[ARM] PR target/69135: Mark ARMv8 vcvt instructions as unconditional
[platform/upstream/gcc.git] / gcc / ChangeLog
1 2016-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2
3         PR target/69135
4         * config/arm/vfp.md (l<vrint_pattern><su_optab><mode>si2): Set "conds"
5         attribute to unconditional.  Remove %? from output template.
6
7 2015-01-19  Wilco Dijkstra  <wdijkstr@arm.com>
8             Jiong Wang  <jiong.wang@arm.com>
9
10         * ccmp.c (expand_ccmp_expr_1): Cost the instruction sequences
11         generated from different expand order.
12
13 2015-01-19  Wilco Dijkstra  <wdijkstr@arm.com>
14
15         * /config/aarch64/aarch64.c (aarch64_if_then_else_costs):
16         Add support for CCMP costing.
17
18 2015-01-19  Wilco Dijkstra  <wdijkstr@arm.com>
19
20         * ccmp.c (ccmp_candidate_p): Remove integer-only restriction.
21         * config/aarch64/aarch64.md (fccmp<mode>): New pattern.
22         (fccmpe<mode>): Likewise.
23         (fcmp): Rename to fcmp and globalize pattern.
24         (fcmpe): Likewise.
25         * config/aarch64/aarch64.c (aarch64_gen_ccmp_first): Add FP support.
26         (aarch64_gen_ccmp_next): Add FP support.
27
28 2015-01-19  Wilco Dijkstra  <wdijkstr@arm.com>
29
30         * target.def (gen_ccmp_first): Update documentation.
31         (gen_ccmp_next): Likewise.
32         * doc/tm.texi (gen_ccmp_first): Update documentation.
33         (gen_ccmp_next): Likewise.
34         * ccmp.c (expand_ccmp_expr): Extract cmp_code from return value of
35         expand_ccmp_expr_1.  Improve comments.
36         * config/aarch64/aarch64.md (ccmp_and): Use if_then_else for ccmp.
37         (ccmp_ior<mode>): Remove pattern.
38         (cmp<mode>): Remove expand.
39         (cmp): Globalize pattern.
40         (cstorecc4): Use cc_register.
41         (mov<mode>cc): Remove ccmp_cc_register check.
42         * config/aarch64/aarch64.c (aarch64_get_condition_code_1):
43         Simplify after removal of CC_DNE/* modes.
44         (aarch64_ccmp_mode_to_code): Remove.
45         (aarch64_print_operand): Remove 'K' case.  Merge 'm' and 'M' cases.
46         In 'k' case use integer as condition.
47         (aarch64_nzcv_codes): Remove inverted cases.
48         (aarch64_code_to_ccmode): Remove.
49         (aarch64_gen_ccmp_first): Use cmp pattern directly.  Return the correct
50         comparison with CC register to be used in folowing CCMP/branch/CSEL.
51         (aarch64_gen_ccmp_next): Use previous comparison and mode in CCMP
52         pattern.  Return the comparison with CC register.  Invert conditions
53         when bitcode is OR.
54         * config/aarch64/aarch64-modes.def: Remove CC_DNE/* modes.
55         * config/aarch64/predicates.md (ccmp_cc_register): Remove.
56
57 2016-01-19  Jan Hubicka  <hubicka@ucw.cz>
58
59         * cgraphunit.c (cgraph_node::reset): Clear thunk info and
60         instrumented_version.
61
62 2016-01-19  Richard Biener  <rguenther@suse.de>
63
64         PR tree-optimization/69336
65         * tree-ssa-scopedtables.c (avail_expr_hash): Handle all
66         handled components with get_ref_base_and_extent.
67         (equal_mem_array_ref_p): Adjust.
68
69 2016-01-19  Jakub Jelinek  <jakub@redhat.com>
70
71         PR debug/65779
72         * shrink-wrap.c: Include valtrack.h.
73         (move_insn_for_shrink_wrap): Add DEBUG argument.  If
74         MAY_HAVE_DEBUG_INSNS, call dead_debug_add on DEBUG_INSNs
75         in between insn and where it will be moved to.  Call
76         dead_debug_insert_temp.
77         (prepare_shrink_wrap): Adjust caller.  Call dead_debug_local_init
78         first and dead_debug_local_finish at the end.
79         For uses and defs bitmap, handle all regs in between REGNO and
80         END_REGNO, not just the first one.
81
82 2016-01-19  Richard Biener  <rguenther@suse.de>
83
84         PR tree-optimization/69352
85         * tree-ssa-scopedtables.c (avail_expr_hash): Check for size == -1.
86         (equal_mem_array_ref_p): Constrain size and max size properly.
87         Compare the reverse flag.
88
89 2016-01-19  Bernd Schmidt  <bschmidt@redhat.com>
90
91         * ira.c (ira): Update regstat data if we deleted insns.
92
93 2016-01-19  Jakub Jelinek  <jakub@redhat.com>
94
95         PR rtl-optimization/68955
96         PR rtl-optimization/64557
97         * dse.c (record_store, check_mem_read_rtx): Don't call get_addr
98         here.  Fix up formatting.
99         * alias.c (get_addr): Handle VALUE +/- CONST_SCALAR_INT_P.
100
101 2016-01-19  Jan Hubicka  <hubicka@ucw.cz>
102
103         PR lto/69133
104         * cgraphunit.c (cgraph_node::expand_thunk): When forcing gimple
105         assume that the node has body.
106         * cgraph.c (cgraph_node::get_untransformed_body): Use gimple_body_p
107         check.
108
109 2016-01-19  Jan Hubicka  <hubicka@ucw.cz>
110
111         * lto-streamer-out.c (lto_output): Do not stream instrumentation
112         thunks.
113
114 2016-01-19  Jan Hubicka  <hubicka@ucw.cz>
115
116         * symtab.c (symtab_node::asm_name): Do not call printable name directly.
117         (symtab_node::name): Report name as unnamed if DECL_NAME is not set.
118
119 2016-01-19  Martin Jambor  <mjambor@suse.cz>
120             Martin Liska  <mliska@suse.cz>
121             Michael Matz <matz@suse.de>
122
123         * Makefile.in (OBJS): Add new source files.
124         (GTFILES): Add hsa.c.
125         * common.opt (disable_hsa): New variable.
126         (-Whsa): New warning.
127         * config.in (ENABLE_HSA): New.
128         * configure.ac: Treat hsa differently from other accelerators.
129         (OFFLOAD_TARGETS): Define ENABLE_OFFLOADING according to
130         $enable_offloading.
131         (ENABLE_HSA): Define ENABLE_HSA according to $enable_hsa.
132         * doc/install.texi (Configuration): Document --with-hsa-runtime,
133         --with-hsa-runtime-include, --with-hsa-runtime-lib and
134         --with-hsa-kmt-lib.
135         * doc/invoke.texi (-Whsa): Document.
136         (hsa-gen-debug-stores): Likewise.
137         * lto-wrapper.c (compile_images_for_offload_targets): Do not attempt
138         to invoke offload compiler for hsa acclerator.
139         * opts.c (common_handle_option): Determine whether HSA offloading
140         should be performed.
141         * params.def (PARAM_HSA_GEN_DEBUG_STORES): New parameter.
142         * builtin-types.def (BT_FN_VOID_UINT_PTR_INT_PTR): New.
143         (BT_FN_VOID_INT_OMPFN_SIZE_PTR_PTR_PTR_UINT_PTR_INT_INT): Removed.
144         (BT_FN_VOID_INT_OMPFN_SIZE_PTR_PTR_PTR_UINT_PTR_PTR): New.
145         * gimple-low.c (lower_stmt): Also handle GIMPLE_OMP_GRID_BODY.
146         * gimple-pretty-print.c (dump_gimple_omp_for): Also handle
147         GF_OMP_FOR_KIND_GRID_LOOP.
148         (dump_gimple_omp_block): Also handle GIMPLE_OMP_GRID_BODY.
149         (pp_gimple_stmt_1): Likewise.
150         * gimple-walk.c (walk_gimple_stmt): Likewise.
151         * gimple.c (gimple_build_omp_grid_body): New function.
152         (gimple_copy): Also handle GIMPLE_OMP_GRID_BODY.
153         * gimple.def (GIMPLE_OMP_GRID_BODY): New.
154         * gimple.h (enum gf_mask): Added GF_OMP_PARALLEL_GRID_PHONY,
155         GF_OMP_FOR_KIND_GRID_LOOP, GF_OMP_FOR_GRID_PHONY and
156         GF_OMP_TEAMS_GRID_PHONY.
157         (gimple_statement_omp_single_layout): Updated comments.
158         (gimple_build_omp_grid_body): New function.
159         (gimple_has_substatements): Also handle GIMPLE_OMP_GRID_BODY.
160         (gimple_omp_for_grid_phony): New function.
161         (gimple_omp_for_set_grid_phony): Likewise.
162         (gimple_omp_parallel_grid_phony): Likewise.
163         (gimple_omp_parallel_set_grid_phony): Likewise.
164         (gimple_omp_teams_grid_phony): Likewise.
165         (gimple_omp_teams_set_grid_phony): Likewise.
166         (gimple_return_set_retbnd): Also handle GIMPLE_OMP_GRID_BODY.
167         * omp-builtins.def (BUILT_IN_GOMP_OFFLOAD_REGISTER): New.
168         (BUILT_IN_GOMP_OFFLOAD_UNREGISTER): Likewise.
169         (BUILT_IN_GOMP_TARGET): Updated type.
170         * omp-low.c: Include symbol-summary.h, hsa.h and params.h.
171         (adjust_for_condition): New function.
172         (get_omp_for_step_from_incr): Likewise.
173         (extract_omp_for_data): Moved parts to adjust_for_condition and
174         get_omp_for_step_from_incr.
175         (build_outer_var_ref): Handle GIMPLE_OMP_GRID_BODY.
176         (fixup_child_record_type): Bail out if receiver_decl is NULL.
177         (scan_sharing_clauses): Handle OMP_CLAUSE__GRIDDIM_.
178         (scan_omp_parallel): Do not create child functions for phony
179         constructs.
180         (check_omp_nesting_restrictions): Handle GIMPLE_OMP_GRID_BODY.
181         (scan_omp_1_op): Checking assert we are not remapping to
182         ERROR_MARK.  Also also handle GIMPLE_OMP_GRID_BODY.
183         (parallel_needs_hsa_kernel_p): New function.
184         (expand_parallel_call): Register apprpriate parallel child
185         functions as HSA kernels.
186         (grid_launch_attributes_trees): New type.
187         (grid_attr_trees): New variable.
188         (grid_create_kernel_launch_attr_types): New function.
189         (grid_insert_store_range_dim): Likewise.
190         (grid_get_kernel_launch_attributes): Likewise.
191         (get_target_argument_identifier_1): Likewise.
192         (get_target_argument_identifier): Likewise.
193         (get_target_argument_value): Likewise.
194         (push_target_argument_according_to_value): Likewise.
195         (get_target_arguments): Likewise.
196         (expand_omp_target): Call get_target_arguments instead of looking
197         up for teams and thread limit.
198         (grid_expand_omp_for_loop): New function.
199         (grid_arg_decl_map): New type.
200         (grid_remap_kernel_arg_accesses): New function.
201         (grid_expand_target_kernel_body): New function.
202         (expand_omp): Call it.
203         (lower_omp_for): Do not emit phony constructs.
204         (lower_omp_taskreg): Do not emit phony constructs but create for them
205         a temporary variable receiver_decl.
206         (lower_omp_taskreg): Do not emit phony constructs.
207         (lower_omp_teams): Likewise.
208         (lower_omp_grid_body): New function.
209         (lower_omp_1): Call it.
210         (grid_reg_assignment_to_local_var_p): New function.
211         (grid_seq_only_contains_local_assignments): Likewise.
212         (grid_find_single_omp_among_assignments_1): Likewise.
213         (grid_find_single_omp_among_assignments): Likewise.
214         (grid_find_ungridifiable_statement): Likewise.
215         (grid_target_follows_gridifiable_pattern): Likewise.
216         (grid_remap_prebody_decls): Likewise.
217         (grid_copy_leading_local_assignments): Likewise.
218         (grid_process_kernel_body_copy): Likewise.
219         (grid_attempt_target_gridification): Likewise.
220         (grid_gridify_all_targets_stmt): Likewise.
221         (grid_gridify_all_targets): Likewise.
222         (execute_lower_omp): Call grid_gridify_all_targets.
223         (make_gimple_omp_edges): Handle GIMPLE_OMP_GRID_BODY.
224         * tree-core.h (omp_clause_code): Added OMP_CLAUSE__GRIDDIM_.
225         (tree_omp_clause): Added union field dimension.
226         * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE__GRIDDIM_.
227         * tree.c (omp_clause_num_ops): Added number of arguments of
228         OMP_CLAUSE__GRIDDIM_.
229         (omp_clause_code_name): Added name of OMP_CLAUSE__GRIDDIM_.
230         (walk_tree_1): Handle OMP_CLAUSE__GRIDDIM_.
231         * tree.h (OMP_CLAUSE_GRIDDIM_DIMENSION): New.
232         (OMP_CLAUSE_SET_GRIDDIM_DIMENSION): Likewise.
233         (OMP_CLAUSE_GRIDDIM_SIZE): Likewise.
234         (OMP_CLAUSE_GRIDDIM_GROUP): Likewise.
235         * passes.def: Schedule pass_ipa_hsa and pass_gen_hsail.
236         * tree-pass.h (make_pass_gen_hsail): Declare.
237         (make_pass_ipa_hsa): Likewise.
238         * ipa-hsa.c: New file.
239         * lto-section-in.c (lto_section_name): Add hsa section name.
240         * lto-streamer.h (lto_section_type): Add hsa section.
241         * timevar.def (TV_IPA_HSA): New.
242         * hsa-brig-format.h: New file.
243         * hsa-brig.c: New file.
244         * hsa-dump.c: Likewise.
245         * hsa-gen.c: Likewise.
246         * hsa.c: Likewise.
247         * hsa.h: Likewise.
248         * toplev.c (compile_file): Call hsa_output_brig.
249         * hsa-regalloc.c: New file.
250
251 2016-01-18  Jeff Law  <law@redhat.com>
252
253         PR tree-optimization/69320
254         * tree-ssa-dom.c (record_edge_info): For comparisons against a boolean
255         ranged object, do nothing if the RHS constant is not [0..1].
256         (optimize_stmt): Comparing a boolean ranged object against a
257         constant outside [0..1] results in a compile-time constant.
258
259         * tree-ssanames.c (ssa_name_has_boolean_range): Remove unnecessary
260         test.
261
262 2016-01-18  Sandra Loosemore <sandra@codesourcery.com>
263
264         * doc/invoke.texi (Invoking GCC): Add new section to menu.
265         (Option Summary): Update to reflect new section and moved options.
266         (C++ Dialect Options): Move -fstats to new section.
267         (Debugging Options): Move all dump, statistics, and other GCC
268         developer options to new section.  Rewrite section introduction
269         and re-order remaining options to put the more basic ones first.
270         (Optimization Options): Move -fira-verbose and -flto-report* to
271         new section.
272         (Developer Options): New section incorporating moved options.
273         * doc/cppopts.texi (-dM): Update cross-reference.
274
275 2016-01-18  Richard Henderson  <rth@redhat.com>
276
277         PR target/69176
278         * config/aarch64/aarch64.md (add<GPI>3): Move long immediate
279         operands to pseudo only if CSE is expected.  Split long immediate
280         operands only after reload, and for the stack pointer.
281         (*add<GPI>3_pluslong): Remove.
282         (*addsi3_aarch64, *adddi3_aarch64): Merge into...
283         (*add<GPI>3_aarch64): ... here.  Add r/rk/Upl alternative.
284         (*addsi3_aarch64_uxtw): Add r/rk/Upl alternative.
285         (*add<GPI>3 peepholes): New.
286         (*add<GPI>3 splitters): New.
287         * config/aarch64/constraints.md (Upl): New.
288         * config/aarch64/predicates.md (aarch64_pluslong_strict_immedate): New.
289
290 2016-01-18  Richard Biener  <rguenther@suse.de>
291
292         PR tree-optimization/69297
293         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Count each scalar
294         stmt at most once.
295         (vect_bb_vectorization_profitable_p): Clear visited flag again.
296
297 2016-01-18  Yuri Rumyantsev  <ysrumyan@gmail.com>
298
299         PR middle-end/68542
300         * fold-const.c (fold_binary_op_with_conditional_arg): Bail out for case
301         of mixind vector and scalar types.
302         (fold_relational_const): Add handling of vector
303         comparison with boolean result.
304         * tree-cfg.c (verify_gimple_comparison): Add argument CODE, allow
305         comparison of vector operands with boolean result for EQ/NE only.
306         (verify_gimple_assign_binary): Adjust call for verify_gimple_comparison.
307         (verify_gimple_cond): Likewise.
308         * tree-vrp.c (extract_code_and_val_from_cond_with_ops): Modify check on
309         valid type of VAL.
310
311 2016-01-18  Joseph Myers  <joseph@codesourcery.com>
312
313         * config/mips/mips.h (ISA_HAS_PAIRED_SINGLE): Require
314         !TARGET_OCTEON.
315
316 2016-01-18  Richard Biener  <rguenther@suse.de>
317
318         PR middle-end/69308
319         * gimple.c (gimple_could_trap_p_1): Handle GIMPLE_COND.
320
321 2016-01-18  Tom de Vries  <tom@codesourcery.com>
322
323         * passes.def: Add pass_parallelize_loops to pass_oacc_kernels.
324
325 2016-01-18  Tom de Vries  <tom@codesourcery.com>
326
327         * omp-low.c (set_oacc_fn_attrib): Make extern.
328         * omp-low.h (set_oacc_fn_attrib): Declare.
329         * tree-parloops.c (struct reduction_info): Add reduc_addr field.
330         (create_call_for_reduction_1): Handle case that reduc_addr is non-NULL.
331         (create_parallel_loop, gen_parallel_loop, try_create_reduction_list):
332         Add and handle function parameter oacc_kernels_p.
333         (find_reduc_addr, get_omp_data_i_param): New function.
334         (ref_conflicts_with_region, oacc_entry_exit_ok_1)
335         (oacc_entry_exit_single_gang, oacc_entry_exit_ok): New function.
336         (parallelize_loops): Add and handle function parameter oacc_kernels_p.
337         Calculate dominance info.  Skip loops that are not in a kernels region
338         in oacc_kernels_p mode.  Skip inner loops of parallelized loops.
339         (pass_parallelize_loops::execute): Call parallelize_loops with
340         oacc_kernels_p argument.
341         (pass_parallelize_loops::clone, pass_parallelize_loops::set_pass_param):
342         New member function.
343         (pass_parallelize_loops::bool oacc_kernels_p): New member var.
344         * passes.def: Add argument to pass_parallelize_loops instantation.
345
346 2016-01-18  Tom de Vries  <tom@codesourcery.com>
347
348         * tree-parloops.c (pass_parallelize_loops::execute): Allow
349         pass_parallelize_loops to be run outside the loop pipeline.
350
351 2016-01-18  Alan Lawrence  <alan.lawrence@arm.com>
352
353         * tree-scalar-evolution.c (follow_copies_to_constant): New.
354         (analyze_initial_condition, analyze_scalar_evolution_1): Call previous.
355
356 2016-01-18  Alan Lawrence  <alan.lawrence@arm.com>
357
358         PR target/63679
359         * tree-ssa-scopedtables.c (avail_expr_hash): Hash MEM_REF and ARRAY_REF
360         using get_ref_base_and_extent.
361         (equal_mem_array_ref_p): New.
362         (hashable_expr_equal_p): Add call to previous.
363
364 2016-01-18  Alan Lawrence  <alan.lawrence@arm.com>
365
366         PR target/63679
367         * tree-sra.c (disqualified_constants, constant_decl_p): New.
368         (sra_initialize): Allocate disqualified_constants.
369         (sra_deinitialize): Free disqualified_constants.
370         (disqualify_candidate): Update disqualified_constants when appropriate.
371         (create_access): Scan for constant-pool entries as we go along.
372         (scalarizable_type_p): Add check against type_contains_placeholder_p.
373         (maybe_add_sra_candidate): Allow constant-pool entries.
374         (load_assign_lhs_subreplacements): Bind debug for constant pool vars.
375         (initialize_constant_pool_replacements): New.
376         (sra_modify_assign): Avoid mangling assignments created by previous,
377         and don't generate writes into constant pool.
378         (sra_modify_function_body): Call initialize_constant_pool_replacements.
379
380 2016-01-18  Ilya Enkovich  <enkovich.gnu@gmail.com>
381
382         * config/i386/i386.c (scalar_to_vector_candidate_p): Support
383         andnot instruction.
384         (scalar_chain::convert_op): Likewise.
385         * config/i386/i386.md (*andndi3_doubleword): New.
386
387 2016-01-18  Richard Biener  <rguenther@suse.de>
388
389         PR tree-optimization/69170
390         * tree-vect-slp.c (vect_build_slp_tree): Verify we are not
391         building a vector from scalar results of a pattern stmt.
392
393 2016-01-18  Jakub Jelinek  <jakub@redhat.com>
394
395         * haifa-sched.c (autopref_multipass_init): Work around
396         -Wmaybe-uninitialized warning.
397
398 2016-01-18  Thomas Preud'homme  <thomas.preudhomme@arm.com>
399
400         * config/arm/arm.c (thumb1_reorg): Check that the comparison is
401         against the constant 0.
402
403 2016-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
404
405         PR tree-optimization/68799
406         * gimple-ssa-strength-reduction.c (create_phi_basis): Directly
407         look up phi candidates in the statement-candidate map.
408         (phi_add_costs): Likewise.
409         (record_phi_increments): Likewise.
410         (phi_incr_cost): Likewise.
411         (ncd_with_phi): Likewise.
412         (all_phi_incrs_profitable): Likewise.
413
414 2016-01-17  Jakub Jelinek  <jakub@redhat.com>
415
416         * omp-low.c (mark_loops_in_oacc_kernels_region): Work around
417         -Wmaybe-uninitialized warning.
418
419 2016-01-16  Sandra Loosemore  <sandra@codesourcery.com>
420
421         * doc/invoke.texi (Invoking GCC): Add new section to menu.
422         (Option Summary): Update to reflect new section and moved options.
423         (C++ Dialect Options): Move -fvtable-verify and related options.
424         (Debugging Options): Move Sanitizer, Pointer Bounds Checker,
425         and profiling-related options.
426         (Optimization Options): Move profile generation options and
427         -fstack-protector and related options.
428         (Instrumentation Options): New section incorporating moved options.
429         (Code Generation Options): Move -finstrument-functions and
430         related options, -fstack-check, -fstack-limit*, and -fbounds-check.
431
432 2016-01-16  Tom de Vries  <tom@codesourcery.com>
433
434         * passes.def: Move pass_expand_omp_ssa out of pass_parallelize_loops.
435
436 2016-01-16  Tom de Vries  <tom@codesourcery.com>
437
438         * omp-low.c (expand_omp_atomic_fetch_op):  Release defs of update stmt.
439
440 2016-01-16  Richard Sandiford  <richard.sandiford@arm.com>
441
442         * hash-table.h (hash_table::empty): Turn into an inline wrapper
443         that checks whether the table is already empty.  Rename the
444         original implementation to...
445         (hash_table::empty_slot): ...this new private function.
446
447 2016-01-15  David Malcolm  <dmalcolm@redhat.com>
448
449         PR diagnostic/68899
450         * diagnostic-show-locus.c (layout::print_source_line): Move x
451         offset of line until after call to
452         get_line_width_without_trailing_whitespace.
453
454 2016-01-15  Jeff Law  <law@redhat.com>
455
456         PR tree-optimization/69270
457         * tree-ssanames.c (ssa_name_has_boolean_range): Moved here from
458         tree-ssa-dom.c.  Improve test for [0..1] ranve from VRP.
459         * tree-ssa-dom.c (ssa_name_has_boolean_range): Remove.
460         * tree-ssanames.h (ssa_name_has_boolean_range): Prototype.
461         * tree-ssa-uncprop.c (associate_equivalences_with_edges): Use
462         ssa_name_has_boolean_range and constant_boolean_node.
463
464 2016-01-15  Vladimir Makarov  <vmakarov@redhat.com>
465
466         PR rtl-optimization/69030
467         * lra-spills.c (remove_pseudos): Check nrefs and make the function
468         returning bool.
469         (spill_pseudos): Delete debug insn for dead pseudo.
470         (lra_spill): Initiate spill_hard_reg and slots memory separately.
471
472 2016-01-15  Jiong Wang  <jiong.wang@arm.com>
473
474         * config/aarch64/aarch64-builtins.c (aarch64_types_unopus_qualifiers):
475         New.
476         (TYPES_UNOPUS): Likewise.
477         * config/aarch64/aarch64-simd-builtins.def (lbtruncuv2sf): Correct
478         builtin type, from UNOP to UNOPUS.
479         (lbtruncuv4sf): Likewise.
480         (lbtruncuv2df): Likewise.
481         (lrounduv2sf): Likewise.
482         (lrounduv4sf): Likewise.
483         (lrounduv2df): Likewise.
484         (lroundusf): Likewise.
485         (lroundusf): Likewise.
486         (lceiluv2sf): Likewise.
487         (lceiluv4sf): Likewise.
488         (lceiluv2df): Likewise.
489         (lceilusf): Likewise.
490         (lceiludf): Likewise.
491         (lflooruv2sf): Likewise.
492         (lflooruv4sf): Likewise.
493         (lflooruv2df): Likewise.
494         (lfloorusf): Likewise.
495         (lfloorudf): Likewise.
496         (lfrintnuv2sf): Likewise.
497         (lfrintnuv4sf): Likewise.
498         (lfrintnuv2df): Likewise.
499         (lfrintnusf): Likewise.
500         (lfrintnudf): Likewise.
501         * config/aarch64/arm_neon.h (vcvt_u32_f32): Remove unncessary type
502         conversion.
503         (vcvtq_u32_f32): Likewise.
504         (vcvtq_u64_f64): Likewise.
505         (vcvta_u32_f32): Likewise.
506         (vcvtaq_u32_f32): Likewise.
507         (vcvtaq_u64_f64): Likewise.
508         (vcvtm_u32_f32): Likewise.
509         (vcvtmq_u32_f32): Likewise.
510         (vcvtmq_u64_f64): Likewise.
511         (vcvtn_u32_f32): Likwise.
512         (vcvtnq_u32_f32): Likewise.
513         (vcvtnq_u64_f64): Likewise.
514         (vcvtp_u32_f32): Likewise.
515         (vcvtpq_u32_f32): Likewise.
516         (vcvtpq_u64_f64): Likewise.
517         (vcvtmd_u64_f64): Likewise.
518         (vcvtms_u32_f32): Likewise.
519         (vcvtad_u64_f64): Likewise.
520         (vcvtas_u32_f32): Likewise.
521         (vcvtnd_u64_f64): Likewise.
522         (vcvtns_u32_f32): Likewise.
523         (vcvtpd_u64_f64): Likewise.
524         (vcvtps_u32_f32): Likewise.
525
526 2016-01-15  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
527
528         * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Handle
529         CSEL of zero_extended registers.
530
531 2016-01-15  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
532
533         * config/aarch64/aarch64.c (aarch64_rtx_costs, COMPARE case):
534         Handle COMPARE of ZERO_EXTRACT against zero form of TST-immediate.
535
536 2016-01-15  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
537
538         * config/aarch64/aarch64.c (aarch64_process_one_target_attr): Return
539         false when argument string is not found in the attributes table
540         at all.
541
542 2016-01-15  David Edelsohn  <dje.gcc@gmail.com>
543
544         PR target/68609
545         * config/rs6000/rs6000.c (rs6000_emit_msub): Delete.
546         (rs6000_emit_swsqrt): Convert to Goldschmidt's Algorithm
547         * config/rs6000/rs6000.md (sqrt<mode>2): Limit swsqrt to high
548         precision estimate.
549
550 2016-01-15  Richard Biener  <rguenther@suse.de>
551
552         PR tree-optimization/66856
553         * tree-vect-loop.c (vect_transform_loop): Free SLP instances here.
554         * tree-vect-slp.c (vect_free_slp_tree): Decrement stmt reference count.
555         (vect_create_new_slp_node): Increment stmt reference count.
556         (vect_get_and_check_slp_defs): Make sure stmts are nor already in
557         an SLP tree before swapping operands.
558         (vect_build_slp_tree): Likewise.
559         (destroy_bb_vec_info): Free stmt info after SLP instances.
560         * tree-vect-stmts.c (new_stmt_vec_info): Initialize reference count.
561         * tree-vectorizer.h (struct _stmt_vec_info): Add num_slp_uses field.
562         (STMT_VINFO_NUM_SLP_USES): New macro.
563
564 2016-01-15  Richard Biener  <rguenther@suse.de>
565
566         PR debug/69137
567         * dwarf2out.c (add_linkage_name_raw): New function split out from ...
568         (add_linkage_name): ... here.
569         (gen_typedef_die): Use add_linkage_name_raw instead of
570         add_linkage_attr to delay DECL_ASSEMBLER_NAME computation
571         if necessary.
572
573 2016-01-15  Cesar Philippidis  <cesar@codesourcery.com>
574
575         * gimplify.c (oacc_default_clause): Decode reference and pointer
576         types for both kernels and parallel regions.
577
578 2016-01-15  Richard Sandiford  <richard.sandiford@arm.com>
579
580         PR middle-end/69246
581         * calls.c (emit_call_1): Force n_popped to zero for sibcalls.
582
583 2016-01-15  Ilya Enkovich  <enkovich.gnu@gmail.com>
584
585         * config/i386/i386.c (scalar_chain::compute_convert_gain): Fix typo.
586         (convert_scalars_to_vector): Likewise.
587
588 2016-01-15  Jonathan Wakely  <jwakely@redhat.com>
589
590         * doc/extend.texi (Type Traits): Fix grammar.
591
592 2016-01-15  Martin Jambor  <mjambor@suse.cz>
593
594         * tree-inline.c (remap_decl): Use existing dclarations if
595         remapping a type and prevent_decl_creation_for_types.
596         (replace_locals_stmt): Do an initial remapping of non-VLA typed
597         decls first.  Do real remapping with
598         prevent_decl_creation_for_types set.
599         * tree-inline.h (copy_body_data): New field
600         prevent_decl_creation_for_types, moved remap_var_for_cilk to avoid
601         padding.
602
603 2016-01-15  Dominik Vogt  <vogt@linux.vnet.ibm.com>
604
605         * config/s390/s390.opt (mmvcle): More verbose help text.
606
607 2016-01-15  Dominik Vogt  <vogt@linux.vnet.ibm.com>
608
609         * config/s390/s390.opt: Add period to -mzvector option text.
610
611 2016-01-15  Richard Biener  <rguenther@suse.de>
612
613         PR tree-optimization/68961
614         * tree-vect-slp.c (vect_analyze_slp_cost_1): Consider cost
615         of invariants in stores again.
616
617 2016-01-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
618
619         * config/arm/t-rtems: Add cortex-m7/fpv5-d16 multilib.
620
621 2016-01-15  Ilya Enkovich  <enkovich.gnu@gmail.com>
622
623         * config/i386/i386.c (ix86_expand_branch): Don't split
624         DI mode xor instruction to SI mode.
625
626 2016-01-15  Jan Hubicka  <hubicka@ucw.cz>
627
628         PR ipa/68148
629         * ipa-icf.c (sem_function::merge): Virtual functions may become
630         reachable even if they address is not taken and there are no
631         idrect calls.
632
633 2016-01-15  Jan Hubicka  <hubicka@ucw.cz>
634
635         * lto-streamer-out.c (subtract_estimated_size): New function.
636         (get_symbol_initial_value): Use it.
637
638 2016-01-15  Christian Bruel  <christian.bruel@st.com>
639
640         PR target/65837
641         * config/arm/arm-builtins.c (ARM_BUILTIN_CRYPTO_BASE): New enum tag.
642         (arm_init_neon_builtins_internal): Rename arm_init_neon_builtins,
643         (arm_init_crypto_builtins_internal): Rename arm_init_crypto_builtins.
644         use add_builtin_function_ext_scope instead of add_builtin_function.
645         (neon_set_p, neon_crypto_set_p): Remove.
646         (arm_init_builtins): Always call arm_init_neon_builtins and
647         arm_init_crypto_builtins.
648         (arm_expand_builtin): Check that builtins are allowed for the arch.
649         * config/arm/arm-protos.h (arm_init_neon_builtins): Remove prototype.
650         * config/arm/arm.c (arm_valid_target_attribute_tree): Remove
651         arm_init_neon_builtins call.
652
653 2016-01-15  Richard Biener  <rguenther@suse.de>
654
655         PR tree-optimization/69117
656         * tree-ssa-sccvn.h (struct vn_ssa_aux): Add info member.
657         * tree-ssa-sccvn.c (set_ssa_val_to): Save and adjust SSA name info
658         of the leader conservatively.
659         (free_scc_vn): Restore original SSA name infos.
660
661 2016-01-14  Jeff Law  <law@redhat.com>
662
663         PR tree-optimization/69270
664         * tree-ssa-dom.c (ssa_name_has_boolean_range): If the type has a
665         single bit of precision, verify it's also unsigned.
666         (record_edge_info): Use constant_boolean_node rather than fold_convert
667         to convert boolean_true/boolean_false to the right type.
668
669 2016-01-14  Richard Henderson  <rth@redhat.com>
670
671         PR rtl-opt/69014
672         * loop-doloop.c (record_reg_sets): New.
673         (doloop_optimize): Reject the transform if the sequence
674         clobbers registers live at the end of the loop block.
675         (doloop_optimize_loops): Enable df_live if needed. 
676
677 2016-01-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
678
679         * config/rs6000/rs6000-builtin.def: Revert 2016-01-13 change.
680         * gcc/config/rs6000/rs6000.c: Likewise.
681         * gcc/config/rs6000/rs6000.h: Likewise.
682         * gcc/config/rs6000/rs6000.md: Likewise.
683         * gcc/doc/extend.texi: Likewsie.
684
685 2016-01-14  Jeff Law  <law@redhat.com>
686
687         * tree-ssa-dom.c (ssa_name_has_boolean_range): Fix comment
688         typo.
689
690 2016-01-14  Richard Henderson  <rth@redhat.com>
691
692         PR c/69272
693         PR tree-opt/68964
694         * trans-mem.c (tm_log_emit_stmt): Fix unit size to bit size.
695         * tree.c (build_tm_vector_builtins): Use builtin_decl_explicit_p
696         instead of builtin_decl_declared_p to test for declaration.
697
698 2016-01-14  Nicklas Bo Jensen  <nbjensen@gmail.com>
699
700         * doc/loop.texi (Loop Analysis and Representation): Document
701         loop_depth function.
702
703 2016-01-14  Tom de Vries  <tom@codesourcery.com>
704
705         PR tree-optimization/68773
706         * omp-low.c (expand_omp_target): Don't set force_output.
707         * varpool.c (varpool_node::get_create): Same.
708         * lto-cgraph.c (input_offload_tables): Mark entries in offload_vars and
709         offload_funcs with force_output.
710
711 2016-01-14  Jakub Jelinek  <jakub@redhat.com>
712
713         PR debug/69244
714         * lra-eliminations.c (move_plus_up): Don't change anything if either
715         the outer or inner subreg mode is not MODE_INT.
716         * dwarf2out.c (mem_loc_descriptor): For SUBREG, if outer mode is
717         integral <= DWARF2_ADDR_SIZE, convert to untyped afterwards.
718
719 2016-01-14  Alan Lawrence  <alan.lawrence@arm.com>
720
721         * doc/md.texi (reduc_smin_@var{m}, reduc_smax_@var{m},
722         reduc_umin_@var{m}, reduc_umax_@var{m}, reduc_splus_@var{m},
723         reduc_uplus_@var{m}): Remove.
724         * expr.c (expand_expr_real_2): Remove expansion path for
725         reduc_[us](min|max|plus) optabs.
726         * optabs-tree.c (scalar_reduc_to_vector): Remove.
727         * optabs-tree.h (scalar_reduc_to_vector): Remove.
728         * optabs.def (reduc_smax_optab, reduc_smin_optab, reduc_splus_optab,
729         reduc_umax_optab, reduc_umin_optab, reduc_uplus_optab): Remove.
730         * tree-vect-loop.c (vectorizable_reduction): Remove test for
731         reduc_[us](min|max|plus) optabs.
732
733 2016-01-14  Alan Lawrence  <alan.lawrence@arm.com>
734
735         * config/mips/mips-ps-3d.md (reduc_splus_v2sf): Remove.
736         (reduc_plus_scal_v2sf): New.
737         (reduc_smax_v2sf): Rename to...
738         (reduc_smax_scal_v2sf): ...here, make result SFmode, add vec_extract.
739         (reduc_smin_v2sf): Rename to...
740         (reduc_smin_scal_v2sf): ...here, make result SFmode, add vec_extract.
741
742 2016-01-14  Jan Hubicka  <hubicka@ucw.cz>
743
744         * alias.c (compare_base_symbol_refs): New function.
745         (rtx_equal_for_memref_p, base_alias_check, memrefs_conflict_p): Use
746         it.
747
748 2016-01-14  Jakub Jelinek  <jakub@redhat.com>
749
750         PR middle-end/68146
751         PR tree-optimization/69155
752         * tree-complex.c: Include cfganal.h.
753         (phis_to_revisit): New variable.
754         (extract_component): Add phiarg_p argument.  Assert that returned
755         SSA_NAME has non-NULL SSA_NAME_DEF_STMT unless phiarg_p is true.
756         (update_phi_components): Partly rewrite to use loop over real/imag
757         components instead of code duplication.  If extract_component returns
758         SSA_NAME with NULL SSA_NAME_DEF_STMT, store SSA_NAME_VAR or
759         create_tmp_reg into the PHI node instead, and mention the phi triplet
760         in phis_to_revisit.
761         (tree_lower_complex): Walk bbs in rpo order.  Adjust phis recorded
762         in phis_to_revisit at the end.
763
764 2016-01-14  Richard Biener  <rguenther@suse.de>
765
766         PR tree-optimization/68060
767         * tree-vect-loop.c (vect_is_simple_reduction): Check the
768         outer loop reduction is only used in the inner loop before
769         detecting a double reduction.
770
771 2016-01-14  Jakub Jelinek  <jakub@redhat.com>
772
773         PR target/68269
774         * combine.c (expand_field_assignment): Punt if compute_mode is
775         unsupported scalar mode.
776
777 2016-01-14  Richard Biener  <rguenther@suse.de>
778
779         PR tree-optimization/66856
780         * tree-vect-slp.c (vect_build_slp_tree): Refactor to build
781         SLP node only if it built successfully.
782         (vect_analyze_slp_instance): Adjust.
783
784 2016-01-14  Jeff Law  <law@redhat.com>
785
786         PR tree-optimization/69270
787         * tree-ssa-dom.c (ssa_name_has_boolean_range): New function.
788         (record_edge_info): Use it.  Convert boolean_{true,false}_node
789         to the type of op0.
790
791 2016-01-13  Jan Hubicka  <hubicka@ucw.cz>
792
793         PR ipa/66487
794         * ipa-polymorphic-call.c (inlined_polymorphic_ctor_dtor_block_p):
795         use block_ultimate_origin
796         (noncall-stmt_may_be_vtbl_ptr_store): Likewise.
797
798 2016-01-13  Sandra Loosemore <sandra@codesourcery.com>
799
800         * doc/invoke.texi (Submodel Options): Rename section to
801         "Machine-Dependent Options" to better reflect its content.
802         Rewrite introductory text to remove archaic CPU names.
803         Update references.
804
805 2016-01-13  Sandra Loosemore <sandra@codesourcery.com>
806
807         * doc/invoke.texi (Code Gen Options): Move section up in file,
808         before target-specific options.  Update menu and option summary
809         to reflect the new section ordering.
810
811 2016-01-13  Jonathan Wakely  <jwakely@redhat.com>
812
813         * doc/invoke.texi (C Dialect Options): Adjust -std default for C++.
814         (C++ Dialect Options): Add cross-reference to -std option.
815         * doc/standards.texi (C++ Language): Document C++14 support.
816
817 2016-01-13  Michael Meissner  <meissner@linux.vnet.ibm.com>
818
819         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_2): Add support
820         for pack/unpack functions for __ibm128.
821         (PACK_IF): Likewise.
822         (UNPACK_IF): Likewise.
823
824         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add
825         support for __ibm128 pack/unpack functions.
826         (rs6000_invalid_builtin): Likewise.
827         (rs6000_init_builtins): Likewise.
828         (rs6000_opt_masks): Likewise.
829
830         * config/rs6000/rs6000.h (MASK_FLOAT128): Add short name.
831         (RS6000_BTM_FLOAT128): Add support for __ibm128 pack/unpack
832         functions
833         (RS6000_BTM_COMMON): Likewise.
834
835         * config/rs6000/rs6000.md (f128_vsx): New mode attribute.
836         (unpack<mode>): Use FMOVE128_FPR iterator instead of FMOVE128, to
837         disallow __builtin_{pack,unpack}_longdouble if long double is IEEE
838         128-bit floating point.  Add support for the double values to be
839         in Altivec registers for TF/IF packing and unpacking, but restrict
840         TD packing sub-fields to be FPR registers.  Don't allow overlapped
841         register support for packing.  Allow pack inputs to be memory
842         locations.  Don't build generator functions for unpack<mode>_dm
843         and unpack<mode>_nodm.
844         (unpack<mode>_dm): Likewise.
845         (unpack<mode>_nodm): Likewise.
846         (pack<mode>): Likewise.
847
848         * config/rs6000/rs6000-builtin.def (__builtin_pack_ibm128): Add
849         built-in functions to pack/unpack explicit __ibm128 values.
850         (__builtin_unpack_ibm128): Likewise.
851
852         * doc/extend.texi (PowerPC Built-in Functions): Document
853         __builtin_pack_ibm128 and __builtin_unpack_ibm128.
854
855 2016-01-13  Bernd Schmidt  <bschmidt@redhat.com>
856
857         PR c/66208
858         * c-common.c (check_function_nonnull): Remove unnecessary declaration.
859         Add new arg loc and pass it down as context.
860         (check_nonnull_arg): Don't mark ctx arg as unused. Use it as a pointer
861         to the location to use for the warning.
862         (check_function_arguments): New arg loc.  All callers changed.  Pass
863         it to check_function_nonnull.
864         * c-common.h (check_function_arguments): Adjust declaration.
865
866 2016-01-13  Jakub Jelinek  <jakub@redhat.com>
867
868         PR tree-optimization/69156
869         * gimple.c (validate_type): Removed.
870         (gimple_builtin_call_types_compatible_p): Use
871         useless_type_conversion_p instead of validate_type.
872         * value-prof.c (gimple_stringop_fixed_value): Fold
873         icall_size to correct type.
874
875 2016-01-13  Jonathan Wakely  <jwakely@redhat.com>
876
877         * doc/extend.texi (__atomic Builtins): Clarify compare_exchange
878         effects.
879
880 2016-01-13  Richard Henderson  <rth@redhat.com>
881
882         PR tree-opt/68964
883         * target.def (builtin_tm_load, builtin_tm_store): Remove.
884         * config/i386/i386.c (ix86_builtin_tm_load): Remove.
885         (ix86_builtin_tm_store): Remove.
886         (TARGET_VECTORIZE_BUILTIN_TM_LOAD): Remove.
887         (TARGET_VECTORIZE_BUILTIN_TM_STORE): Remove.
888         * doc/tm.texi.in (TARGET_VECTORIZE_BUILTIN_TM_LOAD): Remove.
889         (TARGET_VECTORIZE_BUILTIN_TM_STORE): Remove.
890         * doc/tm.texi: Rebuild.
891
892         * gtm-builtins.def (BUILT_IN_TM_MEMCPY_RNWT): New.
893         (BUILT_IN_TM_MEMCPY_RTWN): New.
894         * trans-mem.c (tm_log_emit_stmt): Rearrange code for better
895         fallback from vector to integer helpers.
896         (build_tm_load): Handle vector types directly, instead of
897         via target hook.
898         (build_tm_store): Likewise.
899         (expand_assign_tm): Prepare for register types not handled by
900         the above.  Copy them to memory and use memcpy.
901         * tree.c (tm_define_builtin): New.
902         (find_tm_vector_type): New.
903         (build_tm_vector_builtins): New.
904         (build_common_builtin_nodes): Call it.
905
906 2016-01-13  Uros Bizjak  <ubizjak@gmail.com>
907
908         * config/i386/i386.h (TARGET_FLT_EVAL_METHOD): Return -1 for
909         TARGET_SSE_MATH without TARGET_SSE2.  Rewrite.
910
911 2016-01-13  Tom de Vries  <tom@codesourcery.com>
912
913         PR tree-optimization/69169
914         * tree-ssa-structalias.c (create_variable_info_for_1): Add and handle
915         handled_struct_type param.
916         (create_variable_info_for, intra_create_variable_infos): Call
917         create_variable_info_for_1 with extra arg.
918
919 2016-01-13  Yvan Roux  <yvan.roux@linaro.org>
920
921         * config/arm/arm-arches.def: Remove spurious whitespace in "armv8.1-a"
922         and "armv8.1-a+crc" entries.
923
924 2016-01-13  Alexander Fomin  <alexander.fomin@intel.com>
925
926         PR target/69228
927         * config/i386/sse.md (define_expand "avx512pf_gatherpf<mode>sf"):
928         Change first operand predicate from register_or_constm1_operand
929         to register_operand.
930         (define_expand "avx512pf_gatherpf<mode>df"): Likewise.
931         (define_expand "avx512pf_scatterpf<mode>sf"): Likewise.
932         (define_expand "avx512pf_scatterpf<mode>df"): Likewise.
933         (define_insn "*avx512pf_gatherpf<mode>sf"): Remove.
934         (define_insn "*avx512pf_gatherpf<mode>df"): Likewise.
935         (define_insn "*avx512pf_scatterpf<mode>sf"): Likewise.
936         (define_insn "*avx512pf_scatterpf<mode>df"): Likewise.
937         * config/i386/i386.c (ix86_expand_builtin): Remove first operand
938         comparison with constm1_rtx from vec_prefetch_gen part.
939
940 2016-01-13  Richard Biener  <rguenther@suse.de>
941
942         PR tree-optimization/69013
943         * tree-ssa-uninit.c (prune_uninit_phi_opnds_in_unrealizable_paths):
944         Exchange assert for a test.
945
946 2016-01-13  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
947
948         PR target/69247
949         * config/s390/s390.md (bswaphi splitter): Use simplify_gen_subreg.
950
951 2016-01-13  Richard Biener  <rguenther@suse.de>
952
953         PR tree-optimization/69242
954         * tree-ssa-sccvn.c (visit_reference_op_load): Replace bogus
955         assert with a check.
956
957 2016-01-13  Richard Biener  <rguenther@suse.de>
958
959         PR tree-optimization/69186
960         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment):
961         Properly guard vect_update_misalignment_for_peel call.
962
963 2016-01-12  Jeff Law  <law@redhat.com>
964
965         PR tree-optimization/pr67755
966         * tree-ssa-threadupdate.c (struct ssa_local_info_t): Add new field
967         "need_profile_correction".
968         (thread_block_1): Initialize new field to false by default.  If we
969         have multiple thread paths through a common joiner to different
970         final targets, then set new field to true.
971         (compute_path_counts): Only do count adjustment when it's really
972         needed.
973
974 2016-01-12  Sandra Loosemore  <sandra@codesourcery.com>
975
976         * doc/invoke.texi (Spec Files): Move section down in file, past
977         all command-line option descriptions.
978
979 2016-01-12  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
980
981         PR middle-end/54809
982         * doc/gty.texi: Remove documentation of mark_hook.
983         * gengtype.c (struct write_types_data): Remove code to support
984         mark_hook attribute.
985         (walk_type): Likewise.
986         (write_func_for_structure): Likewise.
987
988 2016-01-12  Sandra Loosemore  <sandra@codesourcery.com>
989
990         * doc/invoke.texi (Option Summary) Move -no-canonical-prefixes to
991         Directory Options, and -specs= to Overall Options.
992         (Overall Options): Adjust similarly.  Reorder to group related
993         options together.  Make -specs= cross-reference the spec file details.
994         (Directory Options): Adjust similarly.
995
996 2016-01-12  Jeff Law  <law@redhat.com>
997
998         * tree-ssa-threadupdate.c: Various whitespace and typo fixes.
999
1000 2016-01-12  Olivier Hainque  <hainque@adacore.com>
1001
1002         * gcc.c (spec_undefvar_allowed): New global.
1003         (process_command): Set to true when running for --version or --help,
1004         alone or together.
1005         (getenv_spec_function): When the variable is not defined, use the
1006         variable name as the variable value if we're allowed not to issue
1007         a fatal error.
1008
1009 2016-01-12  Bin Cheng  <bin.cheng@arm.com>
1010
1011         PR tree-optimization/68911
1012         * tree-vrp.c (adjust_range_with_scev): Check overflow in range
1013         information computed for expression "init + nit * step".
1014
1015 2016-01-12  Sandra Loosemore  <sandra@codesourcery.com>
1016
1017         * doc/invoke.texi (Invoking GCC): Copy-edit.  Incorporate information
1018         about name of GCC executable.  Remove deleted node from menu.
1019         (Directory Options) <-B>: Remove cross-reference to deleted node.
1020         (Target Options): Delete section.
1021
1022 2016-01-12  Christian Bruel  <christian.bruel@st.com>
1023
1024         PR target/69180
1025         * config/arm/arm-c.c (arm_pragma_target_parse): Set NODE_CONDITIONAL
1026         for __ARM_NEON_FP, __ARM_FP, _ARM_FEATURE_LDREX.
1027
1028 2016-01-12  Jakub Jelinek  <jakub@redhat.com>
1029
1030         PR target/69198
1031         * config/i386/i386.c (ix86_expand_special_args_builtin): Ensure
1032         aligned_mem is properly set for AVX512-VL floating point masked
1033         stores.
1034
1035         PR target/69175
1036         * ifcvt.c (cond_exec_process_if_block): When removing the last
1037         insn from then_bb, remove also any possible barriers that follow it.
1038
1039 2016-01-12  H.J. Lu  <hongjiu.lu@intel.com>
1040
1041         PR target/68456
1042         PR target/69226
1043         * config/i386/iamcu.h (SIZE_TYPE): New macro.
1044         (PTRDIFF_TYPE): Likewise.
1045         (WCHAR_TYPE): Likewise.
1046         (WCHAR_TYPE_SIZE): Likewise.
1047         (STDINT_LONG32): Likewise.
1048
1049 2016-01-12  Richard Biener  <rguenther@suse.de>
1050
1051         PR tree-optimization/69053
1052         * tree-vect-loop.c (get_initial_def_for_reduction): Properly
1053         convert initial value for cond reductions.
1054
1055 2016-01-12  Richard Biener  <rguenther@suse.de>
1056
1057         PR tree-optimization/69007
1058         * tree-vect-patterns.c (vect_vect_recog_func_ptrs): Move
1059         widen_sum after dot_prod and sad.
1060
1061 2016-01-12  Richard Biener  <rguenther@suse.de>
1062
1063         PR tree-optimization/69168
1064         * tree-vect-loop.c (vect_analyze_loop_2): Reset both main and
1065         pattern stmt SLP type.
1066         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Patterns may
1067         end up unused so cope with that case.
1068
1069 2016-01-12  Richard Biener  <rguenther@suse.de>
1070
1071         PR tree-optimization/69157
1072         * tree-vect-stmts.c (vectorizable_mask_load_store): Check
1073         stmts def type only during analyze phase.
1074         (vectorizable_call): Likewise.
1075         (vectorizable_simd_clone_call): Likewise.
1076         (vectorizable_conversion): Likewise.
1077         (vectorizable_assignment): Likewise.
1078         (vectorizable_shift): Likewise.
1079         (vectorizable_operation): Likewise.
1080         (vectorizable_store): Likewise.
1081         (vectorizable_load): Likewise.
1082
1083 2016-01-12  Richard Biener  <rguenther@suse.de>
1084
1085         PR tree-optimization/69174
1086         * tree-vect-stmts.c (vect_mark_relevant): Remove excessive vertical
1087         space.
1088         (vectorizable_load): Properly compute the number of loads needed
1089         for permuted strided SLP loads and do not spuriously assign
1090         to SLP_TREE_VEC_STMTS.
1091
1092 2016-01-12  Andris Pavenis  <andris.pavenis@iki.fi>
1093
1094         * config/i386/djgpp.h (PREFERRED_DEBUGGING_TYPE): Define to DWARF2
1095         (TARGET_ASM_OUTPUT_IDENT): Define to default_asm_output_ident_directive
1096         (MD_EXEC_PREFIX): Remove.
1097         (MD_STARTFILE_PREFIX) Removee.
1098         (FILE_NAME_ABSOLUTE_P): Remove.
1099         (CPP_SPEC): Do not read macros from sys/version.h.
1100         (LINK_COMMAND_SPEC): Remove.
1101         (LOCAL_INCLUDE_DIR): Remove.
1102         (TARGET_ASM_NAMED_SECTION): Define to i386_djgpp_asm_named_section
1103         (TARGET_OS_CPP_BUILTINS): Add DJGPP (non ISO only), __DJGPP, __DJGPP__, unix.
1104         (POST_LINK_SPEC): Define to invoke stubify after linker
1105         (LIBSTDCXX): Remove define
1106         (DBX_REGISTER_NUMBER): Define to svr4_dbx_register_map.
1107         (DEFAULT_PCC_STRUCT_RETURN): Define to 1.
1108         (SUBTARGET_OVERRIDE_OPTIONS): Remove warning about -mbnu2210.
1109         (SUBTARGET_OVERRIDE_OPTIONS): Ignore -fPIC and generate message.
1110         (SUBTARGET_OVERRIDE_OPTIONS): Default to DWARF2 debugging info.
1111         (IX86_MAYBE_NO_LIBGCC_TFMODE): Remove.
1112         (i386_djgpp_asm_named_section): Add propotype of new procedure
1113
1114         * config/i386/xm-djgpp.h (NATIVE_SYSTEM_HEADER_DIR): Define.
1115         (MD_EXEC_PREFIX): Define (moved from config/i386/djgpp.h).
1116         (STANDARD_STARTFILE_PREFIX_1): Define (moved from MD_STARTFILE_PREFIX in config/i386/djgpp.h).
1117         (STANDARD_STARTFILE_PREFIX_2): Define identical to STANDARD_STARTFILE_PREFIX_1.
1118         (LOCAL_INCLUDE_DIR): Define (moved from config/i386/djgpp.h).
1119         (GCC_DRIVER_HOST_INITIALIZATION): Fix reporting fatal installation errors.
1120         (MAX_OFILE_ALIGNMENT): Define to 128.
1121         (HAVE_FTW_H): Undefine as DJGPP do not have nftw, but have ftw.h.
1122
1123         * config/i386/djgpp.c: New file. Add implementation of i386_djgpp_asm_named_section.
1124
1125         * config/i386/djgpp.opt: Remove obsolete option -mbnu210.
1126
1127         * config/i386/t-djgpp: New file. Add djgpp.o to EXTRA_OBJS.
1128         Add rule for building djgpp.o.
1129
1130 2016-01-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1131
1132         * config/rs6000/rs6000.c (v2df_reduction_p): New function.
1133         (rtx_is_swappable_p): Reductions are swappable.
1134         (insn_is_swappable_p): V2DF reductions are swappable.
1135
1136 2016-01-11  John David Anglin  <danglin@gcc.gnu.org>
1137
1138         * config/pa/pa.c (pa_emit_move_sequence): Handle floating point
1139         reloads for other unsupported memory operands.
1140
1141 2016-01-12  Kugan Vivekanandarajah  <kuganv@linaro.org>
1142             Jim Wilson  <jim.wilson@linaro.org>
1143
1144         PR target/69194
1145         * config/arm/arm-builtins.c (arm_expand_neon_args): Call
1146         copy_to_mode_reg instead of force_reg.
1147
1148 2016-01-11  H.J. Lu  <hongjiu.lu@intel.com>
1149
1150         PR target/69225
1151         * config/i386/i386.h (TARGET_FLT_EVAL_METHOD): Set to 2 only if
1152         TARGET_80387 is true.
1153
1154 2016-01-11  Jakub Jelinek  <jakub@redhat.com>
1155
1156         PR target/69071
1157         * lra-eliminations.c (move_plus_up): Only move plus up
1158         if subreg of the constant can be simplified into constant
1159         and use the simplified subreg of the constant instead of
1160         the original constant.
1161
1162         * fold-const.c (fold_convertible_p): Don't return true
1163         for conversion of VECTOR_TYPE to same sized integral type.
1164         (fold_convert_loc): Fix up formatting.  Fold conversion of
1165         VECTOR_TYPE to same sized integral type using VIEW_CONVERT_EXPR
1166         instead of NOP_EXPR.
1167
1168         PR tree-optimization/69214
1169         * tree-vrp.c (simplify_cond_using_ranges): Don't propagate
1170         innerop into a comparison if SSA_NAME_OCCURS_IN_ABNORMAL_PHI.
1171         Formatting fix.
1172
1173         PR tree-optimization/69207
1174         * tree-vect-slp.c (vect_get_constant_vectors): For
1175         VECTOR_BOOLEAN_TYPE_P, assert op has integral type instead of
1176         fold_convertible_p to vector_type's element type, and always
1177         use VCE for non-VECTOR_BOOLEAN_TYPE_P.
1178
1179 2016-01-11  Richard Biener  <rguenther@suse.de>
1180
1181         PR tree-optimization/69173
1182         * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Only
1183         fixup the cycle if all stmts are in a pattern.
1184
1185 2016-01-11  Uros Bizjak  <ubizjak@gmail.com>
1186
1187         PR middle-end/68999
1188         * alias.c (base_alias_check): Move check for addresses with
1189         alignment ANDs before the call for compare_base_decls.
1190         (memrefs_conflict_p): Return -1 for different decls
1191         that went through alignment adjustments.
1192
1193 2016-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1194
1195         PR rtl-optimization/68796
1196         * config/aarch64/aarch64.md (*and<mode>_compare0): New pattern.
1197         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Handle HImode
1198         and QImode comparisons against zero with CC_NZmode.
1199         * config/aarch64/iterators.md (short_mask): New mode_attr.
1200
1201 2016-01-11  H.J. Lu  <hongjiu.lu@intel.com>
1202
1203         * config/i386/sse.md (<avx512>_load<mode>_mask): Remove snprintf.
1204         (<avx512>_store<mode>_mask): Likewise.
1205
1206 2016-01-11  Bernd Schmidt  <bschmidt@redhat.com>
1207             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1208
1209         PR rtl-optimization/68841
1210         * ifcvt.c (struct noce_if_info): Add orig_x field.
1211         (bbs_ok_for_cmove_arith): Add to_rename parameter.
1212         Don't record conflicts on to_rename if it's present.
1213         Allow memory destinations in sets.
1214         (noce_try_cmove_arith): Call bbs_ok_for_cmove_arith even on simple
1215         blocks, passing orig_x to the checks.
1216         (noce_process_if_block): Set if_info->orig_x appropriately.
1217
1218 2016-01-11  Tom de Vries  <tom@codesourcery.com>
1219
1220         PR tree-optimization/69069
1221         * tree-parloops.c (create_parallel_loop): Add missing phi args.
1222
1223 2016-01-11  Yuri Rumyantsev  <ysrumyan@gmail.com>
1224
1225         PR rtl-optimization/68920
1226         * config/i386/i386.c (ix86_option_override_internal): Restrict number
1227         of conditional moves for  RTL if-conversion to 1 for
1228         TARGET_ONE_IF_CONV_INSN.
1229         * config/i386/i386.h (TARGET_ONE_IF_CONV_INSN): New macros.
1230         * config/i386/x86-tune.def (X86_TUNE_ONE_IF_CONV_INSN): New macros.
1231         * params.def (PARAM_MAX_RTL_IF_CONVERSION_INSNS) : Introduce new
1232         parameter to restirct number of conditional moves for
1233         RTL if-conversion.
1234         * doc/invoke.texi (max-rtl-if-conversion-insns): Document it.
1235         * ifcvt.c (bb_ok_for_noce_convert_multiple_sets): Limit number of
1236         conditionl moves.
1237
1238 2016-01-11  Alexandre Oliva  <aoliva@redhat.com>
1239
1240         PR bootstrap/69123
1241         * var-tracking.c (drop_overlapping_mem_locs): Operate on all
1242         onepart vars.  Fix typo in comment.  Fix reversed condition in
1243         unshare test.
1244         (dataflow_set_remove_mem_locs): Operate on all onepart vars.
1245
1246         PR bootstrap/69123
1247         * var-tracking.c (dump_onepart_variable_differences): New.
1248         (dataflow_set_different): If a detailed dump is requested,
1249         delay early returns and dump differences between onepart
1250         variables present before and after, and added variables.
1251
1252 2016-01-11  Ilya Enkovich  <enkovich.gnu@gmail.com>
1253
1254         PR target/69010
1255         * expr.c (expand_expr_real_1): For boolean vector constants
1256         with a scalar mode use const_scalar_mask_from_tree.
1257         (const_scalar_mask_from_tree): New.
1258         * optabs.c (expand_vec_cond_mask_expr): Use mask mode
1259         assigned to a mask type to handle constants.
1260
1261 2016-01-11  Martin Jambor  <mjambor@suse.cz>
1262
1263         PR ipa/69044
1264         * ipa-cp.c (estimate_local_effects): Do not clone for removal of
1265         useless parameters if we cannot change function signature.
1266
1267 2016-01-11  Martin Jambor  <mjambor@suse.cz>
1268
1269         PR ipa/66616
1270         * cgraphclones.c (duplicate_thunk_for_node): Copy can_change_signature
1271         flag.
1272
1273 2016-01-11  Tom de Vries  <tom@codesourcery.com>
1274
1275         PR tree-optimization/69109
1276         * tree-parloops.c (try_transform_to_exit_first_loop_alt): Don't allow
1277         latch with phi.
1278
1279 2016-01-11  Tom de Vries  <tom@codesourcery.com>
1280
1281         PR tree-optimization/69108
1282         * tree-parloops.c (gather_scalar_reductions): Handle case that outer phi
1283         res is not used in a phi.
1284
1285 2016-01-11  Yury Gribov  <y.gribov@samsung.com>
1286
1287         PR 67425
1288         * common.opt (frandom-seed): Fix parameter name.
1289         * doc/invoke.texi (frandom-seed): Ditto and describe parameter.
1290
1291 2016-01-11  Tom de Vries  <tom@codesourcery.com>
1292
1293         PR tree-optimization/69058
1294         * tree-parloops.c (pass_parallelize_loops::execute): Return 0 if libgomp
1295         not supported.
1296
1297 2016-01-11  Andrew Burgess  <andrew.burgess@embecosm.com>
1298
1299         * config/arc/arc.opt (mdiv-rem): Add period to the end.
1300         (mcode-density): Likewise.
1301
1302 2016-01-10  Tom de Vries  <tom@codesourcery.com>
1303
1304         PR tree-optimization/69062
1305         * tree-parloops.c (loop_has_phi_with_address_arg): New function.
1306         (parallelize_loops): Don't paralelize loop that has phi with address
1307         arg.
1308
1309 2016-01-10  Tom de Vries  <tom@codesourcery.com>
1310
1311         PR tree-optimization/69039
1312         * tree-parloops.c (try_create_reduction_list): Only allow single exit
1313         phi for reduction.
1314
1315 2016-01-09  John David Anglin  <danglin@gcc.gnu.org>
1316
1317         PR middle-end/68743
1318         * match.pd: Require target has function_c99_misc before doing
1319         truncl(extend(x)) and trunc(extend(x)) -> extend(truncf(x)), etc.
1320         
1321 2016-01-09  Gerald Pfeifer  <gerald@pfeifer.com>
1322
1323         * configure.ac (isl_options_set_schedule_serialize_sccs): Also 
1324         use GMPINC.
1325         * configure: Regenerate.
1326
1327 2016-01-09  Jakub Jelinek  <jakub@redhat.com>
1328
1329         PR middle-end/50865
1330         PR tree-optimization/69097
1331         * fold-const.h (expr_not_equal_to): New prototype.
1332         * fold-const.c: Include stringpool.h and tree-ssanames.h.
1333         (expr_not_equal_to): New function.
1334         * match.pd (X % -Y is the same as X % Y): Don't optimize
1335         unless X is known not to be equal to minimum or Y is known
1336         not to be equal to -1.
1337         * tree-vrp.c (simplify_div_or_mod_using_ranges): Add GSI argument.
1338         fold TRUNC_MOD_EXPR if the second argument is not a power of two.
1339         (simplify_stmt_using_ranges): Adjust caller.
1340         (vrp_finalize): Call set_value_range on SSA_NAMEs before calling
1341         substitute_and_fold.
1342
1343 2016-01-09  Jan Hubicka  <hubicka@ucw.cz>
1344
1345         * ipa-icf.c (sem_item_optimizer::merge_classes): Do not ICE on VAR_DECL
1346         w/o DECL_NAME.
1347
1348 2016-01-08  Jakub Jelinek  <jakub@redhat.com>
1349
1350         PR tree-optimization/69167
1351         * gimple-fold.c (replace_stmt_with_simplification): Also punt if
1352         new SSA_NAME_OCCURS_IN_ABNORMAL_PHI SSA_NAMEs appear in operands of
1353         ops[0] comparison.
1354         * gimple-match-head.c (maybe_push_res_to_seq): Likewise.
1355
1356 2016-01-08  Alan Lawrence  <alan.lawrence@arm.com>
1357             Richard Biener  <rguenther@suse.de>
1358
1359         PR tree-optimization/68707
1360         * tree-vect-slp.c (vect_analyze_slp_instance): Cancel permuted SLP
1361         instances that can be handled via vect_load_lanes.
1362
1363 2016-01-08  Uros Bizjak  <ubizjak@gmail.com>
1364
1365         * symtab.c (symtab_node::equal_address_to): Return -1 instead of 2
1366         if we can't determine address equivalence.
1367         * alias.c (compare_base_decl): Update for changed return value of
1368         symtab_node::equal_address_to.
1369
1370 2016-01-08  Jason Merrill  <jason@redhat.com>
1371
1372         PR c++/68983
1373         PR c++/67557
1374         * function.c (assign_temp): Guard against TREE_ADDRESSABLE types here.
1375         * expr.c (store_field): Not here.
1376         * tree-cfgcleanup.c (fixup_noreturn_call): Don't clear LHS of a
1377         call with TREE_ADDRESSABLE type.
1378         * tree-cfg.c (verify_gimple_call): Adjust.
1379
1380 2016-01-08  Olivier Hainque  <hainque@adacore.com>
1381
1382         * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Don't link shared RTPs with
1383         libc_internal.
1384
1385 2016-01-08  Alan Lawrence  <alan.lawrence@arm.com>
1386
1387         * gcc.target/rs6000/paired.md (reduc_smax_v2sf): Rename to...
1388         (reduc_smax_scal_v2sf): ...here, make result SFmode, extract element.
1389         (reduc_smin_v2sf): Rename to...
1390         (reduc_smin_scal_v2sf): ...here, make result SFmode, extract element.
1391         (reduc_splus_v2sf): Rename to...
1392         (reduc_plus_scal_v2sf): ...here, make result SFmode, extract element.
1393
1394 2016-01-08  Jakub Jelinek  <jakub@redhat.com>
1395
1396         PR tree-optimization/69162
1397         * gimplify.c (gimplify_va_arg_expr): Encode original type of
1398         valist argument in another argument.
1399         (gimplify_modify_expr): Adjust for the above change.  Cleanup.
1400         * tree-stdarg.c (expand_ifn_va_arg_1): Use new 3rd argument
1401         to determine the va_list type, build a MEM_REF instead of
1402         build_fold_indirect_ref.
1403
1404         PR tree-optimization/69172
1405         * gimple-fold.c (gimple_fold_builtin_memory_chk): Pass type to
1406         gimple_build.
1407
1408 2016-01-08  Thomas Preud'homme  <thomas.preudhomme@arm.com>
1409
1410         PR tree-optimization/67781
1411         * tree-ssa-math-opts.c (find_bswap_or_nop): Zero out bytes in cmpxchg
1412         and cmpnop in two steps: first the ones not accessed in original
1413         gimple expression in a endian independent way and then the ones not
1414         accessed in the final result in an endian-specific way.
1415
1416 2016-01-08  Jakub Jelinek  <jakub@redhat.com>
1417
1418         PR tree-optimization/69083
1419         * tree-vect-slp.c (vect_get_constant_vectors): For
1420         VECTOR_BOOLEAN_TYPE_P assert op is fold_convertible_p to vector_type's
1421         element type.  If op is fold_convertible_p to vector_type's element
1422         type, use NOP_EXPR instead of VCE.
1423
1424 2016-01-08  Segher Boessenkool  <segher@kernel.crashing.org>
1425
1426         PR rtl-optimization/67778
1427         PR rtl-optimization/68634
1428         PR rtl-optimization/68909
1429         * shrink-wrap.c (try_shrink_wrapping): Add comment.  Don't pop
1430         block from the stack until done with it.  Remove a superfluous
1431         bitmap set.  Remove a superfluous bitmap test.
1432
1433 2016-01-07  Martin Sebor  <msebor@redhat.com>
1434
1435         PR c/68966
1436         * doc/extend.texi (__atomic Builtins, __sync Builtins): Document
1437         constraint on the type of arguments.
1438
1439 2016-01-07  Andreas Tobler  <andreast@gcc.gnu.org>
1440
1441         * config/arm/freebsd.h: Rename SUBTARGET_OVERRIDE_OPTIONS to
1442         SUBTARGET_OVERRIDE_INTERNAL_OPTIONS. Adjust to check
1443         unaligned_access on the gcc_options set.
1444         * config/arm/arm.c (arm_option_override_internal): Use
1445         SUBTARGET_OVERRIDE_INTERNAL_OPTIONS.
1446
1447 2016-01-07  Uros Bizjak  <ubizjak@gmail.com>
1448
1449         PR target/69140
1450         * config/i386/i386.c (ix86_frame_pointer_required): Enable
1451         frame pointer for TARGET_64BIT_MS_ABI when stack is misaligned.
1452
1453 2016-01-07  Uros Bizjak  <ubizjak@gmail.com>
1454
1455         Revert
1456         2016-01-06  Uros Bizjak  <ubizjak@gmail.com>
1457
1458         PR target/69140
1459         * config/i386/i386.c (ix86_expand_prologue): Declare fs.sp_valid
1460         depending on frame_pointer_needed before remaining integer and SSE
1461         registers are saved.
1462
1463 2016-01-07  Sandra Loosemore  <sandra@codesourcery.com>
1464
1465         PR 1078
1466         * doc/extend.texi (Nvidia PDX Function Attributes): New section.
1467
1468 2016-01-07  H.J. Lu  <hongjiu.lu@intel.com>
1469
1470         PR target/69171
1471         * config/i386/sse.md (<sse>_sqrt<mode>2<mask_name><round_name>):
1472         Use the "xBm" constraint.
1473         (float<sseintvecmodelower><mode>2<mask_name><round_name):
1474         Likewise.
1475         (sse_cvtsi2ss<round_name>): Use round_nimm_scalar_predicate.
1476         (sse_cvtsi2ssq<round_name>): Likewise.
1477         (sse_cvtss2si<round_name>): Likewise.
1478         (sse_cvtss2siq<round_name>): Likewise.
1479         (sse2_cvtsi2sdq<round_name>): Likewise.
1480         (sse2_cvtsd2si<round_name>): Likewise.
1481         (sse2_cvtsd2siq<round_name>): Likewise.
1482         * config/i386/subst.md (round_nimm_scalar_predicate): New
1483         predicate.
1484
1485 2015-12-15  Bernd Schmidt  <bschmidt@redhat.com>
1486
1487         PR middle-end/67639
1488         * varasm.c (make_decl_rtl): Mark invalid register vars as
1489         DECL_EXTERNAL.
1490
1491         PR rtl-optimization/66206
1492         * bt-load.c (find_btr_use): Change first arg to be a pointer to an rtx.
1493         All callers changed.
1494
1495 2016-01-07  Jakub Jelinek  <jakub@redhat.com>
1496
1497         PR tree-optimization/69141
1498         * tree-ssa-pre.c: Include langhooks.h.
1499         (eliminate_dom_walker::before_dom_children): Use
1500         lang_hooks.decl_printable_name instead of
1501         cgraph_node::get ()->name ().
1502
1503         PR middle-end/68960
1504         * gimple-expr.c (copy_var_decl): If var has DECL_USER_ALIGN set, copy
1505         it and DECL_ALIGN too.
1506
1507 2016-01-06  Robert Suchanek  <robert.suchanek@imgtec.com>
1508
1509         * config/mips/mips-ftypes.def: Sort to lexicographical order.
1510
1511 2016-01-06  Uros Bizjak  <ubizjak@gmail.com>
1512
1513         PR target/69140
1514         * config/i386/i386.c (ix86_expand_prologue): Declare fs.sp_valid
1515         depending on frame_pointer_needed before remaining integer and SSE
1516         registers are saved.
1517
1518 2015-01-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1519
1520         * config/rs6000/vsx.md (*p9_vecload_<mode>): Replace VSX_M
1521         mode iterator with VSX_M2.
1522         (*p9_vecstore_<mode>): Likewise.
1523         (*vsx_le_permute_<mode>): Restrict to !TARGET_P9_VECTOR.
1524         (*vsx_le_perm_load_<mode> for VSX_LE_128): Likewise.
1525         (*vsx_le_perm_store_<mode> for VSX_LE_128): Likewise.
1526         (define_split for VSX_LE128 stores): Likewise.
1527         (define_peephole2 for TImode LE swaps): Likewise.
1528         (define_split for VSX_LE128 post-reload stores): Likewise.
1529
1530 2016-01-06  Marek Polacek  <polacek@redhat.com>
1531
1532         PR sanitizer/69099
1533         * convert.c (convert_to_integer_1): Adjust call to
1534         ubsan_instrument_float_cast.  Use NULL_TREE instead of NULL.
1535         * ubsan.c (ubsan_instrument_float_cast): Drop the ARG parameter.  Use
1536         EXPR instead of ARG.
1537         * ubsan.h (ubsan_instrument_float_cast): Adjust declaration.
1538
1539 2016-01-05  Sandra Loosemore  <sandra@codesourcery.com>
1540
1541         PR 1078
1542         * doc/extend.texi (RL78 Variable Attributes): New section.
1543
1544 2016-01-05  Marek Polacek  <polacek@redhat.com>
1545
1546         PR c/69104
1547         * builtins.c (get_memmodel): Use expansion point location rather than
1548         the input location.  Call warning_at rather than warning.
1549         (expand_builtin_atomic_compare_exchange): Likewise.
1550         (expand_builtin_atomic_load): Likewise.
1551         (expand_builtin_atomic_store): Likewise.
1552         (expand_builtin_atomic_clear): Likewise.
1553
1554 2016-01-05  H.J. Lu  <hongjiu.lu@intel.com>
1555
1556         PR target/68991
1557         * config/i386/i386.c (ix86_expand_vector_logical_operator):
1558         Replace nonimmediate_operand with vector_operand.
1559         * config/i386/predicates.md (vector_operand): New predicate.
1560         (general_vector_operand): Replace nonimmediate_operand with
1561         vector_operand.
1562         * config/i386/sse.md: Replace nonimmediate_operand with
1563         vector_operand and m constraint with Bm constraint on SSE
1564         patterns with 16-byte memory operand.
1565         * config/i386/subst.md (round_nimm_predicate): Replace
1566         nonimmediate_operand with vector_operand.
1567         (round_saeonly_nimm_predicate): Likewise.
1568         (round_saeonly_nimm_scalar_predicate): New.
1569
1570 2016-01-05  H.J. Lu  <hongjiu.lu@intel.com>
1571
1572         PR target/68991
1573         * config/i386/constraints.md (Bm): New constraint.
1574         * config/i386/predicates.md (vector_memory_operand): New
1575         predicate.
1576         * config/i386/sse.md: Replace xm with xBm in plusminus and
1577         any_logic patterns.
1578
1579 2016-01-05  Sandra Loosemore  <sandra@codesourcery.com>
1580
1581         PR 1078
1582         * doc/extend.texi (V850 Function Attributes): New section.
1583         (V850 Variable Attributes): New section.
1584
1585 2016-01-05  Sandra Loosemore  <sandra@codesourcery.com>
1586
1587         PR 1078
1588         * doc/extend.texi (MicroBlaze Function Attributes): Document
1589         interrupt_handler and fast_interrupt attributes.
1590
1591 2016-01-05  Sergei Trofimovich  <siarheit@google.com>
1592
1593         PR other/60465
1594         * config/ia64/ia64.c (ia64_expand_load_address): Use gprel64
1595         for local symbolic operands.
1596         * config/ia64/predicates.md (local_symbolic_operand64): New
1597         predicate.
1598
1599 2016-01-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1600
1601         PR rtl-optimization/68651
1602         * combine.c (combine_simplify_rtx): Canonicalize x + x into
1603         x << 1.
1604
1605 2016-01-05  Nathan Sidwell  <nathan@acm.org>
1606
1607         * alias.c (compare_base_decls): Use symtab_node::get.
1608
1609 2016-01-05  Nick Clifton  <nickc@redhat.com>
1610
1611         PR target/68770
1612         * ira-costs.c (copy_cost): Initialise the t_icode field of the
1613         secondary_reload_info structure.
1614
1615         PR target/66655
1616         * config/i386/cygming.h (MAKE_DECL_ONE_ONLY): Define to use weak
1617         decls if weak support is available.
1618
1619 2016-01-04  Martin Sebor  <msebor@redhat.com>
1620
1621         * doc/invoke.texi (Warning Options): Document -Winvalid-memory-model.
1622
1623 2016-01-04  Michael Meissner  <meissner@linux.vnet.ibm.com>
1624
1625         * config/rs6000/rs6000-cpus.def (ISA_3_0_MASKS_SERVER): Add
1626         OPTION_MASK_P9_DFORM.
1627
1628         * config/rs6000/constraints.md (wo constraint): New constraint for
1629         ISA 3.0 (power9).
1630
1631         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Add support
1632         for wo constraint.
1633         (rs6000_init_hard_regno_mode_ok): Likewise.
1634
1635         * config/rs6000/rs6000.h (r6000_reg_class_enum): Add support for
1636         wo constraint.
1637
1638         * config/rs6000/altivec.md (altivec_vperm_<mode>): Clean up vperm
1639         expanders not to have constraints.  Add support for ISA 3.0 xxperm
1640         instruction.  Add support for fusing xxlor with xxperm.
1641         (altivec_vperm_<mode>_internal): Likewise.
1642         (altivec_vperm_v8hiv16qi): Likewise.
1643         (altivec_vperm_<mode>v16q): Likewise.
1644         (altivec_vperm_<mode>_uns): Likewise.
1645         (vperm_v8hiv4si): Likewise.
1646         (vperm_v16qiv8hi): Likewise.
1647
1648         * doc/md.texi (RS/6000 constraints): Document wo constraint.
1649
1650 2016-01-04  Jakub Jelinek  <jakub@redhat.com>
1651
1652         Update copyright years.
1653
1654         * gcc.c (process_command): Update copyright notice dates.
1655         * gcov-dump.c (print_version): Ditto.
1656         * gcov.c (print_version): Ditto.
1657         * gcov-tool.c (print_version): Ditto.
1658         * gengtype.c (create_file): Ditto.
1659         * doc/cpp.texi: Bump @copying's copyright year.
1660         * doc/cppinternals.texi: Ditto.
1661         * doc/gcc.texi: Ditto.
1662         * doc/gccint.texi: Ditto.
1663         * doc/gcov.texi: Ditto.
1664         * doc/install.texi: Ditto.
1665         * doc/invoke.texi: Ditto.
1666
1667 2016-01-04  Eric Botcazou  <ebotcazou@adacore.com>
1668
1669         * config/arm/arm.c (aapcs_vfp_allocate_return_reg): Treat all integer
1670         modes larger than TImode as TImode if NEON is not enabled.
1671
1672 2016-01-04  Eric Botcazou  <ebotcazou@adacore.com>
1673
1674         PR target/69100
1675         * config/sparc/sparc.h (FUNCTION_ARG_REGNO_P): Return true in 64-bit
1676         mode for %f0-%f31 only if TARGET_FPU.
1677
1678 2016-01-04  Eric Botcazou  <ebotcazou@adacore.com>
1679
1680         PR target/69072
1681         * config/sparc/sparc.c (scan_record_type): Take into account subfields
1682         to compute the PACKED_P predicate.
1683         (function_arg_record_value): Minor tweaks.
1684
1685 2016-01-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
1686
1687         * doc/install.texi (--with-multilib-list): Describe the meaning of the
1688         option for arm*-*-* targets.
1689
1690 2016-01-03  Sandra Loosemore  <sandra@codesourcery.com>
1691
1692         * doc/extend.texi (Common Function Attributes): Move docs for
1693         MSP430-specific attributes to....
1694         (MSP430 Function Attributes): ...here.  Delete the redundant
1695         entries and copy-edit the remaining text.
1696         (MSP430 Variable Attributes): Use uniform format for index
1697         entries and add a cross-reference to the corresponding function
1698         attribute docs.
1699
1700 2016-01-03  Vladimír Čunát  <vcunat@gmail.com>
1701
1702         * doc/invoke.texi (RS/6000 and PowerPC Options): Fix
1703         -finite-math typo.
1704         (x86 Options): Likewise.
1705
1706 2016-01-01  Sandra Loosemore  <sandra@codesourcery.com>
1707
1708         PR 1078
1709
1710         * extend.texi (Common Function Attributes) <no_stack_limit>: New.
1711         * invoke.texi (Code Gen Options) <-fno-stack-limit>: Add pointer
1712         to corresponding attribute.
1713
1714 2016-01-01  Sandra Loosemore  <sandra@codesourcery.com>
1715
1716         * doc/extend.texi (Common Function Attributes) <noplt>: Move
1717         to correct alphabetization of table.  Copy-edit and correct
1718         markup.
1719         <stack_protect>: Likewise.
1720         <target_clones>: Likewise.
1721         <simd>: Likewise.
1722         * doc/invoke.texi (Optimize Options) <-fstack-protector-explicit>:
1723         Correct punctuation.
1724         (Code Gen Options) <-fno-plt>: Copy-edit.
1725
1726 2016-01-01  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1727
1728         PR target/68917
1729         * config/tilegx/tilegx.md (clzsi2): Don't create DI subregs of
1730         SI values.  Explicitly convert SI to DI and vice-versa.
1731
1732 2016-01-01  Jakub Jelinek  <jakub@redhat.com>
1733
1734         PR tree-optimization/69070
1735         * tree-ssa-math-opts.c (gimple_expand_builtin_pow): Only test
1736         REAL_VALUE_ISSIGNALING_NAN on arg0 if arg0 is a REAL_CST.
1737
1738         PR sanitizer/69055
1739         * ubsan.c (ubsan_instrument_float_cast): Call
1740         initialize_sanitizer_builtins.
1741
1742         PR target/69015
1743         * ifcvt.c (find_cond_trap): Give up if returnjump_p (jump).
1744 \f
1745 Copyright (C) 2016 Free Software Foundation, Inc.
1746
1747 Copying and distribution of this file, with or without modification,
1748 are permitted in any medium without royalty provided the copyright
1749 notice and this notice are preserved.