_loc_operands_raw): Likewise.
12162 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
12163 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
12164 (hash_loc_operands): Likewise.
12165 (compare_loc_operands): Likewise.
12167 2018-05-14 Mark Wielaard <mark@klomp.org>
12169 * dwarf2out.c (count_index_addrs): New function.
12170 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
12172 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12174 PR tree-optimization/83648
12175 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
12176 return value as malloc candidate.
12178 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12181 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
12182 param as true in call to suggest_attribute.
12184 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
12186 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
12187 -mreadonly-in-sdata.
12189 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12191 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
12193 (aarch64_crypto_aesd_fused): Likewise.
12195 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
12197 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
12198 (movsi_aarch64): Likewise.
12199 (load_pairsi): Likewise.
12200 (load_pairdi): Likewise.
12201 (store_pairsi): Likewise.
12202 (store_pairdi): Likewise.
12203 (load_pairsf): Likewise.
12204 (load_pairdf): Likewise.
12205 (store_pairsf): Likewise.
12206 (store_pairdf): Likewise.
12207 (zero_extend): Likewise.
12208 (trunc): Swap alternatives.
12209 (fcvt_target): Add '?' to prefer w over r.
12211 2018-05-14 Jakub Jelinek <jakub@redhat.com>
12214 * config/i386/i386.md: Disallow non-commutative arithmetics in
12215 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
12216 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
12217 in the peephole2 before it.
12219 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
12221 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
12222 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
12223 (ix86_handle_option): Handle -mcldemote.
12224 * config.gcc: New header.
12225 * config/i386/cldemoteintrin.h: New file.
12226 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
12227 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
12229 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
12230 OPTION_MASK_ISA_CLDEMOTE.
12231 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
12232 (ix86_valid_target_attribute_inner_p): Ditto.
12233 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
12234 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
12235 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
12236 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
12237 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
12239 * config/i386/i386.opt: Add -mcldemote.
12240 * config/i386/x86intrin.h: New header.
12241 * doc/invoke.texi: Add -mcldemote.
12243 2018-05-14 Richard Biener <rguenther@suse.de>
12245 * doc/match-and-simplify.texi: Adjust :s documentation.
12247 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
12249 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
12250 intended memcpy size.
12251 (REORDER_45): Likewise.
12253 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
12255 * sort.cc: New file.
12256 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
12257 * vec.c (qsort_chk): Use gcc_qsort.
12258 * Makefile.in (OBJS-libcommon): Add sort.o.
12259 (build/sort.o): New target. Use it...
12260 (BUILD_RTL): ... here, and...
12261 (build/gencfn-macros): ... here, and...
12262 (build/genmatch): ... here.
12264 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
12265 Chung-Ju Wu <jasonwucj@gmail.com>
12267 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
12268 * config/nds32/nds32-graywolf.md: New file.
12269 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
12270 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
12272 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
12273 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
12274 * config/nds32/nds32.md (pipeline_model): Add graywolf.
12275 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
12276 * config/nds32/pipelines.md: Include n15 settings.
12278 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
12279 Chung-Ju Wu <jasonwucj@gmail.com>
12281 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
12282 * config/nds32/nds32-n13.md: New file.
12283 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
12284 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
12286 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
12287 * config/nds32/nds32.md (pipeline_model): Add n13.
12288 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
12289 * config/nds32/pipelines.md: Include n13 settings.
12291 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
12292 Chung-Ju Wu <jasonwucj@gmail.com>
12294 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
12295 * config/nds32/nds32-n10.md: New file.
12296 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
12297 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
12299 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
12300 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
12301 * config/nds32/nds32.md (pipeline_model): Add n10.
12302 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
12303 * config/nds32/pipelines.md: Include n10 settings.
12305 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
12306 Kito Cheng <kito.cheng@gmail.com>
12307 Chung-Ju Wu <jasonwucj@gmail.com>
12309 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
12310 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
12311 Add enum values for DSP extension instructions.
12312 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
12314 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
12315 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
12316 New code iterators.
12317 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
12318 * config/nds32/nds32-dspext.md: New file for DSP implementation.
12319 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
12320 * config/nds32/nds32-intrinsic.md: Likewise.
12321 * config/nds32/nds32_intrinsic.h: Likewise.
12322 * config/nds32/nds32-md-auxiliary.c: Likewise.
12323 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
12324 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
12325 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
12326 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
12327 * config/nds32/nds32-protos.h: New declarations for DSP extension.
12328 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
12329 TYPE_DMAC in switch statement.
12330 * config/nds32/nds32.c: New checking and implementation for DSP
12331 extension instructions.
12332 * config/nds32/nds32.h: Likewise.
12333 * config/nds32/nds32.md: Likewise.
12334 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
12335 * config/nds32/predicates.md: Implement new predicates for DSP
12338 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
12340 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
12341 Reformat alternatives and attributes so it is easier to identify
12342 which constraints/attributes go with which instruction.
12343 (mov<mode>_hardfloat32, FMOVE64): Likewise.
12344 (mov<mode>_softfloat32, FMOVE64): Likewise.
12345 (mov<mode>_hardfloat64, FMOVE64): Likewise.
12346 (mov<mode>_softfloat64, FMOVE64): Likewise.
12348 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
12350 * doc/extend.texi (PowerPC Built-in Functions): Rename this
12352 (Basic PowerPC Built-in Functions): The new name of the
12353 subsection previously known as "PowerPC Built-in Functions".
12354 (Basic PowerPC Built-in Functions Available on all Configurations):
12356 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
12357 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
12358 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
12359 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
12361 2018-05-11 Martin Jambor <mjambor@suse.cz>
12364 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
12367 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
12370 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
12372 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
12374 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
12375 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
12376 (ix86_handle_option): Handle -mwaitpkg.
12377 * config.gcc: New header.
12378 * config/i386/cpuid.h (bit_WAITPKG): New bit.
12379 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
12380 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
12382 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
12383 OPTION_MASK_ISA_WAITPKG.
12384 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
12385 (ix86_option_override_internal): Add PTA_WAITPKG.
12386 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
12387 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
12388 IX86_BUILTIN_TPAUSE.
12389 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
12390 __builtin_ia32_umwait and __builtin_ia32_tpause.
12391 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
12392 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
12393 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
12394 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
12395 UNSPECV_TPAUSE): New.
12396 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
12397 * config/i386/i386.opt: Add -mwaitpkg.
12398 * config/i386/waitpkgintrin.h: New file.
12399 * config/i386/x86intrin.h: New header.
12400 * doc/invoke.texi: Add -mwaitpkg.
12402 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
12405 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
12407 (cortex-m0): Use armv6s-m isa.
12408 (cortex-m0plus): Likewise.
12409 (cortex-m1): Likewise.
12410 (cortex-m0.small-multiply): Likewise.
12411 (cortex-m0plus.small-multiply): Likewise.
12412 (cortex-m1.small-multiply): Likewise.
12414 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
12415 Jakub Jelinek <jakub@redhat.com>
12417 PR tree-optimization/85692
12418 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
12419 source permute as well.
12421 2018-05-11 Martin Liska <mliska@suse.cz>
12424 * doc/extend.texi: Document LLVM style format for no_sanitize
12427 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
12429 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
12430 mode_supports_vsx_dform_quad to mode_supports_dq_form.
12431 (mode_supports_vsx_dform_quad): Likewise.
12432 (mode_supports_vmx_dform): Move these functions to be next to the
12433 other mode_supports functions.
12434 (mode_supports_dq_form): Likewise.
12435 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
12436 mode_supports_dq_form.
12437 (reg_offset_addressing_ok_p): Likewise.
12438 (offsettable_ok_by_alignment): Likewise.
12439 (rs6000_legitimate_offset_address_p): Likewise.
12440 (legitimate_lo_sum_address_p): Likewise.
12441 (rs6000_legitimize_address): Likewise.
12442 (rs6000_legitimize_reload_address): Likewise.
12443 (rs6000_secondary_reload_inner): Likewise.
12444 (rs6000_preferred_reload_class): Likewise.
12445 (rs6000_output_move_128bit): Likewise.
12447 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
12449 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
12450 Generate SImode target register for null target.
12451 <case IX86_BUILTIN_XGETBV>: Ditto.
12452 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
12453 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
12455 2018-05-10 Carl Love <cel@us.ibm.com>
12457 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
12458 dcbtt and dcbtstt if operands[2] is 0.
12460 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
12463 * config/i386/sse.md (usadv64qi): New expander.
12465 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
12467 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
12468 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
12469 -maltivec=be support.
12470 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
12471 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
12472 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
12473 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
12474 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
12475 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
12476 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
12477 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
12478 altivec_vsumsws): Adjust.
12479 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
12480 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
12481 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
12483 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
12484 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
12485 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
12486 (altivec_lve<VI_char>x): Delete expand.
12487 (*altivec_lve<VI_char>x_internal): Rename to...
12488 (altivec_lve<VI_char>x): ... this.
12489 (altivec_lvxl_<mode>): Delete expand.
12490 (*altivec_lvxl_<mode>_internal): Rename to ...
12491 (altivec_lvxl_<mode>): ... this.
12492 (altivec_stvxl_<mode>): Delete expand.
12493 (*altivec_stvxl_<mode>_internal): Rename to ...
12494 (altivec_stvxl_<mode>): ... this.
12495 (altivec_stve<VI_char>x): Delete expand.
12496 (*altivec_stve<VI_char>x_internal): Rename to ...
12497 (altivec_stve<VI_char>x): ... this.
12498 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
12499 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
12500 reduc_plus_scal_<mode>): Adjust.
12501 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
12503 (rs6000_cpu_cpp_builtins): Adjust.
12504 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
12505 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
12506 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
12507 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
12508 -maltivec=be support.
12509 (rs6000_split_vec_extract_var): Adjust.
12510 (rs6000_split_v4si_init): Adjust.
12511 (swap_selector_for_mode): Delete.
12512 (altivec_expand_lvx_be, altivec_expand_stvx_be,
12513 altivec_expand_stvex_be): Delete.
12514 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
12515 -maltivec=be support.
12516 (rs6000_gimple_fold_builtin): Ditto.
12517 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
12519 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
12520 (TARGET_DIRECT_MOVE_64BIT): Adjust.
12521 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
12522 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
12523 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
12524 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
12525 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
12526 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
12527 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
12528 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
12529 anonymous split): Adjust.
12530 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
12531 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
12533 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
12535 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
12536 when --with-gxx-include-dir is also specified.
12537 * configure: Regenerate.
12539 2018-05-09 Jim Wilson <jimw@sifive.com>
12542 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
12543 * config/riscv/t-withmultilib: New.
12544 * config/riscv/withmultilib.h: New.
12545 * doc/install.texi: Document RISC-V --with-multilib-list support.
12547 2018-05-09 Richard Biener <rguenther@suse.de>
12549 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
12551 (vect_bb_vectorization_profitable_p): Adjust. Compute
12552 actual scalar cost using the cost vector and the add_stmt_cost
12555 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
12557 PR rtl-optimization/85645
12558 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
12559 in the REG_CFA_REGISTER note for LR, don't leave it empty.
12561 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
12563 PR rtl-optimization/85645
12564 * shrink-wrap.c (spread_components): Return a boolean saying if
12565 anything was changed.
12566 (try_shrink_wrapping_separate): Iterate spread_components until
12567 nothing changes anymore.
12569 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
12571 PR rtl-optimization/85645
12572 * regrename.c (build_def_use): Also kill the chains that include the
12573 destination of a REG_CFA_REGISTER note.
12575 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
12577 PR rtl-optimization/85645
12578 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
12579 insn that has a REG_CFA_REGISTER note.
12581 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
12583 * cfgexpand.c (expand_clobber): New function.
12584 (expand_gimple_stmt_1): Use it.
12585 * tree-vect-stmts.c (vect_clobber_variable): New function,
12587 (vectorizable_simd_clone_call): ...here.
12588 (vectorizable_store): Emit a clobber either side of an
12589 IFN_STORE_LANES sequence.
12590 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
12592 2018-05-09 Tom de Vries <tom@codesourcery.com>
12595 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
12596 (define_insn "trap_if_false"): Add exit after trap.
12598 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
12600 PR rtl-optimization/85638
12601 * bb-reorder.c: Include common/common-target.h.
12602 (create_forwarder_block): New function extracted from...
12603 (fix_up_crossing_landing_pad): ...here. Rename into...
12604 (dw2_fix_up_crossing_landing_pad): ...this.
12605 (sjlj_fix_up_crossing_landing_pad): New function.
12606 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
12607 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
12608 from both partitions and exit the loop after one iteration.
12610 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
12613 * doc/extend.texi (PowerPC Built-in Functions): Rename this
12615 (Basic PowerPC Built-in Functions): The new name of the
12616 subsection previously known as "PowerPC Built-in Functions".
12617 (Basic PowerPC Built-in Functions Available on all Configurations):
12619 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
12621 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
12623 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
12625 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
12628 2018-05-08 Jim Wilson <jimw@sifive.com>
12630 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
12631 (LD_EMUL_SUFFIX): New.
12632 (LINK_SPEC): Use it.
12634 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
12636 * doc/extend.texi (PowerPC Built-in Functions): Rename this
12638 (Basic PowerPC Built-in Functions): The new name of the
12639 subsection previously known as "PowerPC Built-in Functions".
12640 (Basic PowerPC Built-in Functions Available on all Configurations):
12642 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
12644 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
12646 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
12648 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
12651 2018-05-08 Jakub Jelinek <jakub@redhat.com>
12654 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
12655 after cmpelim optimization.
12657 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
12659 * config.gcc: Support "goldmont".
12660 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
12661 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
12662 PROCESSOR_GOLDMONT.
12663 * config/i386/i386.c (m_GOLDMONT): Define.
12664 (processor_target_table): Add "goldmont".
12665 (PTA_GOLDMONT): Define.
12666 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
12667 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
12668 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
12669 (fold_builtin_cpu): Add "goldmont".
12670 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
12671 (ix86_option_override_internal): Add "goldmont".
12672 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
12673 (processor_type): Add PROCESSOR_GOLDMONT.
12674 * config/i386/i386.md: Add CPU "glm".
12675 * config/i386/glm.md: New file.
12676 * config/i386/x86-tune.def: Add m_GOLDMONT.
12677 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
12679 2018-05-08 Jakub Jelinek <jakub@redhat.com>
12682 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
12684 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
12685 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
12686 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
12689 * config/i386/i386.c (ix86_fold_builtin): Handle
12690 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
12693 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
12694 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
12696 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
12699 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
12700 (check_arch): Likewise.
12701 (check_fpu): Return the result rather than printing it.
12702 (end arch): Fix operator precedence.
12703 (end cpu): Likewise.
12704 (END): Print the result from check_fpu.
12706 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
12707 Alan Hayward <alan.hayward@arm.com>
12708 David Sherwood <david.sherwood@arm.com>
12710 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
12711 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
12712 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
12713 (*fcmuo<mode>_and): New patterns.
12715 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
12717 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
12718 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
12719 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
12720 (cmp_op, sve_imm_con): New code attributes.
12721 (SVE_COND_INT_CMP, imm_con): Delete.
12722 (cmp_op): Remove above unspecs from int attribute.
12723 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
12725 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
12726 comparison-specific unspecs.
12727 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
12728 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
12729 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
12730 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
12731 (*vec_fcm<cmp_op><mode>): Rename to...
12732 (*fcm<cmp_op><mode>): ...this and adjust likewise.
12733 (*vec_fcmuo<mode>): Rename to...
12734 (*fcmuo<mode>): ...this and adjust likewise.
12735 (*pred_fcm<cmp_op><mode>): New pattern.
12736 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
12737 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
12739 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
12741 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
12742 (aarch64_emit_sve_predicated_cond): New function.
12743 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
12744 (aarch64_emit_unspec_cond_or): Replace with...
12745 (aarch64_emit_sve_or_conds): ...this new function. Use
12746 aarch64_emit_sve_ptrue_op for the individual comparisons and
12747 aarch64_emit_binop to OR them together.
12748 (aarch64_emit_inverted_unspec_cond): Replace with...
12749 (aarch64_emit_sve_inverted_cond): ...this new function. Use
12750 aarch64_emit_sve_ptrue_op for the comparison and
12751 aarch64_emit_unop to invert the result.
12752 (aarch64_expand_sve_vec_cmp_float): Update after the above
12753 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
12755 2018-05-07 Nathan Sidwell <nathan@acm.org>
12757 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
12758 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
12759 (Backwards Compatibility): Likewise.
12761 2018-05-07 Luis Machado <luis.machado@linaro.org>
12765 2018-05-07 Luis Machado <luis.machado@linaro.org>
12767 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
12768 <prefetch_dynamic_strides>: New const bool field.
12769 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
12770 prefetch_dynamic_strides.
12771 (exynosm1_prefetch_tune): Likewise.
12772 (thunderxt88_prefetch_tune): Likewise.
12773 (thunderx_prefetch_tune): Likewise.
12774 (thunderx2t99_prefetch_tune): Likewise.
12775 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
12777 (aarch64_override_options_internal): Update to set
12778 PARAM_PREFETCH_DYNAMIC_STRIDES.
12779 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
12780 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
12781 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
12782 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
12783 prefetch-dynamic-strides setting.
12785 2018-05-07 Luis Machado <luis.machado@linaro.org>
12787 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
12788 <minimum_stride>: New const int field.
12789 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
12790 minimum_stride field.
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 minimum_stride to 2048.
12796 (aarch64_override_options_internal): Update to set
12797 PARAM_PREFETCH_MINIMUM_STRIDE.
12798 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
12799 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
12800 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
12801 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
12802 stride is constant and is below the minimum stride threshold.
12804 2018-05-07 Luis Machado <luis.machado@linaro.org>
12806 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
12809 2018-05-07 Luis Machado <luis.machado@linaro.org>
12811 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
12812 <prefetch_dynamic_strides>: New const bool field.
12813 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
12814 prefetch_dynamic_strides.
12815 (exynosm1_prefetch_tune): Likewise.
12816 (thunderxt88_prefetch_tune): Likewise.
12817 (thunderx_prefetch_tune): Likewise.
12818 (thunderx2t99_prefetch_tune): Likewise.
12819 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
12821 (aarch64_override_options_internal): Update to set
12822 PARAM_PREFETCH_DYNAMIC_STRIDES.
12823 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
12824 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
12825 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
12826 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
12827 prefetch-dynamic-strides setting.
12829 2018-05-07 Luis Machado <luis.machado@linaro.org>
12831 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
12832 <minimum_stride>: New const int field.
12833 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
12834 minimum_stride field.
12835 (exynosm1_prefetch_tune): Likewise.
12836 (thunderxt88_prefetch_tune): Likewise.
12837 (thunderx_prefetch_tune): Likewise.
12838 (thunderx2t99_prefetch_tune): Likewise.
12839 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
12840 (aarch64_override_options_internal): Update to set
12841 PARAM_PREFETCH_MINIMUM_STRIDE.
12842 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
12843 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
12844 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
12845 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
12846 stride is constant and is below the minimum stride threshold.
12848 2018-05-06 Jakub Jelinek <jakub@redhat.com>
12851 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
12852 the type is addressable. Don't force op into register if it has
12855 2018-05-05 Roland McGrath <mcgrathr@google.com>
12858 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
12859 any section for which we don't know a specific type it should have,
12860 regardless of name. Previously this was done only for the exact
12861 names ".init_array", ".fini_array", and ".preinit_array".
12862 (default_elf_asm_named_section): Add comment about
12863 relationship with default_section_type_flags and SECTION_NOTYPE.
12864 (get_section): Don't consider it a type conflict if one side has
12865 SECTION_NOTYPE and the other doesn't, as long as neither has the
12866 SECTION_BSS et al used in the default_section_type_flags logic.
12868 2018-05-05 Tom de Vries <tom@codesourcery.com>
12871 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
12872 (workaround_barsyncs): New function.
12873 (nvptx_reorg): Use workaround_barsyncs.
12874 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
12875 (define_expand "nvptx_membar_cta"): New define_expand.
12876 (define_insn "*nvptx_membar_cta"): New insn.
12878 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
12880 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
12881 To improve optimization opportunities.
12882 * builtin-types.def: The new needed builtin types for the above.
12884 2018-05-04 Richard Biener <rguenther@suse.de>
12886 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
12887 * gimple-ssa-store-merging.c
12888 (imm_store_chain_info::output_merged_store): Remove redundant create,
12889 release split_store vector contents on failure.
12890 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
12891 scalar stmt vector on cache hit.
12893 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
12895 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
12897 * config.gcc (powerpc-xilinx-eabi*): Remove.
12898 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
12900 (fusion_addis_mem_combo_load): Ditto.
12901 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
12903 (rs6000_cpu_cpp_builtins): Ditto.
12904 * config/rs6000/rs6000-linux.c
12905 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
12906 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
12907 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
12909 (rs6000_setup_reg_addr_masks): Ditto.
12910 (rs6000_init_hard_regno_mode_ok): Ditto.
12911 (rs6000_option_override_internal): Ditto.
12912 (legitimate_lo_sum_address_p): Ditto.
12913 (rs6000_legitimize_address): Ditto.
12914 (rs6000_legitimize_reload_address): Ditto.
12915 (rs6000_legitimate_address_p): Ditto.
12916 (abi_v4_pass_in_fpr): Ditto.
12917 (setup_incoming_varargs): Ditto.
12918 (rs6000_gimplify_va_arg): Ditto.
12919 (rs6000_split_multireg_move): Ditto.
12920 (rs6000_savres_strategy): Ditto.
12921 (rs6000_emit_prologue_components): Ditto.
12922 (rs6000_emit_epilogue_components): Ditto.
12923 (rs6000_emit_prologue): Ditto.
12924 (rs6000_emit_epilogue): Ditto.
12925 (rs6000_elf_file_end): Ditto.
12926 (rs6000_function_value): Ditto.
12927 (rs6000_libcall_value): Ditto.
12928 * config/rs6000/rs6000.h: Ditto.
12929 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
12930 (TARGET_MINMAX): ... this. New.
12931 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
12932 * config/rs6000/rs6000.md: Remove Xilinx FP support.
12933 (*movsi_internal1_single): Delete.
12934 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
12935 mfpu=, mxilinx-fpu): Delete.
12936 * config/rs6000/singlefp.h: Delete.
12937 * config/rs6000/sysv4.h: Remove Xilinx FP support.
12938 * config/rs6000/t-rs6000: Ditto.
12939 * config/rs6000/t-xilinx: Delete.
12940 * config/rs6000/titan.md: Adjust for fp_type removal.
12941 * config/rs6000/vsx.md: Remove Xilinx FP support.
12942 (VStype_simple): Delete.
12943 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
12944 * config/rs6000/xfpu.h: Delete.
12945 * config/rs6000/xfpu.md: Delete.
12946 * config/rs6000/xilinx.h: Delete.
12947 * config/rs6000/xilinx.opt: Delete.
12948 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
12949 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
12951 2018-05-04 Tom de Vries <tom@codesourcery.com>
12954 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
12957 2018-05-04 Richard Biener <rguenther@suse.de>
12959 PR middle-end/85627
12960 * tree-complex.c (update_complex_assignment): We are always in SSA form.
12961 (expand_complex_div_wide): Likewise.
12962 (expand_complex_operations_1): Likewise.
12963 (expand_complex_libcall): Preserve EH info of the original stmt.
12964 (tree_lower_complex): Handle removed blocks.
12965 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
12966 on complex multiplication and division libcall builtins.
12968 2018-05-04 Richard Biener <rguenther@suse.de>
12970 PR middle-end/85574
12971 * fold-const.c (negate_expr_p): Restrict negation of operand
12972 zero of a division to when we know that can happen without
12974 (fold_negate_expr_1): Likewise.
12976 2018-05-04 Jakub Jelinek <jakub@redhat.com>
12979 * real.h (real_nextafter): Declare.
12980 * real.c (real_nextafter): New function.
12981 * fold-const-call.c (fold_const_nextafter): New function.
12982 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
12983 CASE_CFN_NEXTTOWARD.
12984 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
12985 even when arg1_mode is different from arg0_mode.
12987 2018-05-03 Nathan Sidwell <nathan@acm.org>
12989 * doc/extend.texi (Deprecated Features): Remove
12990 -ffriend-injection.
12991 (Backwards Compatibility): Likewise.
12992 * doc/invoke.texi (C++ Language Options): Likewise.
12993 (C++ Dialect Options): Likewise.
12995 2018-05-03 Jakub Jelinek <jakub@redhat.com>
12998 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
12999 _mm512_mask_mullox_epi64): New intrinsics.
13001 2018-05-03 Tom de Vries <tom@codesourcery.com>
13004 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
13005 dump files): Add offload-tree.
13007 2018-05-03 Richard Biener <rguenther@suse.de>
13009 PR tree-optimization/85615
13010 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
13011 to loops not nested in BBs loop father to avoid creating multi-entry
13014 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13016 PR tree-optimization/70291
13017 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
13018 arguments. Change return type to tree. Emit libcall as a new
13019 statement rather than replacing existing one when inplace_p is true.
13020 (expand_complex_multiplication_components): New function.
13021 (expand_complex_multiplication): Expand floating-point complex
13022 multiplication using the above.
13023 (expand_complex_division): Rename inner_type parameter to type.
13024 Update expand_complex_libcall call-site.
13025 (expand_complex_operations_1): Update expand_complex_multiplication
13026 and expand_complex_division call-sites.
13028 2018-05-02 Jakub Jelinek <jakub@redhat.com>
13031 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
13032 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
13033 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
13034 the highest significant bit of the shift count mask is clear. In
13035 check whether and[sq]i3 is needed verify that all significant bits
13036 of the shift count other than the highest are set.
13038 2018-05-02 Tom de Vries <tom@codesourcery.com>
13041 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
13042 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
13043 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
13044 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
13045 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
13046 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
13047 __builtin_goacc_parlevel_size.
13049 2018-05-02 Richard Biener <rguenther@suse.de>
13051 PR tree-optimization/85597
13052 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
13053 do not use split vect_get_vec_defs call but call vect_get_slp_defs
13056 2018-05-02 Tom de Vries <tom@codesourcery.com>
13059 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
13060 dump files): Add ltrans-tree.
13062 2018-05-02 Tom de Vries <tom@codesourcery.com>
13065 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
13066 dump files): Add wpa-ipa.
13068 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
13070 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
13071 powerpc*-*-linux*paired* target.
13072 * config/rs6000/750cl.h: Delete.
13073 * config/rs6000/paired.h: Delete.
13074 * config/rs6000/paired.md: Delete.
13075 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
13077 * config/rs6000/rs6000-builtin.def: Remove paired float support.
13078 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
13079 comment. Remove paired float support.
13080 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
13081 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
13083 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
13084 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
13086 * config/rs6000/rs6000.c: Remove paired float support.
13087 (paired_expand_vector_init, paired_expand_vector_move,
13088 paired_emit_vector_compare, paired_emit_vector_cond_expr,
13089 (paired_expand_lv_builtin, paired_expand_stv_builtin,
13090 paired_expand_builtin, paired_expand_predicate_builtin,
13091 paired_init_builtins): Delete.
13092 * config/rs6000/rs6000.h: Remove paired float support.
13093 * config/rs6000/rs6000.md: Remove paired float support.
13094 (move_from_CR_ov_bit): Delete.
13095 * config/rs6000/rs6000.opt (mpaired): Delete.
13096 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
13097 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
13099 2018-05-02 Richard Biener <rguenther@suse.de>
13101 PR middle-end/85567
13102 * gimplify.c (gimplify_save_expr): When in SSA form allow
13103 SAVE_EXPRs to compute to SSA vars.
13105 2018-05-02 Jakub Jelinek <jakub@redhat.com>
13108 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
13109 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
13110 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
13111 clobber operands[2], instead use a new pseudo. Formatting fixes.
13113 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
13115 PR tree-optimization/85586
13116 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
13117 exit early for statements in the same group if the accesses are
13120 2018-05-02 Tom de Vries <tom@codesourcery.com>
13123 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
13126 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
13128 PR tree-optimization/85143
13129 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
13131 2018-05-01 Tom de Vries <tom@codesourcery.com>
13134 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
13135 not found" error message.
13137 2018-05-01 Tom de Vries <tom@codesourcery.com>
13140 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
13141 * vec.c (test_ordered_remove_if): New function.
13142 (vec_c_tests): Call test_ordered_remove_if.
13143 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
13144 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
13145 * tree-vect-patterns.c (vect_pattern_recog_1): Use
13146 VEC_ORDERED_REMOVE_IF.
13148 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13150 PR tree-optimization/82665
13151 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
13152 pointer subtraction where arguments come from a memchr call.
13154 2018-05-01 Jakub Jelinek <jakub@redhat.com>
13156 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
13157 --push-state --as-needed and --pop-state instead of --as-needed and
13158 --no-as-needed if ld supports it.
13159 * configure: Regenerated.
13162 * doc/install.texi2html: Replace _002d with - and _002a with * in
13163 generated html files using sed.
13165 2018-04-30 David Malcolm <dmalcolm@redhat.com>
13168 * gcc-rich-location.c (blank_line_before_p): New function.
13169 (use_new_line): New function.
13170 (gcc_rich_location::add_fixit_insert_formatted): New function.
13171 * gcc-rich-location.h
13172 (gcc_rich_location::add_fixit_insert_formatted): New function.
13174 2018-04-30 David Malcolm <dmalcolm@redhat.com>
13176 * selftest.c (assert_streq): Rename "expected" and "actual" to
13177 "val1" and "val2". Extend NULL-handling to cover both inputs
13178 symmetrically, while still requiring both to be non-NULL for a pass.
13179 * selftest.h (assert_streq): Rename "expected" and "actual" to
13181 (ASSERT_EQ): Likewise.
13182 (ASSERT_EQ_AT): Likewise.
13183 (ASSERT_KNOWN_EQ): Likewise.
13184 (ASSERT_KNOWN_EQ_AT): Likewise.
13185 (ASSERT_NE): Likewise.
13186 (ASSERT_MAYBE_NE): Likewise.
13187 (ASSERT_MAYBE_NE_AT): Likewise.
13188 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
13189 the assertion to pass.
13190 (ASSERT_STREQ_AT): Likewise.
13192 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
13194 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
13195 interaction with -pie.
13197 2018-04-30 David Malcolm <dmalcolm@redhat.com>
13199 * selftest.h: Fix alphabetization of per-source-file selftest
13202 2018-04-30 Jason Merrill <jason@redhat.com>
13204 PR c++/61982 - dead stores to destroyed objects.
13205 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
13208 2018-04-30 Jason Merrill <jason@redhat.com>
13210 * tree.c (build_clobber): New.
13211 * tree.h: Declare it.
13212 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
13214 2018-04-30 David Malcolm <dmalcolm@redhat.com>
13216 * diagnostic-show-locus.c (layout::layout): Update for
13217 location_get_source_line returning a char_span.
13218 (struct char_span): Move to input.h.
13219 (struct correction): Update for fields in char_span becoming
13221 (struct source_line): Update for location_get_source_line
13222 returning a char_span.
13223 (layout::print_line): Likewise.
13224 * edit-context.c (edited_file::print_content): Likewise.
13225 (edited_file::print_diff_hunk): Likewise.
13226 (edited_file::print_run_of_changed_lines): Likewise.
13227 (edited_file::get_num_lines): Likewise.
13228 (edited_line::edited_line): Likewise.
13229 * final.c (asm_show_source): Likewise.
13230 * input.c (location_get_source_line): Convert return type
13231 from const char * to char_span, losing the final "line_len"
13233 (dump_location_info): Update for the above.
13234 (get_substring_ranges_for_loc): Likewise. Use a char_span
13235 when handling the literal within the line.
13236 (test_reading_source_line): Update for location_get_source_line
13237 returning a char_span.
13238 * input.h (class char_span): Move here from
13239 diagnostic-show-locus.c, converting from a struct to a class.
13240 Make data members private.
13241 (char_span::operator bool): New.
13242 (char_span::length): New.
13243 (char_span::get_buffer): New.
13244 (char_span::operator[]): New.
13245 (char_span::subspan): Make const.
13246 (char_span::xstrdup): New.
13247 (location_get_source_line): Convert return type from const char *
13248 to char_span, losing the final "line_size" param.
13250 2018-04-30 Jan Hubicka <jh@suse.cz>
13252 * lto-wrapper.c (ltrans_priorities): New static var.
13253 (cmp_priority): New.
13254 (run_gcc): Read priorities and if doing parallel build order
13255 the Makefile by them.
13257 2018-04-30 David Malcolm <dmalcolm@redhat.com>
13259 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
13261 2018-04-30 Richard Biener <rguenther@suse.de>
13263 * tree-cfg.c (verify_address): Remove base argument, add
13264 flag whether to check TREE_ADDRESSABLE and do that.
13265 (verify_expr): Remove.
13266 (verify_types_in_gimple_reference): Add pieces from verify_expr.
13267 (verify_gimple_assign_single): Likewise.
13268 (verify_gimple_switch): Likewise.
13269 (verify_expr_location_1): Dereference tp once. Add (disabled)
13270 piece from verify_expr.
13271 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
13273 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
13275 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
13277 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
13279 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
13280 (small_data_pattern): Likewise.
13281 (arc_rewrite_small_data): Likewise.
13282 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
13283 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
13284 (get_symbol_alignment): New function.
13285 (legitimate_small_data_address_p): Likewise.
13286 (legitimate_scaled_address): Update, call
13287 legitimate_small_data_address_p.
13288 (output_sdata): New static variable.
13289 (arc_print_operand): Update how we handle small data operands.
13290 (arc_print_operand_address): Likewise.
13291 (arc_legitimate_address_p): Update, use
13292 legitimate_small_data_address_p.
13293 (arc_rewrite_small_data_p): Remove.
13294 (arc_rewrite_small_data_1): Likewise.
13295 (arc_rewrite_small_data): Likewise.
13296 (small_data_pattern): Likewise.
13297 (compact_sda_memory_operand): Update to use
13298 legitimate_small_data_address_p and get_symbol_alignment.
13299 (prepare_move_operands): Don't rewite sdata pattern.
13300 (prepare_extend_operands): Remove.
13301 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
13303 (zero_extendqisi2): Likewise.
13304 (zero_extendhisi2): Likewise.
13305 (extendqihi2): Likewise.
13306 (extendqisi2): Likewise.
13307 (extendhisi2): Likewise.
13308 (addsi3): Likewise.
13309 (subsi3): Likewise.
13310 (andsi3): Likewise.
13311 * config/arc/constraints.md (Usd): Change it to memory constraint.
13313 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
13315 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
13316 as source of std instructions.
13317 * config/arc/arc.md (movsi_insn): Update pattern predicate to
13318 allow 6-bit constants as source for store instructions.
13319 (movdi_insn): Update instruction pattern to allow 6-bit constants
13320 as source for store instructions.
13322 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
13324 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
13326 2018-04-30 Nathan Sidwell <nathan@acm.org>
13327 Sandra Loosemore <sandra@codesourcery.com>
13329 * dumpfile.c (dump_open): Allow '-' for stdout.
13330 * doc/invoke.texi (Developer Options): Document dump filename
13331 determination early. Document stdin/stdout selection.
13333 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
13335 Microblaze Target: PIC data text relative
13337 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
13338 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
13340 * config/microblaze/microblaze.h (microblaze_constant_address_p):
13341 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
13342 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
13343 New addressing mode for data-text relative position indepenedent code.
13344 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
13345 'ADDRESS_SYMBOLIC_TXT_REL'.
13346 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
13347 (microblaze_legitimate_pic_operand): Exclude function calls from
13348 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
13349 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
13351 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
13352 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
13353 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
13354 for 'address + offset'.
13355 (microblaze_expand_prologue): Add new function prologue call for
13357 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
13358 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
13359 table in case of TARGET_PIC_DATA_TEXT_REL.
13360 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
13361 * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
13362 Add new macros 'UNSPEC_TEXT',
13363 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
13364 + exclude function calls from 'UNSPEC_PLT' in case of data text
13366 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
13367 new target hook for generating address diff vector tables in case of
13369 * doc/tm.texi : Regenerate.
13370 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
13371 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
13372 of addr diff vector generation.
13373 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
13374 target hook definition.
13375 * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
13376 Add default function for generate_pic_addr_diff_vec -> flag_pic.
13377 * doc/invoke.texi (Add new pic option): Add new microblaze pic
13378 option for data text relative.
13380 2018-04-30 Richard Biener <rguenther@suse.de>
13382 * tree-chrec.h (evolution_function_is_constant_p): Remove
13384 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
13386 2018-04-30 Richard Biener <rguenther@suse.de>
13389 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
13391 2018-04-30 Richard Biener <rguenther@suse.de>
13393 PR tree-optimization/28364
13394 PR tree-optimization/85275
13395 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
13396 copying first exit test.
13398 2018-04-28 Mark Wielaard <mark@klomp.org>
13400 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
13401 dwarf_version >= 5.
13402 (dwarf_AT): Handle DW_AT_addr_base.
13403 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
13405 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
13408 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
13409 (*ashl<dwi>3_doubleword_mask_1): Ditto.
13410 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
13411 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
13413 2018-04-28 Richard Biener <rguenther@suse.de>
13415 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
13416 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
13417 to reflect use. Only add interesting stmts.
13419 2018-04-27 Martin Jambor <mjambor@suse.cz>
13422 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
13423 the jump function allows for passing through aggregate values.
13425 2018-04-27 David Malcolm <dmalcolm@redhat.com>
13427 * input.h (in_system_header_at): Convert from macro to inline
13429 (from_macro_expansion_at): Likewise.
13430 (from_macro_definition_at): Likewise.
13432 2018-04-27 Jeff Law <law@redhat.com>
13434 * config.gcc: Mark tile* targets as deprecated/obsolete.
13436 2018-04-27 Richard Biener <rguenther@suse.de>
13438 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
13441 2018-04-27 Richard Biener <rguenther@suse.de>
13443 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
13445 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
13447 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
13448 with Yd constraint. Set "preferred_for_speed" attribute from
13449 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
13450 with Yd constraint.
13451 (*movdi_internal): Ditto.
13452 (movti_interunit splitters): Remove
13453 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
13454 (movdi_interunit splitters): Ditto.
13455 * config/i386/constraints.md (Ye): Remove.
13456 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
13458 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13461 * config/aarch64/constraints.md (Usg): Limit to 31.
13462 (Usj): Limit to 63.
13464 2018-04-27 Jakub Jelinek <jakub@redhat.com>
13466 PR tree-optimization/85529
13467 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
13468 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
13469 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
13470 zero extension or masking of the MSB bit.
13471 (optimize_range_tests): Add FIRST_BB argument, pass it through
13472 to optimize_range_tests_var_bound.
13473 (maybe_optimize_range_tests, reassociate_bb): Adjust
13474 optimize_range_tests callers.
13476 2018-04-26 Richard Biener <rguenther@suse.de>
13477 Jakub Jelinek <jakub@redhat.com>
13479 * cgraph.h (symbol_table): Just declare debug method here.
13480 * symtab.c (symbol_table::debug): Define.
13482 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
13484 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
13486 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
13488 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
13489 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
13490 (*movdi_internal): Substitute Yi and Yj constraint with x
13491 and Ym and Yn constraint with y constraint. Update "isa"
13492 attribute and set "preferred_for_speed" attribute from
13493 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
13494 (*movsi_internal): Ditto.
13495 (*movdf_internal): Ditto.
13496 (*movsf_internal): Ditto.
13497 (*zero_extendsidi2): Ditto.
13498 * config/i386/sse.md (vec_set<mode>_0): Ditto.
13499 (sse2_loadld): Ditto.
13500 (*vec_extract<ssevecmodelower>_0): Ditto.
13501 (*vec_extractv4si_0_zext_sse4): Ditto.
13502 (vec_concatv2di): Ditto.
13503 (*vec_dup<mode>): Ditto.
13504 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
13505 * config/i386/constraints.md (Yi): Remove.
13510 2018-04-26 Nathan Sidwell <nathan@acm.org>
13512 * dumpfile.c (dump_open): New.
13513 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
13514 (dump_finish): Detect stdio/stderr by value not name.
13516 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
13518 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
13520 2018-04-26 Tom de Vries <tom@codesourcery.com>
13523 * config/nvptx/nvptx.c (verify_neutering_jumps)
13524 (verify_neutering_labels): New function
13525 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
13527 2018-04-26 Tom de Vries <tom@codesourcery.com>
13530 * config/nvptx/nvptx.c (needs_neutering_p): New function.
13531 (nvptx_single): Use needs_neutering_p to skip over insns that do not
13534 2018-04-26 Richard Biener <rguenther@suse.de>
13535 Tom de Vries <tom@codesourcery.com>
13538 * lto-streamer-out.c (output_function): Fixup loops if required to match
13539 discovery done in the reader.
13541 2018-04-26 Richard Biener <rguenther@suse.de>
13543 PR tree-optimization/85116
13544 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
13545 have a loop exit from the single latch predecessor. Remove
13546 case of header with just condition.
13547 (ch_base::copy_headers): Exclude infinite loops from any
13549 (pass_ch::execute): Record exits.
13551 2018-04-26 Richard Biener <rguenther@suse.de>
13553 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
13554 prologue cost vector and pass it to vect_get_load_cost.
13555 (vect_get_peeling_costs_all_drs): Likewise.
13556 (vect_peeling_hash_get_lowest_cost): Likewise.
13557 (vect_enhance_data_refs_alignment): Likewise.
13559 2018-04-26 Richard Biener <rguenther@suse.de>
13561 PR middle-end/85450
13562 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
13563 checking of integer<->pointer conversions.
13564 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
13565 sign-/zero-extending pointer types.
13566 (expand_omp_for_static_chunk): Likewise.
13568 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
13569 Jean Lee <xiaoyur347@gmail.com>
13571 * config/mips/mips.c (mips_asan_shadow_offset): New function.
13572 (TARGET_ASAN_SHADOW_OFFSET): Define.
13573 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
13574 true for -fsanitize=address.
13576 2018-04-25 Mark Wielaard <mark@klomp.org>
13578 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
13581 2018-04-25 Jakub Jelinek <jakub@redhat.com>
13583 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
13584 than "alu", remove explicit "memory" and "imm_disp" attributes.
13585 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
13587 PR middle-end/85414
13588 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
13589 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
13590 gen_lowpart_no_emit.
13592 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
13595 * config/i386/i386.c (ix86_expand_builtin): Change memory
13596 operand to XI, extend p0 to Pmode.
13597 * config/i386/i386.md: Change unspec volatile and operand
13598 1 mode to XI, change operand 0 mode to P.
13600 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
13602 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
13603 GET_MODE_MASK before any checking.
13604 (nds32_can_use_bset_p): Likewise.
13605 (nds32_can_use_btgl_p): Likewise.
13607 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
13609 * config/nds32/nds32-doubleword.md: New define_split pattern for
13610 illegal register number.
13612 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
13614 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
13616 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
13618 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
13620 2018-04-25 Richard Biener <rguenther@suse.de>
13622 * lto-streamer.h (LTO_major_version): Bump to 8.
13624 2018-04-25 Jakub Jelinek <jakub@redhat.com>
13626 * BASE-VER: Set to 9.0.0.
13628 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
13630 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
13631 in __abskf2 and __powikf2.
13633 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13636 * config/aarch64/constraints.md (Usg, Usj): New constraints.
13637 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
13638 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
13639 Use the above on operand 2. Reindent.
13640 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
13642 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
13645 * common/config/i386/i386-common.c (ix86_handle_option): Don't
13647 * config/i386/i386.opt (mcet): Removed.
13648 * doc/install.texi: Remove -mcet documentation.
13649 * doc/invoke.texi: Likewise.
13651 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
13654 * doc/install.texi: Remove -mcet from bootstrap-cet.
13656 2018-04-24 Jakub Jelinek <jakub@redhat.com>
13659 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
13660 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
13664 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
13665 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
13666 containing a CONST_VECTOR.
13668 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
13670 * doc/install.texi: Update newlib dependency for nvptx.
13672 2018-04-24 Jakub Jelinek <jakub@redhat.com>
13675 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
13676 instead of INTVAL when shifting x left.
13678 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
13680 PR tree-optimization/85478
13681 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
13682 vect_grouped_store_supported for single element vectors.
13684 2018-04-24 Richard Biener <rguenther@suse.de>
13687 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
13688 load cost increase to the case of non-constant step.
13690 2018-04-24 Jakub Jelinek <jakub@redhat.com>
13693 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
13694 destination if any_malformed_asm.
13696 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
13698 PR middle-end/85496
13699 * expr.c (store_field): In the bitfield case, if the value comes from
13700 a function call and is returned in registers by means of a PARALLEL,
13701 do not change the mode of the temporary unless BLKmode and VOIDmode.
13703 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
13705 PR rtl-optimization/85423
13706 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
13707 dependencies to debug insns when the previous insn is non-debug.
13709 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
13711 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
13712 enums into a single definition.
13713 (fls): Fix predicates and printing.
13716 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
13718 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
13719 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
13720 and short u6 immediate.
13721 (check_if_valid_sleep_operand): Remove.
13722 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
13724 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
13726 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
13727 flag_always_save_lp condition.
13728 * config/nds32/nds32.opt (malways-save-lp): New option.
13730 2018-04-22 Shiva Chen <shiva0217@gmail.com>
13732 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
13733 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
13734 * config/nds32/nds32.h
13735 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
13736 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
13738 2018-04-22 Shiva Chen <shiva0217@gmail.com>
13740 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
13741 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
13743 2018-04-22 Shiva Chen <shiva0217@gmail.com>
13744 Chung-Ju Wu <jasonwucj@gmail.com>
13746 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
13748 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
13749 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
13751 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
13753 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
13755 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
13757 * config/nds32/nds32-protos.h (nds32_data_alignment,
13758 nds32_local_alignment): Declare.
13759 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
13760 nds32_local_alignment): New functions.
13761 (TARGET_CONSTANT_ALIGNMENT): Define.
13762 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
13764 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
13766 * config/nds32/nds32.c
13767 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
13768 (TARGET_MODES_TIEABLE_P): Likewise.
13770 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
13772 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
13773 level Ofast and Og.
13775 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
13776 Chung-Ju Wu <jasonwucj@gmail.com>
13778 * config/nds32/constants.md (unspec_volatile_element): Add enum values
13779 for unaligned access.
13780 * config/nds32/nds32-intrinsic.c: Implementation of expanding
13782 * config/nds32/nds32-intrinsic.md: Likewise.
13783 * config/nds32/nds32_intrinsic.h: Likewise.
13784 * config/nds32/nds32.h (nds32_builtins): Likewise.
13785 * config/nds32/nds32.opt (munaligned-access): New option.
13786 * config/nds32/nds32.c (nds32_asm_file_start): Display
13787 flag_unaligned_access status.
13789 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
13791 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
13792 -mno-relax is present.
13793 * config/riscv/linux.h (LINK_SPEC): Ditto.
13795 2018-04-20 Martin Sebor <msebor@redhat.com>
13798 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
13800 (gimple_fold_builtin_stxcpy_chk): Same.
13801 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
13803 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
13806 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
13807 __powikf2 when long double is IEEE 128-bit.
13809 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
13811 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
13812 step to make sure stack always aligned.
13814 2018-04-20 Carl Love <cel@us.ibm.com>
13817 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
13818 size check for arg0.
13820 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
13821 Tom de Vries <tom@codesourcery.com>
13824 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
13825 Emit insns for calls too.
13826 (nvptx_find_par): Always look for worker-level predecessor insn.
13827 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
13829 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
13830 (nvptx_process_pars): Propagate frames for calls.
13832 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
13835 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
13837 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
13838 (ix86_handle_option): Don't handle OPT_mibt.
13839 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
13841 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
13843 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
13844 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
13845 (ix86_target_macros): Define __CET__ with flag_cf_protection
13846 for -fcf-protection.
13847 * config/i386/i386.c (isa2_opts): Remove -mibt.
13848 * config/i386/i386.h (TARGET_IBT): Removed.
13849 (TARGET_IBT_P): Likewise.
13850 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
13851 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
13852 * config/i386/i386.opt (mcet): Update help message.
13853 (mshstk): Likewise.
13855 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
13856 -mcet as an alias for -mshstk.
13858 2018-04-20 Richard Biener <rguenther@suse.de>
13860 PR middle-end/85475
13861 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
13862 complexity by forcing a single use of the multiply operand.
13864 2018-04-20 Martin Jambor <mjambor@suse.cz>
13867 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
13868 recursion dependency to only apply to non-clones.
13870 2018-04-20 Martin Jambor <mjambor@suse.cz>
13873 * ipa-cp.c (create_specialized_node): Check that clones of
13874 self-recursive edges exist during IPA-CP.
13876 2018-04-19 Toon Moene <toon@moene.org>
13878 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
13881 2018-04-19 Jakub Jelinek <jakub@redhat.com>
13883 PR tree-optimization/85467
13884 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
13885 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
13886 VECTOR_CST element to type.
13888 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
13891 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
13892 * config/i386/i386.md (builtin_setjmp_setup): Removed.
13893 (builtin_longjmp): Likewise.
13894 (save_stack_nonlocal): New pattern.
13895 (restore_stack_nonlocal): Likewise.
13897 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
13900 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
13901 Replace ASM_OUTPUT_LABEL with fprintf.
13903 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
13906 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
13907 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
13908 * config/i386/i386-c.c (ix86_target_macros_internal): Also
13909 define __IBT__ and __SHSTK__ for -fcf-protection.
13910 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
13912 (ix86_trampoline_init): Likewise.
13913 (x86_output_mi_thunk): Likewise.
13914 (ix86_notrack_prefixed_insn_p): Likewise.
13915 (ix86_option_override_internal): Don't disallow -fcf-protection.
13916 * config/i386/i386.md (rdssp<mode>): Also enable for
13918 (incssp<mode>): Likewise.
13919 (nop_endbr): Likewise.
13920 * config/i386/i386.opt (mcet): Change help message to built-in
13923 (mshstk): Likewise.
13924 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
13925 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
13926 enable CET built-in functions.
13928 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
13930 * common/config/i386/i386-common.c
13931 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
13932 OPTION_MASK_ISA_MOVDIRI_UNSET,
13933 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
13934 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
13935 * config.gcc (movdirintrin.h): New header.
13936 * config/i386/cpuid.h (bit_MOVDIRI,
13937 bit_MOVDIR64B): New bits.
13938 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
13940 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
13941 (VOID, PVOID, PCVOID)): New function types.
13942 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
13943 __builtin_ia32_directstoreu_u64,
13944 __builtin_ia32_movdir64b): New builtins.
13945 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
13946 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
13948 (ix86_valid_target_attribute_inner_p): Ditto.
13949 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
13950 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
13951 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
13952 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
13953 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
13954 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
13955 (movdiri<mode>, movdir64b_<mode>): New.
13956 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
13957 * config/i386/immintrin.h: Include movdirintrin.h.
13958 * config/i386/movdirintrin.h: New file.
13959 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
13961 2018-04-19 Richard Biener <rguenther@suse.de>
13963 PR middle-end/85455
13964 * cfg.c (clear_bb_flags): When loop state says we have
13965 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
13967 2018-04-19 Richard Biener <rguenther@suse.de>
13969 PR tree-optimization/84737
13970 * tree-vect-data-refs.c (vect_copy_ref_info): New function
13971 copying restrict info.
13972 (vect_setup_realignment): Use it.
13973 * tree-vectorizer.h (vect_copy_ref_info): Declare.
13974 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
13975 the first DR to all generated stores.
13976 (vectorizable_load): Likewise for loads.
13978 2018-04-19 Jakub Jelinek <jakub@redhat.com>
13980 PR tree-optimization/85446
13981 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
13982 the integral and pointer types to have the same precision.
13984 * doc/install.texi: Document --disable-cet being the default and
13987 2018-04-18 Martin Liska <mliska@suse.cz>
13989 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
13992 2018-04-18 Martin Liska <mliska@suse.cz>
13995 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
13998 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
13999 arguments if they are comparable.
14001 2018-04-18 Martin Liska <mliska@suse.cz>
14004 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
14007 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
14010 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
14013 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
14014 ENDBR after calling __morestack.
14016 2018-04-18 David Malcolm <dmalcolm@redhat.com>
14019 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
14020 by using gcc_base_ver to generate a gcc_driver_version, and use
14021 it when generating GCC_DRIVER_NAME.
14022 * configure: Regenerate.
14024 2018-04-18 Jakub Jelinek <jakub@redhat.com>
14027 * config.gcc: Obsolete powerpc*-*-*spe*.
14029 2018-04-17 Jakub Jelinek <jakub@redhat.com>
14032 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
14033 (stabstr_D): Change type of unum from unsigned int to
14034 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
14037 2018-04-17 Jim Wilson <jimw@sifive.com>
14040 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
14041 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
14042 Set arg_pointer_offset after using pretend_args_size.
14044 2018-04-17 Jakub Jelinek <jakub@redhat.com>
14046 PR rtl-optimization/85431
14047 * dse.c (record_store): Ignore zero width stores.
14050 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
14051 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
14052 __builtin_stack_restore rather than after it.
14053 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
14054 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
14055 argument instead of virtual_dynamic_stack_rtx.
14057 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
14059 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
14061 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
14062 Add note to error message to explain internal mapping of overloaded
14063 built-in function name to non-overloaded built-in function name.
14064 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
14067 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
14070 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
14071 where the inputs overlap with the output.
14073 2018-04-17 Jakub Jelinek <jakub@redhat.com>
14075 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
14076 (=v, v) alternative and explicit "memory" attribute.
14077 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
14078 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
14080 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
14081 "sselog1" type instead of "sselog".
14082 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
14083 "sselog". Remove explicit "memory" attribute.
14084 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
14085 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
14087 (vec_extract_hi_v32hi): Merge all alternatives into one, use
14088 "sselog1" type instead of "sselog". Remove explicit "memory"
14090 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
14091 use "sselog1" type instead of "sselog". Remove explicit "memory"
14093 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
14094 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
14096 (vec_extract_hi_v64qi): Merge all alternatives into one, use
14097 "sselog1" type instead of "sselog". Remove explicit "memory"
14099 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
14100 use "sselog1" type instead of "sselog". Remove explicit "memory"
14104 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
14106 PR middle-end/85414
14107 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
14110 2018-04-17 Martin Jambor <mjambor@suse.cz>
14113 * ipa-cp.c (create_specialized_node): Call
14114 expand_all_artificial_thunks if necessary.
14116 2018-04-17 Martin Liska <mliska@suse.cz>
14119 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
14120 in message, remote space in between '_G' and '('.
14122 2018-04-17 Jakub Jelinek <jakub@redhat.com>
14125 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
14126 avx512f_vmcmp<mode>3<round_saeonly_name>,
14127 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
14128 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
14129 avx512f_rndscale<mode><round_saeonly_name>,
14130 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
14131 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
14132 Use %<iptr>2 instead of %2 for -masm=intel.
14133 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
14134 avx512f_vcvttss2usi<round_saeonly_name>,
14135 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
14137 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
14138 avx512f_vcvttsd2usi<round_saeonly_name>,
14139 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
14140 Use %q1 instead of %1 for -masm=intel.
14141 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
14142 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
14143 of %3 for -masm=intel.
14144 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
14146 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
14148 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
14150 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
14151 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
14153 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
14155 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
14156 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
14157 %g1 and one with %0 and %1.
14158 (avx512er_vmrcp28<mode><round_saeonly_name>,
14159 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
14160 %1 for -masm=intel.
14161 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
14162 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
14163 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
14164 of %0 and %{%4%} for -masm=intel.
14165 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
14166 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
14167 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
14168 order of %0 and %{%5%}%{z%} for -masm=intel.
14170 2018-04-17 Jan Hubicka <jh@suse.cz>
14173 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
14175 2018-04-17 Martin Liska <mliska@suse.cz>
14178 * multiple_target.c (create_dispatcher_calls): Set apostrophes
14179 for target_clone error message. Make default implementation
14180 clone to be a local declaration.
14181 (separate_attrs): Add new argument and check for an empty
14183 (expand_target_clones): Handle it.
14184 (ipa_target_clone): Make redirection just for target_clones
14187 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
14188 Tom de Vries <tom@codesourcery.com>
14190 PR middle-end/84955
14191 * omp-expand.c (expand_oacc_for): Add dummy false branch for
14192 tiled basic blocks without omp continue statements.
14194 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
14197 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
14198 vec_extract expression as having side effects to make sure it gets
14201 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
14204 * config/i386/i386.c (get_builtin_code_for_version): Check
14207 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
14210 * config.gcc: Support "skylake".
14211 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
14213 * config/i386/i386.c (m_SKYLAKE): Define.
14214 (processor_target_table): Add "skylake".
14215 (ix86_option_override_internal): Add "skylake".
14216 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
14217 PROCESSOR_CANNONLAKE.
14218 (get_builtin_code_for_version): Fix priority for
14219 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
14220 PROCESSOR_SKYLAKE-AVX512.
14221 * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
14222 (processor_type): Add PROCESSOR_SKYLAKE.
14224 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
14225 Jason Merrill <jason@redhat.com>
14228 * convert.c (convert_to_integer_1): Use direct recursion for
14229 enumeral types and types with a precision less than the number
14230 of bits in their mode.
14232 2018-04-16 Julia Koval <julia.koval@intel.com>
14235 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
14236 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
14238 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
14241 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
14242 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
14243 and -mno-direct-move.
14245 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
14248 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
14249 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
14250 Ensure negative shifts result in {0}.
14252 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
14254 PR rtl-optimization/79916
14255 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
14256 regs (if any) to define how to gnerate SD moves when LRA is in
14259 2018-04-13 Jakub Jelinek <jakub@redhat.com>
14261 PR rtl-optimization/85393
14262 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
14263 * except.c (expand_dw2_landing_pad_for_region): Make static.
14264 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
14265 a label and unconditional jump to old_bb, rather than
14266 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
14269 PR rtl-optimization/85376
14270 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
14271 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
14272 instead of a specific value.
14274 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
14275 Bin Cheng <bin.cheng@arm.com>
14277 PR tree-optimization/82965
14278 PR tree-optimization/83991
14279 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
14280 by_profile_only parameter.
14281 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
14282 information if the loop was predicted to iterate too many times.
14283 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
14285 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
14288 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
14291 2018-04-13 Martin Liska <mliska@suse.cz>
14292 Jakub Jelinek <jakub@redhat.com>
14294 PR middle-end/81657
14295 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
14296 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
14297 * builtins.c (expand_builtin_memory_copy_args): Use
14298 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
14299 handle dest_addr == pc_rtx.
14301 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
14304 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
14305 asked to not generate direct moves.
14306 (fix_trunc<mode>si2_stfiwx): Similar.
14307 (fix_trunc<mode>si2_internal): Similar.
14309 2018-04-12 Jakub Jelinek <jakub@redhat.com>
14312 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
14313 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
14314 lookup if dest in some wider mode is known to be const0_rtx and
14315 if so, record permanent equivalence for it to be ZERO_EXTEND of
14316 the narrower mode destination.
14318 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
14320 * lto-streamer-out.c (output_function): Revert 259346.
14321 * omp-expand.c (expand_oacc_for): Likewise.
14323 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
14325 PR rtl-optimization/85354
14326 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
14327 * sel-sched.c (sel_global_init): ... here.
14329 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
14332 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
14333 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
14334 mode for PE-COFF targets.
14335 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
14336 (i386_pe_asm_lto_end): Likewise.
14337 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
14338 (TARGET_ASM_LTO_END): Likewise.
14339 * config/i386/winnt.c (saved_debug_info_level): New static variable.
14340 (i386_pe_asm_lto_start): New function.
14341 (i386_pe_asm_lto_end): Likewise.
14343 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
14344 Richard Biener <rguenther@suse.de>
14346 PR middle-end/84955
14347 * lto-streamer-out.c (output_function): Fix CFG loop state before
14349 * omp-expand.c (expand_oacc_for): Handle calls to internal
14350 functions like regular functions.
14352 2018-04-12 Richard Biener <rguenther@suse.de>
14355 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
14356 for the early LTO debug to properly generate references to it
14357 during DIE emission. Do not re-use that for the skeleton for
14359 (dwarf2out_early_finish): Likewise.
14361 2018-04-12 Jakub Jelinek <jakub@redhat.com>
14364 * config/i386/sse.md
14365 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
14366 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
14367 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
14368 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
14369 and output is a reg, avoid creating invalid lowpart subreg, but
14370 instead split into a 512-bit move. Don't split if not AVX512VL,
14371 input is xmm16+ reg and output is a mem.
14372 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
14373 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
14374 xmm16+ reg and output is a mem.
14376 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
14378 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
14379 also for flag_dwarf2_cfi_asm.
14381 2018-04-12 Jakub Jelinek <jakub@redhat.com>
14383 PR rtl-optimization/85342
14384 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
14385 a bool scalar var inside of the loop instead. Don't try to update
14386 recog_data.operand after failed apply_change_group.
14388 2018-04-12 Tom de Vries <tom@codesourcery.com>
14391 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
14392 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
14393 array with flexible array member as array without given dimension.
14394 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
14395 argument for undefined param to true.
14397 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
14400 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
14401 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
14402 from PowerPC section.
14403 * config/rs6000/sysv4.opt (mcall-): Improve help text.
14404 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
14405 help text that is too long.
14406 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
14407 help text that is too long.
14408 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
14409 help text that is too long.
14411 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
14413 * config/alpha/alpha.md (stack_probe_internal): Rename
14414 from "probe_stack". Update all callers.
14416 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
14418 PR rtl-optimization/84566
14419 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
14420 sched_macro_fuse_insns.
14422 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
14425 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
14426 (compute_block_dependences): ... from here.
14428 2018-04-11 Jakub Jelinek <jakub@redhat.com>
14430 PR tree-optimization/85331
14431 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
14432 from int to HOST_WIDE_INT.
14434 2018-04-11 Martin Jambor <mjambor@suse.cz>
14437 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
14438 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
14439 not the same as the source val.
14440 (cgraph_edge_brings_value_p): New parameter.
14441 (gather_edges_for_value): Pass destination value to
14442 cgraph_edge_brings_value_p.
14443 (perhaps_add_new_callers): Likewise.
14444 (get_info_about_necessary_edges): Likewise and exclude values brought
14445 only by self-recursive edges.
14446 (create_specialized_node): Redirect only clones of self-calling edges.
14447 (+self_recursive_pass_through_p): New function.
14448 (find_more_scalar_values_for_callers_subset): Use it.
14449 (find_aggregate_values_for_callers_subset): Likewise.
14450 (known_aggs_to_agg_replacement_list): Removed.
14451 (decide_whether_version_node): Re-calculate known constants for all
14452 remaining context clones.
14454 2018-04-11 Richard Biener <rguenther@suse.de>
14457 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
14458 from early DWARF output.
14459 (dwarf2out_early_finish): Output line info unconditionally into
14460 early DWARF and add reference to it.
14462 2018-04-11 Jakub Jelinek <jakub@redhat.com>
14465 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
14466 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
14467 other than V2DFmode using iptr mode attribute.
14468 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
14470 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
14472 PR rtl-optimization/84659
14473 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
14475 2018-04-11 Jakub Jelinek <jakub@redhat.com>
14478 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
14480 (output_loc_list): Pass address of a dummy size variable even in the
14481 locview handling loop.
14482 (index_location_lists): Add comment on why skip_loc_list_entry can't
14485 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
14488 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
14491 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
14494 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
14495 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
14496 and -mstring-compare-inline-limit.
14498 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
14501 * config/rs6000/rs6000.md (allocate_stack): Put the residual size
14502 for stack clash protection in a register whenever we need it to be in
14505 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
14507 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
14508 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
14510 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
14513 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
14515 (mlong-double-): Ditto.
14516 * config/rs6000/sysv4.opt (msdata=): Ditto.
14517 (mtls-size=): Ditto.
14519 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
14521 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
14522 erroneous entries for
14523 "vector int vec_ldl (int, long int *)", and
14524 "vector unsigned int vec_ldl (int, unsigned long int *)".
14525 Add comments and entries for
14526 "vector bool char vec_ldl (int, bool char *)",
14527 "vector bool short vec_ldl (int, bool short *)",
14528 "vector bool int vec_ldl (int, bool int *)",
14529 "vector bool long long vec_ldl (int, bool long long *)",
14530 "vector pixel vec_ldl (int, pixel *)",
14531 "vector long long vec_ldl (int, long long *)",
14532 "vector unsigned long long vec_ldl (int, unsigned long long *)".
14533 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
14534 type tree bool_long_long_type_node and correct definition of
14535 bool_V2DI_type_node to make reference to this new type tree.
14536 (rs6000_mangle_type): Replace erroneous reference to
14537 bool_long_type_node with bool_long_long_type_node.
14538 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
14539 comments to emphasize sign distinctions for char and int types and
14540 replace RS6000_BTI_bool_long constant with
14541 RS6000_BTI_bool_long_long constant. Also add comment to restrict
14542 use of RS6000_BTI_pixel.
14543 (bool_long_type_node): Remove this macro definition.
14544 (bool_long_long_type_node): New macro definition
14546 2018-04-10 Jakub Jelinek <jakub@redhat.com>
14548 PR rtl-optimization/85300
14549 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
14550 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
14551 simplify_unary_operation fails.
14553 2018-04-10 Martin Liska <mliska@suse.cz>
14555 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
14556 cgraph_edge and ipa_ref.
14558 2018-04-10 Jakub Jelinek <jakub@redhat.com>
14562 * config/i386/sse.md
14563 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
14564 computation of the VEC_MERGE selector from mask.
14565 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
14566 Fix decoding of the VEC_MERGE selector into mask.
14568 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
14570 PR tree-optimization/85286
14571 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
14573 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
14575 * final.c (final_1): Set insn_last_address as well as
14576 insn_current_address.
14578 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14581 * explow.c (emit_stack_probe): Call validize_mem on memory location
14582 before passing it to gen_probe_stack. Create address operand and
14583 legitimize it for the probe_stack_address case.
14585 2018-04-09 Jan Hubicka <jh@suse.cz>
14588 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
14589 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
14590 * tree.c (free_lang_data_in_type): Fix handling of binfos;
14592 (free_lang_data): Rebuild type inheritance graph.
14594 2018-04-09 Martin Sebor <msebor@redhat.com>
14596 * invoke.texi (-finline-small-functions): Mention other optimization
14598 (-findirect-inlining, -fpartial-inlining): Same.
14599 (-finline-functions-called-once): Same.
14600 (-freorder-blocks-and-partition): Same.
14602 2018-04-09 Jan Hubicka <jh@suse.cz>
14605 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
14606 jumps; choose last target that matches the criteria (i.e.
14607 no partition changes for non-crossing jumps).
14608 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
14609 support for redirecting crossing jumps to non-crossing.
14611 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
14613 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
14614 also for naked functions.
14616 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
14618 * config/arc/arc.md (add_shift): New pattern.
14619 (add_shift2): Likewise.
14620 (sub_shift): Likewise.
14621 (sub_shift_cmp0_noout): Likewise.
14622 (compare_si_ashiftsi): Likewise.
14623 (xbfu_cmp0_noout): New combine pattern.
14624 (xbfu_cmp0"): Likewise.
14625 (movsi_set_cc_insn): Place the predicable variant first.
14626 (commutative_binary_cmp0_noout): Remove clobber.
14627 (commutative_binary_cmp0): New pattern.
14628 (noncommutative_binary_cmp0): Likewise.
14629 (noncommutative_binary_cmp0_noout): Likewise.
14630 (noncommutative_binary_comparison_result_used): Removed.
14631 (rsub_cmp0): New pattern.
14632 (rsub_cmp0_noout): Likewise.
14633 (extzvsi): Changed, keep only meaningful variants.
14634 (SQH, SEZ): New iterators.
14635 (SQH_postfix): New mode attribute.
14636 (SEZ_prefix): New code attribute.
14637 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
14638 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
14639 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
14640 of numerical value.
14641 (noncommutative_operator): Check the availability of barrel
14644 2018-04-09 Richard Biener <rguenther@suse.de>
14646 PR tree-optimization/85284
14647 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
14648 Only use the niter constraining form of simple_iv when the exit
14649 is always executed.
14651 2018-04-09 Tom de Vries <tom@codesourcery.com>
14654 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
14655 (define_expand "*memory_barrier"): New define_expand.
14656 (define_insn "memory_barrier"): New insn.
14658 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
14660 PR rtl-optimization/80463
14661 PR rtl-optimization/83972
14662 PR rtl-optimization/83480
14664 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
14665 correct producer for the insn.
14666 (tidy_control_flow): Fixup seqnos in case of debug insns.
14668 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
14670 PR rtl-optimization/83913
14672 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
14673 different sched-times when merging exprs.
14675 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
14677 PR rtl-optimization/83962
14679 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
14680 tidy_fallthru_edge and tidy_control_flow.
14682 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
14684 PR rtl-optimization/83530
14686 * sel-sched.c (force_next_insn): New global variable.
14687 (remove_insn_for_debug): When force_next_insn is true, also leave only
14688 next insn in the ready list.
14689 (sel_sched_region): When the region wasn't scheduled, make another pass
14690 over it with force_next_insn set to 1.
14692 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
14694 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
14696 * config/nds32/constants.md (unspec_volatile_element): Add enum values
14697 for interrupt control.
14698 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
14699 functions for interrupt control.
14700 * config/nds32/nds32-intrinsic.md: Likewise.
14701 * config/nds32/nds32_intrinsic.h: Likewise.
14702 * config/nds32/nds32.h (nds32_builtins): Likewise.
14704 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
14706 * config/nds32/nds32.c (nds32_init_machine_status,
14707 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
14708 strict_aligned_p field.
14709 (nds32_expand_to_rtl_hook): New function.
14710 (TARGET_EXPAND_TO_RTL_HOOK): Define.
14711 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
14713 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
14714 Chung-Ju Wu <jasonwucj@gmail.com>
14716 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
14717 * config/nds32/nds32-n7.md: New file.
14718 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
14719 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
14721 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
14722 * config/nds32/nds32.md (pipeline_model): Add n7.
14723 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
14724 * config/nds32/pipelines.md: Include n7 settings.
14726 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
14727 Chung-Ju Wu <jasonwucj@gmail.com>
14729 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
14730 * config/nds32/nds32-e8.md: New file.
14731 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
14732 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
14734 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
14735 * config/nds32/nds32.md (pipeline_model): Add e8.
14736 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
14737 * config/nds32/pipelines.md: Include e8 settings.
14739 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
14740 Chung-Ju Wu <jasonwucj@gmail.com>
14742 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
14743 * config/nds32/nds32-n8.md: New file.
14744 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
14745 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
14747 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
14748 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
14749 * config/nds32/nds32.md (pipeline_model): Add n8.
14750 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
14751 * config/nds32/pipelines.md: Include n8 settings.
14753 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
14754 Chung-Ju Wu <jasonwucj@gmail.com>
14756 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
14757 * config/nds32/nds32-n9-2r1w.md: New file.
14758 * config/nds32/nds32-n9-3r2w.md: New file.
14759 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
14760 nds32_register_ports): New or modify for cpu n9.
14761 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
14763 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
14764 * config/nds32/nds32-utils.c: New file.
14765 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
14766 TARGET_MUL_SLOW): Define.
14767 * config/nds32/nds32.md (pipeline_model): New attribute.
14768 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
14769 New options that support cpu n9.
14770 * config/nds32/pipelines.md: Include n9 settings.
14771 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
14773 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
14775 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
14776 information if necessary.
14777 (output_cond_branch_compare_zero): Likewise.
14778 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
14779 (nds32_target_alignment): Refine for alignment.
14780 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
14781 (FUNCTION_BOUNDARY): Modify.
14782 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
14784 * config/nds32/nds32.opt (malways-align, malign-functions): New.
14786 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
14788 * config/nds32/constants.md (unspec_volatile_element): Add values for
14789 TLB operation and data prefetch.
14790 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
14791 functions for TLB operation and data prefetch.
14792 * config/nds32/nds32-intrinsic.md: Likewise.
14793 * config/nds32/nds32_intrinsic.h: Likewise.
14794 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
14795 (nds32_print_operand): Likewise.
14796 * config/nds32/nds32.h (nds32_builtins): Likewise.
14798 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
14799 Andrew Pinski <pinsika@gcc.gnu.org>
14801 PR middle-end/82976
14802 * match.pd: Use constant_boolean_node of correct type instead of
14803 boolean_true_node or boolean_false_node for simplifying
14804 pointer comparisons to zero.
14806 2018-04-07 Jakub Jelinek <jakub@redhat.com>
14808 PR tree-optimization/80021
14809 * tree.c (verify_type_variant): Make error call in verify_variant_match
14810 translatable and remove final full stop.
14812 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
14814 * config/nds32/constants.md (unspec_volatile_element): Add
14815 UNSPEC_VOLATILE_EH_RETURN.
14816 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
14817 nds32_output_stack_pop): Support dwarf exception handling process.
14818 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
14819 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
14820 exception handling process.
14821 (nds32_compute_stack_frame): Likewise.
14822 (nds32_return_addr_rtx): Likewise.
14823 (nds32_initial_elimination_offset): Likewise.
14824 (nds32_expand_prologue): Likewise.
14825 (nds32_expand_epilogue): Likewise.
14826 (nds32_dynamic_chain_address): New function.
14827 * config/nds32/nds32.h (machine_function): Add fields for dwarf
14828 exception handling.
14829 (DYNAMIC_CHAIN_ADDRESS): Define.
14830 (EH_RETURN_DATA_REGNO): Define.
14831 (EH_RETURN_STACKADJ_RTX): Define.
14832 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
14833 patterns for dwarf exception handling.
14835 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
14837 * config/nds32/nds32.h: Clean up obsolete macros.
14839 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
14841 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
14842 Add enum values for particular instructions.
14843 * config/nds32/nds32-intrinsic.c: Implementation of expanding
14844 particular intrinsic functions.
14845 * config/nds32/nds32-intrinsic.md: Likewise.
14846 * config/nds32/nds32_intrinsic.h: Likewise.
14847 * config/nds32/nds32.h (nds32_builtins): Likewise.
14848 * config/nds32/nds32.md (type): Add pbsad and pbsada.
14849 (btst, ave): New patterns for particular instructions.
14851 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
14853 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
14854 Add enum values for atomic load/store and memory sync.
14855 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
14857 * config/nds32/nds32-intrinsic.md: Likewise.
14858 * config/nds32/nds32_intrinsic.h: Likewise.
14859 * config/nds32/nds32.h (nds32_builtins): Likewise.
14861 2018-04-07 Jakub Jelinek <jakub@redhat.com>
14863 PR tree-optimization/85257
14864 * fold-const.c (native_encode_vector): If not all elts could fit
14865 and off is -1, return 0 rather than offset.
14866 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
14867 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
14868 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
14869 adjust buffer in native_interpret_expr call.
14871 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
14873 * config/nds32/constants.md (unspec_volatile_element): Add cache
14874 control enum values.
14875 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
14876 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
14877 * config/nds32/nds32.c (nds32_cctl_names): New.
14878 (nds32_print_operand): Handle cache control register names.
14879 * config/nds32/nds32.h (nds32_builtins): New enum values.
14880 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
14882 * config/nds32/nds32.md (type): Add mmu.
14883 * config/nds32/pipelines.md (simple_insn): Add mmu.
14885 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
14887 * config/nds32/nds32.md (type): Remove call.
14888 * config/nds32/pipelines.md (simple_insn): Likewise.
14890 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
14892 * config/nds32/constants.md (unspec_volatile_element): Add
14893 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
14894 UNSPEC_VOLATILE_FMFCFG.
14895 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
14896 description for fmfcfg and fmfcsr.
14897 (bdesc_1arg): Add fmtcsr.
14898 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
14899 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
14900 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
14901 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
14902 unspec_fmfcfg): New patterns.
14903 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
14904 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
14905 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
14906 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
14907 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
14908 __nds32__fmfcfg): Define.
14910 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
14912 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
14913 intrinsic register names.
14914 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
14915 intrinsic register enum values and macros.
14917 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
14919 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
14920 for load/store addressing form.
14921 (nds32_print_operand_address): Likewise.
14923 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
14926 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
14927 based on LABEL_REF. Remove useless assertion.
14928 (pic_address_needs_scratch): Fix formatting.
14929 (sparc_legitimize_pic_address): Minor tweaks.
14930 (sparc_delegitimize_address): Adjust assertion accordingly.
14931 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
14932 into symbolic_operand.
14933 (movsi_high_pic_label_ref): Likewise.
14934 (movsi_lo_sum_pic_label_ref): Likewise.
14935 (movdi_pic_label_ref): Likewise.
14936 (movdi_high_pic_label_ref): Likewise.
14937 (movdi_lo_sum_pic_label_ref): Likewise.
14939 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
14941 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
14942 custom LIB_SPEC setup.
14944 2018-04-06 Ruslan Bukin <br@bsdpad.com>
14945 Kito Cheng <kito.cheng@gmail.com>
14947 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
14948 * config/riscv/freebsd.h: New.
14950 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
14952 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
14953 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
14956 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
14957 Kito Cheng <kito.cheng@gmail.com>
14959 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
14960 nds32_output_call, nds32_symbol_binds_local_p): New functions.
14961 * config/nds32/nds32-protos.h (nds32_output_call,
14962 nds32_output_return): Declare.
14963 * config/nds32/nds32.md: Refine all the call and return patterns.
14965 2018-04-06 Jakub Jelinek <jakub@redhat.com>
14968 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
14969 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
14971 PR rtl-optimization/84872
14972 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
14973 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
14974 EDGE_CROSSING edge.
14976 2018-04-06 Tamar Christina <tamar.christina@arm.com>
14978 * expr.c (copy_blkmode_to_reg): Revert 254862.
14979 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
14981 2018-04-06 Richard Biener <rguenther@suse.de>
14983 PR middle-end/85244
14984 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
14985 after seeing a component reference with an adjacent field. Treat
14986 refs to arrays at struct end of external decls similar to
14987 refs to unconstrained commons.
14989 2018-04-06 Jakub Jelinek <jakub@redhat.com>
14992 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
14993 look through SAVE_EXPRs with non-side-effects argument. Adjust
14995 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
14998 2018-04-06 Richard Biener <rguenther@suse.de>
15000 PR middle-end/85180
15001 * alias.c (find_base_term): New wrapper around find_base_term
15002 unwinding CSELIB_VAL_PTR changes.
15003 (find_base_term): Do not restore CSELIB_VAL_PTR during the
15006 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
15008 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
15010 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
15011 constant definitions.
15012 ("nop"): lr 0,0 -> nopr r0
15013 ("nop_lr0", "nop_lr1"): New insn definitions.
15015 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
15017 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
15018 NDS32_V3PUSH_AVAILABLE_P macro.
15020 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
15021 Chung-Ju Wu <jasonwucj@gmail.com>
15023 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
15024 (nds32*-*-*): Add float and fpu_config into supported_defaults.
15025 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
15026 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
15027 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
15028 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
15029 * config/nds32/constraints.md: New constraints and checking for hard
15030 float configuration.
15031 * config/nds32/iterators.md: New mode iterator and attribute for hard
15032 float configuration.
15033 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
15035 * config/nds32/nds32-fpu.md: New file.
15036 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
15037 deal with hard float code generation.
15038 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
15040 (abi_type, float_reg_number): New enum type.
15041 * config/nds32/nds32-predicates.c: New predicates for hard float.
15042 * config/nds32/nds32-protos.h: Declare functions for hard float.
15043 * config/nds32/nds32.c: Implementation for hard float configuration.
15044 * config/nds32/nds32.h: Definitions for hard float configuration.
15045 * config/nds32/nds32.md: Include hard float machine description and
15046 modify patterns for hard float configuration.
15047 * config/nds32/nds32.opt: New options for hard float configuration.
15048 * config/nds32/predicates.md: New predicates for hard float
15051 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
15053 * common/config/nds32/nds32-common.c
15054 (nds32_option_optimization_table): Enable -mreleax-hint by default.
15056 2018-04-05 Jakub Jelinek <jakub@redhat.com>
15058 PR middle-end/85195
15059 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
15060 CONSTRUCTOR_ELT (ctor, ...)->value.
15062 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
15065 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
15067 2018-04-05 Tom de Vries <tom@codesourcery.com>
15070 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
15073 2018-04-05 Shiva Chen <shiva0217@gmail.com>
15074 Kito Cheng <kito.cheng@gmail.com>
15076 * config/nds32/constraints.md (U33): Fine-tune checking condition.
15077 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
15078 * config/nds32/nds32.h (nds32_16bit_address_type): Add
15079 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
15081 2018-04-05 Shiva Chen <shiva0217@gmail.com>
15082 Kito Cheng <kito.cheng@gmail.com>
15084 * config/nds32/constraints.md (Ufe): New memory constraint.
15085 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
15086 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
15087 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
15089 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
15090 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
15092 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
15094 * config/nds32/nds32.md: Use optimize_size in the condition for
15095 alu-shift instructions.
15097 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
15099 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
15101 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
15103 * config/nds32/nds32.md (negsi2): Refine pattern.
15105 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
15106 Chung-Ju Wu <jasonwucj@gmail.com>
15108 * config/nds32/iterators.md (shift_rotate): New code iterator.
15109 (shift): New code attribute.
15110 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
15111 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
15112 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
15113 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
15114 bit-wise operations.
15115 (andsi3, *andsi3): Ditto.
15116 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
15117 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
15118 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
15119 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
15120 nds32_ior_operand, nds32_xor_operand): New predicates.
15122 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
15124 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
15125 (addsi3, subsi3): ... this.
15127 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
15129 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
15131 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
15133 * config/nds32/nds32.md: Adjust indention.
15135 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
15137 * config/nds32/nds32.md (feature): New attribute.
15139 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
15141 * config/nds32/nds32.md (subtype): New attribute.
15143 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
15146 * config/arm/arm-builtins.c (arm_expand_builtin): Change
15147 expansion to perform a bitwise AND of the argument followed by a
15148 boolean negation of the result.
15150 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
15152 PR rtl-optimization/84878
15153 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
15154 the basic block. Assert the use reference is not artificial and that
15155 it has an associated insn.
15157 2018-04-04 Michael Matz <matz@suse.de>
15159 * builtins.c (compute_objsize): Pass correct operand
15160 to array_at_struct_end_p.
15162 2018-04-04 Richard Biener <rguenther@suse.de>
15165 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
15166 from contexts for DINFO_LEVEL_TERSE and below.
15168 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
15170 * config/nds32/nds32-doubleword.md (move_<mode>): Require
15171 resiter_operand condition.
15172 * config/nds32/nds32.md (*move<mode>): Ditto.
15174 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
15175 Monk Chiang <sh.chiang04@gmail.com>
15177 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
15179 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
15181 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
15183 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
15184 Kito Cheng <kito.cheng@gmail.com>
15186 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
15187 nds32_cond_code_str, output_cond_branch,
15188 output_cond_branch_compare_zero, nds32_expand_cbranch,
15189 nds32_expand_cstore, nds32_expand_movcc,
15190 nds32_output_cbranchsi4_equality_zero,
15191 nds32_output_cbranchsi4_equality_reg,
15192 nds32_output_cbranchsi4_equality_reg_or_const_int,
15193 nds32_output_cbranchsi4_greater_less_zero: New functions.
15194 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
15195 nds32_expand_cstore, nds32_expand_movcc,
15196 nds32_output_cbranchsi4_equality_zero,
15197 nds32_output_cbranchsi4_equality_reg,
15198 nds32_output_cbranchsi4_equality_reg_or_const_int,
15199 nds32_output_cbranchsi4_greater_less_zero): Declare.
15200 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
15201 nds32_rimm11s_operand): New predicates.
15202 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
15203 * config/nds32/nds32.md: Rewrite all the branch and conditional move
15206 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
15208 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
15209 * config/nds32/nds32.md: Ditto.
15210 * config/nds32/pipelines.md: Ditto.
15212 2018-04-04 Richard Biener <rguenther@suse.de>
15214 PR tree-optimization/85168
15215 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
15216 propagating abnormals.
15218 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
15220 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
15222 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
15223 Kito Cheng <kito.cheng@gmail.com>
15225 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
15226 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
15227 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
15228 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
15229 * config/nds32/nds32.md (sibcall_internal): New.
15230 (sibcall_register): Remove.
15231 (sibcall_immediate): Remove.
15232 (sibcall_value_internal): New.
15233 (sibcall_value_register): Remove.
15234 (sibcall_value_immediate): Remove.
15235 * config/nds32/predicates.md (nds32_general_register_operand): New.
15236 (nds32_call_address_operand): New.
15238 2018-04-03 Jakub Jelinek <jakub@redhat.com>
15240 PR rtl-optimization/85167
15241 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
15242 bb_defs if *split_p, instead preinitialize it to NULL.
15244 PR tree-optimization/85156
15245 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
15246 evaluating the argument multiple times.
15248 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
15250 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
15252 (_mm_cvtpd_ps): Likewise.
15253 (_mm_cvttpd_epi32): Likewise.
15254 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
15255 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
15256 vector, pixel, and bool following altivec.h include.
15258 2018-04-03 Martin Sebor <msebor@redhat.com>
15260 * doc/extend.texi (Common Function Attributes): Clarify.
15261 (const attribute): Likewise.
15262 (pure attribute): Likewise.
15264 2018-04-03 Jakub Jelinek <jakub@redhat.com>
15267 * config/i386/i386.c (ix86_expand_vector_set): Use
15268 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
15270 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
15272 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
15273 instructions when changing rounding bits to preserve precision bits
15274 in the x87 control word.
15276 2018-04-03 Martin Liska <mliska@suse.cz>
15278 PR tree-optimization/82491
15279 * rtl.h (strip_offset_and_add): Replace += suboffset with
15280 poly_uint64 () + suboffset.
15282 2018-03-29 Martin Liska <mliska@suse.cz>
15283 Martin Jambor <mjambor@suse.cz>
15286 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
15287 param_type is not an integral or pointer type.
15289 2018-04-03 Richard Biener <rguenther@suse.de>
15291 * sese.h (recompute_all_dominators): Remove.
15293 2018-04-02 Martin Sebor <msebor@redhat.com>
15295 * doc/invoke.texi (-Wrestrict): Fix typos.
15297 2018-04-02 Jim Wilson <jimw@sifive.com>
15299 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
15300 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
15301 (<optab>di3, <optab>si3_extend): Likewise.
15302 (<optab>si3_mask, <optab>si3_mask_1): New.
15303 (<optab>di3_mask, <optab>di3_mask_1): New.
15304 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
15305 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
15306 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
15308 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
15310 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
15313 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
15315 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
15316 (nds32_canonicalize_comparison): New function.
15318 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
15319 Kito Cheng <kito.cheng@gmail.com>
15320 Kuan-Lin Chen <kuanlinchentw@gmail.com>
15322 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
15323 * config/nds32/constants.md (unspec_volatile_element): Add
15324 UNSPEC_VOLATILE_RELAX_GROUP.
15325 * config/nds32/nds32-relax-opt.c: New file.
15326 * config/nds32/nds32-predicates.c
15327 (nds32_symbol_load_store_p): New function.
15328 * config/nds32/nds32-protos.h
15329 (nds32_symbol_load_store_p): Declare function.
15330 (make_pass_nds32_relax_opt): Declare new rtl pass function.
15331 * config/nds32/nds32.c
15332 (nds32_register_pass): New function to register pass.
15333 (nds32_register_passes): New function to register passes.
15334 * config/nds32/nds32.md (relax_group): New pattern.
15335 * config/nds32/nds32.opt (mrelax-hint): New option.
15336 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
15338 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
15340 * config/nds32/t-nds32: Modify files dependency.
15342 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
15344 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
15345 (PROFILE_HOOK): Define its implementation.
15347 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
15349 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
15350 type and 32-bit size.
15352 2018-04-01 Jakub Jelinek <jakub@redhat.com>
15354 PR middle-end/85090
15355 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
15356 (V_128_256): New mode iterator.
15357 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
15358 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
15359 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
15361 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
15362 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
15364 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
15367 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
15368 NaN inputs correctly.
15370 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
15373 * config/rs6000/vsx.md (??r): New mode attribute.
15374 (*vsx_mov<mode>_64bit): Use it.
15375 (*vsx_mov<mode>_32bit): Likewise.
15377 2018-03-30 Martin Sebor <msebor@redhat.com>
15379 PR tree-optimization/84818
15380 * builtins.c (check_access): Use warning_n.
15382 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15385 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
15387 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
15390 2018-03-30 Julia Koval <julia.koval@intel.com>
15393 * x86-tune.def (movx, partial_reg_dependency): Enable for
15396 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
15398 PR inline-asm/84985
15399 * lra-constraints.c (process_alt_operands): Move setting
15400 this_alternative_matches below.
15402 2018-03-29 Martin Liska <mliska@suse.cz>
15405 * doc/invoke.texi: Document how LTO works with debug info.
15406 Describe auto-load support of binutils. Mention 'x86-64'
15407 as valid option value of -march option.
15409 2018-03-29 Jakub Jelinek <jakub@redhat.com>
15411 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
15414 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
15415 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
15416 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
15419 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
15422 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
15423 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
15424 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
15425 for RS6000_BTM_POWERPC64.
15426 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
15427 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
15428 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
15431 (DIVDEU): Likewise.
15433 2018-03-28 Carl Love <cel@us.ibm.com>
15436 2017-09-27 Carl Love <cel@us.ibm.com>
15438 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
15439 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
15440 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
15443 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15445 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
15446 instead of __vector bool.
15447 (_mm_max_pu8): Likewise.
15448 (_mm_min_pi16): Likewise.
15450 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
15453 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
15454 (DIVWEUO): Likewise.
15455 (DIVDEO): Likewise.
15456 (DIVDEUO): Likewise.
15457 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
15458 DIVWEUO and DIVDEUO.
15459 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
15460 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
15461 (div_extend): Likewise.
15462 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
15464 (__builtin_divweuo): Likewise.
15465 (__builtin_divdeo): Likewise.
15466 (__builtin_divdeuo): Likewise.
15468 2018-03-28 Jakub Jelinek <jakub@redhat.com>
15471 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
15472 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
15474 PR tree-optimization/82004
15475 * gimple-match-head.c (optimize_pow_to_exp): New function.
15476 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
15477 Don't fold to exp if optimize_pow_to_exp is false.
15479 2018-03-28 Martin Liska <mliska@suse.cz>
15482 * calls.c (initialize_argument_information): Fix trailing space.
15483 * common.opt: Fix typo and provide better explanation for
15484 -fsanitize-coverage option.
15485 * config/i386/i386.opt: Fix typo.
15487 2018-03-28 Jakub Jelinek <jakub@redhat.com>
15488 Martin Liska <mliska@suse.cz>
15491 * gimplify.c (asan_poison_variable): Don't do the check for
15492 gimplify_omp_ctxp here.
15493 (gimplify_decl_expr): Do it here.
15494 (gimplify_target_expr): Likewise.
15496 2018-03-28 Martin Liska <mliska@suse.cz>
15499 * config/i386/i386.c (ix86_function_arg_advance): Do not call
15500 chkp_type_bounds_count if MPX is not enabled.
15502 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
15504 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
15506 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
15509 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
15510 function to create the function decl for complex long double
15511 multiply and divide for -mabi=ieeelongdouble.
15512 (init_float128_ieee): Call it.
15514 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
15517 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
15518 -fcf-protection=branch -mibt.
15519 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
15521 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
15524 * config/arm/arm.c (arm_valid_symbolic_address): Handle
15525 arm_word_relocations.
15527 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
15530 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
15531 extern array declarations.
15533 2018-03-27 Richard Biener <rguenther@suse.de>
15535 PR middle-end/84067
15536 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
15537 explicit single_use checks.
15539 2018-03-27 Richard Biener <rguenther@suse.de>
15541 PR tree-optimization/85082
15542 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
15545 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
15547 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
15548 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
15549 Turn on fasynchronous-unwind-tables and funwind-tables.
15551 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
15554 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
15555 (*bmi_blsr_<mode>_ccz): Ditto.
15557 2018-03-26 Tom de Vries <tom@codesourcery.com>
15559 PR tree-optimization/85063
15560 * omp-general.c (offloading_function_p): New function. Factor out
15562 * omp-offload.c (pass_omp_target_link::gate): ... here.
15563 * omp-general.h (offloading_function_p): Declare.
15564 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
15565 with attribute omp declare target for offloading functions.
15567 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
15569 PR tree-optimization/84005
15570 * tree-data-ref.h (get_base_for_alignment): Declare.
15571 * tree-data-ref.c (get_base_for_alignment_1): New function.
15572 (get_base_for_alignment): Likewise.
15573 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
15574 get_base_for_alignment to find a suitable base object, instead
15575 of always using drb->base_address.
15577 2018-03-23 Jakub Jelinek <jakub@redhat.com>
15579 PR inline-asm/85022
15580 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
15581 known size by default.
15583 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
15585 PR inline-asm/85030
15586 * lra-constraints.c (process_alt_operands): Don't match BLKmode
15587 and non BLKmode operands.
15589 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15592 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
15593 Clean up attributes.
15595 2018-03-23 Richard Biener <rguenther@suse.de>
15598 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
15599 we are going to emit early debug for LTO.
15601 2018-03-23 Jakub Jelinek <jakub@redhat.com>
15603 PR inline-asm/85034
15604 * function.c (match_asm_constraints_1): Don't optimize if input
15605 doesn't satisfy general_operand predicate for output's mode.
15607 PR inline-asm/85022
15608 * alias.c (write_dependence_p): Don't require for x_canonicalized
15609 non-VOIDmode if x has VOIDmode.
15612 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
15613 just don't try to optimize it rather than assert it never happens.
15615 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
15617 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
15618 macro expansions for definition of ST_INTERNAL_<mode> and
15619 LD_INTERNAL_<mode> builtins.
15620 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
15622 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
15624 (altivec_expand_st_builtin): Likewise.
15625 (altivec_expand_builtin): Remove calls to deleted functions.
15626 (rs6000_address_for_altivec): Delete this function.
15627 * config/rs6000/vector.md: Remove expands for
15628 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
15630 2018-03-22 Sudakshina Das <sudi.das@arm.com>
15633 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
15634 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
15635 re-computing once computed.
15636 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
15637 (arm_init_machine_status): Initialize
15638 machine->static_chain_stack_bytes.
15640 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
15643 * doc/extend.texi: Add four new prototypes for vec_ld.
15644 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
15645 definitions for more logical presentation.
15646 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
15647 entries for V1TI variants of __builtin_altivec_ld builtin.
15648 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
15649 handling of V1TI variant of LVX icode pattern.
15650 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
15651 (rs6000_gimple_fold_builtin): Likewise.
15652 (altivec_init_builtins): Add code to define
15653 __builtin_altivec_lvx_v1ti function.
15655 2018-03-22 Jakub Jelinek <jakub@redhat.com>
15657 PR inline-asm/84941
15658 * function.c (match_asm_constraints_1): Don't do the optimization
15659 if input isn't a REG, SUBREG, MEM or constant.
15661 2018-03-22 Tom de Vries <tom@codesourcery.com>
15663 PR tree-optimization/84956
15664 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
15665 bb_has_abnormal_pred.
15667 2018-03-22 Jakub Jelinek <jakub@redhat.com>
15670 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
15671 DECL_INITIAL (decl) to decl at the end.
15672 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
15673 adjust the comment.
15675 2018-03-21 Joseph Myers <joseph@codesourcery.com>
15677 * doc/extend.texi (__builtin_tgmath): Document when complex
15678 integer types are treated as _Complex _Float64.
15680 2018-03-21 Tom de Vries <tom@codesourcery.com>
15682 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
15684 2018-03-21 Jakub Jelinek <jakub@redhat.com>
15686 PR tree-optimization/84960
15687 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
15688 if it is ENTRY block, move them into single succ of ENTRY in that case.
15690 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
15692 PR tree-optimization/84811
15693 * poly-int.h (poly_span_traits): Remove the T3 parameter and
15694 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
15695 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
15696 (known_subrange_p): Update accordingly. Cast each value involved
15697 in the size comparison, rather than casting the result of the
15700 2018-03-21 Jakub Jelinek <jakub@redhat.com>
15702 PR tree-optimization/84982
15703 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
15704 by flipping the least significant bit rather than all bits from
15705 bitpos to bitpos + bitsize - 1.
15707 2018-03-21 Nathan Sidwell <nathan@acm.org>
15709 * doc/extend.texi (Deprecated Features): Remove mention of
15710 long-deleted deprecations.
15712 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15715 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
15716 * configure: Regenerate.
15718 2018-03-21 Tom de Vries <tom@codesourcery.com>
15720 PR tree-optimization/83126
15721 * tree-parloops.c (num_phis): New function.
15722 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
15724 2018-03-21 Nathan Sidwell <nathan@acm.org>
15726 * doc/extend.texi (Deprecated Features): Update deprecated flags,
15727 mention anon-struct/union members and trailing attributes.
15729 2018-03-21 Bin Cheng <bin.cheng@arm.com>
15731 PR tree-optimization/84969
15732 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
15733 builtin memset partitions if they set different rhs values.
15735 2018-03-21 Jakub Jelinek <jakub@redhat.com>
15737 PR rtl-optimization/84989
15738 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
15739 VEC_DUPLICATE with scalar result mode.
15741 2018-03-21 Martin Liska <mliska@suse.cz>
15744 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
15745 not intended return statement.
15747 2018-03-21 Martin Liska <mliska@suse.cz>
15750 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
15751 (chkp_find_bound_slots_1): Limit number of iterations.
15753 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
15756 * Minor grammar fixes for x86 options.
15758 2018-03-20 Jakub Jelinek <jakub@redhat.com>
15761 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
15762 holding REG_CFA_RESTORE notes, instead turn them into a USE.
15764 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
15767 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
15768 (altivec_lvx_<mode>_1op): Likewise.
15769 (altivec_stvx_<mode>_2op): Likewise.
15770 (altivec_stvx_<mode>_1op): Likewise.
15771 (altivec_lvx_<VM2:mode>): New define_expand.
15772 (altivec_stvx_<VM2:mode>): Likewise.
15773 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
15774 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
15775 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
15776 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
15777 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
15778 (rs6000_gen_lvx): Likewise.
15779 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
15780 (altivec_expand_stv_builtin): Likewise.
15781 (altivec_expand_builtin): Likewise.
15782 * config/rs6000/vector.md: Likewise.
15784 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15787 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
15790 2018-03-20 Richard Biener <rguenther@suse.de>
15793 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
15794 sign-conversions as zero, fall back to standard scalar_stmt
15797 2018-03-20 Martin Liska <mliska@suse.cz>
15800 * predict.c (rebuild_frequencies): Handle case when we have
15801 PROFILE_ABSENT, but flag_guess_branch_prob is false.
15803 2018-03-20 Jakub Jelinek <jakub@redhat.com>
15806 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
15807 flag_section_anchors.
15808 * varasm.c (use_blocks_for_decl_p): Remove hack for
15809 dw2_force_const_mem.
15812 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
15814 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
15815 be created, use lowpart_subreg of operands[0] rather than operands[0]
15817 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
15818 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
15819 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
15820 and n constraint instead of aarch64_shift_imm_di and Usd.
15821 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
15822 (*aarch64_<optab>_reg_minus<mode>3): ... this.
15824 2018-03-20 Sudakshina Das <sudi.das@arm.com>
15827 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
15828 to favor GPR over NEON registers.
15829 (<shift>di3_neon): Likewise.
15831 2018-03-20 Tom de Vries <tom@codesourcery.com>
15834 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
15835 (nvptx_process_pars): Emit bar.sync asap and alap.
15837 2018-03-20 Tom de Vries <tom@codesourcery.com>
15840 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
15841 seen_label if seen_label is already set.
15843 2018-03-20 Jakub Jelinek <jakub@redhat.com>
15846 * config/i386/i386.c (fold_builtin_cpu): For features above 31
15847 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
15848 Use 1U instead of 1. Formatting fixes.
15851 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
15852 instead of TREE_TYPE (s1) for the return value.
15854 2018-03-19 Jakub Jelinek <jakub@redhat.com>
15856 PR tree-optimization/84946
15857 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
15858 bitsize + bitsize in poly_uint64 rather than poly_int64.
15861 * dwarf2asm.c: Include fold-const.c.
15862 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
15863 of decl rather than decl itself.
15865 PR rtl-optimization/84643
15866 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
15868 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
15871 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
15872 calling assemble_variable.
15874 2018-03-19 Sudakshina Das <sudi.das@arm.com>
15877 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
15878 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
15880 2018-03-19 Jim Wilson <jimw@sifive.com>
15883 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
15884 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
15885 (riscv_first_stack_step): Likewise.
15886 (riscv_option_override): Use STACK_BOUNDARY instead of
15887 MIN_STACK_BOUNDARY.
15888 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
15889 MIN_STACK_BOUNDARY.
15890 (BIGGEST_ALIGNMENT): Set to 128.
15891 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
15892 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
15895 2018-03-19 Richard Biener <rguenther@suse.de>
15897 PR tree-optimization/84933
15898 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
15899 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
15901 2018-03-19 Richard Biener <rguenther@suse.de>
15903 PR tree-optimization/84859
15904 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
15905 (cond_if_else_store_replacement): Perform sinking operation on
15906 single-store BBs regardless of MAX_STORES_TO_SINK setting.
15907 Generalize what a BB with a single eligible store is.
15909 2018-03-19 Richard Biener <rguenther@suse.de>
15911 PR tree-optimization/84929
15912 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
15913 chrec_is_positive against non-chrec arg.
15915 2018-03-19 Tamar Christina <tamar.christina@arm.com>
15918 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
15920 2018-03-18 Martin Liska <mliska@suse.cz>
15922 PR rtl-optimization/84635
15923 * regrename.c (build_def_use): Use matches_mode only when
15926 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
15928 PR tree-optimization/84913
15929 * tree-vect-loop.c (vectorizable_reduction): Don't try to
15930 vectorize chains of COND_EXPRs.
15932 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
15934 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
15936 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
15938 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
15940 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
15942 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
15944 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
15945 Kito Cheng <kito.cheng@gmail.com>
15947 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
15948 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
15949 (nds32_adjust_reg_alloc_order): New function.
15950 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
15952 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
15954 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
15955 nds32_print_operand, nds32_print_operand_address): Use
15956 HOST_WIDE_INT_PRINT_DEC instead.
15958 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
15960 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
15962 2018-03-17 Jakub Jelinek <jakub@redhat.com>
15965 * config/i386/i386.c (initial_ix86_tune_features,
15966 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
15967 unsigned long long.
15968 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
15969 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
15970 rather than 1u << ix86_tune. Formatting fix.
15971 (ix86_option_override_internal): Change ix86_arch_mask from
15972 unsigned int to unsigned HOST_WIDE_INT, initialize to
15973 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
15974 (ix86_function_specific_restore): Likewise.
15976 2018-03-16 Jakub Jelinek <jakub@redhat.com>
15979 * postreload.c (reload_combine_recognize_pattern): Perform
15980 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
15981 truncate_int_for_mode the result for the destination's mode.
15984 * hsa-gen.c (mem_type_for_type): Fix comment typo.
15985 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
15987 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
15990 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
15993 * lra-assigns.c (lra_split_hard_reg_for): Don't use
15994 regno_allocno_class_array and sorted_pseudos.
15995 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
15996 insns where regno is used.
15998 2018-03-16 Martin Liska <mliska@suse.cz>
16001 * multiple_target.c (create_dispatcher_calls): Redirect
16002 reference in the symbol table.
16004 2018-03-16 Martin Liska <mliska@suse.cz>
16007 * multiple_target.c (create_dispatcher_calls): Redirect also
16010 2018-03-16 Jakub Jelinek <jakub@redhat.com>
16014 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
16015 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
16016 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
16018 2018-03-16 Julia Koval <julia.koval@intel.com>
16020 * doc/invoke.texi (Skylake Server): Add CLWB.
16021 Cannonlake): Remove CLWB.
16023 2018-03-16 Jakub Jelinek <jakub@redhat.com>
16025 PR tree-optimization/84841
16026 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
16028 (FLOAT_ONE_CONST_TYPE): Define.
16029 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
16030 (sort_by_operand_rank): Put entries with higher constant_type last
16031 rather than first to match comments.
16033 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
16035 * config/nios2/nios2.md (movsi_internal): Fix thinko in
16038 2018-03-15 Jakub Jelinek <jakub@redhat.com>
16041 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
16042 check and use address of target always.
16044 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
16047 * config/i386/i386.c (indirect_thunk_needed): Update comments.
16048 (indirect_thunk_bnd_needed): Likewise.
16049 (indirect_thunks_used): Likewise.
16050 (indirect_thunks_bnd_used): Likewise.
16051 (indirect_return_needed): New.
16052 (indirect_return_bnd_needed): Likewise.
16053 (output_indirect_thunk_function): Add a bool argument for
16055 (output_indirect_thunk_function): Don't generate alias for
16056 function return thunk.
16057 (ix86_code_end): Call output_indirect_thunk_function to generate
16058 function return thunks.
16059 (ix86_output_function_return): Set indirect_return_bnd_needed
16060 and indirect_return_needed instead of indirect_thunk_bnd_needed
16061 and indirect_thunk_needed.
16063 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
16065 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
16066 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
16067 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
16069 2018-03-15 David Malcolm <dmalcolm@redhat.com>
16070 Paul Hua <paul.hua.gm@gmail.com>
16073 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
16075 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
16077 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
16078 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
16079 resp. SFmode cases.
16081 2018-03-15 Tamar Christina <tamar.christina@arm.com>
16084 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
16085 instead of GET_MODE_SIZE when comparing Units.
16087 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
16090 * varasm.c (hash_section): Return an unchangeble hash value
16091 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
16092 Return !aarch64_can_use_per_function_literal_pools_p ().
16094 2018-03-15 Jakub Jelinek <jakub@redhat.com>
16097 * optabs.c (emit_conditional_move): Pass address of cmode's copy
16098 rather than address of cmode as last argument to prepare_cmp_insn.
16100 2018-03-15 Julia Koval <julia.koval@intel.com>
16102 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
16103 F_AVX512VNNI, F_AVX512BITALG): New.
16105 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
16108 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
16109 insn for floating-point loads and stores.
16111 2018-03-14 Carl Love <cel@us.ibm.com>
16113 * config/rs6000/rs6000-c.c: Add macro definitions for
16114 ALTIVEC_BUILTIN_VEC_PERMXOR.
16115 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
16116 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
16117 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
16118 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
16120 * config/doc/extend.texi: Add prototypes for vec_permxor.
16122 2018-03-14 David Malcolm <dmalcolm@redhat.com>
16125 * diagnostic-show-locus.c (class layout_point): Convert m_line
16126 from int to linenum_type.
16127 (line_span::comparator): Use linenum "compare" function when
16128 comparing line numbers.
16129 (test_line_span): New function.
16130 (layout_range::contains_point): Convert param "row" from int to
16132 (layout_range::intersects_line_p): Likewise.
16133 (layout::will_show_line_p): Likewise.
16134 (layout::print_source_line): Likewise.
16135 (layout::should_print_annotation_line_p): Likewise.
16136 (layout::print_annotation_line): Likewise.
16137 (layout::print_leading_fixits): Likewise.
16138 (layout::annotation_line_showed_range_p): Likewise.
16139 (struct line_corrections): Likewise for field m_row.
16140 (line_corrections::line_corrections): Likewise for param "row".
16141 (layout::print_trailing_fixits): Likewise.
16142 (layout::get_state_at_point): Likewise.
16143 (layout::get_x_bound_for_row): Likewise.
16144 (layout::print_line): Likewise.
16145 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
16146 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
16147 * input.c (selftest::test_linenum_comparisons): New function.
16148 (selftest::input_c_tests): Call it.
16149 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
16150 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
16151 * selftest.h (ASSERT_GT): New macro.
16152 (ASSERT_GT_AT): New macro.
16153 (ASSERT_LT): New macro.
16154 (ASSERT_LT_AT): New macro.
16156 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
16158 PR rtl-optimization/84780
16159 * combine.c (distribute_links): Don't make a link based on pc_rtx.
16161 2018-03-14 Martin Liska <mliska@suse.cz>
16163 * tree.c (record_node_allocation_statistics): Use
16164 get_stats_node_kind.
16165 (get_stats_node_kind): New function extracted from
16166 record_node_allocation_statistics.
16167 (free_node): Use get_stats_node_kind.
16169 2018-03-14 Richard Biener <rguenther@suse.de>
16171 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
16172 that the value-set of ANTIC_IN doesn't grow.
16175 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
16177 (BB_VISITED_WITH_VISITED_SUCCS): New define.
16178 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
16180 2018-03-14 Julia Koval <julia.koval@intel.com>
16182 * config.gcc (icelake-client, icelake-server): New.
16184 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
16185 (initial_ix86_arch_features): Ditto.
16186 (PTA_SKYLAKE): Add SGX.
16187 (PTA_ICELAKE): Remove.
16188 (PTA_ICELAKE_CLIENT): New.
16189 (PTA_ICELAKE_SERVER): New.
16190 (ix86_option_override_internal): Split up icelake on icelake client and
16192 (get_builtin_code_for_version): Ditto.
16193 (fold_builtin_cpu): Ditto.
16194 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
16195 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
16196 * config/i386/i386.h (processor_type): Ditto.
16197 * doc/invoke.texi: Ditto.
16199 2018-03-14 Jakub Jelinek <jakub@redhat.com>
16202 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
16203 INTEGER_CST offset, add it together with bitpos / 8 and
16204 sign extend based on POINTER_SIZE.
16208 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
16211 * config/i386/constraints.md (Yc): New register constraint.
16212 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
16213 Use Yc constraint for alternative 2 of operand 0. Remove
16214 preferred_for_speed attribute.
16216 2018-03-14 Richard Biener <rguenther@suse.de>
16218 PR tree-optimization/84830
16219 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
16220 with the old one to avoid oscillations.
16222 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
16225 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
16227 (assign_by_spills): Return a flag of reload assignment failure.
16228 Do not process the reload assignment failures. Do not spill other
16229 reload pseudos if they has the same reg class. Update n if
16231 (lra_assign): Add a return arg. Set up from the result of
16232 assign_by_spills call.
16233 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
16234 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
16235 usage_insns if it is not NULL.
16236 (spill_hard_reg_in_range): New function.
16237 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
16238 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
16239 function prototypes.
16240 (lra_assign): Change prototype.
16241 * lra.c (lra): Add code to deal with fails by splitting hard reg
16244 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
16246 * config/riscv/riscv.opt (mrelax): New option.
16247 * config/riscv/riscv.c (riscv_file_start): Emit ".option
16248 "norelax" when riscv_mrelax is disabled.
16249 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
16251 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16254 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
16255 reassociation for int modes.
16257 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
16259 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
16260 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
16262 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
16263 * config/aarch64/aarch64-sve.md
16264 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
16265 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
16266 (*extend<mode><Vwide>2): Rename to...
16267 (aarch64_sve_extend<mode><Vwide>2): ...this.
16268 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
16269 renaming the old pattern to...
16270 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
16272 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
16273 define_expand, renaming the old pattern to...
16274 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
16275 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
16276 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
16277 account when deciding which SVE instruction the optab should use.
16278 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
16280 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
16282 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
16283 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
16284 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
16285 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
16286 (tlsdesc_small_<mode>): Turn a define_expand and use
16287 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
16288 (tlsdesc_small_advsimd_<mode>): ...this.
16289 (tlsdesc_small_sve_<mode>): New pattern.
16291 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
16293 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
16294 (UNSPEC_UMUL_HIGHPART): New constants.
16295 (MUL_HIGHPART): New int iteraor.
16296 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
16297 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
16299 (*<su>mul<mode>3_highpart): New define_insn.
16301 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
16304 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
16307 2018-03-13 Martin Liska <mliska@suse.cz>
16310 * (sem_item_optimizer::sem_item_optimizer): Initialize new
16312 (sem_item_optimizer::~sem_item_optimizer): Release it.
16313 (sem_item_optimizer::merge_classes): Register variable aliases.
16314 (sem_item_optimizer::fixup_pt_set): New function.
16315 (sem_item_optimizer::fixup_points_to_sets): Likewise.
16316 * ipa-icf.h: Declare new variables and functions.
16318 2018-03-13 Jakub Jelinek <jakub@redhat.com>
16320 PR middle-end/84834
16321 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
16322 integer_pow2p@2 and test integer_pow2p in condition.
16323 (A < 0 ? C : 0): Similarly for @1.
16325 PR middle-end/84831
16326 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
16327 characters starting at p contain '\0' character, don't look beyond
16331 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
16332 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
16335 * reg-stack.c (change_stack): Change update_end var from int to
16336 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
16337 also call set_block_for_insn on the newly added insns and rescan.
16340 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
16341 on the last operand.
16344 * tree.c (stabilize_reference_1): Return save_expr (e) for
16345 STATEMENT_LIST even if it doesn't have side-effects.
16347 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
16349 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
16351 2018-03-12 Renlin Li <renlin.li@arm.com>
16353 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
16354 aarch64_output_scalar_simd_mov_immediate.
16356 2018-03-12 Martin Sebor <msebor@redhat.com>
16358 PR tree-optimization/83456
16359 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
16360 for perfectly overlapping calls to memcpy.
16361 (gimple_fold_builtin_memory_chk): Same.
16362 (gimple_fold_builtin_strcpy): Handle no-warning.
16363 (gimple_fold_builtin_stxcpy_chk): Same.
16364 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
16366 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
16368 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
16369 parameter. Use it for SFmode.
16370 (rs6000_function_arg_advance_1): Adjust.
16371 (rs6000_function_arg): Adjust.
16372 (rs6000_gimplify_va_arg): Pass false for that new parameter.
16374 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
16376 PR rtl-optimization/84169
16377 PR rtl-optimization/84780
16378 * combine.c (can_combine_p): Check for a 2-insn combination whether
16379 the destination register is used between the two insns, too.
16381 2018-03-12 Richard Biener <rguenther@suse.de>
16383 PR tree-optimization/84803
16384 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
16385 for refs DR analysis didn't process.
16387 2018-03-12 Richard Biener <rguenther@suse.de>
16389 PR tree-optimization/84777
16390 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
16391 force-vectorize loops ignore whether we are optimizing for size.
16393 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
16395 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
16396 (TARGET_MD_ASM_ADJUST): Define.
16398 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
16399 Kito Cheng <kito.cheng@gmail.com>
16400 Chung-Ju Wu <jasonwucj@gmail.com>
16402 * config/nds32/nds32.c (nds32_compute_stack_frame,
16403 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
16404 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
16405 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
16406 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
16407 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
16408 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
16409 * config/nds32/nds32.md (prologue, epilogue): Use macro
16410 NDS32_V3PUSH_AVAILABLE_P to do checking.
16412 2018-03-11 Jakub Jelinek <jakub@redhat.com>
16415 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
16416 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
16417 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
16418 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
16419 addition of most attributes on !orig_type_die or the attribute not
16420 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
16422 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
16423 Chung-Ju Wu <jasonwucj@gmail.com>
16425 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
16426 __NDS32_VH__ macro.
16427 * config/nds32/nds32.opt (mvh): New option.
16429 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
16430 Chung-Ju Wu <jasonwucj@gmail.com>
16432 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
16434 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
16435 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
16438 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
16439 Chung-Ju Wu <jasonwucj@gmail.com>
16441 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
16443 * config/nds32/nds32-multiple.md (strlensi): New pattern.
16444 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
16446 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
16447 Kito Cheng <kito.cheng@gmail.com>
16448 Chung-Ju Wu <jasonwucj@gmail.com>
16450 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
16451 UNSPEC_FFMISM and UNSPEC_FLMISM.
16452 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
16453 for ffb, ffmism and flmism.
16454 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
16455 (unspec_ffmism): Ditto.
16456 (unspec_flmism): Ditto.
16457 (nds32_expand_builtin_impl): Check if string extension is available.
16458 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
16459 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
16461 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
16464 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
16467 * lra-assigns.c (assign_by_spills): Return a flag of reload
16468 assignment failure. Do not process the reload assignment
16469 failures. Do not spill other reload pseudos if they has the same
16471 (lra_assign): Add a return arg. Set up from the result of
16472 assign_by_spills call.
16473 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
16474 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
16475 usage_insns if it is not NULL.
16476 (spill_hard_reg_in_range): New function.
16477 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
16478 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
16479 function prototypes.
16480 (lra_assign): Change prototype.
16481 * lra.c (lra): Add code to deal with fails by splitting hard reg
16484 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
16487 * config/i386/i386.opt: Replace Enforcment with Enforcement.
16489 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
16492 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
16493 (dw_val_node): Add val_symbolic_view.
16494 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
16495 (symview_upper_bound): New.
16496 (new_line_info_table): Initialize symviews_since_reset.
16497 (dwarf2out_source_line): Count symviews_since_reset and set
16498 symview_upper_bound.
16499 (dw_val_equal_p): Handle symview.
16500 (add_AT_symview): New.
16501 (print_dw_val): Handle symview.
16502 (attr_checksum, attr_checksum_ordered): Likewise.
16503 (same_dw_val_p, size_of_die): Likewise.
16504 (value_format, output_die): Likewise.
16505 (add_high_low_attributes): Use add_AT_symview for entry_view.
16506 (dwarf2out_finish): Reset symview_upper_bound, clear
16509 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
16512 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
16513 Add strict argument and use it.
16514 (rs6000_split_multireg_move): Update for new strict argument.
16515 (mem_operand_gpr): Disallow all non-offsettable addresses.
16516 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
16518 2018-03-09 Jakub Jelinek <jakub@redhat.com>
16521 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
16522 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
16523 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
16526 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
16527 decl, use remap_type if we want to use the type.
16529 2018-03-09 Martin Sebor <msebor@redhat.com>
16531 PR tree-optimization/84526
16532 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
16534 (builtin_access::generic_overlap): Be prepared to handle non-array
16537 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
16539 PR rtl-optimization/84682
16540 * lra-constraints.c (process_address_1): Check is_address flag
16541 for address constraints.
16542 (process_alt_operands): Likewise.
16543 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
16544 preprocess_constraints.
16545 * recog.h (preprocess_constraints): Add oploc parameter.
16547 * recog.c (preprocess_constraints): Test address_operand for
16548 CT_ADDRESS constraints.
16550 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
16553 * lra-assigns.c (assign_by_spills): Return a flag of reload
16554 assignment failure. Do not process the reload assignment
16555 failures. Do not spill other reload pseudos if they has the same
16557 (lra_assign): Add a return arg. Set up from the result of
16558 assign_by_spills call.
16559 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
16560 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
16561 usage_insns if it is not NULL.
16562 (spill_hard_reg_in_range): New function.
16563 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
16564 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
16565 function prototypes.
16566 (lra_assign): Change prototype.
16567 * lra.c (lra): Add code to deal with fails by splitting hard reg
16570 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16573 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
16574 Accept complain bool parameter. Only emit errors if it is true.
16575 (arm_parse_cpu_option_name): Likewise.
16576 (arm_target_thumb_only): Adjust callers of the above.
16577 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
16578 prototype to take a default true bool parameter.
16579 (arm_parse_arch_option_name): Likewise.
16581 2018-03-09 David Malcolm <dmalcolm@redhat.com>
16582 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
16586 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
16587 * configure: Regenerate.
16588 * configure.ac ("linker --version-script option"): New.
16589 ("linker soname option"): New.
16591 2018-03-09 Richard Biener <rguenther@suse.de>
16593 PR tree-optimization/84775
16594 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
16595 immediate uses of predicate stmts and mark them modified.
16598 PR tree-optimization/84178
16599 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
16601 (version_loop_for_if_conversion): Delay update_ssa call.
16602 (tree_if_conversion): Delay update_ssa until after predicate
16605 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
16608 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
16609 when the function accesses prior frames.
16611 2018-03-08 Jakub Jelinek <jakub@redhat.com>
16614 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
16615 gen_llsym, otherwise call maybe_gen_llsym.
16617 PR inline-asm/84742
16618 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
16619 has ',' character inside of it.
16621 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16624 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
16625 as clobbering CC_REGNUM.
16627 2018-03-08 Richard Biener <rguenther@suse.de>
16629 PR middle-end/84552
16630 * tree-scalar-evolution.c: Include tree-into-ssa.h.
16631 (follow_copies_to_constant): Do not follow SSA names registered
16634 2018-03-08 Richard Biener <rguenther@suse.de>
16636 PR tree-optimization/84178
16637 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
16639 (version_loop_for_if_conversion): Delay update_ssa call.
16640 (tree_if_conversion): Delay update_ssa until after predicate
16643 2018-03-08 David Malcolm <dmalcolm@redhat.com>
16645 PR tree-optimization/84178
16646 * tree-if-conv.c (release_bb_predicate): Remove the
16647 the assertion that the stmts have NULL use_ops.
16648 Discard the statements, asserting that they haven't
16649 yet been added to a BB.
16651 2018-03-08 Richard Biener <rguenther@suse.de>
16653 PR tree-optimization/84746
16654 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
16655 (phi_translate): Pass in destination ANTIC_OUT set.
16656 (phi_translate_1): Likewise. For a simplified result lookup
16657 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
16658 (phi_translate_set): Adjust.
16659 (do_pre_regular_insertion): Likewise.
16660 (do_pre_partial_partial_insertion): Likewise.
16662 2018-03-08 Martin Liska <mliska@suse.cz>
16664 PR gcov-profile/84735
16665 * doc/gcov.texi: Document usage of profile files.
16666 * gcov-io.h: Document changes in the format.
16668 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
16672 * dwarf2out.c (struct dw_line_info_table): Update comments for
16674 (FORCE_RESET_NEXT_VIEW): New.
16675 (FORCE_RESETTING_VIEW_P): New.
16676 (RESETTING_VIEW_P): Check for -1 too.
16677 (ZERO_VIEW_P): Likewise.
16678 (new_line_info_table): Force-reset next view.
16679 (dwarf2out_begin_function): Likewise.
16680 (dwarf2out_source_line): Simplify zero_view_p initialization.
16681 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
16682 view directly. Omit view when omitting .loc at line 0.
16684 2018-03-08 Jakub Jelinek <jakub@redhat.com>
16686 PR tree-optimization/84740
16687 * tree-switch-conversion.c (process_switch): Call build_constructors
16688 only if info.phi_count is non-zero.
16690 PR tree-optimization/84739
16691 * tree-tailcall.c (find_tail_calls): Check call arguments against
16692 DECL_ARGUMENTS (current_function_decl) rather than
16693 DECL_ARGUMENTS (func) when checking for tail recursion.
16695 2018-03-07 Jakub Jelinek <jakub@redhat.com>
16697 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
16698 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
16699 Volker Reichelt's entry and add entries for people that perform
16700 GCC fuzzy testing and report numerous bugs.
16702 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
16705 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
16706 readonly data in sdata, if that is disabled.
16707 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
16708 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
16709 -mreadonly-in-sdata option.
16711 2018-03-07 Martin Sebor <msebor@redhat.com>
16713 PR tree-optimization/84468
16714 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
16715 basic block when looking for nul assignment.
16717 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
16720 * except.h (output_function_exception_table): Adjust prototype.
16721 * except.c (output_function_exception_table): Remove FNNAME parameter
16722 and add SECTION parameter. Ouput one part of the table at a time.
16723 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
16724 the first part of the exception table and emit unwind directives.
16725 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
16726 (i386_pe_seh_cold_init): Likewise.
16727 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
16728 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
16729 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
16730 (ix86_output_call_insn): Emit a nop in one more case for SEH.
16731 * config/i386/winnt.c: Include except.h.
16732 (struct seh_frame_state): Add reg_offset, after_prologue and
16733 in_cold_section fields.
16734 (i386_pe_seh_end_prologue): Set seh->after_prologue.
16735 (i386_pe_seh_cold_init): New function.
16736 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
16737 to seh->in_cold_section.
16738 (seh_emit_push): Record the offset of the push.
16739 (seh_emit_save): Record the offet of the save.
16740 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
16741 Test seh->after_prologue to disregard the epilogue.
16742 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
16743 (i386_pe_end_cold_function): New function.
16745 2018-03-07 Jakub Jelinek <jakub@redhat.com>
16748 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
16749 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
16752 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
16754 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
16755 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
16757 PR middle-end/84723
16758 * multiple_target.c: Include tree-inline.h and intl.h.
16759 (expand_target_clones): Diagnose and fail if node->definition and
16760 !tree_versionable_function_p (node->decl).
16762 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
16764 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
16766 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
16767 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
16768 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
16770 2018-03-06 Jakub Jelinek <jakub@redhat.com>
16773 * combine.c (try_combine): Use reg_or_subregno instead of handling
16774 just paradoxical SUBREGs and REGs.
16776 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
16778 * config/arc/arc.c (arc_finalize_pic): Remove function.
16779 (arc_must_save_register): We use single base PIC register, remove
16780 checks to save/restore the PIC register.
16781 (arc_expand_prologue): Likewise.
16782 * config/arc/arc-protos.h (arc_set_default_type_attributes):
16784 (arc_verify_short): Likewise.
16785 (arc_attr_type): Likewise.
16786 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
16787 (walk_stores): Likewise.
16788 (arc_address_cost): Make it static.
16789 (arc_verify_short): Likewise.
16790 (branch_dest): Likewise.
16791 (arc_attr_type): Likewise.
16792 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
16793 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
16794 (arc_final_prescan_insn): Remove inserting the nops due to
16795 hardware hazards. It is done in reorg step.
16796 (insn_length_variant_t): Remove.
16797 (insn_length_parameters_t): Likewise.
16798 (arc_insn_length_parameters): Likewise.
16799 (arc_get_insn_variants): Likewise.
16800 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
16802 2018-03-06 Jakub Jelinek <jakub@redhat.com>
16804 PR inline-asm/84683
16805 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
16808 PR tree-optimization/84687
16809 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
16811 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
16813 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16815 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
16816 Rename to ppc_speculation_barrier.
16817 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
16818 __builtin_ppc_speculation_barrier.
16820 2018-03-05 Jakub Jelinek <jakub@redhat.com>
16823 * combine.c (combine_simplify_rtx): Don't try to simplify if
16824 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
16827 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
16829 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
16830 to 32 bytes when compiling for POWER9.
16832 2018-03-05 Jakub Jelinek <jakub@redhat.com>
16835 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
16836 regparm >= 3 with no arg reg available also for calls with
16837 flag_force_indirect_call. Pass decl to ix86_function_regparm.
16840 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
16842 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
16844 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
16847 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
16849 2018-03-05 Richard Biener <rguenther@suse.de>
16851 PR tree-optimization/84486
16852 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
16853 When inserting a __builtin_assume_aligned call set the LHS
16854 SSA name alignment info accordingly.
16856 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
16858 PR tree-optimization/84114
16859 * config/aarch64/aarch64.c (aarch64_reassociation_width)
16860 Avoid reassociation of FLOAT_MODE addition.
16862 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
16864 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
16865 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
16866 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
16867 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
16868 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
16869 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
16870 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
16872 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
16873 __builtin_ia32_wbinvd): New builtins.
16874 (SPECIAL_ARGS2): New.
16875 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
16876 (SPECIAL_ARGS2): New.
16877 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
16878 (ix86_valid_target_attribute_inner_p): Ditto.
16879 (ix86_init_mmx_sse_builtins): Add special_args2.
16880 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
16881 TARGET_WBNOINVD_P): New.
16882 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
16883 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
16884 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
16885 * config/i386/immintrin.h (_wbinvd): New intrinsic.
16886 * config/i386/pconfigintrin.h: New file.
16887 * config/i386/wbnoinvdintrin.h: Ditto.
16888 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
16890 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
16892 2018-03-05 Richard Biener <rguenther@suse.de>
16894 PR tree-optimization/84670
16895 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
16897 (BB_VISITED_WITH_VISITED_SUCCS): New define.
16898 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
16899 (compute_antic_aux): Only assert the number of values in ANTIC_IN
16900 doesn't grow if all successors (recursively) were visited at least
16903 2018-03-05 Richard Biener <rguenther@suse.de>
16905 PR tree-optimization/84650
16906 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
16907 if executed in the loop pipeline.
16909 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
16911 * doc/configfiles.texi (Configuration Files): Move info about
16912 conditionalizing $target-protos.h to...
16913 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
16914 differs from $target-protos.h.
16916 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
16917 Chung-Ju Wu <jasonwucj@gmail.com>
16919 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
16920 * config/nds32/nds32-multiple.md (setmemsi): Define.
16921 * config/nds32/nds32-memory-manipulation.c
16922 (nds32_gen_dup_4_byte_to_word_value): New.
16923 (emit_setmem_word_loop): New.
16924 (emit_setmem_byte_loop): New.
16925 (nds32_expand_setmem_loop): New.
16926 (nds32_expand_setmem_loop_v3m): New.
16927 (nds32_expand_setmem_unroll): New.
16928 (nds32_expand_setmem): New.
16930 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
16931 Chung-Ju Wu <jasonwucj@gmail.com>
16933 * config/nds32/nds32-memory-manipulation.c
16934 (nds32_emit_load_store): New.
16935 (nds32_emit_post_inc_load_store): New.
16936 (nds32_emit_mem_move): New.
16937 (nds32_emit_mem_move_block): New.
16938 (nds32_expand_movmemsi_loop_unknown_size): New.
16939 (nds32_expand_movmemsi_loop_known_size): New.
16940 (nds32_expand_movmemsi_loop): New.
16941 (nds32_expand_movmemsi_unroll): New.
16942 (nds32_expand_movmemqi): Rename ...
16943 (nds32_expand_movmemsi): ... to this.
16944 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
16945 (movmemsi): ... to this.
16946 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
16947 (nds32_expand_movmemsi): ... to this.
16949 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
16950 Monk Chiang <sh.chiang04@gmail.com>
16951 Chung-Ju Wu <jasonwucj@gmail.com>
16953 * config/nds32/nds32-protos.h
16954 (nds32_expand_load_multiple): New arguments.
16955 (nds32_expand_store_multiple): Ditto.
16956 (nds32_valid_multiple_load_store): Rename ...
16957 (nds32_valid_multiple_load_store_p): ... to this.
16958 * config/nds32/nds32-memory-manipulation.c
16959 (nds32_expand_load_multiple): Refine implementation.
16960 (nds32_expand_store_multiple): Ditto.
16961 * config/nds32/nds32-multiple.md
16962 (load_multiple): Update nds32_expand_load_multiple interface.
16963 (store_multiple): Update nds32_expand_store_multiple interface.
16964 * config/nds32/nds32-predicates.c
16965 (nds32_valid_multiple_load_store): Rename ...
16966 (nds32_valid_multiple_load_store_p): ... to this and refine
16968 * config/nds32/predicates.md
16969 (nds32_load_multiple_and_update_address_operation): New predicate.
16970 (nds32_store_multiple_and_update_address_operation): New predicate.
16972 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
16973 Chung-Ju Wu <jasonwucj@gmail.com>
16975 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
16976 (combo): New attribute.
16977 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
16979 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
16981 * config/nds32/nds32.opt: Change -mcmodel= default value.
16983 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
16984 Monk Chiang <sh.chiang04@gmail.com>
16985 Chung-Ju Wu <jasonwucj@gmail.com>
16987 * config/nds32/constants.md (unspec_element): New enum.
16988 * config/nds32/constraints.md (Umw): New constraint.
16989 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
16990 * config/nds32/nds32-intrinsic.md: Likewise.
16991 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
16992 (nds32_valid_smw_lwm_base_p): New.
16993 (nds32_output_smw_single_word): New.
16994 (nds32_output_lmw_single_word): New.
16995 (nds32_expand_unaligned_load): New.
16996 (nds32_expand_unaligned_store): New.
16997 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
16998 (nds32_output_smw_single_word): Declare.
16999 (nds32_output_lmw_single_word): Declare.
17000 (nds32_expand_unaligned_load): Declare.
17001 (nds32_expand_unaligned_store): Declare.
17002 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
17003 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
17004 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
17005 NDS32_BUILTIN_UASTORE_DW.
17006 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
17009 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
17010 Kito Cheng <kito.cheng@gmail.com>
17011 Chung-Ju Wu <jasonwucj@gmail.com>
17013 * config/nds32/nds32-intrinsic.c
17014 (nds32_expand_builtin_null_ftype_reg): Delete.
17015 (nds32_expand_builtin_reg_ftype_imm): Ditto.
17016 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
17017 (nds32_read_argument): New.
17018 (nds32_legitimize_target): Ditto.
17019 (nds32_legitimize_argument): Ditto.
17020 (nds32_check_constant_argument): Ditto.
17021 (nds32_expand_unop_builtin): Ditto.
17022 (nds32_expand_unopimm_builtin): Ditto.
17023 (nds32_expand_binop_builtin): Ditto.
17024 (nds32_builtin_decl_impl): Ditto.
17025 (builtin_description): Ditto.
17026 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
17027 (nds32_init_builtins_impl): Ditto.
17028 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
17029 (nds32_builtin_decl): New.
17030 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
17031 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
17033 2018-03-02 Jeff Law <law@redhat.com>
17035 * reorg.c (stop_search_p): Handle DEBUG_INSN.
17036 (redundant_insn, fill_simple_delay_slots): Likewise.
17037 (fill_slots_from_thread): Likewise.
17038 * resource.c (mark_referenced_resources): Likewise.
17039 (mark_set_resources, find_dead_or_set_registers): Likewise.
17041 2018-03-02 Jakub Jelinek <jakub@redhat.com>
17043 * substring-locations.h (format_warning_va): Formatting fix for
17044 ATTRIBUTE_GCC_DIAG.
17045 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
17047 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
17048 * substring-locations.c: Include intl.h.
17049 (format_warning_va): Turned into small wrapper around
17050 format_warning_n_va, renamed to ...
17051 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
17052 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
17054 (format_warning_at_substring_n): New function.
17055 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
17056 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
17057 format_warning_at_substring with just a shorter name instead of
17058 const function pointer.
17059 (fmtwarn_n): New function.
17060 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
17061 appropriate, get rid of all the fmtstr temporaries, move conditionals
17062 with G_() wrapped string literals directly into fmtwarn arguments,
17063 cast dir.len to (int), formatting fixes.
17065 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
17067 * doc/invoke.texi: Remove "Cilk Plus" references.
17069 2018-03-02 Jakub Jelinek <jakub@redhat.com>
17070 Richard Biener <rguenther@suse.de>
17073 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
17074 for error or warning attributes if CALL_FROM_THUNK_P is set.
17077 2018-03-02 Jakub Jelinek <jakub@redhat.com>
17080 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
17081 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
17084 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
17085 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
17087 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
17088 instead of -1U in last predictors element's probability member.
17090 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
17093 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
17094 arguments if they are comparable.
17096 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
17098 PR tree-optimization/84634
17099 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
17100 masks and masked_loop_p with a single loop_masks, making sure it's
17101 null for bb vectorization.
17103 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
17105 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
17106 (vect_analyze_data_ref_access): Use loop->safe_len rather than
17107 loop->force_vectorize to check whether there is no alias.
17109 2018-03-02 Jakub Jelinek <jakub@redhat.com>
17112 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
17114 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
17116 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
17117 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
17118 instead of a loop around prev_real_insn.
17119 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
17122 PR inline-asm/84625
17123 * config/i386/i386.c (ix86_print_operand): Use conditional
17124 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
17127 2018-03-02 Richard Biener <rguenther@suse.de>
17129 PR tree-optimization/84427
17130 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
17131 (bitmap_set_subtract_values): Rewrite to handle multiple
17134 (prune_clobbered_mems): Likewise.
17135 (phi_translate): Take edge instead of pred/phiblock.
17136 (phi_translate_1): Likewise.
17137 (phi_translate_set): Likewise. Insert all translated
17138 exprs for a value into the set, keeping possibly multiple
17139 expressions per value.
17140 (compute_antic_aux): Adjust for phi_translate changes.
17141 When intersecting union the expressions and prune those
17142 not in the final value set, keeping possibly multiple
17143 expressions per value. Do not use value-insertion
17144 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
17145 all expressions. Add verification that the value-sets
17146 only shrink during iteration.
17147 (compute_partial_antic_aux): Adjust for the phi_translate changes.
17148 (do_pre_regular_insertion): Likewise.
17149 (do_pre_partial_partial_insertion): Likewise.
17151 2018-03-02 Richard Biener <rguenther@suse.de>
17154 * config/darwin.c (saved_debug_info_level): New static global.
17155 (darwin_asm_lto_start): Disable debug info generation for LTO out.
17156 (darwin_asm_lto_end): Restore debug info generation settings.
17158 2018-03-01 Martin Liska <mliska@suse.cz>
17161 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
17162 volatile arguments.
17164 2018-03-01 Richard Biener <rguenther@suse.de>
17167 * dwarf2out.c (gen_variable_die): Properly handle late VLA
17168 type annotation with LTO when debug was disabled at compile-time.
17170 2018-03-01 Matthew Fortune <mfortune@gmail.com>
17172 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
17174 (mips_final_postscan_insn): Likewise.
17176 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
17178 PR rtl-optimization/84528
17179 * alias.c (init_alias_target): Add commentary.
17180 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
17181 a unique base value if the frame pointer is not eliminated
17182 to the stack pointer.
17184 2018-03-01 Tom de Vries <tom@codesourcery.com>
17186 PR rtl-optimization/83327
17187 * lra-int.h (hard_regs_spilled_into): Declare.
17188 * lra.c (hard_regs_spilled_into): Define.
17189 (init_reg_info): Init hard_regs_spilled_into.
17190 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
17191 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
17192 (process_bb_lives): Handle hard_regs_spilled_into.
17193 (lra_create_live_ranges_1): Before doing liveness propagation, clear
17194 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
17196 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
17198 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
17199 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
17200 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
17201 * config/rs6000/aix72.h: New file.
17203 2018-02-28 Jakub Jelinek <jakub@redhat.com>
17205 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
17206 instead of warning_at with conditional singular and plural messages
17210 * stor-layout.c (update_alignment_for_field): For
17211 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
17212 && !DECL_PACKED (field), do the alignment update, just use
17213 only desired_align instead of MAX (type_align, desired_align)
17215 (place_field): Don't do known_align < desired_align handling
17216 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
17217 is non-NULL, instead do it after rli->prev_field handling and
17218 only if not within a bitfield word. For DECL_PACKED (field)
17219 use type_align of BITS_PER_UNIT.
17221 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
17223 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
17224 superfluous parentheses and trailing spaces.
17226 2018-02-28 Richard Biener <rguenther@suse.de>
17228 PR tree-optimization/84584
17229 * graphite-scop-detection.c (scop_detection::add_scop): Discard
17230 SCoPs with fake exit edge.
17232 2018-02-28 Martin Liska <mliska@suse.cz>
17235 * timevar.c (timer::print): Fix format to properly print 100%
17238 2018-02-28 Richard Biener <rguenther@suse.de>
17240 PR middle-end/84607
17241 * genmatch.c (capture_info::walk_match): Do not mark
17242 captured expressions without operands as expr_p given
17243 they act more like predicates and should be subject to
17244 "lost tail" side-effect preserving.
17246 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
17248 PR rtl-optimization/81611
17249 * auto-inc-dec.c (attempt_change): Move dead note from
17250 mem_insn if it's the next use of regno
17251 (find_address): Take address use of reg holding
17252 non-incremented value. Add parm to limit search to the named
17254 (merge_in_block): Attempt to use a mem insn that is the next
17255 use of the original regno.
17257 2018-02-27 Martin Sebor <msebor@redhat.com>
17260 * doc/invoke.texi (-Wmissing-attributes): New option.
17261 * print-tree.c (print_node): Handle DECL_UNINLINABLE.
17263 2018-02-27 Martin Sebor <msebor@redhat.com>
17265 PR translation/84207
17266 * diagnostic-core.h (warning_n, error_n, inform_n): Change
17267 n argument to unsigned HOST_WIDE_INT.
17268 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
17269 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
17270 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
17271 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
17273 2018-02-27 Richard Biener <rguenther@suse.de>
17275 PR tree-optimization/84512
17276 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
17277 Do not use the estimate returned from record_stmt_cost for
17278 the scalar iteration cost but sum properly using add_stmt_cost.
17280 2018-02-27 Richard Biener <rguenther@suse.de>
17282 PR tree-optimization/84466
17283 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
17284 Adjust last change to less strictly validate use operands.
17286 2018-02-27 Martin Liska <mliska@suse.cz>
17288 PR gcov-profile/84548
17289 * gcov.c (process_file): Allow partial overlap and consider it
17290 also as group functions.
17291 (output_lines): Properly calculate range of lines for a group.
17293 2018-02-27 Martin Liska <mliska@suse.cz>
17295 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
17296 'ggc' suffixes. Change first column width.
17297 (timer::print): Fix formatting of the column.
17299 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
17301 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
17302 preserve inline entry blocks for the sake of debug inline
17303 entry point markers alone.
17304 (remove_unused_locals): Suggest in comments a better place to
17305 force the preservation of inline entry blocks that are
17306 otherwise unused, but do not preserve them.
17308 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
17310 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
17312 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
17315 * config/i386/constraints.md (Bs): Replace
17316 ix86_indirect_branch_register with
17317 TARGET_INDIRECT_BRANCH_REGISTER.
17319 * config/i386/i386.md (indirect_jump): Likewise.
17320 (tablejump): Likewise.
17321 (*sibcall_memory): Likewise.
17322 (*sibcall_value_memory): Likewise.
17323 Peepholes of indirect call and jump via memory: Likewise.
17324 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
17325 (*sibcall_value_GOT_32): Likewise.
17326 * config/i386/predicates.md (indirect_branch_operand): Likewise.
17327 (GOT_memory_operand): Likewise.
17328 (call_insn_operand): Likewise.
17329 (sibcall_insn_operand): Likewise.
17330 (GOT32_symbol_operand): Likewise.
17331 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
17333 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
17335 PR rtl-optimization/83496
17336 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
17337 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
17338 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
17339 redundant insn, if any.
17340 (relax_delay_slots): Likewise.
17341 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
17343 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
17345 PR tree-optimization/83965
17346 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
17347 that grouped statements are part of a reduction chain. Return
17348 true if the statement is not marked as a reduction itself but
17349 is part of a group.
17350 (vect_recog_dot_prod_pattern): Don't check whether the statement
17351 is part of a group here.
17352 (vect_recog_sad_pattern): Likewise.
17353 (vect_recog_widen_sum_pattern): Likewise.
17355 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
17358 * final.c (rest_of_clean_state): Also look for calls inside sequences.
17360 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
17363 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
17365 (ix86_output_indirect_function_return): New prototype.
17366 (ix86_split_simple_return_pop_internal): Likewise.
17367 * config/i386/i386.c (indirect_return_via_cx): New.
17368 (indirect_return_via_cx_bnd): Likewise.
17369 (indirect_thunk_name): Handle return va CX_REG.
17370 (output_indirect_thunk_function): Create alias for
17371 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
17372 (ix86_output_indirect_jmp): Remove the bool argument.
17373 (ix86_output_indirect_function_return): New function.
17374 (ix86_split_simple_return_pop_internal): Likewise.
17375 * config/i386/i386.md (*indirect_jump): Don't pass false
17376 to ix86_output_indirect_jmp.
17377 (*tablejump_1): Likewise.
17378 (simple_return_pop_internal): Change it to define_insn_and_split.
17379 Call ix86_split_simple_return_pop_internal to split it for
17380 -mfunction-return=.
17381 (simple_return_indirect_internal): Call
17382 ix86_output_indirect_function_return instead of
17383 ix86_output_indirect_jmp.
17385 2018-02-26 Jakub Jelinek <jakub@redhat.com>
17388 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
17389 memset and value initialization afterwards.
17391 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
17393 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
17395 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
17398 * common/config/aarch64/aarch64-common.c
17399 (aarch_option_optimization_table[]): Switch
17400 off fomit-frame-pointer
17402 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
17403 Chung-Ju Wu <jasonwucj@gmail.com>
17405 * config/nds32/nds32-multiple.md (load_multiple): Disallow
17407 (store_multiple): Ditto.
17409 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
17411 * config.gcc: Add --with-cpu support for nds32 target.
17412 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
17413 * config/nds32/nds32.opt: Add -mcpu= option.
17415 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
17417 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
17418 isel=yes): Warn for these deprecated options.
17420 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
17422 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
17423 ISA_2_5_MASKS_EMBEDDED.
17425 2018-02-23 Jakub Jelinek <jakub@redhat.com>
17427 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
17428 p->max as pointers rather than using iterative_hash_expr.
17430 2018-02-23 Carl Love <cel@us.ibm.com>
17432 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
17433 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
17435 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
17436 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
17437 P8V_BUILTIN_VEC_VUNSIGNED2.
17439 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
17442 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
17443 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
17445 * lra-constraints.c (curr_insn_transform): Set up
17446 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
17448 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
17449 * lra-eliminations.c (spill_pseudos): Ditto.
17450 (process_insn_for_elimination): Ditto.
17451 * lra-lives.c (reg_early_clobber_p): Use the new macros.
17452 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
17453 LRA_NON_CLOBBERED_ALT.
17455 2018-02-22 Martin Sebor <msebor@redhat.com>
17457 PR tree-optimization/84480
17458 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
17459 to maybe_diag_stxncpy_trunc. Call it.
17460 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
17461 from gimple_fold_builtin_strcpy. Print inlining stack.
17462 (handle_builtin_stxncpy): Print inlining stack.
17463 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
17465 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
17468 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
17469 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
17470 and -fcheck-pointer-bounds are used together.
17471 (indirect_thunk_prefix): New enum.
17472 (indirect_thunk_need_prefix): New function.
17473 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
17474 "_nt" instead of "_bnd" for NOTRACK prefix.
17475 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
17476 (output_indirect_thunk_function): Likewise.
17478 (ix86_code_end): Update output_indirect_thunk_function calls.
17479 (ix86_output_indirect_branch_via_reg): Replace
17480 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
17481 (ix86_output_indirect_branch_via_push): Likewise.
17482 (ix86_output_function_return): Likewise.
17483 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
17484 incompatible with -fcf-protection=branch and
17485 -fcheck-pointer-bounds.
17487 2018-02-22 Steve Ellcey <sellcey@cavium.com>
17490 * config/aarch64/aarch64.c (aarch64_print_address_internal):
17491 Change gcc_assert call to output_operand_lossage.
17493 2018-02-22 Steve Ellcey <sellcey@cavium.com>
17495 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
17497 2018-02-22 DJ Delorie <dj@redhat.com>
17498 Sebastian Perta <sebastian.perta@renesas.com>
17499 Oleg Endo <olegendo@gcc.gnu.org>
17501 * config/rx/rx.c (rx_rtx_costs): New function.
17502 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
17504 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
17506 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
17508 2018-02-22 Martin Liska <mliska@suse.cz>
17511 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
17512 Add "native" as a possible value.
17514 2018-02-22 Martin Liska <mliska@suse.cz>
17517 * config/i386/i386.c (ix86_option_override_internal):
17518 Add "native" as a possible value for -march and -mtune.
17520 2018-02-22 Jakub Jelinek <jakub@redhat.com>
17523 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
17524 to all type variants.
17526 PR tree-optimization/84503
17527 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
17528 width as info->bitpos + info->bitsize - start.
17529 (merged_store_group::merge_overlapping): Simplify width computation.
17530 (check_no_overlap): New function.
17531 (imm_store_chain_info::try_coalesce_bswap): Compute expected
17532 start + width and last_order of the group, fail if check_no_overlap
17534 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
17535 to group if check_no_overlap fails.
17537 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
17539 * config/rs6000/altivec.md: Delete contraint arguments to
17540 define_expand, define_split, and define_peephole2, and in
17541 define_insn_and_split if always unused.
17542 * config/rs6000/darwin.md: Ditto.
17543 * config/rs6000/dfp.md: Ditto.
17544 * config/rs6000/rs6000.md: Ditto.
17545 * config/rs6000/sync.md: Ditto.
17546 * config/rs6000/vector.md: Ditto.
17547 * config/rs6000/vsx.md: Ditto.
17549 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
17551 * config/rs6000/altivec.md: Write output control strings as braced
17552 blocks instead of double-quoted strings.
17553 * config/rs6000/darwin.md: Ditto.
17554 * config/rs6000/rs6000.md: Ditto.
17555 * config/rs6000/vector.md: Ditto.
17556 * config/rs6000/vsx.md: Ditto.
17558 2018-02-21 Jason Merrill <jason@redhat.com>
17560 PR c++/84314 - ICE with templates and fastcall attribute.
17561 * attribs.c (build_type_attribute_qual_variant): Remove assert.
17563 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
17565 * ipa-cp.c (determine_versionability): Fix comment typos.
17567 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
17570 * ipa-cp.c (determine_versionability): Do not version functions caling
17573 2018-02-21 Martin Liska <mliska@suse.cz>
17576 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
17577 Add "native" as a possible value.
17578 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
17579 the macro when native cpu detection is available.
17581 2018-02-21 Martin Liska <mliska@suse.cz>
17584 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
17585 Add "native" as a possible value.
17586 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
17587 when native cpu detection is available.
17589 2018-02-21 Jakub Jelinek <jakub@redhat.com>
17590 Martin Sebor <msebor@redhat.com>
17592 PR tree-optimization/84478
17593 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
17595 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
17596 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
17597 support which is conservatively correct, for 2 only stay conservative
17598 for maxlen. Formatting and comment capitalization fixes. Add STRICT
17599 argument to the 2 argument get_range_strlen, adjust 6 arg
17600 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
17602 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
17603 (gimple_fold_builtin_strlen): Pass true as last argument to
17606 2018-02-20 Martin Sebor <msebor@redhat.com>
17608 PR middle-end/84095
17609 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
17610 (builtin_memref::set_base_and_offset): Same. Handle inner references.
17611 (builtin_memref::builtin_memref): Factor out parts into
17612 set_base_and_offset and call it.
17614 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
17616 PR middle-end/84406
17617 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
17618 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
17619 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
17620 search at the associated MODE_INT.
17622 2018-02-20 Jeff Law <law@redhat.com>
17624 PR middle-end/82123
17625 PR tree-optimization/81592
17626 PR middle-end/79257
17627 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
17628 for range data rather than using global data.
17629 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
17630 range data rather than using global data.
17631 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
17632 pass it to children as needed.
17633 (struct directive::fmtresult): Similarly.
17634 (struct directive::set_width): Similarly.
17635 (struct directive::set_precision): Similarly.
17636 (format_integer, format_directive, parse_directive): Similarly.
17637 (format_none): Accept unnamed vr_values parameter.
17638 (format_percent, format_floating, format_character): Similarly.
17639 (format_string, format_plain): Similarly.
17640 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
17641 the EVRP range analyzer for range data rather than using global data.
17642 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
17643 gimple-ssa-evrp-analyze.h
17644 (class sprintf_dom_walker): Add after_dom_children member function.
17645 Add evrp_range_analyzer member.
17646 (sprintf_dom_walker::before_dom_children): Call into the EVRP
17647 range analyzer as needed.
17648 (sprintf_dom_walker::after_dom_children): New member function.
17649 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
17651 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
17652 (evrp_range_analyzer::pop_to_marker): Likewise.
17654 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
17656 PR tree-optimization/84419
17657 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
17658 with the required type if its current type is compatible but
17661 2018-02-20 Jakub Jelinek <jakub@redhat.com>
17663 PR middle-end/82004
17664 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
17665 after vectorization.
17667 2018-02-20 Martin Liska <mliska@suse.cz>
17670 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
17671 possible values if we don't have a hint.
17673 2018-02-20 Martin Liska <mliska@suse.cz>
17677 * final.c (shorten_branches): Build align_tab array with one
17679 * opts.c (finish_options): Add alignment option limit check.
17680 (MAX_CODE_ALIGN): Likewise.
17681 (MAX_CODE_ALIGN_VALUE): Likewise.
17682 * doc/invoke.texi: Document maximum allowed option value for
17683 all -falign-* options.
17685 2018-02-19 Jakub Jelinek <jakub@redhat.com>
17688 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
17689 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
17690 * var-tracking.c (emit_note_insn_var_location): Remove all references
17691 to NOTE_INSN_CALL_ARG_LOCATION.
17692 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
17693 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
17694 Use copy_rtx_if_shared.
17695 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
17696 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
17697 (dwarf2out_var_location): Remove handling of
17698 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
17700 * final.c (final_scan_insn): Remove all references to
17701 NOTE_INSN_CALL_ARG_LOCATION.
17702 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
17703 before dumping final insns.
17704 * except.c (emit_note_eh_region_end): Remove all references to
17705 NOTE_INSN_CALL_ARG_LOCATION.
17706 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
17707 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
17708 * config/arc/arc.c (hwloop_optimize): Likewise.
17709 * config/arm/arm.c (create_fix_barrier): Likewise.
17710 * config/s390/s390.c (s390_chunkify_start): Likewise.
17711 * config/sh/sh.c (find_barrier): Likewise.
17712 * config/i386/i386.c (rest_of_insert_endbranch,
17713 ix86_seh_fixup_eh_fallthru): Likewise.
17714 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
17715 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
17716 * config/frv/frv.c (frv_function_prologue): Likewise.
17717 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
17719 (note_outside_basic_block_p): Remove all references to
17720 NOTE_INSN_CALL_ARG_LOCATION.
17721 * gengtype.c (adjust_field_rtx_def): Likewise.
17722 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
17724 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
17725 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
17728 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
17731 PR tree-optimization/84452
17732 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
17733 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
17736 2018-02-19 Martin Liska <mliska@suse.cz>
17739 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
17741 2018-02-19 Martin Liska <mliska@suse.cz>
17742 Richard Sandiford <richard.sandiford@linaro.org>
17744 PR tree-optimization/82491
17745 * gimple-fold.c (get_base_constructor): Make earlier bail out
17748 2018-02-19 Carl Love <cel@us.ibm.com>
17750 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
17751 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
17753 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
17754 P8V_BUILTIN_VEC_NEG.
17756 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
17758 * config/rl78/rl78.md (movdf): New define expand.
17760 2018-02-19 Martin Liska <mliska@suse.cz>
17763 * doc/invoke.texi: Fix typo.
17764 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
17766 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
17768 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
17769 handle rs6000_single_float and rs6000_double_float specially for
17772 2018-02-16 Jeff Law <law@redhat.com>
17774 * config/rx/rx.c (add_pop_cfi_notes): New function.;
17775 (pop_regs): Use it.
17777 2018-02-16 Jakub Jelinek <jakub@redhat.com>
17780 * ipa-inline.c (inline_small_functions): Fix a typo.
17782 2018-02-16 Nathan Sidwell <nathan@acm.org>
17784 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
17786 2018-02-16 Carl Love <cel@us.ibm.com>
17788 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
17789 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
17790 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
17791 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
17792 expansion to P8V_BUILTIN_VEC_FLOAT2.
17794 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
17796 PR rtl-optimization/70023
17797 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
17798 src_regno into account.
17800 2018-02-16 Carl Love <cel@us.ibm.com>
17802 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
17803 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
17804 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
17805 * config/rs6000/rs6000.c: Remove case statements for
17806 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
17807 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
17808 and P9V_BUILTIN_VEC_VINSERT4B.
17809 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
17810 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
17811 * config/rs6000/vsx.md:
17812 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
17815 2018-02-16 Carl Love <cel@us.ibm.com>
17817 * config/rs6000/altivec.h: Add builtin names vec_extract4b
17819 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
17821 * config/rs6000/rs6000-c.c: Add the definitions for
17822 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
17823 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
17824 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
17825 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
17826 definition for insert4b and define insn *insert3b_internal.
17827 * doc/extend.texi: Add documentation for vec_extract4b.
17829 2018-02-16 Nathan Sidwell <nathan@acm.org>
17831 * doc/extend.texi (Backwards Compatibility): Mention friend
17832 injection. Note for-scope is deprecated.
17833 * doc/invoke.texi (-ffriend-injection): Deprecate.
17835 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
17837 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
17838 that moved to I2, also allow destinations that are a paradoxical
17839 subreg (instead of a normal reg).
17841 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
17844 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
17847 2018-02-16 Richard Biener <rguenther@suse.de>
17849 PR tree-optimization/84037
17850 PR tree-optimization/84016
17852 * config/i386/i386.c (ix86_builtin_vectorization_cost):
17853 Adjust vec_construct for the fact we need additional higher latency
17854 128bit inserts for AVX256 and AVX512 vector builds.
17855 (ix86_add_stmt_cost): Scale vector construction cost for
17858 2018-02-16 Richard Biener <rguenther@suse.de>
17860 PR tree-optimization/84417
17861 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
17862 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
17863 (non_rewritable_lvalue_p): Likewise, use poly-ints.
17865 2018-02-16 Martin Liska <mliska@suse.cz>
17868 * internal-fn.def (ASAN_CHECK): Set proper flags.
17869 (ASAN_MARK): Likewise.
17871 2018-02-16 Julia Koval <julia.koval@intel.com>
17873 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
17874 from PTA_CANNONLAKE.
17876 2018-02-16 Jakub Jelinek <jakub@redhat.com>
17879 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
17880 Use ++iter rather than iter++ for std::list iterators.
17881 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
17882 defer deleting them until all nodes in the forest are processed. Do
17883 free even leaf nodes. Change to_process into auto_vec.
17886 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
17887 * vec.h (vec_default_construct): Use memset instead of placement new
17888 if BROKEN_VALUE_INITIALIZATION is defined.
17889 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
17890 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
17893 PR rtl-optimization/83723
17894 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
17895 * lra.c (lra_substitute_pseudo): Likewise. If true, use
17896 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
17898 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
17900 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
17902 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
17904 PR rtl-optimization/81443
17905 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
17906 from inner REGs to paradoxical SUBREGs.
17908 2018-02-16 Richard Biener <rguenther@suse.de>
17910 PR tree-optimization/84399
17911 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
17912 For operands we can analyze at their definition make sure we can
17913 analyze them at each use as well.
17915 2018-02-16 Richard Biener <rguenther@suse.de>
17917 PR tree-optimization/84190
17918 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
17919 volatile accesses if the decl isn't volatile.
17921 2018-02-15 Jason Merrill <jason@redhat.com>
17923 PR c++/84314 - ICE with templates and fastcall attribute.
17924 * attribs.c (build_type_attribute_qual_variant): Don't clobber
17925 TYPE_CANONICAL on an existing type.
17927 2018-02-15 Jakub Jelinek <jakub@redhat.com>
17929 PR tree-optimization/84383
17930 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
17931 dstoff nor call operand_equal_p if dstbase is NULL.
17933 PR tree-optimization/84334
17934 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
17935 also a CONSTANT_CLASS_P, punt.
17937 2018-02-14 Jim Wilson <jimw@sifive.com>
17939 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
17940 first SMALL_OPERAND check. New local min_second_step. Move assert
17941 to where locals are set. Add TARGET_RVC support.
17942 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
17944 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
17946 * doc/invoke.texi: Correct -Wformat-overflow code sample.
17948 2018-02-14 Martin Sebor <msebor@redhat.com>
17950 PR tree-optimization/83698
17951 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
17952 arrays constrain the offset range to their bounds.
17953 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
17954 (builtin_access::overlap): Avoid setting the size of overlap if it's
17956 (maybe_diag_overlap): Also consider arrays when deciding what values
17957 of offsets to include in diagnostics.
17959 2018-02-14 Martin Sebor <msebor@redhat.com>
17962 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
17963 that correspond to the kind of a declaration.
17965 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
17968 * config/pa/pa.md: Load address of PIC label using the linkage table
17969 if the label is nonlocal.
17971 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
17973 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
17974 warning message if user requests -maltivec=be.
17975 * doc/invoke.texi: Document deprecation of -maltivec=be.
17977 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
17980 * config/rs6000/rs6000-c.c: Update definitions for
17981 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
17982 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
17984 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
17987 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
17988 add _get_ssp intrinsics. Remove argument from
17989 __builtin_ia32_rdssp[d|q].
17990 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
17991 * config/i386/i386-builtin.def: Remove argument from
17992 __builtin_ia32_rdssp[d|q].
17993 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
17994 ix86_expand_special_args_builtin for _rdssp[d|q].
17995 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
17996 Clear register before usage.
17997 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
17998 Add documentation for new _get_ssp and _inc_ssp intrinsics.
18000 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
18002 PR tree-optimization/84357
18003 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
18004 operand 1 of an ARRAY_REF too.
18006 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
18009 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
18010 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
18012 (set_of_reg): New struct.
18013 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
18014 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
18015 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
18017 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
18018 Split into bitclr, bitset, bitinvert patterns if appropriate.
18019 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
18020 use rx_fuse_in_memory_bitop.
18021 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
18022 to named insn, correct maximum insn length.
18024 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
18027 * machmode.def: Define a complex mode for PARTIAL_INT.
18028 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
18030 * doc/rtl.texi: Document CSPImode.
18031 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
18033 (msp430_hard_regno_nregs_with_padding): Likewise.
18035 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
18038 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
18040 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
18042 PR rtl-optimization/84169
18043 * combine.c (try_combine): New variable split_i2i3. Set it to true if
18044 we generated a parallel as new i3 and we split that to new i2 and i3
18045 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
18046 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
18047 those to i2, not i1. Partially rewrite this scan code.
18049 2018-02-13 Jakub Jelinek <jakub@redhat.com>
18052 * stor-layout.c (place_field): For variable length fields, adjust
18053 offset_align afterwards not just based on the field's alignment,
18054 but also on the size.
18056 PR middle-end/84309
18057 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
18058 of exps and logs in the use_exp2 case.
18060 2018-02-13 Jeff Law <law@redhat.com>
18062 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
18063 entry for "vector".
18065 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
18068 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
18072 * common.opt (gas-loc-support, gas-locview-support): New.
18073 (ginline-points, ginternal-reset-location-views): New.
18074 * doc/invoke.texi: Document them. Use @itemx where intended.
18075 (gvariable-location-views): Adjust.
18076 * target.def (reset_location_view): New.
18077 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
18078 (TARGET_RESET_LOCATION_VIEW): New.
18079 * doc/tm.texi: Rebuilt.
18080 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
18081 (dwarf2out_default_as_locview_support): New.
18082 (output_asm_line_debug_info): Use option variables.
18083 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
18084 (output_loc_list): Likewise.
18085 (add_high_low_attributes): Check option variables.
18086 Don't output entry view attribute in strict mode.
18087 (gen_inlined_subroutine_die): Check option variables.
18088 (dwarf2out_inline_entry): Likewise.
18089 (init_sections_and_labels): Likewise.
18090 (dwarf2out_early_finish): Likewise.
18091 (maybe_reset_location_view): New, from...
18092 (dwarf2out_var_location): ... here. Call it.
18093 * debug.h (dwarf2out_default_as_loc_support): Declare.
18094 (dwarf2out_default_as_locview_support): Declare.
18095 * hooks.c (hook_int_rtx_insn_0): New.
18096 * hooks.h (hook_int_rtx_insn_0): Declare.
18097 * toplev.c (process_options): Take -gas-loc-support and
18098 -gas-locview-support from dwarf2out. Enable
18099 -gvariable-location-views by default only with locview
18100 assembler support. Enable -ginternal-reset-location-views by
18101 default only if the target defines the corresponding hook.
18102 Enable -ginline-points by default if location views are
18103 enabled; force it disabled if statement frontiers are
18105 * tree-inline.c (expand_call_inline): Check option variables.
18106 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
18108 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
18110 PR tree-optimization/84321
18111 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
18112 handling. Also check whether the anti-range contains any values
18113 that satisfy the mask; switch to a VR_RANGE if not.
18115 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
18118 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
18120 2018-02-13 Martin Jambor <mjambor@suse.cz>
18123 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
18124 of call statements, also set location of a load to a temporary.
18126 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
18128 * config/rl78/rl78.c (add_vector_labels): New function.
18129 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
18130 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
18131 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
18132 which checks that no arguments are passed.
18133 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
18134 * doc/extend.texi: Documentation for the new attribute.
18136 2018-02-13 Andreas Schwab <schwab@suse.de>
18138 * config/riscv/linux.h (CPP_SPEC): Define.
18140 2018-02-13 Jakub Jelinek <jakub@redhat.com>
18143 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
18144 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
18145 OPTION_MASK_ISA_AES as first argument to def_builtin_const
18146 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
18147 instead of OPTION_MASK_ISA_PCLMUL as first argument to
18148 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
18149 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
18150 temporarily for AES and PCLMUL builtins.
18152 PR tree-optimization/84339
18153 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
18154 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
18157 PR middle-end/84309
18158 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
18159 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
18160 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
18162 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
18164 * omp-simd-clone.h: New file.
18165 * omp-simd-clone.c: Include omp-simd-clone.h.
18166 (expand_simd_clones): No longer static.
18167 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
18168 cgraph.h and omp-simd-clone.h.
18169 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
18170 (vect_recog_widen_shift_pattern): Formatting fix.
18171 (vect_pattern_recog_1): Don't check optab for calls.
18174 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
18175 operands[2] into a REG before using gen_lowpart on it.
18177 2018-02-12 Jeff Law <law@redhat.com>
18180 * config/sh/sh.c (find_barrier): Consider a sibling call
18183 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
18184 successfully back substituting a reg.
18186 2018-02-12 Richard Biener <rguenther@suse.de>
18188 PR tree-optimization/84037
18189 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
18190 parameter, move visited init to caller.
18191 (vect_slp_analyze_operations): Separate cost from validity
18192 check, initialize visited once for all instances.
18193 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
18195 * tree-vect-stmts.c (vect_model_simple_cost): Make early
18197 (vect_model_promotion_demotion_cost): Likewise.
18198 (vectorizable_bswap): Guard cost modeling with !slp_node
18199 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
18201 (vectorizable_call): Likewise.
18202 (vectorizable_conversion): Likewise.
18203 (vectorizable_assignment): Likewise.
18204 (vectorizable_shift): Likewise.
18205 (vectorizable_operation): Likewise.
18206 (vectorizable_store): Likewise.
18207 (vectorizable_load): Likewise.
18208 (vectorizable_condition): Likewise.
18209 (vectorizable_comparison): Likewise.
18211 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
18214 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
18215 (ASAN_MARK): Fix fnspec to account for return value, change pointer
18216 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
18218 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
18220 PR middle-end/83665
18221 * params.def (inline-min-speedup): Increase from 8 to 15.
18222 (max-inline-insns-auto): Decrease from 40 to 30.
18223 * ipa-split.c (consider_split): Add some buffer for function to
18224 be considered inlining candidate.
18225 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
18228 2018-02-12 Richard Biener <rguenther@suse.de>
18230 PR tree-optimization/84037
18231 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
18232 matched stmts if we cannot swap the non-matched ones.
18234 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
18236 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
18237 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
18238 _mm_maskz_scalef_round_ss): New intrinsics.
18239 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
18240 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
18241 __builtin_ia32_scalefss_round): Remove.
18242 (__builtin_ia32_scalefsd_mask_round,
18243 __builtin_ia32_scalefss_mask_round): New intrinsics.
18244 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
18245 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
18246 ((match_operand:VF_128 2 "<round_nimm_predicate>"
18247 "<round_constraint>")): Changed to ...
18248 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
18249 "<round_scalar_constraint>")): ... this.
18250 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
18251 %0, %1, %2<round_op3>}"): Changed to ...
18252 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
18253 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
18254 %2<round_scalar_mask_op3>}"): ... this.
18255 * config/i386/subst.md (round_scalar_nimm_predicate): New.
18257 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
18259 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
18260 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
18261 (_mm_maskz_sqrt_round_ss): New intrinsics.
18262 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
18263 (__builtin_ia32_sqrtsd_mask_round)
18264 (__builtin_ia32_sqrtss_mask_round): New builtins.
18265 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
18266 (__builtin_ia32_sqrtss_round): Remove.
18267 (__builtin_ia32_sqrtsd_mask_round)
18268 (__builtin_ia32_sqrtss_mask_round): New builtins.
18269 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
18270 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
18271 ((match_operand:VF_128 1 "vector_operand"
18272 "xBm,<round_constraint>")): Changed to ...
18273 ((match_operand:VF_128 1 "vector_operand"
18274 "xBm,<round_scalar_constraint>")): ... this.
18275 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
18276 %0, %2, %<iptr>1<round_op3>}): Changed to ...
18277 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
18278 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
18279 %<iptr>1<round_scalar_mask_op3>}): ... this.
18280 ((set_attr "prefix" "<round_prefix>")): Changed to ...
18281 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
18283 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
18286 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
18287 Cast vec_cmpeq result to correct type.
18288 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
18289 Cast vec_cmpgt result to correct type.
18291 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
18293 * final.c (final_scan_insn_1): Renamed from...
18294 (final_scan_insn): ... this. New wrapper, to recover
18295 seen from the outermost call in recursive ones.
18296 * config/sparc/sparc.c (output_return): Drop seen from call.
18297 (output_sibcall): Likewise.
18298 * config/visium/visium.c (output_branch): Likewise.
18300 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
18302 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
18305 2018-02-10 Alan Modra <amodra@gmail.com>
18308 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
18309 Specify LR as an input.
18311 2018-02-10 Jakub Jelinek <jakub@redhat.com>
18314 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
18315 remove_member_access_dummy_vars): New functions.
18316 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
18317 lower_omp_1, execute_lower_omp): Use them.
18319 PR rtl-optimization/84308
18320 * shrink-wrap.c (spread_components): Release todo vector.
18322 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
18324 PR rtl-optimization/57193
18325 * ira-color.c (struct allocno_color_data): Add member
18326 conflict_allocno_hard_prefs.
18327 (update_conflict_allocno_hard_prefs): New.
18328 (bucket_allocno_compare_func): Add a preference based on
18329 conflict_allocno_hard_prefs.
18330 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
18331 (color_allocnos): Remove a dead code. Initiate
18332 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
18334 2018-02-09 Jakub Jelinek <jakub@redhat.com>
18337 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
18338 constraint from =wa to wa. Avoid a subreg on the output operand,
18339 instead use a pseudo and subreg it in a move.
18340 (p9_xxbrd_<mode>): Changed to ...
18341 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
18342 (p9_xxbrd_v2df): New expander.
18343 (p9_xxbrw_<mode>): Changed to ...
18344 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
18345 (p9_xxbrw_v4sf): New expander.
18347 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
18349 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
18351 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
18354 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
18355 multiply in 32-bit mode.
18356 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
18357 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
18360 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
18362 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
18363 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
18364 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
18365 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
18367 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
18370 * dwarf2out.c (is_trivial_indirect_ref): New function.
18371 (dwarf2out_late_global_decl): Do not generate a location
18372 attribute for variables that have a non-trivial DECL_VALUE_EXPR
18373 and that are not defined in the current unit.
18375 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
18377 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
18378 instead of a libcall for UNORDERED.
18380 2018-02-09 Tamar Christina <tamar.christina@arm.com>
18383 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
18384 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
18386 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
18389 * config/s390/s390.c (s390_set_current_function): Invoke
18390 s390_indirect_branch_settings also if fndecl didn't change.
18392 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
18394 * config/rs6000/rs6000.md (blockage): Set length to zero.
18396 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
18398 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
18400 2018-02-09 Jakub Jelinek <jakub@redhat.com>
18403 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
18404 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
18408 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
18409 PARALLEL incoming that failed vt_get_decl_and_offset check.
18411 PR middle-end/84237
18412 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
18413 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
18415 (get_variable_section): For decls in named .bss* sections pass true as
18416 second argument to bss_initializer_p.
18418 2018-02-09 Marek Polacek <polacek@redhat.com>
18419 Jakub Jelinek <jakub@redhat.com>
18422 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
18423 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
18424 Sync some changes from cxx_fold_indirect_ref.
18426 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
18428 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
18430 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
18431 (BLOCK_INLINE_ENTRY_LABEL): New.
18432 (dwarf2out_var_location): Disregard inline entry markers.
18433 (inline_entry_data): New struct.
18434 (inline_entry_data_hasher): New hashtable type.
18435 (inline_entry_data_hasher::hash): New.
18436 (inline_entry_data_hasher::equal): New.
18437 (inline_entry_data_table): New variable.
18438 (add_high_low_attributes): Add DW_AT_entry_pc and
18439 DW_AT_GNU_entry_view attributes if a pending entry is found
18440 in inline_entry_data_table. Add old entry_pc attribute only
18441 if debug nonbinding markers are disabled.
18442 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
18443 markers are enabled.
18444 (block_within_block_p, dwarf2out_inline_entry): New.
18445 (dwarf2out_finish): Check that no entries remained in
18446 inline_entry_data_table.
18447 * final.c (reemit_insn_block_notes): Handle inline entry notes.
18448 (final_scan_insn, notice_source_line): Likewise.
18449 (rest_of_clean_state): Skip inline entry markers.
18450 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
18452 * gimple.c (gimple_build_debug_inline_entry): New.
18453 * gimple.h (enum gimple_debug_subcode): Add
18454 GIMPLE_DEBUG_INLINE_ENTRY.
18455 (gimple_build_debug_inline_entry): Declare.
18456 (gimple_debug_inline_entry_p): New.
18457 (gimple_debug_nonbind_marker_p): Adjust.
18458 * insn-notes.def (INLINE_ENTRY): New.
18459 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
18460 inline entry marker notes.
18461 (print_insn): Likewise.
18462 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
18463 (INSN_DEBUG_MARKER_KIND): Likewise.
18464 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
18465 * tree-inline.c (expand_call_inline): Build and insert
18466 debug_inline_entry stmt.
18467 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
18468 inline entry blocks early, if nonbind markers are enabled.
18469 (dump_scope_block): Dump fragment info.
18470 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
18471 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
18472 (gimple_build_debug_inline_entry): New.
18473 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
18474 Enable/disable inline entry points too.
18475 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
18476 (DEBUG_INSN): Describe inline entry markers.
18478 * common.opt (gvariable-location-views): New.
18479 (gvariable-location-views=incompat5): New.
18480 * config.in: Rebuilt.
18481 * configure: Rebuilt.
18482 * configure.ac: Test assembler for view support.
18483 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
18484 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
18485 * dwarf2out.c (var_loc_view): New typedef.
18486 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
18487 (dwarf2out_locviews_in_attribute): New.
18488 (dwarf2out_locviews_in_loclist): New.
18489 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
18490 (enum dw_line_info_opcode): Add LI_adv_address.
18491 (struct dw_line_info_table): Add view.
18492 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
18493 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
18494 (zero_view_p): New variable.
18495 (ZERO_VIEW_P): New macro.
18496 (output_asm_line_debug_info): New.
18497 (struct var_loc_node): Add view.
18498 (add_AT_view_list, AT_loc_list): New.
18499 (add_var_loc_to_decl): Add view param. Test it against last.
18500 (new_loc_list): Add view params. Record them.
18501 (AT_loc_list_ptr): Handle loc and view lists.
18502 (view_list_to_loc_list_val_node): New.
18503 (print_dw_val): Handle dw_val_class_view_list.
18504 (size_of_die): Likewise.
18505 (value_format): Likewise.
18506 (loc_list_has_views): New.
18507 (gen_llsym): Set vl_symbol too.
18508 (maybe_gen_llsym, skip_loc_list_entry): New.
18509 (dwarf2out_maybe_output_loclist_view_pair): New.
18510 (output_loc_list): Output view list or entries too.
18511 (output_view_list_offset): New.
18512 (output_die): Handle dw_val_class_view_list.
18513 (output_dwarf_version): New.
18514 (output_compilation_unit_header): Use it.
18515 (output_skeleton_debug_sections): Likewise.
18516 (output_rnglists, output_line_info): Likewise.
18517 (output_pubnames, output_aranges): Update version comments.
18518 (output_one_line_info_table): Output view numbers in asm comments.
18519 (dw_loc_list): Determine current endview, pass it to new_loc_list.
18520 Call maybe_gen_llsym.
18521 (loc_list_from_tree_1): Adjust.
18522 (add_AT_location_description): Create view list attribute if
18523 needed, check it's absent otherwise.
18524 (convert_cfa_to_fb_loc_list): Adjust.
18525 (maybe_emit_file): Call output_asm_line_debug_info for test.
18526 (dwarf2out_var_location): Reset views as needed. Precompute
18527 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
18528 attribute. Set view.
18529 (new_line_info_table): Reset next view.
18530 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
18531 (dwarf2out_source_line): Likewise. Output view resets and labels to
18532 the assembler, or select appropriate line info opcodes.
18533 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
18534 (optimize_string_length): Catch it. Adjust.
18535 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
18536 dw_val_class_view_list, and remove it if no longer needed.
18537 (hash_loc_list): Hash view numbers.
18538 (loc_list_hasher::equal): Compare them.
18539 (optimize_location_lists): Check whether a view list symbol is
18540 needed, and whether the locview attribute is present, and
18541 whether they match. Remove the locview attribute if no longer
18543 (index_location_lists): Call skip_loc_list_entry for test.
18544 (dwarf2out_finish): Call output_asm_line_debug_info for test.
18545 Use output_dwarf_version.
18546 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
18547 (struct dw_val_node): Add val_view_list.
18548 * final.c (SEEN_NEXT_VIEW): New.
18549 (set_next_view_needed): New.
18550 (clear_next_view_needed): New.
18551 (maybe_output_next_view): New.
18552 (final_start_function): Rename to...
18553 (final_start_function_1): ... this. Take pointer to FIRST,
18554 add SEEN parameter. Emit param bindings in the initial view.
18555 (final_start_function): Reintroduce SEEN-less interface.
18556 (final): Rename to...
18557 (final_1): ... this. Take SEEN parameter. Output final pending
18558 next view at the end.
18559 (final): Reintroduce seen-less interface.
18560 (final_scan_insn): Output pending next view before switching
18561 sections or ending a block. Mark the next view as needed when
18562 outputting variable locations. Notify debug backend of section
18563 changes, and of location view changes.
18564 (rest_of_handle_final): Adjust.
18565 * toplev.c (process_options): Autodetect value for debug variable
18566 location views option. Warn on incompat5 without -gdwarf-5.
18567 * doc/invoke.texi (gvariable-location-views): New.
18568 (gvariable-location-views=incompat5): New.
18569 (gno-variable-location-views): New.
18571 2018-02-08 David Malcolm <dmalcolm@redhat.com>
18573 PR tree-optimization/84136
18574 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
18575 that the result of find_edge is non-NULL.
18577 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
18580 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
18581 storing integer register in SImode. Fix cost of 256 and 512
18582 byte aligned SSE register store.
18584 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
18586 * config/i386/i386.c (ix86_multiplication_cost): Fix
18587 multiplication cost for TARGET_AVX512DQ.
18589 2018-02-08 Marek Polacek <polacek@redhat.com>
18591 PR tree-optimization/84238
18592 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
18595 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
18597 PR tree-optimization/84265
18598 * tree-vect-stmts.c (vectorizable_store): Don't treat
18599 VMAT_CONTIGUOUS accesses as grouped.
18600 (vectorizable_load): Likewise.
18602 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
18604 PR tree-optimization/81635
18605 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
18606 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
18607 (test_round_for_mask): New functions.
18608 (wide_int_cc_tests): Call test_round_for_mask.
18609 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
18610 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
18611 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
18612 range returned by get_range_info.
18614 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
18617 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
18618 * symtab.c: Include builtins.h
18619 (symtab_node::output_to_lto_symbol_table_p): Move here
18620 from lto-streamer-out.c:output_symbol_p.
18621 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
18622 (output_symbol_p): Move all logic to symtab.c
18623 (produce_symtab): Update.
18625 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
18627 * config/s390/s390-opts.h (enum indirect_branch): Define.
18628 * config/s390/s390-protos.h (s390_return_addr_from_memory)
18629 (s390_indirect_branch_via_thunk)
18630 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
18631 (enum s390_indirect_branch_type): Define.
18632 * config/s390/s390.c (struct s390_frame_layout, struct
18633 machine_function): Remove.
18634 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
18635 (indirect_branch_table_label_no, indirect_branch_table_name):
18637 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
18638 (enum s390_indirect_branch_option): Define.
18639 (s390_return_addr_from_memory): New function.
18640 (s390_handle_string_attribute): New function.
18641 (s390_attribute_table): Add new attribute handler.
18642 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
18643 (s390_indirect_branch_via_thunk): New function.
18644 (s390_indirect_branch_via_inline_thunk): New function.
18645 (s390_function_ok_for_sibcall): When jumping via thunk disallow
18646 sibling call optimization for non z10 compiles.
18647 (s390_emit_call): Force indirect branch target to be a single
18648 register. Add r1 clobber for non-z10 compiles.
18649 (s390_emit_epilogue): Emit return jump via return_use expander.
18650 (s390_reorg): Handle JUMP_INSNs as execute targets.
18651 (s390_option_override_internal): Perform validity checks for the
18652 new command line options.
18653 (s390_indirect_branch_attrvalue): New function.
18654 (s390_indirect_branch_settings): New function.
18655 (s390_set_current_function): Invoke s390_indirect_branch_settings.
18656 (s390_output_indirect_thunk_function): New function.
18657 (s390_code_end): Implement target hook.
18658 (s390_case_values_threshold): Implement target hook.
18659 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
18661 * config/s390/s390.h (struct s390_frame_layout)
18662 (struct machine_function): Move here from s390.c.
18663 (TARGET_INDIRECT_BRANCH_NOBP_RET)
18664 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
18665 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
18666 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
18667 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
18668 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
18669 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
18670 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
18671 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
18672 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
18673 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
18674 (mnemonic attribute): Add values which aren't recognized
18676 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
18677 pattern for branch conversion. Fix mnemonic attribute.
18678 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
18679 indirect branch via thunk if requested.
18680 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
18681 ("*indirect_jump"): Disable for branch conversion using out of
18683 ("indirect_jump_via_thunk<mode>_z10")
18684 ("indirect_jump_via_thunk<mode>")
18685 ("indirect_jump_via_inlinethunk<mode>_z10")
18686 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
18687 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
18688 ("casesi_jump_via_inlinethunk<mode>_z10")
18689 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
18690 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
18691 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
18692 ("*indirect2_jump"): Disable for branch conversion.
18693 ("casesi_jump"): Turn into expander and expand patterns for branch
18695 ("return_use"): New expander.
18696 ("*return"): Emit return via thunk and rename it to ...
18697 ("*return<mode>"): ... this one.
18698 * config/s390/s390.opt: Add new options and and enum for the
18701 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
18703 * lra-constraints.c (match_reload): Unconditionally use
18704 gen_lowpart_SUBREG, rather than selecting between that
18705 and equivalent gen_rtx_SUBREG code.
18707 2018-02-08 Richard Biener <rguenther@suse.de>
18709 PR tree-optimization/84233
18710 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
18711 changed flag instead of boguously re-using phi_inserted.
18713 2018-02-08 Martin Jambor <mjambor@suse.cz>
18715 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
18716 static local variables.
18718 2018-02-08 Richard Biener <rguenther@suse.de>
18720 PR tree-optimization/84278
18721 * tree-vect-stmts.c (vectorizable_store): When looking for
18722 smaller vector types to perform grouped strided loads/stores
18723 make sure the mode is supported by the target.
18724 (vectorizable_load): Likewise.
18726 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
18728 * config/aarch64/aarch64.c (aarch64_components_for_bb):
18729 Increase LDP/STP opportunities by adding adjacent callee-saves.
18731 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
18733 PR rtl-optimization/84068
18734 PR rtl-optimization/83459
18735 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
18737 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
18739 PR tree-optimization/84224
18740 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
18741 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
18742 non-zero arguments.
18744 2018-02-07 Iain Sandoe <iain@codesourcery.com>
18747 * config/rs6000/altivec.md (*restore_world): Remove LR use.
18748 * config/rs6000/predicates.md (restore_world_operation): Adjust op
18749 count, remove one USE.
18751 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
18753 * doc/install.texi (Configuration): Document the
18754 --with-long-double-format={ibm,ieee} PowerPC configuration
18758 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
18759 Convert from define_expand to be define_insn_and_split. Rework
18760 float/double/_Float128 conversions to QI/HI/SImode to work with
18761 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
18762 conversions to QI/HImode types did a store and then a load to
18763 truncate the value. For conversions to VSX registers, don't split
18764 the insn, instead emit the code directly. Use the code iterator
18765 any_fix to combine signed and unsigned conversions.
18766 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
18767 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
18768 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
18769 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
18770 (fix_<mode>di2_hw): Likewise.
18771 (fixuns_<mode>di2_hw): Likewise.
18772 (fix_<mode>si2_hw): Likewise.
18773 (fixuns_<mode>si2_hw): Likewise.
18774 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
18775 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
18776 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
18777 fix<uns>_trunc<SFDF:mode>si2_p8.
18778 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
18779 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
18780 (fix<uns>_<mode>_mem): Likewise.
18781 (fctiw<u>z_<mode>_mem): Likewise.
18782 (fix<uns>_<mode>_mem): Likewise.
18783 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
18784 the register allocator from doing a direct move to the GPRs to do
18785 a store, and instead use the ISA 3.0 store byte/half-word from
18786 vector register instruction. For IEEE 128-bit floating point,
18787 also optimize stores of 32-bit ints.
18788 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
18790 2018-02-07 Alan Hayward <alan.hayward@arm.com>
18792 * genextract.c (push_pathstr_operand): New function to support
18794 (walk_rtx): Call push_pathstr_operand.
18795 (print_path): Support [a-zA-Z].
18797 2018-02-07 Richard Biener <rguenther@suse.de>
18799 PR tree-optimization/84037
18800 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
18801 (cse_and_gimplify_to_preheader): Declare.
18802 (vect_get_place_in_interleaving_chain): Likewise.
18803 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
18805 (_loop_vec_info::~_loop_vec_info): Delete it.
18806 (cse_and_gimplify_to_preheader): New function.
18807 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
18808 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
18809 (vectorizable_load): Likewise. For grouped stores always base
18810 the IV on the first element.
18811 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
18812 condition before gimplifying.
18814 2018-02-07 Jakub Jelinek <jakub@redhat.com>
18816 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
18817 *DIV_EXPR and *MOD_EXPR.
18819 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
18822 * config/i386/i386.c (ix86_option_override_internal): Mask out
18823 the CF_SET bit when checking -fcf-protection.
18825 2018-02-07 Tom de Vries <tom@codesourcery.com>
18828 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
18831 2018-02-07 Richard Biener <rguenther@suse.de>
18833 PR tree-optimization/84204
18834 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
18837 PR tree-optimization/84205
18838 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
18839 special-case isl_ast_op_zdiv_r.
18841 PR tree-optimization/84223
18842 * graphite-scop-detection.c (gather_bbs::before_dom_children):
18843 Only add conditions from within the region.
18844 (gather_bbs::after_dom_children): Adjust.
18846 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
18849 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
18850 * config/avr/avr.md: Only post-reload split REG-REG moves if
18851 either register is GENERAL_REG_P.
18853 2018-02-07 Jakub Jelinek <jakub@redhat.com>
18855 PR tree-optimization/84235
18856 * tree-ssa-scopedtables.c
18857 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
18858 if the subtraction is performed in floating point type where NaNs are
18859 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
18860 build 1. Formatting fix.
18862 2018-02-06 Jakub Jelinek <jakub@redhat.com>
18865 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
18866 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
18867 and skip it regardless of bb boundaries. Use CALL_P macro,
18868 don't test INSN_P (insn) together with CALL_P or JUMP_P check
18869 unnecessarily, formatting fix.
18871 2018-02-06 Michael Collison <michael.collison@arm.com>
18873 * config/arm/thumb2.md:
18874 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
18875 (*thumb_mov_notscc): Ditto.
18877 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
18880 * config/rs6000/rs6000.md (su code attribute): Use "u" for
18881 unsigned_fix, not "s".
18883 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
18885 * configure.ac (gcc_fn_eh_frame_ro): New function.
18886 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
18887 correct .eh_frame permissions.
18888 * configure: Regenerate.
18890 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
18892 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
18893 irrelevant options.
18895 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18897 * config/rs6000/rs6000.c (rs6000_option_override_internal):
18898 Display warning message for -mno-speculate-indirect-jumps.
18900 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
18902 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
18904 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
18906 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
18908 PR tree-optimization/84225
18909 * tree-eh.c (find_trapping_overflow): Only call
18910 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
18912 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
18915 * config/i386/i386.c: Reimplement the check of possible options
18916 -mibt/-mshstk conbination. Change error messages.
18917 * doc/invoke.texi: Fix a typo: remove extra '='.
18919 2018-02-06 Marek Polacek <polacek@redhat.com>
18921 PR tree-optimization/84228
18922 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
18924 2018-02-06 Tamar Christina <tamar.christina@arm.com>
18927 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
18928 emitted arch directives.
18929 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
18930 __ARM_FEATURE_COPROC before changing architectures.
18932 2018-02-06 Richard Biener <rguenther@suse.de>
18934 * config/i386/i386.c (print_reg): Fix typo.
18935 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
18937 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
18939 * configure: Regenerate.
18941 2018-02-05 Martin Sebor <msebor@redhat.com>
18943 PR tree-optimization/83369
18944 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
18947 2018-02-05 Martin Liska <mliska@suse.cz>
18949 * doc/invoke.texi: Cherry-pick upstream r323995.
18951 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
18953 * ira.c (ira_init_register_move_cost): Adjust comment.
18955 2018-02-05 Martin Liska <mliska@suse.cz>
18957 PR gcov-profile/84137
18958 * doc/gcov.texi: Fix typo in documentation.
18960 2018-02-05 Martin Liska <mliska@suse.cz>
18962 PR gcov-profile/83879
18963 * doc/gcov.texi: Document necessity of --dynamic-list-data when
18964 using dlopen functionality.
18966 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
18968 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
18969 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
18970 _mm_maskz_range_ss, _mm_mask_range_round_ss,
18971 _mm_maskz_range_round_ss): New intrinsics.
18972 (__builtin_ia32_rangesd128_round)
18973 (__builtin_ia32_rangess128_round): Remove.
18974 (__builtin_ia32_rangesd128_mask_round,
18975 __builtin_ia32_rangess128_mask_round): New builtins.
18976 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
18977 __builtin_ia32_rangess128_round): Remove.
18978 (__builtin_ia32_rangesd128_mask_round,
18979 __builtin_ia32_rangess128_mask_round): New builtins.
18980 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
18981 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
18982 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
18983 "<round_saeonly_constraint>")): Changed to ...
18984 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
18985 "<round_saeonly_scalar_constraint>")): ... this.
18986 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
18987 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
18988 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
18989 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
18990 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
18992 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
18994 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
18996 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
18997 Remove all values except native, 8540 and 8548.
18999 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
19001 * config/i386/i386.c (ix86_output_function_return): Pass
19002 INVALID_REGNUM, instead of -1, as invalid register number to
19003 indirect_thunk_name and output_indirect_thunk.
19005 2018-02-02 Julia Koval <julia.koval@intel.com>
19007 * config.gcc: Add -march=icelake.
19008 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
19009 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
19010 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
19011 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
19012 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
19013 (processor_target_table): Add icelake.
19014 (ix86_option_override_internal): Handle new PTAs.
19015 (get_builtin_code_for_version): Handle icelake.
19016 (M_INTEL_COREI7_ICELAKE): New.
19017 (fold_builtin_cpu): Handle icelake.
19018 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
19019 * doc/invoke.texi: Add -march=icelake.
19021 2018-02-02 Julia Koval <julia.koval@intel.com>
19023 * config/i386/i386.c (ix86_option_override_internal): Change flags type
19024 to wide_int_bitmask.
19025 * wide-int-bitmask.h: New.
19027 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
19030 * config/i386/i386.md: Replace Pmode with word_mode in
19031 builtin_setjmp_setup and builtin_longjmp to support x32.
19033 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
19037 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
19039 (rs6000_supported_cpu_names): New static variable.
19040 (linux_cpu_translation_table): Likewise.
19041 (elf_platform) <cpu>: Define new static variable and use it.
19042 Translate kernel AT_PLATFORM name to canonical name if needed.
19043 Error if platform name is unknown.
19045 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
19048 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
19050 2018-02-01 Jeff Law <law@redhat.com>
19053 * config/i386/i386.c (release_scratch_register_on_entry): Add new
19054 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
19055 the scratch if RELEASE_VIA_POP is false.
19056 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
19057 If we have to save a temporary register, decrement SIZE appropriately.
19058 Pass new arguments to release_scratch_register_on_entry.
19059 (ix86_adjust_stack_and_probe): Likewise.
19060 (ix86_emit_probe_stack_range): Pass new arguments to
19061 release_scratch_register_on_entry.
19063 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
19065 PR rtl-optimization/84157
19066 * combine.c (change_zero_ext): Use REG_P predicate in
19067 front of HARD_REGISTER_P predicate.
19069 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
19071 * config/avr/avr.c (avr_option_override): Move disabling of
19072 -fdelete-null-pointer-checks to...
19073 * common/config/avr/avr-common.c (avr_option_optimization_table):
19076 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
19078 PR tree-optimization/81635
19079 * tree-data-ref.c (split_constant_offset_1): For types that
19080 wrap on overflow, try to use range info to prove that wrapping
19083 2018-02-01 Renlin Li <renlin.li@arm.com>
19086 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
19087 TAILCALL_ADDR_REGS.
19088 (aarch64_register_move_cost): Likewise.
19089 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
19090 TAILCALL_ADDR_REGS.
19091 (REG_CLASS_NAMES): Likewise.
19092 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
19093 TAILCALL_ADDR_REGS. Remove IP registers.
19094 * config/aarch64/aarch64.md (Ucs): Update register constraint.
19096 2018-02-01 Richard Biener <rguenther@suse.de>
19098 * domwalk.h (dom_walker::dom_walker): Add additional constructor
19099 for specifying RPO order and allow NULL for that.
19100 * domwalk.c (dom_walker::dom_walker): Likewise.
19101 (dom_walker::walk): Handle NULL RPO order.
19102 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
19104 (rewrite_update_dom_walker): Likewise.
19105 (mark_def_dom_walker): Likewise.
19107 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
19109 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
19110 (aarch64_maybe_expand_sve_subreg_move): Declare.
19111 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
19112 * config/aarch64/predicates.md (aarch64_any_register_operand): New
19114 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
19115 that are semantically a reverse operation.
19116 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
19117 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
19118 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
19120 (aarch64_can_change_mode_class): For big-endian, forbid changes
19121 between two SVE modes if they have different element sizes.
19123 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
19125 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
19126 the TImode handling for big-endian targets.
19128 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
19130 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
19131 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
19133 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
19134 Remove BSWAP handing for big-endian targets and use the form of
19135 LD1RQ appropariate for the mode.
19137 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
19139 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
19140 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
19141 duplicated element.
19143 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
19146 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
19147 check for operands that need to go through aarch64_sve_reload_be.
19149 2018-02-01 Jakub Jelinek <jakub@redhat.com>
19151 PR tree-optimization/81661
19152 PR tree-optimization/84117
19153 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
19154 * tree-eh.c: Include gimplify.h.
19155 (find_trapping_overflow, replace_trapping_overflow,
19156 rewrite_to_non_trapping_overflow): New functions.
19157 * tree-vect-loop.c: Include tree-eh.h.
19158 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
19159 * tree-data-ref.c: Include tree-eh.h.
19160 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
19162 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
19164 PR rtl-optimization/84123
19165 * combine.c (change_zero_ext): Check if hard register satisfies
19166 can_change_dest_mode before calling gen_lowpart_SUBREG.
19168 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
19171 * ira.c (ira_init_register_move_cost): Remove assert.
19173 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
19175 PR rtl-optimization/84071
19176 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
19177 * doc/tm.texi: Regenerate.
19179 2018-01-31 Richard Biener <rguenther@suse.de>
19181 PR tree-optimization/84132
19182 * tree-data-ref.c (analyze_miv_subscript): Properly
19183 check whether evolution_function_is_affine_multivariate_p
19184 before calling gcd_of_steps_may_divide_p.
19186 2018-01-31 Julia Koval <julia.koval@intel.com>
19189 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
19190 * config/i386/i386.md (rdpid_rex64) New.
19191 (rdpid): Make 32bit only.
19193 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
19196 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
19197 an IDENTIFIER_NODE for FUNCTION_TYPE's.
19199 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
19202 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
19204 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
19206 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
19208 PR rtl-optimization/84071
19209 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
19210 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
19212 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
19214 * config/arc/arc.c (arc_handle_aux_attribute): New function.
19215 (arc_attribute_table): Add 'aux' attribute.
19216 (arc_in_small_data_p): Consider aux like variables.
19217 (arc_is_aux_reg_p): New function.
19218 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
19219 (arc_get_aux_arg): New function.
19220 (prepare_move_operands): Handle aux-register access.
19221 (arc_handle_aux_attribute): New function.
19222 * doc/extend.texi (ARC Variable attributes): Add subsection.
19224 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
19226 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
19227 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
19228 (arc_attribute_table): Add 'uncached' attribute.
19229 (arc_print_operand): Print '.di' flag for uncached memory
19231 (arc_in_small_data_p): Do not consider for small data the uncached
19233 (arc_is_uncached_mem_p): New function.
19234 * config/arc/predicates.md (compact_store_memory_operand): Check
19235 for uncached memory accesses.
19236 (nonvol_nonimm_operand): Likewise.
19237 * doc/extend.texi (ARC Type Attribute): New subsection.
19239 2018-01-31 Jakub Jelinek <jakub@redhat.com>
19242 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
19243 falign-loops=): Add Optimization flag.
19245 2018-01-30 Jeff Law <law@redhat.com>
19248 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
19249 INT_REGISTERS_SAVED. Check it prior to calling
19250 get_scratch_register_on_entry.
19251 (ix86_adjust_stack_and_probe): Similarly.
19252 (ix86_emit_probe_stack_range): Similarly.
19253 (ix86_expand_prologue): Corresponding changes.
19255 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
19258 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
19259 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
19261 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
19264 * lra-constraints.c (curr_insn_transform): Process AND in the
19267 2018-01-30 Jakub Jelinek <jakub@redhat.com>
19269 PR rtl-optimization/83986
19270 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
19271 dependence against last_pending_memory_flush in addition to
19272 pending_jump_insns.
19274 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
19276 PR tree-optimization/81611
19277 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
19280 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
19283 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
19286 2018-01-30 Richard Biener <rguenther@suse.de>
19287 Jakub Jelinek <jakub@redhat.com>
19289 PR tree-optimization/84111
19290 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
19291 inner loops added during recursion, as they don't have up-to-date
19294 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
19297 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
19298 (can_inline_edge_by_limits_p): ... here.
19299 (can_early_inline_edge_p, check_callers,
19300 update_caller_keys, update_callee_keys, recursive_inlining,
19301 add_new_edges_to_heap, speculation_useful_p,
19302 inline_small_functions,
19303 inline_small_functions, flatten_function,
19304 inline_to_all_callers_1): Update.
19306 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
19308 * profile-count.c (profile_count::combine_with_ipa_count): Handle
19311 2018-01-30 Richard Biener <rguenther@suse.de>
19313 PR tree-optimization/83008
19314 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
19315 invariant and constant vector uses in stmts when they need
19316 more than one stmt.
19318 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
19321 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
19322 * configure: Regenerate.
19324 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
19326 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
19328 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
19329 Use gen_rtx_REG rather than gen_lowpart.
19331 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
19333 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
19334 rather than 0 when creating partial subregs.
19336 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
19338 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
19341 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
19344 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
19345 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
19346 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
19347 flags. This restores the settings used before the 2017-07-24.
19348 Turning off pre increment/decrement/modify allows IVOPTS to
19349 optimize DF/SF loops where the index is an int.
19351 2018-01-29 Richard Biener <rguenther@suse.de>
19352 Kelvin Nilsen <kelvin@gcc.gnu.org>
19355 * tree-vect-stmts.c (vectorizable_call): Don't call
19356 targetm.vectorize_builtin_md_vectorized_function if callee is
19359 2018-01-22 Carl Love <cel@us.ibm.com>
19361 * doc/extend.tex: Fix typo in second arg in
19362 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
19364 2018-01-29 Richard Biener <rguenther@suse.de>
19366 PR tree-optimization/84086
19367 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
19368 (flush_ssaname_freelist): When SSA names were released reset
19369 the SCEV hash table.
19371 2018-01-29 Richard Biener <rguenther@suse.de>
19373 PR tree-optimization/84057
19374 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
19375 removed paths when removing edges.
19377 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
19379 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
19380 -mfunction-return=@var{choice}.
19382 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
19384 PR diagnostic/84034
19385 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
19386 Handle CR like TAB.
19387 (layout::print_source_line): Likewise.
19388 (test_get_line_width_without_trailing_whitespace): Add test cases.
19390 2018-01-27 Jakub Jelinek <jakub@redhat.com>
19392 PR middle-end/84040
19393 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
19396 2018-01-26 Jim Wilson <jimw@sifive.com>
19398 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
19400 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
19403 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19405 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
19406 and CMP + SUB-immediate -> SUBS.
19408 2018-01-26 Martin Sebor <msebor@redhat.com>
19410 PR tree-optimization/83896
19411 * tree-ssa-strlen.c (get_string_len): Rename...
19412 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
19413 Avoid assuming length is constant.
19414 (handle_char_store): Use HOST_WIDE_INT for string length.
19416 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
19419 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
19420 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
19422 2018-01-26 Richard Biener <rguenther@suse.de>
19424 PR rtl-optimization/84003
19425 * dse.c (record_store): Only record redundant stores when
19426 the earlier store aliases at least all accesses the later one does.
19428 2018-01-26 Jakub Jelinek <jakub@redhat.com>
19430 PR rtl-optimization/83985
19431 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
19432 REG_CFA_RESTORE insns.
19433 (delete_unmarked_insns): Don't ignore separate shrink wrapping
19434 REG_CFA_RESTORE insns here.
19437 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
19438 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
19440 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
19442 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
19443 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
19444 (arc_init): Likewise.
19445 (arc_override_options): Likewise.
19446 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
19448 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
19450 * config/arc/arc.h (TARGET_DBNZ): Define.
19451 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
19452 properly set the tune attribute.
19453 (dbnz): Use TARGET_DBNZ guard.
19454 * config/arc/arc.opt (mtune): Add core3 option.
19456 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
19458 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
19459 recognize new pic like addresses.
19460 (arc_delegitimize_address): Clean up.
19462 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
19464 * config/arc/arc-arches.def: Option mrf16 valid for all
19466 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
19467 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
19468 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
19469 * config/arc/arc-tables.opt: Regenerate.
19470 * config/arc/arc.c (arc_conditional_register_usage): Handle
19471 reduced register file case.
19472 (arc_file_start): Set must have build attributes.
19473 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
19474 mrf16 option value.
19475 * config/arc/arc.opt (mrf16): Add new option.
19476 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
19477 * config/arc/genmultilib.awk: Handle new mrf16 option.
19478 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
19479 * config/arc/t-multilib: Regenerate.
19480 * doc/invoke.texi (ARC Options): Document mrf16 option.
19482 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
19484 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
19485 * config/arc/arc.c (arc_handle_secure_attribute): New function.
19486 (arc_attribute_table): Add 'secure_call' attribute.
19487 (arc_print_operand): Print secure call operand.
19488 (arc_function_ok_for_sibcall): Don't optimize tail calls when
19490 (arc_is_secure_call_p): New function. * config/arc/arc.md
19491 (call_i): Add support for sjli instruction.
19492 (call_value_i): Likewise.
19493 * config/arc/constraints.md (Csc): New constraint.
19495 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
19496 John Eric Martin <John.Martin@emmicro-us.com>
19498 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
19499 * config/arc/arc.c (_arc_jli_section): New struct.
19500 (arc_jli_section): New type.
19501 (rc_jli_sections): New static variable.
19502 (arc_handle_jli_attribute): New function.
19503 (arc_attribute_table): Add jli_always and jli_fixed attribute.
19504 (arc_file_end): New function.
19505 (TARGET_ASM_FILE_END): Define.
19506 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
19507 (arc_add_jli_section): New function.
19508 (jli_call_scan): Likewise.
19509 (arc_reorg): Call jli_call_scan.
19510 (arc_output_addsi): Remove 'S' from printing asm operand.
19511 (arc_is_jli_call_p): New function.
19512 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
19514 (movhi_insn): Likewise.
19515 (movsi_insn): Likewise.
19516 (movsi_set_cc_insn): Likewise.
19517 (loadqi_update): Likewise.
19518 (load_zeroextendqisi_update): Likewise.
19519 (load_signextendqisi_update): Likewise.
19520 (loadhi_update): Likewise.
19521 (load_zeroextendhisi_update): Likewise.
19522 (load_signextendhisi_update): Likewise.
19523 (loadsi_update): Likewise.
19524 (loadsf_update): Likewise.
19525 (movsicc_insn): Likewise.
19526 (bset_insn): Likewise.
19527 (bxor_insn): Likewise.
19528 (bclr_insn): Likewise.
19529 (bmsk_insn): Likewise.
19530 (bicsi3_insn): Likewise.
19531 (cmpsi_cc_c_insn): Likewise.
19532 (movsi_ne): Likewise.
19533 (movsi_cond_exec): Likewise.
19534 (clrsbsi2): Likewise.
19535 (norm_f): Likewise.
19545 (call_i): Remove 'S' asm letter, add jli instruction.
19546 (call_value_i): Likewise.
19547 * config/arc/arc.op (mjli-always): New option.
19548 * config/arc/constraints.md (Cji): New constraint.
19549 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
19551 (subsf3_fpx): Likewise.
19552 (mulsf3_fpx): Likewise.
19553 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
19555 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
19556 function attrbutes.
19557 * doc/invoke.texi (ARC): Document mjli-always option.
19559 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
19561 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
19562 avoid addition with 0 and use incw and decw where possible.
19564 2018-01-26 Richard Biener <rguenther@suse.de>
19566 PR tree-optimization/81082
19567 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
19568 association if it requires casting to unsigned.
19569 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
19570 from fold_plusminus_mult_expr to catch important cases late when
19571 range info is available.
19573 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
19575 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
19576 * configure.ac (hidden_linkonce): New test.
19577 * configure: Regenerate.
19578 * config.in: Regenerate.
19580 2018-01-26 Julia Koval <julia.koval@intel.com>
19582 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
19583 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
19584 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
19585 _mm_mask_bitshuffle_epi64_mask): Fix type.
19586 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
19587 USI_FTYPE_V4DI_V4DI_USI): Remove.
19588 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
19589 __builtin_ia32_vpshufbitqmb256_mask,
19590 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
19591 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
19592 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
19594 2018-01-26 Alan Modra <amodra@gmail.com>
19597 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
19598 UNSPEC_VBPERMQ. Sort other unspecs.
19600 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
19602 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
19604 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
19606 PR middle-end/83055
19607 * predict.c (drop_profile): Do not push/pop cfun; update also
19609 (handle_missing_profiles): Fix logic looking for zero profiles.
19611 2018-01-25 Jakub Jelinek <jakub@redhat.com>
19613 PR middle-end/83977
19614 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
19615 on functions with #pragma omp declare simd or functions with simd
19617 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
19618 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
19619 Remove trailing \n from warning_at calls.
19621 2018-01-25 Tom de Vries <tom@codesourcery.com>
19624 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
19625 for neutered workers.
19627 2018-01-24 Joseph Myers <joseph@codesourcery.com>
19630 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
19631 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
19633 2018-01-24 Jeff Law <law@redhat.com>
19636 * i386.c (get_probe_interval): Move to earlier point.
19637 (ix86_compute_frame_layout): If -fstack-clash-protection and
19638 the frame is larger than the probe interval, then use pushes
19639 to save registers rather than reg->mem moves.
19640 (ix86_expand_prologue): Remove conditional for int_registers_saved
19643 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
19646 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
19647 min/max for never referenced object.
19649 2018-01-24 Jakub Jelinek <jakub@redhat.com>
19651 PR middle-end/83977
19652 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
19654 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
19655 attributes from DECL_ATTRIBUTES (decl) without affecting
19656 DECL_ATTRIBUTES (current_function_decl).
19657 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
19658 functions with non-NULL DECL_ABSTRACT_ORIGIN.
19660 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
19662 PR tree-optimization/83979
19663 * fold-const.c (fold_comparison): Use constant_boolean_node
19664 instead of boolean_{true,false}_node.
19666 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
19668 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
19671 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
19673 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
19674 Simplify the clause that sets the length attribute.
19675 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
19676 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
19677 clause that sets the length attribute.
19678 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
19680 2018-01-24 Tom de Vries <tom@codesourcery.com>
19683 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
19684 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
19685 Add strict parameter.
19686 (prevent_branch_around_nothing): Insert dummy insn between branch to
19687 label and label with no ptx insn inbetween.
19688 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
19690 2018-01-24 Tom de Vries <tom@codesourcery.com>
19693 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
19694 for neutered threads in warp.
19695 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
19697 2018-01-24 Richard Biener <rguenther@suse.de>
19699 PR tree-optimization/83176
19700 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
19703 2018-01-24 Richard Biener <rguenther@suse.de>
19705 PR tree-optimization/82819
19706 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
19707 code generating pluses that are no-ops in the target precision.
19709 2018-01-24 Richard Biener <rguenther@suse.de>
19711 PR middle-end/84000
19712 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
19714 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
19716 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
19717 to merge probabilities.
19718 * predict.c (probably_never_executed): Also mark as cold functions
19719 with global 0 profile and guessed local profile.
19720 * profile-count.c (profile_probability::combine_with_count): New
19722 * profile-count.h (profile_probability::operator*,
19723 profile_probability::operator*=, profile_probability::operator/,
19724 profile_probability::operator/=): Reduce precision to adjusted
19725 and set value to guessed on contradictory divisions.
19726 (profile_probability::combine_with_freq): Remove.
19727 (profile_probability::combine_wiht_count): Declare.
19728 (profile_count::force_nonzero):: Set to adjusted.
19729 (profile_count::probability_in):: Set quality to adjusted.
19730 * tree-ssa-tail-merge.c (replace_block_by): Use
19731 combine_with_count.
19733 2018-01-23 Andrew Waterman <andrew@sifive.com>
19734 Jim Wilson <jimw@sifive.com>
19736 * config/riscv/riscv.c (riscv_stack_boundary): New.
19737 (riscv_option_override): Set riscv_stack_boundary. Handle
19738 riscv_preferred_stack_boundary_arg.
19739 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
19740 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
19741 (STACK_BOUNDARY): Set to riscv_stack_boundary.
19742 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
19743 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
19744 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
19746 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
19749 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
19750 of struct ix86_frame.
19751 (ix86_expand_epilogue): Likewise. Add a local variable for
19752 the reg_save_offset field in struct ix86_frame.
19754 2018-01-23 Bin Cheng <bin.cheng@arm.com>
19756 PR tree-optimization/82604
19757 * tree-loop-distribution.c (enum partition_kind): New enum item
19758 PKIND_PARTIAL_MEMSET.
19759 (partition_builtin_p): Support above new enum item.
19760 (generate_code_for_partition): Ditto.
19761 (compute_access_range): Differentiate cases that equality can be
19762 proven at all loops, the innermost loops or no loops.
19763 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
19764 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
19765 (finalize_partitions, distribute_loop): Don't fuse partition of
19766 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
19767 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
19768 parloop is enabled.
19770 2018-01-23 Martin Liska <mliska@suse.cz>
19772 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
19773 order to ignore the predictor.
19774 (PRED_POLYMORPHIC_CALL): Likewise.
19775 (PRED_RECURSIVE_CALL): Likewise.
19777 2018-01-23 Martin Liska <mliska@suse.cz>
19779 * tree-profile.c (tree_profiling): Print function header to
19780 aware reader which function we are working on.
19781 * value-prof.c (gimple_find_values_to_profile): Do not print
19782 not interesting value histograms.
19784 2018-01-23 Martin Liska <mliska@suse.cz>
19786 * profile-count.h (enum profile_quality): Add
19787 profile_uninitialized as the first value. Do not number values
19788 as they are zero based.
19789 (profile_count::verify): Update sanity check.
19790 (profile_probability::verify): Likewise.
19792 2018-01-23 Nathan Sidwell <nathan@acm.org>
19794 * doc/invoke.texi (ffor-scope): Deprecate.
19796 2018-01-23 David Malcolm <dmalcolm@redhat.com>
19798 PR tree-optimization/83510
19799 * domwalk.c (set_all_edges_as_executable): New function.
19800 (dom_walker::dom_walker): Convert bool param
19801 "skip_unreachable_blocks" to enum reachability. Move setup of
19802 edge flags to set_all_edges_as_executable and only do it when
19803 reachability is REACHABLE_BLOCKS.
19804 * domwalk.h (enum dom_walker::reachability): New enum.
19805 (dom_walker::dom_walker): Convert bool param
19806 "skip_unreachable_blocks" to enum reachability.
19807 (set_all_edges_as_executable): New decl.
19808 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
19809 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
19811 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
19812 but converting true to REACHABLE_BLOCKS.
19813 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
19815 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
19816 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
19817 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
19819 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
19820 if check_all_array_refs will be called.
19822 2018-01-23 David Malcolm <dmalcolm@redhat.com>
19824 * tree.c (selftest::test_location_wrappers): Add more test
19827 2018-01-23 David Malcolm <dmalcolm@redhat.com>
19829 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
19830 (selftest::test_bit_in_range): Likewise.
19832 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
19835 * doc/sourcebuild.texi (vect_float): Say that the selector
19836 only describes the situation when -funsafe-math-optimizations is on.
19837 (vect_float_strict): Document.
19839 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
19841 PR tree-optimization/83965
19842 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
19843 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
19844 instead of checking only for a reduction.
19845 (vect_recog_widen_sum_pattern): Likewise.
19847 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
19849 * predict.c (probably_never_executed): Only use precise profile info.
19850 (compute_function_frequency): Skip after inlining hack since we now
19851 have quality checking.
19853 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
19855 * profile-count.h (profile_probability::very_unlikely,
19856 profile_probability::unlikely, profile_probability::even): Set
19857 precision to guessed.
19859 2018-01-23 Richard Biener <rguenther@suse.de>
19861 PR tree-optimization/83963
19862 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
19863 Properly terminate dominator walk when crossing the exit edge not
19864 when visiting its source block.
19866 2018-01-23 Jakub Jelinek <jakub@redhat.com>
19869 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
19870 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
19872 2018-01-22 Jakub Jelinek <jakub@redhat.com>
19874 PR tree-optimization/83957
19875 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
19876 semicolon after for body surrounded by braces.
19878 PR tree-optimization/83081
19879 * profile-count.h (profile_probability::split): New method.
19880 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
19881 Use profile_probability::split.
19882 (do_compare_rtx_and_jump): Fix adjustment of probabilities
19883 when splitting a single conditional jump into 2.
19885 2018-01-22 David Malcolm <dmalcolm@redhat.com>
19887 PR tree-optimization/69452
19888 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
19891 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
19893 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
19894 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
19895 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
19897 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
19899 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
19901 * config/rl78/rl78.md (movdi): New define_expand.
19902 * config/rl78/rl78.c (rl78_split_movdi): New function.
19904 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
19907 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
19909 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
19910 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
19911 128-bit to produce an UNSPEC move to get the double word with the
19912 signbit and then a shift directly to do signbit.
19913 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
19914 implementation with a new version that just does either a direct
19915 move or a regular move. Move memory interface to separate insns.
19916 Move insns so they are next to the expander.
19917 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
19918 with signbit move. Split big and little endian case.
19919 (signbit<mode>2_dm_mem_le): Likewise.
19920 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
19921 (signbit<mode>2_dm2): Likewise.
19923 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
19925 * config/rl78/rl78.md (anddi3): New define_expand.
19927 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
19929 * config/rl78/rl78.md (umindi3): New define_expand.
19931 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
19933 * config/rl78/rl78.md (smindi3): New define_expand.
19935 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
19937 * config/rl78/rl78.md (smaxdi3): New define_expand.
19939 2018-01-22 Carl Love <cel@us.ibm.com>
19941 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
19942 LVX_V1TI): Add macro expansion.
19943 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
19944 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
19945 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
19946 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
19947 Change check to determine if the instruction is a byte reversing
19948 entry. Fix typo in comment.
19949 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
19950 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
19951 Add def_builtin calls for new builtins.
19952 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
19953 Add define_insn expansion.
19955 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
19957 * config/rl78/rl78.md (umaxdi3): New define_expand.
19959 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
19961 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
19962 for non-QImode registers.
19964 2018-01-22 Richard Biener <rguenther@suse.de>
19966 PR tree-optimization/83963
19967 * graphite-scop-detection.c (scop_detection::get_sese): Delay
19968 including the loop exit block.
19969 (scop_detection::merge_sese): Likewise.
19970 (scop_detection::add_scop): Do it here instead.
19972 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19974 * doc/sourcebuild.texi (arm_softfloat): Document.
19976 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
19979 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
19980 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
19981 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
19983 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
19984 David Edelsohn <dje.gcc@gmail.com>
19987 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
19988 Change "crset eq" to "crset 2".
19989 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
19990 (*call_indirect_aix<mode>_nospec): Likewise.
19991 (*call_value_indirect_aix<mode>_nospec): Likewise.
19992 (*call_indirect_elfv2<mode>_nospec): Likewise.
19993 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
19994 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
19995 change assembly output from . to $.
19996 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
19997 (indirect_jump<mode>_nospec): Change assembly output from . to $.
19998 (*tablejump<mode>_internal1_nospec): Likewise.
20000 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
20003 * config/sh/sh_optimize_sett_clrt.cc:
20004 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
20006 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
20008 PR tree-optimization/83940
20009 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
20010 offset_dt to vect_constant_def rather than vect_unknown_def_type.
20011 (vect_check_load_store_mask): Add a mask_dt_out parameter and
20012 use it to pass back the definition type.
20013 (vect_check_store_rhs): Likewise rhs_dt_out.
20014 (vect_build_gather_load_calls): Add a mask_dt argument and use
20015 it instead of a call to vect_is_simple_use.
20016 (vectorizable_store): Update calls to vect_check_load_store_mask
20017 and vect_check_store_rhs. Use the dt returned by the latter instead
20018 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
20019 instead of calls to vect_is_simple_use. Pass the scalar rather
20020 than the vector operand to vect_is_simple_use when handling
20021 second and subsequent copies of an rhs value.
20022 (vectorizable_load): Update calls to vect_check_load_store_mask
20023 and vect_build_gather_load_calls. Use the cached mask_dt and
20024 gs_info.offset_dt instead of calls to vect_is_simple_use.
20026 2018-01-20 Jakub Jelinek <jakub@redhat.com>
20028 PR middle-end/83945
20029 * tree-emutls.c: Include gimplify.h.
20030 (lower_emutls_2): New function.
20031 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
20032 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
20033 it before further processing.
20036 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
20037 UINTVAL (trueop1) instead of INTVAL (op1).
20039 2018-01-19 Jakub Jelinek <jakub@redhat.com>
20043 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
20044 INCOMING_FRAME_SP_OFFSET if not defined.
20045 (scan_trace): Add ENTRY argument. If true and
20046 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
20047 emit a note to adjust the CFA offset.
20048 (create_cfi_notes): Adjust scan_trace callers.
20049 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
20050 INCOMING_FRAME_SP_OFFSET in the CIE.
20051 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
20052 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
20054 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
20055 * doc/tm.texi: Regenerated.
20057 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
20059 PR rtl-optimization/83147
20060 * lra-constraints.c (remove_inheritance_pseudos): Use
20061 lra_substitute_pseudo_within_insn.
20063 2018-01-19 Tom de Vries <tom@codesourcery.com>
20064 Cesar Philippidis <cesar@codesourcery.com>
20067 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
20069 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
20072 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
20073 spaces for function labels.
20075 2018-01-19 Martin Liska <mliska@suse.cz>
20077 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
20078 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
20079 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
20080 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
20081 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
20082 (PRED_CONST_RETURN): Change from 69 to 65.
20083 (PRED_NULL_RETURN): Change from 91 to 71.
20084 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
20085 (PRED_LOOP_GUARD): Change from 66 to 73.
20087 2018-01-19 Martin Liska <mliska@suse.cz>
20089 * predict.c (predict_insn_def): Add new assert.
20090 (struct branch_predictor): Change type to signed integer.
20091 (test_prediction_value_range): Amend test to cover
20092 PROB_UNINITIALIZED.
20093 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
20094 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
20095 (PRED_LOOP_ITERATIONS_MAX): Likewise.
20096 (PRED_LOOP_IV_COMPARE): Likewise.
20097 * predict.h (PROB_UNINITIALIZED): Define new constant.
20099 2018-01-19 Martin Liska <mliska@suse.cz>
20101 * predict.c (dump_prediction): Add new format for
20102 analyze_brprob.py script which is enabled with -details
20104 * profile-count.h (precise_p): New function.
20106 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
20108 PR tree-optimization/83922
20109 * tree-vect-loop.c (vect_verify_full_masking): Return false if
20110 there are no statements that need masking.
20111 (vect_active_double_reduction_p): New function.
20112 (vect_analyze_loop_operations): Use it when handling phis that
20113 are not in the loop header.
20115 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
20117 PR tree-optimization/83914
20118 * tree-vect-loop.c (vectorizable_induction): Don't convert
20119 init_expr or apply the peeling adjustment for inductions
20120 that are nested within the vectorized loop.
20122 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20124 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
20127 2018-01-18 Jakub Jelinek <jakub@redhat.com>
20131 * function.h (gimplify_parameters): Add gimple_seq * argument.
20132 * function.c: Include gimple.h and options.h.
20133 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
20134 for the added local temporaries if needed.
20135 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
20136 if there are any parameter cleanups, wrap whole body into a
20137 try/finally with the cleanups.
20139 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
20142 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
20143 Use GET_MODE_CLASS for scalar floating point.
20145 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
20149 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
20150 Fix call of call_cgraph_insertion_hooks.
20152 2018-01-18 Martin Sebor <msebor@redhat.com>
20154 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
20156 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
20159 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
20162 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
20165 * common.opt: (-ffile-prefix-map): New option.
20166 * opts.c (common_handle_option): Defer it.
20167 * opts-global.c (handle_common_deferred_options): Handle it.
20168 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
20169 * file-prefix-map.h: New file.
20170 (remap_debug_filename, add_debug_prefix_map): ...here.
20171 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
20172 * final.c (debug_prefix_map, add_debug_prefix_map
20173 remap_debug_filename): Move to...
20174 * file-prefix-map.c: New file.
20175 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
20176 generalize, get rid of alloca(), use strrchr() instead of strchr().
20177 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
20178 Implement in terms of add_prefix_map().
20179 (remap_macro_filename, remap_debug_filename): Implement in term of
20181 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
20182 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
20183 * dbxout.c: Include file-prefix-map.h.
20184 * varasm.c: Likewise.
20185 * vmsdbgout.c: Likewise.
20186 * xcoffout.c: Likewise.
20187 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
20188 * doc/cppopts.texi (-fmacro-prefix-map): Document.
20189 * doc/invoke.texi (-ffile-prefix-map): Document.
20190 (-fdebug-prefix-map): Update description.
20192 2018-01-18 Martin Liska <mliska@suse.cz>
20194 * config/i386/i386.c (indirect_thunk_name): Document that also
20196 (output_indirect_thunk): Document why both instructions
20197 (pause and lfence) are generated.
20199 2018-01-18 Richard Biener <rguenther@suse.de>
20201 PR tree-optimization/83887
20202 * graphite-scop-detection.c
20203 (scop_detection::get_nearest_dom_with_single_entry): Remove.
20204 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
20205 (scop_detection::merge_sese): Re-implement with a flood-fill
20206 algorithm that properly finds a SESE region if it exists.
20208 2018-01-18 Jakub Jelinek <jakub@redhat.com>
20211 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
20212 pointer_diff optimizations use view_convert instead of convert.
20214 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
20216 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
20217 Generate different code for -mno-speculate-indirect-jumps.
20218 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
20219 (*call_indirect_aix<mode>): Disable for
20220 -mno-speculate-indirect-jumps.
20221 (*call_indirect_aix<mode>_nospec): New define_insn.
20222 (*call_value_indirect_aix<mode>): Disable for
20223 -mno-speculate-indirect-jumps.
20224 (*call_value_indirect_aix<mode>_nospec): New define_insn.
20225 (*sibcall_nonlocal_sysv<mode>): Generate different code for
20226 -mno-speculate-indirect-jumps.
20227 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
20229 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
20231 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
20232 long double type, set the flags for noting the default long double
20233 type, even if we don't pass or return a long double type.
20235 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
20238 * ipa-inline.c (flatten_function): Do not overwrite final inlining
20241 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
20243 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
20244 support for merge[hl].
20245 (fold_mergehl_helper): New helper function.
20246 (tree-vector-builder.h): New #include for tree_vector_builder usage.
20247 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
20248 (altivec_vmrglw_direct): Add xxmrglw insn.
20250 2018-01-17 Andrew Waterman <andrew@sifive.com>
20252 * config/riscv/riscv.c (riscv_conditional_register_usage): If
20253 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
20255 2018-01-17 David Malcolm <dmalcolm@redhat.com>
20258 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
20259 call the lto_location_cache before reading the
20260 DECL_SOURCE_LOCATION of the types.
20262 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
20263 Richard Sandiford <richard.sandiford@linaro.org>
20265 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
20266 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
20267 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
20268 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
20269 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
20271 * config/aarch64/constraints.md (aarch64_movti_operand):
20273 * config/aarch64/predicates.md (Uti): Add new constraint.
20275 2018-01-17 Carl Love <cel@us.ibm.com>
20277 * config/rs6000/vsx.md (define_expand xl_len_r,
20278 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
20279 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
20281 (define_expand, define_insn): Move the shift left from the
20282 define_insn to the define_expand for lxvl and stxvl instructions.
20283 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
20284 and XL_LEN_R definitions to PURE.
20286 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
20288 * config/i386/i386.c (indirect_thunk_name): Declare regno
20289 as unsigned int. Compare regno with INVALID_REGNUM.
20290 (output_indirect_thunk): Ditto.
20291 (output_indirect_thunk_function): Ditto.
20292 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
20293 in the call to output_indirect_thunk_function.
20295 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
20297 PR middle-end/83884
20298 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
20299 rather than the size of inner_type to determine the stack slot size
20300 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
20302 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
20305 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
20308 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
20310 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
20311 endian Linux systems to optionally enable multilibs for selecting
20312 the long double type if the user configured an explicit type.
20313 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
20314 have no long double multilibs if not defined.
20315 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
20316 warn if the user used -mabi={ieee,ibm}longdouble and we built
20317 multilibs for long double.
20318 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
20319 appropriate multilib option.
20320 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
20322 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
20323 for building long double multilibs.
20324 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
20326 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
20328 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
20331 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
20333 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
20336 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
20339 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
20342 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
20344 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
20345 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
20347 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
20349 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
20350 different rtl trees depending on TARGET_64BIT.
20351 (rs6000_gen_lvx): Likewise.
20353 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
20355 * config/visium/visium.md (nop): Tweak comment.
20356 (hazard_nop): Likewise.
20358 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
20360 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
20361 -mspeculate-indirect-jumps.
20362 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
20363 for -mno-speculate-indirect-jumps.
20364 (*call_indirect_elfv2<mode>_nospec): New define_insn.
20365 (*call_value_indirect_elfv2<mode>): Disable for
20366 -mno-speculate-indirect-jumps.
20367 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
20368 (indirect_jump): Emit different RTL for
20369 -mno-speculate-indirect-jumps.
20370 (*indirect_jump<mode>): Disable for
20371 -mno-speculate-indirect-jumps.
20372 (*indirect_jump<mode>_nospec): New define_insn.
20373 (tablejump): Emit different RTL for
20374 -mno-speculate-indirect-jumps.
20375 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
20376 (tablejumpsi_nospec): New define_expand.
20377 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
20378 (tablejumpdi_nospec): New define_expand.
20379 (*tablejump<mode>_internal1): Disable for
20380 -mno-speculate-indirect-jumps.
20381 (*tablejump<mode>_internal1_nospec): New define_insn.
20382 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
20385 2018-01-16 Artyom Skrobov tyomitch@gmail.com
20387 * caller-save.c (insert_save): Drop unnecessary parameter. All
20390 2018-01-16 Jakub Jelinek <jakub@redhat.com>
20391 Richard Biener <rguenth@suse.de>
20394 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
20395 return early, inline manually is_gimple_sizepos. Make sure if we
20396 call gimplify_expr we don't end up with a gimple constant.
20397 * tree.c (variably_modified_type_p): Don't return true for
20398 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
20399 * gimplify.h (is_gimple_sizepos): Remove.
20401 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
20403 PR tree-optimization/83857
20404 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
20405 vectorizable_live_operation for pure SLP statements.
20406 (vectorizable_live_operation): Handle PHIs.
20408 2018-01-16 Richard Biener <rguenther@suse.de>
20410 PR tree-optimization/83867
20411 * tree-vect-stmts.c (vect_transform_stmt): Precompute
20412 nested_in_vect_loop_p since the scalar stmt may get invalidated.
20414 2018-01-16 Jakub Jelinek <jakub@redhat.com>
20417 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
20418 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
20419 If off is not INTEGER_CST, issue a may not be aligned warning
20420 rather than isn't aligned. Use isn%'t rather than isn't.
20421 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
20423 <case MULT_EXPR>: Improve the case when bottom and one of the
20424 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
20425 operand, in that case check if the other operand is multiple of
20426 bottom divided by the INTEGER_CST operand.
20428 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
20431 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
20432 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
20433 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
20434 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
20435 * config/pa/pa.c (pa_function_arg_advance): Likewise.
20436 (pa_function_arg, pa_arg_partial_bytes): Likewise.
20437 (pa_function_arg_size): New function.
20439 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
20441 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
20442 in a separate statement.
20444 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
20446 PR tree-optimization/83847
20447 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
20448 group gathers and scatters.
20450 2018-01-16 Jakub Jelinek <jakub@redhat.com>
20452 PR rtl-optimization/86620
20453 * params.def (max-sched-ready-insns): Bump minimum value to 1.
20455 PR rtl-optimization/83213
20456 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
20457 to last if both are JUMP_INSNs.
20459 PR tree-optimization/83843
20460 * gimple-ssa-store-merging.c
20461 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
20462 store_immediate_info for bswap/nop orig_stores.
20464 2018-01-15 Andrew Waterman <andrew@sifive.com>
20466 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
20468 <UDIV>: Increase cost if !TARGET_DIV.
20470 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
20472 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
20473 (define_attr "cr_logical_3op"): New.
20474 (cceq_ior_compare): Adjust.
20475 (cceq_ior_compare_complement): Adjust.
20476 (*cceq_rev_compare): Adjust.
20477 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
20478 (is_cracked_insn): Adjust.
20479 (insn_must_be_first_in_group): Adjust.
20480 * config/rs6000/40x.md: Adjust.
20481 * config/rs6000/440.md: Adjust.
20482 * config/rs6000/476.md: Adjust.
20483 * config/rs6000/601.md: Adjust.
20484 * config/rs6000/603.md: Adjust.
20485 * config/rs6000/6xx.md: Adjust.
20486 * config/rs6000/7450.md: Adjust.
20487 * config/rs6000/7xx.md: Adjust.
20488 * config/rs6000/8540.md: Adjust.
20489 * config/rs6000/cell.md: Adjust.
20490 * config/rs6000/e300c2c3.md: Adjust.
20491 * config/rs6000/e500mc.md: Adjust.
20492 * config/rs6000/e500mc64.md: Adjust.
20493 * config/rs6000/e5500.md: Adjust.
20494 * config/rs6000/e6500.md: Adjust.
20495 * config/rs6000/mpc.md: Adjust.
20496 * config/rs6000/power4.md: Adjust.
20497 * config/rs6000/power5.md: Adjust.
20498 * config/rs6000/power6.md: Adjust.
20499 * config/rs6000/power7.md: Adjust.
20500 * config/rs6000/power8.md: Adjust.
20501 * config/rs6000/power9.md: Adjust.
20502 * config/rs6000/rs64.md: Adjust.
20503 * config/rs6000/titan.md: Adjust.
20505 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
20507 * config/i386/predicates.md (indirect_branch_operand): Rewrite
20508 ix86_indirect_branch_register logic.
20510 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
20512 * config/i386/constraints.md (Bs): Update
20513 ix86_indirect_branch_register check. Don't check
20514 ix86_indirect_branch_register with GOT_memory_operand.
20516 * config/i386/predicates.md (GOT_memory_operand): Don't check
20517 ix86_indirect_branch_register here.
20518 (GOT32_symbol_operand): Likewise.
20520 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
20522 * config/i386/predicates.md (constant_call_address_operand):
20523 Rewrite ix86_indirect_branch_register logic.
20524 (sibcall_insn_operand): Likewise.
20526 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
20528 * config/i386/constraints.md (Bs): Replace
20529 ix86_indirect_branch_thunk_register with
20530 ix86_indirect_branch_register.
20532 * config/i386/i386.md (indirect_jump): Likewise.
20533 (tablejump): Likewise.
20534 (*sibcall_memory): Likewise.
20535 (*sibcall_value_memory): Likewise.
20536 Peepholes of indirect call and jump via memory: Likewise.
20537 * config/i386/i386.opt: Likewise.
20538 * config/i386/predicates.md (indirect_branch_operand): Likewise.
20539 (GOT_memory_operand): Likewise.
20540 (call_insn_operand): Likewise.
20541 (sibcall_insn_operand): Likewise.
20542 (GOT32_symbol_operand): Likewise.
20544 2018-01-15 Jakub Jelinek <jakub@redhat.com>
20546 PR middle-end/83837
20547 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
20548 type rather than type addr's type points to.
20549 (expand_omp_atomic_mutex): Likewise.
20550 (expand_omp_atomic): Likewise.
20552 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
20555 * config/i386/i386.c (output_indirect_thunk_function): Use
20556 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
20557 for __x86_return_thunk.
20559 2018-01-15 Richard Biener <rguenther@suse.de>
20561 PR middle-end/83850
20562 * expmed.c (extract_bit_field_1): Fix typo.
20564 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20567 * config/arm/iterators.md (VF): New mode iterator.
20568 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
20569 Remove integer-related logic from pattern.
20570 (neon_vabd<mode>_3): Likewise.
20572 2018-01-15 Jakub Jelinek <jakub@redhat.com>
20574 PR middle-end/82694
20575 * common.opt (fstrict-overflow): No longer an alias.
20576 (fwrapv-pointer): New option.
20577 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
20578 also for pointer types based on flag_wrapv_pointer.
20579 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
20580 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
20581 opts->x_flag_wrapv got set.
20582 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
20583 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
20584 POINTER_TYPE_OVERFLOW_UNDEFINED.
20585 * match.pd: Likewise in address comparison pattern.
20586 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
20588 2018-01-15 Richard Biener <rguenther@suse.de>
20591 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
20592 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
20593 Reset type names to their identifier if their TYPE_DECL doesn't
20594 have linkage (and thus is used for ODR and devirt).
20595 (save_debug_info_for_decl): Remove.
20596 (save_debug_info_for_type): Likewise.
20597 (add_tree_to_fld_list): Adjust.
20598 * tree-pretty-print.c (dump_generic_node): Make dumping of
20599 type names more robust.
20601 2018-01-15 Richard Biener <rguenther@suse.de>
20603 * BASE-VER: Bump to 8.0.1.
20605 2018-01-14 Martin Sebor <msebor@redhat.com>
20608 * builtins.c (check_access): Avoid warning when the no-warning bit
20611 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
20613 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
20614 * ira-color (allocno_hard_regs_compare): Likewise.
20616 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
20619 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
20620 Use .pushsection/.popsection.
20622 2018-01-14 Martin Sebor <msebor@redhat.com>
20625 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
20627 2018-01-14 Jakub Jelinek <jakub@redhat.com>
20629 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
20630 entry from extra_headers.
20631 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
20632 extra_headers, make the list bitwise identical to the i?86-*-* one.
20634 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
20636 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
20637 -mcmodel=large with -mindirect-branch=thunk,
20638 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
20639 -mfunction-return=thunk-extern.
20640 * doc/invoke.texi: Document -mcmodel=large is incompatible with
20641 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
20642 -mfunction-return=thunk and -mfunction-return=thunk-extern.
20644 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
20646 * config/i386/i386.c (print_reg): Print the name of the full
20647 integer register without '%'.
20648 (ix86_print_operand): Handle 'V'.
20649 * doc/extend.texi: Document 'V' modifier.
20651 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
20653 * config/i386/constraints.md (Bs): Disallow memory operand for
20654 -mindirect-branch-register.
20656 * config/i386/predicates.md (indirect_branch_operand): Likewise.
20657 (GOT_memory_operand): Likewise.
20658 (call_insn_operand): Likewise.
20659 (sibcall_insn_operand): Likewise.
20660 (GOT32_symbol_operand): Likewise.
20661 * config/i386/i386.md (indirect_jump): Call convert_memory_address
20662 for -mindirect-branch-register.
20663 (tablejump): Likewise.
20664 (*sibcall_memory): Likewise.
20665 (*sibcall_value_memory): Likewise.
20666 Disallow peepholes of indirect call and jump via memory for
20667 -mindirect-branch-register.
20668 (*call_pop): Replace m with Bw.
20669 (*call_value_pop): Likewise.
20670 (*sibcall_pop_memory): Replace m with Bs.
20671 * config/i386/i386.opt (mindirect-branch-register): New option.
20672 * doc/invoke.texi: Document -mindirect-branch-register option.
20674 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
20676 * config/i386/i386-protos.h (ix86_output_function_return): New.
20677 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
20678 set function_return_type.
20679 (indirect_thunk_name): Add ret_p to indicate thunk for function
20681 (output_indirect_thunk_function): Pass false to
20682 indirect_thunk_name.
20683 (ix86_output_indirect_branch_via_reg): Likewise.
20684 (ix86_output_indirect_branch_via_push): Likewise.
20685 (output_indirect_thunk_function): Create alias for function
20686 return thunk if regno < 0.
20687 (ix86_output_function_return): New function.
20688 (ix86_handle_fndecl_attribute): Handle function_return.
20689 (ix86_attribute_table): Add function_return.
20690 * config/i386/i386.h (machine_function): Add
20691 function_return_type.
20692 * config/i386/i386.md (simple_return_internal): Use
20693 ix86_output_function_return.
20694 (simple_return_internal_long): Likewise.
20695 * config/i386/i386.opt (mfunction-return=): New option.
20696 (indirect_branch): Mention -mfunction-return=.
20697 * doc/extend.texi: Document function_return function attribute.
20698 * doc/invoke.texi: Document -mfunction-return= option.
20700 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
20702 * config/i386/i386-opts.h (indirect_branch): New.
20703 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
20704 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
20705 with local indirect jump when converting indirect call and jump.
20706 (ix86_set_indirect_branch_type): New.
20707 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
20708 (indirectlabelno): New.
20709 (indirect_thunk_needed): Likewise.
20710 (indirect_thunk_bnd_needed): Likewise.
20711 (indirect_thunks_used): Likewise.
20712 (indirect_thunks_bnd_used): Likewise.
20713 (INDIRECT_LABEL): Likewise.
20714 (indirect_thunk_name): Likewise.
20715 (output_indirect_thunk): Likewise.
20716 (output_indirect_thunk_function): Likewise.
20717 (ix86_output_indirect_branch_via_reg): Likewise.
20718 (ix86_output_indirect_branch_via_push): Likewise.
20719 (ix86_output_indirect_branch): Likewise.
20720 (ix86_output_indirect_jmp): Likewise.
20721 (ix86_code_end): Call output_indirect_thunk_function if needed.
20722 (ix86_output_call_insn): Call ix86_output_indirect_branch if
20724 (ix86_handle_fndecl_attribute): Handle indirect_branch.
20725 (ix86_attribute_table): Add indirect_branch.
20726 * config/i386/i386.h (machine_function): Add indirect_branch_type
20727 and has_local_indirect_jump.
20728 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
20730 (tablejump): Likewise.
20731 (*indirect_jump): Use ix86_output_indirect_jmp.
20732 (*tablejump_1): Likewise.
20733 (simple_return_indirect_internal): Likewise.
20734 * config/i386/i386.opt (mindirect-branch=): New option.
20735 (indirect_branch): New.
20738 (thunk-inline): Likewise.
20739 (thunk-extern): Likewise.
20740 * doc/extend.texi: Document indirect_branch function attribute.
20741 * doc/invoke.texi: Document -mindirect-branch= option.
20743 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
20746 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
20748 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
20750 * ipa-inline.c (want_inline_small_function_p): Return false if
20751 inlining has already failed with CIF_FINAL_ERROR.
20752 (update_caller_keys): Call want_inline_small_function_p before
20754 (update_callee_keys): Likewise.
20756 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
20758 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
20760 (rs6000_quadword_masked_address_p): Likewise.
20761 (quad_aligned_load_p): Likewise.
20762 (quad_aligned_store_p): Likewise.
20763 (const_load_sequence_p): Add comment to describe the outer-most loop.
20764 (mimic_memory_attributes_and_flags): New function.
20765 (rs6000_gen_stvx): Likewise.
20766 (replace_swapped_aligned_store): Likewise.
20767 (rs6000_gen_lvx): Likewise.
20768 (replace_swapped_aligned_load): Likewise.
20769 (replace_swapped_load_constant): Capitalize argument name in
20770 comment describing this function.
20771 (rs6000_analyze_swaps): Add a third pass to search for vector loads
20772 and stores that access quad-word aligned addresses and replace
20773 with stvx or lvx instructions when appropriate.
20774 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
20775 New function prototype.
20776 (rs6000_quadword_masked_address_p): Likewise.
20777 (rs6000_gen_lvx): Likewise.
20778 (rs6000_gen_stvx): Likewise.
20779 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
20780 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
20781 when memory address is aligned.
20782 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
20783 this split to select lvx instruction when memory address is aligned.
20784 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
20785 instruction when memory address is aligned.
20786 (*vsx_le_perm_load_v16qi): Likewise.
20787 (four unnamed splitters): Modify to select the stvx instruction
20788 when memory is aligned.
20790 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
20792 * predict.c (determine_unlikely_bbs): Handle correctly BBs
20793 which appears in the queue multiple times.
20795 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20796 Alan Hayward <alan.hayward@arm.com>
20797 David Sherwood <david.sherwood@arm.com>
20799 * tree-vectorizer.h (vec_lower_bound): New structure.
20800 (_loop_vec_info): Add check_nonzero and lower_bounds.
20801 (LOOP_VINFO_CHECK_NONZERO): New macro.
20802 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
20803 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
20804 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
20805 fields. Make seg_len the distance travelled, not including the
20807 (dr_direction_indicator): Declare.
20808 (dr_zero_step_indicator): Likewise.
20809 (dr_known_forward_stride_p): Likewise.
20810 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
20812 (runtime_alias_check_p): Allow runtime alias checks with
20814 (operator ==): Compare access_size and align.
20815 (prune_runtime_alias_test_list): Rework for new distinction between
20816 the access_size and seg_len.
20817 (create_intersect_range_checks_index): Likewise. Cope with polynomial
20819 (get_segment_min_max): New function.
20820 (create_intersect_range_checks): Use it.
20821 (dr_step_indicator): New function.
20822 (dr_direction_indicator): Likewise.
20823 (dr_zero_step_indicator): Likewise.
20824 (dr_known_forward_stride_p): Likewise.
20825 * tree-loop-distribution.c (data_ref_segment_size): Return
20826 DR_STEP * (niters - 1).
20827 (compute_alias_check_pairs): Update call to the dr_with_seg_len
20829 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
20830 (vect_preserves_scalar_order_p): New function, split out from...
20831 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
20832 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
20833 (vect_vfa_access_size): New function.
20834 (vect_vfa_align): Likewise.
20835 (vect_compile_time_alias): Take access_size_a and access_b arguments.
20836 (dump_lower_bound): New function.
20837 (vect_check_lower_bound): Likewise.
20838 (vect_small_gap_p): Likewise.
20839 (vectorizable_with_step_bound_p): Likewise.
20840 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
20841 depencies if the vectorization factor is 1. Convert the checks
20842 for nonzero steps into checks on the bounds of DR_STEP. Try using
20843 a bunds check for variable steps if the minimum required step is
20844 relatively small. Update calls to the dr_with_seg_len
20845 constructor and to vect_compile_time_alias.
20846 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
20848 (vect_loop_versioning): Call it.
20849 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
20851 (vect_estimate_min_profitable_iters): Account for any bounds checks.
20853 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20854 Alan Hayward <alan.hayward@arm.com>
20855 David Sherwood <david.sherwood@arm.com>
20857 * doc/sourcebuild.texi (vect_scatter_store): Document.
20858 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
20860 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
20862 * genopinit.c (main): Add supports_vec_scatter_store and
20863 supports_vec_scatter_store_cached to target_optabs.
20864 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
20865 IFN_MASK_SCATTER_STORE.
20866 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
20868 * internal-fn.h (internal_store_fn_p): Declare.
20869 (internal_fn_stored_value_index): Likewise.
20870 * internal-fn.c (scatter_store_direct): New macro.
20871 (expand_scatter_store_optab_fn): New function.
20872 (direct_scatter_store_optab_supported_p): New macro.
20873 (internal_store_fn_p): New function.
20874 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
20875 IFN_MASK_SCATTER_STORE.
20876 (internal_fn_mask_index): Likewise.
20877 (internal_fn_stored_value_index): New function.
20878 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
20879 for scatter stores.
20880 * optabs-query.h (supports_vec_scatter_store_p): Declare.
20881 * optabs-query.c (supports_vec_scatter_store_p): New function.
20882 * tree-vectorizer.h (vect_get_store_rhs): Declare.
20883 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
20884 true for scatter stores.
20885 (vect_gather_scatter_fn_p): Handle scatter stores too.
20886 (vect_check_gather_scatter): Consider using scatter stores if
20887 supports_vec_scatter_store_p.
20888 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
20889 scatter stores too.
20890 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
20891 internal_fn_stored_value_index.
20892 (check_load_store_masking): Handle scatter stores too.
20893 (vect_get_store_rhs): Make public.
20894 (vectorizable_call): Use internal_store_fn_p.
20895 (vectorizable_store): Handle scatter store internal functions.
20896 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
20897 when deciding whether the end of the group has been reached.
20898 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
20899 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
20900 (mask_scatter_store<mode>): New insns.
20902 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20903 Alan Hayward <alan.hayward@arm.com>
20904 David Sherwood <david.sherwood@arm.com>
20906 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
20907 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
20908 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
20910 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
20911 Use vect_truncate_gather_scatter_offset if we can't treat the
20912 operation as a normal gather load or scatter store.
20913 (get_group_load_store_type): Take the gather_scatter_info
20914 as argument. Try using a gather load or scatter store for
20915 single-element groups.
20916 (get_load_store_type): Update calls to get_group_load_store_type
20917 and vect_use_strided_gather_scatters_p.
20919 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20920 Alan Hayward <alan.hayward@arm.com>
20921 David Sherwood <david.sherwood@arm.com>
20923 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
20924 optional tree argument.
20925 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
20927 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
20928 but continue to use the current value as a fallback.
20929 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
20930 to compare the updates.
20931 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
20932 (get_load_store_type): Use it when handling a strided access.
20933 (vect_get_strided_load_store_ops): New function.
20934 (vect_get_data_ptr_increment): Likewise.
20935 (vectorizable_load): Handle strided gather loads. Always pass
20936 a step to vect_create_data_ref_ptr and bump_vector_ptr.
20938 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20939 Alan Hayward <alan.hayward@arm.com>
20940 David Sherwood <david.sherwood@arm.com>
20942 * doc/md.texi (gather_load@var{m}): Document.
20943 (mask_gather_load@var{m}): Likewise.
20944 * genopinit.c (main): Add supports_vec_gather_load and
20945 supports_vec_gather_load_cached to target_optabs.
20946 * optabs-tree.c (init_tree_optimization_optabs): Use
20947 ggc_cleared_alloc to allocate target_optabs.
20948 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
20949 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
20951 * internal-fn.h (internal_load_fn_p): Declare.
20952 (internal_gather_scatter_fn_p): Likewise.
20953 (internal_fn_mask_index): Likewise.
20954 (internal_gather_scatter_fn_supported_p): Likewise.
20955 * internal-fn.c (gather_load_direct): New macro.
20956 (expand_gather_load_optab_fn): New function.
20957 (direct_gather_load_optab_supported_p): New macro.
20958 (direct_internal_fn_optab): New function.
20959 (internal_load_fn_p): Likewise.
20960 (internal_gather_scatter_fn_p): Likewise.
20961 (internal_fn_mask_index): Likewise.
20962 (internal_gather_scatter_fn_supported_p): Likewise.
20963 * optabs-query.c (supports_at_least_one_mode_p): New function.
20964 (supports_vec_gather_load_p): Likewise.
20965 * optabs-query.h (supports_vec_gather_load_p): Declare.
20966 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
20967 and memory_type field.
20968 (NUM_PATTERNS): Bump to 15.
20969 * tree-vect-data-refs.c: Include internal-fn.h.
20970 (vect_gather_scatter_fn_p): New function.
20971 (vect_describe_gather_scatter_call): Likewise.
20972 (vect_check_gather_scatter): Try using internal functions for
20973 gather loads. Recognize existing calls to a gather load function.
20974 (vect_analyze_data_refs): Consider using gather loads if
20975 supports_vec_gather_load_p.
20976 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
20977 (vect_get_gather_scatter_offset_type): Likewise.
20978 (vect_convert_mask_for_vectype): Likewise.
20979 (vect_add_conversion_to_patterm): Likewise.
20980 (vect_try_gather_scatter_pattern): Likewise.
20981 (vect_recog_gather_scatter_pattern): New pattern recognizer.
20982 (vect_vect_recog_func_ptrs): Add it.
20983 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
20984 internal_fn_mask_index and internal_gather_scatter_fn_p.
20985 (check_load_store_masking): Take the gather_scatter_info as an
20986 argument and handle gather loads.
20987 (vect_get_gather_scatter_ops): New function.
20988 (vectorizable_call): Check internal_load_fn_p.
20989 (vectorizable_load): Likewise. Handle gather load internal
20991 (vectorizable_store): Update call to check_load_store_masking.
20992 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
20993 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
20994 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
20995 (aarch64_gather_scale_operand_d): New predicates.
20996 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
20997 (mask_gather_load<mode>): New insns.
20999 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21000 Alan Hayward <alan.hayward@arm.com>
21001 David Sherwood <david.sherwood@arm.com>
21003 * optabs.def (fold_left_plus_optab): New optab.
21004 * doc/md.texi (fold_left_plus_@var{m}): Document.
21005 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
21006 * internal-fn.c (fold_left_direct): Define.
21007 (expand_fold_left_optab_fn): Likewise.
21008 (direct_fold_left_optab_supported_p): Likewise.
21009 * fold-const-call.c (fold_const_fold_left): New function.
21010 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
21011 * tree-parloops.c (valid_reduction_p): New function.
21012 (gather_scalar_reductions): Use it.
21013 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
21014 (vect_finish_replace_stmt): Declare.
21015 * tree-vect-loop.c (fold_left_reduction_fn): New function.
21016 (needs_fold_left_reduction_p): New function, split out from...
21017 (vect_is_simple_reduction): ...here. Accept reductions that
21018 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
21019 (vect_force_simple_reduction): Also store the reduction type in
21020 the assignment's STMT_VINFO_REDUC_TYPE.
21021 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
21022 (merge_with_identity): New function.
21023 (vect_expand_fold_left): Likewise.
21024 (vectorize_fold_left_reduction): Likewise.
21025 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
21026 scalar phi in place for it. Check for target support and reject
21027 cases that would reassociate the operation. Defer the transform
21028 phase to vectorize_fold_left_reduction.
21029 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
21030 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
21031 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
21033 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21035 * tree-if-conv.c (predicate_mem_writes): Remove redundant
21036 call to ifc_temp_var.
21038 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21039 Alan Hayward <alan.hayward@arm.com>
21040 David Sherwood <david.sherwood@arm.com>
21042 * target.def (legitimize_address_displacement): Take the original
21043 offset as a poly_int.
21044 * targhooks.h (default_legitimize_address_displacement): Update
21046 * targhooks.c (default_legitimize_address_displacement): Likewise.
21047 * doc/tm.texi: Regenerate.
21048 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
21049 as an argument, moving assert of ad->disp == ad->disp_term to...
21050 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
21051 Try calling targetm.legitimize_address_displacement before expanding
21052 the address rather than afterwards, and adjust for the new interface.
21053 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
21054 Match the new hook interface. Handle SVE addresses.
21055 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
21056 new hook interface.
21058 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21060 * Makefile.in (OBJS): Add early-remat.o.
21061 * target.def (select_early_remat_modes): New hook.
21062 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
21063 * doc/tm.texi: Regenerate.
21064 * targhooks.h (default_select_early_remat_modes): Declare.
21065 * targhooks.c (default_select_early_remat_modes): New function.
21066 * timevar.def (TV_EARLY_REMAT): New timevar.
21067 * passes.def (pass_early_remat): New pass.
21068 * tree-pass.h (make_pass_early_remat): Declare.
21069 * early-remat.c: New file.
21070 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
21072 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
21074 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21075 Alan Hayward <alan.hayward@arm.com>
21076 David Sherwood <david.sherwood@arm.com>
21078 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
21079 vfm1 with a bound_epilog parameter.
21080 (vect_do_peeling): Update calls accordingly, and move the prologue
21081 call earlier in the function. Treat the base bound_epilog as 0 for
21082 fully-masked loops and retain vf - 1 for other loops. Add 1 to
21083 this base when peeling for gaps.
21084 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
21085 with fully-masked loops.
21086 (vect_estimate_min_profitable_iters): Handle the single peeled
21087 iteration in that case.
21089 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21090 Alan Hayward <alan.hayward@arm.com>
21091 David Sherwood <david.sherwood@arm.com>
21093 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
21094 single-element interleaving even if the size is not a power of 2.
21095 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
21096 accesses for single-element interleaving if the group size is
21099 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21100 Alan Hayward <alan.hayward@arm.com>
21101 David Sherwood <david.sherwood@arm.com>
21103 * doc/md.texi (fold_extract_last_@var{m}): Document.
21104 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
21105 * optabs.def (fold_extract_last_optab): New optab.
21106 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
21107 * internal-fn.c (fold_extract_direct): New macro.
21108 (expand_fold_extract_optab_fn): Likewise.
21109 (direct_fold_extract_optab_supported_p): Likewise.
21110 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
21111 * tree-vect-loop.c (vect_model_reduction_cost): Handle
21112 EXTRACT_LAST_REDUCTION.
21113 (get_initial_def_for_reduction): Do not create an initial vector
21114 for EXTRACT_LAST_REDUCTION reductions.
21115 (vectorizable_reduction): Leave the scalar phi in place for
21116 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
21117 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
21118 epilogue code for EXTRACT_LAST_REDUCTION and defer the
21119 transform phase to vectorizable_condition.
21120 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
21122 (vect_finish_stmt_generation): ...here.
21123 (vect_finish_replace_stmt): New function.
21124 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
21125 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
21127 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
21129 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21130 Alan Hayward <alan.hayward@arm.com>
21131 David Sherwood <david.sherwood@arm.com>
21133 * doc/md.texi (extract_last_@var{m}): Document.
21134 * optabs.def (extract_last_optab): New optab.
21135 * internal-fn.def (EXTRACT_LAST): New internal function.
21136 * internal-fn.c (cond_unary_direct): New macro.
21137 (expand_cond_unary_optab_fn): Likewise.
21138 (direct_cond_unary_optab_supported_p): Likewise.
21139 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
21140 loops using EXTRACT_LAST.
21141 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
21142 (extract_last_<mode>): ...this optab.
21143 (vec_extract<mode><Vel>): Update accordingly.
21145 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21146 Alan Hayward <alan.hayward@arm.com>
21147 David Sherwood <david.sherwood@arm.com>
21149 * target.def (empty_mask_is_expensive): New hook.
21150 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
21151 * doc/tm.texi: Regenerate.
21152 * targhooks.h (default_empty_mask_is_expensive): Declare.
21153 * targhooks.c (default_empty_mask_is_expensive): New function.
21154 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
21155 if the target says that empty masks are expensive.
21156 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
21158 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
21160 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21161 Alan Hayward <alan.hayward@arm.com>
21162 David Sherwood <david.sherwood@arm.com>
21164 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
21165 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
21166 (vect_use_loop_mask_for_alignment_p): New function.
21167 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
21168 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
21169 niters_skip argument. Make sure that the first niters_skip elements
21170 of the first iteration are inactive.
21171 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
21172 Update call to vect_set_loop_masks_directly.
21173 (get_misalign_in_elems): New function, split out from...
21174 (vect_gen_prolog_loop_niters): ...here.
21175 (vect_update_init_of_dr): Take a code argument that specifies whether
21176 the adjustment should be added or subtracted.
21177 (vect_update_init_of_drs): Likewise.
21178 (vect_prepare_for_masked_peels): New function.
21179 (vect_do_peeling): Skip prologue peeling if we're using a mask
21180 instead. Update call to vect_update_inits_of_drs.
21181 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
21183 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
21184 alignment. Do not include the number of peeled iterations in
21185 the minimum threshold in that case.
21186 (vectorizable_induction): Adjust the start value down by
21187 LOOP_VINFO_MASK_SKIP_NITERS iterations.
21188 (vect_transform_loop): Call vect_prepare_for_masked_peels.
21189 Take the number of skipped iterations into account when calculating
21191 * tree-vect-stmts.c (vect_gen_while_not): New function.
21193 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21194 Alan Hayward <alan.hayward@arm.com>
21195 David Sherwood <david.sherwood@arm.com>
21197 * doc/sourcebuild.texi (vect_fully_masked): Document.
21198 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
21199 default value to 0.
21200 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
21202 (vect_analyze_loop_2): ...here. Don't check the vectorization
21203 factor against the number of loop iterations if the loop is
21206 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21207 Alan Hayward <alan.hayward@arm.com>
21208 David Sherwood <david.sherwood@arm.com>
21210 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
21211 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
21212 (dump_groups): Update accordingly.
21213 (iv_use::mem_type): New member variable.
21214 (address_p): New function.
21215 (record_use): Add a mem_type argument and initialize the new
21217 (record_group_use): Add a mem_type argument. Use address_p.
21218 Remove obsolete null checks of base_object. Update call to record_use.
21219 (find_interesting_uses_op): Update call to record_group_use.
21220 (find_interesting_uses_cond): Likewise.
21221 (find_interesting_uses_address): Likewise.
21222 (get_mem_type_for_internal_fn): New function.
21223 (find_address_like_use): Likewise.
21224 (find_interesting_uses_stmt): Try find_address_like_use before
21225 calling find_interesting_uses_op.
21226 (addr_offset_valid_p): Use the iv mem_type field as the type
21227 of the addressed memory.
21228 (add_autoinc_candidates): Likewise.
21229 (get_address_cost): Likewise.
21230 (split_small_address_groups_p): Use address_p.
21231 (split_address_groups): Likewise.
21232 (add_iv_candidate_for_use): Likewise.
21233 (autoinc_possible_for_pair): Likewise.
21234 (rewrite_groups): Likewise.
21235 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
21236 (determine_group_iv_cost): Update after split of USE_ADDRESS.
21237 (get_alias_ptr_type_for_ptr_address): New function.
21238 (rewrite_use_address): Rewrite address uses in calls that were
21239 identified by find_address_like_use.
21241 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21242 Alan Hayward <alan.hayward@arm.com>
21243 David Sherwood <david.sherwood@arm.com>
21245 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
21247 * gimple-expr.h (is_gimple_addressable: Likewise.
21248 * gimple-expr.c (is_gimple_address): Likewise.
21249 * internal-fn.c (expand_call_mem_ref): New function.
21250 (expand_mask_load_optab_fn): Use it.
21251 (expand_mask_store_optab_fn): Likewise.
21253 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21254 Alan Hayward <alan.hayward@arm.com>
21255 David Sherwood <david.sherwood@arm.com>
21257 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
21258 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
21259 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
21260 (cond_umax@var{mode}): Document.
21261 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
21262 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
21263 (cond_umin_optab, cond_umax_optab): New optabs.
21264 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
21265 (COND_IOR, COND_XOR): New internal functions.
21266 * internal-fn.h (get_conditional_internal_fn): Declare.
21267 * internal-fn.c (cond_binary_direct): New macro.
21268 (expand_cond_binary_optab_fn): Likewise.
21269 (direct_cond_binary_optab_supported_p): Likewise.
21270 (get_conditional_internal_fn): New function.
21271 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
21272 Cope with reduction statements that are vectorized as calls rather
21274 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
21275 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
21276 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
21277 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
21278 (UNSPEC_COND_EOR): New unspecs.
21279 (optab): Add mappings for them.
21280 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
21281 (sve_int_op, sve_fp_op): New int attributes.
21283 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21284 Alan Hayward <alan.hayward@arm.com>
21285 David Sherwood <david.sherwood@arm.com>
21287 * optabs.def (while_ult_optab): New optab.
21288 * doc/md.texi (while_ult@var{m}@var{n}): Document.
21289 * internal-fn.def (WHILE_ULT): New internal function.
21290 * internal-fn.h (direct_internal_fn_supported_p): New override
21291 that takes two types as argument.
21292 * internal-fn.c (while_direct): New macro.
21293 (expand_while_optab_fn): New function.
21294 (convert_optab_supported_p): Likewise.
21295 (direct_while_optab_supported_p): New macro.
21296 * wide-int.h (wi::udiv_ceil): New function.
21297 * tree-vectorizer.h (rgroup_masks): New structure.
21298 (vec_loop_masks): New typedef.
21299 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
21300 and fully_masked_p.
21301 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
21302 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
21303 (vect_max_vf): New function.
21304 (slpeel_make_loop_iterate_ntimes): Delete.
21305 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
21306 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
21307 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
21308 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
21309 internal-fn.h, stor-layout.h and optabs-query.h.
21310 (vect_set_loop_mask): New function.
21311 (add_preheader_seq): Likewise.
21312 (add_header_seq): Likewise.
21313 (interleave_supported_p): Likewise.
21314 (vect_maybe_permute_loop_masks): Likewise.
21315 (vect_set_loop_masks_directly): Likewise.
21316 (vect_set_loop_condition_masked): Likewise.
21317 (vect_set_loop_condition_unmasked): New function, split out from
21318 slpeel_make_loop_iterate_ntimes.
21319 (slpeel_make_loop_iterate_ntimes): Rename to..
21320 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
21321 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
21322 (vect_do_peeling): Update call accordingly.
21323 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
21325 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
21326 mask_compare_type, can_fully_mask_p and fully_masked_p.
21327 (release_vec_loop_masks): New function.
21328 (_loop_vec_info): Use it to free the loop masks.
21329 (can_produce_all_loop_masks_p): New function.
21330 (vect_get_max_nscalars_per_iter): Likewise.
21331 (vect_verify_full_masking): Likewise.
21332 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
21333 retries, and free the mask rgroups before retrying. Check loop-wide
21334 reasons for disallowing fully-masked loops. Make the final decision
21335 about whether use a fully-masked loop or not.
21336 (vect_estimate_min_profitable_iters): Do not assume that peeling
21337 for the number of iterations will be needed for fully-masked loops.
21338 (vectorizable_reduction): Disable fully-masked loops.
21339 (vectorizable_live_operation): Likewise.
21340 (vect_halve_mask_nunits): New function.
21341 (vect_double_mask_nunits): Likewise.
21342 (vect_record_loop_mask): Likewise.
21343 (vect_get_loop_mask): Likewise.
21344 (vect_transform_loop): Handle the case in which the final loop
21345 iteration might handle a partial vector. Call vect_set_loop_condition
21346 instead of slpeel_make_loop_iterate_ntimes.
21347 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
21348 (check_load_store_masking): New function.
21349 (prepare_load_store_mask): Likewise.
21350 (vectorizable_store): Handle fully-masked loops.
21351 (vectorizable_load): Likewise.
21352 (supportable_widening_operation): Use vect_halve_mask_nunits for
21354 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
21355 (vect_gen_while): New function.
21356 * config/aarch64/aarch64.md (umax<mode>3): New expander.
21357 (aarch64_uqdec<mode>): New insn.
21359 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21360 Alan Hayward <alan.hayward@arm.com>
21361 David Sherwood <david.sherwood@arm.com>
21363 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
21364 (reduc_xor_scal_optab): New optabs.
21365 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
21366 (reduc_xor_scal_@var{m}): Document.
21367 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
21368 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
21369 internal functions.
21370 * fold-const-call.c (fold_const_call): Handle them.
21371 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
21372 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
21373 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
21374 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
21375 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
21376 (UNSPEC_XORV): New unspecs.
21377 (optab): Add entries for them.
21378 (BITWISEV): New int iterator.
21379 (bit_reduc_op): New int attributes.
21381 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21382 Alan Hayward <alan.hayward@arm.com>
21383 David Sherwood <david.sherwood@arm.com>
21385 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
21386 * internal-fn.def (VEC_SHL_INSERT): New internal function.
21387 * optabs.def (vec_shl_insert_optab): New optab.
21388 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
21389 (duplicate_and_interleave): Likewise.
21390 * tree-vect-loop.c: Include internal-fn.h.
21391 (neutral_op_for_slp_reduction): New function, split out from
21392 get_initial_defs_for_reduction.
21393 (get_initial_def_for_reduction): Handle option 2 for variable-length
21394 vectors by loading the neutral value into a vector and then shifting
21395 the initial value into element 0.
21396 (get_initial_defs_for_reduction): Replace the code argument with
21397 the neutral value calculated by neutral_op_for_slp_reduction.
21398 Use gimple_build_vector for constant-length vectors.
21399 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
21400 but the first group_size elements have a neutral value.
21401 Use duplicate_and_interleave otherwise.
21402 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
21403 Update call to get_initial_defs_for_reduction. Handle SLP
21404 reductions for variable-length vectors by creating one vector
21405 result for each scalar result, with the elements associated
21406 with other scalar results stubbed out with the neutral value.
21407 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
21408 Require IFN_VEC_SHL_INSERT for double reductions on
21409 variable-length vectors, or SLP reductions that have
21410 a neutral value. Require can_duplicate_and_interleave_p
21411 support for variable-length unchained SLP reductions if there
21412 is no neutral value, such as for MIN/MAX reductions. Also require
21413 the number of vector elements to be a multiple of the number of
21414 SLP statements when doing variable-length unchained SLP reductions.
21415 Update call to vect_create_epilog_for_reduction.
21416 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
21417 and remove initial values.
21418 (duplicate_and_interleave): Make public.
21419 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
21420 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
21422 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21423 Alan Hayward <alan.hayward@arm.com>
21424 David Sherwood <david.sherwood@arm.com>
21426 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
21427 (can_duplicate_and_interleave_p): New function.
21428 (vect_get_and_check_slp_defs): Take the vector of statements
21429 rather than just the current one. Remove excess parentheses.
21430 Restriction rejectinon of vect_constant_def and vect_external_def
21431 for variable-length vectors to boolean types, or types for which
21432 can_duplicate_and_interleave_p is false.
21433 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
21434 (duplicate_and_interleave): New function.
21435 (vect_get_constant_vectors): Use gimple_build_vector for
21436 constant-length vectors and suitable variable-length constant
21437 vectors. Use duplicate_and_interleave for other variable-length
21438 vectors. Don't defer the update when inserting new statements.
21440 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21441 Alan Hayward <alan.hayward@arm.com>
21442 David Sherwood <david.sherwood@arm.com>
21444 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
21445 min_profitable_iters doesn't go negative.
21447 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21448 Alan Hayward <alan.hayward@arm.com>
21449 David Sherwood <david.sherwood@arm.com>
21451 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
21452 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
21453 * optabs.def (vec_mask_load_lanes_optab): New optab.
21454 (vec_mask_store_lanes_optab): Likewise.
21455 * internal-fn.def (MASK_LOAD_LANES): New internal function.
21456 (MASK_STORE_LANES): Likewise.
21457 * internal-fn.c (mask_load_lanes_direct): New macro.
21458 (mask_store_lanes_direct): Likewise.
21459 (expand_mask_load_optab_fn): Handle masked operations.
21460 (expand_mask_load_lanes_optab_fn): New macro.
21461 (expand_mask_store_optab_fn): Handle masked operations.
21462 (expand_mask_store_lanes_optab_fn): New macro.
21463 (direct_mask_load_lanes_optab_supported_p): Likewise.
21464 (direct_mask_store_lanes_optab_supported_p): Likewise.
21465 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
21467 (vect_load_lanes_supported): Likewise.
21468 * tree-vect-data-refs.c (strip_conversion): New function.
21469 (can_group_stmts_p): Likewise.
21470 (vect_analyze_data_ref_accesses): Use it instead of checking
21471 for a pair of assignments.
21472 (vect_store_lanes_supported): Take a masked_p parameter.
21473 (vect_load_lanes_supported): Likewise.
21474 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
21475 vect_store_lanes_supported and vect_load_lanes_supported.
21476 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
21477 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
21478 parameter. Don't allow gaps for masked accesses.
21479 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
21480 and vect_load_lanes_supported.
21481 (get_load_store_type): Take a masked_p parameter and update
21482 call to get_group_load_store_type.
21483 (vectorizable_store): Update call to get_load_store_type.
21484 Handle IFN_MASK_STORE_LANES.
21485 (vectorizable_load): Update call to get_load_store_type.
21486 Handle IFN_MASK_LOAD_LANES.
21488 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21489 Alan Hayward <alan.hayward@arm.com>
21490 David Sherwood <david.sherwood@arm.com>
21492 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
21494 * config/aarch64/aarch64-protos.h
21495 (aarch64_sve_struct_memory_operand_p): Declare.
21496 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
21497 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
21498 (VPRED, vpred): Handle SVE structure modes.
21499 * config/aarch64/constraints.md (Utx): New constraint.
21500 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
21501 (aarch64_sve_struct_nonimmediate_operand): New predicates.
21502 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
21503 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
21504 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
21505 structure modes. Split into pieces after RA.
21506 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
21507 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
21509 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
21510 SVE structure modes.
21511 (aarch64_classify_address): Likewise.
21512 (sizetochar): Move earlier in file.
21513 (aarch64_print_operand): Handle SVE register lists.
21514 (aarch64_array_mode): New function.
21515 (aarch64_sve_struct_memory_operand_p): Likewise.
21516 (TARGET_ARRAY_MODE): Redefine.
21518 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21519 Alan Hayward <alan.hayward@arm.com>
21520 David Sherwood <david.sherwood@arm.com>
21522 * target.def (array_mode): New target hook.
21523 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
21524 * doc/tm.texi: Regenerate.
21525 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
21526 * hooks.c (hook_optmode_mode_uhwi_none): New function.
21527 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
21528 targetm.array_mode.
21529 * stor-layout.c (mode_for_array): Likewise. Support polynomial
21532 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21533 Alan Hayward <alan.hayward@arm.com>
21534 David Sherwood <david.sherwood@arm.com>
21536 * fold-const.c (fold_binary_loc): Check the argument types
21537 rather than the result type when testing for a vector operation.
21539 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21541 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
21542 * doc/tm.texi: Regenerate.
21544 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21545 Alan Hayward <alan.hayward@arm.com>
21546 David Sherwood <david.sherwood@arm.com>
21548 * doc/invoke.texi (-msve-vector-bits=): Document new option.
21549 (sve): Document new AArch64 extension.
21550 * doc/md.texi (w): Extend the description of the AArch64
21551 constraint to include SVE vectors.
21552 (Upl, Upa): Document new AArch64 predicate constraints.
21553 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
21555 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
21556 (msve-vector-bits=): New option.
21557 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
21558 SVE when these are disabled.
21559 (sve): New extension.
21560 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
21561 modes. Adjust their number of units based on aarch64_sve_vg.
21562 (MAX_BITSIZE_MODE_ANY_MODE): Define.
21563 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
21564 aarch64_addr_query_type.
21565 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
21566 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
21567 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
21568 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
21569 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
21570 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
21571 (aarch64_simd_imm_zero_p): Delete.
21572 (aarch64_check_zero_based_sve_index_immediate): Declare.
21573 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
21574 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
21575 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
21576 (aarch64_sve_float_mul_immediate_p): Likewise.
21577 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
21578 rather than an rtx.
21579 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
21580 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
21581 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
21582 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
21583 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
21584 (aarch64_regmode_natural_size): Likewise.
21585 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
21586 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
21588 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
21589 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
21590 for VG and the SVE predicate registers.
21591 (V_ALIASES): Add a "z"-prefixed alias.
21592 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
21593 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
21594 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
21595 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
21596 (REG_CLASS_NAMES): Add entries for them.
21597 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
21598 and the predicate registers.
21599 (aarch64_sve_vg): Declare.
21600 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
21601 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
21602 (REGMODE_NATURAL_SIZE): Define.
21603 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
21605 * config/aarch64/aarch64.c: Include cfgrtl.h.
21606 (simd_immediate_info): Add a constructor for series vectors,
21607 and an associated step field.
21608 (aarch64_sve_vg): New variable.
21609 (aarch64_dbx_register_number): Handle VG and the predicate registers.
21610 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
21611 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
21612 (VEC_ANY_DATA, VEC_STRUCT): New constants.
21613 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
21614 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
21615 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
21616 (aarch64_get_mask_mode): New functions.
21617 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
21618 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
21619 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
21620 predicate modes and predicate registers. Explicitly restrict
21621 GPRs to modes of 16 bytes or smaller. Only allow FP registers
21622 to store a vector mode if it is recognized by
21623 aarch64_classify_vector_mode.
21624 (aarch64_regmode_natural_size): New function.
21625 (aarch64_hard_regno_caller_save_mode): Return the original mode
21627 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
21628 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
21629 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
21630 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
21632 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
21633 does not overlap dest if the function is frame-related. Handle
21635 (aarch64_split_add_offset): New function.
21636 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
21637 them aarch64_add_offset.
21638 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
21639 and update call to aarch64_sub_sp.
21640 (aarch64_add_cfa_expression): New function.
21641 (aarch64_expand_prologue): Pass extra temporary registers to the
21642 functions above. Handle the case in which we need to emit new
21643 DW_CFA_expressions for registers that were originally saved
21644 relative to the stack pointer, but now have to be expressed
21645 relative to the frame pointer.
21646 (aarch64_output_mi_thunk): Pass extra temporary registers to the
21648 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
21649 IP0 and IP1 values for SVE frames.
21650 (aarch64_expand_vec_series): New function.
21651 (aarch64_expand_sve_widened_duplicate): Likewise.
21652 (aarch64_expand_sve_const_vector): Likewise.
21653 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
21654 Handle SVE constants. Use emit_move_insn to move a force_const_mem
21655 into the register, rather than emitting a SET directly.
21656 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
21657 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
21658 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
21659 (offset_9bit_signed_scaled_p): New functions.
21660 (aarch64_replicate_bitmask_imm): New function.
21661 (aarch64_bitmask_imm): Use it.
21662 (aarch64_cannot_force_const_mem): Reject expressions involving
21663 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
21664 (aarch64_classify_index): Handle SVE indices, by requiring
21665 a plain register index with a scale that matches the element size.
21666 (aarch64_classify_address): Handle SVE addresses. Assert that
21667 the mode of the address is VOIDmode or an integer mode.
21668 Update call to aarch64_classify_symbol.
21669 (aarch64_classify_symbolic_expression): Update call to
21670 aarch64_classify_symbol.
21671 (aarch64_const_vec_all_in_range_p): New function.
21672 (aarch64_print_vector_float_operand): Likewise.
21673 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
21674 "vN" for FP registers with SVE modes. Handle (const ...) vectors
21675 and the FP immediates 1.0 and 0.5.
21676 (aarch64_print_address_internal): Handle SVE addresses.
21677 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
21678 (aarch64_regno_regclass): Handle predicate registers.
21679 (aarch64_secondary_reload): Handle big-endian reloads of SVE
21681 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
21682 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
21683 (aarch64_convert_sve_vector_bits): New function.
21684 (aarch64_override_options): Use it to handle -msve-vector-bits=.
21685 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
21686 rather than an rtx.
21687 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
21688 Handle SVE vector and predicate modes. Accept VL-based constants
21689 that need only one temporary register, and VL offsets that require
21690 no temporary registers.
21691 (aarch64_conditional_register_usage): Mark the predicate registers
21692 as fixed if SVE isn't available.
21693 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
21694 Return true for SVE vector and predicate modes.
21695 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
21696 rather than an unsigned int. Handle SVE modes.
21697 (aarch64_preferred_simd_mode): Update call accordingly. Handle
21699 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
21701 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
21702 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
21703 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
21704 (aarch64_sve_float_mul_immediate_p): New functions.
21705 (aarch64_sve_valid_immediate): New function.
21706 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
21707 Explicitly reject structure modes. Check for INDEX constants.
21708 Handle PTRUE and PFALSE constants.
21709 (aarch64_check_zero_based_sve_index_immediate): New function.
21710 (aarch64_simd_imm_zero_p): Delete.
21711 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
21712 vector modes. Accept constants in the range of CNT[BHWD].
21713 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
21714 ask for an Advanced SIMD mode.
21715 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
21716 (aarch64_simd_vector_alignment): Handle SVE predicates.
21717 (aarch64_vectorize_preferred_vector_alignment): New function.
21718 (aarch64_simd_vector_alignment_reachable): Use it instead of
21720 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
21721 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
21723 (MAX_VECT_LEN): Delete.
21724 (expand_vec_perm_d): Add a vec_flags field.
21725 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
21726 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
21727 (aarch64_evpc_ext): Don't apply a big-endian lane correction
21729 (aarch64_evpc_rev): Rename to...
21730 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
21731 (aarch64_evpc_rev_global): New function.
21732 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
21733 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
21735 (aarch64_evpc_sve_tbl): New function.
21736 (aarch64_expand_vec_perm_const_1): Update after rename of
21737 aarch64_evpc_rev. Handle SVE permutes too, trying
21738 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
21739 than aarch64_evpc_tbl.
21740 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
21741 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
21742 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
21743 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
21744 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
21745 (aarch64_expand_sve_vcond): New functions.
21746 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
21747 of aarch64_vector_mode_p.
21748 (aarch64_dwarf_poly_indeterminate_value): New function.
21749 (aarch64_compute_pressure_classes): Likewise.
21750 (aarch64_can_change_mode_class): Likewise.
21751 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
21752 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
21753 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
21754 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
21755 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
21756 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
21757 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
21758 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
21760 (Dn, Dl, Dr): Accept const as well as const_vector.
21761 (Dz): Likewise. Compare against CONST0_RTX.
21762 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
21763 of "vector" where appropriate.
21764 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
21765 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
21766 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
21767 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
21768 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
21769 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
21770 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
21771 (v_int_equiv): Extend to SVE modes.
21772 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
21774 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
21775 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
21776 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
21777 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
21778 (SVE_COND_FP_CMP): New int iterators.
21779 (perm_hilo): Handle the new unpack unspecs.
21780 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
21782 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
21783 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
21784 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
21785 (aarch64_equality_operator, aarch64_constant_vector_operand)
21786 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
21787 (aarch64_sve_nonimmediate_operand): Likewise.
21788 (aarch64_sve_general_operand): Likewise.
21789 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
21790 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
21791 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
21792 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
21793 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
21794 (aarch64_sve_float_arith_immediate): Likewise.
21795 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
21796 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
21797 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
21798 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
21799 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
21800 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
21801 (aarch64_sve_float_arith_operand): Likewise.
21802 (aarch64_sve_float_arith_with_sub_operand): Likewise.
21803 (aarch64_sve_float_mul_operand): Likewise.
21804 (aarch64_sve_vec_perm_operand): Likewise.
21805 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
21806 (aarch64_mov_operand): Accept const_poly_int and const_vector.
21807 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
21808 as well as const_vector.
21809 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
21810 in file. Use CONST0_RTX and CONSTM1_RTX.
21811 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
21812 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
21813 Use aarch64_simd_imm_zero.
21814 * config/aarch64/aarch64-sve.md: New file.
21815 * config/aarch64/aarch64.md: Include it.
21816 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
21817 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
21818 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
21819 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
21820 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
21821 (sve): New attribute.
21822 (enabled): Disable instructions with the sve attribute unless
21824 (movqi, movhi): Pass CONST_POLY_INT operaneds through
21825 aarch64_expand_mov_immediate.
21826 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
21827 CNT[BHSD] immediates.
21828 (movti): Split CONST_POLY_INT moves into two halves.
21829 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
21830 Split additions that need a temporary here if the destination
21831 is the stack pointer.
21832 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
21833 (*add<mode>3_poly_1): New instruction.
21834 (set_clobber_cc): New expander.
21836 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21838 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
21839 parameter and use it instead of GET_MODE_SIZE (innermode). Use
21840 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
21841 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
21842 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
21843 Change innermode from fixed_mode_size to machine_mode.
21844 (simplify_subreg): Update call accordingly. Handle a constant-sized
21845 subreg of a variable-length CONST_VECTOR.
21847 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21848 Alan Hayward <alan.hayward@arm.com>
21849 David Sherwood <david.sherwood@arm.com>
21851 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
21852 (add_offset_to_base): New function, split out from...
21853 (create_mem_ref): ...here. When handling a scale other than 1,
21854 check first whether the address is valid without the offset.
21855 Add it into the base if so, leaving the index and scale as-is.
21857 2018-01-12 Jakub Jelinek <jakub@redhat.com>
21860 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
21861 fold_for_warn before checking if arg2 is INTEGER_CST.
21863 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
21865 * config/rs6000/predicates.md (load_multiple_operation): Delete.
21866 (store_multiple_operation): Delete.
21867 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
21868 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
21869 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
21870 guarded by TARGET_STRING.
21871 (rs6000_output_load_multiple): Delete.
21872 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
21873 OPTION_MASK_STRING / TARGET_STRING handling.
21874 (print_operand) <'N', 'O'>: Add comment that these are unused now.
21875 (const rs6000_opt_masks) <"string">: Change mask to 0.
21876 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
21877 (MASK_STRING): Delete.
21878 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
21880 (load_multiple): Delete.
21887 (store_multiple): Delete.
21894 (movmemsi_8reg): Delete.
21895 (corresponding unnamed define_insn): Delete.
21896 (movmemsi_6reg): Delete.
21897 (corresponding unnamed define_insn): Delete.
21898 (movmemsi_4reg): Delete.
21899 (corresponding unnamed define_insn): Delete.
21900 (movmemsi_2reg): Delete.
21901 (corresponding unnamed define_insn): Delete.
21902 (movmemsi_1reg): Delete.
21903 (corresponding unnamed define_insn): Delete.
21904 * config/rs6000/rs6000.opt (mno-string): New.
21905 (mstring): Replace by deprecation warning stub.
21906 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
21908 2018-01-12 Jakub Jelinek <jakub@redhat.com>
21910 * regrename.c (regrename_do_replace): If replacing the same
21911 reg multiple times, try to reuse last created gen_raw_REG.
21914 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
21915 main to workaround a bug in GDB.
21917 2018-01-12 Tom de Vries <tom@codesourcery.com>
21920 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
21922 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
21924 PR rtl-optimization/80481
21925 * ira-color.c (get_cap_member): New function.
21926 (allocnos_conflict_by_live_ranges_p): Use it.
21927 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
21928 (setup_slot_coalesced_allocno_live_ranges): Ditto.
21930 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
21933 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
21934 (*saddl_se_1): Ditto.
21935 (*ssubsi_1): Ditto.
21936 (*ssubl_se_1): Ditto.
21938 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
21940 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
21941 rather than wi::to_widest for DR_INITs.
21942 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
21943 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
21944 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
21946 (vect_analyze_group_access_1): Note that here.
21948 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
21950 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
21951 polynomial type sizes.
21953 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
21955 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
21956 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
21957 (gimple_add_tmp_var): Likewise.
21959 2018-01-12 Martin Liska <mliska@suse.cz>
21961 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
21962 (gimple_alloc_sizes): Likewise.
21963 (dump_gimple_statistics): Use PRIu64 in printf format.
21964 * gimple.h: Change uint64_t to int.
21966 2018-01-12 Martin Liska <mliska@suse.cz>
21968 * tree-core.h: Use uint64_t instead of int.
21969 * tree.c (tree_node_counts): Likewise.
21970 (tree_node_sizes): Likewise.
21971 (dump_tree_statistics): Use PRIu64 in printf format.
21973 2018-01-12 Martin Liska <mliska@suse.cz>
21975 * Makefile.in: As qsort_chk is implemented in vec.c, add
21976 vec.o to linkage of gencfn-macros.
21977 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
21978 passing the info to record_node_allocation_statistics.
21979 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
21981 * ggc-common.c (struct ggc_usage): Add operator== and use
21982 it in operator< and compare function.
21983 * mem-stats.h (struct mem_usage): Likewise.
21984 * vec.c (struct vec_usage): Remove operator< and compare
21985 function. Can be simply inherited.
21987 2018-01-12 Martin Jambor <mjambor@suse.cz>
21990 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
21991 * tree-ssa-math-opts.c: Include domwalk.h.
21992 (convert_mult_to_fma_1): New function.
21993 (fma_transformation_info): New type.
21994 (fma_deferring_state): Likewise.
21995 (cancel_fma_deferring): New function.
21996 (result_of_phi): Likewise.
21997 (last_fma_candidate_feeds_initial_phi): Likewise.
21998 (convert_mult_to_fma): Added deferring logic, split actual
21999 transformation to convert_mult_to_fma_1.
22000 (math_opts_dom_walker): New type.
22001 (math_opts_dom_walker::after_dom_children): New method, body moved
22002 here from pass_optimize_widening_mul::execute, added deferring logic
22004 (pass_optimize_widening_mul::execute): Moved most of code to
22005 math_opts_dom_walker::after_dom_children.
22006 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
22007 * config/i386/i386.c (ix86_option_override_internal): Added
22008 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
22010 2018-01-12 Richard Biener <rguenther@suse.de>
22013 * dwarf2out.c (gen_variable_die): Do not reset old_die for
22014 inline instance vars.
22016 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
22019 * config/rx/rx.c (rx_is_restricted_memory_address):
22020 Handle SUBREG case.
22022 2018-01-12 Richard Biener <rguenther@suse.de>
22024 PR tree-optimization/80846
22025 * target.def (split_reduction): New target hook.
22026 * targhooks.c (default_split_reduction): New function.
22027 * targhooks.h (default_split_reduction): Declare.
22028 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
22029 target requests first reduce vectors by combining low and high
22031 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
22032 (get_vectype_for_scalar_type_and_size): Export.
22033 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
22034 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
22035 * doc/tm.texi: Regenerate.
22036 * config/i386/i386.c (ix86_split_reduction): Implement
22037 TARGET_VECTORIZE_SPLIT_REDUCTION.
22039 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
22042 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
22043 in PIC mode except for TARGET_VXWORKS_RTP.
22044 * config/sparc/sparc.c: Include cfgrtl.h.
22045 (TARGET_INIT_PIC_REG): Define.
22046 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
22047 (sparc_pic_register_p): New predicate.
22048 (sparc_legitimate_address_p): Use it.
22049 (sparc_legitimize_pic_address): Likewise.
22050 (sparc_delegitimize_address): Likewise.
22051 (sparc_mode_dependent_address_p): Likewise.
22052 (gen_load_pcrel_sym): Remove 4th parameter.
22053 (load_got_register): Adjust call to above. Remove obsolete stuff.
22054 (sparc_expand_prologue): Do not call load_got_register here.
22055 (sparc_flat_expand_prologue): Likewise.
22056 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
22057 (sparc_use_pseudo_pic_reg): New function.
22058 (sparc_init_pic_reg): Likewise.
22059 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
22060 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
22062 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
22064 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
22065 Add item for branch_cost.
22067 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
22069 PR rtl-optimization/83565
22070 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
22071 not extend the result to a larger mode for rotate operations.
22072 (num_sign_bit_copies1): Likewise.
22074 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
22077 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
22079 Use values-Xc.o for -pedantic.
22080 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
22082 2018-01-12 Martin Liska <mliska@suse.cz>
22085 * ipa-devirt.c (final_warning_record::grow_type_warnings):
22087 (possible_polymorphic_call_targets): Use it.
22088 (ipa_devirt): Likewise.
22090 2018-01-12 Martin Liska <mliska@suse.cz>
22092 * profile-count.h (enum profile_quality): Use 0 as invalid
22093 enum value of profile_quality.
22095 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
22097 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
22098 -mext-string options.
22100 2018-01-12 Richard Biener <rguenther@suse.de>
22102 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
22103 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
22104 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
22106 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
22108 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
22110 * configure.ac (--with-long-double-format): Add support for the
22111 configuration option to change the default long double format on
22113 * config.gcc (powerpc*-linux*-*): Likewise.
22114 * configure: Regenerate.
22115 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
22116 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
22117 used without modification.
22119 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
22121 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
22122 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
22123 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
22124 MISC_BUILTIN_SPEC_BARRIER.
22125 (rs6000_init_builtins): Likewise.
22126 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
22128 (speculation_barrier): New define_insn.
22129 * doc/extend.texi: Document __builtin_speculation_barrier.
22131 2018-01-11 Jakub Jelinek <jakub@redhat.com>
22134 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
22135 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
22136 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
22138 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
22139 integral modes instead of "ss" and "sd".
22140 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
22141 vectors with 32-bit and 64-bit elements.
22142 (vecdupssescalarmodesuffix): New mode attribute.
22143 (vec_dup<mode>): Use it.
22145 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
22148 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
22149 frame if argument is passed on stack.
22151 2018-01-11 Jakub Jelinek <jakub@redhat.com>
22154 * ree.c (combine_reaching_defs): Optimize also
22155 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
22156 reg2=any_extend(exp); reg1=reg2;, formatting fix.
22158 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
22160 PR middle-end/83189
22161 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
22163 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
22165 PR middle-end/83718
22166 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
22167 after they are computed.
22169 2018-01-11 Bin Cheng <bin.cheng@arm.com>
22171 PR tree-optimization/83695
22172 * gimple-loop-linterchange.cc
22173 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
22174 reset cached scev information after interchange.
22175 (pass_linterchange::execute): Remove call to scev_reset_htab.
22177 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22179 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
22180 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
22181 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
22182 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
22183 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
22184 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
22185 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
22186 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
22187 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
22188 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
22189 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
22190 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
22191 (V_lane_reg): Likewise.
22192 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
22194 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
22195 (vfmal_lane_low<mode>_intrinsic,
22196 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
22197 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
22198 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
22199 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
22200 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
22201 vfmsl_lane_high<mode>_intrinsic): New define_insns.
22203 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22205 * config/arm/arm-cpus.in (fp16fml): New feature.
22206 (ALL_SIMD): Add fp16fml.
22207 (armv8.2-a): Add fp16fml as an option.
22208 (armv8.3-a): Likewise.
22209 (armv8.4-a): Add fp16fml as part of fp16.
22210 * config/arm/arm.h (TARGET_FP16FML): Define.
22211 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
22213 * config/arm/arm-modes.def (V2HF): Define.
22214 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
22215 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
22216 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
22217 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
22218 vfmsl_low, vfmsl_high): New set of builtins.
22219 * config/arm/iterators.md (PLUSMINUS): New code iterator.
22220 (vfml_op): New code attribute.
22221 (VFMLHALVES): New int iterator.
22222 (VFML, VFMLSEL): New mode attributes.
22223 (V_reg): Define mapping for V2HF.
22224 (V_hi, V_lo): New mode attributes.
22225 (VF_constraint): Likewise.
22226 (vfml_half, vfml_half_selector): New int attributes.
22227 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
22229 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
22230 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
22232 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
22233 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
22234 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
22235 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
22237 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
22238 Document new effective target and option set.
22240 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22242 * config/arm/arm-cpus.in (armv8_4): New feature.
22243 (ARMv8_4a): New fgroup.
22244 (armv8.4-a): New arch.
22245 * config/arm/arm-tables.opt: Regenerate.
22246 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
22247 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
22248 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
22249 Add matching rules for -march=armv8.4-a and extensions.
22250 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
22252 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
22255 * config/rx/rx.md (BW): New mode attribute.
22256 (sync_lock_test_and_setsi): Add mode suffix to insn output.
22258 2018-01-11 Richard Biener <rguenther@suse.de>
22260 PR tree-optimization/83435
22261 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
22262 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
22263 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
22265 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
22266 Alan Hayward <alan.hayward@arm.com>
22267 David Sherwood <david.sherwood@arm.com>
22269 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
22271 (aarch64_classify_address): Initialize it. Track polynomial offsets.
22272 (aarch64_print_address_internal): Use it to check for a zero offset.
22274 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
22275 Alan Hayward <alan.hayward@arm.com>
22276 David Sherwood <david.sherwood@arm.com>
22278 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
22279 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
22280 Return a poly_int64 rather than a HOST_WIDE_INT.
22281 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
22282 rather than a HOST_WIDE_INT.
22283 * config/aarch64/aarch64.h (aarch64_frame): Protect with
22284 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
22285 hard_fp_offset, frame_size, initial_adjust, callee_offset and
22286 final_offset from HOST_WIDE_INT to poly_int64.
22287 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
22288 to_constant when getting the number of units in an Advanced SIMD
22290 (aarch64_builtin_vectorized_function): Check for a constant number
22292 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
22294 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
22295 attribute instead of GET_MODE_NUNITS.
22296 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
22297 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
22298 GET_MODE_SIZE for fixed-size registers.
22299 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
22300 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
22301 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
22302 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
22303 (aarch64_print_operand, aarch64_print_address_internal)
22304 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
22305 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
22306 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
22307 Handle polynomial GET_MODE_SIZE.
22308 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
22309 wider than SImode without modification.
22310 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
22311 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
22312 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
22313 passing and returning SVE modes.
22314 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
22315 rather than GEN_INT.
22316 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
22317 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
22318 (aarch64_allocate_and_probe_stack_space): Likewise.
22319 (aarch64_layout_frame): Cope with polynomial offsets.
22320 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
22321 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
22322 polynomial offsets.
22323 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
22324 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
22325 poly_int64 rather than a HOST_WIDE_INT.
22326 (aarch64_get_separate_components, aarch64_process_components)
22327 (aarch64_expand_prologue, aarch64_expand_epilogue)
22328 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
22329 (aarch64_anchor_offset): New function, split out from...
22330 (aarch64_legitimize_address): ...here.
22331 (aarch64_builtin_vectorization_cost): Handle polynomial
22332 TYPE_VECTOR_SUBPARTS.
22333 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
22335 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
22336 number of elements from the PARALLEL rather than the mode.
22337 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
22338 rather than GET_MODE_BITSIZE.
22339 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
22340 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
22341 (aarch64_expand_vec_perm_const_1): Handle polynomial
22342 d->perm.length () and d->perm elements.
22343 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
22344 Apply to_constant to d->perm elements.
22345 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
22346 polynomial CONST_VECTOR_NUNITS.
22347 (aarch64_move_pointer): Take amount as a poly_int64 rather
22349 (aarch64_progress_pointer): Avoid temporary variable.
22350 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
22351 the mode attribute instead of GET_MODE.
22353 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
22354 Alan Hayward <alan.hayward@arm.com>
22355 David Sherwood <david.sherwood@arm.com>
22357 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
22358 x exists before using it.
22359 (aarch64_add_constant_internal): Rename to...
22360 (aarch64_add_offset_1): ...this. Replace regnum with separate
22361 src and dest rtxes. Handle the case in which they're different,
22362 including when the offset is zero. Replace scratchreg with an rtx.
22363 Use 2 additions if there is no spare register into which we can
22364 move a 16-bit constant.
22365 (aarch64_add_constant): Delete.
22366 (aarch64_add_offset): Replace reg with separate src and dest
22367 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
22368 Use aarch64_add_offset_1.
22369 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
22370 an rtx rather than an int. Take the delta as a poly_int64
22371 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
22372 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
22373 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
22374 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
22375 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
22376 and aarch64_add_sp.
22377 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
22378 aarch64_add_constant.
22380 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
22382 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
22383 Use scalar_float_mode.
22385 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
22387 * config/aarch64/aarch64-simd.md
22388 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
22389 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
22390 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
22391 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
22392 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
22393 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
22394 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
22395 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
22396 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
22397 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
22399 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
22402 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
22403 targ_options->x_arm_arch_string is non NULL.
22405 2018-01-11 Tamar Christina <tamar.christina@arm.com>
22407 * config/aarch64/aarch64.h
22408 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
22410 2018-01-11 Sudakshina Das <sudi.das@arm.com>
22413 * expmed.c (emit_store_flag_force): Swap if const op0
22414 and change VOIDmode to mode of op0.
22416 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
22418 PR rtl-optimization/83761
22419 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
22420 than bytes to mode_for_size.
22422 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
22424 PR middle-end/83189
22425 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
22426 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
22429 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
22431 PR middle-end/83575
22432 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
22433 when in layout mode.
22434 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
22435 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
22438 2018-01-10 Michael Collison <michael.collison@arm.com>
22440 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
22441 * config/aarch64/aarch64-option-extension.def: Add
22442 AARCH64_OPT_EXTENSION of 'fp16fml'.
22443 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
22444 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
22445 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
22446 * config/aarch64/constraints.md (Ui7): New constraint.
22447 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
22448 (VFMLA_SEL_W): Ditto.
22451 (VFMLA16_LOW): New int iterator.
22452 (VFMLA16_HIGH): Ditto.
22453 (UNSPEC_FMLAL): New unspec.
22454 (UNSPEC_FMLSL): Ditto.
22455 (UNSPEC_FMLAL2): Ditto.
22456 (UNSPEC_FMLSL2): Ditto.
22457 (f16mac): New code attribute.
22458 * config/aarch64/aarch64-simd-builtins.def
22459 (aarch64_fmlal_lowv2sf): Ditto.
22460 (aarch64_fmlsl_lowv2sf): Ditto.
22461 (aarch64_fmlalq_lowv4sf): Ditto.
22462 (aarch64_fmlslq_lowv4sf): Ditto.
22463 (aarch64_fmlal_highv2sf): Ditto.
22464 (aarch64_fmlsl_highv2sf): Ditto.
22465 (aarch64_fmlalq_highv4sf): Ditto.
22466 (aarch64_fmlslq_highv4sf): Ditto.
22467 (aarch64_fmlal_lane_lowv2sf): Ditto.
22468 (aarch64_fmlsl_lane_lowv2sf): Ditto.
22469 (aarch64_fmlal_laneq_lowv2sf): Ditto.
22470 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
22471 (aarch64_fmlalq_lane_lowv4sf): Ditto.
22472 (aarch64_fmlsl_lane_lowv4sf): Ditto.
22473 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
22474 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
22475 (aarch64_fmlal_lane_highv2sf): Ditto.
22476 (aarch64_fmlsl_lane_highv2sf): Ditto.
22477 (aarch64_fmlal_laneq_highv2sf): Ditto.
22478 (aarch64_fmlsl_laneq_highv2sf): Ditto.
22479 (aarch64_fmlalq_lane_highv4sf): Ditto.
22480 (aarch64_fmlsl_lane_highv4sf): Ditto.
22481 (aarch64_fmlalq_laneq_highv4sf): Ditto.
22482 (aarch64_fmlsl_laneq_highv4sf): Ditto.
22483 * config/aarch64/aarch64-simd.md:
22484 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
22485 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
22486 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
22487 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
22488 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
22489 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
22490 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
22491 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
22492 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
22493 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
22494 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
22495 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
22496 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
22497 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
22498 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
22499 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
22500 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
22501 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
22502 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
22503 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
22504 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
22505 (vfmlsl_low_u32): Ditto.
22506 (vfmlalq_low_u32): Ditto.
22507 (vfmlslq_low_u32): Ditto.
22508 (vfmlal_high_u32): Ditto.
22509 (vfmlsl_high_u32): Ditto.
22510 (vfmlalq_high_u32): Ditto.
22511 (vfmlslq_high_u32): Ditto.
22512 (vfmlal_lane_low_u32): Ditto.
22513 (vfmlsl_lane_low_u32): Ditto.
22514 (vfmlal_laneq_low_u32): Ditto.
22515 (vfmlsl_laneq_low_u32): Ditto.
22516 (vfmlalq_lane_low_u32): Ditto.
22517 (vfmlslq_lane_low_u32): Ditto.
22518 (vfmlalq_laneq_low_u32): Ditto.
22519 (vfmlslq_laneq_low_u32): Ditto.
22520 (vfmlal_lane_high_u32): Ditto.
22521 (vfmlsl_lane_high_u32): Ditto.
22522 (vfmlal_laneq_high_u32): Ditto.
22523 (vfmlsl_laneq_high_u32): Ditto.
22524 (vfmlalq_lane_high_u32): Ditto.
22525 (vfmlslq_lane_high_u32): Ditto.
22526 (vfmlalq_laneq_high_u32): Ditto.
22527 (vfmlslq_laneq_high_u32): Ditto.
22528 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
22529 (AARCH64_FL_FOR_ARCH8_4): New.
22530 (AARCH64_ISA_F16FML): New ISA flag.
22531 (TARGET_F16FML): New feature flag for fp16fml.
22532 (doc/invoke.texi): Document new fp16fml option.
22534 2018-01-10 Michael Collison <michael.collison@arm.com>
22536 * config/aarch64/aarch64-builtins.c:
22537 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
22538 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
22539 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
22540 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
22541 (AARCH64_ISA_SHA3): New ISA flag.
22542 (TARGET_SHA3): New feature flag for sha3.
22543 * config/aarch64/iterators.md (sha512_op): New int attribute.
22544 (CRYPTO_SHA512): New int iterator.
22545 (UNSPEC_SHA512H): New unspec.
22546 (UNSPEC_SHA512H2): Ditto.
22547 (UNSPEC_SHA512SU0): Ditto.
22548 (UNSPEC_SHA512SU1): Ditto.
22549 * config/aarch64/aarch64-simd-builtins.def
22550 (aarch64_crypto_sha512hqv2di): New builtin.
22551 (aarch64_crypto_sha512h2qv2di): Ditto.
22552 (aarch64_crypto_sha512su0qv2di): Ditto.
22553 (aarch64_crypto_sha512su1qv2di): Ditto.
22554 (aarch64_eor3qv8hi): Ditto.
22555 (aarch64_rax1qv2di): Ditto.
22556 (aarch64_xarqv2di): Ditto.
22557 (aarch64_bcaxqv8hi): Ditto.
22558 * config/aarch64/aarch64-simd.md:
22559 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
22560 (aarch64_crypto_sha512su0qv2di): Ditto.
22561 (aarch64_crypto_sha512su1qv2di): Ditto.
22562 (aarch64_eor3qv8hi): Ditto.
22563 (aarch64_rax1qv2di): Ditto.
22564 (aarch64_xarqv2di): Ditto.
22565 (aarch64_bcaxqv8hi): Ditto.
22566 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
22567 (vsha512h2q_u64): Ditto.
22568 (vsha512su0q_u64): Ditto.
22569 (vsha512su1q_u64): Ditto.
22570 (veor3q_u16): Ditto.
22571 (vrax1q_u64): Ditto.
22572 (vxarq_u64): Ditto.
22573 (vbcaxq_u16): Ditto.
22574 * config/arm/types.md (crypto_sha512): New type attribute.
22575 (crypto_sha3): Ditto.
22576 (doc/invoke.texi): Document new sha3 option.
22578 2018-01-10 Michael Collison <michael.collison@arm.com>
22580 * config/aarch64/aarch64-builtins.c:
22581 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
22582 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
22583 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
22584 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
22585 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
22586 (AARCH64_ISA_SM4): New ISA flag.
22587 (TARGET_SM4): New feature flag for sm4.
22588 * config/aarch64/aarch64-simd-builtins.def
22589 (aarch64_sm3ss1qv4si): Ditto.
22590 (aarch64_sm3tt1aq4si): Ditto.
22591 (aarch64_sm3tt1bq4si): Ditto.
22592 (aarch64_sm3tt2aq4si): Ditto.
22593 (aarch64_sm3tt2bq4si): Ditto.
22594 (aarch64_sm3partw1qv4si): Ditto.
22595 (aarch64_sm3partw2qv4si): Ditto.
22596 (aarch64_sm4eqv4si): Ditto.
22597 (aarch64_sm4ekeyqv4si): Ditto.
22598 * config/aarch64/aarch64-simd.md:
22599 (aarch64_sm3ss1qv4si): Ditto.
22600 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
22601 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
22602 (aarch64_sm4eqv4si): Ditto.
22603 (aarch64_sm4ekeyqv4si): Ditto.
22604 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
22605 (sm3part_op): Ditto.
22606 (CRYPTO_SM3TT): Ditto.
22607 (CRYPTO_SM3PART): Ditto.
22608 (UNSPEC_SM3SS1): New unspec.
22609 (UNSPEC_SM3TT1A): Ditto.
22610 (UNSPEC_SM3TT1B): Ditto.
22611 (UNSPEC_SM3TT2A): Ditto.
22612 (UNSPEC_SM3TT2B): Ditto.
22613 (UNSPEC_SM3PARTW1): Ditto.
22614 (UNSPEC_SM3PARTW2): Ditto.
22615 (UNSPEC_SM4E): Ditto.
22616 (UNSPEC_SM4EKEY): Ditto.
22617 * config/aarch64/constraints.md (Ui2): New constraint.
22618 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
22619 * config/arm/types.md (crypto_sm3): New type attribute.
22620 (crypto_sm4): Ditto.
22621 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
22622 (vsm3tt1aq_u32): Ditto.
22623 (vsm3tt1bq_u32): Ditto.
22624 (vsm3tt2aq_u32): Ditto.
22625 (vsm3tt2bq_u32): Ditto.
22626 (vsm3partw1q_u32): Ditto.
22627 (vsm3partw2q_u32): Ditto.
22628 (vsm4eq_u32): Ditto.
22629 (vsm4ekeyq_u32): Ditto.
22630 (doc/invoke.texi): Document new sm4 option.
22632 2018-01-10 Michael Collison <michael.collison@arm.com>
22634 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
22635 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
22636 (AARCH64_FL_FOR_ARCH8_4): New.
22637 (AARCH64_FL_V8_4): New flag.
22638 (doc/invoke.texi): Document new armv8.4-a option.
22640 2018-01-10 Michael Collison <michael.collison@arm.com>
22642 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
22643 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
22644 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
22645 * config/aarch64/aarch64-option-extension.def: Add
22646 AARCH64_OPT_EXTENSION of 'sha2'.
22647 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
22648 (crypto): Disable sha2 and aes if crypto disabled.
22649 (crypto): Enable aes and sha2 if enabled.
22650 (simd): Disable sha2 and aes if simd disabled.
22651 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
22653 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
22654 (TARGET_SHA2): New feature flag for sha2.
22655 (TARGET_AES): New feature flag for aes.
22656 * config/aarch64/aarch64-simd.md:
22657 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
22658 conditional on TARGET_AES.
22659 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
22660 (aarch64_crypto_sha1hsi): Make pattern conditional
22662 (aarch64_crypto_sha1hv4si): Ditto.
22663 (aarch64_be_crypto_sha1hv4si): Ditto.
22664 (aarch64_crypto_sha1su1v4si): Ditto.
22665 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
22666 (aarch64_crypto_sha1su0v4si): Ditto.
22667 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
22668 (aarch64_crypto_sha256su0v4si): Ditto.
22669 (aarch64_crypto_sha256su1v4si): Ditto.
22670 (doc/invoke.texi): Document new aes and sha2 options.
22672 2018-01-10 Martin Sebor <msebor@redhat.com>
22674 PR tree-optimization/83781
22675 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
22678 2018-01-11 Martin Sebor <msebor@gmail.com>
22679 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
22681 PR tree-optimization/83501
22682 PR tree-optimization/81703
22684 * tree-ssa-strlen.c (get_string_cst): Rename...
22685 (get_string_len): ...to this. Handle global constants.
22686 (handle_char_store): Adjust.
22688 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
22689 Jim Wilson <jimw@sifive.com>
22691 * config/riscv/riscv-protos.h (riscv_output_return): New.
22692 * config/riscv/riscv.c (struct machine_function): New naked_p field.
22693 (riscv_attribute_table, riscv_output_return),
22694 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
22695 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
22696 (riscv_compute_frame_info): Only compute frame->mask if not a naked
22698 (riscv_expand_prologue): Add early return for naked function.
22699 (riscv_expand_epilogue): Likewise.
22700 (riscv_function_ok_for_sibcall): Return false for naked function.
22701 (riscv_set_current_function): New.
22702 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
22703 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
22704 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
22705 * doc/extend.texi (RISC-V Function Attributes): New.
22707 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
22709 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
22710 check for 128-bit long double before checking TCmode.
22711 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
22712 128-bit long doubles before checking TFmode or TCmode.
22713 (FLOAT128_IBM_P): Likewise.
22715 2018-01-10 Martin Sebor <msebor@redhat.com>
22717 PR tree-optimization/83671
22718 * builtins.c (c_strlen): Unconditionally return zero for the empty
22720 Use -Warray-bounds for warnings.
22721 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
22722 for non-constant array indices with COMPONENT_REF, arrays of
22723 arrays, and pointers to arrays.
22724 (gimple_fold_builtin_strlen): Determine and set length range for
22725 non-constant character arrays.
22727 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
22729 PR middle-end/81897
22730 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
22733 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
22735 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
22737 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
22740 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
22741 VECTOR_MEM_ALTIVEC_OR_VSX_P.
22742 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
22743 indexed_or_indirect_operand predicate.
22744 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
22745 (*vsx_le_perm_load_v8hi): Likewise.
22746 (*vsx_le_perm_load_v16qi): Likewise.
22747 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
22748 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
22749 (*vsx_le_perm_store_v8hi): Likewise.
22750 (*vsx_le_perm_store_v16qi): Likewise.
22751 (eight unnamed splitters): Likewise.
22753 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
22755 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
22756 * config/rs6000/emmintrin.h: Likewise.
22757 * config/rs6000/mmintrin.h: Likewise.
22758 * config/rs6000/xmmintrin.h: Likewise.
22760 2018-01-10 David Malcolm <dmalcolm@redhat.com>
22763 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
22765 * tree.c (tree_nop_conversion): Return true for location wrapper
22767 (maybe_wrap_with_location): New function.
22768 (selftest::check_strip_nops): New function.
22769 (selftest::test_location_wrappers): New function.
22770 (selftest::tree_c_tests): Call it.
22771 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
22772 (maybe_wrap_with_location): New decl.
22773 (EXPR_LOCATION_WRAPPER_P): New macro.
22774 (location_wrapper_p): New inline function.
22775 (tree_strip_any_location_wrapper): New inline function.
22777 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
22780 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
22781 stack_realign_offset for the largest alignment of stack slot
22783 (ix86_find_max_used_stack_alignment): New function.
22784 (ix86_finalize_stack_frame_flags): Use it. Set
22785 max_used_stack_alignment if we don't realign stack.
22786 * config/i386/i386.h (machine_function): Add
22787 max_used_stack_alignment.
22789 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
22791 * config/arm/arm.opt (-mbranch-cost): New option.
22792 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
22795 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
22798 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
22799 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
22801 2018-01-10 Richard Biener <rguenther@suse.de>
22804 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
22805 early out so it also covers the case where we have a non-NULL
22808 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
22810 PR tree-optimization/83753
22811 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
22812 for non-strided grouped accesses if the number of elements is 1.
22814 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
22817 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
22818 * i386.h (TARGET_USE_GATHER): Define.
22819 * x86-tune.def (X86_TUNE_USE_GATHER): New.
22821 2018-01-10 Martin Liska <mliska@suse.cz>
22824 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
22825 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
22827 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
22828 CLEANUP_NO_PARTITIONING is not set.
22830 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
22832 * doc/rtl.texi: Remove documentation of (const ...) wrappers
22833 for vectors, as a partial revert of r254296.
22834 * rtl.h (const_vec_p): Delete.
22835 (const_vec_duplicate_p): Don't test for vector CONSTs.
22836 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
22837 * expmed.c (make_tree): Likewise.
22840 * common.md (E, F): Use CONSTANT_P instead of checking for
22842 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
22843 checking for CONST_VECTOR.
22845 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
22847 PR middle-end/83575
22848 * predict.c (force_edge_cold): Handle in more sane way edges
22849 with no prediction.
22851 2018-01-09 Carl Love <cel@us.ibm.com>
22853 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
22855 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
22856 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
22857 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
22858 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
22859 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
22860 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
22861 * config/rs6000/rs6000-protos.h: Add extern defition for
22862 rs6000_generate_float2_double_code.
22863 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
22865 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
22866 (float2_v2df): Add define_expand.
22868 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
22871 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
22872 op_mode in the force_to_mode call.
22874 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
22876 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
22877 instead of checking each element individually.
22878 (aarch64_evpc_uzp): Likewise.
22879 (aarch64_evpc_zip): Likewise.
22880 (aarch64_evpc_ext): Likewise.
22881 (aarch64_evpc_rev): Likewise.
22882 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
22883 instead of checking each element individually. Return true without
22885 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
22886 whether all selected elements come from the same input, instead of
22887 checking each element individually. Remove calls to gen_rtx_REG,
22888 start_sequence and end_sequence and instead assert that no rtl is
22891 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
22893 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
22894 order of HIGH and CONST checks.
22896 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
22898 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
22899 if the destination isn't an SSA_NAME.
22901 2018-01-09 Richard Biener <rguenther@suse.de>
22903 PR tree-optimization/83668
22904 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
22906 (canonicalize_loop_form): ... here, renamed from ...
22907 (canonicalize_loop_closed_ssa_form): ... this and amended to
22908 swap successor edges for loop exit blocks to make us use
22909 the RPO order we need for initial schedule generation.
22911 2018-01-09 Joseph Myers <joseph@codesourcery.com>
22913 PR tree-optimization/64811
22914 * match.pd: When optimizing comparisons with Inf, avoid
22915 introducing or losing exceptions from comparisons with NaN.
22917 2018-01-09 Martin Liska <mliska@suse.cz>
22920 * asan.c (shadow_mem_size): Add gcc_assert.
22922 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
22924 Don't save registers in main().
22927 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
22928 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
22929 * config/avr/avr.c (avr_set_current_function): Don't error if
22930 naked, OS_task or OS_main are specified at the same time.
22931 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
22933 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
22935 * common/config/avr/avr-common.c (avr_option_optimization_table):
22936 Switch on -mmain-is-OS_task for optimizing compilations.
22938 2018-01-09 Richard Biener <rguenther@suse.de>
22940 PR tree-optimization/83572
22941 * graphite.c: Include cfganal.h.
22942 (graphite_transform_loops): Connect infinite loops to exit
22943 and remove fake edges at the end.
22945 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
22947 * ipa-inline.c (edge_badness): Revert accidental checkin.
22949 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
22952 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
22953 symbols; not inline clones.
22955 2018-01-09 Jakub Jelinek <jakub@redhat.com>
22958 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
22959 hard registers. Formatting fixes.
22961 PR preprocessor/83722
22962 * gcc.c (try_generate_repro): Pass
22963 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
22964 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
22967 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
22968 Kito Cheng <kito.cheng@gmail.com>
22970 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
22971 (riscv_leaf_function_p): Delete.
22972 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
22974 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
22976 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
22978 (do_ifelse): New function.
22979 (do_isel): New function.
22980 (do_sub3): New function.
22981 (do_add3): New function.
22982 (do_load_mask_compare): New function.
22983 (do_overlap_load_compare): New function.
22984 (expand_compare_loop): New function.
22985 (expand_block_compare): Call expand_compare_loop() when appropriate.
22986 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
22987 option description.
22988 (-mblock-compare-inline-loop-limit): New option.
22990 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
22993 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
22994 Reverse order of second and third operands in first alternative.
22995 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
22996 of first and second elements in UNSPEC_VPERMR vector.
22997 (altivec_expand_vec_perm_le): Likewise.
22999 2018-01-08 Jeff Law <law@redhat.com>
23001 PR rtl-optimizatin/81308
23002 * tree-switch-conversion.c (cfg_altered): New file scoped static.
23003 (process_switch): If group_case_labels makes a change, then set
23005 (pass_convert_switch::execute): If a switch is converted, then
23006 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
23008 PR rtl-optimization/81308
23009 * recog.c (split_all_insns): Conditionally cleanup the CFG after
23012 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
23014 PR target/83663 - Revert r255946
23015 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
23016 generation for cases where splatting a value is not useful.
23017 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
23018 across a vec_duplicate and a paradoxical subreg forming a vector
23019 mode to a vec_concat.
23021 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23023 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
23024 -march=armv8.3-a variants.
23025 * config/arm/t-multilib: Likewise.
23026 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
23028 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
23030 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
23032 (cceq_ior_compare_complement): Give it a name so I can use it, and
23033 change boolean_or_operator predicate to boolean_operator so it can
23034 be used to generate a crand.
23035 (eqne): New code iterator.
23036 (bd/bd_neg): New code_attrs.
23037 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
23038 a single define_insn.
23039 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
23040 decrement (bdnzt/bdnzf/bdzt/bdzf).
23041 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
23042 with the new names of the branch decrement patterns, and added the
23043 names of the branch decrement conditional patterns.
23045 2018-01-08 Richard Biener <rguenther@suse.de>
23047 PR tree-optimization/83563
23048 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
23051 2018-01-08 Richard Biener <rguenther@suse.de>
23053 PR middle-end/83713
23054 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
23056 2018-01-08 Richard Biener <rguenther@suse.de>
23058 PR tree-optimization/83685
23059 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
23060 references to abnormals.
23062 2018-01-08 Richard Biener <rguenther@suse.de>
23065 * dwarf2out.c (output_indirect_strings): Handle empty
23066 skeleton_debug_str_hash.
23067 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
23069 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
23071 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
23072 (emit_store_direct): Likewise.
23073 (arc_trampoline_adjust_address): Likewise.
23074 (arc_asm_trampoline_template): New function.
23075 (arc_initialize_trampoline): Use asm_trampoline_template.
23076 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
23077 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
23078 * config/arc/arc.md (flush_icache): Delete pattern.
23080 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
23082 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
23083 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
23086 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
23089 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
23090 by not USED_FOR_TARGET.
23091 (make_pass_resolve_sw_modes): Likewise.
23093 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
23095 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
23098 2018-01-08 Richard Biener <rguenther@suse.de>
23100 PR middle-end/83580
23101 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
23103 2018-01-08 Richard Biener <rguenther@suse.de>
23105 PR middle-end/83517
23106 * match.pd ((t * 2) / 2) -> t): Add missing :c.
23108 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
23110 PR middle-end/81897
23111 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
23112 basic blocks with a small number of successors.
23113 (convert_control_dep_chain_into_preds): Improve handling of
23115 (dump_predicates): Split apart into...
23116 (dump_pred_chain): ...here...
23117 (dump_pred_info): ...and here.
23118 (can_one_predicate_be_invalidated_p): Add debugging printfs.
23119 (can_chain_union_be_invalidated_p): Improve check for invalidation
23121 (uninit_uses_cannot_happen): Avoid unnecessary if
23122 convert_control_dep_chain_into_preds yielded nothing.
23124 2018-01-06 Martin Sebor <msebor@redhat.com>
23126 PR tree-optimization/83640
23127 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
23128 subtracting negative offset from size.
23129 (builtin_access::overlap): Adjust offset bounds of the access to fall
23130 within the size of the object if possible.
23132 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
23134 PR rtl-optimization/83699
23135 * expmed.c (extract_bit_field_1): Restrict the vector usage of
23136 extract_bit_field_as_subreg to cases in which the extracted
23137 value is also a vector.
23139 * lra-constraints.c (process_alt_operands): Test for the equivalence
23140 substitutions when detecting a possible reload cycle.
23142 2018-01-06 Jakub Jelinek <jakub@redhat.com>
23145 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
23146 by default if flag_selective_schedling{,2}. Formatting fixes.
23148 PR rtl-optimization/83682
23149 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
23150 if it has non-VECTOR_MODE element mode.
23151 (vec_duplicate_p): Likewise.
23153 PR middle-end/83694
23154 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
23155 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
23157 2018-01-05 Jakub Jelinek <jakub@redhat.com>
23160 * config/i386/i386-builtin.def
23161 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
23162 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
23163 Require also OPTION_MASK_ISA_AVX512F in addition to
23164 OPTION_MASK_ISA_GFNI.
23165 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
23166 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
23167 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
23168 to OPTION_MASK_ISA_GFNI.
23169 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
23170 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
23171 OPTION_MASK_ISA_AVX512BW.
23172 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
23173 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
23174 addition to OPTION_MASK_ISA_GFNI.
23175 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
23176 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
23177 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
23178 to OPTION_MASK_ISA_GFNI.
23179 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
23180 a requirement for all ISAs rather than any of them with a few
23182 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
23184 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
23185 bitmasks to be enabled with 3 exceptions, instead of requiring any
23186 enabled ISA with lots of exceptions.
23187 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
23188 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
23189 Change avx512bw in isa attribute to avx512f.
23190 * config/i386/sgxintrin.h: Add license boilerplate.
23191 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
23192 to __AVX512F__ and __AVX512VL to __AVX512VL__.
23193 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
23194 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
23196 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
23197 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
23198 temporarily sse2 rather than sse if not enabled already.
23201 * config/i386/sse.md (VI248_VLBW): Rename to ...
23202 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
23203 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
23204 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
23205 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
23206 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
23207 mode iterator instead of VI248_VLBW.
23209 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
23211 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
23212 (record_modified): Skip clobbers; add debug output.
23213 (param_change_prob): Use sreal frequencies.
23215 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
23217 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
23218 punt for user-aligned variables.
23220 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
23222 * tree-chrec.c (chrec_contains_symbols): Return true for
23225 2018-01-05 Sudakshina Das <sudi.das@arm.com>
23228 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
23229 of (x|y) == x for BICS pattern.
23231 2018-01-05 Jakub Jelinek <jakub@redhat.com>
23233 PR tree-optimization/83605
23234 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
23235 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
23238 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
23240 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
23241 * config/epiphany/rtems.h: New file.
23243 2018-01-04 Jakub Jelinek <jakub@redhat.com>
23244 Uros Bizjak <ubizjak@gmail.com>
23247 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
23248 QIreg_operand instead of register_operand predicate.
23249 * config/i386/i386.c (ix86_rop_should_change_byte_p,
23250 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
23251 comments instead of -fmitigate[-_]rop.
23253 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
23256 * cgraphunit.c (symbol_table::compile): Switch to text_section
23257 before calling assembly_start debug hook.
23258 * run-rtl-passes.c (run_rtl_passes): Likewise.
23261 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
23263 * tree-vrp.c (extract_range_from_binary_expr_1): Check
23264 range_int_cst_p rather than !symbolic_range_p before calling
23265 extract_range_from_multiplicative_op_1.
23267 2018-01-04 Jeff Law <law@redhat.com>
23269 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
23270 redundant test in assertion.
23272 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
23274 * doc/rtl.texi: Document machine_mode wrapper classes.
23276 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
23278 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
23279 using tree_to_uhwi.
23281 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
23283 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
23284 the VEC_PERM_EXPR fold to fail.
23286 2018-01-04 Jakub Jelinek <jakub@redhat.com>
23289 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
23290 to switched_sections.
23292 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
23295 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
23296 test for d.testing.
23298 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
23301 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
23302 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
23304 2018-01-04 Jakub Jelinek <jakub@redhat.com>
23307 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
23308 is BLKmode and bitpos not zero or mode change is needed.
23310 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
23313 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
23316 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
23319 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
23320 instead of MULT rtx. Update all corresponding splitters.
23321 (*saddl_se): Ditto.
23322 (*ssub<modesuffix>): Ditto.
23323 (*ssubl_se): Ditto.
23324 (*cmp_sadd_di): Update split patterns.
23325 (*cmp_sadd_si): Ditto.
23326 (*cmp_sadd_sidi): Ditto.
23327 (*cmp_ssub_di): Ditto.
23328 (*cmp_ssub_si): Ditto.
23329 (*cmp_ssub_sidi): Ditto.
23330 * config/alpha/predicates.md (const23_operand): New predicate.
23331 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
23332 Look for ASHIFT, not MULT inner operand.
23333 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
23335 2018-01-04 Martin Liska <mliska@suse.cz>
23337 PR gcov-profile/83669
23338 * gcov.c (output_intermediate_file): Add version to intermediate
23340 * doc/gcov.texi: Document new field 'version' in intermediate
23341 file format. Fix location of '-k' option of gcov command.
23343 2018-01-04 Martin Liska <mliska@suse.cz>
23346 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
23348 2018-01-04 Jakub Jelinek <jakub@redhat.com>
23350 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
23352 2018-01-03 Martin Sebor <msebor@redhat.com>
23354 PR tree-optimization/83655
23355 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
23356 checking calls with invalid arguments.
23358 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23360 * tree-vect-stmts.c (vect_get_store_rhs): New function.
23361 (vectorizable_mask_load_store): Delete.
23362 (vectorizable_call): Return false for masked loads and stores.
23363 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
23364 instead of gimple_assign_rhs1.
23365 (vectorizable_load): Handle IFN_MASK_LOAD.
23366 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
23368 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23370 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
23372 (vectorizable_mask_load_store): ...here.
23373 (vectorizable_load): ...and here.
23375 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23377 * tree-vect-stmts.c (vect_build_all_ones_mask)
23378 (vect_build_zero_merge_argument): New functions, split out from...
23379 (vectorizable_load): ...here.
23381 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23383 * tree-vect-stmts.c (vect_check_store_rhs): New function,
23385 (vectorizable_mask_load_store): ...here.
23386 (vectorizable_store): ...and here.
23388 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23390 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
23392 (vectorizable_mask_load_store): ...here.
23394 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23396 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
23397 (vect_model_store_cost): Take a vec_load_store_type instead of a
23399 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
23400 (vect_model_store_cost): Take a vec_load_store_type instead of a
23402 (vectorizable_mask_load_store): Update accordingly.
23403 (vectorizable_store): Likewise.
23404 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
23406 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23408 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
23409 IFN_MASK_LOAD calls here rather than...
23410 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
23412 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23413 Alan Hayward <alan.hayward@arm.com>
23414 David Sherwood <david.sherwood@arm.com>
23416 * expmed.c (extract_bit_field_1): For vector extracts,
23417 fall back to extract_bit_field_as_subreg if vec_extract
23420 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23421 Alan Hayward <alan.hayward@arm.com>
23422 David Sherwood <david.sherwood@arm.com>
23424 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
23425 they are variable or constant sized.
23426 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
23427 slots for constant-sized data.
23429 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23430 Alan Hayward <alan.hayward@arm.com>
23431 David Sherwood <david.sherwood@arm.com>
23433 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
23434 handling COND_EXPRs with boolean comparisons, try to find a better
23435 basis for the mask type than the boolean itself.
23437 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23439 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
23440 is calculated and how it can be overridden.
23441 * genmodes.c (max_bitsize_mode_any_mode): New variable.
23442 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
23444 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
23447 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23448 Alan Hayward <alan.hayward@arm.com>
23449 David Sherwood <david.sherwood@arm.com>
23451 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
23452 Remove the mode argument.
23453 (aarch64_simd_valid_immediate): Remove the mode and inverse
23455 * config/aarch64/iterators.md (bitsize): New iterator.
23456 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
23457 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
23458 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
23459 aarch64_simd_valid_immediate.
23460 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
23461 (aarch64_reg_or_bic_imm): Likewise.
23462 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
23463 with an insn_type enum and msl with a modifier_type enum.
23464 Replace element_width with a scalar_mode. Change the shift
23465 to unsigned int. Add constructors for scalar_float_mode and
23466 scalar_int_mode elements.
23467 (aarch64_vect_float_const_representable_p): Delete.
23468 (aarch64_can_const_movi_rtx_p)
23469 (aarch64_simd_scalar_immediate_valid_for_move)
23470 (aarch64_simd_make_constant): Update call to
23471 aarch64_simd_valid_immediate.
23472 (aarch64_advsimd_valid_immediate_hs): New function.
23473 (aarch64_advsimd_valid_immediate): Likewise.
23474 (aarch64_simd_valid_immediate): Remove mode and inverse
23475 arguments. Rewrite to use the above. Use const_vec_duplicate_p
23476 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
23477 and aarch64_float_const_representable_p on the result.
23478 (aarch64_output_simd_mov_immediate): Remove mode argument.
23479 Update call to aarch64_simd_valid_immediate and use of
23480 simd_immediate_info.
23481 (aarch64_output_scalar_simd_mov_immediate): Update call
23484 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23485 Alan Hayward <alan.hayward@arm.com>
23486 David Sherwood <david.sherwood@arm.com>
23488 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
23489 (mode_nunits): Likewise CONST_MODE_NUNITS.
23490 * machmode.def (ADJUST_NUNITS): Document.
23491 * genmodes.c (mode_data::need_nunits_adj): New field.
23492 (blank_mode): Update accordingly.
23493 (adj_nunits): New variable.
23494 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
23496 (emit_mode_size_inline): Set need_bytesize_adj for all modes
23497 listed in adj_nunits.
23498 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
23499 listed in adj_nunits. Don't emit case statements for such modes.
23500 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
23501 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
23502 nothing if adj_nunits is nonnull.
23503 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
23504 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
23505 (emit_mode_fbit): Update use of print_maybe_const_decl.
23506 (emit_move_size): Likewise. Treat the array as non-const
23508 (emit_mode_adjustments): Handle adj_nunits.
23510 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23512 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
23513 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
23514 (VECTOR_MODES): Use it.
23515 (make_vector_modes): Take the prefix as an argument.
23517 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23518 Alan Hayward <alan.hayward@arm.com>
23519 David Sherwood <david.sherwood@arm.com>
23521 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
23522 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
23523 for MODE_VECTOR_BOOL.
23524 * machmode.def (VECTOR_BOOL_MODE): Document.
23525 * genmodes.c (VECTOR_BOOL_MODE): New macro.
23526 (make_vector_bool_mode): New function.
23527 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
23529 * lto-streamer-in.c (lto_input_mode_table): Likewise.
23530 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
23532 * stor-layout.c (int_mode_for_mode): Likewise.
23533 * tree.c (build_vector_type_for_mode): Likewise.
23534 * varasm.c (output_constant_pool_2): Likewise.
23535 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
23536 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
23537 for MODE_VECTOR_BOOL.
23538 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
23539 of mode class checks.
23540 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
23541 instead of a list of mode class checks.
23542 (expand_vector_scalar_condition): Likewise.
23543 (type_for_widest_vector_mode): Handle BImode as an inner mode.
23545 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23546 Alan Hayward <alan.hayward@arm.com>
23547 David Sherwood <david.sherwood@arm.com>
23549 * machmode.h (mode_size): Change from unsigned short to
23551 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
23552 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
23553 or if measurement_type is not polynomial.
23554 (fixed_size_mode::includes_p): Check for constant-sized modes.
23555 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
23556 return a poly_uint16 rather than an unsigned short.
23557 (emit_mode_size): Change the type of mode_size from unsigned short
23558 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
23559 (emit_mode_adjustments): Cope with polynomial vector sizes.
23560 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
23562 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
23564 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
23565 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
23566 * caller-save.c (setup_save_areas): Likewise.
23567 (replace_reg_with_saved_mem): Likewise.
23568 * calls.c (emit_library_call_value_1): Likewise.
23569 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
23570 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
23571 (gen_lowpart_for_combine): Likewise.
23572 * convert.c (convert_to_integer_1): Likewise.
23573 * cse.c (equiv_constant, cse_insn): Likewise.
23574 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
23575 (cselib_subst_to_values): Likewise.
23576 * dce.c (word_dce_process_block): Likewise.
23577 * df-problems.c (df_word_lr_mark_ref): Likewise.
23578 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
23579 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
23580 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
23581 (rtl_for_decl_location): Likewise.
23582 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
23583 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
23584 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
23585 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
23586 (expand_expr_real_1): Likewise.
23587 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
23588 (pad_below): Likewise.
23589 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
23590 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
23591 * ira.c (get_subreg_tracking_sizes): Likewise.
23592 * ira-build.c (ira_create_allocno_objects): Likewise.
23593 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
23594 (ira_sort_regnos_for_alter_reg): Likewise.
23595 * ira-costs.c (record_operand_costs): Likewise.
23596 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
23597 (resolve_simple_move): Likewise.
23598 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
23599 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
23600 (lra_constraints): Likewise.
23601 (CONST_POOL_OK_P): Reject variable-sized modes.
23602 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
23603 (add_pseudo_to_slot, lra_spill): Likewise.
23604 * omp-low.c (omp_clause_aligned_alignment): Likewise.
23605 * optabs-query.c (get_best_extraction_insn): Likewise.
23606 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
23607 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
23608 (expand_mult_highpart, valid_multiword_target_p): Likewise.
23609 * recog.c (offsettable_address_addr_space_p): Likewise.
23610 * regcprop.c (maybe_mode_change): Likewise.
23611 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
23612 * regrename.c (build_def_use): Likewise.
23613 * regstat.c (dump_reg_info): Likewise.
23614 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
23615 (find_reloads, find_reloads_subreg_address): Likewise.
23616 * reload1.c (eliminate_regs_1): Likewise.
23617 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
23618 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
23619 (simplify_binary_operation_1, simplify_subreg): Likewise.
23620 * targhooks.c (default_function_arg_padding): Likewise.
23621 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
23622 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
23623 (verify_gimple_assign_ternary): Likewise.
23624 * tree-inline.c (estimate_move_cost): Likewise.
23625 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
23626 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
23627 (get_address_cost_ainc): Likewise.
23628 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
23629 (vect_supportable_dr_alignment): Likewise.
23630 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
23631 (vectorizable_reduction): Likewise.
23632 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
23633 (vectorizable_operation, vectorizable_load): Likewise.
23634 * tree.c (build_same_sized_truth_vector_type): Likewise.
23635 * valtrack.c (cleanup_auto_inc_dec): Likewise.
23636 * var-tracking.c (emit_note_insn_var_location): Likewise.
23637 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
23638 (ADDR_VEC_ALIGN): Likewise.
23640 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23641 Alan Hayward <alan.hayward@arm.com>
23642 David Sherwood <david.sherwood@arm.com>
23644 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
23646 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
23647 or if measurement_type is polynomial.
23648 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
23649 * combine.c (make_extraction): Likewise.
23650 * dse.c (find_shift_sequence): Likewise.
23651 * dwarf2out.c (mem_loc_descriptor): Likewise.
23652 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
23653 (extract_bit_field, extract_low_bits): Likewise.
23654 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
23655 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
23656 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
23657 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
23658 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
23659 * reload.c (find_reloads): Likewise.
23660 * reload1.c (alter_reg): Likewise.
23661 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
23662 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
23663 * tree-if-conv.c (predicate_mem_writes): Likewise.
23664 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
23665 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
23666 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
23667 * valtrack.c (dead_debug_insert_temp): Likewise.
23668 * varasm.c (mergeable_constant_section): Likewise.
23669 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
23671 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23672 Alan Hayward <alan.hayward@arm.com>
23673 David Sherwood <david.sherwood@arm.com>
23675 * expr.c (expand_assignment): Cope with polynomial mode sizes
23676 when assigning to a CONCAT.
23678 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23679 Alan Hayward <alan.hayward@arm.com>
23680 David Sherwood <david.sherwood@arm.com>
23682 * machmode.h (mode_precision): Change from unsigned short to
23684 (mode_to_precision): Return a poly_uint16 rather than an unsigned
23686 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
23687 or if measurement_type is not polynomial.
23688 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
23689 in which the mode is already known to be a scalar_int_mode.
23690 * genmodes.c (emit_mode_precision): Change the type of mode_precision
23691 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
23693 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
23694 for GET_MODE_PRECISION.
23695 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
23696 for GET_MODE_PRECISION.
23697 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
23699 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
23700 (expand_field_assignment, make_extraction): Likewise.
23701 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
23702 (get_last_value): Likewise.
23703 * convert.c (convert_to_integer_1): Likewise.
23704 * cse.c (cse_insn): Likewise.
23705 * expr.c (expand_expr_real_1): Likewise.
23706 * lra-constraints.c (simplify_operand_subreg): Likewise.
23707 * optabs-query.c (can_atomic_load_p): Likewise.
23708 * optabs.c (expand_atomic_load): Likewise.
23709 (expand_atomic_store): Likewise.
23710 * ree.c (combine_reaching_defs): Likewise.
23711 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
23712 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
23713 * tree.h (type_has_mode_precision_p): Likewise.
23714 * ubsan.c (instrument_si_overflow): Likewise.
23716 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23717 Alan Hayward <alan.hayward@arm.com>
23718 David Sherwood <david.sherwood@arm.com>
23720 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
23721 polynomial numbers of units.
23722 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
23723 (valid_vector_subparts_p): New function.
23724 (build_vector_type): Remove temporary shim and take the number
23725 of units as a poly_uint64 rather than an int.
23726 (build_opaque_vector_type): Take the number of units as a
23727 poly_uint64 rather than an int.
23728 * tree.c (build_vector_from_ctor): Handle polynomial
23729 TYPE_VECTOR_SUBPARTS.
23730 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
23731 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
23732 (build_vector_from_val): If the number of units is variable,
23733 use build_vec_duplicate_cst for constant operands and
23734 VEC_DUPLICATE_EXPR otherwise.
23735 (make_vector_type): Remove temporary is_constant ().
23736 (build_vector_type, build_opaque_vector_type): Take the number of
23737 units as a poly_uint64 rather than an int.
23738 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
23740 * cfgexpand.c (expand_debug_expr): Likewise.
23741 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
23742 (store_constructor, expand_expr_real_1): Likewise.
23743 (const_scalar_mask_from_tree): Likewise.
23744 * fold-const-call.c (fold_const_reduction): Likewise.
23745 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
23746 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
23747 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
23748 (fold_relational_const): Likewise.
23749 (native_interpret_vector): Likewise. Change the size from an
23750 int to an unsigned int.
23751 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
23752 TYPE_VECTOR_SUBPARTS.
23753 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
23754 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
23755 duplicating a non-constant operand into a variable-length vector.
23756 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
23757 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
23758 * ipa-icf.c (sem_variable::equals): Likewise.
23759 * match.pd: Likewise.
23760 * omp-simd-clone.c (simd_clone_subparts): Likewise.
23761 * print-tree.c (print_node): Likewise.
23762 * stor-layout.c (layout_type): Likewise.
23763 * targhooks.c (default_builtin_vectorization_cost): Likewise.
23764 * tree-cfg.c (verify_gimple_comparison): Likewise.
23765 (verify_gimple_assign_binary): Likewise.
23766 (verify_gimple_assign_ternary): Likewise.
23767 (verify_gimple_assign_single): Likewise.
23768 * tree-pretty-print.c (dump_generic_node): Likewise.
23769 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
23770 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
23771 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
23772 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
23773 (vect_shift_permute_load_chain): Likewise.
23774 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
23775 (expand_vector_condition, optimize_vector_constructor): Likewise.
23776 (lower_vec_perm, get_compute_type): Likewise.
23777 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
23778 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
23779 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
23780 (vect_recog_mask_conversion_pattern): Likewise.
23781 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
23782 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
23783 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
23784 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
23785 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
23786 (vectorizable_shift, vectorizable_operation, vectorizable_store)
23787 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
23788 (supportable_widening_operation): Likewise.
23789 (supportable_narrowing_operation): Likewise.
23790 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
23792 * varasm.c (output_constant): Likewise.
23794 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23795 Alan Hayward <alan.hayward@arm.com>
23796 David Sherwood <david.sherwood@arm.com>
23798 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
23799 so that both the length == 3 and length != 3 cases set up their
23800 own permute vectors. Add comments explaining why we know the
23801 number of elements is constant.
23802 (vect_permute_load_chain): Likewise.
23804 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23805 Alan Hayward <alan.hayward@arm.com>
23806 David Sherwood <david.sherwood@arm.com>
23808 * machmode.h (mode_nunits): Change from unsigned char to
23810 (ONLY_FIXED_SIZE_MODES): New macro.
23811 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
23812 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
23813 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
23815 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
23816 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
23817 or if measurement_type is not polynomial.
23818 * genmodes.c (ZERO_COEFFS): New macro.
23819 (emit_mode_nunits_inline): Make mode_nunits_inline return a
23821 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
23822 Use ZERO_COEFFS when emitting initializers.
23823 * data-streamer.h (bp_pack_poly_value): New function.
23824 (bp_unpack_poly_value): Likewise.
23825 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
23826 for GET_MODE_NUNITS.
23827 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
23828 for GET_MODE_NUNITS.
23829 * tree.c (make_vector_type): Remove temporary shim and make
23830 the real function take the number of units as a poly_uint64
23831 rather than an int.
23832 (build_vector_type_for_mode): Handle polynomial nunits.
23833 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
23834 * emit-rtl.c (const_vec_series_p_1): Likewise.
23835 (gen_rtx_CONST_VECTOR): Likewise.
23836 * fold-const.c (test_vec_duplicate_folding): Likewise.
23837 * genrecog.c (validate_pattern): Likewise.
23838 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
23839 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
23840 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
23841 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
23842 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
23843 * rtlanal.c (subreg_get_info): Likewise.
23844 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
23845 (vect_grouped_load_supported): Likewise.
23846 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
23847 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
23848 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
23849 (simplify_const_unary_operation, simplify_binary_operation_1)
23850 (simplify_const_binary_operation, simplify_ternary_operation)
23851 (test_vector_ops_duplicate, test_vector_ops): Likewise.
23852 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
23853 instead of CONST_VECTOR_NUNITS.
23854 * varasm.c (output_constant_pool_2): Likewise.
23855 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
23856 explicit-encoded elements in the XVEC for variable-length vectors.
23858 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23860 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
23862 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23863 Alan Hayward <alan.hayward@arm.com>
23864 David Sherwood <david.sherwood@arm.com>
23866 * coretypes.h (fixed_size_mode): Declare.
23867 (fixed_size_mode_pod): New typedef.
23868 * builtins.h (target_builtins::x_apply_args_mode)
23869 (target_builtins::x_apply_result_mode): Change type to
23870 fixed_size_mode_pod.
23871 * builtins.c (apply_args_size, apply_result_size, result_vector)
23872 (expand_builtin_apply_args_1, expand_builtin_apply)
23873 (expand_builtin_return): Update accordingly.
23875 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23877 * cse.c (hash_rtx_cb): Hash only the encoded elements.
23878 * cselib.c (cselib_hash_rtx): Likewise.
23879 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
23880 CONST_VECTOR encoding.
23882 2018-01-03 Jakub Jelinek <jakub@redhat.com>
23883 Jeff Law <law@redhat.com>
23886 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
23887 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
23888 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
23889 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
23892 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
23893 explicitly probe *sp in a noreturn function if there were any callee
23894 register saves or frame pointer is needed.
23896 2018-01-03 Jakub Jelinek <jakub@redhat.com>
23899 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
23900 BLKmode for ternary, binary or unary expressions.
23903 * var-tracking.c (delete_vta_debug_insn): New inline function.
23904 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
23905 insns from get_insns () to NULL instead of each bb separately.
23906 Use delete_vta_debug_insn. No longer static.
23907 (vt_debug_insns_local, variable_tracking_main_1): Adjust
23908 delete_vta_debug_insns callers.
23909 * rtl.h (delete_vta_debug_insns): Declare.
23910 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
23911 instead of variable_tracking_main.
23913 2018-01-03 Martin Sebor <msebor@redhat.com>
23915 PR tree-optimization/83603
23916 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
23917 arguments past the endof the argument list in functions declared
23918 without a prototype.
23919 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
23920 Avoid checking when arguments are null.
23922 2018-01-03 Martin Sebor <msebor@redhat.com>
23925 * doc/extend.texi (attribute const): Fix a typo.
23926 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
23927 issuing -Wsuggest-attribute for void functions.
23929 2018-01-03 Martin Sebor <msebor@redhat.com>
23931 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
23932 offset_int::from instead of wide_int::to_shwi.
23933 (maybe_diag_overlap): Remove assertion.
23934 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
23935 * gimple-ssa-sprintf.c (format_directive): Same.
23936 (parse_directive): Same.
23937 (sprintf_dom_walker::compute_format_length): Same.
23938 (try_substitute_return_value): Same.
23940 2018-01-03 Jeff Law <law@redhat.com>
23942 PR middle-end/83654
23943 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
23944 non-constant residual for zero at runtime and avoid probing in
23945 that case. Reorganize code for trailing problem to mirror handling
23948 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
23950 PR tree-optimization/83501
23951 * tree-ssa-strlen.c (get_string_cst): New.
23952 (handle_char_store): Call get_string_cst.
23954 2018-01-03 Martin Liska <mliska@suse.cz>
23956 PR tree-optimization/83593
23957 * tree-ssa-strlen.c: Include tree-cfg.h.
23958 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
23959 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
23960 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
23962 (strlen_dom_walker::before_dom_children): Call
23963 gimple_purge_dead_eh_edges. Dump tranformation with details
23965 (strlen_dom_walker::before_dom_children): Update call by adding
23966 new argument cleanup_eh.
23967 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
23969 2018-01-03 Martin Liska <mliska@suse.cz>
23972 * cif-code.def (VARIADIC_THUNK): New enum value.
23973 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
23976 2018-01-03 Jan Beulich <jbeulich@suse.com>
23978 * sse.md (mov<mode>_internal): Tighten condition for when to use
23979 vmovdqu<ssescalarsize> for TI and OI modes.
23981 2018-01-03 Jakub Jelinek <jakub@redhat.com>
23983 Update copyright years.
23985 2018-01-03 Martin Liska <mliska@suse.cz>
23988 * ipa-visibility.c (function_and_variable_visibility): Skip
23989 functions with noipa attribure.
23991 2018-01-03 Jakub Jelinek <jakub@redhat.com>
23993 * gcc.c (process_command): Update copyright notice dates.
23994 * gcov-dump.c (print_version): Ditto.
23995 * gcov.c (print_version): Ditto.
23996 * gcov-tool.c (print_version): Ditto.
23997 * gengtype.c (create_file): Ditto.
23998 * doc/cpp.texi: Bump @copying's copyright year.
23999 * doc/cppinternals.texi: Ditto.
24000 * doc/gcc.texi: Ditto.
24001 * doc/gccint.texi: Ditto.
24002 * doc/gcov.texi: Ditto.
24003 * doc/install.texi: Ditto.
24004 * doc/invoke.texi: Ditto.
24006 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24008 * vector-builder.h (vector_builder::m_full_nelts): Change from
24009 unsigned int to poly_uint64.
24010 (vector_builder::full_nelts): Update prototype accordingly.
24011 (vector_builder::new_vector): Likewise.
24012 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
24013 (vector_builder::operator ==): Likewise.
24014 (vector_builder::finalize): Likewise.
24015 * int-vector-builder.h (int_vector_builder::int_vector_builder):
24016 Take the number of elements as a poly_uint64 rather than an
24018 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
24019 from unsigned int to poly_uint64.
24020 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
24021 (vec_perm_indices::new_vector): Likewise.
24022 (vec_perm_indices::length): Likewise.
24023 (vec_perm_indices::nelts_per_input): Likewise.
24024 (vec_perm_indices::input_nelts): Likewise.
24025 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
24026 number of elements per input as a poly_uint64 rather than an
24027 unsigned int. Use the original encoding for variable-length
24028 vectors, rather than clamping each individual element.
24029 For the second and subsequent elements in each pattern,
24030 clamp the step and base before clamping their sum.
24031 (vec_perm_indices::series_p): Handle polynomial element counts.
24032 (vec_perm_indices::all_in_range_p): Likewise.
24033 (vec_perm_indices_to_tree): Likewise.
24034 (vec_perm_indices_to_rtx): Likewise.
24035 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
24036 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
24037 (tree_vector_builder::new_binary_operation): Handle polynomial
24038 element counts. Return false if we need to know the number
24039 of elements at compile time.
24040 * fold-const.c (fold_vec_perm): Punt if the number of elements
24041 isn't known at compile time.
24043 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24045 * vec-perm-indices.h (vec_perm_builder): Change element type
24046 from HOST_WIDE_INT to poly_int64.
24047 (vec_perm_indices::element_type): Update accordingly.
24048 (vec_perm_indices::clamp): Handle polynomial element_types.
24049 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
24050 (vec_perm_indices::all_in_range_p): Likewise.
24051 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
24053 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
24054 polynomial vec_perm_indices element types.
24055 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
24056 * fold-const.c (fold_vec_perm): Likewise.
24057 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
24058 * tree-vect-generic.c (lower_vec_perm): Likewise.
24059 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
24060 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
24061 element type to HOST_WIDE_INT.
24063 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24064 Alan Hayward <alan.hayward@arm.com>
24065 David Sherwood <david.sherwood@arm.com>
24067 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
24068 rather than an int. Use plus_constant.
24069 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
24070 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
24072 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24073 Alan Hayward <alan.hayward@arm.com>
24074 David Sherwood <david.sherwood@arm.com>
24076 * calls.c (emit_call_1, expand_call): Change struct_value_size from
24077 a HOST_WIDE_INT to a poly_int64.
24079 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24080 Alan Hayward <alan.hayward@arm.com>
24081 David Sherwood <david.sherwood@arm.com>
24083 * calls.c (load_register_parameters): Cope with polynomial
24084 mode sizes. Require a constant size for BLKmode parameters
24085 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
24086 forces a parameter to be padded at the lsb end in order to
24087 fill a complete number of words, require the parameter size
24088 to be ordered wrt UNITS_PER_WORD.
24090 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24091 Alan Hayward <alan.hayward@arm.com>
24092 David Sherwood <david.sherwood@arm.com>
24094 * reload1.c (spill_stack_slot_width): Change element type
24095 from unsigned int to poly_uint64_pod.
24096 (alter_reg): Treat mode sizes as polynomial.
24098 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24099 Alan Hayward <alan.hayward@arm.com>
24100 David Sherwood <david.sherwood@arm.com>
24102 * reload.c (complex_word_subreg_p): New function.
24103 (reload_inner_reg_of_subreg, push_reload): Use it.
24105 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24106 Alan Hayward <alan.hayward@arm.com>
24107 David Sherwood <david.sherwood@arm.com>
24109 * lra-constraints.c (process_alt_operands): Reject matched
24110 operands whose sizes aren't ordered.
24111 (match_reload): Refer to this check here.
24113 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24114 Alan Hayward <alan.hayward@arm.com>
24115 David Sherwood <david.sherwood@arm.com>
24117 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
24118 that the mode size is in the set {1, 2, 4, 8, 16}.
24120 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24121 Alan Hayward <alan.hayward@arm.com>
24122 David Sherwood <david.sherwood@arm.com>
24124 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
24125 Use plus_constant instead of gen_rtx_PLUS.
24127 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24128 Alan Hayward <alan.hayward@arm.com>
24129 David Sherwood <david.sherwood@arm.com>
24131 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
24132 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
24133 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
24134 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
24135 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
24136 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
24137 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
24138 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
24139 * config/i386/i386.c (ix86_push_rounding): ...this new function.
24140 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
24142 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
24143 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
24144 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
24145 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
24146 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
24147 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
24148 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
24149 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
24150 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
24151 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
24153 * expr.c (emit_move_resolve_push): Treat the input and result
24154 of PUSH_ROUNDING as a poly_int64.
24155 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
24156 (emit_push_insn): Likewise.
24157 * lra-eliminations.c (mark_not_eliminable): Likewise.
24158 * recog.c (push_operand): Likewise.
24159 * reload1.c (elimination_effects): Likewise.
24160 * rtlanal.c (nonzero_bits1): Likewise.
24161 * calls.c (store_one_arg): Likewise. Require the padding to be
24162 known at compile time.
24164 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24165 Alan Hayward <alan.hayward@arm.com>
24166 David Sherwood <david.sherwood@arm.com>
24168 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
24169 Use plus_constant instead of gen_rtx_PLUS.
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 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
24176 rather than an int.
24178 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24179 Alan Hayward <alan.hayward@arm.com>
24180 David Sherwood <david.sherwood@arm.com>
24182 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
24183 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
24184 via stack temporaries. Treat the mode size as polynomial too.
24186 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24187 Alan Hayward <alan.hayward@arm.com>
24188 David Sherwood <david.sherwood@arm.com>
24190 * expr.c (expand_expr_real_2): When handling conversions involving
24191 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
24192 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
24193 as a poly_uint64 too.
24195 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24196 Alan Hayward <alan.hayward@arm.com>
24197 David Sherwood <david.sherwood@arm.com>
24199 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
24201 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24202 Alan Hayward <alan.hayward@arm.com>
24203 David Sherwood <david.sherwood@arm.com>
24205 * combine.c (can_change_dest_mode): Handle polynomial
24206 REGMODE_NATURAL_SIZE.
24207 * expmed.c (store_bit_field_1): Likewise.
24208 * expr.c (store_constructor): Likewise.
24209 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
24210 and polynomial REGMODE_NATURAL_SIZE.
24211 (gen_lowpart_common): Likewise.
24212 * reginfo.c (record_subregs_of_mode): Likewise.
24213 * rtlanal.c (read_modify_subreg_p): Likewise.
24215 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24216 Alan Hayward <alan.hayward@arm.com>
24217 David Sherwood <david.sherwood@arm.com>
24219 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
24220 numbers of elements.
24222 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24223 Alan Hayward <alan.hayward@arm.com>
24224 David Sherwood <david.sherwood@arm.com>
24226 * match.pd: Cope with polynomial numbers of vector elements.
24228 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24229 Alan Hayward <alan.hayward@arm.com>
24230 David Sherwood <david.sherwood@arm.com>
24232 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
24233 in a POINTER_PLUS_EXPR.
24235 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24236 Alan Hayward <alan.hayward@arm.com>
24237 David Sherwood <david.sherwood@arm.com>
24239 * omp-simd-clone.c (simd_clone_subparts): New function.
24240 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
24241 (ipa_simd_modify_function_body): Likewise.
24243 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24244 Alan Hayward <alan.hayward@arm.com>
24245 David Sherwood <david.sherwood@arm.com>
24247 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
24248 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
24249 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
24250 (expand_vector_condition, vector_element): Likewise.
24251 (subparts_gt): New function.
24252 (get_compute_type): Use subparts_gt.
24253 (count_type_subparts): Delete.
24254 (expand_vector_operations_1): Use subparts_gt instead of
24255 count_type_subparts.
24257 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24258 Alan Hayward <alan.hayward@arm.com>
24259 David Sherwood <david.sherwood@arm.com>
24261 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
24262 (vect_compile_time_alias): ...this new function. Do the calculation
24263 on poly_ints rather than trees.
24264 (vect_prune_runtime_alias_test_list): Update call accordingly.
24266 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24267 Alan Hayward <alan.hayward@arm.com>
24268 David Sherwood <david.sherwood@arm.com>
24270 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
24272 (vect_schedule_slp_instance): Likewise.
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-slp.c (vect_get_and_check_slp_defs): Reject
24279 constant and extern definitions for variable-length vectors.
24280 (vect_get_constant_vectors): Note that the number of units
24281 is known to be constant.
24283 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24284 Alan Hayward <alan.hayward@arm.com>
24285 David Sherwood <david.sherwood@arm.com>
24287 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
24288 of units as polynomial. Choose between WIDE and NARROW based
24291 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24292 Alan Hayward <alan.hayward@arm.com>
24293 David Sherwood <david.sherwood@arm.com>
24295 * tree-vect-stmts.c (simd_clone_subparts): New function.
24296 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
24298 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24299 Alan Hayward <alan.hayward@arm.com>
24300 David Sherwood <david.sherwood@arm.com>
24302 * tree-vect-stmts.c (vectorizable_call): Treat the number of
24303 vectors as polynomial. Use build_index_vector for
24304 IFN_GOMP_SIMD_LANE.
24306 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24307 Alan Hayward <alan.hayward@arm.com>
24308 David Sherwood <david.sherwood@arm.com>
24310 * tree-vect-stmts.c (get_load_store_type): Treat the number of
24311 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
24312 for variable-length vectors.
24313 (vectorizable_mask_load_store): Treat the number of units as
24314 polynomial, asserting that it is constant if the condition has
24315 already been enforced.
24316 (vectorizable_store, vectorizable_load): Likewise.
24318 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24319 Alan Hayward <alan.hayward@arm.com>
24320 David Sherwood <david.sherwood@arm.com>
24322 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
24323 of units as polynomial. Punt if we can't tell at compile time
24324 which vector contains the final result.
24326 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24327 Alan Hayward <alan.hayward@arm.com>
24328 David Sherwood <david.sherwood@arm.com>
24330 * tree-vect-loop.c (vectorizable_induction): Treat the number
24331 of units as polynomial. Punt on SLP inductions. Use an integer
24332 VEC_SERIES_EXPR for variable-length integer reductions. Use a
24333 cast of such a series for variable-length floating-point
24336 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24337 Alan Hayward <alan.hayward@arm.com>
24338 David Sherwood <david.sherwood@arm.com>
24340 * tree.h (build_index_vector): Declare.
24341 * tree.c (build_index_vector): New function.
24342 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
24343 of units as polynomial, forcibly converting it to a constant if
24344 vectorizable_reduction has already enforced the condition.
24345 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
24346 to create a {1,2,3,...} vector.
24347 (vectorizable_reduction): Treat the number of units as polynomial.
24348 Choose vectype_in based on the largest scalar element size rather
24349 than the smallest number of units. Enforce the restrictions
24352 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24353 Alan Hayward <alan.hayward@arm.com>
24354 David Sherwood <david.sherwood@arm.com>
24356 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
24357 number of units as polynomial.
24359 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24360 Alan Hayward <alan.hayward@arm.com>
24361 David Sherwood <david.sherwood@arm.com>
24363 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
24364 * target.def (autovectorize_vector_sizes): Return the vector sizes
24365 by pointer, using vector_sizes rather than a bitmask.
24366 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
24367 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
24368 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
24370 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
24371 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
24372 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
24373 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
24374 * omp-general.c (omp_max_vf): Likewise.
24375 * omp-low.c (omp_clause_aligned_alignment): Likewise.
24376 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
24377 * tree-vect-loop.c (vect_analyze_loop): Likewise.
24378 * tree-vect-slp.c (vect_slp_bb): Likewise.
24379 * doc/tm.texi: Regenerate.
24380 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
24382 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
24383 the vector size as a poly_uint64 rather than an unsigned int.
24384 (current_vector_size): Change from an unsigned int to a poly_uint64.
24385 (get_vectype_for_scalar_type): Update accordingly.
24386 * tree.h (build_truth_vector_type): Take the size and number of
24387 units as a poly_uint64 rather than an unsigned int.
24388 (build_vector_type): Add a temporary overload that takes
24389 the number of units as a poly_uint64 rather than an unsigned int.
24390 * tree.c (make_vector_type): Likewise.
24391 (build_truth_vector_type): Take the number of units as a poly_uint64
24392 rather than an unsigned int.
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 * target.def (get_mask_mode): Take the number of units and length
24399 as poly_uint64s rather than unsigned ints.
24400 * targhooks.h (default_get_mask_mode): Update accordingly.
24401 * targhooks.c (default_get_mask_mode): Likewise.
24402 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
24403 * doc/tm.texi: Regenerate.
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 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
24410 * omp-general.c (omp_max_vf): Likewise.
24411 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
24412 (expand_omp_simd): Handle polynomial safelen.
24413 * omp-low.c (omplow_simd_context): Add a default constructor.
24414 (omplow_simd_context::max_vf): Change from int to poly_uint64.
24415 (lower_rec_simd_input_clauses): Update accordingly.
24416 (lower_rec_input_clauses): Likewise.
24418 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24419 Alan Hayward <alan.hayward@arm.com>
24420 David Sherwood <david.sherwood@arm.com>
24422 * tree-vectorizer.h (vect_nunits_for_cost): New function.
24423 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
24424 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
24425 (vect_analyze_slp_cost): Likewise.
24426 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
24427 (vect_model_load_cost): Likewise.
24429 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24430 Alan Hayward <alan.hayward@arm.com>
24431 David Sherwood <david.sherwood@arm.com>
24433 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
24434 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
24435 from an unsigned int * to a poly_uint64_pod *.
24436 (calculate_unrolling_factor): New function.
24437 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
24439 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24440 Alan Hayward <alan.hayward@arm.com>
24441 David Sherwood <david.sherwood@arm.com>
24443 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
24444 from an unsigned int to a poly_uint64.
24445 (_loop_vec_info::slp_unrolling_factor): Likewise.
24446 (_loop_vec_info::vectorization_factor): Change from an int
24448 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
24449 (vect_get_num_vectors): New function.
24450 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
24451 (vect_get_num_copies): Use vect_get_num_vectors.
24452 (vect_analyze_data_ref_dependences): Change max_vf from an int *
24453 to an unsigned int *.
24454 (vect_analyze_data_refs): Change min_vf from an int * to a
24456 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
24457 than an unsigned HOST_WIDE_INT.
24458 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
24459 (vect_analyze_data_ref_dependence): Change max_vf from an int *
24460 to an unsigned int *.
24461 (vect_analyze_data_ref_dependences): Likewise.
24462 (vect_compute_data_ref_alignment): Handle polynomial vf.
24463 (vect_enhance_data_refs_alignment): Likewise.
24464 (vect_prune_runtime_alias_test_list): Likewise.
24465 (vect_shift_permute_load_chain): Likewise.
24466 (vect_supportable_dr_alignment): Likewise.
24467 (dependence_distance_ge_vf): Take the vectorization factor as a
24468 poly_uint64 rather than an unsigned HOST_WIDE_INT.
24469 (vect_analyze_data_refs): Change min_vf from an int * to a
24471 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
24472 vfm1 as a poly_uint64 rather than an int. Make the same change
24473 for the returned bound_scalar.
24474 (vect_gen_vector_loop_niters): Handle polynomial vf.
24475 (vect_do_peeling): Likewise. Update call to
24476 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
24477 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
24479 * tree-vect-loop.c (vect_determine_vectorization_factor)
24480 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
24481 (vect_get_known_peeling_cost): Likewise.
24482 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
24483 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
24484 (vect_transform_loop): Likewise. Use the lowest possible VF when
24485 updating the upper bounds of the loop.
24486 (vect_min_worthwhile_factor): Make static. Return an unsigned int
24487 rather than an int.
24488 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
24489 polynomial unroll factors.
24490 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
24491 (vect_make_slp_decision): Likewise.
24492 (vect_supported_load_permutation_p): Likewise, and polynomial
24494 (vect_analyze_slp_cost): Handle polynomial vf.
24495 (vect_slp_analyze_node_operations): Likewise.
24496 (vect_slp_analyze_bb_1): Likewise.
24497 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
24498 than an unsigned HOST_WIDE_INT.
24499 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
24500 (vectorizable_load): Handle polynomial vf.
24501 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
24503 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
24505 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24506 Alan Hayward <alan.hayward@arm.com>
24507 David Sherwood <david.sherwood@arm.com>
24509 * match.pd: Handle bit operations involving three constants
24510 and try to fold one pair.
24512 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24514 * tree-vect-loop-manip.c: Include gimple-fold.h.
24515 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
24516 niters_maybe_zero parameters. Handle other cases besides a step of 1.
24517 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
24518 Add a path that uses a step of VF instead of 1, but disable it
24520 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
24521 and niters_no_overflow parameters. Update calls to
24522 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
24523 Create a new SSA name if the latter choses to use a ste other
24524 than zero, and return it via niters_vector_mult_vf_var.
24525 * tree-vect-loop.c (vect_transform_loop): Update calls to
24526 vect_do_peeling, vect_gen_vector_loop_niters and
24527 slpeel_make_loop_iterate_ntimes.
24528 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
24529 (vect_gen_vector_loop_niters): Update declarations after above changes.
24531 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
24533 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
24534 128-bit round to integer instructions.
24535 (ceil<mode>2): Likewise.
24536 (btrunc<mode>2): Likewise.
24537 (round<mode>2): Likewise.
24539 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
24541 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
24542 unaligned VSX load/store on P8/P9.
24543 (expand_block_clear): Allow the use of unaligned VSX
24544 load/store on P8/P9.
24546 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
24548 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
24550 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
24551 swap associated with both a load and a store.
24553 2018-01-02 Andrew Waterman <andrew@sifive.com>
24555 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
24556 * config/riscv/riscv.md (clear_cache): Use it.
24558 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
24560 * web.c: Remove out-of-date comment.
24562 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24564 * expr.c (fixup_args_size_notes): Check that any existing
24565 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
24566 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
24567 (emit_single_push_insn): ...here.
24569 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24571 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
24572 (const_vector_encoded_nelts): New function.
24573 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
24574 (const_vector_int_elt, const_vector_elt): Declare.
24575 * emit-rtl.c (const_vector_int_elt_1): New function.
24576 (const_vector_elt): Likewise.
24577 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
24578 of CONST_VECTOR_ELT.
24580 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24582 * expr.c: Include rtx-vector-builder.h.
24583 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
24584 directly on the tree encoding.
24585 (const_vector_from_tree): Likewise.
24586 * optabs.c: Include rtx-vector-builder.h.
24587 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
24588 sequence of "u" values.
24589 * vec-perm-indices.c: Include rtx-vector-builder.h.
24590 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
24591 directly on the vec_perm_indices encoding.
24593 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24595 * doc/rtl.texi (const_vector): Describe new encoding scheme.
24596 * Makefile.in (OBJS): Add rtx-vector-builder.o.
24597 * rtx-vector-builder.h: New file.
24598 * rtx-vector-builder.c: Likewise.
24599 * rtl.h (rtx_def::u2): Add a const_vector field.
24600 (CONST_VECTOR_NPATTERNS): New macro.
24601 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
24602 (CONST_VECTOR_DUPLICATE_P): Likewise.
24603 (CONST_VECTOR_STEPPED_P): Likewise.
24604 (CONST_VECTOR_ENCODED_ELT): Likewise.
24605 (const_vec_duplicate_p): Check for a duplicated vector encoding.
24606 (unwrap_const_vec_duplicate): Likewise.
24607 (const_vec_series_p): Check for a non-duplicated vector encoding.
24608 Say that the function only returns true for integer vectors.
24609 * emit-rtl.c: Include rtx-vector-builder.h.
24610 (gen_const_vec_duplicate_1): Delete.
24611 (gen_const_vector): Call gen_const_vec_duplicate instead of
24612 gen_const_vec_duplicate_1.
24613 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
24614 (gen_const_vec_duplicate): Use rtx_vector_builder.
24615 (gen_const_vec_series): Likewise.
24616 (gen_rtx_CONST_VECTOR): Likewise.
24617 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
24618 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
24619 Build a new vector rather than modifying a CONST_VECTOR in-place.
24620 (handle_special_swappables): Update call accordingly.
24621 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
24622 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
24623 Build a new vector rather than modifying a CONST_VECTOR in-place.
24624 (handle_special_swappables): Update call accordingly.
24626 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24628 * simplify-rtx.c (simplify_const_binary_operation): Use
24629 CONST_VECTOR_ELT instead of XVECEXP.
24631 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24633 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
24634 the selector elements to be different from the data elements
24635 if the selector is a VECTOR_CST.
24636 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
24637 ssizetype for the selector.
24639 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24641 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
24642 before testing each element individually.
24643 * tree-vect-generic.c (lower_vec_perm): Likewise.
24645 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24647 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
24648 * selftest-run-tests.c (selftest::run_tests): Call it.
24649 * vector-builder.h (vector_builder::operator ==): New function.
24650 (vector_builder::operator !=): Likewise.
24651 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
24652 (vec_perm_indices::all_from_input_p): New function.
24653 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
24654 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
24655 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
24656 instead of reading the VECTOR_CST directly. Detect whether both
24657 vector inputs are the same before constructing the vec_perm_indices,
24658 and update the number of inputs argument accordingly. Use the
24659 utility functions added above. Only construct sel2 if we need to.
24661 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24663 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
24664 the broadcast of the low byte.
24665 (expand_mult_highpart): Use an explicit encoding for the permutes.
24666 * optabs-query.c (can_mult_highpart_p): Likewise.
24667 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
24668 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
24669 (vectorizable_bswap): Likewise.
24670 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
24671 explicit encoding for the power-of-2 permutes.
24672 (vect_permute_store_chain): Likewise.
24673 (vect_grouped_load_supported): Likewise.
24674 (vect_permute_load_chain): Likewise.
24676 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24678 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
24679 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
24680 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
24681 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
24682 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
24683 (vect_gen_perm_mask_any): Likewise.
24685 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24687 * int-vector-builder.h: New file.
24688 * vec-perm-indices.h: Include int-vector-builder.h.
24689 (vec_perm_indices): Redefine as an int_vector_builder.
24690 (auto_vec_perm_indices): Delete.
24691 (vec_perm_builder): Redefine as a stand-alone class.
24692 (vec_perm_indices::vec_perm_indices): New function.
24693 (vec_perm_indices::clamp): Likewise.
24694 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
24695 (vec_perm_indices::new_vector): New function.
24696 (vec_perm_indices::new_expanded_vector): Update for new
24697 vec_perm_indices class.
24698 (vec_perm_indices::rotate_inputs): New function.
24699 (vec_perm_indices::all_in_range_p): Operate directly on the
24700 encoded form, without computing elided elements.
24701 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
24702 encoding. Update for new vec_perm_indices class.
24703 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
24704 the given vec_perm_builder.
24705 (expand_vec_perm_var): Update vec_perm_builder constructor.
24706 (expand_mult_highpart): Use vec_perm_builder instead of
24707 auto_vec_perm_indices.
24708 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
24709 vec_perm_indices instead of auto_vec_perm_indices. Use a single
24710 or double series encoding as appropriate.
24711 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
24712 vec_perm_indices instead of auto_vec_perm_indices.
24713 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
24714 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
24715 (vect_permute_store_chain): Likewise.
24716 (vect_grouped_load_supported): Likewise.
24717 (vect_permute_load_chain): Likewise.
24718 (vect_shift_permute_load_chain): Likewise.
24719 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
24720 (vect_transform_slp_perm_load): Likewise.
24721 (vect_schedule_slp_instance): Likewise.
24722 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
24723 (vectorizable_mask_load_store): Likewise.
24724 (vectorizable_bswap): Likewise.
24725 (vectorizable_store): Likewise.
24726 (vectorizable_load): Likewise.
24727 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
24728 vec_perm_indices instead of auto_vec_perm_indices. Use
24729 tree_to_vec_perm_builder to read the vector from a tree.
24730 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
24731 vec_perm_builder instead of a vec_perm_indices.
24732 (have_whole_vector_shift): Use vec_perm_builder and
24733 vec_perm_indices instead of auto_vec_perm_indices. Leave the
24734 truncation to calc_vec_perm_mask_for_shift.
24735 (vect_create_epilog_for_reduction): Likewise.
24736 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
24737 from auto_vec_perm_indices to vec_perm_indices.
24738 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
24739 instead of changing individual elements.
24740 (aarch64_vectorize_vec_perm_const): Use new_vector to install
24741 the vector in d.perm.
24742 * config/arm/arm.c (expand_vec_perm_d::perm): Change
24743 from auto_vec_perm_indices to vec_perm_indices.
24744 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
24745 instead of changing individual elements.
24746 (arm_vectorize_vec_perm_const): Use new_vector to install
24747 the vector in d.perm.
24748 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
24749 Update vec_perm_builder constructor.
24750 (rs6000_expand_interleave): Likewise.
24751 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
24752 (rs6000_expand_interleave): Likewise.
24754 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24756 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
24757 to qimode could truncate the indices.
24758 * optabs.c (expand_vec_perm_var): Likewise.
24760 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24762 * Makefile.in (OBJS): Add vec-perm-indices.o.
24763 * vec-perm-indices.h: New file.
24764 * vec-perm-indices.c: Likewise.
24765 * target.h (vec_perm_indices): Replace with a forward class
24767 (auto_vec_perm_indices): Move to vec-perm-indices.h.
24768 * optabs.h: Include vec-perm-indices.h.
24769 (expand_vec_perm): Delete.
24770 (selector_fits_mode_p, expand_vec_perm_var): Declare.
24771 (expand_vec_perm_const): Declare.
24772 * target.def (vec_perm_const_ok): Replace with...
24773 (vec_perm_const): ...this new hook.
24774 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
24775 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
24776 * doc/tm.texi: Regenerate.
24777 * optabs.def (vec_perm_const): Delete.
24778 * doc/md.texi (vec_perm_const): Likewise.
24779 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
24780 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
24781 expand_vec_perm for constant permutation vectors. Assert that
24782 the mode of variable permutation vectors is the integer equivalent
24783 of the mode that is being permuted.
24784 * optabs-query.h (selector_fits_mode_p): Declare.
24785 * optabs-query.c: Include vec-perm-indices.h.
24786 (selector_fits_mode_p): New function.
24787 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
24788 is defined, instead of checking whether the vec_perm_const_optab
24789 exists. Use targetm.vectorize.vec_perm_const instead of
24790 targetm.vectorize.vec_perm_const_ok. Check whether the indices
24791 fit in the vector mode before using a variable permute.
24792 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
24793 vec_perm_indices instead of an rtx.
24794 (expand_vec_perm): Replace with...
24795 (expand_vec_perm_const): ...this new function. Take the selector
24796 as a vec_perm_indices rather than an rtx. Also take the mode of
24797 the selector. Update call to shift_amt_for_vec_perm_mask.
24798 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
24799 Use vec_perm_indices::new_expanded_vector to expand the original
24800 selector into bytes. Check whether the indices fit in the vector
24801 mode before using a variable permute.
24802 (expand_vec_perm_var): Make global.
24803 (expand_mult_highpart): Use expand_vec_perm_const.
24804 * fold-const.c: Includes vec-perm-indices.h.
24805 * tree-ssa-forwprop.c: Likewise.
24806 * tree-vect-data-refs.c: Likewise.
24807 * tree-vect-generic.c: Likewise.
24808 * tree-vect-loop.c: Likewise.
24809 * tree-vect-slp.c: Likewise.
24810 * tree-vect-stmts.c: Likewise.
24811 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
24813 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
24814 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
24815 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
24816 (aarch64_vectorize_vec_perm_const): ...this new function.
24817 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
24818 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
24819 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
24820 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
24821 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
24822 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
24823 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
24825 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
24826 check for NEON modes.
24827 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
24828 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
24829 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
24830 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
24832 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
24833 the old VEC_PERM_CONST conditions.
24834 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
24835 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
24836 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
24837 (ia64_vectorize_vec_perm_const_ok): Merge into...
24838 (ia64_vectorize_vec_perm_const): ...this new function.
24839 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
24840 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
24841 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
24842 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
24843 * config/mips/mips.c (mips_expand_vec_perm_const)
24844 (mips_vectorize_vec_perm_const_ok): Merge into...
24845 (mips_vectorize_vec_perm_const): ...this new function.
24846 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
24847 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
24848 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
24849 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
24850 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
24851 (rs6000_expand_vec_perm_const): Delete.
24852 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
24854 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
24855 (altivec_expand_vec_perm_const_le): Take each operand individually.
24856 Operate on constant selectors rather than rtxes.
24857 (altivec_expand_vec_perm_const): Likewise. Update call to
24858 altivec_expand_vec_perm_const_le.
24859 (rs6000_expand_vec_perm_const): Delete.
24860 (rs6000_vectorize_vec_perm_const_ok): Delete.
24861 (rs6000_vectorize_vec_perm_const): New function.
24862 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
24863 an element count and rtx array.
24864 (rs6000_expand_extract_even): Update call accordingly.
24865 (rs6000_expand_interleave): Likewise.
24866 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
24867 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
24868 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
24869 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
24870 (rs6000_expand_vec_perm_const): Delete.
24871 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
24872 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
24873 (altivec_expand_vec_perm_const_le): Take each operand individually.
24874 Operate on constant selectors rather than rtxes.
24875 (altivec_expand_vec_perm_const): Likewise. Update call to
24876 altivec_expand_vec_perm_const_le.
24877 (rs6000_expand_vec_perm_const): Delete.
24878 (rs6000_vectorize_vec_perm_const_ok): Delete.
24879 (rs6000_vectorize_vec_perm_const): New function. Remove stray
24880 reference to the SPE evmerge intructions.
24881 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
24882 an element count and rtx array.
24883 (rs6000_expand_extract_even): Update call accordingly.
24884 (rs6000_expand_interleave): Likewise.
24885 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
24886 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
24888 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
24890 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24892 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
24893 vector mode and that that mode matches the mode of the data
24895 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
24896 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
24897 directly using expand_vec_perm_1 when forcing selectors into
24899 (expand_vec_perm_var): New function, split out from expand_vec_perm.
24901 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24903 * optabs-query.h (can_vec_perm_p): Delete.
24904 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
24905 * optabs-query.c (can_vec_perm_p): Split into...
24906 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
24907 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
24908 particular selector is valid.
24909 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
24910 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
24911 (vect_grouped_load_supported): Likewise.
24912 (vect_shift_permute_load_chain): Likewise.
24913 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
24914 (vect_transform_slp_perm_load): Likewise.
24915 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
24916 (vectorizable_bswap): Likewise.
24917 (vect_gen_perm_mask_checked): Likewise.
24918 * fold-const.c (fold_ternary_loc): Likewise. Don't take
24919 implementations of variable permutation vectors into account
24920 when deciding which selector to use.
24921 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
24922 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
24923 with a false third argument.
24924 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
24925 to test whether the constant selector is valid and can_vec_perm_var_p
24926 to test whether a variable selector is valid.
24928 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24930 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
24931 * optabs-query.c (can_vec_perm_p): Likewise.
24932 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
24933 instead of vec_perm_indices.
24934 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
24935 (vect_gen_perm_mask_checked): Likewise,
24936 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
24937 (vect_gen_perm_mask_checked): Likewise,
24939 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24941 * optabs-query.h (qimode_for_vec_perm): Declare.
24942 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
24943 (qimode_for_vec_perm): ...this new function.
24944 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
24946 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
24948 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
24949 does not have a conditional at the top.
24951 2018-01-02 Richard Biener <rguenther@suse.de>
24953 * ipa-inline.c (big_speedup_p): Fix expression.
24955 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
24958 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
24961 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
24965 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
24966 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
24967 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
24968 cond_taken_branch_cost 3->4.
24970 2018-01-01 Jakub Jelinek <jakub@redhat.com>
24972 PR tree-optimization/83581
24973 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
24974 TODO_cleanup_cfg if any changes have been made.
24976 PR middle-end/83608
24977 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
24978 convert_modes if target mode has the right side, but different mode
24981 PR middle-end/83609
24982 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
24983 last argument when extracting from CONCAT. If either from_real or
24984 from_imag is NULL, use expansion through memory. If result is not
24985 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
24986 the parts directly to inner mode, if even that fails, use expansion
24989 PR middle-end/83623
24990 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
24991 check for bswap in mode rather than HImode and use that in expand_unop
24994 Copyright (C) 2018 Free Software Foundation, Inc.
24996 Copying and distribution of this file, with or without modification,
24997 are permitted in any medium without royalty provided the copyright
24998 notice and this notice are preserved.