1 2016-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4 * config/arm/vfp.md (l<vrint_pattern><su_optab><mode>si2): Set "conds"
5 attribute to unconditional. Remove %? from output template.
7 2015-01-19 Wilco Dijkstra <wdijkstr@arm.com>
8 Jiong Wang <jiong.wang@arm.com>
10 * ccmp.c (expand_ccmp_expr_1): Cost the instruction sequences
11 generated from different expand order.
13 2015-01-19 Wilco Dijkstra <wdijkstr@arm.com>
15 * /config/aarch64/aarch64.c (aarch64_if_then_else_costs):
16 Add support for CCMP costing.
18 2015-01-19 Wilco Dijkstra <wdijkstr@arm.com>
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.
25 * config/aarch64/aarch64.c (aarch64_gen_ccmp_first): Add FP support.
26 (aarch64_gen_ccmp_next): Add FP support.
28 2015-01-19 Wilco Dijkstra <wdijkstr@arm.com>
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
54 * config/aarch64/aarch64-modes.def: Remove CC_DNE/* modes.
55 * config/aarch64/predicates.md (ccmp_cc_register): Remove.
57 2016-01-19 Jan Hubicka <hubicka@ucw.cz>
59 * cgraphunit.c (cgraph_node::reset): Clear thunk info and
62 2016-01-19 Richard Biener <rguenther@suse.de>
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.
69 2016-01-19 Jakub Jelinek <jakub@redhat.com>
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.
82 2016-01-19 Richard Biener <rguenther@suse.de>
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.
89 2016-01-19 Bernd Schmidt <bschmidt@redhat.com>
91 * ira.c (ira): Update regstat data if we deleted insns.
93 2016-01-19 Jakub Jelinek <jakub@redhat.com>
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.
101 2016-01-19 Jan Hubicka <hubicka@ucw.cz>
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
109 2016-01-19 Jan Hubicka <hubicka@ucw.cz>
111 * lto-streamer-out.c (lto_output): Do not stream instrumentation
114 2016-01-19 Jan Hubicka <hubicka@ucw.cz>
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.
119 2016-01-19 Martin Jambor <mjambor@suse.cz>
120 Martin Liska <mliska@suse.cz>
121 Michael Matz <matz@suse.de>
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
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
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
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
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.
248 * toplev.c (compile_file): Call hsa_output_brig.
249 * hsa-regalloc.c: New file.
251 2016-01-18 Jeff Law <law@redhat.com>
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.
259 * tree-ssanames.c (ssa_name_has_boolean_range): Remove unnecessary
262 2016-01-18 Sandra Loosemore <sandra@codesourcery.com>
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
272 (Developer Options): New section incorporating moved options.
273 * doc/cppopts.texi (-dM): Update cross-reference.
275 2016-01-18 Richard Henderson <rth@redhat.com>
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.
290 2016-01-18 Richard Biener <rguenther@suse.de>
292 PR tree-optimization/69297
293 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Count each scalar
295 (vect_bb_vectorization_profitable_p): Clear visited flag again.
297 2016-01-18 Yuri Rumyantsev <ysrumyan@gmail.com>
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
311 2016-01-18 Joseph Myers <joseph@codesourcery.com>
313 * config/mips/mips.h (ISA_HAS_PAIRED_SINGLE): Require
316 2016-01-18 Richard Biener <rguenther@suse.de>
319 * gimple.c (gimple_could_trap_p_1): Handle GIMPLE_COND.
321 2016-01-18 Tom de Vries <tom@codesourcery.com>
323 * passes.def: Add pass_parallelize_loops to pass_oacc_kernels.
325 2016-01-18 Tom de Vries <tom@codesourcery.com>
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):
343 (pass_parallelize_loops::bool oacc_kernels_p): New member var.
344 * passes.def: Add argument to pass_parallelize_loops instantation.
346 2016-01-18 Tom de Vries <tom@codesourcery.com>
348 * tree-parloops.c (pass_parallelize_loops::execute): Allow
349 pass_parallelize_loops to be run outside the loop pipeline.
351 2016-01-18 Alan Lawrence <alan.lawrence@arm.com>
353 * tree-scalar-evolution.c (follow_copies_to_constant): New.
354 (analyze_initial_condition, analyze_scalar_evolution_1): Call previous.
356 2016-01-18 Alan Lawrence <alan.lawrence@arm.com>
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.
364 2016-01-18 Alan Lawrence <alan.lawrence@arm.com>
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.
380 2016-01-18 Ilya Enkovich <enkovich.gnu@gmail.com>
382 * config/i386/i386.c (scalar_to_vector_candidate_p): Support
384 (scalar_chain::convert_op): Likewise.
385 * config/i386/i386.md (*andndi3_doubleword): New.
387 2016-01-18 Richard Biener <rguenther@suse.de>
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.
393 2016-01-18 Jakub Jelinek <jakub@redhat.com>
395 * haifa-sched.c (autopref_multipass_init): Work around
396 -Wmaybe-uninitialized warning.
398 2016-01-18 Thomas Preud'homme <thomas.preudhomme@arm.com>
400 * config/arm/arm.c (thumb1_reorg): Check that the comparison is
401 against the constant 0.
403 2016-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
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.
414 2016-01-17 Jakub Jelinek <jakub@redhat.com>
416 * omp-low.c (mark_loops_in_oacc_kernels_region): Work around
417 -Wmaybe-uninitialized warning.
419 2016-01-16 Sandra Loosemore <sandra@codesourcery.com>
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.
432 2016-01-16 Tom de Vries <tom@codesourcery.com>
434 * passes.def: Move pass_expand_omp_ssa out of pass_parallelize_loops.
436 2016-01-16 Tom de Vries <tom@codesourcery.com>
438 * omp-low.c (expand_omp_atomic_fetch_op): Release defs of update stmt.
440 2016-01-16 Richard Sandiford <richard.sandiford@arm.com>
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.
447 2016-01-15 David Malcolm <dmalcolm@redhat.com>
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.
454 2016-01-15 Jeff Law <law@redhat.com>
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.
464 2016-01-15 Vladimir Makarov <vmakarov@redhat.com>
466 PR rtl-optimization/69030
467 * lra-spills.c (remove_pseudos): Check nrefs and make the function
469 (spill_pseudos): Delete debug insn for dead pseudo.
470 (lra_spill): Initiate spill_hard_reg and slots memory separately.
472 2016-01-15 Jiong Wang <jiong.wang@arm.com>
474 * config/aarch64/aarch64-builtins.c (aarch64_types_unopus_qualifiers):
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
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.
526 2016-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
528 * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Handle
529 CSEL of zero_extended registers.
531 2016-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
533 * config/aarch64/aarch64.c (aarch64_rtx_costs, COMPARE case):
534 Handle COMPARE of ZERO_EXTRACT against zero form of TST-immediate.
536 2016-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
538 * config/aarch64/aarch64.c (aarch64_process_one_target_attr): Return
539 false when argument string is not found in the attributes table
542 2016-01-15 David Edelsohn <dje.gcc@gmail.com>
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
550 2016-01-15 Richard Biener <rguenther@suse.de>
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.
564 2016-01-15 Richard Biener <rguenther@suse.de>
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
573 2016-01-15 Cesar Philippidis <cesar@codesourcery.com>
575 * gimplify.c (oacc_default_clause): Decode reference and pointer
576 types for both kernels and parallel regions.
578 2016-01-15 Richard Sandiford <richard.sandiford@arm.com>
581 * calls.c (emit_call_1): Force n_popped to zero for sibcalls.
583 2016-01-15 Ilya Enkovich <enkovich.gnu@gmail.com>
585 * config/i386/i386.c (scalar_chain::compute_convert_gain): Fix typo.
586 (convert_scalars_to_vector): Likewise.
588 2016-01-15 Jonathan Wakely <jwakely@redhat.com>
590 * doc/extend.texi (Type Traits): Fix grammar.
592 2016-01-15 Martin Jambor <mjambor@suse.cz>
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
603 2016-01-15 Dominik Vogt <vogt@linux.vnet.ibm.com>
605 * config/s390/s390.opt (mmvcle): More verbose help text.
607 2016-01-15 Dominik Vogt <vogt@linux.vnet.ibm.com>
609 * config/s390/s390.opt: Add period to -mzvector option text.
611 2016-01-15 Richard Biener <rguenther@suse.de>
613 PR tree-optimization/68961
614 * tree-vect-slp.c (vect_analyze_slp_cost_1): Consider cost
615 of invariants in stores again.
617 2016-01-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
619 * config/arm/t-rtems: Add cortex-m7/fpv5-d16 multilib.
621 2016-01-15 Ilya Enkovich <enkovich.gnu@gmail.com>
623 * config/i386/i386.c (ix86_expand_branch): Don't split
624 DI mode xor instruction to SI mode.
626 2016-01-15 Jan Hubicka <hubicka@ucw.cz>
629 * ipa-icf.c (sem_function::merge): Virtual functions may become
630 reachable even if they address is not taken and there are no
633 2016-01-15 Jan Hubicka <hubicka@ucw.cz>
635 * lto-streamer-out.c (subtract_estimated_size): New function.
636 (get_symbol_initial_value): Use it.
638 2016-01-15 Christian Bruel <christian.bruel@st.com>
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.
653 2016-01-15 Richard Biener <rguenther@suse.de>
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.
661 2016-01-14 Jeff Law <law@redhat.com>
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.
669 2016-01-14 Richard Henderson <rth@redhat.com>
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.
677 2016-01-14 Michael Meissner <meissner@linux.vnet.ibm.com>
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.
685 2016-01-14 Jeff Law <law@redhat.com>
687 * tree-ssa-dom.c (ssa_name_has_boolean_range): Fix comment
690 2016-01-14 Richard Henderson <rth@redhat.com>
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.
698 2016-01-14 Nicklas Bo Jensen <nbjensen@gmail.com>
700 * doc/loop.texi (Loop Analysis and Representation): Document
703 2016-01-14 Tom de Vries <tom@codesourcery.com>
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.
711 2016-01-14 Jakub Jelinek <jakub@redhat.com>
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.
719 2016-01-14 Alan Lawrence <alan.lawrence@arm.com>
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.
733 2016-01-14 Alan Lawrence <alan.lawrence@arm.com>
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.
742 2016-01-14 Jan Hubicka <hubicka@ucw.cz>
744 * alias.c (compare_base_symbol_refs): New function.
745 (rtx_equal_for_memref_p, base_alias_check, memrefs_conflict_p): Use
748 2016-01-14 Jakub Jelinek <jakub@redhat.com>
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
761 (tree_lower_complex): Walk bbs in rpo order. Adjust phis recorded
762 in phis_to_revisit at the end.
764 2016-01-14 Richard Biener <rguenther@suse.de>
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.
771 2016-01-14 Jakub Jelinek <jakub@redhat.com>
774 * combine.c (expand_field_assignment): Punt if compute_mode is
775 unsupported scalar mode.
777 2016-01-14 Richard Biener <rguenther@suse.de>
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.
784 2016-01-14 Jeff Law <law@redhat.com>
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
791 2016-01-13 Jan Hubicka <hubicka@ucw.cz>
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.
798 2016-01-13 Sandra Loosemore <sandra@codesourcery.com>
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.
805 2016-01-13 Sandra Loosemore <sandra@codesourcery.com>
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.
811 2016-01-13 Jonathan Wakely <jwakely@redhat.com>
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.
817 2016-01-13 Michael Meissner <meissner@linux.vnet.ibm.com>
819 * config/rs6000/rs6000-builtin.def (BU_FLOAT128_2): Add support
820 for pack/unpack functions for __ibm128.
822 (UNPACK_IF): Likewise.
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.
830 * config/rs6000/rs6000.h (MASK_FLOAT128): Add short name.
831 (RS6000_BTM_FLOAT128): Add support for __ibm128 pack/unpack
833 (RS6000_BTM_COMMON): Likewise.
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.
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.
852 * doc/extend.texi (PowerPC Built-in Functions): Document
853 __builtin_pack_ibm128 and __builtin_unpack_ibm128.
855 2016-01-13 Bernd Schmidt <bschmidt@redhat.com>
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.
866 2016-01-13 Jakub Jelinek <jakub@redhat.com>
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.
875 2016-01-13 Jonathan Wakely <jwakely@redhat.com>
877 * doc/extend.texi (__atomic Builtins): Clarify compare_exchange
880 2016-01-13 Richard Henderson <rth@redhat.com>
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.
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
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.
906 2016-01-13 Uros Bizjak <ubizjak@gmail.com>
908 * config/i386/i386.h (TARGET_FLT_EVAL_METHOD): Return -1 for
909 TARGET_SSE_MATH without TARGET_SSE2. Rewrite.
911 2016-01-13 Tom de Vries <tom@codesourcery.com>
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.
919 2016-01-13 Yvan Roux <yvan.roux@linaro.org>
921 * config/arm/arm-arches.def: Remove spurious whitespace in "armv8.1-a"
922 and "armv8.1-a+crc" entries.
924 2016-01-13 Alexander Fomin <alexander.fomin@intel.com>
927 * config/i386/sse.md (define_expand "avx512pf_gatherpf<mode>sf"):
928 Change first operand predicate from register_or_constm1_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.
940 2016-01-13 Richard Biener <rguenther@suse.de>
942 PR tree-optimization/69013
943 * tree-ssa-uninit.c (prune_uninit_phi_opnds_in_unrealizable_paths):
944 Exchange assert for a test.
946 2016-01-13 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
949 * config/s390/s390.md (bswaphi splitter): Use simplify_gen_subreg.
951 2016-01-13 Richard Biener <rguenther@suse.de>
953 PR tree-optimization/69242
954 * tree-ssa-sccvn.c (visit_reference_op_load): Replace bogus
957 2016-01-13 Richard Biener <rguenther@suse.de>
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.
963 2016-01-12 Jeff Law <law@redhat.com>
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
974 2016-01-12 Sandra Loosemore <sandra@codesourcery.com>
976 * doc/invoke.texi (Spec Files): Move section down in file, past
977 all command-line option descriptions.
979 2016-01-12 Trevor Saunders <tbsaunde+gcc@tbsaunde.org>
982 * doc/gty.texi: Remove documentation of mark_hook.
983 * gengtype.c (struct write_types_data): Remove code to support
985 (walk_type): Likewise.
986 (write_func_for_structure): Likewise.
988 2016-01-12 Sandra Loosemore <sandra@codesourcery.com>
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.
996 2016-01-12 Jeff Law <law@redhat.com>
998 * tree-ssa-threadupdate.c: Various whitespace and typo fixes.
1000 2016-01-12 Olivier Hainque <hainque@adacore.com>
1002 * gcc.c (spec_undefvar_allowed): New global.
1003 (process_command): Set to true when running for --version or --help,
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
1009 2016-01-12 Bin Cheng <bin.cheng@arm.com>
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".
1015 2016-01-12 Sandra Loosemore <sandra@codesourcery.com>
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.
1022 2016-01-12 Christian Bruel <christian.bruel@st.com>
1025 * config/arm/arm-c.c (arm_pragma_target_parse): Set NODE_CONDITIONAL
1026 for __ARM_NEON_FP, __ARM_FP, _ARM_FEATURE_LDREX.
1028 2016-01-12 Jakub Jelinek <jakub@redhat.com>
1031 * config/i386/i386.c (ix86_expand_special_args_builtin): Ensure
1032 aligned_mem is properly set for AVX512-VL floating point masked
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.
1039 2016-01-12 H.J. Lu <hongjiu.lu@intel.com>
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.
1049 2016-01-12 Richard Biener <rguenther@suse.de>
1051 PR tree-optimization/69053
1052 * tree-vect-loop.c (get_initial_def_for_reduction): Properly
1053 convert initial value for cond reductions.
1055 2016-01-12 Richard Biener <rguenther@suse.de>
1057 PR tree-optimization/69007
1058 * tree-vect-patterns.c (vect_vect_recog_func_ptrs): Move
1059 widen_sum after dot_prod and sad.
1061 2016-01-12 Richard Biener <rguenther@suse.de>
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.
1069 2016-01-12 Richard Biener <rguenther@suse.de>
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.
1083 2016-01-12 Richard Biener <rguenther@suse.de>
1085 PR tree-optimization/69174
1086 * tree-vect-stmts.c (vect_mark_relevant): Remove excessive vertical
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.
1092 2016-01-12 Andris Pavenis <andris.pavenis@iki.fi>
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
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.
1123 * config/i386/djgpp.c: New file. Add implementation of i386_djgpp_asm_named_section.
1125 * config/i386/djgpp.opt: Remove obsolete option -mbnu210.
1127 * config/i386/t-djgpp: New file. Add djgpp.o to EXTRA_OBJS.
1128 Add rule for building djgpp.o.
1130 2016-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
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.
1136 2016-01-11 John David Anglin <danglin@gcc.gnu.org>
1138 * config/pa/pa.c (pa_emit_move_sequence): Handle floating point
1139 reloads for other unsupported memory operands.
1141 2016-01-12 Kugan Vivekanandarajah <kuganv@linaro.org>
1142 Jim Wilson <jim.wilson@linaro.org>
1145 * config/arm/arm-builtins.c (arm_expand_neon_args): Call
1146 copy_to_mode_reg instead of force_reg.
1148 2016-01-11 H.J. Lu <hongjiu.lu@intel.com>
1151 * config/i386/i386.h (TARGET_FLT_EVAL_METHOD): Set to 2 only if
1152 TARGET_80387 is true.
1154 2016-01-11 Jakub Jelinek <jakub@redhat.com>
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.
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.
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.
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.
1179 2016-01-11 Richard Biener <rguenther@suse.de>
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.
1185 2016-01-11 Uros Bizjak <ubizjak@gmail.com>
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.
1193 2016-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
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.
1201 2016-01-11 H.J. Lu <hongjiu.lu@intel.com>
1203 * config/i386/sse.md (<avx512>_load<mode>_mask): Remove snprintf.
1204 (<avx512>_store<mode>_mask): Likewise.
1206 2016-01-11 Bernd Schmidt <bschmidt@redhat.com>
1207 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
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.
1218 2016-01-11 Tom de Vries <tom@codesourcery.com>
1220 PR tree-optimization/69069
1221 * tree-parloops.c (create_parallel_loop): Add missing phi args.
1223 2016-01-11 Yuri Rumyantsev <ysrumyan@gmail.com>
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
1234 * doc/invoke.texi (max-rtl-if-conversion-insns): Document it.
1235 * ifcvt.c (bb_ok_for_noce_convert_multiple_sets): Limit number of
1238 2016-01-11 Alexandre Oliva <aoliva@redhat.com>
1241 * var-tracking.c (drop_overlapping_mem_locs): Operate on all
1242 onepart vars. Fix typo in comment. Fix reversed condition in
1244 (dataflow_set_remove_mem_locs): Operate on all onepart vars.
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.
1252 2016-01-11 Ilya Enkovich <enkovich.gnu@gmail.com>
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.
1261 2016-01-11 Martin Jambor <mjambor@suse.cz>
1264 * ipa-cp.c (estimate_local_effects): Do not clone for removal of
1265 useless parameters if we cannot change function signature.
1267 2016-01-11 Martin Jambor <mjambor@suse.cz>
1270 * cgraphclones.c (duplicate_thunk_for_node): Copy can_change_signature
1273 2016-01-11 Tom de Vries <tom@codesourcery.com>
1275 PR tree-optimization/69109
1276 * tree-parloops.c (try_transform_to_exit_first_loop_alt): Don't allow
1279 2016-01-11 Tom de Vries <tom@codesourcery.com>
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.
1285 2016-01-11 Yury Gribov <y.gribov@samsung.com>
1288 * common.opt (frandom-seed): Fix parameter name.
1289 * doc/invoke.texi (frandom-seed): Ditto and describe parameter.
1291 2016-01-11 Tom de Vries <tom@codesourcery.com>
1293 PR tree-optimization/69058
1294 * tree-parloops.c (pass_parallelize_loops::execute): Return 0 if libgomp
1297 2016-01-11 Andrew Burgess <andrew.burgess@embecosm.com>
1299 * config/arc/arc.opt (mdiv-rem): Add period to the end.
1300 (mcode-density): Likewise.
1302 2016-01-10 Tom de Vries <tom@codesourcery.com>
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
1309 2016-01-10 Tom de Vries <tom@codesourcery.com>
1311 PR tree-optimization/69039
1312 * tree-parloops.c (try_create_reduction_list): Only allow single exit
1315 2016-01-09 John David Anglin <danglin@gcc.gnu.org>
1318 * match.pd: Require target has function_c99_misc before doing
1319 truncl(extend(x)) and trunc(extend(x)) -> extend(truncf(x)), etc.
1321 2016-01-09 Gerald Pfeifer <gerald@pfeifer.com>
1323 * configure.ac (isl_options_set_schedule_serialize_sccs): Also
1325 * configure: Regenerate.
1327 2016-01-09 Jakub Jelinek <jakub@redhat.com>
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.
1343 2016-01-09 Jan Hubicka <hubicka@ucw.cz>
1345 * ipa-icf.c (sem_item_optimizer::merge_classes): Do not ICE on VAR_DECL
1348 2016-01-08 Jakub Jelinek <jakub@redhat.com>
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
1354 * gimple-match-head.c (maybe_push_res_to_seq): Likewise.
1356 2016-01-08 Alan Lawrence <alan.lawrence@arm.com>
1357 Richard Biener <rguenther@suse.de>
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.
1363 2016-01-08 Uros Bizjak <ubizjak@gmail.com>
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.
1370 2016-01-08 Jason Merrill <jason@redhat.com>
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.
1380 2016-01-08 Olivier Hainque <hainque@adacore.com>
1382 * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Don't link shared RTPs with
1385 2016-01-08 Alan Lawrence <alan.lawrence@arm.com>
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.
1394 2016-01-08 Jakub Jelinek <jakub@redhat.com>
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.
1404 PR tree-optimization/69172
1405 * gimple-fold.c (gimple_fold_builtin_memory_chk): Pass type to
1408 2016-01-08 Thomas Preud'homme <thomas.preudhomme@arm.com>
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.
1416 2016-01-08 Jakub Jelinek <jakub@redhat.com>
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.
1424 2016-01-08 Segher Boessenkool <segher@kernel.crashing.org>
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.
1433 2016-01-07 Martin Sebor <msebor@redhat.com>
1436 * doc/extend.texi (__atomic Builtins, __sync Builtins): Document
1437 constraint on the type of arguments.
1439 2016-01-07 Andreas Tobler <andreast@gcc.gnu.org>
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.
1447 2016-01-07 Uros Bizjak <ubizjak@gmail.com>
1450 * config/i386/i386.c (ix86_frame_pointer_required): Enable
1451 frame pointer for TARGET_64BIT_MS_ABI when stack is misaligned.
1453 2016-01-07 Uros Bizjak <ubizjak@gmail.com>
1456 2016-01-06 Uros Bizjak <ubizjak@gmail.com>
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.
1463 2016-01-07 Sandra Loosemore <sandra@codesourcery.com>
1466 * doc/extend.texi (Nvidia PDX Function Attributes): New section.
1468 2016-01-07 H.J. Lu <hongjiu.lu@intel.com>
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):
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
1485 2015-12-15 Bernd Schmidt <bschmidt@redhat.com>
1488 * varasm.c (make_decl_rtl): Mark invalid register vars as
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.
1495 2016-01-07 Jakub Jelinek <jakub@redhat.com>
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 ().
1504 * gimple-expr.c (copy_var_decl): If var has DECL_USER_ALIGN set, copy
1505 it and DECL_ALIGN too.
1507 2016-01-06 Robert Suchanek <robert.suchanek@imgtec.com>
1509 * config/mips/mips-ftypes.def: Sort to lexicographical order.
1511 2016-01-06 Uros Bizjak <ubizjak@gmail.com>
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.
1518 2015-01-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
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.
1530 2016-01-06 Marek Polacek <polacek@redhat.com>
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.
1539 2016-01-05 Sandra Loosemore <sandra@codesourcery.com>
1542 * doc/extend.texi (RL78 Variable Attributes): New section.
1544 2016-01-05 Marek Polacek <polacek@redhat.com>
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.
1554 2016-01-05 H.J. Lu <hongjiu.lu@intel.com>
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
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.
1570 2016-01-05 H.J. Lu <hongjiu.lu@intel.com>
1573 * config/i386/constraints.md (Bm): New constraint.
1574 * config/i386/predicates.md (vector_memory_operand): New
1576 * config/i386/sse.md: Replace xm with xBm in plusminus and
1579 2016-01-05 Sandra Loosemore <sandra@codesourcery.com>
1582 * doc/extend.texi (V850 Function Attributes): New section.
1583 (V850 Variable Attributes): New section.
1585 2016-01-05 Sandra Loosemore <sandra@codesourcery.com>
1588 * doc/extend.texi (MicroBlaze Function Attributes): Document
1589 interrupt_handler and fast_interrupt attributes.
1591 2016-01-05 Sergei Trofimovich <siarheit@google.com>
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
1599 2016-01-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1601 PR rtl-optimization/68651
1602 * combine.c (combine_simplify_rtx): Canonicalize x + x into
1605 2016-01-05 Nathan Sidwell <nathan@acm.org>
1607 * alias.c (compare_base_decls): Use symtab_node::get.
1609 2016-01-05 Nick Clifton <nickc@redhat.com>
1612 * ira-costs.c (copy_cost): Initialise the t_icode field of the
1613 secondary_reload_info structure.
1616 * config/i386/cygming.h (MAKE_DECL_ONE_ONLY): Define to use weak
1617 decls if weak support is available.
1619 2016-01-04 Martin Sebor <msebor@redhat.com>
1621 * doc/invoke.texi (Warning Options): Document -Winvalid-memory-model.
1623 2016-01-04 Michael Meissner <meissner@linux.vnet.ibm.com>
1625 * config/rs6000/rs6000-cpus.def (ISA_3_0_MASKS_SERVER): Add
1626 OPTION_MASK_P9_DFORM.
1628 * config/rs6000/constraints.md (wo constraint): New constraint for
1631 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Add support
1633 (rs6000_init_hard_regno_mode_ok): Likewise.
1635 * config/rs6000/rs6000.h (r6000_reg_class_enum): Add support for
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.
1648 * doc/md.texi (RS/6000 constraints): Document wo constraint.
1650 2016-01-04 Jakub Jelinek <jakub@redhat.com>
1652 Update copyright years.
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.
1667 2016-01-04 Eric Botcazou <ebotcazou@adacore.com>
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.
1672 2016-01-04 Eric Botcazou <ebotcazou@adacore.com>
1675 * config/sparc/sparc.h (FUNCTION_ARG_REGNO_P): Return true in 64-bit
1676 mode for %f0-%f31 only if TARGET_FPU.
1678 2016-01-04 Eric Botcazou <ebotcazou@adacore.com>
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.
1685 2016-01-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
1687 * doc/install.texi (--with-multilib-list): Describe the meaning of the
1688 option for arm*-*-* targets.
1690 2016-01-03 Sandra Loosemore <sandra@codesourcery.com>
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
1700 2016-01-03 Vladimír Čunát <vcunat@gmail.com>
1702 * doc/invoke.texi (RS/6000 and PowerPC Options): Fix
1704 (x86 Options): Likewise.
1706 2016-01-01 Sandra Loosemore <sandra@codesourcery.com>
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.
1714 2016-01-01 Sandra Loosemore <sandra@codesourcery.com>
1716 * doc/extend.texi (Common Function Attributes) <noplt>: Move
1717 to correct alphabetization of table. Copy-edit and correct
1719 <stack_protect>: Likewise.
1720 <target_clones>: Likewise.
1722 * doc/invoke.texi (Optimize Options) <-fstack-protector-explicit>:
1723 Correct punctuation.
1724 (Code Gen Options) <-fno-plt>: Copy-edit.
1726 2016-01-01 Bernd Edlinger <bernd.edlinger@hotmail.de>
1729 * config/tilegx/tilegx.md (clzsi2): Don't create DI subregs of
1730 SI values. Explicitly convert SI to DI and vice-versa.
1732 2016-01-01 Jakub Jelinek <jakub@redhat.com>
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.
1739 * ubsan.c (ubsan_instrument_float_cast): Call
1740 initialize_sanitizer_builtins.
1743 * ifcvt.c (find_cond_trap): Give up if returnjump_p (jump).
1745 Copyright (C) 2016 Free Software Foundation, Inc.
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.