1 2011-12-06 Georg-Johann Lay <avr@gjlay.de>
3 Forward-port from gcc-4_6-branch r181936 2011-12-02.
4 * config/avr/avr-mcus.def (avr_mcu_types): ATtiny4313 and
5 AT86RF401 have a 16-bit SP (their manual is bogus).
7 2011-12-06 Georg-Johann Lay <avr@gjlay.de>
9 Forward-port from gcc-4_6-branch r181936 2011-12-02.
12 * config/avr/avr.md (movhi_sp_r): Set insn condition to
14 * config/avr/avr.c (output_movhi): Use "clr%B0" instead of "in
15 %B0,__SP_H__" if AVR_HAVE_8BIT_SP.
16 (avr_file_start): Only print "__SP_H__ = 0x3e" if !AVR_HAVE_8BIT_SP.
18 2011-12-06 Georg-Johann Lay <avr@gjlay.de>
22 * config/avr/avr.h (ADDR_SPACE_PGM, ADDR_SPACE_PGM1,
23 ADDR_SPACE_PGM2, ADDR_SPACE_PGM3, ADDR_SPACE_PGM4,
24 ADDR_SPACE_PGM5, ADDR_SPACE_PGMX): Write as enum.
25 (avr_addrspace_t): New typedef.
26 (avr_addrspace): New declaration.
27 * config/avr/avr-c.c (avr_toupper): New static function.
28 (avr_register_target_pragmas, avr_cpu_cpp_builtins): Use
29 avr_addrspace to get address space information.
30 * config/avr/avr.c (avr_addrspace): New variable.
31 (avr_out_lpm, avr_pgm_check_var_decl, avr_insert_attributes,
32 avr_asm_named_section, avr_section_type_flags,
33 avr_asm_select_section, avr_addr_space_address_mode,
34 avr_addr_space_convert, avr_emit_movmemhi): Use it.
35 (avr_addr_space_pointer_mode): Forward to avr_addr_space_address_mode.
36 (avr_pgm_segment): Remove.
38 2011-12-06 Richard Guenther <rguenther@suse.de>
41 * varasm.c (assemble_alias): Move DECL_EXTERNAL implementation
43 * passes.c (rest_of_decl_compilation): ... here.
45 2011-12-06 Richard Guenther <rguenther@suse.de>
48 * gimple-fold.c (gimplify_and_update_call_from_tree): Guard
49 vdef check for the fact we do not have virtual operands when
52 2011-12-06 Richard Guenther <rguenther@suse.de>
54 PR tree-optimization/51245
55 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_constant_for_pieces):
57 (vn_reference_lookup_3): Use it. Properly valueize all refs
60 2011-12-06 Iain Sandoe <iains@gcc.gnu.org>
62 * config/rs6000/rs6000.c (darwin_rs6000_override_options): Initialize
65 2011-12-06 Richard Guenther <rguenther@suse.de>
67 PR tree-optimization/51363
68 * gimple.c (gimple_rhs_has_side_effects): Remove.
69 * gimple.h (gimple_rhs_has_side_effects): Likewise.
70 * tree-ssa-dom.c (optimize_stmt): Simplify conditional.
72 2011-12-06 Alan Modra <amodra@gmail.com>
75 * config/rs6000/rs6000.c (rs6000_emit_prologue <TARGET_SPE_ABI>):
76 Do not mark r11 setup as frame-related. Pass correct offset to
77 rs6000_emit_savres_rtx. Correct out-of-line rs6000_frame_related
78 arguments. Correct sp_offset. Remove "offset" fudge from
79 in-line rs6000_frame_related call. Rename misleading variable.
80 Fix comments and whitespace. Tidy some expressions.
81 (rs6000_emit_epilogue <TARGET_SPE_ABI>): Always set frame_reg_rtx
82 to r11 in out-of-line case. Correct sp_offset. Pass correct
83 offset to rs6000_emit_savres_rtx. Rename misleading variable.
84 Fix comments and whitespace. Tidy some expressions.
85 (rs6000_emit_epilogue <non-TARGET_SPE_ABI>): Add sp_offset
86 adjustment when !saving_GPRs_inline. Correct register mode
87 used in address calcs.
88 (rs6000_emit_epilogue <non-TARGET_SPE_ABI>): Similarly when
89 !restoring_GPRs_inline.
91 2011-12-06 Ramana Radhakrishnan <ramana.radhakrishnan@linaro.org>
93 * config/arm/vfp.md (*combine_vcvt_f64_<FCVTI32typename>): Fix
94 formatting character for vmov.f64 case.
96 2011-12-05 Jakub Jelinek <jakub@redhat.com>
98 PR tree-optimization/51396
99 * tree-ssa-math-opts.c (convert_mult_to_fma): Don't optimize
100 if MUL_RESULT has zero uses.
103 * c-decl.c (pop_scope): Don't add DECL_EXTERNAL decls
104 for debug info if scope is file_scope.
107 * c-decl.c (c_finish_incomplete_decl, finish_decl): Call
108 relayout_decl instead of layout_decl.
110 2011-12-05 Richard Sandiford <rdsandiford@googlemail.com>
112 * resource.c (init_resource_info): Only consider EXIT_IGNORE_STACK
113 if there is in epilogue.
115 2011-12-05 Richard Sandiford <rdsandiford@googlemail.com>
117 * optabs.c (maybe_emit_sync_lock_test_and_set): Pass a null target
118 to emit_library_call_value.
119 (expand_atomic_compare_and_swap): Likewise.
121 2011-12-05 Vladimir Makarov <vmakarov@redhat.com>
124 * ira-int.h (struct ira_object): Remove add_data.
125 (OBJECT_ADD_DATA): Remove.
127 * ira-build.c (ira_create_object): Remove OBJECT_ADD_DATA
130 * ira-color.c (object_hard_regs_t, object_hard_regs): Rename to
131 allocno_hard_regs_t, allocno_hard_regs.
132 (object_hard_regs_node_t, object_hard_regs_node): Rename to
133 allocno_hard_regs_node_t and allocno_hard_regs_node.
134 (struct allocno_color_data): Add new member last_process. Move
135 profitable_hard_regs, hard_regs_node, and hard_regs_subnodes_start
136 from object_color_data.
137 (object_color_data_t, object_color_data, OBJECT_COLOR_DATA):
139 (curr_allocno_process): New static variable.
140 (object_hard_regs_eq, object_hard_regs_htab): Rename to
141 allocno_hard_regs_eq and allocno_hard_regs_htab.
142 (init_object_hard_regs, finish_object_hard_regs): Rename to
143 init_allocno_hard_regs and finish_allocno_hard_regs.
144 (object_hard_regs_compare, object_hard_regs_node_t): Rename to
145 allocno_hard_regs_compare and allocno_hard_regs_node_t.
146 (create_new_object_hard_regs_node): Rename to
147 create_new_allocno_hard_regs_node.
148 (add_new_object_hard_regs_node_to_forest): Rename to
149 add_new_allocno_hard_regs_node_to_forest.
150 (add_object_hard_regs_to_forest, collect_object_hard_regs_cover):
151 Rename to add_allocno_hard_regs_to_forest and collect_allocno_hard_regs_cover.
152 (setup_object_hard_regs_nodes_parent): Rename to setup_allocno_hard_regs_nodes_parent.
153 (remove_unused_object_hard_regs_nodes): Rename to remove_unused_allocno_hard_regs_nodes.
154 (enumerate_object_hard_regs_nodes, object_hard_regs_nodes_num):
155 Rename to enumerate_allocno_hard_regs_nodes and allocno_hard_regs_nodes_num.
156 (object_hard_regs_nodes, object_hard_regs_subnode_t): Rename to
157 allocno_hard_regs_nodes and allocno_hard_regs_subnode_t.
158 (object_hard_regs_subnode, object_hard_regs_subnodes): Rename to
159 allocno_hard_regs_subnode and allocno_hard_regs_subnodes.
160 (object_hard_regs_subnode_index): Rename to allocno_hard_regs_subnode_index.
161 (setup_object_hard_regs_subnode_index): Rename to
162 setup_allocno_hard_regs_subnode_index.
163 (get_object_hard_regs_subnodes_num): Rename to get_allocno_hard_regs_subnodes_num.
164 (form_object_hard_regs_nodes_forest): Rename to form_allocno_hard_regs_nodes_forest.
165 (finish_object_hard_regs_nodes_tree): Rename to form_allocno_hard_regs_nodes_forest
166 (finish_object_hard_regs_nodes_forest): Rename to
167 finish_allocno_hard_regs_nodes_forest.
168 (setup_left_conflict_sizes_p): Use allocno data instead of object
169 ones. Process conflict allocnos once.
170 (update_left_conflict_sizes_p): Use allocno data instead of object
171 ones. Change prototype signature.
172 (empty_profitable_hard_regs): Use allocno data instead of object
174 (setup_profitable_hard_regs): Ditto.
175 (get_conflict_profitable_regs): Rename to
176 get_conflict_and_start_profitable_regs. Use allocno data for
177 profitable regs calculation.
178 (check_hard_reg_p): Change prototype signature. Check profitable
179 regs for allocno not the objects.
180 (assign_hard_reg): Process conflict allocnos only once for
181 updating conflict costs.
182 (setup_allocno_available_regs_num): Use allocno data instead of
183 object ones. Modify debug output.
184 (color_pass): Remove initialization and finalization of object
187 2011-12-05 Kazu Hirata <kazu@codesourcery.com>
190 * config/arm/arm.md (*minmax_arithsi): Always require the else
191 clause in the MINUS case.
193 2011-12-05 Richard Guenther <rguenther@suse.de>
195 PR tree-optimization/50904
196 * tree-ssa-loop-im.c (struct mem_ref): Remove vops member.
197 (MEM_ANALYZABLE): New.
198 (memory_references): Remove clobbered_vops and vop_ref_map
199 members, add all_refs_stored_in_loop member.
200 (memref_free): Adjust.
201 (mem_ref_alloc): Likewise.
202 (gather_mem_refs_stmt): Do not record clobbers, instead
203 record refs for unanalyzable stmts.
204 (gather_mem_refs_in_loops): Do not propagate clobbers.
205 (struct vop_to_refs_elt, vtoe_hash, vtoe_eq, vtoe_free,
206 record_vop_access, get_vop_accesses, get_vop_stores,
207 add_vop_ref_mapping): Remove.
208 (create_vop_ref_mapping_loop): Adjust to simply record all
210 (analyze_memory_references): Adjust.
211 (refs_independent_p): Check for not analyzable refs.
212 (can_sm_ref_p): Likewise.
213 (ref_indep_loop_p_1): Simplify.
214 (tree_ssa_lim_finalize): Adjust.
216 * tree-ssa-loop-im.c (stmt_cost): Simplify, use LIM_EXPENSIVE
217 rather than magic constants. Assign zero cost to PAREN_EXPR
218 and SSA_NAME copies. Assign cost proportional to the vector
219 size for vector constructors.
221 2011-12-05 Richard Guenther <rguenther@suse.de>
223 * tree-ssa-alias.h (struct ao_ref_s): Add volatile_p field.
224 * tree-ssa-alias.c (ao_ref_init): Initialize it.
225 (ao_ref_init_from_ptr_and_size): Likewise.
226 (refs_may_alias_p_1): Two volatile accesses conflict.
227 (ref_maybe_used_by_call_p_1): Likewise.
228 (call_may_clobber_ref_p_1): Likewise.
229 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Initialize
232 2011-12-05 Richard Guenther <rguenther@suse.de>
234 * tree-ssa.c (verify_ssa): Verify SSA names in the loop
235 over all SSA names. Remove SSA operand checking, call
237 * tree-ssa-operands.h (verify_ssa_operands): Declare.
238 * tree-ssa-operands.c (verify_ssa_operands): New function.
240 2011-12-05 Ramana Radhakrishnan <ramana.radhakrishnan@linaro.org>
242 * config/arm/arm.c (vfp3_const_double_for_fract_bits): Define.
243 * config/arm/arm-protos.h (vfp3_const_double_for_fract_bits): Declare.
244 * config/arm/constraints.md ("Dt"): New constraint.
245 * config/arm/predicates.md
246 (const_double_vcvt_power_of_two_reciprocal): New.
247 * config/arm/vfp.md (FCVTI32typename): New.
248 (FCVT): New iterator.
249 (*combine_vcvt_f32_<FCVTI32typename>): New.
250 (*combine_vcvt_f64_<FCVTI32typename>): New.
252 2011-12-05 Richard Guenther <rguenther@suse.de>
254 * cgraph.c (cgraph_create_edge_1): Initialize
255 call_stmt_cannot_inline_p from the stmt if possible.
256 (cgraph_make_edge_direct): Likewise.
257 * gimple-streamer-in.c (input_gimple_stmt): Do not
258 call gimple_call_set_cannot_inline.
259 * gimple.h (enum gf_mask): Remove GF_CALL_CANNOT_INLINE, shift
261 (gimple_call_set_cannot_inline): Remove.
262 (gimple_call_cannot_inline_p): Likewise.
263 * ipa-inline-analysis.c (initialize_inline_failed): Look
264 at the edge call_stmt_cannot_inline_p flag.
265 * ipa-inline.c (can_inline_edge_p): Likewise.
266 (early_inliner): Only update the edge flag.
267 * ipa-prop.c (update_indirect_edges_after_inlining): Likewise.
268 (ipa_modify_call_arguments): Do not call gimple_call_set_cannot_inline.
269 * cgraphunit.c (assemble_thunk): Likewise.
270 * gimple-fold.c (gimple_fold_call): Likewise.
272 * tree.h (CALL_CANNOT_INLINE_P): Remove.
273 * tree-mudflap.c (mf_xform_statements): Do not modify alloca calls.
274 * builtins.c (expand_builtin_alloca): With -fmudflap do not expand
276 * cfgexpand.c (expand_call_stmt): Do not set CALL_CANNOT_INLINE_P.
277 * gimple.c (gimple_build_call_from_tree): Do not read
278 CALL_CANNOT_INLINE_P.
279 * gimplify.c (gimplify_call_expr): Do not copy CALL_CANNOT_INLINE_P.
281 2011-12-05 Jakub Jelinek <jakub@redhat.com>
282 Eric Botcazou <ebotcazou@adacore.com>
286 * calls.c (internal_arg_pointer_exp_state): New variable.
287 (internal_arg_pointer_based_exp_1,
288 internal_arg_pointer_exp_scan): New functions.
289 (internal_arg_pointer_based_exp): New function.
290 (mem_overlaps_already_clobbered_arg_p): Use it.
291 (expand_call): Free internal_arg_pointer_exp_state.cache vector
292 and clear internal_arg_pointer_exp_state.scan_start.
294 2011-12-04 Kaz Kojima <kkojima@gcc.gnu.org>
296 * config/sh/linux.h (TARGET_DEFAULT): Add MASK_SOFT_ATOMIC.
297 * config/sh/sync.md: New file.
298 * config/sh/sh.md: Include sync.md.
299 * config/sh/sh.opt (msoft-atomic): New option.
300 * doc/invoke.texi (SH Options): Document it.
302 2011-12-04 Nathan Sidwell <nathan@acm.org>
304 * gcov-io.h (struct gcov_info): Replace trailing array with
306 * profile.c (branch_prob): Only call renamed
307 coverage_begin_function once.
308 * coverage.h (coverage_begin_output): Rename to ...
309 (coverage_begin_function): ... here.
310 * coverage.c (struct function_list): Rename to ...
311 (struct coverage_data): ... this. Update all uses.
312 (gcov_info_var, gcov_fn_info_type, gcov_fn_info_ptr_type): New
314 (bbg_file_opened, bbg_function_announced): Remove.
315 (get_coverage_counts): Adjust message.
316 (coverage_begin_ouput): Rename to ...
317 (coverage_begin_function): ... here. Move file opening to
318 coverage_init. Adjust for being called only once.
319 (coverage_end_function): Remove bbg file and inhibit further
320 output here on error.
321 (build_info_type): Adjust for change to pointer to array.
322 (build_info): Receive array of function pointers and adjust.
323 (create_coverage): Break into ...
324 (coverage_obj_init, coverage_obj_fn, coverage_obj_finish):
325 ... these, and adjust.
326 (coverage_init): Open the notes file here. Tidy.
327 (coverage_finish): Call coverage_obj_init etc.
329 2011-12-04 Ira Rosen <ira.rosen@linaro.org>
332 * tree-vect-loop.c (vect_create_epilog_for_reduction): Create exit
333 phi nodes for outer loop in case of double reduction.
335 2011-12-04 Jérémie Detrey <Jeremie.Detrey@loria.fr>
338 * config/i386/avxintrin.h (_mm256_insert_epi64): Declare second
339 parameter as long long.
341 2011-12-04 Richard Sandiford <rdsandiford@googlemail.com>
344 * optabs.c (init_sync_libfuncs): Use "or" rather than "ior"
345 in the external names.
347 2011-12-03 Jack Howarth <howarth@bromo.med.uc.edu>
349 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Pass -no_pie for
350 non-PIC code when targeting 10.7 or later.
352 2011-12-03 Iain Sandoe <iains@gcc.gnu.org>
354 * config/darwin.h (STARTFILE_SPEC): Do not use -lbundle1.o when
356 (DARWIN_DYLIB1_SPEC): Do not use -ldylib1.10.5.o when Darwin >= 10.
357 (DARWIN_CRT1_SPEC): Use -lcrt1.10.6.o when Darwin >= 10.
359 2011-12-03 Jakub Jelinek <jakub@redhat.com>
361 * fold-const.c (fold_unary_loc): Fold VEC_UNPACK_LO_EXPR,
362 VEC_UNPACK_HI_EXPR, VEC_UNPACK_FLOAT_LO_EXPR and
363 VEC_UNPACK_FLOAT_HI_EXPR with VECTOR_CST argument.
364 (fold_binary_loc): Fold VEC_PACK_TRUNC_EXPR,
365 VEC_PACK_FIX_TRUNC_EXPR, VEC_WIDEN_MULT_LO_EXPR
366 and VEC_WIDEN_MULT_HI_EXPR with VECTOR_CST arguments.
369 * tree-ssa.c (target_for_debug_bind): Also allow is_gimple_reg_type
370 vars that aren't referenced.
371 (tree-ssa-live.c (remove_unused_locals): Don't clear TREE_ADDRESSABLE
372 of unreferenced local vars.
373 * cfgexpand.c (expand_debug_expr): For DEBUG_IMPLICIT_PTR allow also
374 TREE_ADDRESSABLE vars that satisfy target_for_debug_bind.
376 2011-12-03 Anatoly Sokolov <aesok@post.ru>
378 * config/arm/arm.h (LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P): Remove.
379 * config/arm/arm-protos.h (aapcs_libcall_value): Remove.
380 * config/arm/arm.c (TARGET_FUNCTION_VALUE_REGNO_P): Define.
381 (arm_libcall_value_1, arm_function_value_regno_p): New function.
382 (arm_function_value, arm_libcall_value): Use arm_libcall_value_1.
383 (aapcs_libcall_value): Make static.
384 (arm_libcall_value): Add static qualifier.
386 2011-12-02 Jakub Jelinek <jakub@redhat.com>
389 * config/i386/sse.md (mul<mode>3 with VI1_AVX2 iterator): For
390 V32QImode use { 0,2,..,14,32,34,..,46,16,18,..,30,48,50,..,62 }
391 permutation instead of extract even permutation.
393 2011-12-02 Nathan Sidwell <nathan@acm.org>
395 * gcov.c (struct arc_info): Add is_throw field.
396 (struct (block_info): Add exceptional field, reduce flags size to
398 (struct function_info): Add has_catch field.
399 (struct line_info): Add unexceptional field.
400 (process_file): Call find_exception_blocks if necessary.
401 (read_graph_file): Adjust. Note if an exceptional edge is seen.
402 (find_exception_blocks): New.
403 (add_line_counts): Set line's unexceptional flag if not
405 (output_branch_count): Note exceptional arcs, lines and blocks.
406 * gcov-dump.c (tag_arcs): Decode arc flags.
407 * doc/gcov.texi: Document '=====' lines.
409 2011-12-02 Anatoly Sokolov <aesok@post.ru>
411 * config/ia64/ia64.h (PRINT_OPERAND, PRINT_OPERAND_ADDRESS,
412 PRINT_OPERAND_PUNCT_VALID_P): Remove macros.
413 * config/ia64/ia64-protos.h (ia64_print_operand,
414 ia64_print_operand_address): Remove.
415 * config/ia64/ia64.c (TARGET_PRINT_OPERAND,
416 TARGET_PRINT_OPERAND_ADDRESS, TARGET_PRINT_OPERAND_PUNCT_VALID_P):
418 (ia64_print_operand_punct_valid_p): New function.
419 (ia64_print_operand, ia64_print_operand_address): Make static.
421 2011-12-02 Michael Meissner <meissner@the-meissners.org>
424 * config/rs6000/rs6000.c (def_builtin): Use the correct field to
425 figure out the function's attributes.
426 (struct builtin_description): Mark mask field as const since we no
427 longer need to modify it for SPE and PAIRED builtins.
429 2011-12-02 Joseph Myers <joseph@codesourcery.com>
433 2008-09-18 Andrew Pinski <andrew_pinski@playstation.sony.com>
436 * loop-doloop.c (doloop_modify): New argument zero_extend_p and
437 zero extend count after the correction to it is done.
438 (doloop_optimize): Update call to doloop_modify, don't zero extend
441 2008-11-03 Andrew Pinski <andrew_pinski@playstation.sony.com>
444 * loop-doloop.c (doloop_modify): Add from_mode argument that says what
446 (doloop_optimize): Update call to doloop_modify.
448 2011-12-02 Richard Guenther <rguenther@suse.de>
451 * ipa.c (varpool_externally_visible_p): Register variables
452 are always externally visible.
454 2011-12-02 Sameera Deshpande <sameera.deshpande@arm.com>
456 * gimple-pretty-print.c (dump_gimple_call): Dereference fn only if
459 2011-12-02 Martin Jambor <mjambor@suse.cz>
461 PR tree-optimization/50622
462 * tree-sra.c (load_assign_lhs_subreplacements): Force gimple operand
463 if both lacc and racc are grp_partial_lhs.
465 2011-12-01 Kaz Kojima <kkojima@gcc.gnu.org>
468 * config/sh/sh.c (expand_ashiftrt): Handle TARGET_SH2A same as
470 (shl_sext_kind): Likewise.
471 * config/sh/sh.h (SH_DYNAMIC_SHIFT_COST): Likewise.
472 * config/sh/sh.md (ashlsi3_sh2a, ashrsi3_sh2a, lshrsi3_sh2a): Remove.
473 (ashlsi3_std): Handle TARGET_SH2A same as TARGET_SH3.
475 (ashrsi3_d): Likewise.
476 (lshrsi3_d): Likewise.
479 2011-12-01 Diego Novillo <dnovillo@google.com>
482 * ipa-inline.c (can_inline_edge_p): If the edge E has a statement,
483 use the statement's inline indicator instead of E's.
484 Remove consistency check.
486 2011-12-01 Diego Novillo <dnovillo@google.com>
491 2011-11-29 Diego Novillo <dnovillo@google.com>
493 * gimple.c (gimple_call_set_cannot_inline): Move from gimple.h.
494 Update field call_stmt_cannot_inline_p from call
495 graph edge, if needed.
496 * gimple.h (gimple_call_set_cannot_inline): Move to gimple.c.
498 2011-12-01 Nathan Sidwell <nathan@acm.org>
500 PR gcov-profile/51113
501 * coverage.c (build_var): Keep coverage variables static.
503 2011-12-01 Jakub Jelinek <jakub@redhat.com>
505 PR tree-optimization/51356
506 * tree-vect-patterns.c (vect_recog_bool_pattern): Give up if
507 vectype doesn't have VECTOR_MODE_P.
510 * tree-ssa-dce.c (remove_dead_stmt): Add a debug stmt when removing
511 as unnecessary a store to a variable with gimple reg type.
512 * tree-ssa-live.c (remove_unused_locals): Clear TREE_ADDRESSABLE bit
513 on local unreferenced variables.
514 * cfgexpand.c (expand_gimple_basic_block): Don't emit DEBUG_INSNs
515 for !target_for_debug_bind variables.
517 2011-12-01 Patrick Marlier <patrick.marlier@gmail.com>
520 * cgraph.h (cgraph_call_node_duplication_hooks): Declare.
521 * cgraph.c (cgraph_call_node_duplication_hooks): Make global.
522 * cgraphunit.c (cgraph_copy_node_for_versioning): Call it.
524 2011-12-01 Andrew Pinski <apinski@cavium.com>
527 * tree.c (free_lang_data_in_decl): Clear FIELD_DECL's
530 2011-12-01 Jakub Jelinek <jakub@redhat.com>
532 PR tree-optimization/51246
533 * tree-predcom.c (replace_ref_with): Handle also clobber on the rhs.
535 PR rtl-optimization/51014
536 * loop-unroll.c (apply_opt_in_copies): Ignore label DEBUG_INSNs
537 both from bb and orig_bb.
539 2011-12-01 Joern Rennecke <joern.rennecke@embecosm.com>
541 PR tree-optimization/50802
542 * tree-vrp.c (simplify_conversion_using_ranges): Rewrite test
543 considering what happens to ranges during sign changes and/or
544 intermediate narrowing conversions.
546 2011-11-30 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
549 * config/pa/pa.md (in_branch_delay): Disallow frame related insns.
550 (in_nullified_branch_delay): Likewise.
551 (in_call_delay): Likewise.
553 2011-11-30 Richard Henderson <rth@redhat.com>
555 * predict.c (expr_expected_value_1): Assume compare-and-swap builtin
556 boolean return is true.
558 2011-11-30 Jakub Jelinek <jakub@redhat.com>
560 PR rtl-optimization/51044
561 * dse.c (emit_inc_dec_insn_before): Ensure new_insn is an instruction
562 rather than just insn body.
565 * function.c (thread_prologue_and_epilogue_insns): If
566 stack_realign_drap, add drap_reg to SET_UP_BY_PROLOGUE
570 * tree-eh.c (cleanup_empty_eh_merge_phis): Add check to
571 avoid creating duplicate edges here.
572 (cleanup_empty_eh_unsplit): And remove it in the caller.
574 2011-11-30 Andrew Pinski <apinski@cavium.com>
577 * c-parser.c (c_parser_postfix_expression): Check groktypename results
578 before looking at the main variant.
580 2011-11-30 Jakub Jelinek <jakub@redhat.com>
582 PR rtl-optimization/48721
583 * sched-deps.c (sched_analyze_insn): For SIBLING_CALL_P set
584 reg_pending_barrier to TRUE_BARRIER.
586 PR rtl-optimization/48721
587 * sched-ebb.c (begin_move_insn): Insert empty unreachable
588 block after BARRIER if insn is followed by it.
590 2011-11-30 Richard Henderson <rth@redhat.com>
592 * config/sparc/sparc-opts.h (enum sparc_memory_model_type): New.
593 * config/sparc/sparc.opt (mmemory-model=): New option.
594 * doc/invoke.texi (Sparc Options): Document it.
595 * config/sparc/sparc.c (sparc_option_override): Provide default
596 for sparc_memory_model.
597 (sparc_emit_membar_for_model): Omit barrier combinations that are
598 implied by the memory model.
600 2011-11-30 Richard Henderson <rth@redhat.com>
602 * config/sparc/sync.md (I124MODE): New mode iterator.
603 (memory_barrier): Remove.
604 (atomic_exchangesi): New.
605 (swapsi): Rename from *swapsi.
606 (atomic_test_and_set<I124MODE>): New.
607 (ldstubqi): Rename from *ldstubqi, merge with expander.
608 (ldstub<I24MODE>): Rename from *ldstub<I24MODE>, merge with expander.
609 (sync_lock_test_and_setsi): Remove.
610 (sync_lock_test_and_set<I12MODE>): Remove.
612 2011-11-30 Richard Henderson <rth@redhat.com>
614 * config/sparc/constraints.md ("w"): New.
615 * config/sparc/predicates.md (mem_noofs_operand): New.
616 * config/sparc/sparc.c (sparc_expand_compare_and_swap_12): Make static.
617 Generate the boolean result of the atomic_compare_exchange.
618 (sparc_expand_compare_and_swap): New.
619 * config/sparc/sync.md (atomic_compare_and_swap<I>): New.
620 (atomic_compare_and_swap<I48MODE>_1): Rename from
621 *sync_compare_and_swap<I48MODE>, use mem_noofs_operand.
622 (*atomic_compare_and_swapdi_v8plus): Rename from
623 *sync_compare_and_swapdi_v8plus, use mem_noofs_operand.
624 (sync_compare_and_swap<I12MODE>): Remove.
625 (sync_compare_and_swap<I48MODE>): Remove.
627 2011-11-30 Richard Henderson <rth@redhat.com>
629 * config/sparc/predicates.md (register_or_v9_zero_operand): New.
630 * config/sparc/sparc.md (UNSPEC_ATOMIC): New.
631 * config/sparc/sync.md (atomic_load<I>): New.
632 (atomic_loaddi_1, atomic_store<I>, atomic_storedi_1): New.
634 2011-11-30 Richard Henderson <rth@redhat.com>
636 * config/sparc/predicates.md (zero_or_v7_operand): New.
637 * config/sparc/sparc.c (sparc_emit_membar_for_model): New.
638 * config/sparc/sparc-protos.h: Update.
639 * config/sparc/sync.md (mem_thread_fence): New.
640 (memory_barrier): Use sparc_emit_membar_for_model.
641 (membar, *membar_empty, *membar_storestore, *membar_storeload): New.
642 (*membar_v8): Accept and ignore the membar mask.
643 (*membar): Accept and print the membar mask.
645 2011-11-30 Matthew Gretton-Dann <matthew.gretton-dann@arm.com>
647 * config/arm/arm.c (arm_issue_rate): Cortex-A15 can triple issue.
648 * config/arm/arm.md (mul64): New attribute.
649 (generic_sched): Cortex-A15 is not scheduled generically.
650 (cortex-a15.md): Include.
651 * config/arm/cortex-a15.md: New machine description.
652 * config/arm/t-arm (MD_INCLUDES): Add cortex-a15.md.
654 2011-11-30 Matthew Gretton-Dann <matthew.gretton-dann@arm.com>
656 * config/arm/t-arm (MD_INCLUDES): Ensure all md files are listed.
658 2011-11-30 Iain Sandoe <iains@gcc.gnu.org>
660 * doc/tm.texi.in (NEXT_OBJC_RUNTIME): Update macro description.
661 * doc/tm.texi: Regenerate.
662 * flags.h (flag_next_runtime): Remove references.
663 * toplev.c: Likewise.
664 * defaults.h (NEXT_OBJC_RUNTIME): Provide default.
665 * gccspec.c (lang_specific_driver): Adjust use of NEXT_OBJC_RUNTIME.
666 * config/darwin.c (darwin_override_options): Provide default
667 Objective-C abi settings and target conflict checks.
668 * config/darwin.h (SUBTARGET_C_COMMON_OVERRIDE_OPTIONS): Remove unused
669 code. Adjust indenting.
671 2011-11-29 Joseph Myers <joseph@codesourcery.com>
673 * config/sh/sh.h (enum reg_class, REG_CLASS_NAMES)
674 (REG_CLASS_CONTENTS): Add NON_SP_REGS.
675 (REGCLASS_HAS_GENERAL_REG): Handle NON_SP_REGS.
676 * config/sh/constraints.md (u): New constraint.
678 2011-11-29 Oleg Endo <oleg.endo@t-online.de>
681 * config/sh/sh.c (sh_secondary_reload): Add case when FPUL
682 register is being loaded from a pseudo in memory.
684 2011-11-29 DJ Delorie <dj@redhat.com>
686 * config.gcc (rl78-*-elf): New case.
687 * doc/extend.texi: Add RL78 documentation.
688 * doc/invoke.texi: Likewise.
689 * doc/md.texi: Likewise.
690 * doc/contrib.texi: Add RL78.
691 * doc/install.texi: Add rl78-*-elf.
692 * config/rl78: New directory for the Renesas RL78.
694 2011-11-29 Jakub Jelinek <jakub@redhat.com>
696 PR tree-optimization/51247
697 * tree-vrp.c (extract_range_from_assert): For signed 1-bit precision
698 types instead of adding 1 subtract -1 and instead of subtracting 1
699 add -1 to avoid overflows.
701 2011-11-29 Andrew MacLeod <amacleod@redhat.com>
704 * optabs.c (maybe_optimize_fetch_op): New. Look for more optimal
705 instructions for a FECTH_OP or OP_FECTH sequence.
706 (expand_atomic_fetch_op): Call maybe_optimize_fetch_op.
708 2011-11-29 Uros Bizjak <ubizjak@gmail.com>
710 * config/i386/sync.md (UNSPEC_LDA, UNSPEC_STA): New unspecs.
711 (movdi_via_fpu): Remove.
712 (loaddi_via_fpu): New insn pattern.
713 (storedi_via_fpu): Ditto.
714 (atomic_loaddi_fpu): Use loaddi_via_fpu and storedi_via_fpu.
715 (atomic_storedi_fpu): Ditto.
716 * reg-stack.c (get_true_reg): Handle UNSPEC_LDA.
717 (subst_stack_regs_pat): Handle UNSPEC_STA.
719 2011-11-29 Uros Bizjak <ubizjak@gmail.com>
721 * config/i386/i386.md (*floathi<mode>2_i387_with_temp): Do not
722 allocate scratch memory for alternative 0.
724 2011-11-29 Michael Meissner <meissner@linux.vnet.ibm.com>
726 * config/rs6000/rs6000-builtins.def: Completely rewrite builtin
727 handling to centralize more of the builtins in this file. Change
728 some builtin enumerations to be more consistant. Use a new mask
729 to hold the current builtins, including SPE and PAIRED builtins
730 which no longer are set via target_flags masks. Add
731 -mdebug=builtin debug support. For power machines, define all
732 Altivec and VSX buitins when the compiler starts, but don't allow
733 the use of a builtin unless the appropriate switch is used, or
734 #pragma GCC target is used to change the options. If the user
735 uses #pragma GCC target, update the appropriate hardware macros.
736 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Ditto.
737 * config/rs6000/rs6000.opt (rs6000_builtin_mask): Ditto.
738 * config/rs6000/rs6000-c.c (rs6000_macro_to_expand): Ditto.
739 (rs6000_define_or_undefine_macro): Ditto.
740 (rs6000_target_modify_macros): Ditto.
741 (rs6000_cpu_cpp_builtins): Ditto.
742 (altivec_overloaded_builtins): Ditto.
743 (altivec_build_resolved_builtin): Ditto.
744 * config/rs6000/rs6000.c (rs6000_target_modify_macros_ptr): Ditto.
745 (rs6000_builtin_info): Ditto.
746 (def_builtin): Ditto.
747 (enable_mask_for_builtins): Ditto.
748 (DEBUG_FMT_X): Ditto.
749 (rs6000_debug_reg_global): Ditto.
750 (rs6000_builtin_mask_calculate): Ditto.
751 (rs6000_option_override_internal): Ditto.
752 (rs6000_builtin_conversion): Ditto.
753 (rs6000_builtin_vectorized_function): Ditto.
757 (builtin_description_predicates): Ditto.
758 (bdesc_altivec_preds): Ditto.
759 (bdesc_spe_predicates): Ditto.
760 (bdesc_spe_evsel): Ditto.
761 (bdesc_paired_preds): Ditto.
764 (rs6000_overloaded_builtin_p): Ditto.
765 (rs6000_expand_unop_builtin): Ditto.
766 (bdesc_2arg_spe): Ditto.
767 (spe_expand_builtin): Ditto.
768 (rs6000_invalid_builtin): Ditto.
769 (rs6000_expand_builtin): Ditto.
770 (rs6000_init_builtins): Ditto.
771 (spe_init_builtins): Ditto.
772 (paired_init_builtins): Ditto.
773 (altivec_init_builtins): Ditto.
774 (builtin_function_type): Ditto.
775 (rs6000_common_init_builtins): Ditto.
776 (rs6000_builtin_reciprocal): Ditto.
777 (rs6000_builtin_mask_names): Ditto.
778 (rs6000_pragma_target_parse): Ditto.
779 (rs6000_function_specific_print): Ditto.
780 * config/rs6000/rs6000.h (MASK_DEBUG_BUILTIN): Ditto.
781 (MASK_DEBUG_ALL): Ditto.
782 (TARGET_DEBUG_BUILTIN): Ditto.
783 (TARGET_EXTRA_BUILTINS): Ditto.
784 (REGISTER_TARGET_PRAGMAS): Ditto.
785 (enum rs6000_btc): Ditto.
786 (RS6000_BTC_*): Ditto.
787 (RS6000_BTM_*): Ditto.
788 (enum rs6000_builtins): Ditto.
789 * config/rs6000/rs6000-protos.h (rs6000_overloaded_builtin_p): Ditto.
790 (rs6000_target_modify_macros): Ditto.
791 (rs6000_target_modify_macros_ptr): Ditto.
793 * config/rs6000/vector.md (unsfloat<VEC_int><mode>2): Use the
794 standard name for converting vector unsigned values to floating
795 point. Delete the TARGET_VECTORIZE_BUILTIN_CONVERSION hook, which
796 is handled by machine independent code.
797 * config/rs6000/rs6000.c (TARGET_VECTORIZE_BUILTIN_CONVERSION): Ditto.
798 (rs6000_builtin_conversion): Ditto.
800 2011-11-29 Dodji Seketeli <dodji@redhat.com>
802 * tree-diagnostic (struct loc_t): Rename into struct loc_map_pair.
803 (maybe_unwind_expanded_macro_loc): Adjust.
805 2011-11-29 David Edelsohn <dje.gcc@gmail.com>
806 Aldy Hernandez <aldyh@redhat.com>
808 * trans-mem.c (ipa_tm_create_version_alias): Mangle new_decl
810 (ipa_tm_create_version): Same.
812 2011-11-29 Bernd Schmidt <bernds@codesourcery.com>
814 * haifa-sched.c (recompute_todo_spec): Simplify and correct the
815 code checking for a clobber of a condition register when deciding
816 whether to predicate.
818 2011-11-29 Diego Novillo <dnovillo@google.com>
820 * gimple.c (gimple_call_set_cannot_inline): Move from gimple.h.
821 Update field call_stmt_cannot_inline_p from call graph edge, if needed.
822 * gimple.h (gimple_call_set_cannot_inline): Move to gimple.c.
824 2011-11-29 Sameera Deshpande <sameera.deshpande@arm.com>
826 * config/arm/arm.md (arm_movdi): Update NEG_POOL_RANGE.
827 (movdf_soft_insn): Likewise.
828 * config/arm/fpa.md (thumb2_movdf_fpa): Likewise.
829 * config/arm/neon.md (neon_mov<mode>): Likewise.
830 * config/arm/vfp.md (movdi_vfp): Likewise.
831 (movdi_vfp_cortexa8): Likewise.
832 (movdf_vfp): Likewise.
834 2011-11-29 Jakub Jelinek <jakub@redhat.com>
837 * expr.c (expand_expr_addr_expr_1): Don't call force_operand for
840 2011-11-29 Ira Rosen <ira.rosen@linaro.org>
842 PR tree-optimization/51301
843 * tree-vect-patterns.c (vect_recog_over_widening_pattern): Check that
844 the last statement doesn't convert to a bigger type than the original
845 type of the computation.
847 2011-11-28 Richard Henderson <rth@redhat.com>
849 * config/rs6000/rs6000.c (rs6000_expand_atomic_compare_and_swap):
850 Handle overlap between retval and oldval. Always compute boolval
853 2011-11-28 Jakub Jelinek <jakub@redhat.com>
856 * tree-ssa.c (execute_update_addresses_taken): Remove
857 var ={v} {CLOBBER} stmts instead of rewriting them into
858 var_N ={v} {CLOBBER}.
861 * function.c (convert_jumps_to_returns): When redirecting an edge
862 succ to EXIT_BLOCK_PTR, clear EDGE_CROSSING flag.
864 PR tree-optimization/50078
865 * tree-ssa-forwprop.c (forward_propagate_addr_expr_1): Copy over
866 TREE_THIS_VOLATILE also from the old to new lhs resp. rhs.
868 PR tree-optimization/50682
869 * tree-eh.c (maybe_remove_unreachable_handlers): New function.
870 * tree-flow.h (maybe_remove_unreachable_handlers): New prototype.
871 * tree-optimize.c (execute_cleanup_cfg_post_optimizing): Call it.
873 2011-11-28 Georg-Johann Lay <avr@gjlay.de>
875 * doc/extend.texi (AVR Built-in Functions): Add documentation for
876 __builtin_avr_map8 and __builtin_avr_map16.
878 * config/avr/avr.md: Document new %t and %T asm output codes.
879 (define_c_enum "unspec"): Add UNSPEC_MAP_BITS.
880 (adjust_len): Add map_bits.
881 (map_bitsqi, map_bitshi): New insns.
882 * config/avr/avr-protos.h (avr_out_map_bits): New.
883 * config/avr/avr-protos.c (print_operand): Implement %t and %T.
884 (adjust_insn_length): Handle ADJUST_LEN_MAP_BITS.
885 (avr_double_int_push_digit): New function.
886 (avr_map, avr_revert_map, avr_swap_map, avr_id_map): New functions.
887 (avr_sig_map, avr_map_hamming_byte): New functions.
888 (avr_out_swap_bits, avr_out_revert_bits, avr_move_bits,
889 avr_out_map_bits): New functions.
890 (enum avr_builtin_id): Add AVR_BUILTIN_MAP8, AVR_BUILTIN_MAP16.
891 (avr_init_builtins): Populate __builtin_avr_map8, __builtin_avr_map16.
892 (bdesc_2arg): Add __builtin_avr_map8, __builtin_avr_map16 ...
893 (avr_expand_builtin): ...and expand them.
894 * config/avr/avr-c.c (avr_cpu_cpp_builtins): New built-in defines:
895 __BUILTIN_AVR_MAP8, __BUILTIN_AVR_MAP16.
897 2011-11-27 Richard Sandiford <rdsandiford@googlemail.com>
900 * config/mips/mips.c (mips_rtx_costs): Handle baddu.
902 2011-11-27 Richard Sandiford <rdsandiford@googlemail.com>
904 * config/mips/mips.c (mips16_unextended_reference_p): Test for BLKmode.
906 2011-11-27 Richard Sandiford <rdsandiford@googlemail.com>
908 * config/mips/mips-protos.h (mips_emit_binary): Declare.
909 * config/mips/mips.c (mips_emit_binary): Make global.
910 (mips_set_mips16_mode): Turn off -mfix-r4000 in MIPS16 mode.
911 (mips_conditional_register_usage): Don't treat LO and HI as
912 register operands in MIPS16 mode.
913 (mips_mulsidi3_gen_fn): Use {u,}mulsidi3_{32,64}bit_mips16
915 * config/mips/predicates.md (muldiv_target_operand): New predicate.
916 (move_operand): Allow hilo_operand.
917 * config/mips/mips.md (mul<mode>3): Explicitly specify LO as the
918 target of MIPS16 multiplies, then move it into the target register.
919 (mul<mode>3_internal, *macc2, *msac2): Use muldiv_target_operand.
920 (<u>mulsidi3_32bit_mips16): New expander.
921 (<u>mulsidi3_32bit): Use muldiv_target_operand.
922 (<u>mulsidi3_32bit_r4000): Disable for ISA_HAS_DSP.
923 (<u>mulsidi3_64bit): Require !TARGET_MIPS16. Split into
924 <u>mulsidi3_64bit_split.
925 (<u>mulsidi3_64bit_mips16): New expander.
926 (<u>mulsidi3_64bit_split): Likewise, using expansions from
927 two previous define_splits.
928 (<u>mulsidi3_64bit_hilo, *muls<u>_di, <u>msubsidi4): Use
929 muldiv_target_operand.
930 (<su>mulsi3_highpart): Use <su>mulsi3_highpart_split for MIPS16 code.
931 (<su>mulsi3_highpart_internal): Require !TARGET_MIPS16.
932 Split into <su>mulsi3_highpart_split.
933 (<su>mulsi3_highpart_split): New expander.
934 (<su>muldi3_highpart): Turn into a define_expand.
935 Use <su>muldi3_highpart_split for MIPS16 code.
936 (<su>muldi3_highpart_internal): Renamed from <su>muldi3_highpart.
937 Require !TARGET_MIPS16. Split into <su>muldi3_highpart_split.
938 (<su>muldi3_highpart_split): New expander.
939 (<u>mulditi3): Explicitly specify LO as the target of MIPS16
940 multiplies, then move it into the target register.
941 (<u>mulditi3_internal, <u>maddsidi4): Use muldiv_target_operand.
942 (divmod<mode>4, udivmod<mode>4): Turn into define_expands.
943 Use <u>divmod<mode>4_split for MIPS16 code, then explicitly
944 move LO into operand 0.
945 (divmod<mode>4_internal, udivmod<mode>4_internal): Renamed
946 from <u>divmod<mode>4. Use muldiv_target_operand.
947 Require !TARGET_MIPS16. Split into <u>divmod<mode>4_split.
948 (<u>divmod<mode>4_split): New expander.
949 (<u>divmod<GPR:mode>4_hilo_<HILO:mode>): Use muldiv_target_operand.
950 (mfhi<GPR:mode>_<HILO:mode>): Use hilo_operand.
952 2011-11-27 Richard Sandiford <rdsandiford@googlemail.com>
954 * hard-reg-set.h (target_hard_regs): Add x_accessible_reg_set
955 and x_operand_reg_set.
956 (accessible_reg_set, operand_reg_set): New macros.
957 * reginfo.c (init_reg_sets): Initialize accessible_reg_set and
959 (saved_accessible_reg_set, saved_operand_reg_set): New variables.
960 (save_register_info): Save them.
961 (restore_register_info): Restore them.
962 (init_reg_sets_1): Limit operand_reg_set to accessible_reg_set.
963 Remove NO_REGS registers from operand_reg_set. Treat members
964 of operand_reg_set as fixed.
965 * recog.c (general_operand): Check operand_reg_set rather than NO_REGS.
966 (register_operand, nonmemory_operand): Likewise.
967 * varasm.c (make_decl_rtl): Always use DECL_MODE as the mode of
968 register variables. Check accessible_reg_set and operand_reg_set.
969 * config/mips/mips.c (mips_conditional_register_usage): Remove
970 inaccessible register from accessible_reg_set, rather than just
973 2011-11-27 Gerald Pfeifer <gerald@pfeifer.com>
975 * config/pa/pa-linux.h (TARGET_GAS): Remove comment.
977 2011-11-26 Richard Henderson <rth@redhat.com>
979 * config/arm/arm.c (arm_init_libfuncs): Call init_sync_libfuncs
982 2011-11-26 Richard Henderson <rth@redhat.com>
984 * optabs.c (expand_atomic_fetch_op): Always return result.
986 * optabs.c (init_sync_libfuncs_1): Include max in iteration.
988 * omp-low.c (expand_omp_atomic): Assume anything aligned to
989 BIGGEST_ALIGNMENT is aligned.
991 2011-11-26 Richard Henderson <rth@redhat.com>
993 * config/m68k/m68k.md (UNSPECV_CAS_1, UNSPECV_CAS_2): New.
994 (UNSPECV_TAS_1, UNSPECV_TAS_2): New.
995 (I): New mode iterator.
996 (xz): New mode attribute.
997 * config/m68k/sync.md: New file.
999 * config/m68k/linux.h (TARGET_INIT_LIBFUNCS): New.
1000 * config/m68k/m68k.c (m68k_init_sync_libfuncs): New.
1002 2011-11-26 Nathan Sidwell <nathan@acm.org>
1004 PR gcov-profile/51297
1005 * gcov.c (main): Allocate initial names and sources arrays.
1006 (find_source): Don't check for null name or source arrays here.
1008 2011-11-26 Jakub Jelinek <jakub@redhat.com>
1010 PR rtl-optimization/49912
1011 * cfgrtl.c (rtl_verify_flow_info_1): Ignore also EDGE_PRESERVE bit
1012 when counting n_branch.
1014 2011-11-26 Uros Bizjak <ubizjak@gmail.com>
1016 * config/i386/sync.md (movdi_via_fpu): Add %Z insn suffixes.
1018 2011-11-26 Joern Rennecke <joern.rennecke@embecosm.com>
1021 * calls.c (mem_overlaps_already_clobbered_arg_p):
1022 Return false if no outgoing arguments have been stored so far.
1024 2011-11-25 Georg-Johann Lay <avr@gjlay.de>
1026 * config/avr/avr.c (print_operand): Support code = 'i' for CONST_INT.
1028 2011-11-24 Andrew MacLeod <amacleod@redhat.com>
1031 * tree.h (is_builtin_name): No longer external.
1032 * builtins.c (is_builtin_name): Make static.
1033 * varasm.c (incorporeal_function_p): __sync and __atomic external calls
1034 are not incorporeal and may need asm label processing.
1036 2011-11-25 Georg-Johann Lay <avr@gjlay.de>
1039 * config/avr/avr-protos.h (avr_log_t): Add field .builtin.
1040 * config/avr/avr-log.c (avr_log_set_avr_log): Initialize it.
1041 Don't bypass TARGET_ALL_DEBUG. Print self-info with ?.
1042 (avr_log_vadump): Support %D and %X to print double_int.
1043 (avr_double_int_pop_digit): New static function.
1044 (avr_dump_double_int_hex): New static function.
1046 2011-11-24 Enkovich Ilya <ilya.enkovich@intel.com>
1049 * config/i386/i386.c (distance_non_agu_define_in_bb): Fix insn
1052 2011-11-24 Andrew MacLeod <amacleod@redhat.com>
1054 * optab.c (maybe_emit_atomic_exchange): New. Try to emit an
1055 atomic_exchange pattern.
1056 (maybe_emit_sync_lock_test_and_set): New. Try to emit an exchange
1057 using __sync_lock_test_and_set.
1058 (maybe_emit_compare_and_swap_exchange_loop): New. Try to emit an
1059 exchange using a compare_and_swap loop.
1060 (expand_sync_lock_test_and_set): New. Expand sync_lock_test_and_set.
1061 (expand_atomic_test_and_set): New. Expand test_and_set operation.
1062 (expand_atomic_exchange): Use new maybe_emit_* functions.
1063 (expand_atomic_store): Use new maybe_emit_* functions.
1064 * builtins.c (expand_builtin_sync_lock_test_and_set): Call
1065 expand_sync_lock_test_and_set routine.
1066 (expand_builtin_atomic_exchange): Remove parameter from call.
1067 (expand_builtin_atomic_clear): Use atomic_clear pattern if present.
1068 (expand_builtin_atomic_test_and_set): Add target and simply call
1069 expand_atomic_test_and_set.
1070 (expand_builtin): Add target to expand_builtin_atomic_test_and_set.
1071 * expr.h (expand_atomic_exchange): Add parameter.
1072 (expand_sync_lock_test_and_set): New prototype.
1073 (expand_atomic_test_and_set, expand_atomic_clear): New prototypes.
1075 2011-11-24 H.J. Lu <hongjiu.lu@intel.com>
1078 * config/i386/i386.h (processor_costs): Revert revision 181357.
1079 * config/i386/i386.c (cost models): Likewise.
1080 (core_cost): Likewise.
1081 (promote_duplicated_reg): Likewise.
1082 (promote_duplicated_reg_to_size): Likewise.
1083 (processor_target): Likewise.
1084 (expand_set_or_movmem_via_loop_with_iter): Likewise.
1085 (expand_set_or_movmem_via_loop): Likewise.
1086 (emit_strset): Likewise.
1087 (expand_movmem_epilogue): Likewise.
1088 (expand_setmem_epilogue): Likewise.
1089 (expand_movmem_prologue): Likewise.
1090 (expand_setmem_prologue): Likewise.
1091 (expand_constant_movmem_prologue): Likewise.
1092 (expand_constant_setmem_prologue): Likewise.
1093 (decide_alg): Likewise.
1094 (decide_alignment): Likewise.
1095 (ix86_expand_movmem): Likewise.
1096 (ix86_expand_setmem): Likewise.
1097 (ix86_slow_unaligned_access): Likewise.
1098 * config/i386/i386.md (strset): Likewise.
1099 * config/i386/sse.md (vec_dupv4si): Likewise.
1100 (vec_dupv2di): Likewise.
1102 2011-11-24 Georg-Johann Lay <avr@gjlay.de>
1104 * config/avr/avr.md (*rotlhi2.15): Set length to 4.
1106 2011-11-24 Razya Ladelsky <razya@il.ibm.com>
1108 * tree-data-ref.c (initialize_data_dependence_relation): Update
1109 comment for the self dependence case.
1110 (compute_self_dependence): Remove.
1111 * tree-vect-data-refs.c (vect_analyze_data_refs): Remove call to
1112 compute_self_dependenc.
1114 2011-11-24 Alan Modra <amodra@gmail.com>
1116 * c-decl.c (warn_if_shadowing): Don't warn if shadowed
1117 identifier is from system header.
1119 2011-11-23 Jakub Jelinek <jakub@redhat.com>
1122 * config/i386/i386.c (decide_alg): Initialize *dynamic_check
1125 2011-11-23 Vladimir Makarov <vmakarov@redhat.com>
1127 PR rtl-optimization/48455
1128 * doc/invoke.texi (-fira-region): Document default values.
1130 * flags-types.h (enum ira_region): Add new value
1131 IRA_REGION_AUTODETECT.
1133 * common.opt (fira-region): Set up initial value to
1134 IRA_REGION_AUTODETECT.
1136 * toplev.c (process_options): Set up flag_ira_region depending on
1139 * ira.c (ira.c): Remove optimize guard for ira_build.
1141 2011-11-23 Chung-Lin Tang <cltang@codesourcery.com>
1143 PR rtl-optimization/50496
1144 * jump.c (redirect_jump): Assert fail on nlabel == NULL_RTX
1145 only after epilogue is created. Add comments.
1147 2011-11-22 Richard Henderson <rth@redhat.com>
1149 * config/ia64/ia64.c (ia64_expand_atomic_op): Add model parameter.
1150 Generate the barrier required for the memory model.
1151 (rtx_needs_barrier): Handle UNSPEC_FETCHADD_REL, UNSPEC_CMPXCHG_REL.
1152 * config/ia64/ia64-protos.h: Update.
1153 * config/ia64/ia64.md (UNSPEC_FETCHADD_REL): New.
1154 (UNSPEC_CMPXCHG_REL): New.
1155 * config/ia64/sync.md (mem_thread_fence): New.
1156 (atomic_load<IMODE>, atomic_store<IMODE>): New.
1157 (atomic_compare_and_swap<IMODE>): New.
1158 (cmpxchg_acq_<I124MODE>, cmpxchg_acq_di): New.
1159 (atomic_exchange<IMODE>): New.
1160 (xchg_acq_<IMODE>): Rename from sync_lock_test_and_set<IMODE>.
1161 (atomic_<FETCHOP><IMODE>, atomic_nand<IMODE>): New.
1162 (atomic_fetch_<FETCHOP><IMODE>, atomic_fetch_nand<IMODE>): New.
1163 (atomic_<FETCHOP>_fetch<IMODE>, atomic_nand_fetch<IMODE>): New.
1164 (fetchadd_rel_<I48MODE>): New.
1165 (sync_<FETCHOP><IMODE>, sync_nand<IMODE>): Remove.
1166 (sync_old_<FETCHOP><IMODE>, sync_old_nand<IMODE>): Remove.
1167 (sync_new_<FETCHOP><IMODE>, sync_new_nand<IMODE>): Remove.
1168 (sync_compare_and_swap<IMODE>): Remove.
1169 (sync_lock_release<IMODE>): Remove.
1171 * config/ia64/ia64.md: Use define_c_enum for UNSPEC_*
1172 and UNSPECV_* constants.
1174 2011-11-22 Tom de Vries <tom@codesourcery.com>
1176 PR rtl-optimization/50764
1177 * sched-deps.c (sched_analyze_insn): Make sure frame-related insns are
1178 not moved past preceding jump.
1180 2011-11-22 Jeff Law <law@redhat.com>
1182 * doc/contrib.texi: Add entry for David Binderman.
1184 2011-11-22 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
1186 * postreload.c (reload_combine): Mark reg_state as invalid at
1187 volatile insns if there has been a use already.
1189 2011-11-21 Aldy Hernandez <aldyh@redhat.com>
1191 * opts.c (finish_options): Do not fail for -fgnu-tm.
1192 * gimple-streamer-out.c (output_gimple_stmt): Handle
1194 * gimple-streamer-in.c (input_gimple_stmt): Same.
1195 * lto-cgraph.c (input_overwrite_node): Read tm_clone bit.
1196 (lto_output_node): Write tm_clone bit.
1198 2011-11-22 Ian Lance Taylor <iant@google.com>
1200 * doc/install.texi (Configuration): Correct doc of
1201 --enable-build-poststage1-with-cxx: it is not experimental.
1203 2011-11-22 Jakub Jelinek <jakub@redhat.com>
1205 PR tree-optimization/51074
1206 * fold-const.c (fold_binary_loc): Fix up VEC_INTERLEAVE_*_EXPR
1207 handling for BYTES_BIG_ENDIAN.
1208 * optabs.c (can_vec_perm_for_code_p): Likewise.
1210 2011-11-22 Michael Matz <matz@suse.de>
1213 * trans-mem.c (expand_block_tm): Ignore clobbers.
1215 2011-11-22 Iain Sandoe <iains@gcc.gnu.org>
1217 * config/darwin.h (ENDFILE_SPEC): New.
1218 (DARWIN_CRT1_SPEC): Add crttms.o.
1219 (STARTFILE_SPEC): Likewise for dynamic libs and bundles.
1220 * config/i386.h (ENDFILE_SPEC): Amend to include crttme.o.
1222 2011-11-22 Michael Matz <matz@suse.de>
1225 * tree.c (iterative_hash_expr): Remove break after return.
1227 2011-11-22 Iain Sandoe <iains@gcc.gnu.org>
1229 * target.def (tm_clone_table_section): New hook.
1230 * doc/tm.texi.in (TARGET_ASM_TM_CLONE_TABLE_SECTION): Define.
1231 * doc/tm.texi: Regenerate.
1232 * varasm.c (dump_tm_clone_pairs): Use target tm_clone_table_section.
1233 * output.h (default_clone_table_section): New prototype.
1234 * config/darwin.h (TARGET_ASM_TM_CLONE_TABLE_SECTION): New.
1235 * config/darwin-protos.h (darwin_tm_clone_table_section): New prototype.
1236 * config/darwin.c (darwin_tm_clone_table_section): New.
1238 2011-11-21 Torvald Riegel <triegel@redhat.com>
1240 * trans-mem.c (diagnose_tm_1): Print an expression instead of a
1241 declaration in error messages for indirect calls.
1243 2011-11-21 David S. Miller <davem@davemloft.net>
1245 * config/sparc/sparc.c (sparc_regmode_natural_size): New function
1246 implementing REGMODE_NATURAL_SIZE taking into consideration vector
1248 (sparc_modes_tieable_p): Similarly for MODES_TIEABLE_P.
1249 * config/sparc/sparc-protos.h (sparc_regmode_natural_size,
1250 sparc_modes_tieable_p): Declare.
1251 * gcc/config/sparc/sparc.h (REGMODE_NATURAL_SIZE,
1252 MODES_TIEABLE_P): Use new helper functions.
1255 2011-11-16 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
1258 * expmed.c (store_bit_field_1): Use extract_bit_field on big
1259 endian targets if the source cannot be exactly covered by word
1262 2011-11-21 Jakub Jelinek <jakub@redhat.com>
1265 * var-tracking.c (loc_cmp): For ENTRY_VALUEs use ENTRY_VALUE_EXP
1266 macro instead of XEXP.
1268 2011-11-21 Andreas Tobler <andreast@fgznet.ch>
1270 * configure: Regenerate.
1272 2011-11-21 Georg-Johann Lay <avr@gjlay.de>
1274 * config/avr/avr.c (output_reload_in_const): Loading a byte with 0
1275 must not affect cc0.
1276 * config/avr/avr.md (*movhi, *movpsi, *movsi, *movsf): Zero to any
1277 register does not change cc0. Same for any constant to ld-register.
1279 2011-11-21 Uros Bizjak <ubizjak@gmail.com>
1281 * config/i386/mmx.md (unspec) <UNSPEC_MOVNTQ, UNSPEC_PFRCP,
1282 UNSPEC_PFRCPIT1, UNSPEC_PFRCPIT2, UNSPEC_PFRSQRT, UNSPEC_PFRSQIT1>:
1283 Move from config/i386/i386.md
1284 (unspecv) <UNSPECV_EMMS, UNSPECV_FEMMS>: Ditto.
1285 * config/i386/sse.md (unspec) <UNSPEC_MOVNT,UNSPEC_MOVU, UNSPEC_LDDQU,
1286 UNSPEC_PSHUFB, UNSPEC_PSIGN, UNSPEC_PALIGNR, UNSPEC_EXTRQI,
1287 UNSPEC_EXTRQ, UNSPEC_INSERTQI, UNSPEC_INSERTQ, UNSPEC_BLENDV,
1288 UNSPEC_INSERTPS, UNSPEC_DP, UNSPEC_MOVNTDQA, UNSPEC_MPSADBW,
1289 UNSPEC_PHMINPOSUW, UNSPEC_PTEST, UNSPEC_PCMPESTR, UNSPEC_PCMPISTR,
1290 UNSPEC_FMADDSUB, UNSPEC_XOP_UNSIGNED_CMP, UNSPEC_XOP_TRUEFALSE,
1291 UNSPEC_XOP_PERMUTE, UNSPEC_FRCZ, UNSPEC_AESENC, UNSPEC_AESENCLAST,
1292 UNSPEC_AESDEC, UNSPEC_AESDECLAST, UNSPEC_AESIMC,
1293 UNSPEC_AESKEYGENASSIST, UNSPEC_PCLMUL, UNSPEC_PCMP, UNSPEC_VPERMIL,
1294 UNSPEC_VPERMIL2, UNSPEC_VPERMIL2F128, UNSPEC_CAST, UNSPEC_VTESTP,
1295 UNSPEC_VCVTPH2PS, UNSPEC_VCVTPS2PH, UNSPEC_VPERMSI, UNSPEC_VPERMDF,
1296 UNSPEC_VPERMSF, UNSPEC_VPERMTI, UNSPEC_GATHER, UNSPEC_VSIBADDR>: Ditto.
1297 (unspecv) <UNSPECV_LDMXCSR, UNSPECV_STMXCSR, UNSPECV_CLFLUSH,
1298 UNSPECV_MONITOR, UNSPECV_MWAIT, UNSPECV_VZEROALL, UNSPECV_VZEROUPPER>:
1300 * config/i386/sync.md (unspec) <UNSPEC_LFENCE, UNSPEC_SFENCE,
1301 UNSPEC_MFENCE, UNSPEC_MOVA>: Ditto.
1302 (unspecv) <UNSPECV_CMPXCHG_1, UNSPECV_CMPXCHG_2, UNSPECV_CMPXCHG_3,
1303 UNSPECV_CMPXCHG_4, UNSPECV_XCHG, UNSPECV_LOCK>: Ditto.
1304 (sse2_lfence): Move from config/i386/sse.md.
1305 (*sse2_lfence): Ditto.
1306 (sse_sfence): Ditto.
1307 (*sse_sfence): Ditto.
1308 (sse2_mfence): Ditto.
1309 (mfence_sse2): Ditto. Rename from *sse2_mfence. Enable also
1311 (mem_thread_fence): Use mfence_sse2.
1313 2011-11-21 Georg-Johann Lay <avr@gjlay.de>
1315 * config/avr/avr.h (struct base_arch_s): Add field sfr_offset.
1316 * config/avr/avr-devices.c: Ditto. And initialize it.
1317 * config/avr/avr-c.c (avr_cpu_cpp_builtins): New built-in define
1319 * config/avr/avr-protos.h (out_movqi_r_mr, out_movqi_mr_r): Remove.
1320 (out_movhi_r_mr, out_movhi_mr_r): Remove.
1321 (out_movsi_r_mr, out_movsi_mr_r): Remove.
1322 * config/avr/avr.md (*cbi, *sbi): Use %i instead of %m-0x20.
1323 (*insv.io, *insv.not.io): Ditto.
1324 * config/avr/avr.c (out_movsi_r_mr, out_movsi_mr_r): Make static.
1325 (print_operand): Implement "%i" to print address as I/O address.
1326 (output_movqi): Clean up call of out_movqi_mr_r.
1327 (output_movhi): Clean up call of out_movhi_mr_r.
1328 (avr_file_start): Use avr_current_arch->sfr_offset instead of
1329 magic -0x20. Use TMP_REGNO, ZERO_REGNO instead of 0, 1.
1330 (avr_out_sbxx_branch): Use %i instead of %m-0x20.
1331 (out_movqi_r_mr, out_movqi_mr_r): Ditto. And make static.
1332 (out_movhi_r_mr, out_movhi_mr_r): Ditto. And use avr_asm_len.
1333 (out_shift_with_cnt): Clean up code: Use avr_asm_len.
1334 (output_movsisf): Use output_reload_insisf for all CONSTANT_P sources.
1335 (avr_out_movpsi): USE avr_out_reload_inpsi for all CONSTANT_P sources.
1336 Clean up call of avr_out_store_psi.
1337 (output_reload_in_const): Don't cut symbols longer than 2 bytes.
1338 (output_reload_insisf): Filter CONST_INT_P or CONST_DOUBLE_P to
1339 try if setting pre-cleared register is advantageous.
1340 (avr_out_plus_1): Use gen_int_mode instead of GEN_INT.
1342 2011-11-20 Joey Ye <joey.ye@arm.com>
1344 * expr.c (expand_expr_real_1): Correctly handle strict volatile
1345 bitfield loads smaller than mode size.
1347 2011-11-20 Richard Henderson <rth@redhat.com>
1349 * config/mips/mips.c (mips_init_libfuncs): Call init_sync_libfuncs.
1351 2011-11-20 Uros Bizjak <ubizjak@gmail.com>
1354 * config/i386/i386.c (ix86_expand_vcond): Generate TARGET_XOP
1355 patterns for supported mode only.
1358 * config/i386/i386.c (ix86_expand_builtin)
1359 <IX86_BUILTIN_GATHERALTSIV4DI>: Use CODE_FOR_avx2_gathersiv4di.
1361 2011-11-20 Uros Bizjak <ubizjak@gmail.com>
1363 * config/i386/i386.md (UNSPEC_MOVNTI): Remove.
1364 (UNSPEC_MOVNTQ): New unspec.
1365 * config/i386/mmx.md (sse_movntq): Rename from sse_movntdi.
1366 Use UNSPEC_MOVNTQ instead of UNSPEC_MOVNT.
1367 * config/i386/sse.md (sse2_movnti<mode>): Use UNSPEC_MOVNT instead of
1369 (STORENT_MODE): Add DI and V4DI modes.
1371 2011-11-20 Nathan Sidwell <nathan@acm.org>
1373 PR gcov-profile/51113
1374 * coverage.c (build_var): Propagate visibility for public
1375 decls. Use make_decl_one_only for PIC correctness.
1378 * gcc.misc-tests/gcov-13.c: Check gcovpart-13b coverage
1379 * gcc.misc-tests/gcov-16.c: New.
1380 * gcc.misc-tests/gcov-17.c: New.
1381 * g++.dg/gcov/gcov-8.C: New.
1382 * g++.dg/gcov/gcov-9.C: New.
1383 * g++.dg/gcov/gcov-10.C: New.
1385 2011-11-19 Eric Botcazou <ebotcazou@adacore.com>
1387 PR rtl-optimization/51187
1388 * reorg.c (relax_delay_slots): Do not consider a jump useless if there
1389 is a barrier between the jump and its target label.
1391 2011-11-19 Patrick Marlier <patrick.marlier@gmail.com>
1394 * tracer.c (ignore_bb_p): Don't copy GIMPLE_TRANSACTION.
1396 2011-11-19 Eric Botcazou <ebotcazou@adacore.com>
1398 * expmed.c (store_bit_field_1): Revert bogus formatting change.
1400 2011-11-19 Richard Earnshaw <rearnsha@arm.com>
1403 * config/arm/arm.c (neon_disambiguate_copy): Correctly handle partial
1404 overlap of src and dest operands.
1406 2011-11-19 Iain Sandoe <iains@gcc.gnu.org>
1408 * config/darwin.h (ASM_DEBUG_SPEC): New.
1409 * config/darwin9.h (ASM_DEBUG_SPEC): New.
1411 2011-11-18 Jan Hubicka <jh@suse.cz>
1413 * config/i386/i386.c (atom_cost): Disable SSE loop until alignment
1416 2011-11-18 Anatoly Sokolov <aesok@post.ru>
1418 * config/cris/cris.c (cris_function_value_regno_p): Make static.
1419 (TARGET_FUNCTION_VALUE_REGNO_P): Define.
1420 * config/cris/cris.h (FUNCTION_VALUE_REGNO_P): Remove.
1421 * config/cris/cris-protos.h (cris_function_value_regno_p): Remove.
1423 2011-11-18 H.J. Lu <hongjiu.lu@intel.com>
1426 * doc/extend.texi: Document __builtin_ia32_movnti64.
1428 * config/i386/emmintrin.h (_mm_stream_si64): New.
1430 * config/i386/i386-builtin-types.def: Add VOID_FTYPE_PLONGLONG_LONGLONG.
1432 * config/i386/i386.c (ix86_builtins): Add IX86_BUILTIN_MOVNTI64.
1433 (bdesc_special_args): Update __builtin_ia32_movnti. Add
1434 __builtin_ia32_movnti64.
1435 (ix86_expand_special_args_builtin): Handle
1436 VOID_FTYPE_PLONGLONG_LONGLONG.
1438 * config/i386/i386.md (UNSPEC_MOVNTI): New.
1440 * config/i386/sse.md (sse2_movntsi): Renamed to ...
1441 (sse2_movnti<mode>): This.
1443 2011-11-18 Georg-Johann Lay <avr@gjlay.de>
1446 * config/avr/avr.h (base_arch_s): Add field n_segments.
1447 (ADDR_SPACE_PGM1, ADDR_SPACE_PGM2, ADDR_SPACE_PGM3,
1448 ADDR_SPACE_PGM4, ADDR_SPACE_PGM5, ADDR_SPACE_PGMX): New address spaces.
1449 (AVR_HAVE_ELPM, AVR_HAVE_ELPMX): New defines.
1450 (INIT_EXPANDERS): New define.
1451 * config/avr/avr-protos.h (avr_mem_pgmx_p): New.
1452 (avr_init_expanders): New.
1453 (avr_emit_movmemhi, avr_out_movmem): New.
1454 (avr_xload_libgcc_p): New.
1455 * config/avr/avr-c.c (avr_register_target_pragmas): Register
1456 address spaces __pgm1, __pgm2, __pgm3, __pgm4 __pgm5, __pgmx.
1457 (avr_cpu_cpp_builtins): Add built-in defines __PGM1,
1458 __PGM2, __PGM3, __PGM4, __PGM5, __PGMX.
1459 * config/avr/avr-devices.c (avr_arch_types): Set field n_segments.
1461 * config/avr/avr.c (AVR_SECTION_PROGMEM): Change define to cover
1462 3 bits instead of just 1.
1463 (xstring_empty, xstring_e, rampz_rtx): New static GTYed variables.
1464 (progmem_section): Change from section to array of sections.
1465 (progmem_section_prefix): New static variable.
1466 (avr_file_start): Print set for __RAMPZ__
1467 (avr_option_override): Move initialization of RTXes from here...
1468 (avr_init_expanders): ...to this new function.
1469 (avr_pgm_segment): New static function.
1470 (avr_decl_pgm_p): Handle error_mark_node.
1471 (avr_mem_pgmx_p, avr_decl_pgmx_p): New static functions.
1472 (avr_out_xload, avr_find_unused_d_reg): New static functions.
1473 (expand_prologue, expand_epilogue): Use rampz_rtx.
1474 (print_operand): Hande CONST_STRING.
1475 (avr_xload_libgcc_p): New static function.
1476 (avr_out_lpm_no_lpmx, avr_out_lpm): Handle ELPM.
1477 (avr_progmem_p): Return 2 for 24-bit flash address space.
1478 (avr_out_sbxx_branch): Clean-up code from ASn macros.
1479 (out_movqi_r_mr, out_movqi_mr_r): Ditto. And recognize RAMPZ's
1480 address and print symbolically.
1481 (avr_asm_named_section, avr_section_type_flags,
1482 avr_encode_section_info, avr_asm_select_section,
1483 avr_addr_space_address_mode, avr_addr_space_pointer_mode,
1484 avr_addr_space_legitimate_address_p, avr_addr_space_convert,
1485 avr_addr_space_legitimize_address): Handle new address spaces.
1486 (avr_output_progmem_section_asm_op): New static function.
1487 (avr_asm_init_sections): Initialize progmem_section[].
1488 (adjust_insn_length): Handle ADJUST_LEN_XLOAD, ADJUST_LEN_MOVMEM.
1489 (avr_const_address_lo16): New static function.
1490 (avr_assemble_integer): Use it to handle 3-byte integers.
1491 (avr_emit_movmemhi, avr_out_movmem): New functions.
1493 * config/avr/predicates.md (nox_general_operand): Handle new
1495 * config/avr/avr.md (unspec): Add UNSPEC_MOVMEM.
1496 (adjust_len): Add xload, movmem.
1497 (SP_ADDR): New define_constants.
1498 (isa): Add "lpm", "lpmx", "elpm", "elpmx".
1499 (enabled): Handle them.
1500 (load<mode>_libgcc): New expander.
1501 (*load.<mode>.libgcc): Rename to load_<mode>_libgcc.
1502 (xload8_A, xload<mode>_A): New insn-and-splits.
1503 (xload_8, xload_<mode>_libgcc, xload_<mode>, loadmem_elpm): New insns.
1504 (mov<mode>): Handle new address spaces.
1505 (movmemhi): Rewrite using avr_emit_movmemhi.
1506 (MOVMEM_r_d): New mode attribute.
1507 (movmem_<mode>, movmem_qi_elpm): New insns.
1508 (setmemhi, *clrmemqi, *clrmemhi, strlenhi, *strlenhi): Unquote
1509 C-code. Use label instead of hard-coded instrunction lengths.
1511 2011-11-18 Martin Jambor <mjambor@suse.cz>
1513 PR tree-optimization/50605
1514 * gimple.c (is_gimple_ip_invariant_address): Also handle MEM_REFs
1515 of IPA invariant decls.
1517 2011-11-18 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1519 * tree-outof-ssa.c (insert_back_edge_copies): Add call to
1520 mark_dfs_back_edges.
1522 2011-11-18 Iain Sandoe <iains@gcc.gnu.org>
1525 * configure.ac: Remove ranlib special-casing for Darwin.
1526 * configure: Regenerate.
1528 2011-11-18 Uros Bizjak <ubizjak@gmail.com>
1530 PR tree-optimization/51118
1531 * fold-const.c (fold_checksum_tree): Check for TS_TYPED structure
1532 before using TREE_TYPE accessor on expr.
1534 2011-11-17 Jan Hubicka <jh@suse.cz>
1537 * i386.c (atom_cost): Fix 32bit memset description.
1538 (expand_set_or_movmem_via_loop_with_iter): Output proper bounds check
1540 (expand_movmem_epilogue): Handle epilogues up to size 15 w/o producing
1542 (decide_alg): sse_loop is not useable wthen SSE2 is disabled;
1543 when not optimizing always use rep movsb or lincall; do not produce
1544 word sized loops when optimizing memset for size (to avoid need
1545 for large constants).
1546 (ix86_expand_movmem): Get into sync with ix86_expand_setmem;
1547 choose unroll factors better; always do 128bit moves when producing
1548 SSE loops; do not produce loopy epilogue when size is too small.
1549 (promote_duplicated_reg_to_size): Do not look into desired alignments
1550 when doing vector expansion.
1551 (ix86_expand_setmem): Track better when promoted value is available;
1552 choose unroll factors more sanely. Output loopy epilogue only
1555 2011-11-17 Steve Ellcey <sje@cup.hp.com>
1558 * output.h (fprint_w): Remove.
1559 * final.c (fprint_w): Remove.
1560 (output_addr_const): Change fprint_w back to fprintf.
1562 2011-11-17 Andrew Pinski <apinski@cavium.com>
1564 * config/host-linux.c (TRY_EMPTY_VM_SPACE): Define for MIPS.
1566 2011-11-17 Andrew MacLeod <amacleod@redhat.com>
1568 * builtins.c (expand_builtin): Remove 4th parameter representing
1569 weak/strong mode when __atomic_compare_exchange becomes a library call.
1571 2011-11-17 Richard Henderson <rth@redhat.com>
1573 * config/rs6000/rs6000.c (rs6000_expand_atomic_compare_and_swap): Get
1574 new pseudo for target after convert_modes.
1575 (rs6000_expand_atomic_exchange, rs6000_expand_atomic_op): Likewise.
1577 2011-11-17 Richard Henderson <rth@redhat.com>
1579 * config/rs6000/rs6000.c (rs6000_expand_atomic_compare_and_swap): Get
1580 new pseudo for target after convert_modes.
1581 (rs6000_expand_atomic_exchange, rs6000_expand_atomic_op): Likewise.
1583 2011-11-17 Richard Henderson <rth@redhat.com>
1585 * builtins.c (expand_builtin_mem_thread_fence): Remove.
1586 (expand_builtin_mem_signal_fence): Remove.
1587 (expand_builtin_atomic_thread_fence): Use expand_mem_thread_fence.
1588 (expand_builtin_sync_synchronize): Likewise.
1589 (expand_builtin_atomic_signal_fence): Use expand_mem_signal_fence.
1590 * optabs.c (expand_asm_memory_barrier): Split out from
1591 expand_builtin_mem_signal_fence.
1592 (expand_mem_thread_fence): New, a combination of code from
1593 expand_builtin_mem_thread_fence and expand_builtin_sync_synchronize.
1594 (expand_mem_signal_fence): Moved and renamed from
1595 expand_builtin_mem_signal_fence.
1596 (expand_atomic_exchange): Use expand_mem_thread_fence.
1597 (expand_atomic_load, expand_atomic_store): Likewise.
1598 * expr.h, optabs.h: Update decls.
1600 2011-11-17 Bin Cheng <bin.cheng@arm.com>
1602 PR rtl-optimization/50663
1603 * cprop.c (implicit_set_indexes): New global variable.
1604 (insert_set_in_table): Add additional parameter and record implicit
1606 (hash_scan_set): Add additional parameter and pass it to above.
1607 (hash_scan_insn): Pass false to hash_scan_set.
1608 (compute_hash_table_work): Pass true to hash_scan_set.
1609 (compute_cprop_data): Add implicit set to AVIN of block which the
1610 implicit set is recorded for.
1611 (one_cprop_pass): Handle implicit_set_indexes array.
1613 2011-11-17 Michael Matz <matz@suse.de>
1617 * tree-ssa-live.c (mark_all_vars_used_1): Recurse only for decls of
1619 (remove_unused_locals): Ditto.
1621 2011-11-16 Richard Henderson <rth@redhat.com>
1623 * config/alpha/alpha.c (alpha_split_compare_and_swap_12): Copy
1624 ALIAS_SET_MEMORY_BARRIER when creating a new memory.
1625 (alpha_split_atomic_exchange_12): Likewise.
1626 * config/rs6000/rs6000.c (rs6000_adjust_atomic_subword): Likewise.
1628 2011-11-16 Joseph Myers <joseph@codesourcery.com>
1630 * config/i386/i386elf.h (PREFERRED_DEBUGGING_TYPE): Remove.
1632 2011-11-16 David Edelsohn <dje.gcc@gmail.com>
1634 * config/rs6000/rs6000.c (rs6000_xcoff_section_type_flags):
1635 Default to minimum alignment if decl is NULL.
1637 2011-11-16 Uros Bizjak <ubizjak@gmail.com>
1639 * config/i386/sse.md (round<mode>2_vec_pack_sfix): Optimize V2DFmode
1641 (<sse4_1>_round<ssemodesuffix>_vec_pack_sfix<avxsizesuffix>): Ditto.
1643 2011-11-16 Venkataramanan Kumar <venkataramanan.kumar@amd.com>
1645 * doc/invoke.texi: Document AMD bdver1 and btver1.
1647 2011-11-16 Richard Earnshaw <rearnsha@arm.com>
1648 Bernd Schmidt <bernds@coudesourcery.com>
1649 Sebastian Huber <sebastian.huber@embedded-brains.de>
1652 * config/arm/arm.c (store_multiple_sequence): Avoid cases where
1653 the base reg is stored iff compiling for Thumb1.
1655 2011-11-16 Razya Ladelsky <razya@il.ibm.com>
1657 PR tree-optimization/49960
1658 * tree-data-ref.c (initialize_data_dependence_relation): Add
1660 Remove call to compute_self_dependence.
1661 (compute_affine_dependence): Remove the !DDR_SELF_REFERENCE condition.
1662 (compute_self_dependence): Remove old code. Add call to
1663 compute_affine_dependence.
1664 (compute_all_dependences): Remove call to compute_self_dependence.
1665 Add call to compute_affine_dependence.
1667 2011-11-16 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
1670 * expmed.c (store_bit_field_1): Use extract_bit_field on big
1671 endian targets if the source cannot be exactly covered by word
1674 2011-11-15 Joseph Myers <joseph@codesourcery.com>
1676 * config/i386/i386elf.h (ASM_OUTPUT_ASCII): Change STRING_LIMIT to
1679 2011-11-15 Richard Henderson <rth@redhat.com>
1681 * config/alpha/alpha.c (alpha_pre_atomic_barrier): New.
1682 (alpha_post_atomic_barrier): New.
1683 (alpha_split_atomic_op): New memmodel argument; honor it.
1684 (alpha_split_compare_and_swap): Take array of operands. Honor
1685 memmodel; always set bool output
1686 (alpha_expand_compare_and_swap_12): Similarly.
1687 (alpha_split_compare_and_swap_12): Similarly.
1688 (alpha_split_atomic_exchange): Similarly. Rename from
1689 alpha_split_lock_test_and_set.
1690 (alpha_expand_atomic_exchange_12): Similarly. Rename from
1691 alpha_expand_lock_test_and_set_12.
1692 (alpha_split_atomic_exchange_12): Similarly. Rename from
1693 alpha_split_lock_test_and_set_12.
1694 * config/alpha/alpha-protos.h: Update.
1695 * config/alpha/alpha.md (UNSPECV_CMPXCHG): New.
1696 * config/alpha/constraints.md ("w"): New.
1697 * config/alpha/predicates.md (mem_noofs_operand): New.
1698 * config/alpha/sync.md (atomic_compare_and_swap<mode>): Rename from
1699 sync_compare_and_swap<mode>; add the new parameters.
1700 (atomic_exchange<mode>): Update from sync_test_and_set<mode>.
1701 (atomic_fetch_<op><mode>): Update from sync_old_<op><mode>.
1702 (atomic_<op>_fetch<mode>): Update from sync_new_<op><mode>.
1703 (atomic_<op><mode>): Update from sync_<op><mode>.
1705 2011-11-16 Tom de Vries <tom@codesourcery.com>
1707 * tree-ssa-tail-merge.c (replace_block_by): Add frequency of bb2 to bb1.
1709 2011-11-16 Alan Modra <amodra@gmail.com>
1711 PR rtl-optimization/51051
1713 * function.c (thread_prologue_and_epilogue_insns): Guard
1714 emitting return with single_succ_p test.
1716 2011-11-15 Joern Rennecke <joern.rennecke@embecosm.com>
1718 * config/epiphany/epiphany.md (mov<mode>cc): Fix code to
1719 get mode from CMP_OP1 if CMP_OP0 is VOIDmode.
1721 2011-11-15 Uros Bizjak <ubizjak@gmail.com>
1723 * config/i386/sse.md (vec_pack_trunc_v2df): Optimize sequence for AVX.
1724 (vec_pack_sfix_trunc_v2df): Ditto.
1725 (vec_pack_sfix_v2df): Ditto.
1726 (vec_pack_sfix_trunc_v4df): Generate fix_truncv4dfv4si2 and
1727 avx_vec_concatv8si patterns.
1728 (vec_pack_sfix_v4df): Generate avx_cvtpd2dq256 and
1729 avx_vec_concatv8si patterns.
1731 2011-11-15 Torvald Riegel <triegel@redhat.com>
1733 * c-parser.c (c_parser_transaction_expression): Require parentheses
1734 when parsing transaction expressions.
1736 2011-11-15 Tristan Gingold <gingold@adacore.com>
1738 * incpath.c (get_added_cpp_dirs): New function.
1739 * incpath.h (get_added_cpp_dirs): Declare.
1740 * config/vms/vms-c.c (vms_c_register_includes): New function.
1741 (vms_std_modules): New variable.
1742 * config/vms/vms.h (TARGET_EXTRA_INCLUDES): Define.
1743 (vms_c_register_includes): Declare.
1745 2011-11-15 Tristan Gingold <gingold@adacore.com>
1747 * c-family/c-pragma.h (pragma_extern_prefix): Declare.
1748 * c-family/c-pragma.c (pragma_extern_prefix): Make it public.
1749 * config/vms/vms-c.c (saved_extern_prefix): New variable.
1750 (vms_pragma_extern_prefix): New function.
1751 (vms_c_register_pragma): Register vms_pragma_extern_prefix.
1753 2011-11-15 Georg-Johann Lay <avr@gjlay.de>
1756 * config/avr/avr.h (ADDR_SPACE_PGM): New address spaces.
1757 (REGISTER_TARGET_PRAGMAS): New define.
1758 * config/avr/avr-protos.h (avr_mem_pgm_p): New.
1759 (avr_load_libgcc_p): New.
1760 (avr_register_target_pragmas): New.
1761 (asm_output_external_libcall): Remove.
1762 (avr_log_t): Add field "progmem". Order alphabetically.
1763 * config/avr/avr-log.c (avr_log_set_avr_log): Set avr_log.progmem.
1764 * config/avr/avr-c.c (langhooks.h): New include.
1765 (avr_register_target_pragmas): New function. Register address
1767 (avr_cpu_cpp_builtins): Add built-in define __PGM.
1769 * config/avr/avr.c: Include "c-family/c-common.h".
1770 (TARGET_LEGITIMATE_ADDRESS_P): Remove define.
1771 (TARGET_LEGITIMIZE_ADDRESS): Remove define.
1772 (TARGET_ADDR_SPACE_SUBSET_P): Define to...
1773 (avr_addr_space_subset_p): ...this new static function.
1774 (TARGET_ADDR_SPACE_CONVERT): Define to...
1775 (avr_addr_space_convert): ...this new static function.
1776 (TARGET_ADDR_SPACE_ADDRESS_MODE): Define to...
1777 (avr_addr_space_address_mode): ...this new static function.
1778 (TARGET_ADDR_SPACE_POINTER_MODE): Define to...
1779 (avr_addr_space_pointer_mode): ...this new static function.
1780 (TARGET_ADDR_SPACE_LEGITIMATE_ADDRESS_P): Define to...
1781 (avr_addr_space_legitimate_address_p): ...this new static function.
1782 (TARGET_ADDR_SPACE_LEGITIMIZE_ADDRESS): Define to...
1783 (avr_addr_space_legitimize_address): ...this new static function.
1784 (avr_mode_code_base_reg_class): Handle address spaces.
1785 (avr_regno_mode_code_ok_for_base_p): Ditto.
1786 (lpm_addr_reg_rtx, lpm_reg_rtx, all_regs_rtx): New static variables.
1787 (avr_option_override): Initialize them.
1788 (output_reload_in_const): Use all_regs_rtx. Fix signedness of loop
1790 (avr_pgm_segment): New static function.
1791 (avr_decl_pgm_p, avr_mem_pgm_p): New static functions.
1792 (avr_out_lpm, avr_out_lpm_no_lpmx): New static functions.
1793 (output_movqi, output_movhi, output_movsisf, avr_out_movpsi): Call
1794 avr_out_lpm to handle loads from progmem.
1795 (avr_load_libgcc_p): New static function.
1796 (avr_progmem_p): Test if decl is in flash.
1797 (avr_pgm_pointer_const_p): New static function.
1798 (avr_nonconst_pointer_addrspace): New static function.
1799 (avr_pgm_check_var_decl): New static function.
1800 (avr_insert_attributes): Use it. Change error message to report
1801 cause (progmem or address space) when code wants to write to flash.
1802 (avr_section_type_flags): Unset section flag SECTION_BSS for
1805 * config/avr/predicates.md (nop_general_operand): New predicate.
1806 (nox_general_operand): New predicate.
1807 * config/avr/avr.md (LPM_REGNO): New define_constant.
1808 (load<mode>_libgcc): New expander.
1809 (*load.<mode>.libgcc): New insn.
1810 (mov<mode>): Handle loads from non-generic AS.
1811 (movmemhi): Ditto. Propagate address space information to newly
1813 (movqi_insn, *movhi, *movpsi, *movsi, *movsf): Change predicate #1
1814 to nox_general_operand.
1815 (ashrqi3, ashrhi3, ashrsi3): Change predicate #1 to
1816 nop_general_operand.
1817 (ashlqi3, *ashlqi3, ashlhi3, ashlsi3): Ditto.
1818 (lshrqi3, *lshrqi3, lshrhi3, lshrsi3): Ditto.
1819 (split-lpmx): New split.
1820 (*ashlhi3_const, *ashlsi3_const, *ashrhi3_const, *ashrsi3_const,
1821 *lshrhi3_const, *lshrsi3_const): Indent, unquote C.
1823 2011-11-15 Maxim Kuvyrkov <maxim@codesourcery.com>
1825 * ipa-cp.c (ipa_value_from_jfunc): Make global.
1826 (ipa_cst_from_jfunc): Remove, use ipa_value_from_jfunc instead.
1827 (get_indirect_edge_target): Rename, make global.
1828 (devirtualization_time_bonus, estimate_local_effects,)
1829 (ipcp_discover_new_direct_edges): Update.
1830 * ipa-inline-analysis.c (evaluate_conditions_for_edge):
1831 Generalize to also handle types. Rename to ...
1832 (evaluate_properties_for_edge): Use instead of
1833 evaluate_conditions_for_edge.
1834 (estimate_edge_devirt_benefit): New function.
1835 (estimate_calls_size_and_time): Use it.
1836 (estimate_node_size_and_time, estimate_ipcp_clone_size_and_time,)
1837 (inline_merge_summary): Update.
1838 (do_estimate_edge_time, do_estimate_edge_growth): Update. Calculate
1839 parameter information at the call site and pass it on to subroutines.
1840 * tree-inline.c (estimate_num_insns): Distinguish between direct and
1842 (init_inline_once): Set size and time costs or indirect calls.
1843 * tree-inline.h (eni_weights): Add indirect_call_cost.
1845 2011-11-15 Tom de Vries <tom@codesourcery.com>
1847 PR tree-optimization/51005
1848 * tree-ssa-tail-merge.c (delete_basic_block_same_succ): Rename to
1849 mark_basic_block_deleted.
1850 (update_worklist): Inline purge_bbs.
1851 (purge_bbs, unlink_virtual_phi, update_vuses, vop_at_entry)
1852 (delete_block_update_dominator_info): Remove.
1853 (replace_block_by): Remove update_vops parameter. Partially evaluate
1854 for update_vops == false.
1855 (apply_clusters): Remove update_vops parameter. Remove update_vops
1856 argument in replace_block_by call.
1857 (update_debug_stmts): Remove MAY_HAVE_DEBUG_STMTS test.
1858 (tail_merge_optimize): Remove update_vops argument to apply_clusters.
1859 Remove call to purge_bbs. Add calls to calculate_dominance_info and
1860 free_dominance_info. Add MAY_HAVE_DEBUG_STMTS before calling
1861 update_debug_stmts. Mark vop var for renaming, if necessary.
1863 2011-11-15 Bernd Schmidt <bernds@codesourcery.com>
1865 PR rtl-optimization/51051
1866 * cfgrtl.c (cfg_layout_can_merge_blocks_p): Return FALSE if the
1867 move would cause fallthrough into the exit block.
1869 2011-11-14 Richard Henderson <rth@redhat.com>
1871 * config/rs6000/rs6000.c (emit_load_locked): Assert the mode is
1873 (emit_store_conditional): Likewise.
1874 (rs6000_pre_atomic_barrier, rs6000_post_atomic_barrier): New.
1875 (rs6000_adjust_atomic_subword): New.
1876 (rs6000_mask_atomic_subword, rs6000_finish_atomic_subword): New.
1877 (rs6000_expand_atomic_op): Rename from rs6000_emit_sync; rewrite
1878 for pre-reload expansion.
1879 (rs6000_split_compare_and_swap, rs6000_split_compare_and_swapqhi,
1880 rs6000_expand_compare_and_swapqhi): Merge into ...
1881 (rs6000_expand_atomic_compare_and_swap): ... here. New function.
1882 rs6000_split_lock_test_and_set; expand immediately. Handle
1884 * config/rs6000/rs6000.md (UNSPEC_LWSYNC): Move and rename
1885 from UNSPECV_LWSYNC.
1886 * config/rs6000/sync.md (fetchopsi_constr, fetchopdi_constr): Remove.
1887 (mem_thread_fence): New.
1888 (hwsync): Rename from memory_barrier.
1889 (*hwsync): Rename from *sync_internal.
1890 (lwsync, *lwsync): Mirror hwsync implementation.
1891 (isync): Don't reference memory.
1893 (atomic_load<INT>, atomic_store<INT>): New.
1894 (ATOMIC): New mode iterator.
1895 (load_locked<ATOMIC>): Rename from load_locked_<GPR>.
1896 (store_conditional<ATOMIC>): Rename from store_conditional_<GPR>.
1897 (sync_compare_and_swap<GPR>): Remove.
1898 (sync_compare_and_swaphi, sync_compare_and_swapqi): Remove.
1899 (sync_compare_and_swapqhi_internal): Remove.
1900 (sync_lock_test_and_set<GPR>): Remove.
1901 (sync_<FETCHOP><INT1>): Remove.
1902 (*sync_<FETCHOP>si_internal, *sync_<FETCHOP>di_internal): Remove.
1903 (sync_nand<INT1>, *sync_nand<GPR>_internal): Remove.
1904 (sync_old_<FETCHOP><GPR>): Remove.
1905 (*sync_old_<FETCHOP>si_internal): Remove.
1906 (*sync_old_<FETCHOP>di_internal): Remove.
1907 (sync_old_nand<INT1>): Remove.
1908 (*sync_old_nand<GPR>_internal): Remove.
1909 (sync_new_<FETCHOP><GPR>): Remove.
1910 (*sync_new_<FETCHOP>si_internal): Remove.
1911 (*sync_new_<FETCHOP>di_internal): Remove.
1912 (sync_new_nand<INT1>): Remove.
1913 (*sync_new_nand<GPR>_internal): Remove.
1914 (*atomic_andsi, *atomic_anddi): Remove.
1915 (*sync_addshort_internal, *sync_subshort_internal): Remove.
1916 (*sync_andsi_internal, *sync_boolsi_internal): Remove.
1917 (*sync_boolcshort_internal): Remove.
1918 (sync_lock_release<INT1>): Remove.
1919 (atomic_compare_and_swap<INT1>): New.
1920 (atomic_exchange<INT1>): New.
1921 (atomic_<FETCHOP><INT1>, atomic_nand<INT1>): New.
1922 (atomic_fetch_<FETCHOP><INT1>, atomic_fetch_nand<INT1>): New.
1923 (atomic_<FETCHOP>_fetch<INT1>, atomic_nand_fetch<INT1>): New.
1925 2011-11-14 Uros Bizjak <ubizjak@gmail.com>
1927 * config/i386/sse.md (round<mode>2): Use register_operand for
1928 operand 1 predicate.
1929 (round<mode>2_sfix): Ditto.
1930 (round<mode>2_vec_pack_sfix): Ditto for operand 1 and operand 2.
1932 2011-11-14 Nathan Sidwell <nathan@acm.org>
1934 * gcov.c (canonicalize_name): Protect use of S_ISLNK.
1936 2011-11-14 Jan Hubicka <jh@suse.cz>
1939 * config/i386/i386.c (core cost model): Correct pasto.
1941 2011-11-14 Uros Bizjak <ubizjak@gmail.com>
1943 * config/i386/sse.md (round<mode>2_sfix): New expander.
1944 (round<mode>2_vec_pack_sfix): Ditto.
1945 (<sse4_1>_round<ssemodesuffix>_sfix<avxsizesuffix>): Ditto.
1946 (<sse4_1>_round<ssemodesuffix>_vec_pack_sfix<avxsizesuffix>): Ditto.
1947 * config/i386/builtin-types.def (V4SI_FTYPE_V4SF_ROUND,
1948 V8SI_FTYPE_V8SF_ROUND, V4SI_FTYPE_V2DF_V2DF_ROUND,
1949 V8SI_FTYPE_V4DF_V4DF_ROUND): New builtin types.
1950 * config/i386/i386.c (ix86_builtins): Add
1951 IX86_BUILTIN_{FLOORPD,CEILPD,ROUNDPD_AZ}_VEC_PACK_SFIX{,256} and
1952 IX86_BUILTIN_{FLOORPS,CEILPS,ROUNDPS_AZ}_SFIX{,256} defines.
1953 (bdesc_args): Add __builtin_ia32_{floorpd,ceilpd}_vec_pack_sfix{,256},
1954 __builtin_ia32_roundpd_az_vec_pack_sfix{,256},
1955 __builtin_ia32_{floorps,ceilps}_sfix{,256}and
1956 __builtin_ia32_roundps_az_sfix{,256} descriptions.
1957 (ix86_expand_sse_round_vec_pack_sfix): New.
1958 (ix86_expand_args_builtin): Handle V4SI_FTYPE_V4SF_ROUND,
1959 V8SI_FTYPE_V8SF_ROUND, V4SI_FTYPE_V2DF_V2DF_ROUND and
1960 V8SI_FTYPE_V4DF_V4DF_ROUND types. Check last argument of
1961 CODE_FOR_sse4_1_roundpd_vec_pack_sfix, CODE_FOR_sse4_1_roundps_sfix,
1962 CODE_FOR_avx_roundpd_vec_pack_sfix256 and CODE_FOR_avx_roundps_sfix256.
1963 (ix86_builtin_vectorized_function): Handle
1964 BUILT_IN_{I,L,LL}FLOOR{,F}, BUILT_IN_{I,L,LL}CEIL{,F} and
1965 BUILT_IN_{I,L,LL}ROUND{,F}
1967 2011-11-14 Jan Hubicka <jh@suse.cz>
1970 * cgraph.h (varpool_externally_visible_p): Declare.
1971 (varpool_add_new_variable): Declare.
1972 * tree-emultls.c (get_emutls_init_templ_addr,
1973 new_emutls_decl): Use varpool_add_new_variable.
1974 * ipa.c (varpool_externally_visible_p): Export.
1975 * varpool.c (varpool_add_new_variable): New function.
1977 2011-11-14 Michael Zolotukhin <michael.v.zolotukhin@gmail.com>
1978 Jan Hubicka <jh@suse.cz>
1980 * config/i386/i386.h (processor_costs): Add second dimension to
1981 stringop_algs array.
1982 * config/i386/i386.c (cost models): Initialize second dimension of
1983 stringop_algs arrays.
1984 (core_cost): New costs based on generic64 costs with updated stringop
1986 (promote_duplicated_reg): Add support for vector modes, add
1988 (promote_duplicated_reg_to_size): Likewise.
1989 (processor_target): Set core costs for core variants.
1990 (expand_set_or_movmem_via_loop_with_iter): New function.
1991 (expand_set_or_movmem_via_loop): Enable reuse of the same iters in
1992 different loops, produced by this function.
1993 (emit_strset): New function.
1994 (expand_movmem_epilogue): Add epilogue generation for bigger sizes,
1995 use SSE-moves where possible.
1996 (expand_setmem_epilogue): Likewise.
1997 (expand_movmem_prologue): Likewise for prologue.
1998 (expand_setmem_prologue): Likewise.
1999 (expand_constant_movmem_prologue): Likewise.
2000 (expand_constant_setmem_prologue): Likewise.
2001 (decide_alg): Add new argument align_unknown. Fix algorithm of
2002 strategy selection if TARGET_INLINE_ALL_STRINGOPS is set; Skip sse_loop
2003 (decide_alignment): Update desired alignment according to chosen move
2005 (ix86_expand_movmem): Change unrolled_loop strategy to use SSE-moves.
2006 (ix86_expand_setmem): Likewise.
2007 (ix86_slow_unaligned_access): Implementation of new hook
2008 slow_unaligned_access.
2009 * config/i386/i386.md (strset): Enable half-SSE moves.
2010 * config/i386/sse.md (vec_dupv4si): Add expand for vec_dupv4si.
2011 (vec_dupv2di): Add expand for vec_dupv2di.
2013 2011-11-14 Dimitrios Apostolou <jimis@gmx.net>
2017 * config/elfos.h (TARGET_ASM_INTERNAL_LABEL): Remove.
2018 * varasm.c (default_elf_internal_label): Remove.
2020 2011-11-14 Kai Tietz <ktietz@redhat.com>
2022 * gcov.c (generate_results): Add missing semicolon and correct indent.
2024 2011-11-14 Ira Rosen <ira.rosen@linaro.org>
2027 * tree-vect-stmts.c (vectorizable_condition): Initialize comp_vectype.
2029 2011-11-14 Mingjie Xing <mingjie.xing@gmail.com>
2031 * doc/invoke.texi (Wunused-result): Change @pxref{Variable Attributes}
2032 to @pxref{Function Attributes}.
2034 2011-11-13 Oleg Endo <oleg.endo@t-online.de>
2037 * config/sh/sh.h (IS_LITTLE_ENDIAN_OPTION, UNSUPPORTED_SH2A):
2039 (DRIVER_SELF_SPECS): Use new macros to filter out unsupported options
2040 taking the default configuration into account.
2042 2011-11-13 Jonathan Wakely <jwakely.gcc@gmail.com>
2044 * doc/install.texi (Specific): Fix anchor for
2045 x86_64-*-solaris2.1[0-9]*.
2047 2011-11-13 Nathan Sidwell <nathan@acm.org>
2049 * gcov.c (source_prefix, source_length): New globals.
2050 (flag_relative_only): Likewise.
2051 (print_usage, options, process_args): Update.
2052 (generate_results): Use coverage.name, check
2053 flag_relative_only. Adjust messages.
2054 (find_source): Check source_prefix.
2055 (output_lines): Use coverage.name, adjust messages.
2056 * doc/gcov.texi (Invoking Gcov): Document new options.
2058 2011-11-12 Jason Merrill <jason@redhat.com>
2061 * gimplify.c (gimplify_target_expr): Add a clobber to the cleanup.
2062 (gimplify_modify_expr): Don't try to simplify it.
2064 2011-11-12 Dimitrios Apostolou <jimis@gmx.net>
2067 * config/elfos.h (STRING_ASM_OP): Define.
2068 * varasm.c (default_elf_asm_output_limited_string): Use it.
2071 * gcc/configure.ac: Check for stpcpy declaration.
2072 * system.h: Declare it if not.
2073 * config.in, configure: Regenerate.
2075 2011-11-12 Richard Henderson <rth@redhat.com>
2077 * config/rs6000/rs6000.md (fix_trunc<SFDF>si2_stfiwx): Use
2078 nonimmediate_operand for the destination.
2079 (fixuns_trunc<SFDF>si2_stfiwx): Likewise.
2081 * builtins.c (expand_builtin_compare_and_swap): If target is const0,
2082 don't pass the target to expand_atomic_compare_and_swap.
2083 (expand_builtin_atomic_compare_exchange): Likewise.
2085 * config/sh/linux.h (TARGET_INIT_LIBFUNCS): New.
2086 * config/sh/sh.c (sh_init_sync_libfuncs): New.
2088 * config/pa/pa-linux.h (TARGET_SYNC_LIBCALL): New.
2089 * config/pa/pa.h (TARGET_SYNC_LIBCALL): New default.
2090 * config/pa/pa.c (pa_init_libfuncs): Use init_sync_libfuncs.
2092 2011-11-12 Iain Sandoe <iains@gcc.gnu.org>
2095 * config/rs6000/rs6000.c (rs6000_legitimize_reload_address):
2096 Only expand a symbol ref. into an access when the entity is defined
2099 2011-11-12 Jan Hubicka <jh@suse.cz>
2102 * cgraphunit.c (output_weakrefs): Output really only weakrefs.
2104 2011-11-12 Nathan Sidwell <nathan@acm.org>
2106 * gcov.c (struct name_map): New.
2107 (names, n_names, a_names): New global vars.
2108 (print_usage): Adjust usage.
2109 (generate_results): Canonicalize main file name.
2110 (release_structures): Adjust.
2111 (name_search, name_sort): New callbacks.
2112 (find_source): Look for and create a canonical name.
2113 (canonicalize_name): New.
2114 (make_gcov_file_name): Reimplement and fix mangling.
2116 * doc/gcov.texi: Update documentation about path preservation.
2118 2011-11-11 David S. Miller <davem@davemloft.net>
2120 * config/sparc/linux.h (ASM_GENERATE_INTERNAL_LABEL): Delete.
2121 * config/sparc/linux64.h (ASM_GENERATE_INTERNAL_LABEL): Delete.
2123 2011-11-11 Jakub Jelinek <jakub@redhat.com>
2125 * config/i386/i386-protos.h (ix86_maybe_emit_epilogue_vzeroupper):
2127 * config/i386/i386.c (ix86_maybe_emit_epilogue_vzeroupper): New
2129 (ix86_expand_epilogue): Use it.
2130 * config/i386/i386.md (return, simple_return): Call it in the
2133 PR tree-optimization/51091
2134 * tree-stdarg.c (execute_optimize_stdarg): Ignore TREE_CLOBBER_P
2135 rhs also in the va_list_simple_ptr case.
2137 PR tree-optimization/51058
2138 * tree-vect-slp.c (vect_remove_slp_scalar_calls): New function.
2139 (vect_schedule_slp): Call it.
2140 * tree-vect-stmts.c (vectorizable_call): If slp_node != NULL,
2141 don't replace scalar calls with setting of their lhs to zero here.
2143 PR tree-optimization/51074
2144 * fold-const.c (vec_cst_ctor_to_array, fold_vec_perm): New functions.
2145 (fold_binary_loc): Handle VEC_EXTRACT_EVEN_EXPR,
2146 VEC_EXTRACT_ODD_EXPR, VEC_INTERLEAVE_HIGH_EXPR and
2147 VEC_INTERLEAVE_LOW_EXPR with VECTOR_CST or CONSTRUCTOR operands.
2148 (fold_ternary_loc): Handle VEC_PERM_EXPR with VECTOR_CST or
2149 CONSTRUCTOR operands.
2150 * tree-ssa-propagate.c (valid_gimple_rhs_p): Handle ternary
2152 * tree-vect-generic.c (lower_vec_perm): Mask sel_int elements
2153 to 0 .. 2 * elements - 1.
2155 2011-11-11 Richard Henderson <rth@redhat.com>
2157 * gimple-pretty-print.c (dump_gimple_omp_atomic_load): Dump needed.
2158 (dump_gimple_omp_atomic_store): Likewise.
2159 * optabs.c (can_atomic_exchange_p): New.
2160 * optabs.h (can_atomic_exchange_p): Declare.
2161 * omp-low.c (expand_omp_atomic_load): Implement.
2162 (expand_omp_atomic_store): Likewise.
2163 (expand_omp_atomic): Update for new arguments to load/store.
2165 2011-11-11 Eric Botcazou <ebotcazou@adacore.com>
2167 * cprop.c: Adjust outdated comments throughout.
2168 (hash_scan_set): Rename PAT parameter into SET.
2169 (cprop_pavloc): Rename into...
2170 (cprop_avloc): ...this.
2171 (cprop_absaltered): Rename into...
2172 (cprop_kill): ...this.
2173 (alloc_cprop_mem): Adjust for above renaming.
2174 (free_cprop_mem): Likewise.
2175 (compute_cprop_data): Likewise.
2176 (compute_local_properties): Rename TRANSP parameter into KILL and
2177 adjust throughout. Rework comments.
2178 (try_replace_reg): Fix long line.
2179 (cprop_jump): Likewise.
2180 (constprop_register): Fix prototype and take INSN last.
2181 (cprop_insn): Adjust calls to above function. Fix long lines.
2182 (bypass_block): Likewise.
2183 (one_cprop_pass): Likewise.
2185 2011-11-11 Uros Bizjak <ubizjak@gmail.com>
2187 * config/alpha/elf.h (ELF_ASCII_ESCAPES): Rename from ESCAPES.
2188 (ELF_STRING_LIMIT): Rename from STRING_LIMIT.
2190 2011-11-11 Jason Merrill <jason@redhat.com>
2192 * varasm.c (default_elf_asm_output_limited_string): #ifdef
2194 (default_elf_asm_output_ascii): Likewise.
2196 2011-11-11 Uros Bizjak <ubizjak@gmail.com>
2199 * config/i386/constraints.md (j): New address constraint.
2200 * config/i386/predicates.md (lea_address_operand): Redefine as
2202 * config/i386/i386.md (*lea_3_zext): Use "j" constraint for operand 1.
2203 (*lea_4_zext): Ditto.
2205 2011-11-11 David S. Miller <davem@davemloft.net>
2208 2011-11-05 David S. Miller <davem@davemloft.net>
2210 * config/sparc/sparc.md (UNSPEC_SHORT_LOAD): New unspec.
2211 (zero-extend_v8qi_vis, zero_extend_v4hi_vis): New expanders.
2212 (*zero_extend_v8qi_<P:mode>_insn,
2213 *zero_extend_v4hi_<P:mode>_insn): New insns.
2214 * config/sparc/sparc.c (vector_init_move_words,
2215 vector_init_prepare_elts, sparc_expand_vector_init_vis2,
2216 sparc_expand_vector_init_vis1): New functions.
2217 (vector_init_bshuffle): Rewrite to handle more cases and make use
2218 of locs[] array prepared by vector_init_prepare_elts.
2219 (vector_init_fpmerge, vector_init_faligndata): Delete.
2220 (sparc_expand_vector_init): Rewrite using new infrastructure.
2222 2011-11-11 Jakub Jelinek <jakub@redhat.com>
2224 * opts-common.c (generate_canonical_option): Free opt_text
2225 it it has been allocated here and not stored anywhere.
2227 * tree-vect-slp.c (vect_free_slp_tree): Also free SLP_TREE_CHILDREN
2229 (vect_create_new_slp_node): Don't allocate node before checking stmt
2231 (vect_free_oprnd_info): Remove FREE_DEF_STMTS argument, always
2232 free def_stmts vectors and additionally free oprnd_info.
2233 (vect_build_slp_tree): Adjust callers. Call it even if
2234 stop_recursion. If vect_create_new_slp_node or
2235 vect_build_slp_tree fails, properly handle freeing memory.
2236 If it succeeded, clear def_stmts in oprnd_info.
2238 2011-11-11 Dimitrios Apostolou <jimis@gmx.net>
2240 * defaults.h (ASM_OUTPUT_LABEL, ASM_OUTPUT_INTERNAL_LABEL): Expand on
2241 multiple lines for readability.
2242 * output.h (fprint_whex, fprint_w, fprint_ul, sprint_ul): Declare.
2243 (ASM_OUTPUT_LABELREF): Call fputs instead of asm_fprintf.
2244 * dwarf2asm.c (dw2_assemble_integer, dw2_asm_output_data,
2245 dw2_asm_data_uleb128, dw2_asm_delta_uleb128,
2246 dw2_asm_delta_sleb128): Convert fprintf calls to the new functions.
2247 * dwarf2out.c (dwarf2out_source_line): Call fprint_ul instead of
2249 * final.c (fprint_whex, fprint_w, fprint_ul, sprint_ul): New functions
2250 serving as fast replacements for fprintf integer to string conversions.
2251 (sprint_ul_rev): Internal helper for the above.
2252 (output_addr_const) <CONST_INT>: Don't use fprintf.
2253 * varasm.c: Fix some whitespace inconsistencies.
2254 (default_elf_asm_output_limited_string): New.
2255 (default_elf_asm_output_ascii): Likewise.
2256 (default_elf_internal_label): Likewise.
2257 * config/elfos.h (ASM_GENERATE_INTERNAL_LABEL): Use sprint_ul and
2258 stpcpy instead of sprintf.
2259 (TARGET_ASM_INTERNAL_LABEL): Define as default_elf_internal_label.
2260 (ELF_ASCII_ESCAPES, ELF_STRING_LIMIT): Rename from old ESCAPES and
2261 STRING_LIMIT macros.
2262 (ASM_OUTPUT_LIMITED_STRING, ASM_OUTPUT_ASCII): Call functions that
2263 provide the same functionality.
2264 * config/i386/i386.c (print_reg): Call fprint_ul instead of fprintf.
2266 2011-11-10 Andrew MacLeod <amacleod@redhat.com>
2268 * doc/extend.texi: Document __atomic_test_and_set and __atomic_clear.
2270 2011-11-10 Andrew MacLeod <amacleod@redhat.com>
2273 * builtins.c (expand_builtin_atomic_clear): New. Expand atomic_clear.
2274 (expand_builtin_atomic_test_and_set): New. Expand atomic test_and_set.
2275 (expand_builtin): Add cases for test_and_set and clear.
2276 * sync-builtins.def (BUILT_IN_ATOMIC_TEST_AND_SET): New.
2277 (BUILT_IN_ATOMIC_CLEAR): New.
2279 2011-11-10 Roberto Agostino Vitillo <ravitillo@lbl.gov>
2282 * dwarf2out.c (set_cur_line_info_table): Restore the last is_stmt
2283 value in the current line table.
2285 2011-11-10 Nathan Sidwell <nathan@acm.org>
2287 * gcov.c (struct function_info): Make src an index, not a pointer.
2288 (struct source_info): Remove index and next source fields.
2289 (fn_end): New static var.
2290 (sources_index): Remove.
2291 (sources): Now a pointer to an array, not a list.
2292 (n_sources, a_sources): New.
2293 (process_file): Adjust for changes to read_graph_file. Insert
2294 functions into source lists and check line numbers here.
2295 (generate_results): Only allocate lines for sources with
2296 contents. Adjust for source array.
2297 (release_structures): Likewise.
2298 (find_source): Return source index, adjust for source array.
2299 (read_graph_file): Return function list. Don't insert into source
2301 (read_count_file): Take list of functions.
2302 (solve_flow_graph): Reverse the arc lists here.
2303 (add_line_counts): Adjust for source array.
2305 2011-11-10 Jakub Jelinek <jakub@redhat.com>
2308 * tree-object-size.c (addr_object_size): Check TREE_CODE of
2309 MEM_REF's operand rather than code of the MEM_REF itself.
2311 2011-11-10 Andrew MacLeod <amacleod@redhat.com>
2313 PR rtl-optimization/51040
2314 * optabs.c (expand_atomic_fetch_op): Patchup code for NAND should
2315 be AND followed by NOT.
2316 * builtins.c (expand_builtin_atomic_fetch_op): Patchup code for NAND
2317 should be AND followed by NOT.
2319 2011-11-10 Jakub Jelinek <jakub@redhat.com>
2321 * vec.h (VEC_BASE): If base is at offset 0 in the structure,
2322 use &(P)->base even if P is NULL.
2324 PR rtl-optimization/51023
2325 * combine.c (simplify_comparison) <case SIGN_EXTEND>: Don't use
2326 val_signbit_known_clear_p for signed comparison narrowing
2327 optimization. Don't check for non-VOIDmode, use
2328 HWI_COMPUTABLE_MODE_P macro.
2329 <case ZERO_EXTEND>: Don't check for non-VOIDmode.
2330 Optimize even when const_op is equal to GET_MODE_MASK (mode),
2331 don't optimize if const_op is negative.
2333 2011-11-10 Richard Guenther <rguenther@suse.de>
2335 PR tree-optimization/51042
2336 * tree-ssa-pre.c (phi_translate_1): Avoid recursing on
2337 self-referential expressions. Refactor code to avoid duplication.
2339 2011-11-10 Richard Guenther <rguenther@suse.de>
2341 PR tree-optimization/51070
2342 * tree-loop-distribution.c (generate_builtin): Do not replace
2343 the loop with a builtin if the partition contains statements which
2344 results are used outside of the loop.
2345 (pass_loop_distribution): Verify and collect.
2347 2011-11-10 Richard Guenther <rguenther@suse.de>
2349 PR tree-optimization/51030
2350 * tree-ssa-phiopt.c (jump_function_from_stmt): New function.
2351 (value_replacement): Use it to handle trivial non-empty
2352 intermediate blocks.
2354 2011-11-10 Richard Guenther <rguenther@suse.de>
2357 * gimple.c (gimple_has_side_effects): Remove checking code
2358 that doesn't belong here.
2360 2011-11-10 Ira Rosen <ira.rosen@linaro.org>
2362 PR tree-optimization/51058
2363 * tree-vect-slp.c (vect_get_constant_vectors): Handle CALL_EXPR.
2365 2011-11-10 Jakub Jelinek <jakub@redhat.com>
2367 PR tree-optimization/51000
2368 * tree-vect-patterns.c (vect_recog_bool_pattern): If adding
2369 a pattern stmt for a bool store, adjust DR_STMT too.
2370 Don't handle bool conversions to single bit precision lhs.
2371 * tree-vect-stmts.c (vect_remove_stores): If next is a pattern
2372 stmt, remove its related stmt and free its stmt_vinfo.
2373 (free_stmt_vec_info): Free also pattern stmt's vinfo and
2374 pattern def stmt's vinfo.
2375 * tree-vect-loop.c (destroy_loop_vec_info): Don't try to
2376 free pattern stmt's vinfo here.
2377 (vect_transform_loop): When calling vect_remove_stores,
2378 do gsi_next first and don't call gsi_remove. If not strided
2379 store, free stmt vinfo for gsi_stmt (si) rather than stmt.
2381 2011-11-09 Jakub Jelinek <jakub@redhat.com>
2384 * config/i386/i386.c (expand_vec_perm_interleave2): If d->vmode is
2385 V4SImode, !TARGET_SSE2 and punpck[lh]* is needed, change dremap.vmode
2388 * function.h (requires_stack_frame_p): New prototype.
2389 * function.c (requires_stack_frame_p): No longer static.
2390 * config/i386/i386.c (ix86_finalize_stack_realign_flags): If
2391 stack_realign_fp was just a conservative guess for a function
2392 which doesn't use sp/fp/argp at all, clear frame_pointer_needed
2393 and stack realignment.
2395 2011-11-09 Paolo Carlini <paolo.carlini@oracle.com>
2397 PR preprocessor/51061
2398 * incpath.c (merge_include_chains): Make sure to not pass null
2399 pointers to free_path.
2401 2011-11-09 Paolo Carlini <paolo.carlini@oracle.com>
2403 * doc/invoke.texi ([Option Summary, C++ Language Options]):
2404 Add -Wzero-as-null-pointer-constant.
2406 2011-11-07 Matthew Gretton-Dann <matthew.gretton-dann@arm.com>
2408 * config/arm/arm-cores.def: Add -mcpu=cortex-a7.
2409 * config/arm/arm-tables.opt: Regenerate.
2410 * config/arm/arm-tune.md: Likewise.
2411 * config/arm/bpabi.h (BE8_LINK_SPEC): Add Cortex A-7.
2412 * doc/invoke.texi: Document -mcpu=cortex-a7.
2414 2011-11-09 Richard Guenther <rguenther@suse.de>
2417 2011-11-09 Richard Guenther <rguenther@suse.de>
2419 * tree-cfg.c (verify_gimple_call): Verify that
2420 gimple_call_cannot_inline_p is returning a conservative
2421 correct result according to gimple_check_call_matching_types.
2423 2011-11-09 Richard Guenther <rguenther@suse.de>
2425 PR tree-optimization/51039
2426 * tree-cfg.c (verify_gimple_call): Verify that
2427 gimple_call_cannot_inline_p is returning a conservative
2428 correct result according to gimple_check_call_matching_types.
2429 * ipa-inline-analysis.c (estimate_function_body_sizes): Remove
2430 code dealing with un-inlinablility.
2431 * gimple-streamer-in.c (input_gimple_stmt): Update the
2434 2011-11-09 Richard Guenther <rguenther@suse.de>
2436 PR tree-optimization/51039
2437 * tree-inline.c (setup_one_parameter): Always perform a
2438 valid gimple type change.
2439 (declare_return_variable): Likewise.
2441 2011-11-09 Jakub Jelinek <jakub@redhat.com>
2443 * config/rs6000/vector.md (vcondv4sfv4si, vcondv4siv4sf,
2444 vconduv4sfv4si): New patterns.
2445 * config/rs6000/rs6000.c (rs6000_emit_vector_cond_expr): Handle
2446 different dest_mode from comparison mode.
2448 2011-11-09 Richard Guenther <rguenther@suse.de>
2450 * gimple-fold.c (canonicalize_constructor_val): Mark
2451 address-taken variables addressable.
2453 2011-11-09 Ira Rosen <ira.rosen@linaro.org>
2455 PR tree-optimization/51015
2456 * tree-vect-loop.c (vect_determine_vectorization_factor): Expect
2457 vectype to be set for pattern def stmts.
2459 2011-11-09 Alan Modra <amodra@gmail.com>
2461 * function.c (bb_active_p): Delete.
2462 (dup_block_and_redirect, active_insn_between): New functions.
2463 (convert_jumps_to_returns, emit_return_for_exit): New functions,
2465 (thread_prologue_and_epilogue_insns): ..here. Delete
2466 shadowing variables. Don't do prologue register clobber tests
2467 when shrink wrapping already failed. Delete all last_bb_active
2468 code. Instead compute tail block candidates for duplicating
2469 exit path. Remove these from antic set. Duplicate tails when
2470 reached from both blocks needing a prologue/epilogue and
2471 blocks not needing such.
2472 * ifcvt.c (dead_or_predicable): Test both flag_shrink_wrap and
2474 * bb-reorder.c (get_uncond_jump_length): Make global.
2475 * bb-reorder.h (get_uncond_jump_length): Declare.
2476 * cfgrtl.c (rtl_create_basic_block): Comment typo fix.
2477 (rtl_split_edge): Likewise. Warning fix.
2478 (rtl_duplicate_bb): New function.
2479 (rtl_cfg_hooks): Enable can_duplicate_block_p and duplicate_block.
2480 * Makefile.in (function.o): Update dependencies.
2482 2011-11-08 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
2483 Georg-Johann Lay <avr@gjlay.de>
2485 * doc/tm.texi.in (MODE_CODE_BASE_REG_CLASS): Add address space
2487 (REGNO_MODE_CODE_OK_FOR_BASE_P): Likewise.
2488 * doc/tm.texi: Regenerate.
2490 * config/cris/cris.h (MODE_CODE_BASE_REG_CLASS): Add address
2492 (REGNO_MODE_CODE_OK_FOR_BASE_P): Likewise.
2493 * config/bfin/bfin.h (MODE_CODE_BASE_REG_CLASS): Likewise.
2494 (REGNO_MODE_CODE_OK_FOR_BASE_P): Likewise.
2495 * config/avr/avr.h (MODE_CODE_BASE_REG_CLASS): Add address space
2497 (REGNO_MODE_CODE_OK_FOR_BASE_P): Ditto.
2498 * config/avr/avr-protos.h (avr_mode_code_base_reg_class): Ditto.
2499 (avr_regno_mode_code_ok_for_base_p): Ditto.
2500 * config/avr/avr.c (avr_mode_code_base_reg_class): Ditto.
2501 (avr_regno_mode_code_ok_for_base_p): Ditto.
2502 (avr_reg_ok_for_addr_p): Pass AS down to
2503 avr_regno_mode_code_ok_for_base_p.
2505 * addresses.h (base_reg_class): Add address space argument.
2506 Pass to MODE_CODE_BASE_REG_CLASS.
2507 (ok_for_base_p_1): Add address space argument. Pass to
2508 REGNO_MODE_CODE_OK_FOR_BASE_P.
2509 (regno_ok_for_base_p): Add address space argument. Pass to
2512 * regrename.c (scan_rtx_address): Add address space argument.
2513 Pass address space to regno_ok_for_base_p and base_reg_class.
2514 Update recursive calls.
2515 (scan_rtx): Pass address space to scan_rtx_address.
2516 (build_def_use): Likewise.
2517 * regcprop.c (replace_oldest_value_addr): Add address space
2518 argument. Pass to regno_ok_for_base_p and base_reg_class.
2519 Update recursive calls.
2520 (replace_oldest_value_mem): Pass address space to
2521 replace_oldest_value_addr.
2522 (copyprop_hardreg_forward_1): Likewise.
2524 * reload.c (find_reloads_address_1): Add address space argument.
2525 Pass address space to base_reg_class and regno_ok_for_base_p.
2526 Update recursive calls.
2527 (find_reloads_address): Pass address space to base_reg_class,
2528 regno_ok_for_base_p, and find_reloads_address_1.
2529 (find_reloads): Pass address space to base_reg_class.
2530 (find_reloads_subreg_address): Likewise.
2532 * ira-costs.c (record_reg_classes): Update calls to base_reg_class.
2533 (ok_for_base_p_nonstrict): Add address space argument. Pass to
2535 (record_address_regs): Add address space argument. Pass to
2536 base_reg_class and ok_for_base_p_nonstrict. Update recursive calls.
2537 (record_operand_costs): Pass address space to record_address_regs.
2538 (scan_one_insn): Likewise.
2540 * caller-save.c (init_caller_save): Update call to base_reg_class.
2541 * ira-conflicts.c (ira_build_conflicts): Likewise.
2542 * reload1.c (maybe_fix_stack_asms): Likewise.
2544 2011-11-08 Michael Matz <matz@suse.de>
2546 * gengtype.c (write_field_root): Avoid out-of-scope access of newv.
2548 * tree-stdarg.c (execute_optimize_stdarg): Accept clobbers.
2550 * tree.h (TREE_CLOBBER_P): New macro.
2551 * gimple.h (gimple_clobber_p): New inline function.
2552 * gimplify.c (gimplify_bind_expr): Add clobbers for all variables
2553 that go out of scope and live in memory.
2554 * tree-ssa-operands.c (get_expr_operands): Transfer volatility also
2556 * cfgexpand.c (decl_to_stack_part): New static variable.
2557 (add_stack_var): Allocate it, and remember mapping.
2558 (fini_vars_expansion): Deallocate it.
2559 (stack_var_conflict_p): Add early outs.
2560 (visit_op, visit_conflict, add_scope_conflicts_1,
2561 add_scope_conflicts): New static functions.
2562 (expand_used_vars_for_block): Don't call add_stack_var_conflict, tidy.
2563 (expand_used_vars): Add scope conflicts.
2564 (expand_gimple_stmt_1): Expand clobbers to nothing.
2565 (expand_debug_expr): Ditto.
2567 * tree-pretty-print.c (dump_generic_node): Dump clobbers nicely.
2568 * tree-ssa-live.c (remove_unused_locals): Remove clobbers that
2569 refer to otherwise unused locals.
2570 * tree-sra.c (build_accesses_from_assign): Ignore clobbers.
2571 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Clobbers of
2572 SSA names aren't necessary.
2573 (propagate_necessity): Accept and ignore constructors on the rhs, tidy.
2574 * gimple.c (walk_gimple_op): Accept constructors like mem_rhs.
2575 * tree-ssa-structalias.c (find_func_aliases): Clobbers don't store
2577 * tree-ssa-sccvn.c (vn_reference_lookup_3): Ditto, in particular they
2578 don't zero-initialize something.
2579 * tree-ssa-phiopt.c (cond_if_else_store_replacement_1): Ignore
2580 clobber RHS, we don't want PHI nodes with those.
2582 2011-11-08 Jakub Jelinek <jakub@redhat.com>
2584 * config/i386/i386.c (ix86_preferred_simd_mode): Even for TARGET_AVX
2585 if not TARGET_PREFER_AVX128 return 32-byte integer vectors.
2587 2011-11-08 Richard Guenther <rguenther@suse.de>
2589 PR tree-optimization/51012
2590 * ipa-prop.c (update_indirect_edges_after_inlining): Fixup
2591 non-inlinable state.
2592 * cgraph.c (cgraph_make_edge_direct): Likewise.
2594 2011-11-08 Eric Botcazou <ebotcazou@adacore.com>
2596 PR rtl-optimization/47698
2597 * ifcvt.c (noce_operand_ok): Move around comment.
2599 2011-11-08 Richard Guenther <rguenther@suse.de>
2602 * lto-opts.c (append_to_collect_gcc_options): Split out from...
2603 (lto_write_options): ... here. Prepend frontend specific flags.
2605 2011-11-08 Jakub Jelinek <jakub@redhat.com>
2607 * config/i386/sse.md (*avx_vperm2f128<mode>3_nozero): Emit mask
2608 0x12 and 0x20 as vinsert[fi]128 instead of vperm2[fi]128.
2610 * tree-vect-stmts.c (vectorizable_call): Add SLP_NODE argument.
2611 Handle vectorization of SLP calls.
2612 (vect_analyze_stmt): Adjust caller, add call to it for SLP too.
2613 (vect_transform_stmt): Adjust vectorizable_call caller, remove
2615 * tree-vect-slp.c (vect_get_and_check_slp_defs): For calls start
2617 (vect_build_slp_tree): Allow CALL_EXPR.
2619 2011-11-08 Richard Guenther <rguenther@suse.de>
2621 * gimple-fold.c (canonicalize_constructor_val): Make sure
2622 we have referenced vars setup before adding to them.
2624 2011-11-07 Richard Henderson <rth@redhat.com>
2625 Aldy Hernandez <aldyh@redhat.com>
2626 Andrew MacLeod <amacleod@redhat.com>
2627 Torvald Riegel <triegel@redhat.com>
2629 Merged from transactional-memory.
2631 * gtm-builtins.def: New file.
2632 * trans-mem.c: New file.
2633 * trans-mem.h: New file.
2635 * opts.c (finish_options): Error out when using -flto and -fgnu-tm.
2637 * config/i386/i386.c: Define TARGET_VECTORIZE* transactional variants.
2638 (ix86_handle_tm_regparm_attribute, struct bdesc_tm,
2639 ix86_builtin_tm_load, ix86_builtin_tm_store,
2640 ix86_init_tm_builtins): New.
2641 (ix86_init_builtins): Initialize TM builtins.
2642 (struct ix86_attribute_table): Add "*tm regparm".
2643 * config/i386/i386-builtin-types.def (PV2SI): Define.
2645 Define V2SI_FTYPE_PCV2SI.
2646 Define V4SF_FTYPE_PCV4SF.
2647 Define V8SF_FTYPE_PCV8SF.
2648 Define VOID_PV2SI_V2SI.
2650 * doc/invoke.texi (C Dialect Options): Document -fgnu-tm and
2651 tm-max-aggregate-size.
2652 * doc/tm.texi.in: Add TARGET_VECTORIZE_BUILTIN_TM_LOAD and
2653 TARGET_VECTORIZE_BUILTIN_TM_STORE hooks.
2654 * doc/tm.texi: Regenerate.
2656 * attribs.c (apply_tm_attr): New.
2657 (init_attributes): Allow '*' prefix for overrides.
2658 (register_attribute): Likewise.
2659 * builtin-attrs.def (ATTR_TM_TMPURE, ATTR_TM_REGPARM): New.
2660 (ATTR_TM_NOTHROW_LIST, ATTR_TM_TMPURE_NOTHROW_LIST,
2661 ATTR_TM_PURE_TMPURE_NOTHROW_LIST, ATTR_TM_NORETURN_NOTHROW_LIST,
2662 ATTR_TM_CONST_NOTHROW_LIST, ATTR_TMPURE_MALLOC_NOTHROW_LIST,
2663 ATTR_TMPURE_NOTHROW_LIST): New.
2664 * builtin-types.def (BT_FN_I[1248]_VPTR, BT_FN_FLOAT_VPTR,
2665 BT_FN_DOUBLE_VPTR, BT_FN_LDOUBLE_VPTR, BT_FN_VOID_VPTR_I[1248],
2666 BT_FN_VOID_VPTR_FLOAT, BT_FN_VOID_VPTR_DOUBLE,
2667 BT_FN_VOID_VPTR_LDOUBLE, BT_FN_VOID_VPTR_SIZE): New.
2668 * builtins.def: Include gtm-builtins.def. Add comments regarding
2669 transactional memory synchronization.
2670 (DEF_TM_BUILTIN): New.
2671 * c-parser.c (struct c_parser): Add in_transaction.
2672 (c_parser_transaction, c_parser_transaction_expression,
2673 c_parser_transaction_cancel, c_parser_transaction_attributes): New.
2674 (c_parser_attribute_any_word): Split out from c_parser_attributes.
2675 (c_parser_statement_after_labels): Handle RID_TRANSACTION*.
2676 (c_parser_unary_expression): Same.
2677 * c-tree.h (c_finish_transaction): Declare.
2678 * c-typeck.c (c_finish_transaction): New.
2679 (build_function_call_vec): Call tm_malloc_replacement.
2680 * calls.c (is_tm_builtin): New.
2681 (flags_from_decl_or_type): Add ECF_TM_BUILTIN and ECF_TM_PURE.
2682 * cfgbuild.c (make_edges): Add edges for REG_TM notes.
2683 * cfgexpand.c (expand_call_stmt): Call
2684 mark_transaction_restart_calls.
2685 (gimple_expand_cfg): Free the tm_restart map.
2686 (mark_transaction_restart_calls): New.
2687 * cfgrtl.c (purge_dead_edges): Look for REG_TM notes.
2688 * cgraph.c (dump_cgraph_node): Handle tm_clone.
2689 * cgraph.h (struct cgraph_node): Add tm_clone field.
2690 (decl_is_tm_clone): New.
2691 (struct cgraph_local_info): Add tm_may_enter_irr.
2692 (cgraph_copy_node_for_versioning): Declare.
2693 * cgraphunit.c (cgraph_copy_node_for_versioning): Export;
2694 copy analyzed from old version.
2695 * combine.c (distribute_notes): Handle REG_TM notes.
2696 * common.opt: Add -fgnu-tm.
2697 * crtstuff.c (__TMC_LIST__, __TMC_END__): New.
2698 (__do_global_dtors_aux): Deregister clone table.
2699 (frame_dummy): Register clone table.
2700 * emit-rtl.c (try_split): Handle REG_TM. Early return if no function
2702 * gimple-low.c (lower_stmt): Handle GIMPLE_EH_ELSE and
2704 (gimple_stmt_may_fallthru): Handle GIMPLE_EH_ELSE.
2705 * gimple-pretty-print.c: Include trans-mem.h.
2706 (dump_gimple_fmt): Add %x.
2707 (dump_gimple_call): Dump arguments for calls to _ITM_beginTransaction.
2708 (dump_gimple_eh_else, dump_gimple_transaction): New.
2709 (dump_gimple_stmt): Handle GIMPLE_EH_ELSE and GIMPLE_TRANSACTION.
2710 * gimple.c (gimple_build_eh_else, gimple_build_transaction): New.
2711 (walk_gimple_seq): Honor removed_stmt. Document usage of removed_stmt
2713 (walk_gimple_op): Handle GIMPLE_TRANSACTION.
2714 (walk_gimple_stmt): Initialize and honor removed_stmt.
2715 Handle GIMPLE_EH_ELSE and GIMPLE_TRANSACTION.
2716 (gimple_copy): Handle GIMPLE_EH_ELSE and GIMPLE_TRANSACTION.
2717 * gimple.def (GIMPLE_TRANSACTION, GIMPLE_EH_ELSE): New.
2718 * gimple.h (struct gimple_statement_eh_else,
2719 gimple_statement_transaction, GTMA_*): New.
2720 (gimple_statement_d): Add gimple_statement_eh_else and
2722 (gimple_build_eh_else, gimple_build_transaction,
2723 gimple_fold_call, diagnose_tm_safe_errors): Declare.
2724 (get_call_expr_in): Remove prototype.
2725 (gimple_has_substatements): Add GIMPLE_EH_ELSE and GIMPLE_TRANSACTION.
2726 (gimple_eh_else_n_body, gimple_eh_else_e_body,
2727 gimple_eh_else_set_n_body, gimple_eh_else_set_e_body,
2728 gimple_transaction_body, gimple_transaction_label,
2729 gimple_transaction_label_ptr, gimple_transaction_subcode,
2730 gimple_transaction_set_body, gimple_transaction_set_label,
2731 gimple_transaction_set_subcode): New.
2732 (struct walk_stmt_info): Use BOOL_BITFIELD; add removed_stmt.
2733 * gimplify.c (create_tmp_var_name): Use clean_symbol_name.
2734 (voidify_wrapper_expr): Handle TRANSACTION_EXPR.
2735 (gimplify_transaction): New.
2736 (gimplify_expr): Handle TRANSACTION_EXPR.
2737 * gsstruct.def (GSS_EH_ELSE, GSS_TRANSACTION): New.
2738 * ipa-inline.c (can_inline_edge_p): Do not inline TM safe calling
2740 * Makefile.in: Add trans-mem.o and dependencies.
2741 (BUILTINS_DEF): Add gtm-builtins.def.
2742 (gimple-pretty-print.o): Depend on TRANS_MEM_H.
2743 (GTFILES): Add trans-mem.c.
2744 * omp-low.c (WALK_SUBSTMTS): Add GIMPLE_TRANSACTION.
2745 * output.h (record_tm_clone_pair, finish_tm_clone_pairs,
2746 get_tm_clone_pair): Declare.
2747 * params.def (PARAM_TM_MAX_AGGREGATE_SIZE): New.
2748 * passes.c (init_optimization_passes): Place transactional memory
2750 * print-tree.c (print_node): Dump tm-clone.
2751 * recog.c (peep2_attempt): Handle REG_TM.
2752 * reg-notes.def (TM): New.
2753 * rtlanal.c (alloc_reg_note): Handle REG_TM.
2754 * target.def (builtin_tm_load, builtin_tm_store): New.
2755 * targhooks.c (default_builtin_tm_load_store): New.
2756 * targhooks.h (default_builtin_tm_load_store): Declare.
2757 * timevar.def (TV_TRANS_MEM): New.
2758 * toplev.c (compile_file): Call finish_tm_clone_pairs.
2759 * tree-cfg.c (make_edges): Handle GIMPLE_TRANSACTION.
2760 (cleanup_dead_labels): Handle GIMPLE_TRANSACTION. Avoid unnecessary
2761 writes into the statements to update labels.
2762 (is_ctrl_altering_stmt): Add TM ending statements. Handle
2764 (verify_gimple_transaction): New.
2765 (verify_gimple_stmt): Handle GIMPLE_TRANSACTION.
2766 (verify_gimple_in_seq_2): Handle GIMPLE_EH_ELSE and GIMPLE_TRANSACTION.
2767 (gimple_redirect_edge_and_branch): Handle TM_TRANSACTION.
2768 (dump_function_to_file): Display [tm-clone] if applicable.
2769 * tree-eh.c (struct_ptr_eq): Make inline and move to tree.h.
2770 (struct_ptr_hash): Same.
2771 (collect_finally_tree): Handle GIMPLE_EH_ELSE.
2772 (replace_goto_queue_1): Likewise.
2774 (honor_protect_cleanup_actions): Handle GIMPLE_EH_ELSE.
2775 (lower_try_finally_nofallthru): Likewise.
2776 (lower_try_finally_onedest): Likewise.
2777 (lower_try_finally_copy): Likewise.
2778 (lower_try_finally_switch): Likewise.
2779 (lower_try_finally): Likewise.
2780 (decide_copy_try_finally): Likewise.
2781 (lower_eh_constructs_2): Likewise.
2782 (refactor_eh_r): Likewise.
2783 * tree-flow.h (struct gimple_df): Add tm_restart field.
2784 Define tm_restart_node.
2785 * tree-inline.c (remap_gimple_stmt): Handle GIMPLE_TRANSACTION.
2786 (estimate_num_insns): Likewise.
2787 (init_inline_once): Init tm_cost.
2788 * tree-inline.h (struct eni_weights_d): Add tm_cost.
2789 * tree-pass.h (pass_diagnose_tm_blocks, pass_lower_tm, pass_tm_init,
2790 pass_tm_mark, pass_tm_memopt, pass_tm_edges, pass_ipa_tm): Declare.
2791 * tree-pretty-print.c (dump_generic_node): Handle TRANSACTION_EXPR.
2792 * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Handle
2793 BUILT_IN_TM_MEMSET, BUILT_IN_TM_MEMCPY, BUILT_IN_TM_MEMMOVE.
2794 Add support for TM vector loads. Add support for TM logging builtins.
2795 (call_may_clobber_ref_p_1): Add support for vector stores.
2796 * tree-ssa-structalias.c (find_func_aliases): Add support for TM
2797 vector stores and loads. Handle BUILT_IN_TM_MEMSET,
2798 BUILT_IN_TM_MEMCPY, BUILT_IN_TM_MEMMOVE.
2799 * tree.c (strip_invariant_refs): Moved from gimple.c to here.
2800 (local_define_builtin): Handle ECF_TM_PURE.
2801 (build_common_builtin_nodes): Set __builtin_eh_pointer to ECF_TM_PURE.
2802 * tree.def (TRANSACTION_EXPR): New.
2803 * tree.h (strip_invariant_refs): Moved from gimple.h to here.
2804 (TRANSACTION_EXPR_BODY, TRANSACTION_EXPR_CHECK,
2805 TRANSACTION_EXPR_OUTER, TRANSACTION_EXPR_RELAXED,
2806 BUILTIN_TM_LOAD_STORE_P, BUILTIN_TM_LOAD_P, BUILTIN_TM_STORE_P,
2807 CASE_BUILT_IN_TM_LOAD, CASE_BUILT_IN_TM_STORE): New.
2808 (ECF_TM_PURE, ECF_TM_BUILTIN): New.
2809 (struct tree_function_decl): Add tm_clone_flag.
2810 (struct_ptr_eq, struct_ptr_hash): New.
2811 (apply_tm_attr): Declare.
2812 (is_tm_safe_or_pure): New.
2813 (build_tm_abort_call, is_tm_safe, is_tm_pure,
2814 is_tm_may_cancel_outer, is_tm_ending_fndecl, record_tm_replacement,
2815 tm_malloc_replacement): Declare.
2816 * varasm.c (tm_clone_hash): New.
2817 (record_tm_clone_pair, finish_tm_clone_pairs, get_tm_clone_pair,
2818 dump_tm_clone_to_vec, dump_tm_clone_pairs, tm_alias_pair_cmp): New.
2819 (struct tm_alias_pair): New. Declare VEC types for object.
2821 2011-11-07 Richard Henderson <rth@redhat.com>
2823 * optabs.h (OTI_sync_compare_and_swap, OTI_sync_lock_test_and_set,
2824 OTI_sync_old_add, OTI_sync_old_sub, OTI_sync_old_ior,
2825 OTI_sync_old_and, OTI_sync_old_xor, OTI_sync_old_nand,
2826 OTI_sync_new_add, OTI_sync_new_sub, OTI_sync_new_ior,
2827 OTI_sync_new_and, OTI_sync_new_xor, OTI_sync_new_nand): Move and
2828 rename from the direct_optab_index enum.
2829 (sync_compare_and_swap_optab, sync_lock_test_and_set_optab,
2830 sync_old_add_optab, sync_old_sub_optab, sync_old_ior_optab,
2831 sync_old_and_optab, sync_old_xor_optab, sync_old_nand_optab,
2832 sync_new_add_optab, sync_new_sub_optab, sync_new_ior_optab,
2833 sync_new_and_optab, sync_new_xor_optab, sync_new_nand_optab): Read
2834 from the optab_table, not the direct_optab_table.
2835 (init_sync_libfuncs): Declare.
2836 (can_compare_and_swap_p): Update parameters.
2837 * optabs.c (init_sync_libfuncs_1, init_sync_libfuncs): New.
2838 (can_compare_and_swap_p): Add allow_libcall parameter; if true,
2839 test for the legacy compare-and-swap libcall.
2840 (expand_atomic_exchange): Use the legacy test-and-set libcall.
2841 (expand_atomic_compare_and_swap): Use the legacy CAS libcall.
2842 (struct atomic_op_functions): Update for optab type changes.
2843 (maybe_emit_op): Likewise.
2844 (expand_atomic_fetch_op): Use the legacy fetch-op libcalls.
2845 * builtins.c (fold_builtin_atomic_always_lock_free): Update call
2846 to can_compare_and_swap_p.
2847 * omp-low.c (expand_omp_atomic_fetch_op): Likewise.
2848 (expand_omp_atomic_pipeline): Likewise.
2849 * genopinit.c (optabs): Make sync_old_*_optab, sync_new_*_optab,
2850 sync_compare_and_swap_optab, sync_lock_test_and_set_optab regular
2852 * doc/md.texi (sync_compare_and_swap): Update docs for libcalls.
2854 2011-11-07 Jakub Jelinek <jakub@redhat.com>
2856 * config/i386/i386-bultin-types.def (V8SI_FTYPE_V4DF_V4DF): Add.
2857 * config/i386/i386.c (enum ix86_builtins): Add
2858 IX86_BUILTIN_VEC_PACK_SFIX256.
2859 (bdesc_args): Add __builtin_ia32_vec_pack_sfix256.
2860 (ix86_expand_args_builtin): Handle V8SI_FTYPE_V4DF_V4DF.
2861 (ix86_builtin_vectorized_function): Also vectorize lrint using
2862 256-bit vectors for -mavx.
2864 2011-11-07 Anatoly Sokolov <aesok@post.ru>
2866 * config/cris/constraints.md: New file.
2867 * config/cris/cris.h (REG_CLASS_FROM_LETTER, CONSTRAINT_LEN,
2868 CRIS_CONST_OK_FOR_LETTER_P, CONST_OK_FOR_CONSTRAINT_P,
2869 CONST_DOUBLE_OK_FOR_LETTER_P, EXTRA_MEMORY_CONSTRAINT,
2870 EXTRA_CONSTRAINT, EXTRA_CONSTRAINT_Q, EXTRA_CONSTRAINT_R,
2871 EXTRA_CONSTRAINT_T, EXTRA_CONSTRAINT_S, EXTRA_CONSTRAINT_U): Remove.
2872 * config/cris/cris.c: Incule "tm-constrs.h".
2873 (cris_print_operand): Use satisfies_constraint_O.
2874 (cris_normal_notice_update_cc, cris_rtx_costs): Use
2875 satisfies_constraint_I.
2876 (cris_address_cost): Use satisfies_constraint_L.
2877 * config/cris/cris.md: Include "constraints.md".
2878 (*mov_side<mode>, *mov_sidesisf, *mov_side<mode>_mem,
2879 *mov_sidesisf_mem, *clear_side<mode>, *ext_sideqihi,
2880 *ext_side<mode>si, *op_side<mode>, *op_swap_side<mode>,
2881 *extopqihi_side, *extop<mode>si_side, *extopqihi_swap_side,
2882 *extop<mode>si_swap_side): Use satisfies_constraint_N and
2883 satisfies_constraint_J.
2884 (moversideqi movemsideqi mover2side peephole2): Use
2885 satisfies_constraint_N and satisfies_constraint_J.
2886 (andu peephole2): Use satisfies_constraint_I and
2887 satisfies_constraint_O.
2889 2011-11-07 Uros Bizjak <ubizjak@gmail.com>
2891 * config/i386/i386.c (ix86_builtin_vectorized_function): Handle
2892 BUILT_IN_IRINT, BUILT_IN_IRINTF, BUILT_IN_LLRINT and BUILT_IN_LLRINTF.
2894 2011-11-07 Andrew MacLeod <amacleod@redhat.com>
2896 * optabs.c (get_atomic_op_for_code): Fill in optab table at runtime so
2897 SWITCHABLE_TARGET can change the values during compilation.
2898 (expand_atomic_fetch_op): Handle parameter change ripples for
2899 get_atomic_op_for_code call.
2901 2011-11-07 Andrew MacLeod <amacleod@redhat.com>
2903 * doc/extend.texi: Docuemnt behaviour change for __atomic_exchange and
2905 * optabs.c (expand_atomic_exchange): Expand to __sync_lock_test_and_set
2906 only when originated from that builtin.
2907 (expand_atomic_store): Expand to __sync_lock_release when originated
2909 * builtins.c (expand_builtin_sync_lock_test_and_set): Add flag that
2910 expand_atomic_exchange call originated from here.
2911 (expand_builtin_sync_lock_release): Add flag that expand_atomic_store
2912 call originated from here.
2913 (expand_builtin_atomic_exchange): Add origination flag.
2914 (expand_builtin_atomic_store): Add origination flag.
2915 * expr.h (expand_atomic_exchange, expand_atomic_store): Add boolean
2916 parameters to indicate implementation fall back options.
2918 2011-11-07 Georg-Johann Lay <avr@gjlay.de>
2920 * config/avr/avr.c (output_reload_in_const): Can handle CONSTANT_P
2921 now, not only CONST_INT and CONST_DOUBLE.
2922 (output_movqi): Use output_reload_in_const.
2923 (output_reload_inhi): Ditto.
2924 (output_reload_insisf): Move assertion to output_reload_in_const.
2925 (avr_out_reload_inpsi): Ditto.
2927 2011-11-07 Nathan Sidwell <nathan@acm.org>
2929 * gcov.c (object_summary): Replace with ...
2930 (object_runs): ... this.
2931 (process_file): Remove functions with no data.
2932 (generate_results): Ignore files with no lines.
2933 (release_function): New helper, broken out of ...
2934 (release_structures): ... here. Use it.
2935 (read_count_file): Adjust for new data file format.
2936 (output_lines): Use object_runs.
2937 * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Obsolete.
2938 (struct gcov_ctr_info): Move definition.
2939 (struct gcov_fn_info): Add key field, use gcov_ctr_info for
2941 (struct gcov_info): Add merge function array, remove mask and
2942 counts. Trailing array is array of pointers to function info.
2943 * coverage.c (struct function_list): Replace counter numbers with
2944 counter arrays. Add fndecl field. GTYify.
2945 (counts_entry): Remove chain workspace.
2946 (functions_head): GTYify.
2947 (prg_n_ctrs): Remove.
2949 (tree_ctr_tables): Remove.
2950 (read_counts_file): Cope with blank entries and expect program
2951 summaries before functions. Don't warn on missing entries.
2952 (coverage_counter_alloc): Allocate individual function arrays.
2953 (tree_coverage_counter_ref, tree_coverage_counter_addr): Adjust
2954 for individual function arrays.
2955 (coverage_end_function): GTYify function list object. Finalize
2956 function's counter arrays.
2957 (build_var): New. Create a counter-related variable with
2958 appropriate linkage.
2959 (build_fn_info_type): Adjust for new runtime structure.
2960 (build_fn_info_value): Rename to ...
2961 (build_fn_info): ... here. Build new format data.
2962 (build_ctr_info_type, build_ctr_info_value): Remove.
2963 (build_info_type): New. Build new format data structure.
2964 (build_info): Adjust for new format data.
2965 (create_coverage): Likewise.
2966 * gcov-dump.c (tag_function): Recognize placeholders.
2968 2011-11-07 Georg-Johann Lay <avr@gjlay.de>
2970 * config/avr/constraints.md (Cm2): New constraint for int -2.
2971 * config/avr/avr.md (addqi3): Use it. New alternatives for +/-2.
2972 (*negqihi2): New insn.
2974 2011-11-07 H.J. Lu <hongjiu.lu@intel.com>
2976 * dwarf2cfi.c (dwarf2out_frame_debug_expr): Check
2977 HARD_FRAME_POINTER_REGNUM instead of hard_frame_pointer_rtx in Rule 18.
2979 2011-11-07 Jakub Jelinek <jakub@redhat.com>
2981 * config/i386/i386.c (ix86_expand_builtin): If gather mask
2982 argument is known to have all high bits set, pass pc_rtx as
2983 second argument to the expander instead of op0.
2984 * config/i386/sse.md (*avx2_gathersi<mode>_2,
2985 *avx2_gatherdi<mode>_2): New patterns.
2986 * config/i386/avx2intrin.h (_mm256_i32gather_pd,
2987 _mm256_i64gather_pd, _mm256_i32gather_ps): Set mask using
2988 _mm256_cmp_pd with zero vector arguments and _CMP_EQ_OQ instead of
2991 PR tree-optimization/50789
2992 * tree-vect-stmts.c (process_use): Add force argument, avoid
2993 exist_non_indexing_operands_for_use_p check if true.
2994 (vect_mark_stmts_to_be_vectorized): Adjust callers. Handle
2995 STMT_VINFO_GATHER_P.
2996 (gen_perm_mask): New function.
2997 (perm_mask_for_reverse): Use it.
2998 (reverse_vec_element): Rename to...
2999 (permute_vec_elements): ... this. Add Y and MASK_VEC arguments,
3000 generalize for any permutations.
3001 (vectorizable_load): Adjust caller. Handle STMT_VINFO_GATHER_P.
3002 * target.def (TARGET_VECTORIZE_BUILTIN_GATHER): New hook.
3003 * doc/tm.texi.in (TARGET_VECTORIZE_BUILTIN_GATHER): Document it.
3004 * doc/tm.texi: Regenerate.
3005 * tree-data-ref.c (initialize_data_dependence_relation,
3006 compute_self_dependence): No longer static.
3007 * tree-data-ref.h (initialize_data_dependence_relation,
3008 compute_self_dependence): New prototypes.
3009 * tree-vect-data-refs.c (vect_check_gather): New function.
3010 (vect_analyze_data_refs): Detect possible gather load data refs.
3011 * tree-vectorizer.h (struct _stmt_vec_info): Add gather_p field.
3012 (STMT_VINFO_GATHER_P): Define.
3013 (vect_check_gather): New prototype.
3014 * config/i386/i386-builtin-types.def: Add types for alternate
3016 * config/i386/sse.md (AVXMODE48P_DI): Remove.
3017 (VEC_GATHER_MODE): Rename mode_attr to...
3018 (VEC_GATHER_IDXSI): ... this.
3019 (VEC_GATHER_IDXDI, VEC_GATHER_SRCDI): New mode_attrs.
3020 (avx2_gathersi<mode>, *avx2_gathersi<mode>): Use <VEC_GATHER_IDXSI>
3021 instead of <VEC_GATHER_MODE>.
3022 (avx2_gatherdi<mode>): Use <VEC_GATHER_IDXDI> instead of
3023 <AVXMODE48P_DI> and <VEC_GATHER_SRCDI> instead of VEC_GATHER_MODE
3024 on src and mask operands.
3025 (*avx2_gatherdi<mode>): Likewise. Use VEC_GATHER_MODE iterator
3026 instead of AVXMODE48P_DI.
3027 (avx2_gatherdi<mode>256, *avx2_gatherdi<mode>256): Removed.
3028 * config/i386/i386.c (enum ix86_builtins): Add
3029 IX86_BUILTIN_GATHERALTSIV4DF, IX86_BUILTIN_GATHERALTDIV8SF,
3030 IX86_BUILTIN_GATHERALTSIV4DI and IX86_BUILTIN_GATHERALTDIV8SI.
3031 (ix86_init_mmx_sse_builtins): Create those builtins.
3032 (ix86_expand_builtin): Handle those builtins and adjust expansions
3033 of other gather builtins.
3034 (ix86_vectorize_builtin_gather): New function.
3035 (TARGET_VECTORIZE_BUILTIN_GATHER): Define.
3037 2011-11-07 Uros Bizjak <ubizjak@gmail.com>
3039 * config/i386/f16cintrin.h: Remove extra _X86INTRIN_H_INCLUDED check.
3041 2011-11-07 Tristan Gingold <gingold@adacore.com>
3043 * config.gcc (*-*-*vms*): Set c_target_objs, cxx_target_objs.
3044 * config/vms/t-vms: Add vms-c.o rule.
3045 * config/vms/vms-c.c: New file.
3046 * config/vms/vms-protos.h (vms_c_register_pragma): New prototype.
3047 * config/vms/vms.h (REGISTER_TARGET_PRAGMAS): Define.
3049 2011-11-07 Tristan Gingold <gingold@adacore.com>
3051 * config/alpha/vms.h (TARGET_OBJECT_SUFFIX,
3052 TARGET_EXECUTABLE_SUFFIX, TARGET_OS_CPP_BUILTINS,
3053 TARGET_ABI_OPEN_VMS, LONG_TYPE_SIZE, ADA_LONG_TYPE_SIZE,
3054 POINTER_SIZE, POINTERS_EXTEND_UNSIGNED): Move to config/vms/vms.h
3055 (SUBTARGET_OS_CPP_BUILTINS): Define.
3056 (TARGET_DEFAULT): Tune according to POINTER_SIZE.
3057 (MASK_RETURN_ADDR): Define in 64 bit mode.
3058 * config/ia64/vms.h: Likewise.
3059 * config/vms/vms.h: New file.
3060 * config/vms/vms64.h: New file.
3061 * config/alpha/vms64.h: Removed.
3062 * config/ia64/vms64.h: Removed.
3063 * config.gcc (alpha64-dec-*vms*, alpha*-dec-*vms*, ia64-hp-*vms*):
3064 Adjust for above change.
3066 2011-11-07 Enkovich Ilya <ilya.enkovich@intel.com>
3069 * config/i386/i386-protos.h (ix86_use_lea_for_mov): New.
3070 * config/i386/i386.c (ix86_use_lea_for_mov): Likewise.
3071 * config/i386/i386.md (movsi_internal): Emit lea if profitable.
3072 (movdi_internal_rex64): Likewise.
3074 2011-11-07 Sergey Ostanevich <sergos.gnu@gmail.com>
3076 PR rtl-optimization/47698
3077 * ifcvt.c (noce_operand_ok): Return false for mems with side effects.
3079 2011-11-07 Tristan Gingold <gingold@adacore.com>
3081 * common/config/alpha/alpha-common.c (alpha_option_init_struct):
3083 (TARGET_OPTION_INIT_STRUCT): Define.
3084 * config/alpha/alpha.c (TARGET_MIN_ANCHOR_OFFSET)
3085 (TARGET_MAX_ANCHOR_OFFSET)
3086 (TARGET_USE_BLOCKS_FOR_CONSTANT_P): Redefine.
3088 2011-11-06 Quentin Neill <quentin.neill@amd.com>
3090 Fix r180999, update ChangeLog
3091 * config.gcc: Add f16cintrin.h.
3092 * config/i386/f16cintrin.h: Add missing endif.
3094 2011-11-06 Sebastian Huber <sebastian.huber@embedded-brains.de>
3096 * config.gcc (arm*-*-rtemseabi*): New target.
3097 * config/arm/rtems-eabi.h: New.
3098 * config/arm/t-rtems-eabi: New.
3100 2011-11-06 David S. Miller <davem@davemloft.net>
3102 * config/sparc/sparc.md (UNSPEC_SHORT_LOAD): Delete.
3103 (zero_extend_v8qi_vis, zero_extend_v4hi_vis,
3104 *zero_extend_v8qi_<P:mode>_insn,
3105 *zero_extend_v4hi_<P:mode>_insn): Express using vec_merge
3106 and vec_duplicate instead of using an UNSPEC.
3108 2011-11-07 Alan Modra <amodra@gmail.com>
3111 * config/rs6000/rs6000.c (rs6000_emit_stack_reset): Always emit
3112 blockage for ABI_V4.
3114 2011-11-06 Dave Korn <dave.korn.cygwin@gmail.com>
3116 * config/i386/cygwin.h (LIBGCJ_SONAME): Updated to match recent
3117 libgcj version bump.
3118 * config/i386/mingw32.h (LIBGCJ_SONAME): Likewise.
3120 2011-11-06 Eric Botcazou <ebotcazou@adacore.com>
3122 * gcse.c: Adjust outdated comments throughout.
3123 (struct mem_conflict_info): New structure.
3124 (mems_conflict_for_gcse_p): Use it to communicate with caller.
3125 (load_killed_in_block_p): Pass it to note_stores.
3126 (hash_expr): Remove superfluous line break.
3127 (hash_scan_set): Rename PAT parameter into SET.
3128 (hash_scan_insn): Reorder cases.
3129 (canon_list_insert): Fix long line.
3130 (edge_list): Delete.
3131 (prune_expressions): Rename E local variable into EXPR.
3132 (compute_pre_data): Return struct edge_list * object.
3133 (pre_expr_reaches_here_p_work): Fix formatting.
3134 (process_insert_insn): Move around comment.
3135 (pre_edge_insert): Fix long line.
3136 (pre_insert_copies): Likewise.
3137 (gcse_emit_move_after): Swap SRC and DEST parameters.
3138 (pre_delete): Adjust call to gcse_emit_move_after.
3139 (pre_gcse): Take struct edge_list * parameter. Fix long line.
3140 (one_pre_gcse_pass): Use flag_gcse_lm condition for all routines.
3141 Use a local list of edges.
3142 (hoist_code): Fix long line. Adjust call to gcse_emit_move_after.
3143 (pre_ldst_expr_hash): Fix long line.
3144 (free_ldst_mems): Rename into...
3145 (free_ld_motion_mems): ...this.
3146 (first_ls_expr): Delete.
3147 (next_ls_expr): Likewise.
3148 (print_ldst_list): Do not use above two functions.
3149 (simple_mem): Adjust interface.
3150 (compute_ld_motion_mems): Fix formatting.
3151 (update_ld_motion_stores): Reuse local variable.
3153 2011-11-06 Joseph Myers <joseph@codesourcery.com>
3155 * c-decl.c (shadow_tag_warned, grokdeclarator): Handle _Alignas
3157 (build_null_declspecs): Initialize align_log and alignas_p fields.
3158 (declspecs_add_alignas): New.
3159 * c-parser.c (c_token_starts_declspecs): Handle RID_ALIGNAS.
3160 (c_parser_declspecs): Handle _Alignas specifiers.
3161 (c_parser_alignas_specifier): New.
3162 (c_parser_alignof_expression): Diagnose alignof use for non-C1X.
3163 Diagnose _Alignof (expression).
3164 * c-tree.h (struct c_declspecs): Add align_log and alignas_p fields.
3165 (declspecs_add_alignas): Declare.
3166 * ginclude/stddef.h (max_align_t): Define for C1X and C++11.
3167 * ginclude/stdalign.h: New.
3168 * Makefile.in (USER_H): Add stdalign.h.
3170 2011-11-06 Joern Rennecke <joern.rennecke@embecosm.com>
3172 * regset.h (fixed_reg_set_regset): Declare.
3173 * dse.c: Include regset.h .
3174 (struct insn_info): Add member fixed_regs_live.
3175 (note_add_store_info): New typedef.
3176 (note_add_store): New function.
3177 (emit_inc_dec_insn_before): Expect arg to be of type insn_info_t.
3178 Use gen_add3_insn / gen_move_insn.
3179 Check new insn for unwanted clobbers before emitting it.
3180 (check_for_inc_dec): Rename to...
3181 (check_for_inc_dec_1:) ... this. Return bool. Take insn_info
3182 parameter. Changed all callers in file.
3183 (check_for_inc_dec, copy_fixed_regs): New functions.
3184 (scan_insn): Set fixed_regs_live field of insn_info.
3185 * rtl.h (check_for_inc_dec): Update prototype.
3186 * postreload.c (reload_cse_simplify): Take new signature of
3187 check_ind_dec into account.
3188 * reginfo.c (fixed_reg_set_regset): New variable.
3189 (init_reg_sets_1): Initialize it.
3191 2011-11-06 Jakub Jelinek <jakub@redhat.com>
3193 * tree-cfg.c (gimple_can_merge_blocks_p): For -O0 don't remove
3196 2011-11-06 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
3198 * config/pa/pa.c (pa_hpux_init_libfuncs): Rename to pa_init_libfuncs.
3199 Remove dependence of declaration and target define on definition of
3200 HPUX_LONG_DOUBLE_LIBRARY. Update implementation.
3202 2011-11-06 Andrew Macleod <amacleod@redhat.com>
3203 Richard Henderson <rth@redhat.com>
3204 Aldy Hernandez <aldyh@redhat.com>
3206 Merged from cxx-mem-model.
3208 * cppbuiltin.c (define__GNUC__): Define __ATOMIC memory models
3209 * coretypes.h (enum memmodel): New. enumerated memory model type.
3210 * Makefile.in (cppbuiltin.o) Add missing dependency on $(TREE_H)
3211 * genopinit,c (optabs): Add atomic direct optab handlers.
3212 * sync-builtins.def (BUILT_IN_ATOMIC_*): New builtins.
3213 * builtin-types.def (BT_CONST_VOLATILE_PTR,
3214 BT_FN_I{1,2,4,8,16}_CONST_VPTR_INT, BT_FN_VOID_VPTR_INT,
3215 BT_FN_BOOL_VPTR_INT, BT_FN_BOOL_SIZE_CONST_VPTR,
3216 BT_FN_I{1,2,4,8,16}_VPTR_I{1,2,4,8,16}_INT,
3217 BT_FN_VOID_VPTR_I{1,2,4,8,16}_INT, BT_FN_VOID_SIZE_VPTR_PTR_INT,
3218 BT_FN_VOID_SIZE_CONST_VPTR_PTR_INT, BT_FN_VOID_SIZE_VPTR_PTR_PTR_INT,
3219 BT_FN_BOOL_VPTR_PTR_I{1,2,4,8,16}_BOOL_INT_INT): New builtin types.
3220 * expr.h (expand_atomic_*): Add prototypes.
3221 (expand_{bool,val}_compare_and_swap): Remove prototypes.
3222 * c-typeck.c (build_function_call_vec): Don't reprocess __atomic
3224 * common.opt (Winvalid-memory-model): New warning flag.
3225 (finline-atomics): New. Flag to disable atomic inlining.
3226 * params.h (ALLOW_LOAD_DATA_RACES): New.
3227 (ALLOW_PACKED_LOAD_DATA_RACES): New.
3228 (ALLOW_PACKED_STORE_DATA_RACES): New.
3229 * params.def (PARAM_ALLOW_LOAD_DATA_RACES): New.
3230 (PARAM_ALLOW_PACKED_LOAD_DATA_RACES): New.
3231 (PARAM_ALLOW_PACKED_STORE_DATA_RACES): New.
3232 * builtins.c (is_builtin_name): Handle __atomic.
3233 (get_memmodel): New. Extract memory model.
3234 (expand_expr_force_mode): New. Factor out common code for ensuring an
3235 integer argument is in the proper mode.
3236 (expand_builtin_sync_operation): Remove ignore param. Always call
3237 expand_atomic_fetch_op instead of the old expanders.
3238 (expand_builtin_compare_and_swap,
3239 expand_builtin_sync_lock_test_and_set): Use expand_expr_force_mode,
3240 call atomic expanders instead of sync expanders.
3241 (expand_builtin_sync_lock_release): Call atomic_store expander.
3242 (expand_builtin_atomic_compare_exchange, expand_builtin_atomic_load,
3243 expand_builtin_atomic_store, expand_builtin_atomic_fetch_op): New.
3244 (expand_builtin_atomic_exchange): New.
3245 (fold_builtin_atomic_always_lock_free,
3246 expand_builtin_atomic_always_lock_free,
3247 fold_builtin_atomic_is_lock_free, expand_builtin_atomic_is_lock_free):
3249 (expand_builtin_mem_thread_fence, expand_builtin_atomic_thread_fence,
3250 expand_builtin_atomic_signal_fence): New.
3251 (expand_builtin_mem_signal_fence): New.
3252 (expand_builtin): Add cases for BUILT_IN_ATOMIC_*.
3253 (fold_builtin_2): Add cases for BUILT_IN_ATOMIC_{IS,ALWAYS}_LOCK_FREE.
3254 * optabs.h (DOI_atomic_*): Define new atomics.
3255 (atomic_*_optab): Define.
3256 (can_compare_and_swap_p, expand_atomic_compare_and_swap): New
3258 * optabs.c (expand_sync_operation, expand_sync_fetch_operation): Remove.
3259 (expand_sync_lock_test_and_set): Remove.
3260 (expand_atomic_load, expand_atomic_store): New.
3261 (expand_atomic_exchange): New.
3262 (expand_atomic_compare_and_swap): New. Implements
3263 atomic_compare_exchange via compare and swap.
3264 (struct atomic_op_functions): Opcode table struct for fetch ops.
3265 (get_atomic_op_for_code): New. Return an opcode table entry.
3266 (maybe_emit_op): New. Try to emit a fetch op.
3267 (expand_atomic_fetch_op): New.
3268 (expand_val_compare_and_swap_1): Remove.
3269 (expand_val_compare_and_swap, expand_bool_compare_and_swap): Remove.
3270 (expand_atomic_compare_and_swap): Rename from
3271 expand_atomic_compare_exchange. Rewrite to return both success and
3272 oldval return values; expand via both atomic and sync optabs.
3273 (can_compare_and_swap_p): New.
3274 (expand_compare_and_swap_loop): Use expand_atomic_compare_and_swap.
3275 (maybe_gen_insn): Handle 7 and 8 operands.
3276 * omp-low.c (expand_omp_atomic_fetch_op): Don't test individual
3277 fetch_op optabs, only test can_compare_and_swap_p. Use __atomic
3278 builtins instead of __sync builtins.
3279 (expand_omp_atomic_pipeline): Use can_compare_and_swap_p.
3280 * doc/extend.texi: Document __atomic built-in functions.
3281 * doc/invoke.texi: Document data race parameters.
3282 * doc/md.texi: Document atomic patterns.
3283 * config/i386/i386.md (UNSPEC_MOVA): New.
3284 (UNSPECV_CMPXCHG): Split into ...
3285 (UNSPECV_CMPXCHG_1, UNSPECV_CMPXCHG_2,
3286 UNSPECV_CMPXCHG_3, UNSPECV_CMPXCHG_4): New.
3287 * config/i386/sync.md (ATOMIC): New mode iterator.
3288 (atomic_load<ATOMIC>, atomic_store<ATOMIC>): New.
3289 (atomic_loaddi_fpu, atomic_storedi_fpu, movdi_via_fpu): New.
3290 (mem_thread_fence): Rename from memory_barrier.
3291 Handle the added memory model parameter.
3292 (mfence_nosse): Rename from memory_barrier_nosse.
3293 (sync_compare_and_swap<CASMODE>): Split into ...
3294 (atomic_compare_and_swap<SWI124>): this and ...
3295 (atomic_compare_and_swap<CASMODE>): this. Handle the new parameters.
3296 (atomic_compare_and_swap_single<SWI>): Rename from
3297 sync_compare_and_swap<SWI>; rewrite to use split unspecs.
3298 (atomic_compare_and_swap_double<DCASMODE>): Rename from
3299 sync_double_compare_and_swap<DCASMODE>; rewrite to use split unspecs.
3300 (*atomic_compare_and_swap_doubledi_pic): Rename from
3301 sync_double_compare_and_swapdi_pic; rewrite to use split unspecs.
3302 (atomic_fetch_add<SWI>): Rename from sync_old_add<SWI>; add memory
3304 (*atomic_fetch_add_cmp<SWI>): Similarly.
3305 (atomic_add<SWI>, atomic<any_logic><SWI>): Similarly.
3306 (atomic_sub<SWI>): Similarly. Use x86_maybe_negate_const_int.
3307 (sync_lock_test_and_set<SWI>): Merge with ...
3308 (atomic_exchange<SWI>): ... this.
3310 2011-11-6 Richard Guenther <rguenther@suse.de>
3312 * ipa-prop.c (ipa_modify_call_arguments): Re-compute inlinable flag.
3314 2011-11-06 Ira Rosen <ira.rosen@linaro.org>
3316 * tree-vectorizer.h (vectorizable_condition): Add argument.
3317 * tree-vect-loop.c (vectorizable_reduction): Fail for condition
3318 in SLP. Update calls to vectorizable_condition.
3319 * tree-vect-stmts.c (vect_is_simple_cond): Add basic block info to
3320 the arguments. Pass it to vect_is_simple_use_1.
3321 (vectorizable_condition): Add slp_node to the arguments. Support
3322 vectorization of basic blocks. Fail for reduction in SLP. Update
3323 calls to vect_is_simple_cond and vect_is_simple_use. Support SLP:
3324 call vect_get_slp_defs to get vector operands.
3325 (vect_analyze_stmt): Update calls to vectorizable_condition.
3326 (vect_transform_stmt): Likewise.
3327 * tree-vect-slp.c (vect_create_new_slp_node): Handle COND_EXPR.
3328 (vect_get_and_check_slp_defs): Handle COND_EXPR. Allow pattern
3330 (vect_build_slp_tree): Handle COND_EXPR.
3331 (vect_analyze_slp_instance): Push pattern statements to root node.
3332 (vect_get_constant_vectors): Fix comments. Handle COND_EXPR.
3334 2011-11-05 David S. Miller <davem@davemloft.net>
3336 * config/sparc/sparc.md (UNSPEC_SHORT_LOAD): New unspec.
3337 (zero-extend_v8qi_vis, zero_extend_v4hi_vis): New expanders.
3338 (*zero_extend_v8qi_<P:mode>_insn,
3339 *zero_extend_v4hi_<P:mode>_insn): New insns.
3340 * config/sparc/sparc.c (vector_init_move_words,
3341 vector_init_prepare_elts, sparc_expand_vector_init_vis2,
3342 sparc_expand_vector_init_vis1): New functions.
3343 (vector_init_bshuffle): Rewrite to handle more cases and make use
3344 of locs[] array prepared by vector_init_prepare_elts.
3345 (vector_init_fpmerge, vector_init_faligndata): Delete.
3346 (sparc_expand_vector_init): Rewrite using new infrastructure.
3348 2011-11-05 Joern Rennecke <joern.rennecke@embecosm.com>
3350 * config.gcc (epiphany-*-*): New architecture.
3351 (epiphany-*-elf): New configuration.
3352 * config/epiphany, common/config/epiphany : New directories.
3353 * doc/extend.texi (disinterrupt attribute): Add Epiphany.
3354 (interrupt attribute): Add Epiphany.
3355 (long_call, short_call attribute): Add Epiphany.
3356 * doc/invoke.texi (Options): Add Epiphany options.
3357 * doc/md.texi (Machine Constraints): Add Epiphany constraints.
3358 * doc/install.texi (Options specification):
3359 Add --with-stack-offset=@var{num} description.
3360 (host/target specific issues): Add epiphany-*-elf.
3361 * doc/contrib.texi (Contributors): Mention Epiphany port.
3363 2011-11-05 Jakub Jelinek <jakub@redhat.com>
3365 PR tree-optimization/50693
3366 * tree-cfg.c (gimple_can_merge_blocks_p): Allow merging with
3367 non-forced user labels.
3368 (gimple_merge_blocks): Turn non-forced user labels into
3369 debug bind stmt with the label as first operand and reset value.
3370 (gimple_duplicate_bb): Don't duplicate label debug stmts.
3371 * dwarf2out.c (gen_label_die): Handle NOTE_INSN_DELETED_DEBUG_LABEL.
3372 * final.c (final_scan_insn): Likewise.
3373 (rest_of_clean_state): Don't dump NOTE_INSN_DELETED_DEBUG_LABEL.
3374 * var-tracking.c (debug_label_num): New variable.
3375 (delete_debug_insns): Don't delete DEBUG_INSNs for LABEL_DECLs,
3376 instead turn them into NOTE_INSN_DELETED_DEBUG_LABEL notes.
3377 * cfglayout.c (skip_insns_after_block, duplicate_insn_chain): Handle
3378 NOTE_INSN_DELETED_DEBUG_LABEL.
3379 (duplicate_insn_chain): Don't duplicate LABEL_DECL DEBUG_INSNs.
3380 * insn-notes.def (DELETED_DEBUG_LABEL): New note kind.
3381 * print-rtl.c (print_rtx): Handle NOTE_INSN_DELETED_DEBUG_LABEL.
3382 * gengtype.c (adjust_field_rtx_def): Likewise.
3383 * config/i386/i386.c (ix86_output_function_epilogue): For MachO
3384 clear CODE_LABEL_NUMBER of NOTE_INSN_DELETED_DEBUG_LABEL
3385 if their are at the end of function and nop hasn't been emitted.
3386 * config/rs6000/rs6000.c (rs6000_output_function_epilogue): Likewise.
3388 2011-11-05 Georg-Johann Lay <avr@gjlay.de>
3390 PR rtl-optimization/50448
3391 * cprop.c (try_replace_reg): Also try to replace uses of FROM that
3394 2011-11-05 Peter Dufault <dufault@hda.com>,
3395 Sebastian Huber <sebastian.huber@embedded-brains.de>
3397 * config/rs6000/rtems.h (SUBSUBTARGET_OVERRIDE_OPTIONS):
3398 Extend rs6000_spe handling.
3400 2011-11-05 Ralf Corsépius <ralf.corsepius@rtems.org>
3402 * config/rs6000/t-rtems: Add -mcpu=8540/-mfloat-gprs=double multilib.
3403 Remove -mcpu=601 multilib.
3404 Remove -Dmpc8260 multilib.
3405 * config/rs6000/rtems.h: Allow --float-gprs=... to override grps
3408 2011-11-05 Quentin Neill <quentin.neill@amd.com>
3410 Piledriver f16cintrin.h fix.
3411 * config/i386/f16cintrin.h: Contents moved from immintrin.h.
3412 * config/i386/immintrin.h: Include f16cintrin.h.
3413 * config.gcc (i[34567]86-*-*, x86_64-*-*): Add f16cintrin.h.
3416 2011-11-04 Eric Botcazou <ebotcazou@adacore.com>
3419 * c-parser.c (c_parser_postfix_expression) <RID_OFFSETOF>: Adjust call
3421 * c-typeck.c (build_unary_op) <ADDR_EXPR>: Call fold_offsetof_1.
3423 2011-11-04 Alan Modra <amodra@gmail.com>
3425 * reload1.c (gen_reload): Don't use REGNO on SUBREGs.
3426 * print-rtl.c (print_rtx): Don't segfault on negative regno.
3428 2011-11-04 David S. Miller <davem@davemloft.net>
3431 * config/sparc/sparc.c (sparc_expand_conditional_move): Handle the
3432 fact that sparc_emit_float_lib_cmp modifies the comparison in
3435 2011-11-04 Ralf Corsépius <ralf.corsepius@rtems.org>
3437 * config/lm32/t-rtems: New.
3438 * config.gcc (lm32-*-rtems*): Add t-rtems.
3440 2011-11-04 Eric Botcazou <ebotcazou@adacore.com>
3443 * config/sparc/sparc.h (ASM_CPU_SPEC): Pass -Av8 if -mcpu=v8.
3445 2011-11-04 Jiangning Liu <jiangning.liu@arm.com>
3447 PR rtl-optimization/38644
3448 * config/arm/arm.c (thumb1_expand_epilogue): Add memory barrier
3449 for epilogue having stack adjustment.
3451 2011-11-04 Georg-Johann Lay <avr@gjlay.de>
3454 * config/avr/avr-modes.def: New file defining PSImode.
3455 * config/avr/avr-c.c (__INT24_MAX__, __INT24_MIN__,
3456 __UINT24_MAX__): New built-in defines.
3457 * config/avr/avr.md (adjust_len): Add tstpsi, mov24, reload_in24,
3458 ashlpsi, ashrpsi, lshrpsi.
3459 (QISI, QIDI, HISI, HIDI, MPUSH, rotx, rotsmode): Add PSI.
3460 (MOVMODE): New mode iterator.
3461 (movpsi): New expander.
3462 (movqi, movhi, movsi, movsf, movpsi): Write as one using MOVMODE.
3463 (*reload_inpsi, *movpsi): New insns.
3464 (*reload_inpsi): New RTL peephole.
3465 (addpsi3, *addpsi3_zero_extend.qi, *addpsi3_zero_extend.hi,
3466 *addpsi3_sign_extend.hi): New insns.
3467 (subpsi3, *subpsi3_zero_extend.qi, *subpsi3_zero_extend.hi,
3468 *subpsi3_sign_extend.hi): New insns.
3469 (divmodpsi4, udivmodpsi4): New define insn-and-split.
3470 (*divmodpsi4_call, *udivmodpsi4_call): New insns.
3471 (andpsi3, iorpsi3, xorpsi3): New insns.
3472 (*rotlpsi2.1, *rotlpsi2.23): New insns.
3473 (*rotw<mode>): Insn condition only allow even-sized modes.
3474 (*rotb<mode>): Insn condition allows odd-sized modes.
3475 (ashlpsi3, ashrpsi3, lshrpsi3, *addpsi3.lt0): New insns.
3476 (negpsi2, one_cmplpsi2): New insns.
3477 (extendqipsi2, extendhipsi2, extendpsisi2): New insns.
3478 (zero_extendqipsi2, zero_extendhipsi2, zero_extendpsisi2): New
3480 (*cmppsi, *negated_tstpsi, *reversed_tstpsi): New insns.
3481 (cbranchpsi4): New expander.
3482 * config/avr/constraints.md (Ca3, Co3, Cx3): New constraints.
3483 * config/avr/avr-protos.h (avr_out_tstpsi, avr_out_movpsi,
3484 avr_out_ashlpsi3, avr_out_ashrpsi3, avr_out_lshrpsi3,
3485 avr_out_reload_inpsi): New prototypes.
3487 * config/avr/avr.c (TARGET_SCALAR_MODE_SUPPORTED_P): Define to...
3488 (avr_scalar_mode_supported_p): ...this new static function.
3489 (avr_asm_len): Always return "".
3490 (avr_out_load_psi, avr_out_store_psi): New static functions.
3491 (avr_out_movpsi, avr_out_reload_inpsi): New functions.
3492 (avr_out_tstpsi): New function.
3493 (avr_out_ashlpsi3, avr_out_ashrpsi3, avr_out_lshrpsi3): New functions.
3494 (avr_out_plus_1, output_reload_in_const): Handle 3-byte types.
3495 (avr_simplify_comparison_p): Ditto.
3496 (adjust_insn_length): Handle ADJUST_LEN_RELOAD_IN24,
3497 ADJUST_LEN_MOV24, ADJUST_LEN_TSTPSI, ADJUST_LEN_ASHLPSI,
3498 ADJUST_LEN_ASHRPSI, ADJUST_LEN_LSHRPSI.
3499 (avr_rtx_costs_1): Report PSI costs.
3500 (avr_libcall_value): Handle odd-sized parameters.
3501 (avr_init_builtin_int24): New static function to define built-in
3502 24-bit types __int24 and __uint24.
3503 (avr_init_builtins): Use it.
3505 2011-11-04 Thomas Doerfler <thomas.doerfler@embedded-brains.de>
3508 * config/arm/rtems-elf.h, config/arm/t-rtems: Add optional
3509 support for VFP floating point model.
3511 2011-11-04 Tristan Gingold <gingold@adacore.com>
3513 * config/alpha/vms.h (ASM_OUTPUT_DEF): Do not switch section.
3515 2011-11-04 Ira Rosen <ira.rosen@linaro.org>
3518 2011-10-24 Ira Rosen <ira.rosen@linaro.org>
3520 PR tree-optimization/50730
3521 * tree-vect-data-refs.c (vect_analyze_data_refs): Stop basic block
3522 analysis if encountered unsupported data-ref.
3524 2011-11-04 Jakub Jelinek <jakub@redhat.com>
3526 * config/i386/i386.c (ix86_expand_vector_convert_uns_vsivsf): New
3528 * config/i386/i386-protos.h (ix86_expand_vector_convert_uns_vsivsf):
3530 * config/i386/sse.md (floatuns<sseintvecmodelower><mode>2): Use it.
3531 For floatunsv8siv8sf2 require TARGET_AVX2.
3533 * config/i386/i386.c (ix86_expand_adjust_ufix_to_sfix_si): Add
3534 XORP argument. Subtract 0x1p31 instead of 0x1p32. Use normal
3535 signalling comparison instead of non-signalling. Store into
3536 *XORP pseudo holding 0x80000000 integers if 0x1p31 has been
3537 subtracted and 0 otherwise.
3538 * config/i386/i386-protos.h (ix86_expand_adjust_ufix_to_sfix_si):
3540 * config/i386/sse.md (fixuns_trunc<mode><sseintvecmodelower>2): Enable
3541 already for TARGET_SSE2. Xor in vector initialized by
3542 ix86_expand_adjust_ufix_to_sfix_si at the end.
3543 (vec_pack_ufix_trunc_<mode>): Likewise.
3545 * tree-vect-stmts.c (vectorizable_conversion): Rewritten to handle
3546 not just FLOAT_EXPR and FIX_TRUNC_EXPR, but also CONVERT_EXPR_CODE_P,
3547 WIDEN_MULT_EXPR and WIDEN_LSHIFT_EXPR to handle what
3548 vectorizable_type_demotion and vectorizable_type_promotion did.
3549 Additionally handle FLOAT_EXPR and FIX_TRUNC_EXPR where the integer
3550 is {,un}signed {char,short}.
3551 (vect_create_vectorized_demotion_stmts): Fix comment typo. For
3552 recursive calls unconditionally use VEC_PACK_TRUNC_EXPR.
3553 Push vec_dest back to the vec_dsts vector at the end.
3554 (vect_create_vectorized_promotion_stmts): Don't recurse, do just
3555 one step. Removed multi_step_cvt, vec_dsts, slp_node and
3556 prev_stmt_info arguments, add vec_dest argument. Push always
3557 into vec_tmp, not just when multi_step_cvt != 0, replace *vec_oprdn0
3558 with vec_tmp at the end after freeing old *vec_oprnd0 vector.
3559 (vectorizable_type_demotion, vectorizable_type_promotion): Removed.
3560 (vect_analyze_stmt): Don't call vectorizable_type_demotion and
3561 vectorizable_type_promotion. Call vectorizable_conversion even
3562 for SLP bb vectorization.
3563 (vect_transform_stmt): Call vectorizable_conversion instead of
3564 vectorizable_type_demotion and vectorizable_type_promotion.
3565 (supportable_widening_operation): Clear *multi_step_cvt first,
3566 simplify c1/c2 computation, free *interm_types vector on failure.
3567 (supportable_narrowing_operation): Clear *multi_step_cvt first,
3568 free *interm_types vector on failure, handle multi-step
3571 2011-11-04 Tristan Gingold <gingold@adacore.com>
3573 * config/alpha/alpha.c (alpha_write_linkage): Remove fundecl
3574 argument. Conditionally generate crash debug info. Adjust
3575 for alpha_funcs_tree removal.
3576 (machine_function): Add links field.
3577 (alpha_start_function): Conditionally generate crash debug info.
3578 (alpha_end_function): Adjust call to alpha_write_linkage.
3579 (alpha_funcs): Remove.
3580 (links_kind): Remove.
3581 (alpha_links): Remove num, target and lkind field. Add func field.
3582 (alpha_links_tree): Remove.
3583 (alpha_funcs_tree): Remove.
3584 (alpha_need_linkage): Remove.
3585 (alpha_use_linkage): Change prototype. Adjust.
3586 (alpha_write_one_linkage): Use ASM_OUTPUT_INTERNAL_LABEL.
3587 Use SYMBOL_REF_EXTERNAL_P and SYMBOL_REF_LOCAL_P macro.
3588 * config/alpha/alpha-protos.h (alpha_use_linkage): Update.
3589 (alpha_need_linkage): Remove.
3590 * config/alpha/alpha.md: Update calls to alpha_use_linkage.
3591 Adjust calls to alpha_need_linkage.
3593 2011-11-03 Uros Bizjak <ubizjak@gmail.com>
3595 * sched-vis.c (print_value): Handle STRICT_LOW_PART.
3597 2011-11-03 Uros Bizjak <ubizjak@gmail.com>
3599 * config/i386/i386.md (lround<X87MODEF:mode><SWI248x:mode>2,
3600 rint<mode>2, floor<mode>2, lfloor<MODEF:mode><SWI48:mode>2,
3601 btrunc<mode>2, lwp_lwpval<mode>3): Use operands[N] instead of operandN.
3603 2011-11-03 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3606 * config/arm/t-bpabi: New file.
3607 * config.gcc (arm*-*-linux*): Add arm/t-bpabi to tmake_file for
3609 (arm*-*-uclinux*): Add arm/t-bpabi to tmake_file for
3610 arm*-*-uclinux*eabi.
3611 (arm*-*-eabi*, arm*-*-symbianelf*): Add arm/t-bpabi to tmake_file
3614 2011-11-03 Michael Matz <matz@suse.de>
3617 * configure.ac: Check for -fno-exceptions -fno-rtti.
3618 * configure: Regenerate.
3619 * Makefile.in (NOEXCEPTION_FLAGS): New flag.
3620 (ALL_CXXFLAGS): Use it.
3622 2011-11-03 Uros Bizjak <ubizjak@gmail.com>
3624 * config/i386/i386.md: Use {} for multi-line preparation statements.
3626 2011-11-03 Eric Botcazou <ebotcazou@adacore.com>
3628 * config/sparc/sparc.md (movtf_insn_sp32_no_fpu): Consolidate into...
3629 (movtf_insn_sp32): ...this.
3630 (movtf_insn_sp64_no_fpu): Consolidate into...
3631 (movtf_insn_sp64): ...this.
3632 (movtf_insn_sp64_hq): Do not test TARGET_FPU.
3633 * config/sparc/sparc.c (sparc_legitimate_address_p): Likewise.
3635 2011-11-03 Tristan Gingold <gingold@adacore.com>
3637 * config/vms/vms.c (vms_patch_builtins): Fix typo.
3639 2011-11-03 Richard Guenther <rguenther@suse.de>
3642 * lto-opts.c: Re-implement.
3643 * lto-streamer.h (lto_register_user_option): Remove.
3644 (lto_read_file_options): Likewise.
3645 (lto_reissue_options): Likewise.
3646 (lto_clear_user_options): Likewise.
3647 (lto_clear_file_options): Likewise.
3648 * opts-global.c (post_handling_callback): Remove.
3649 (set_default_handlers): Do not set post_handling_callback.
3650 (decode_options): Remove LTO specific code.
3651 * lto-wrapper.c (merge_and_complain): New function.
3652 (run_gcc): Read all input file options and
3653 prepend a merged set before the linker driver options.
3654 * gcc.c (driver_post_handling_callback): Remove.
3655 (set_option_handlers): Do not set post_handling_callback.
3656 * opts-common.c (handle_option): Do not call post_handling_callback.
3657 * opts.h (struct cl_option_handlers): Remove post_handling_callback.
3659 2011-11-03 Richard Guenther <rguenther@suse.de>
3661 * collect2.c (main): Guard object_nbr variable with TARGET_AIX_VERSION.
3663 2011-11-03 Martin Jambor <mjambor@suse.cz>
3665 * ipa-prop.c (type_change_info): New fields offset, object,
3666 known_current_type and multiple_types_encountered.
3667 (extr_type_from_vtbl_ptr_store): New function.
3668 (check_stmt_for_type_change): Use it, set multiple_types_encountered if
3669 the result is different from the previous one.
3670 (detect_type_change): Renamed to detect_type_change_1. New parameter
3671 comp_type. Set up new fields in tci, build known type jump
3672 functions if the new type can be identified.
3673 (detect_type_change): New function.
3674 * tree.h (DECL_CONTEXT): Comment new use.
3676 2011-11-03 Richard Guenther <rguenther@suse.de>
3679 * lto-wrapper.c (get_options_from_collect_gcc_options): Properly
3680 decode an encoded literal '.
3682 2011-11-03 Tristan Gingold <gingold@adacore.com>
3684 * collect2.c (main): Add support of -f (response file) on AIX.
3686 2011-11-03 Ira Rosen <ira.rosen@linaro.org>
3688 PR tree-optimization/50912
3689 * tree-vectorizer.h (slp_void_p): New.
3690 (struct _slp_tree): Replace left and right with children. Update
3692 (struct _slp_oprnd_info): New.
3693 (vect_get_vec_defs): Declare.
3694 (vect_get_slp_defs): Update arguments.
3695 * tree-vect-loop.c (vect_create_epilog_for_reduction): Call
3696 vect_get_vec_defs instead of vect_get_slp_defs.
3697 (vectorizable_reduction): Likewise.
3698 * tree-vect-stmts.c (vect_get_vec_defs): Remove static, add argument.
3699 Update call to vect_get_slp_defs.
3700 (vectorizable_conversion): Update call to vect_get_vec_defs.
3701 (vectorizable_assignment, vectorizable_shift,
3702 vectorizable_operation): Likewise.
3703 (vectorizable_type_demotion): Call vect_get_vec_defs instead of
3705 (vectorizable_type_promotion, vectorizable_store): Likewise.
3706 (vect_analyze_stmt): Fix typo.
3707 * tree-vect-slp.c (vect_free_slp_tree): Update SLP tree traversal.
3708 (vect_print_slp_tree, vect_mark_slp_stmts,
3709 vect_mark_slp_stmts_relevant, vect_slp_rearrange_stmts,
3710 vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations,
3711 vect_schedule_slp_instance): Likewise.
3712 (vect_create_new_slp_node): New.
3713 (vect_create_oprnd_info, vect_free_oprnd_info): Likewise.
3714 (vect_get_and_check_slp_defs): Pass information about defs using
3715 oprnds_info, allow any number of operands.
3716 (vect_build_slp_tree): Likewise. Update calls to
3717 vect_get_and_check_slp_defs. Fix comments.
3718 (vect_analyze_slp_instance): Move node creation to
3719 vect_create_new_slp_node.
3720 (vect_get_slp_defs): Allow any number of operands.
3722 2011-11-02 Peter Bergner <bergner@vnet.ibm.com>
3723 Iain Sandoe <iains@gcc.gnu.org>
3725 * config/rs6000/rs6000.c (USE_HIDDEN_LINKONCE): New define.
3726 (get_ppc476_thunk_name): Use it.
3727 (rs6000_code_end): Likewise.
3728 (macho_branch_islands): Fix typo.
3730 2011-11-02 Paolo Carlini <paolo.carlini@oracle.com>
3731 Jason Merrill <jason@redhat.com>
3734 * configure.ac: Add -Wno-narrowing to warning options.
3735 * doc/invoke.texi ([-Wnarrowing], [-Wc++0x-compat]): Update.
3737 2011-11-02 Eric Botcazou <ebotcazou@adacore.com>
3740 * config/sparc/sparc.md (movsf_insn): Reindent constraints.
3741 (movdf_insn_sp32): Likewise. Remove redundant G constraint.
3742 (movdf_insn_sp64): Likewise.
3743 (DFmode splitter): Do not test TARGET_FPU.
3744 (movtf_insn_sp32): Reindent constraints.
3745 (movtf_insn_sp32_no_fpu): Likewise.
3746 (movtf_insn_sp64): Likewise.
3747 (movtf_insn_sp64_hq): Likewise.
3748 (movtf_insn_sp64_no_fpu): Likewise.
3750 2011-11-02 Paolo Carlini <paolo.carlini@oracle.com>
3753 * builtins.c (fold_builtin_memchr): Fix cast.
3755 2011-11-02 Teresa Johnson <tejohnson@google.com>
3757 * config/i386/predicates.md (promotable_binary_operator): Add minus
3758 to the list of promotable operators.
3760 2011-11-02 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3762 * gthr-single.h, gthr.h: Move to ../libgcc.
3763 * gthr-aix.h: Move to ../libgcc/config/rs6000.
3764 * gthr-dce.h: Move to ../libgcc/config/pa.
3765 * gthr-lynx.h: Move to ../libgcc/config.
3766 * gthr-mipssde.h: Move to ../libgcc/config/mips.
3767 * gthr-posix.h: Move to ../libgcc/config.
3768 * gthr-rtems.h: Likewise.
3769 * gthr-tpf.h: Move to ../libgcc/config/s390.
3770 * gthr-vxworks.h: Move to ../libgcc/config.
3771 * gthr-win32.h: Move to ../libgcc/config/i386.
3772 * configure.ac (gthread_flags): Remove
3773 (gthr-default.h): Don't create.
3774 (thread_file): Don't substitute.
3775 * configure: Regenerate.
3776 * Makefile.in (GCC_THREAD_FILE): Remove.
3777 (GTHREAD_FLAGS): Remove.
3778 (libgcc.mvars): Remove GTHREAD_FLAGS.
3779 * config/t-vxworks (EXTRA_HEADERS): Remove.
3781 2011-11-02 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3782 Paolo Bonzini <bonzini@gnu.org>
3784 * configure.ac (libgcc_tm_file_list, libgcc_tm_include_list): Remove.
3785 * configure: Regenerate.
3786 * Makefile.in (libgcc_tm_file_list, libgcc_tm_include_list): Remove.
3787 (TM_H): Remove libgcc_tm.h, $(libgcc_tm_file_list).
3788 (libgcc_tm.h, cs-libgcc_tm.h): Remove.
3789 (clean): Remove libgcc_tm.h
3790 * mkconfig.sh: Don't include libgcc_tm.h in tm.h.
3791 * config.gcc (libgcc_tm_file): Remove.
3792 (arm*-*-linux*): Remove libgcc_tm_file for arm*-*-linux-*eabi.
3793 (arm*-*-uclinux*): Remove libgcc_tm_file for arm*-*-uclinux*eabi.
3794 (arm*-*-eabi*, arm*-*-symbianelf*): Remove libgcc_tm_file.
3795 (avr-*-rtems*): Likewise.
3796 (avr-*-*): Likewise.
3797 (frv-*-elf): Likewise.
3798 (frv-*-*linux*): Likewise.
3799 (h8300-*-rtems*): Likewise.
3800 (h8300-*-elf*): Likewise.
3801 (i[34567]86-*-darwin*): Likewise.
3802 (x86_64-*-darwin*): Likewise.
3803 (rx-*-elf*): Likewise.
3804 (tic6x-*-elf): Likewise.
3805 (tic6x-*-uclinux): Likewise.
3806 (i[34567]86-*-linux*, x86_64-*-linux*): Likewise.
3808 2011-11-02 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3810 * Makefile.in (LIBGCC2_DEBUG_CFLAGS LIBGCC2_CFLAGS)
3811 (LIBGCC2_INCLUDES, TARGET_LIBGCC2_CFLAGS, LIB2FUNCS_EXTRA)
3812 (LIB2FUNCS_STATIC_EXTRA, LIB2FUNCS_EXCLUDE, T, T_TARGET)
3813 (INCLUDES_FOR_TARGET): Remove.
3814 (LIBGCC2_CFLAGS): Don't export.
3815 (LIB2FUNCS_ST, LIB2_DIVMOD_FUNCS, LIB2ADD, LIB2ADD_ST, srcdirify):
3817 (libgcc-support): Remove $(LIB2ADD), $(LIB2ADD_ST) dependencies.
3818 (libgcc.mvars): Likewise.
3819 Don't emit LIB2FUNCS_ST, LIB2FUNCS_EXCLUDE, LIB2ADD, LIB2ADD_ST,
3820 LIB2_SIDITI_CONV_FUNCS, LIB2_DIVMOD_FUNCS, LIBGCC2_CFLAGS,
3821 TARGET_LIBGCC2_CFLAGS.
3823 * libgcc2.c, libgcc2.h, gbl-ctors.h, longlong.h: Move to ../libgcc.
3824 * config/darwin-64.c: Move to ../libgcc/config.
3825 * config/divmod.c, config/floatunsidf.c, config/floatunsisf.c,
3826 config/floatunsitf.c, config/floatunsixf.c, config/udivmod.c,
3827 config/udivmodsi4.c: Move to ../libgcc/config.
3828 * config/gthr-posix.c: Move to ../libgcc/config/alpha.
3829 * config/memcmp.c, config/memcpy.c, config/memmove.c,
3830 config/memset.c: Move to ../libgcc/config.
3831 * config/t-darwin (TARGET_LIBGCC2_CFLAGS): Remove.
3832 * config/t-freebsd: Remove.
3833 * config/t-freebsd-thread: Move to ../libgcc/config.
3834 * config/t-libgcc-pic: Move to ../libgcc/config.
3835 * config/t-libunwind (TARGET_LIBGCC2_CFLAGS): Remove.
3836 * config/t-linux: Remove.
3837 * config/t-lynx (TARGET_LIBGCC2_CFLAGS, LIBGCC, INSTALL_LIBGCC):
3839 * config/t-openbsd-thread: Move to ../libgcc/config.
3840 * config/t-rtems (LIBGCC2_INCLUDES): Remove.
3841 * config/t-sol2 (TARGET_LIBGCC2_CFLAGS): Remove.
3842 * config/t-svr4: Remove.
3843 * config/t-vxworks (LIBGCC, INSTALL_LIBGCC, TARGET_LIBGCC2_CFLAGS)
3844 (LIBGCC2_DEBUG_CFLAGS, LIB2FUNCS_EXTRA, LIBGCC2_INCLUDES): Remove.
3845 * config/vxlib.c, config/vxlib-tls.c: Move to ../libgcc/config.
3846 * config/alpha/qrnnd.asm: Move to ../libgcc/config/alpha/qrnnd.S.
3847 * config/alpha/t-alpha, config/alpha/t-ieee: Remove.
3848 * config/alpha/t-vms (LIB2FUNCS_EXTRA, LIBGCC, INSTALL_LIBGCC): Remove.
3849 * config/alpha/vms-gcc_shell_handler.c: Move to ../libgcc/config/alpha.
3850 * config/arm/bpabi.c, config/arm/unaligned-funcs.c,
3851 config/arm/fp16.c, config/arm/linux-atomic.c,
3852 config/arm/linux-atomic-64bit.c: Move to ../libgcc/config/arm.
3853 * config/arm/t-arm-elf (LIBGCC, INSTALL_LIBGCC)
3854 (TARGET_LIBGCC2_CFLAGS): Remove.
3855 * config/arm/t-bpabi, config/arm/t-linux: Remove.
3856 * config/arm/t-linux-eabi (TARGET_LIBGCC2_CFLAGS)
3857 (LIB2FUNCS_STATIC_EXTRA): Remove.
3858 * config/arm/t-netbsd: Remove.
3859 * config/arm/t-strongarm-elf (LIBGCC, INSTALL_LIBGCC)
3860 (TARGET_LIBGCC2_CFLAGS): Remove.
3861 * config/arm/t-symbian (LIB2FUNCS_STATIC_EXTRA): Remove.
3862 * config/arm/t-wince-pe (LIBGCC, INSTALL_LIBGCC)
3863 (TARGET_LIBGCC2_CFLAGS): Remove.
3864 * config/avr/t-avr (LIB2FUNCS_EXCLUDE, TARGET_LIBGCC2_CFLAGS)
3865 (LIBGCC, INSTALL_LIBGCC): Remove.
3866 * config/bfin/t-bfin-elf (TARGET_LIBGCC2_CFLAGS): Remove.
3867 * config/bfin/t-bfin-linux: Likewise.
3868 * config/bfin/t-bfin-uclinux: Likewise.
3869 * config/c6x/eqd.c, config/c6x/eqf.c, config/c6x/ged.c,
3870 config/c6x/gef.c, config/c6x/gtd.c, config/c6x/gtf.c,
3871 config/c6x/led.c, config/c6x/lef.c, config/c6x/ltd.c,
3872 config/c6x/ltf.c: Move to ../libgcc/config/c6x.
3873 * config/c6x/t-c6x-elf (LIB2FUNCS_EXCLUDE, LIB2FUNCS_EXTRA): Remove.
3874 * config/c6x/t-c6x-uclinux (TARGET_LIBGCC2_CFLAGS): Remove.
3875 * config/cris/arit.c: Move to ../libgcc/config/cris.
3876 * config/cris/cris_abi_symbol.c: Remove.
3877 * config/cris/cris.h: Remove obsolete comment.
3878 * config/cris/mulsi3.asm: Move to ../libgcc/config/cris/mulsi3.S.
3879 * config/cris/t-cris (LIB2FUNCS_EXTRA, CRIS_LIB1CSRC)
3880 ($(LIB2FUNCS_EXTRA)): Remove.
3881 * config/cris/t-elfmulti (LIB2FUNCS_STATIC_EXTRA, INSTALL_LIBGCC)
3883 * config/cris/t-linux (TARGET_LIBGCC2_CFLAGS): Remove.
3884 * config/fr30/t-fr30: Remove.
3885 * config/frv/cmovd.c, config/frv/cmovh.c, config/frv/cmovw.c,
3886 config/frv/modi.c, config/frv/uitod.c, config/frv/uitof.c,
3887 config/frv/ulltod.c, config/frv/ulltof.c, config/frv/umodi.c: Move
3888 to ../libgcc/config/frv.
3889 * config/frv/t-frv (LIB2FUNCS_EXTRA, TARGET_LIBGCC2_CFLAGS)
3890 (cmovh.c, cmovw.c, cmovd.c, modi.c, umodi.c, uitof.c, uitod.c)
3891 (ulltof.c, LIBGCC, INSTALL_LIBGCC): Remove.
3892 * config/frv/t-linux (TARGET_LIBGCC2_CFLAGS): Remove.
3893 * config/h8300/clzhi2.c, config/h8300/ctzhi2.c,
3894 config/h8300/fixunssfsi.c, config/h8300/parityhi2.c,
3895 config/h8300/popcounthi2.c: Move to ../libgcc/config/h8300.
3896 * config/h8300/t-h8300 (LIB2FUNCS_EXTRA, TARGET_LIBGCC2_CFLAGS)
3897 (LIBGCC, INSTALL_LIBGCC): Remove.
3898 * config/i386/gthr-win32.c: Move to ../libgcc/config/i386.
3899 * config/i386/t-cygming (LIBGCC2_INCLUDES): Remove.
3900 * config/i386/t-cygwin: Remove.
3901 * config/i386/t-darwin (LIB2_SIDITI_CONV_FUNCS, LIB2FUNCS_EXTRA)
3902 (LIB2FUNCS_EXCLUDE): Remove.
3903 * config/i386/t-darwin64 (LIB2_SIDITI_CONV_FUNCS, LIB2FUNCS_EXTRA)
3904 (LIBGCC, INSTALL_LIBGCC): Remove.
3905 * config/i386/t-gthr-win32: Move to ../libgcc/config/i386.
3906 * config/i386/t-linux64 (LIBGCC, INSTALL_LIBGCC): Remove.
3907 * config/i386/t-mingw-w32: Likewise.
3908 * config/i386/t-mingw-w64: Likewise.
3909 * config/i386/t-openbsd: Likewise.
3910 * config/i386/t-nto: Remove.
3911 * config/ia64/quadlib.c: Move to ../libgcc/config/ia64.
3912 * config/ia64/t-hpux (LIBGCC, INSTALL_LIBGCC, LIB2FUNCS_EXTRA)
3913 (quadlib.c): Remove.
3914 * config/ia64/t-ia64: Remove comment.
3915 * config/iq2000/lib2extra-funcs.c: Move to
3916 ../libgcc/config/iq2000/lib2funcs.c.
3917 * config/iq2000/t-iq2000: Remove.
3918 * config/m32c/m32c-lib2.c: Move to ../libgcc/config/m32c/lib2funcs.c.
3919 * config/m32c/m32c-lib2-trapv.c: Move to ../libgcc/config/m32c/trapv.c.
3920 * config/m32r/t-linux (TARGET_LIBGCC2_CFLAGS): Remove.
3921 * config/m32c/t-m32c (LIB2FUNCS_EXTRA): Remove.
3922 * config/m32r/t-m32r (TARGET_LIBGCC2_CFLAGS, LIBGCC)
3923 (INSTALL_LIBGCC): Remove.
3924 * config/m68k/fpgnulib.c: Move to ../libgcc/config/m68k.
3925 * config/m68k/t-floatlib: Remove.
3926 * config/m68k/t-mlibs (LIBGCC, INSTALL_LIBGCC): Remove.
3927 * config/mcore/t-mcore (TARGET_LIBGCC2_CFLAGS): Remove.
3929 (LIBGCC, INSTALL_LIBGCC): Remove.
3930 * config/mep/mep-lib2.c: Move to ../libgcc/config/mep/lib2funcs.c.
3931 * config/mep/mep-tramp.c: Move to ../libgcc/config/mep/tramp.c.
3932 * config/mep/t-mep (LIB2FUNCS_EXTRA): Remove.
3933 * config/mips/t-elf (TARGET_LIBGCC2_CFLAGS, LIBGCC)
3934 (INSTALL_LIBGCC): Remove.
3935 * config/mips/t-isa3264: Likewise.
3936 * config/mips/t-mips (LIB2_SIDITI_CONV_FUNCS): Remove.
3937 * config/mips/t-r3900 (TARGET_LIBGCC2_CFLAGS, LIBGCC)
3938 (INSTALL_LIBGCC): Remove.
3939 * config/mips/t-sde (LIBGCC, INSTALL_LIBGCC): Remove.
3940 * config/mips/t-sr71k (TARGET_LIBGCC2_CFLAGS, LIBGCC)
3941 (INSTALL_LIBGCC): Remove.
3942 * config/mips/t-vr (TARGET_LIBGCC2_CFLAGS)
3943 (LIB2FUNCS_STATIC_EXTRA): Remove.
3944 * config/mips/vr4120-div.S: Move to ../libgcc/config/mips.
3945 * config/mmix/t-mmix (TARGET_LIBGCC2_CFLAGS): Remove.
3946 * config/mn10300/t-mn10300 (LIBGCC, INSTALL_LIBGCC): Remove.
3947 * config/pa/fptr.c, config/pa/linux-atomic.c: Move to
3948 ../libgcc/config/pa.
3949 * config/pa/lib2funcs.asm: Move to ../libgcc/config/pa/lib2funcs.S.
3950 * config/pa/quadlib.c: Move to ../libgcc/config/pa.
3951 * config/pa/t-dce-thr (LIBGCC, INSTALL_LIBGCC): Remove.
OSDN Git Service
