1 2007-02-10 Richard Henderson <rth@redhat.com>, Jakub Jelinek <jakub@redhat.com>, Alexandre Oliva <aoliva@redhat.com>
3 * Makefile.in (libgcc-support, libgcc.mvars): Add emutls.c.
4 * builtin-types.def (BT_WORD): Make unsigned.
5 (BT_FN_VOID_PTR_WORD_WORD_PTR): New.
6 * builtins.def (BUILT_IN_EMUTLS_GET_ADDRESS): New.
7 (BUILT_IN_EMUTLS_REGISTER_COMMON): New.
8 * c-decl.c (grokdeclarator): Don't error if !have_tls.
9 * c-parser.c (c_parser_omp_threadprivate): Likewise.
10 * dwarf2out.c (loc_descriptor_from_tree_1): Don't do anything for
12 * expr.c (emutls_var_address): New.
13 (expand_expr_real_1): Expand emulated tls.
14 (expand_expr_addr_expr_1): Likewise.
15 * libgcc-std.ver: Add __emutls_get_address, __emutls_register_common.
16 * output.h (emutls_finish): Declare.
17 * toplev.c (compile_file): Call it.
18 * tree-ssa-address.c (gen_addr_rtx): Check for const-ness of the
19 address before wrapping in CONST.
20 * varasm.c (emutls_htab, emutls_object_type): New.
21 (EMUTLS_VAR_PREFIX, EMUTLS_TMPL_PREFIX): New.
22 (get_emutls_object_name, get_emutls_object_type): New.
23 (get_emutls_init_templ_addr, emutls_decl): New.
24 (emutls_common_1, emutls_finish): New.
25 (assemble_variable): When emulating tls, swap decls; generate
26 constructor for the emutls objects.
27 (do_assemble_alias): When emulating tls, swap decl and target name.
28 (default_encode_section_info): Don't add SYMBOL_FLAG_TLS_SHIFT
30 * varpool.c (decide_is_variable_needed): Look at force_output.
31 Recurse for emulated tls.
32 (cgraph_varpool_remove_unreferenced_decls): Remove checks redundant
33 with decide_is_variable_needed.
35 * config/sparc/sol2.h (ASM_DECLARE_OBJECT_NAME): Only emit
36 tls_object for real tls.
38 2007-02-10 Kaz Kojima <kkojima@gcc.gnu.org>
40 PR rtl-optimization/29599
41 * reload1.c (eliminate_regs_in_insn): Take the destination
42 mode into account when computing the offset.
44 2007-02-09 Stuart Hastings <stuart@apple.com>
45 Richard Henderson <rth@redhat.com>
47 * gcc/config/i386/i386.h (TARGET_KEEPS_VECTOR_ALIGNED_STACK): New.
48 * gcc/config/i386/darwin.h: (TARGET_KEEPS_VECTOR_ALIGNED_STACK): New.
49 * gcc/config/i386/i386.md (fixuns_trunc<mode>si2, fixuns_truncsfhi2,
50 fixuns_truncdfhi2): New.
51 (fix_truncsfdi_sse): Call ix86_expand_convert_sign_didf_sse.
52 (floatunsdidf2): Call ix86_expand_convert_uns_didf_sse.
53 (floatunssisf2): Add call to ix86_expand_convert_uns_sisf_sse.
54 (floatunssidf2): Allow nonimmediate source.
55 * gcc/config/i386/sse.md (movdi_to_sse): New. (vec_concatv2di): Drop '*'.
56 * gcc/config/i386/i386-protos.h (ix86_expand_convert_uns_si_sse,
57 ix86_expand_convert_uns_didf_sse, ix86_expand_convert_uns_sidf_sse,
58 ix86_expand_convert_uns_sisf_sse, ix86_expand_convert_sign_didf_sse): New.
59 * gcc/config/i386/i386.c (ix86_expand_convert_uns_si_sse,
60 ix86_expand_convert_uns_didf_sse, ix86_expand_convert_uns_sidf_sse,
61 ix86_expand_convert_uns_sisf_sse, ix86_expand_convert_sign_didf_sse,
62 ix86_build_const_vector, ix86_expand_vector_init_one_nonzero): New.
63 (ix86_build_signbit_mask): Fix decl of v, refactor to call ix86_build_const_vector.
64 (x86_emit_floatuns): Rewrite.
66 2007-02-10 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
68 * genautomata.c (longest_path_length): Delete unused function.
69 (struct state): Delete unused longest_path_length.
70 (UNDEFINED_LONGEST_PATH_LENGTH): Delete unused macro.
71 (get_free_state): Delete unused.
73 2007-02-09 Jan Hubicka <jh@suse.cz>
75 * params.def (PARAM_INLINE_UNIT_GROWTH): Set to 30.
76 * doc/invoke.texi (inline-unit-growth): Update default value.
78 * Makefile.in (passes.o, ipa-inline.o): Add dependencies.
79 * cgraphbuild.c (build_cgraph_edges): Compute frequencies.
80 (rebuild_cgraph_edges): Likewise.
81 * cgraph.c (cgraph_set_call_stmt): Add new argument frequency.
82 (dump_cgraph_node): Dump frequencies.
83 (cgraph_clone_edge): Add frequency scales.
84 (cgraph_clone_node): Add freuqnecy.
85 * cgraph.h (cgraph_edge): Add freuqnecy argument.
86 (CGRAPH_FREQ_BASE, CGRAPH_FREQ_MAX): New constants.
87 (cgraph_create_edge, cgraph_clone_edge, cgraph_clone_node): Update.
88 * tree-pass.h (TODO_rebuild_frequencies): New constant.
89 * cgraphunit.c (verify_cgraph_node): Verify frequencies.
90 (cgraph_copy_node_for_versioning): Update call of cgraph_clone_edge.
91 (save_inline_function_body): Likewise.
92 * ipa-inline.c: inluce rtl.h
93 (cgraph_clone_inlined_nods): Update call of cgraph_clone_node.
94 (cgraph_edge_badness): Use frequencies.
95 (cgraph_decide_recursive_inlining): Update clonning.
96 (cgraph_decide_inlining_of_small_function): Dump frequency.
97 * predict.c (estimate_bb_frequencies): Export.
98 * predict.h (estimate_bb_frequencies): Declare.
99 * tree-inline.c (copy_bb): Watch overflows.
100 (expand_call_inline): Update call of cgraph_create_edge.
101 (optimize_inline_calls): Use TODO flags to update frequnecies.
102 * passes.h: Include predict.h
103 (init_optimization_passes): Move profile ahead.
104 (execute_function_todo): Handle TODO_rebuild_frequencies.
106 2007-02-09 Roger Sayle <roger@eyesopen.com>
108 * config/alpha/alpha.c (emit_insxl): Force the first operand of
109 the insbl or inswl pattern into a register.
111 2007-02-09 Roger Sayle <roger@eyesopen.com>
113 * config/ia64/ia64.md (bswapdi2): New define_insn.
115 2007-02-09 Richard Henderson <rth@redhat.com>
117 * config/i386/constraints.md (Ym): New constraint.
118 * config/i386/i386.md (movsi_1): Change Y2 to Yi constraints.
119 (movdi_1_rex64): Split sse and xmm general register moves from
120 memory move alternatives. Use conditional register constraints.
121 (movsf_1, movdf_integer): Likewise.
122 (zero_extendsidi2_32, zero_extendsidi2_rex64): Likewise.
123 (movdf_integer_rex64): New.
124 (pushsf_rex64): Fix output constraints.
125 * config/i386/sse.md (sse2_loadld): Split rm alternative, use Yi.
126 (sse2_stored): Likewise.
127 (sse2_storeq_rex64): New.
128 * config/i386/i386.c (x86_inter_unit_moves): Enable for not
130 (ix86_secondary_memory_needed): Don't bypass TARGET_INTER_UNIT_MOVES
131 for optimize_size. Remove SF/DFmode hack.
133 2007-02-09 Dwarakanath Rajagopal <dwarak.rajagopal@amd.com>
135 * config/i386/driver-i386.c: Turn on -mtune=native for AMDFAM10.
138 2007-02-09 Nathan Sidwell <nathan@codesourcery.com>
139 Richard Sandiford <richard@codesourcery.com>
141 * config.gcc (m68010-*-netbsdelf*, m68k*-*-netbsdelf*)
142 (m68k*-*-openbsd*, m68k-*-linux*): Set default_cf_cpu.
143 (m68k-*-aout*, m68k-*-coff*, m68k-*-uclinux*, m68k-*-rtems*): Add
144 m68k/t-mlib to tmake_file.
145 (m68020-*-elf*, m68k-*-elf*): Likewise. Add t-m68kbare as well.
146 (m68k*-*-*): Use --with-arch to pick a default for --with-cpu.
147 (m680[012]0-*-*, m68k*-*-*): Add support for --with-arch.
148 Allow it to be cf or m68k. Set m68k_arch_family. If that
149 variable is not empty, add t-$m68k_arch_family to tmake_file.
150 Add t-mlibs to tmake_file.
151 * doc/install.texi: Document --with-arch=m68k and --with-arch=cf.
152 * config/m68k/t-cf: New file.
153 * config/m68k/t-m68k: Likewise.
154 * config/m68k/t-mlibs: Likewise.
155 * config/m68k/t-m68kbare (MULTILIB_OPTIONS, MULTILIB_DIRNAMES)
156 (MULTILIB_MATCHES, MULTILIB_EXCEPTIONS): Delete.
157 (M68K_MLIB_DIRNAMES, M68K_MLIB_OPTIONS): Define.
158 * config/m68k/t-m68kelf (MULTILIB_OPTIONS, MULTILIB_DIRNAMES)
159 (MULTILIB_MATCHES, MULTILIB_EXCEPTIONS, LIBGCC, INSTALL_LIBGCC):
161 * config/m68k/t-openbsd (MULTILIB_OPTIONS, LIBGCC): Delete.
162 (INSTALL_LIBGCC): Delete.
163 (M68K_MLIB_DIRNAMES, M68K_MLIB_OPTIONS): Define.
164 * config/m68k/t-rtems (MULTILIB_OPTIONS, MULTILIB_DIRNAMES)
165 (MULTILIB_MATCHES, MULTILIB_EXCEPTIONS): Delete.
166 (M68K_MLIB_CPU): Define.
167 * config/m68k/t-uclinux (MULTILIB_OPTIONS, MULTILIB_DIRNAMES)
168 (MULTILIB_MATCHES, MULTILIB_EXCEPTIONS): Delete.
169 (M68K_MLIB_CPU, M68K_MLIB_OPTIONS, M68K_MLIB_DIRNAMES): Define.
171 2007-02-09 Zdenek Dvorak <dvorakz@suse.cz>
172 Richard Guenther <rguenther@suse.de>
175 * fold-const.c (fold_comparison): Handle obfuscated comparisons
176 against INT_MIN/INT_MAX.
177 * tree-ssa-loop-ivcanon.c (remove_empty_loop): Print to dump
178 file if a loop is removed.
180 2007-02-09 Joseph Myers <joseph@codesourcery.com>
182 * calls.c (store_one_arg): Pass correct alignment to
183 emit_push_insn for non-BLKmode values.
184 * expr.c (emit_push_insn): If STRICT_ALIGNMENT, copy to an
185 unaligned stack slot via a suitably aligned slot.
187 2007-02-08 DJ Delorie <dj@redhat.com>
189 * config/m32c/m32c.c (m32c_unpend_compare): Add default to silence
191 (legal_subregs): Use unsigned char, make const.
192 (m32c_illegal_subreg_p): Use ARRAY_SIZE. Delete unused variables.
194 2007-02-08 Paul Brook <paul@codesourcery.com>
196 * config/arm/lib1funcs.asm (RETLDM): Pop directly into PC when no
197 special interworking needed.
199 2007-02-08 Harsha Jagasia <harsha.jagasia@amd.com>
201 * config/i386/xmmintrin.h: Make inclusion of emmintrin.h
202 conditional to __SSE2__.
203 (Entries below should have been added to first ChangeLog
204 entry for amdfam10 dated 2007-02-05)
205 * config/i386/emmintrin.h: Generate #error if __SSE2__ is not
207 * config/i386/pmmintrin.h: Generate #error if __SSE3__ is not
209 * config/i386/tmmintrin.h: Generate #error if __SSSE3__ is not
212 2007-02-08 DJ Delorie <dj@redhat.com>
214 * config/m32c/m32c-protos.h (m32c_illegal_subreg_p): New.
215 * config/m32c/m32c.c (legal_subregs): New.
216 (m32c_illegal_subreg_p): New.
217 * config/m32c/predicates.md (m32c_any_operand): Use it to reject
218 unsupported subregs of hard regs.
220 2007-02-08 Jan Hubicka <jh@suse.cz>
222 * tree-cfg.c (bsi_replace): Shortcut when replacing the statement with
223 the same one; always update histograms.
225 2007-02-08 Diego Novillo <dnovillo@redhat.com>
227 * passes.c (init_optimization_passes): Tidy comment.
229 2007-02-08 Roger Sayle <roger@eyesopen.com>
231 * simplify-rtx.c (simplify_unary_operation_1) <POPCOUNT>: We can
232 strip zero_extend, bswap and rotates from POCOUNT's argument.
233 <PARITY>: Likewise, we can strip not, bswap, sign_extend,
234 zero_extend and rotates from PARITY's argument.
235 <BSWAP>: A byte-swap followed by a byte-swap is an identity.
236 (simplify_const_unary_operation) <BSWAP>: Evaluate the byte-swap
237 of an integer constant at compile-time.
239 2007-02-08 Diego Novillo <dnovillo@redhat.com>
242 * tree-flow.h (struct var_ann_d): Remove field 'is_used'.
244 * tree-ssa-alias.c (compute_is_aliased): Remove. Update all
247 * tree-ssa-live.c (remove_unused_locals): Do not remove
248 TREE_ADDRESSABLE variables.
249 * tree-ssa-structalias.c (compute_points_to_sets): Tidy.
250 * tree-ssa-operands.c (add_virtual_operand): Remove argument
251 FOR_CLOBBER. Update all users.
252 If VAR has an associated alias set, add a virtual operand for
253 it if no alias is found to conflict with the memory reference.
255 2007-02-07 Jan Hubicka <jh@suse.cz>
256 Robert Kidd <rkidd@crhc.uiuc.edu>
258 * value-prof.c (visit_hist, free_hist): Return 1 instead of 0.
260 2007-02-07 Ian Lance Taylor <iant@google.com>
262 * lower-subreg.c (simple_move): Reject PARTIAL_INT modes.
264 2007-02-07 Roger Sayle <roger@eyesopen.com>
266 * config/rs6000/rs6000.md (ctz<mode>2, ffs<mode>2, popcount<mode>2,
267 parity<mode>2, smulsi3_highpart, abstf2_internal, allocate_stack,
268 tablejumpdi, movsi_to_cr_one): Remove constraints from
269 define_expand's match_operands.
271 2007-02-07 Roger Sayle <roger@eyesopen.com>
273 * global.c (compute_regsets): Move declatation of "i" inside of
274 #ifdef ELIMINABLE_REGS to avoid unused variable bootstrap failure.
276 2007-02-07 Jakub Jelinek <jakub@redhat.com>
279 * gimplify.c (gimplify_scan_omp_clauses): Remove special casing
280 of INDIRECT_REF <RESULT_DECL>.
282 * config/i386/i386.c (override_options): Set PTA_SSSE3 for core2.
284 2007-02-06 J"orn Rennecke <joern.rennecke@arc.com>
285 Kaz Kojima <kkojima@gcc.gnu.org>
288 * config/sh/sh.c (expand_cbranchdi4): Use scratch register
290 (sh_initialize_trampoline): Add parentheses to avoid a warning.
292 2007-02-06 Zdenek Dvorak <dvorakz@suse.cz>
294 * doc/loop.texi: Document possibility not to perform disambiguation
295 of loops with multiple latches.
296 * cfgloopmanip.c (alp_enum_p): Removed.
297 (add_loop): Handle subloops. Use get_loop_body_with_size.
298 (create_preheader): Do not allow ENTRY_BLOCK_PTR to be preheader.
299 * cfghooks.c (redirect_edge_and_branch_force): Set dominator for
300 the new forwarder block.
301 (make_forwarder_block): Only call new_bb_cbk if it is not NULL.
302 Handle the case latch is NULL.
303 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Avoid cfg modifications
304 when marking loop exits.
305 * ifcvt.c (if_convert): Ditto. Mark loop exits even if cfg cannot
307 * loop-init.c (loop_optimizer_init): Do not modify cfg. Call
308 disambiguate_loops_with_multiple_latches.
309 * tree-cfgcleanup.c (cleanup_tree_cfg_loop): Calculate dominators
310 before fix_loop_structure.
311 * cfgloop.c: Include pointer-set.h and output.h.
312 (canonicalize_loop_headers, HEADER_BLOCK, LATCH_EDGE,
313 update_latch_info, mfb_keep_just, mfb_keep_nonlatch): Removed.
314 (get_loop_latch_edges, find_subloop_latch_edge_by_profile,
315 find_subloop_latch_edge_by_ivs, find_subloop_latch_edge,
316 mfb_redirect_edges_in_set, form_subloop, merge_latch_edges,
317 disambiguate_multiple_latches, get_loop_body_with_size,
318 disambiguate_loops_with_multiple_latches): New functions.
319 (flow_loop_dump): Dump multiple latch edges.
320 (flow_loop_nodes_find): Handle loops with multiple latches.
321 (flow_loops_find): Ditto. Do not call canonicalize_loop_headers.
322 (glb_enum_p): Modified.
323 (get_loop_body): Use get_loop_body_with_size.
324 * cfgloop.h (LOOPS_HAVE_RECORDED_EXITS): New flag.
325 (AVOID_CFG_MODIFICATIONS): New constant.
326 (disambiguate_loops_with_multiple_latches, add_loop,
327 get_loop_body_with_size): Declare.
328 * Makefile.in (cfgloop.o): Add pointer-set.h and output.h.
330 2007-02-06 Seongbae Park <seongbae.park@gmail.com>
333 * global.c (compute_regsets): New function.
334 (global_alloc): Refactored ELIMINABLE_REGSET
335 and NO_GLOBAL_ALLOC_REGS computation out.
336 (rest_of_handle_global_alloc): Call compute_regsets()
337 for non-optimizing case.
339 2007-02-06 Richard Henderson <rth@redhat.com>
341 * config/i386/constraints.md (Y2): Rename from Y.
342 (Yi): New constraint.
343 * config/i386/i386.md (movsi_1, movdi_2, pushdf_nointeger,
344 pushdf_integer, movdf_nointeger, movdf_integer, zero_extendsidi2_32,
345 zero_extendsidi2_rex64, truncxfdf2_mixed): Change Y constraints to Y2.
346 (extendsfdf2_mixed, extendsfdf2_sse, truncdfsf_fast_mixed,
347 truncdfsf_fast_sse, truncdfsf_mixed, fix_truncdfdi_sse,
348 fix_truncdfsi_sse, floatsidf2_mixed, floatsidf2_sse,
349 floatdidf2_mixed, floatdidf2_sse, absnegdf2_mixed,
350 absnegdf2_sse, sse_setccdf, fop_df_comm_mixed, fop_df_comm_sse,
351 fop_df_1_mixed, fop_df_1_sse): Change Y constraints to x.
352 * config/i386/mmx.md (mov<MMXMODEI>_internal_rex64,
353 mov<MMXMODEI>_internal, movv2sf_internal_rex64, movv2sf_internal,
354 vec_extractv2si_1): Change Y constraints to Y2.
355 * config/i386/sse.md (vec_setv4sf_0, vec_concatv2df, vec_dupv4si,
356 vec_dupv2di, sse2_concatv2si, vec_concatv4si_1, vec_concatv2di):
357 Change Y constraints to Y2.
358 (sse2_loadld): Change Y constraints to x.
360 2007-02-06 Roger Sayle <roger@eyesopen.com>
362 * config/rs6000/rs6000.md (popcount<mode>2): Rewrite.
363 (parity<mode>2): New define_expand using rs6000_emit_parity.
364 * config/rs6000/rs6000.c (rs6000_emit_popcount,
365 rs6000_emit_parity): New functions.
366 * config/rs6000/rs6000-protos.h (rs6000_emit_popcount,
367 rs6000_emit_parity): Prototype here.
369 2007-02-06 Ian Lance Taylor <iant@google.com>
371 * lower-subreg.c (simple_move_operand): Reject CONST.
372 (resolve_clobber): Call validate_change rather than directly
373 assigning to XEXP (pat, 0).
375 2006-02-06 Paolo Bonzini <bonzini@gnu.org>
377 * Makefile.in (tree-ssa-loop-ivopts.o): Add pointer-set.h dependency.
378 (tree-ssa-reassoc.o): Add pointer-set.h dependency.
379 (tree-cfg.o): Remove hashtab.h dependency.
381 * tree-ssa-loop-ivopts.c: Include pointer-set.h.
382 (struct ivopts_data): Change niters to pointer_map_t.
383 (struct nfe_cache_elt, nfe_hash, nfe_eq): Delete.
384 (niter_for_exit): Create pointer_map on demand. Change for
386 (tree_ssa_iv_optimize_init): Initialize data->niters to NULL.
387 (free_loop_data): Destroy data->niters if created and reset field.
388 (tree_ssa_iv_optimize_finalize): Don't delete data->niters here.
389 (tree_ssa_iv_optimize_loop): Check for presence of stale data.
391 * tree-ssa-reassoc.c: Include pointer-set.h.
392 (bb_rank): Change to long *.
393 (operand_rank): Change to pointer_map_t.
394 (find_operand_rank): Return long, -1 if not found. Declare as inline.
395 (insert_operand_rank): Accept long.
396 (operand_entry_hash, operand_entry_eq): Remove.
397 (get_rank): Return long. Adjust for changes above.
398 (init_reassoc): Change rank type to long. Adjust creation of bb_rank
400 (fini_reassoc): Delete operand_rank with pointer_map_destroy.
402 * tree-ssa-structalias.c (vi_for_tree): Change to pointer_map.
403 (struct tree_vi, tree_vi_t, tree_vi_hash, tree_vi_eq): Delete.
404 (insert_vi_for_tree): Rewrite for pointer_map API. Assert argument
406 (lookup_vi_for_tree): Rewrite for pointer_map API. Return varinfo_t
407 directly since it cannot be NULL.
408 (get_vi_for_tree): Rewrite for pointer_map API.
409 (find_what_p_points_to): Adjust for change to lookup_vi_for_tree.
410 (init_alias_vars): Create vi_for_tree as pointer_map.
411 (delete_points_to_sets): Delete vi_for_tree using pointer_map_destroy.
413 * tree-cfg.c: Don't include hashtab.h.
414 (edge_to_cases): Declare as pointer_map.
415 (struct edge_to_cases_elt, edge_to_cases_hash, edge_to_cases_eq):
417 (edge_to_cases_cleanup): Rewrite as pointer_map_traverse callback.
418 (start_recording_case_labels): Create edge_to_cases as pointer_map.
419 (end_recoding_case_labels): Cleanup edge_to_cases manually before
421 (record_switch_edge): Delete.
422 (get_cases_for_edge): Adjust for pointer_map API, inline
423 record_switch_edge (rewritten for new API), remove goto.
425 2006-02-06 Paolo Bonzini <bonzini@gnu.org>
427 * Makefile.in (tree-nested.o): Add pointer-set.h dependency.
428 * tree-nested.c: Include pointer-set.h.
429 (var_map_elt, var_map_eq, var_map_hash): Delete.
430 (struct nesting_info): Remove GTY marker. Change the two htab_t's
432 (nesting_info_bitmap_obstack): New.
433 (lookup_field_for_decl): Adjust for pointer_map API.
434 (lookup_tramp_for_decl): Adjust for pointer_map API.
435 (get_nonlocal_debug_decl): Adjust for pointer_map API.
436 (get_local_debug_decl): Adjust for pointer_map API.
437 (convert_nl_goto_reference): Adjust for pointer_map API.
438 (convert_nl_goto_receiver): Adjust for pointer_map API.
439 (create_nesting_tree): Create outside GGC space. Create bitmap on
440 the new obstack. Create field_map and var_map as pointer_maps.
441 (free_nesting_tree): Adjust for changes to create_nesting_tree.
443 (lower_nested_functions): Move root here, no need to NULL it.
444 Initialize and release the obstack.
446 2007-02-06 Paolo Bonzini <bonzini@gnu.org>
448 * tree.c (tree_int_map_hash, tree_int_map_eq, tree_int_map_marked_p):
449 Remove prototypes and make them non-static.
450 (struct tree_int_map): Remove.
451 * tree.h (struct tree_int_map): Move here, turning TO into an
453 (tree_int_map_hash, tree_int_map_eq, tree_int_map_marked_p): Declare.
455 * tree.h (TREE_COMPLEXITY): Remove.
456 (struct tree_exp): Remove complexity field.
457 * tree.c (build1_stat): Don't set it.
459 2007-02-06 Dorit Nuzman <dorit@il.ibm.com>
460 Victor Kaplansky <victork@il.ibm.com>
462 * tree-vectorizer.c (vect_is_simple_use): Support induction.
463 (vect_is_simple_reduction): Support reduction with induction as
465 (vect_is_simple_iv_evolution): Fix formatting.
466 * tree-vect-analyze.c (vect_mark_stmts_to_be_vectorized): Fix
467 formatting. Don't mark induction phis for vectorization.
468 (vect_analyze_scalar_cycles): Analyze all inductions, then reductions.
469 * tree-vect-transform.c (get_initial_def_for_induction): New function.
470 (vect_get_vec_def_for_operand): Support induction.
471 (vect_get_vec_def_for_stmt_copy): Fix formatting and add check for
473 (vectorizable_reduction): Support reduction with induction as one of
475 (vectorizable_type_demotion): Use def-type of stmt argument rather
478 * tree-ssa-loop.c (gate_scev_const_prop): Return the value of
479 flag_tree_scev_cprop.
480 * common.opt (tree-scev-cprop): New flag.
482 * tree-vect-transform.c (vect_create_destination_var): Use 'kind' in
483 call to vect_get_new_vect_var.
485 2007-02-06 Ira Rosen <irar@il.ibm.com>
487 * tree-vect-patterns.c (vect_recog_widen_mult_pattern): Check that
489 (vect_pattern_recog_1): Likewise.
491 2007-02-05 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
493 * fold-const.c (negate_expr_p): Handle CONJ_EXPR.
494 (fold_negate_expr): Likewise.
496 2007-02-05 Alexandre Oliva <aoliva@redhat.com>
499 * dwarf2out.c (modified_type_die): Follow DECL_ORIGINAL_TYPE
500 even if cv-qualification is the same.
502 2007-02-05 Geoffrey Keating <geoffk@apple.com>
504 * config/rs6000/darwin-tramp.asm (__trampoline_setup): Call
505 __enable_execute_stack on completion.
507 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
509 * config/i386/athlon.md (athlon_fldxf_k8, athlon_fld_k8,
510 athlon_fstxf_k8, athlon_fst_k8, athlon_fist, athlon_fmov,
511 athlon_fadd_load, athlon_fadd_load_k8, athlon_fadd, athlon_fmul,
512 athlon_fmul_load, athlon_fmul_load_k8, athlon_fsgn,
513 athlon_fdiv_load, athlon_fdiv_load_k8, athlon_fdiv_k8,
514 athlon_fpspc_load, athlon_fpspc, athlon_fcmov_load,
515 athlon_fcmov_load_k8, athlon_fcmov_k8, athlon_fcomi_load_k8,
516 athlon_fcomi, athlon_fcom_load_k8, athlon_fcom): Added amdfam10.
518 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
520 * config/i386/i386.md (x86_sahf_1, cmpfp_i_mixed, cmpfp_i_sse,
521 cmpfp_i_i387, cmpfp_iu_mixed, cmpfp_iu_sse, cmpfp_iu_387,
522 swapsi, swaphi_1, swapqi_1, swapdi_rex64, fix_truncsfdi_sse,
523 fix_truncdfdi_sse, fix_truncsfsi_sse, fix_truncdfsi_sse,
524 x86_fldcw_1, floatsisf2_mixed, floatsisf2_sse, floatdisf2_mixed,
525 floatdisf2_sse, floatsidf2_mixed, floatsidf2_sse,
526 floatdidf2_mixed, floatdidf2_sse, muldi3_1_rex64, mulsi3_1,
527 mulsi3_1_zext, mulhi3_1, mulqi3_1, umulqihi3_1, mulqihi3_insn,
528 umulditi3_insn, umulsidi3_insn, mulditi3_insn, mulsidi3_insn,
529 umuldi3_highpart_rex64, umulsi3_highpart_insn,
530 umulsi3_highpart_zext, smuldi3_highpart_rex64,
531 smulsi3_highpart_insn, smulsi3_highpart_zext, x86_64_shld,
532 x86_shld_1, x86_64_shrd, sqrtsf2_mixed, sqrtsf2_sse,
533 sqrtsf2_i387, sqrtdf2_mixed, sqrtdf2_sse, sqrtdf2_i387,
534 sqrtextendsfdf2_i387, sqrtxf2, sqrtextendsfxf2_i387,
535 sqrtextenddfxf2_i387): Added amdfam10_decode.
537 * config/i386/athlon.md (athlon_idirect_amdfam10,
538 athlon_ivector_amdfam10, athlon_idirect_load_amdfam10,
539 athlon_ivector_load_amdfam10, athlon_idirect_both_amdfam10,
540 athlon_ivector_both_amdfam10, athlon_idirect_store_amdfam10,
541 athlon_ivector_store_amdfam10): New define_insn_reservation.
542 (athlon_idirect_loadmov, athlon_idirect_movstore): Added
545 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
547 * config/i386/athlon.md (athlon_call_amdfam10,
548 athlon_pop_amdfam10, athlon_lea_amdfam10): New
549 define_insn_reservation.
550 (athlon_branch, athlon_push, athlon_leave_k8, athlon_imul_k8,
551 athlon_imul_k8_DI, athlon_imul_mem_k8, athlon_imul_mem_k8_DI,
552 athlon_idiv, athlon_idiv_mem, athlon_str): Added amdfam10.
554 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
556 * config/i386/athlon.md (athlon_sseld_amdfam10,
557 athlon_mmxld_amdfam10, athlon_ssest_amdfam10,
558 athlon_mmxssest_short_amdfam10): New define_insn_reservation.
560 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
562 * config/i386/athlon.md (athlon_sseins_amdfam10): New
563 define_insn_reservation.
564 * config/i386/i386.md (sseins): Added sseins to define_attr type
565 and define_attr unit.
566 * config/i386/sse.md: Set type attribute to sseins for insertq
569 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
571 * config/i386/athlon.md (sselog_load_amdfam10, sselog_amdfam10,
572 ssecmpvector_load_amdfam10, ssecmpvector_amdfam10,
573 ssecomi_load_amdfam10, ssecomi_amdfam10,
574 sseaddvector_load_amdfam10, sseaddvector_amdfam10): New
575 define_insn_reservation.
576 (ssecmp_load_k8, ssecmp, sseadd_load_k8, seadd): Added amdfam10.
578 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
580 * config/i386/athlon.md (cvtss2sd_load_amdfam10,
581 cvtss2sd_amdfam10, cvtps2pd_load_amdfam10, cvtps2pd_amdfam10,
582 cvtsi2sd_load_amdfam10, cvtsi2ss_load_amdfam10,
583 cvtsi2sd_amdfam10, cvtsi2ss_amdfam10, cvtsd2ss_load_amdfam10,
584 cvtsd2ss_amdfam10, cvtpd2ps_load_amdfam10, cvtpd2ps_amdfam10,
585 cvtsX2si_load_amdfam10, cvtsX2si_amdfam10): New
586 define_insn_reservation.
588 * config/i386/sse.md (cvtsi2ss, cvtsi2ssq, cvtss2si,
589 cvtss2siq, cvttss2si, cvttss2siq, cvtsi2sd, cvtsi2sdq,
590 cvtsd2si, cvtsd2siq, cvttsd2si, cvttsd2siq,
591 cvtpd2dq, cvttpd2dq, cvtsd2ss, cvtss2sd,
592 cvtpd2ps, cvtps2pd): Added amdfam10_decode attribute.
594 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
596 * config/i386/athlon.md (athlon_ssedivvector_amdfam10,
597 athlon_ssedivvector_load_amdfam10, athlon_ssemulvector_amdfam10,
598 athlon_ssemulvector_load_amdfam10): New define_insn_reservation.
599 (athlon_ssediv, athlon_ssediv_load_k8, athlon_ssemul,
600 athlon_ssemul_load_k8): Added amdfam10.
602 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
604 * config/i386/i386.h (TARGET_SSE_UNALIGNED_MOVE_OPTIMAL): New macro.
605 (x86_sse_unaligned_move_optimal): New variable.
607 * config/i386/i386.c (x86_sse_unaligned_move_optimal): Enable for
609 (ix86_expand_vector_move_misalign): Add code to generate movupd/movups
610 for unaligned vector SSE double/single precision loads for AMDFAM10.
612 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
614 * config/i386/i386.h (TARGET_AMDFAM10): New macro.
615 (TARGET_CPU_CPP_BUILTINS): Add code for amdfam10.
616 Define TARGET_CPU_DEFAULT_amdfam10.
617 (TARGET_CPU_DEFAULT_NAMES): Add amdfam10.
618 (processor_type): Add PROCESSOR_AMDFAM10.
620 * config/i386/i386.md: Add amdfam10 as a new cpu attribute to match
621 processor_type in config/i386/i386.h.
622 Enable imul peepholes for TARGET_AMDFAM10.
624 * config.gcc: Add support for --with-cpu option for amdfam10.
626 * config/i386/i386.c (amdfam10_cost): New variable.
627 (m_AMDFAM10): New macro.
628 (m_ATHLON_K8_AMDFAM10): New macro.
629 (x86_use_leave, x86_push_memory, x86_movx, x86_unroll_strlen,
630 x86_cmove, x86_3dnow_a, x86_deep_branch, x86_use_simode_fiop,
631 x86_promote_QImode, x86_integer_DFmode_moves,
632 x86_partial_reg_dependency, x86_memory_mismatch_stall,
633 x86_accumulate_outgoing_args, x86_arch_always_fancy_math_387,
634 x86_sse_partial_reg_dependency, x86_sse_typeless_stores,
635 x86_use_ffreep, x86_use_incdec, x86_four_jump_limit,
636 x86_schedule, x86_use_bt, x86_cmpxchg16b, x86_pad_returns):
637 Enable/disable for amdfam10.
638 (override_options): Add amdfam10_cost to processor_target_table.
639 Set up PROCESSOR_AMDFAM10 for amdfam10 entry in
640 processor_alias_table.
641 (ix86_issue_rate): Add PROCESSOR_AMDFAM10.
642 (ix86_adjust_cost): Add code for amdfam10.
644 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
646 * config/i386/i386.opt: Add new Advanced Bit Manipulation (-mabm)
647 instruction set feature flag. Add new (-mpopcnt) flag for popcnt
648 instruction. Add new SSE4A (-msse4a) instruction set feature flag.
649 * config/i386/i386.h: Add builtin definition for SSE4A.
650 * config/i386/i386.md: Add support for ABM instructions
652 * config/i386/sse.md: Add support for SSE4A instructions
653 (movntss, movntsd, extrq, insertq).
654 * config/i386/i386.c: Add support for ABM and SSE4A builtins.
655 Add -march=amdfam10 flag.
656 * config/i386/ammintrin.h: Add support for SSE4A intrinsics.
657 * doc/invoke.texi: Add documentation on flags for sse4a, abm, popcnt
659 * doc/extend.texi: Add documentation for SSE4A builtins.
661 2007-02-05 Bob Wilson <bob.wilson@acm.org>
663 * config/xtensa/xtensa.c (constantpool_mem_p): Skip over SUBREGs.
665 2007-02-05 Richard Guenther <rguenther@suse.de>
667 * tree-vectorizer.h (vectorizable_function): Add argument type
668 argument, change return type.
669 * tree-vect-patterns.c (vect_recog_pow_pattern): Adjust caller.
670 * tree-vect-transform.c (vectorizable_function): Handle extra
671 argument, return vectorized function decl.
672 (build_vectorized_function_call): Remove.
673 (vectorizable_call): Handle calls with result and argument types
674 differing. Handle loop vectorization factor correctly.
675 * targhooks.c (default_builtin_vectorized_function): Adjust for
677 * targhooks.h (default_builtin_vectorized_function): Likewise.
678 * target.h (builtin_vectorized_function): Add argument type
680 * config/i386/i386.c (ix86_builtin_vectorized_function): Handle
681 extra argument, allow vectorizing of lrintf.
682 * doc/tm.texi (TARGET_VECTORIZE_BUILTIN_VECTORIZED_FUNCTION): Adjust
683 documentation of target hook.
685 2007-02-05 Hans-Peter Nilsson <hp@axis.com>
688 * config/cris/cris.md ("*andsi_movu", "*andsi_clear", "*andhi_movu")
689 ("*andhi_clear", andu (casesi+45)): For size-changed operand where
690 memory is allowed, require !side_effects_p, not just !MEM_VOLATILE_P.
692 2007-02-05 Roger Sayle <roger@eyesopen.com>
694 * fold-const.c (fold_unary) <REAL_PART>: Test for availability of
695 BUILT_IN_COS before simplifying REAL_PART(CEXPI)) to COS.
696 <IMAG_PART>: Likewise, check for availability of BUILT_IN_SIN.
697 * builtins.c (fold_builtin_sincos): Check for TARGET_C99_FUNCTIONS
698 before canonicalizing sincos to cexpi.
699 (fold_builtin_cexp): Likewise, for canonicalizing cexp to cexpi.
701 2007-02-05 Roger Sayle <roger@eyesopen.com>
703 * config/alpha/alpha.c (alpha_add_builtins): New Helper function.
704 Set TREE_READONLY and TREE_NOTHROW directly, not via attributes.
705 (alpha_init_builtins): Use alpha_add_builtins to process tables.
707 2007-02-05 Roger Sayle <roger@eyesopen.com>
709 * mips-tfile.c (initialize_init_file): Correct endianness test.
711 2007-02-05 Kazu Hirata <kazu@codesourcery.com>
713 * config/m68k/m68k.md (pushdi-1, pushdi, movsi+1): Don't use
716 2007-02-05 Richard Sandiford <richard@codesourcery.com>
718 * dwarf2out.c (dwarf2out_frame_debug_expr): Record the register
719 saves in a PARALLEL before the register assignments.
721 2007-02-05 Richard Sandiford <richard@codesourcery.com>
723 * doc/tm.texi (DWARF_ALT_FRAME_RETURN_COLUMN): Do not require
724 DWARF_FRAME_RETURN_COLUMN to be a general register.
725 * dwarf2out.c (init_return_column_size): New function, split from...
726 (expand_builtin_init_dwarf_reg_sizes): ...here. Allow both
727 DWARF_FRAME_RETURN_COLUMN and DWARF_ALT_FRAME_RETURN_COLUMN
728 to be nongeneral registers.
729 * config/m68k/m68k.h (DWARF_FRAME_REGNUM): Only map FP and
731 (DWARF_FRAME_REGISTERS, DWARF_FRAME_RETURN_COLUMN): Define.
732 (DWARF_ALT_FRAME_RETURN_COLUMN): Define.
734 2007-02-04 Zdenek Dvorak <dvorakz@suse.cz>
736 * cfgcleanup.c (try_optimize_cfg): Avoid removing ENTRY_BLOCK_PTR.
738 2007-02-04 Zdenek Dvorak <dvorakz@suse.cz>
740 * cfgloopmanip.c (loop_delete_branch_edge): Removed.
741 (remove_path): Use can_remove_branch_p and remove_branch instead
742 of loop_delete_branch_edge.
743 * tree-ssa-loop-manip.c (scale_dominated_blocks_in_loop): New function.
744 (tree_transform_and_unroll_loop): Remove dead branches immediately.
745 Update profile using scale_dominated_blocks_in_loop.
746 * cfghooks.c (can_remove_branch_p, remove_branch): New functions.
747 * cfghooks.h (struct cfg_hooks): Add can_remove_branch_p.
748 (can_remove_branch_p, remove_branch): Declare.
749 * tree-cfg.c (tree_can_remove_branch_p): New function.
750 (tree_cfg_hooks): Add tree_can_remove_branch_p.
751 * cfgrtl.c (rtl_can_remove_branch_p): New function.
752 (rtl_cfg_hooks, cfg_layout_rtl_cfg_hook): Add rtl_can_remove_branch_p.
754 2007-02-05 Jan Hubicka <jh@suse.cz>
757 * ipa-inline.c (cgraph_clone_inlined_nodes): When there are unanalyzed
758 nodes in cgraph, don't remove offline copy of the function.
760 2007-02-04 Jan Hubicka <jh@suse.cz>
762 * tree-sra.c (sra_walk_expr): Add linebreaks. BITFIELD_REFs into
763 vectors might cause maybe_lookup_element_for_expr to be called
764 on non-sra-candidate.
766 2007-02-04 Kazu Hirata <kazu@codesourcery.com>
768 * config/bfin/bfin-modes.def, config/bfin/bfin.c,
769 config/bfin/bfin.md, config/bfin/predicates.md: Follow
770 spelling conventions.
772 2007-02-04 Richard Guenther <rguenther@suse.de>
775 * fold-const.c (try_move_mult_to_index): Make sure to not
776 overflow one dimension of a multi-dimensional array access.
778 2007-02-04 Jan Hubicka <jh@suse.cz>
780 * passes.c (init_optimization_passes): Reindent.
782 2007-02-04 Jan Hubicka <jh@suse.cz>
783 Eric Botcazou <ebotcazou@adacore.com>
785 * tree-optimize.c (has_abnormal_outgoing_edge_p): Move to...
786 (execute_fixup_cfg): Break out the abnormal goto code.
787 * tree-inline.c (has_abnormal_outgoing_edge_p): ...here.
788 (make_nonlocal_label_edges): Move here from execute_fixup_cfg.
789 (optimize_inline_calls): Call make_nonlocal_label_edges.
791 2007-02-04 Jan Hubicka <jh@suse.cz>
793 * tree-ssa-copyrename.c (copy_rename_partition_coalesce): Return
794 true when something was changed.
795 (rename_ssa_copies): When something was changed, do
796 TODO_remove_unused_locals.
797 * tree-ssa-forwprop.c (tree_ssa_forward_propagate_single_use_value):
798 add TODO_remove_unused_locals when instruction was removed.
800 2007-02-04 Jan Hubicka <jh@suse.cz>
802 * ipa-inline.c (try_inline): Improve debug output; work on already
804 (cgraph_decide_inlining_incrementally): Indent; improve debug output;
805 call try_inline for already inlined edges too when flattening;
806 inline also functions that make callee growth but overall unit size
809 2007-02-04 Kazu Hirata <kazu@codesourcery.com>
811 * config/m32c/bitops.md, config/m32c/jump.md,
812 config/m32c/m32c.c, config/m32c/m32c.h, config/m32r/m32r.c,
813 config/m32r/m32r.h, config/m32r/m32r.md,
814 config/m32r/predicates.md, config/m68hc11/larith.asm,
815 config/m68hc11/m68hc11.c, config/m68hc11/m68hc11.h,
816 config/m68k/m68k.h, config/mcore/mcore.md, config/mips/4k.md,
817 config/mips/mips-protos.h, config/mips/mips.c,
818 config/mips/mips.h, config/mips/mips.md, config/mips/mips16.S,
819 config/mn10300/mn10300.h, config/mn10300/predicates.md,
820 config/mt/mt.c, config/mt/mt.h, config/mt/mt.md: Follow
821 spelling conventions.
823 * config/v850/v850.c, config/v850/v850.h, config/v850/v850.md:
824 Follow spelling conventions.
826 2007-02-03 Douglas Gregor <doug.gregor@gmail.com>
828 * c-opts.c (c_common_post_options): If C++0x mode is enabled, don't
829 warn about C++0x compatibility.
831 2007-02-04 Kazu Hirata <kazu@codesourcery.com>
833 * config/h8300/h8300.c, config/h8300/h8300.h,
834 config/h8300/h8300.md: Follow spelling conventions.
836 2007-02-03 Uros Bizjak <ubizjak@gmail.com>
839 * combine.c (try_combine): Do not substitute source operand
840 with constants wider than 2 * HOST_BITS_PER_WIDE_INT.
842 2007-02-03 Jan Hubicka <jh@suse.cz>
844 PR gcov-profile/30650
845 * value-prof.c (stringop_block_profile): Fix handling of size counter;
846 do not divide by zero for never executed counters.
847 (tree_find_values_to_profile): Fix counters.
848 * gcov-ui.h (GCOV_COUNTER_AVERAGE, GCOV_COUNTER_IOR): Fix comments.
850 2007-02-03 Ian Lance Taylor <iant@google.com>
852 * lower-subreg.c (simple_move_operand): New static function,
853 broken out of simple_move. Reject LABEL_REF, SYMBOL_REF, and HIGH
855 (simple_move): Call simple_move_operand.
856 (find_decomposable_subregs): Add special handling of MEMs.
857 (can_decompose_p): Rename from cannot_decompose_p. Reverse
858 meaning of return value. If we see a hard register, test whether
859 it can store a word_mode value. Change all callers.
861 2007-02-03 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
863 * pa.md (addvdi3, addvsi3, subvdi3, subvsi3, negvdi2, negvsi2): New
864 ftrapv insns and expanders.
865 (subdi3): Change define_expand operand 1 to arith11_operand, and
866 operand 2 to reg_or_0_operand. Change constraints of 64-bit insn
867 pattern to handle reg_or_0 operands. Revise 32-bit insn pattern to
868 handle 11-bit constants and reg_or_0 operands in operands 1 and 2,
872 * varasm.c (notice_global_symbol): Treat global objects as weak when
875 2007-02-03 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
877 * emit-rtl.c (dconstpi): Delete.
879 (init_emit_once): Delete dconstpi and init dconstsqrt2.
880 * real.h (dconstpi): Delete.
882 * builtins.c (fold_builtin_cabs): Use dconstsqrt2.
883 (fold_builtin_hypot): Likewise.
885 2007-02-03 Tom Tromey <tromey@redhat.com>
888 * gcc.c (cpp_unique_options): Any of -ggdb3, -gstabs3,
889 -gcoff3, -gxcoff3, -gvms3 implies -dD.
891 2007-02-03 Kazu Hirata <kazu@codesourcery.com>
893 * c-decl.c, config/avr/avr.c, config/avr/avr.h,
894 config/m68k/m68k.c, config/m68k/netbsd-elf.h,
895 config/mn10300/mn10300.c, config/pdp11/pdp11.h,
896 config/rs6000/cell.md, config/rs6000/darwin.h,
897 config/sh/sh.md, config/sh/sh4-300.md, config/spu/spu.c,
898 config/spu/spu.md, cselib.c, expr.c, haifa-sched.c, hwint.h,
899 jump.c, reload.c, sched-deps.c, sched-int.h, tree-inline.c,
900 tree-profile.c, tree-ssa-live.h, tree-vrp.c: Fix comment
901 typos. Follow spelling conventions.
902 * doc/invoke.texi: Follow spelling conventions.
904 2007-02-03 Roger Sayle <roger@eyesopen.com>
906 * simplify-rtx.c (simplify_relational_operation_1): Implement some
907 canonicalization transformations that attempt to simplify integer
908 constant comparisons to become comparisons against zero.
910 2007-02-03 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
912 * builtins.c (fold_builtin_cabs): Fold cabs(x+xi) into
914 * fold-const.c (fold_binary): Fix comment typos. Fold complex
915 (x,0)-(0,y) into (x,-y). Likewise (0,y)-(x,0) into (-x,y).
917 2007-02-02 Mike Stump <mrs@apple.com>
919 * config/darwin9.h (DARWIN_LINKER_GENERATES_ISLANDS): Add.
920 * config/rs6000/rs6000.c (DARWIN_GENERATE_ISLANDS): Add.
921 (output_call): Use DARWIN_GENERATE_ISLANDS to decide when to
922 generate a branch island.
924 2007-02-02 Bob Wilson <bob.wilson@acm.org>
926 * config/xtensa/xtensa.c (smalloffset_mem_p): Use BASE_REG_P.
927 (xtensa_legitimate_address_p): New.
928 (xtensa_legitimize_address): New.
929 (xtensa_output_addr_const_extra): New.
930 * config/xtensa/xtensa.h (REG_OK_STRICT_FLAG): Define.
932 (REG_OK_FOR_BASE_P): Use BASE_REG_P.
933 (GO_IF_LEGITIMATE_ADDRESS): Move code to xtensa_legitimate_address_p.
934 (LEGITIMIZE_ADDRESS): Move code to xtensa_legitimize_address.
935 (OUTPUT_ADDR_CONST_EXTRA): Move code to xtensa_output_addr_const_extra.
936 * config/xtensa/xtensa-protos.h (xtensa_legitimate_address_p): New.
937 (xtensa_legitimize_address): New.
938 (xtensa_output_addr_const_extra): New.
940 2007-02-02 Steve Ellcey <sje@cup.hp.com>
942 * config/ia64/ia64.c (ia64_print_operand): Fix compare strings.
944 2007-02-02 Ian Lance Taylor <iant@google.com>
946 * expmed.c (expand_divmod): Add comment.
948 2007-02-02 Kazu Hirata <kazu@codesourcery.com>
950 * emit-rtl.c (renumber_insns): Remove.
951 * flags.h: Remove the extern for flag_renumber_insns.
952 * rtl.h: Remove the prototype for renumber_insns.
953 * toplev.c (flag_renumber_insns): Remove.
955 2007-02-02 Hui-May Chang <hm.chang@apple.com>
957 Revert for x86 darwin:
958 2005-06-19 Uros Bizjak <uros@kss-loka.si>
960 * config/i386/i386.c (ix86_function_arg_regno_p): Put back the
961 code before the following patch under TARGET_MACHO.
962 (ix86_function_value_regno_p): Likewise.
964 2007-02-02 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
966 * fold-const.c (negate_expr_p, fold_negate_expr): Handle
969 2007-02-02 Maxim Kuvyrkov <mkuvyrkov@ispras.ru>
972 * config/ia64/ia64.c (ia64_speculate_insn): Restrict to memory loads to
973 general or fp registers. Add comments.
974 * config/ia64/ia64.md (reg_pred_prefix): Add comment.
976 2007-02-02 Paolo Bonzini <bonzini@gnu.org>
978 * pointer-set.c (insert_aux): Only return insertion slot.
979 (pointer_set_insert): Adjust.
980 (pointer_set_traverse, struct pointer_map_t, pointer_map_create,
981 pointer_map_destroy, pointer_map_insert, pointer_map_contains,
982 pointer_map_traverse): New.
983 * pointer-set.h (pointer_set_traverse, struct pointer_map_t,
984 pointer_map_create, pointer_map_destroy, pointer_map_insert,
985 pointer_map_contains, pointer_map_traverse): Declare.
987 2007-02-02 Jakub Jelinek <jakub@redhat.com>
990 * builtins.c (fold_builtin_sprintf): Do not attempt to optimize
991 sprintf (str, "%s"). Do not optimize sprintf (str, "nopercent", p++).
993 2007-02-02 Maxim Kuvyrkov <mkuvyrkov@ispras.ru>
995 * sched-int.h (ds_to_dk, dk_to_ds): Declare functions.
997 (struct _dep): New type.
998 (dep_t): New typedef.
999 (DEP_PRO, DEP_CON, DEP_KIND): New access macros.
1000 (DEP_STATUS): New access macro. The macro with the same name was
1001 renamed to DEP_LINK_STATUS.
1002 (dep_init): Declare function
1004 (struct _dep_link): New type.
1005 (dep_link_t): New typedef.
1006 (DEP_LINK_NODE, DEP_LINK_NEXT, DEP_LINK_PREV_NEXTP): New access macros.
1007 (DEP_LINK_DEP, DEP_LINK_PRO, DEP_LINK_CON, DEP_LINK_KIND): New macros.
1008 (DEP_LINK_STATUS): New macro.
1009 (debug_dep_links): New debug function.
1011 (struct _deps_list): New type.
1012 (deps_list_t): New typedef.
1013 (DEPS_LIST_FIRST): New access macro.
1014 (FOR_EACH_DEP_LINK): New cycle macro.
1015 (create_deps_list, free_deps_list, delete_deps_list): Declare
1017 (deps_list_empty_p, debug_deps_list, add_back_dep_to_deps_list): Ditto.
1018 (find_link_by_pro_in_deps_list, find_link_by_con_in_deps_list): Ditto.
1019 (copy_deps_list_change_con): Ditto.
1021 (move_dep_link): Declare function.
1023 (struct _dep_node): New type.
1024 (dep_node_t): New typedef.
1025 (DEP_NODE_BACK, DEP_NODE_DEP, DEP_NODE_FORW): New access macros.
1027 (struct haifa_insn_data.back_deps): New field to hold backward
1028 dependencies of the insn.
1029 (struct haifa_insn_data.depend): Rename to forw_deps. Change its type
1031 (struct haifa_insn_data.resolved_deps): Rename to resolved_back_deps.
1032 Change its type to deps_list_t.
1033 (INSN_BACK_DEPS): New access macro to use instead of LOG_LINKS.
1034 (INSN_DEPEND): Rename to INSN_FORW_DEPS.
1035 (RESOLVED_DEPS): Rename to INSN_RESOLVED_BACK_DEPS.
1037 (INSN_COST): Move to haifa-sched.c. Use insn_cost () instead.
1039 (DEP_STATUS): Rename to DEP_LINK_STATUS. Fix typo in the comment.
1041 (add_forw_dep, delete_back_forw_dep, insn_cost): Update declaration and
1043 (dep_cost): Declare.
1045 * sched-deps.c (CHECK): New macro to (en/dis)able sanity checks.
1046 (ds_to_dk, dk_to_ds): New functions.
1048 (init_dep_1): New static function.
1049 (init_dep): New function.
1050 (copy_dep): New static function.
1052 (dep_link_consistent_p, attach_dep_link, add_to_deps_list): New static
1054 (detach_dep_link): New static function.
1055 (move_dep_link): New function.
1057 (dep_links_consistent_p, dump_dep_links): New static functions.
1058 (debug_dep_links): New debugging function.
1060 (deps_obstack, dl_obstack, dn_obstack): New static variables.
1062 (alloc_deps_list, init_deps_list): New static functions.
1063 (create_deps_list): New function.
1064 (clear_deps_list): New static function.
1065 (free_deps_list, delete_deps_list, deps_list_empty_p): New functions.
1066 (deps_list_consistent_p, dump_deps_list): New static functions.
1067 (debug_deps_list): New function.
1068 (add_back_dep_to_deps_list, find_link_by_pro_in_deps_list): New
1070 (find_link_by_con_in_deps_list, copy_deps_list_change_con): Ditto.
1072 (maybe_add_or_update_back_dep_1, add_or_update_back_dep_1): Update to
1073 use new scheduler dependencies lists.
1074 (add_back_dep, delete_all_dependences, fixup_sched_groups): Ditto.
1075 (sched_analyze): Ditto. Initialize dependencies lists.
1076 (add_forw_dep, compute_forward_dependences): Update to use new
1077 scheduler dependencies lists.
1079 (init_dependency_caches): Init deps_obstack.
1080 (free_dependency_caches): Free deps_obstack.
1082 (adjust_add_sorted_back_dep, adjust_back_add_forw_dep): Update to use
1083 new scheduler dependencies lists.
1084 (delete_forw_dep, add_or_update_back_forw_dep): Ditto.
1085 (add_back_forw_dep, delete_back_forw_dep): Ditto.
1087 * sched-rgn.c (set_spec_fed, find_conditional_protection, is_pfree):
1088 Update to use new scheduler dependencies lists.
1089 (is_conditionally_protected, is_prisky, add_branch_dependences): Ditto.
1090 (debug_dependencies): Ditto.
1091 (schedule_region): Update comments.
1093 * sched-ebb.c (earliest_block_with_similiar_load): Update to use new
1094 scheduler dependencies lists.
1095 (schedule_ebb): Update comments.
1097 * rtl.def (DEPS_LIST): Remove.
1099 * lists.c (unused_deps_list): Remove.
1100 (free_list): Update assertions.
1102 (alloc_DEPS_LIST, free_DEPS_LIST_list, free_DEPS_LIST_node): Remove.
1103 (remove_free_DEPS_LIST_elem, copy_DEPS_LIST_list): Ditto.
1105 * rtl.h (free_DEPS_LIST_list, alloc_DEPS_LIST): Remove declarations.
1106 (remove_free_DEPS_LIST_elem, copy_DEPS_LIST_list): Ditto.
1108 * haifa-sched.c (comments): Update.
1109 (insn_cost1): Remove. Inline the code into insn_cost ().
1110 (insn_cost): Update to use new scheduler dependencies lists. Move
1111 processing of the dependency cost to dep_cost ().
1112 (dep_cost): New function. Use it instead of insn_cost () when
1113 evaluating cost of the dependency. Use compatible interface to
1114 interact with the target.
1115 (priority): Update to use new scheduler dependencies lists.
1116 (rank_for_schedule): Ditto. Optimize heuristic that prefers the insn
1117 with greater number of insns that depend on the insn.
1118 (schedule_insn): Update to use new scheduler dependencies lists. Add
1119 code to free backward dependencies lists. Inline and optimize code
1120 from resolve_dep () - see PR28071.
1121 (ok_for_early_queue_removal): Update to use new scheduler dependencies
1122 lists. Update call to targetm.sched.is_costly_dependence hook.
1124 (fix_inter_tick, try_ready, fix_tick_ready): Update to use new
1125 scheduler dependencies lists.
1127 (resolve_dep): Remove. Move the logic to schedule_insn ().
1128 (init_h_i_d): Initialize dependencies lists.
1130 (process_insn_depend_be_in_spec): Rename to
1131 process_insn_forw_deps_be_in_spec. Update to use new scheduler
1133 (add_to_speculative_block, create_check_block_twin, fix_recovery_deps):
1134 Update to use new scheduler dependencies lists.
1135 (clear_priorities, calc_priorities, add_jump_dependencies): Ditto.
1137 * ddg.c (create_ddg_dependence, create_ddg_dep_no_link): Update to use
1138 new scheduler dependencies lists.
1139 (build_intra_loop_deps): Ditto.
1141 * target.h (struct _dep): Declare to use in
1142 gcc_target.sched.is_costly_dependence.
1143 (struct gcc_target.sched.adjust_cost): Fix typo.
1144 (struct gcc_target.sched.is_costly_dependence): Change signature to use
1145 single dep_t parameter instead of an equivalent triad.
1146 (struct gcc_target.sched.adjust_cost_2): Remove.
1148 * target-def.h (TARGET_SCHED_ADJUST_COST_2): Remove.
1150 * reg-notes.def (DEP_TRUE, DEP_OUTPUT, DEP_ANTI): Update comments.
1152 * doc/tm.texi (TARGET_SCHED_IS_COSTLY_DEPENDENCE): Update
1154 (TARGET_SCHED_ADJUST_COST_2): Remove documentation.
1156 * doc/rtl.texi (LOG_LINKS): Remove part about instruction scheduler.
1157 (REG_DEP_TRUE): Document.
1159 * config/ia64/ia64.c (ia64_adjust_cost_2): Rename to ia64_adjust_cost.
1160 Change signature to correspond to the targetm.sched.adjust_cost hook.
1161 Update use in TARGET_SCHED_ADJUST_COST_2.
1162 (TARGET_SCHED_ADJUST_COST_2): Rename to TARGET_SCHED_ADJUST_COST.
1163 (ia64_dependencies_evaluation_hook, ia64_dfa_new_cycle): Update to use
1164 new scheduler dependencies lists.
1165 (ia64_gen_check): Ditto.
1167 * config/mips/mips.c (vr4130_swap_insns_p): Update to use new scheduler
1170 * config/rs6000/rs6000.c (rs6000_is_costly_dependence): Change
1171 signature to correspond to the targetm.sched.is_costly_dependence hook.
1172 (is_costly_group): Update to use new scheduler dependencies lists.
1174 * config/spu/spu.c (spu_sched_adjust_cost): Use insn_cost () function
1175 instead of INSN_COST () macro.
1177 2007-02-01 Ian Lance Taylor <iant@google.com>
1179 * lower-subreg.c (resolve_clobber): Handle a subreg of a concatn.
1181 2007-02-01 Guy Martin <gmsoft@gentoo.org>
1183 * pa.md (tp_load): Correct mfctl instruction syntax.
1185 2007-02-01 Geoffrey Keating <geoffk@apple.com>
1187 * config/rs6000/rs6000.c (rs6000_stack_info): Correct
1188 altivec_padding_size calculation on AIX. Improve comment, add
1189 assert to verify that it's right.
1191 * config/rs6000/darwin.h (MD_UNWIND_SUPPORT): Don't define for
1194 2007-2-01 Seongbae Park <seongbae.park@gmail.com>
1197 * global.c (global_alloc): Add mising initialization of
1200 2007-02-01 Roger Sayle <roger@eyesopen.com>
1202 * alias.c (init_alias_analysis): Correct whitespace.
1203 * bb-reorder.c (fix_edges_for_rarely_executed_code,
1204 partition_hot_cold_basic_blocks): Likewise.
1205 * builtins.c (expand_builtin_printf, expand_builtin_fprintf,
1206 expand_builtin_sprintf, fold_builtin_carg, fold_builtin_sprintf,
1207 maybe_emit_sprintf_chk_warning, fold_builtin_sprintf_chk,
1208 fold_builtin_snprintf_chk, fold_builtin_printf,
1209 fold_builtin_fprintf, do_mpfr_ckconv, do_mpfr_arg1, do_mpfr_arg2,
1210 do_mpfr_arg3, do_mpfr_sincos): Likewise.
1211 * cfgcleanup.c (cleanup_cfg): Likewise.
1212 * cfgexpand.c (tree_expand_cfg): Likewise.
1213 * fold-const.c (fold_binary) <RDIV_EXPR>: Likewise.
1214 * function.c (get_next_funcdef_no): Likewise.
1215 * gengtype.c (main): Likewise.
1216 * genmodes.c (main): Likewise.
1217 * gcse.c (bypass_conditional_jumps, print_ldst_list): Likewise.
1218 * haifa-sched.c (schedule_block, extend_h_i_d): Likewise.
1219 * ifcvt.c (noce_emit_move_insn): Likewise.
1220 * modulo-sched.c (generate_prolog_epilog, sms_schedule_by_order):
1222 * stor-layout.c (get_best_mode): Likewise.
1223 * tree-ssa-loop-niter.c (get_val_for): Likewise.
1224 * tree-ssa-structalias.c (get_varinfo, get_varinfo_fc,
1225 scc_visit, do_ds_constraint, do_complex_constraint, label_visit,
1226 perform_var_substitution, solve_graph): Likewise.
1227 * tree-vrp.c (vrp_finalize): Likewise.
1229 2007-02-01 Ian Lance Taylor <iant@google.com>
1231 * lower-subreg.c (simplify_gen_subreg_concatn): If we ask for the
1232 high part of a paradoxical subreg, return a constant zero.
1234 2007-02-01 Zdenek Dvorak <dvorakz@suse.cz>
1236 * toplev.c (lang_dependent_init): Call init_set_costs.
1237 * loop-init.c (loop_optimizer_init): Do not call init_set_costs.
1239 2007-02-01 Richard Guenther <rguenther@suse.de>
1242 * fold-const.c (fold_negate_expr): Allow negating a
1243 constant if overflow does not change.
1245 2007-02-01 Ralf Wildenhues <Ralf.Wildenhues@gmx.de>
1247 * doc/c-tree.texi (Expression trees): Improve markup.
1248 * doc/tm.texi (Register Classes, Addressing Modes)
1249 (Floating Point): Fix spacing after abbreviations. Fix some
1252 2007-02-01 Ben Elliston <bje@au.ibm.com>
1254 * doc/invoke.texi: Replace "bugfix" with "bug fix" throughout.
1255 * doc/contrib.texi: Likewise.
1256 * doc/install.texi: Likewise.
1258 2007-01-31 Richard Henderson <rth@redhat.com>
1259 Ian Lance Taylor <iant@google.com>
1261 * lower-subreg.c: New file.
1262 * rtl.def (CONCATN): Define.
1263 * passes.c (init_optimization_passes): Add pass_lower_subreg and
1265 * emit-rtl.c (update_reg_offset): New static function, broken out
1266 of gen_rtx_REG_offset.
1267 (gen_rtx_REG_offset): Call update_reg_offset.
1268 (gen_reg_rtx_offset): New function.
1269 * regclass.c: Revert patch of 2006-03-05, restoring
1271 (clear_reg_info_regno): New function.
1272 * dwarf2out.c (concatn_loc_descriptor): New static function.
1273 (loc_descriptor): Handle CONCATN.
1274 * common.opt (fsplit_wide_types): New option.
1275 * opts.c (decode_options): Set flag_split_wide_types when
1277 * timevar.def (TV_LOWER_SUBREG): Define.
1278 * rtl.h (gen_reg_rtx_offset): Declare.
1279 (reg_scan_update): Declare.
1280 * regs.h (clear_reg_info_regno): Declare.
1281 * tree-pass.h (pass_lower_subreg): Declare.
1282 (pass_lower_subreg2): Declare.
1283 * doc/invoke.texi (Option Summary): List -fno-split-wide-types.
1284 (Optimize Options): Add -fsplit-wide-types to -O1 list. Document
1286 * doc/rtl.texi (Regs and Memory): Document concat and concatn.
1287 * Makefile.in (OBJS-common): Add lower-subreg.o.
1288 (lower-subreg.o): New target.
1290 2007-01-31 Kazu Hirata <kazu@codesourcery.com>
1292 * config/sh/sh.h (HAVE_SECONDARY_RELOADS): Remove.
1294 2007-01-31 Anatoly Sokolov <aesok@post.ru>
1297 * config/avr/avr.c (DWARF2_ADDR_SIZE): Define.
1299 2007-01-31 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1302 * builtins.c (fold_builtin_sqrt): Use MPFR for constant args.
1304 2007-01-31 Zdenek Dvorak <dvorakz@suse.cz>
1306 * cfgloop.h: Include vec-prim.h.
1307 (enum li_flags): Remove LI_ONLY_OLD.
1308 (loop_iterator): Changed.
1309 (fel_next, fel_init): Iterate over loop tree.
1310 (FOR_EACH_LOOP_BREAK): New macro.
1311 * loop-unswitch.c (unswitch_loops): Do not pass LI_ONLY_OLD to
1313 * tree-ssa-loop-unswitch.c (tree_ssa_unswitch_loops): Ditto.
1314 * modulo-sched.c (sms_schedule): Ditto.
1315 * tree-vectorizer.c (vectorize_loops): Ditto.
1316 * doc/loop.texi: Update information on loop numbering and behavior of
1317 FOR_EACH_LOOP wrto new loops.
1318 * tree-scalar-evolution.c (compute_overall_effect_of_inner_loop,
1319 add_to_evolution_1): Test nestedness of loops instead of comparing
1321 * tree-chrec.c (chrec_fold_plus_poly_poly,
1322 chrec_fold_multiply_poly_poly, chrec_evaluate,
1323 hide_evolution_in_other_loops_than_loop, chrec_component_in_loop_num,
1324 reset_evolution_in_loop): Ditto.
1325 * Makefile.in (CFGLOOP_H): Add vecprim.h dependency.
1327 2007-01-31 Dirk Mueller <dmueller@suse.de>
1329 * c-common.c (warn_about_parentheses): Separate warning about
1330 un-parenthized sequence of comparison operators from the one
1331 which is supposed to warn about x <= y <= z.
1333 2007-01-31 Uros Bizjak <ubizjak@gmail.com>
1335 * optabs.h (enum optab_index): Add new OTI_isinf.
1336 (isinf_optab): Define corresponding macro.
1337 * optabs.c (init_optabs): Initialize isinf_optab.
1338 * genopinit.c (optabs): Implement isinf_optab using isinf?f2
1340 * builtins.c (mathfn_built_in): Handle BUILT_IN_ISINF{,F,L}.
1341 (expand_builtin_interclass_mathfn): Expand BUILT_IN_ISINF{,F,L}
1343 (expand_builtin): Expand BUILT_IN_ISINF{,F,L} using
1344 expand_builtin_interclass_mathfn.
1345 * reg_stack.c (subst_stack_regs_pat): Handle UNSPEC_FXAM.
1346 * config/i386/i386.md (UNSPEC_FXAM): New constant.
1347 (fxam<mode>2_i387): New insn pattern.
1348 (isinf<mode>2) New expander to implement isinf, isinff and isinfl
1349 built-in functions as x87 inline asm.
1351 2007-01-31 Kazu Hirata <kazu@codesourcery.com>
1353 * gcc/config/arm/unwind-arm.h (_sleb128_t, _uleb128_t): New.
1355 2007-01-30 Eric Christopher <echristo@apple.com>
1357 * config.gcc: Add geode.
1359 2007-01-31 Kazu Hirata <kazu@codesourcery.com>
1361 * cgraphunit.c, config/arm/arm.c, config/m68k/m68k.c,
1362 ipa-inline.c, tree-profile.c, tree-ssa-live.c,
1363 tree-ssa-math-opts.c, tree-ssanames.c, tree-vect-analyze.c,
1364 value-prof.c: Fix comment typos.
1366 2007-01-30 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
1369 * doc/invoke.texi (Wpointer-arith): Document warning.
1371 2007-01-30 Janis Johnson
1373 * doc/extend.texi (Decimal Floating Types): Remove decfloat.h from
1374 the list of discrepancies from the draft TR.
1376 2007-01-30 Dirk Mueller <dmueller@suse.de>
1379 * doc/invoke.texi (-Wreturn-type): Update description to
1382 2007-01-30 Richard Sandiford <richard@codesourcery.com>
1384 * cfgrtl.c (try_redirect_by_replacing_jump): Check only_sets_cc0_p.
1386 2007-01-30 Uros Bizjak <ubizjak@gmail.com>
1388 * builtins.c (expand_builtin_int_interclass_roundingfn): New function
1389 to handle optabs that operate on floating point input argument and
1390 output to integer output.
1391 (expand_builtin_mathfn) [BUILT_IN_ILOGB]: Move from here ...
1392 (expand_builtin_interclass_mathfn) [BUILT_IN_ILOGB]: ... to here.
1393 (expand_builtin): Expand BUILT_IN_ILOGB{,F,L} using
1394 expand_builtin_interclass_mathfn ().
1395 * config/i386/i386.md (fxtractxf3_i387): Rename from *fxtractxf3_i387.
1397 (ilogbxf2, ilogb<mode>2): New expanders to implement ilogb, ilogbf and
1398 ilogbl built-in functions as x87 intrinsics.
1400 2007-01-30 Richard Guenther <rguenther@suse.de>
1403 * dwarf2out.c (reference_to_unused): Query varpool if the
1404 variable was output.
1406 2007-01-30 Richard Guenther <rguenther@suse.de>
1409 * passes.c (execute_one_pass): Reset in_gimple_form to not
1410 confuse non-unit-at-a-time mode.
1412 2007-01-29 Roger Sayle <roger@eyesopen.com>
1413 Richard Guenther <rguenther@suse.de>
1415 * fold-const.c (round_up): Make HIGH an unsigned HOST_WIDE_INT to
1416 avoid undefined behaviour on overflow. Use force_fit_type_double
1417 to construct the constant with the specified TREE_OVERFLOW.
1419 2007-01-29 Janis Johnson <janis187@us.ibm.com>
1421 * config/dfp-bit.c: Add parameterized support for fp exceptions.
1422 * config/dfp-bit.h: Ditto.
1424 2007-01-29 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
1426 * c-decl.c (pop_scope): Replace warnings with call to
1427 warn_for_unused_label.
1428 * c-common.h (warn_for_unused_label): Declare.
1429 * c-common.c (warn_for_unused_label): Define.
1431 2007-01-29 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
1433 * tree-optimize.c (update_inlined_to_pointers): Delete unused
1436 2007-01-29 Janis Johnson <janis187@us.ibm.com>
1438 * Makefile.in (USER_H): Remove decfloat.h.
1439 * ginclude/decfloat.h: Delete, moving contents to ...
1440 * ginclude/float.h: Add support for decimal floating point,
1441 guarded by __STDC_WANT_DEC_FP__.
1443 2007-01-29 Mike Stump <mrs@apple.com>
1445 * doc/gccint.texi (Top): Rename Loop Representation to Loop
1446 Analysis and Representation to resolve case insensitive conflict.
1447 * doc/loop.texi (Loop Analysis and Representation): Likewise.
1449 2007-01-28 Daniel Berlin <dberlin@dberlin.org>
1451 * tree.h (struct tree_memory_tag): Add aliases member.
1452 (MTAG_ALIASES): New macro.
1453 * tree-ssa-alias.c (alias_bitmap_obstack): New variable.
1454 (add_may_alias): Remove pointer-set. Update for may_aliases being
1456 (mark_aliases_call_clobbered): Update for may_aliases being a
1458 (compute_tag_properties): Ditto.
1459 (create_partition_for): Ditto.
1460 (compute_memory_partitions): Ditto.
1461 (dump_may_aliases_for): Ditto.
1462 (is_aliased_with): Ditto.
1463 (add_may_alias_for_new_tag): Ditto.
1464 (rewrite_alias_set_for): Rewrite for may_aliases being a bitmap.
1465 (compute_is_aliased): New function.
1466 (compute_may_aliases): Call compute_is_aliased).
1467 (init_alias_info): Initialize alias_bitmap_obstack.
1468 (union_alias_set_into): New function.
1469 (compute_flow_sensitive_aliasing): Use union_aliases_into.
1470 (have_common_aliases_p): Rewrite to take two bitmaps and use
1472 (compute_flow_insensitive_aliasing): Stop using pointer-sets.
1474 (finalize_ref_all_pointers): Update for add_may_alias changes.
1475 (new_type_alias): Ditto.
1476 * tree-flow-inline.h (may_aliases): Return a bitmap.
1477 * tree-dfa.c (dump_variable): Check for MTAG_P'ness.
1478 * tree-ssa.c (verify_flow_insensitive_alias_info): Update for
1479 may_aliases being a bitmap.
1480 * tree-flow.h (struct var_ann_d): Remove may_aliases member.
1481 may_aliases now returns a bitmap.
1482 * tree-ssa-structalias.c (merge_smts_into): Update for may_aliases
1484 * tree-ssa-operands.c (add_virtual_operand): Update for
1485 may_aliases being a bitmap.
1487 2007-01-29 Daniel Berlin <dberlin@dberlin.org>
1489 PR tree-optimization/30630
1490 * tree-ssa-structalias.c (do_complex_constraint): Mark correct
1491 variable as changed.
1493 2007-01-29 Simon Martin <simartin@users.sourceforge.net>
1496 * gimplify.c (gimplify_target_expr): Make sure that the TARGET_EXPR is
1497 expanded only once even if an error occurs.
1499 2007-01-29 Ben Elliston <bje@au.ibm.com>
1501 * gcov-io.h (__gcov_indirect_call_profiler): Declare.
1502 (__gcov_average_profiler): Likewise.
1503 (__gcov_ior_profiler): Likewise.
1504 (__gcov_merge_ior): Likewise.
1506 2007-01-28 Jan Hubicka <jh@suse.cz>
1508 * builtins.c (expand_builtin_memset): Fix typo in my last patch.
1509 * value-prof.c (stringop_block_profile): Likewise.
1511 2007-01-28 Jan Hubicka <jh@suse.cz>
1513 * expr.c (emit_block_move_via_movmem, emit_block_move_via_libcall): Add
1514 variant handling histograms; add wrapper.
1515 (clear_storage_via_libcall): Export.
1516 (emit_block_move_hints): Break out from ...; add histograms.
1517 (emit_block_move): ... this one.
1518 (clear_storage_hints): Break out from ...; add histograms.
1519 (clear_storage): ... this one.
1520 (set_storage_via_memset): Handle histogram.
1521 * expr.h (emit_block_move_via_libcall, emit_block_move_hints): Declare.
1522 (clear_storage_hints, clear_storage_via_libcall): Declare.
1523 (set_storage_via_setmem): Update prototype.
1524 * doc/md.texi (movmem, setmem): Document new arguments.
1526 * value-prof.c (dump_histogram_value, tree_find_values_to_profile): Add
1528 (stringop_block_profile): New global function.
1529 (tree_stringops_values_to_profile): Profile block size and alignment.
1530 * value-prof.h (enum hist_type): add HIST_TYPE_AVERAGE and
1532 (struct profile_hooks): Add gen_average_profiler and gen_ior_profiler.
1533 (stringop_block_profile): Declare.
1534 * builtins.c: Include value-prof.h.
1535 (expand_builtin_memcpy, expand_builtin_memset): Pass block profile.
1536 * gcov-ui.h (GCOV_COUNTER_NAMES): Add new counter.
1537 (GCOV_COUNTER_AVERAGE, GCOV_COUNTER_IOR): New constants.
1538 (GCOV_COUNTERS, GCOV_LAST_VALUE_COUNTER): Update.
1539 * profile.c (instrument_values): Add new counters.
1540 * cfgexpand.c (expand_gimple_basic_block): Propagate histograms to
1542 * tree-profile.c (tree_average_profiler_fn, tree_ior_profiler_fn): New.
1543 (tree_init_edge_profiler): Build new profilers.
1544 (tree_gen_average_profiler, tree_gen_ior_profiler): New.
1545 (pass_tree_profile): Add dump.
1546 (tree_profile_hooks): Update.
1547 * Makefile.in (LIBGCOV): Add new constants.
1548 * libgcov.c (__gcov_merge_ior, __gcov_average_profiler,
1549 __gcov_ior_profiler): New.
1550 * i386.md (movmem/setmem expanders): Add new optional arguments.
1552 2007-01-28 David Edelsohn <edelsohn@gnu.org>
1554 * doc/md.texi (Standard Pattern Names): Document blockage pattern.
1556 2007-01-28 Zdenek Dvorak <dvorakz@suse.cz>
1558 * tree-ssa-loop-unswitch.c: Include tree-inline.h.
1559 (tree_unswitch_single_loop): Pass eni_size_weights to
1560 tree_num_loop_insns.
1561 * tree-ssa-loop-manip.c: Include tree-inline.h.
1562 (can_unroll_loop_p): Pass eni_size_weights to
1563 tree_num_loop_insns.
1564 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p):
1565 Pass eni_size_weights to estimate_num_insns.
1566 * tree.h (init_inline_once): Export.
1567 * toplev.c (backend_init): Call init_inline_once.
1568 * cgraphunit.c (cgraph_process_new_functions,
1569 cgraph_analyze_function): Pass eni_inlining_weights to
1571 * ipa-inline.c (compute_inline_parameters): Ditto.
1572 * tree-ssa-loop-ivcanon.c (tree_num_loop_insns): Pass weights
1573 to estimate_num_insns.
1574 (try_unroll_loop_completely): Pass eni_size_weights to
1575 tree_num_loop_insns.
1576 * tree-eh.c (decide_copy_try_finally): Pass eni_size_weights
1577 ot estimate_num_insns.
1578 * tree-ssa-loop-prefetch.c: Include tree-inline.h.
1579 (loop_prefetch_arrays): Pass eni_time_weights to tree_num_loop_insns.
1580 * tree-inline.c (eni_inlining_weights, eni_size_weights,
1581 eni_time_weights): New variables.
1582 (init_inline_once): Initialize them.
1583 (struct eni_data): Mew.
1584 (estimate_num_insns_1, estimate_num_insns): Use weights.
1585 * tree-inline.h (struct eni_weights_d): New.
1586 (eni_inlining_weights, eni_size_weights, eni_time_weights): Declare.
1587 (estimate_num_insns): Declaration changed.
1588 * cfgloop.h (tree_num_loop_insns): Declaration changed.
1589 * Makefile.in (tree-ssa-loop-unswitch.o, tree-ssa-loop-prefetch.o,
1590 tree-ssa-loop-manip.o): Add TREE_INLINE_H dependency.
1592 2007-01-28 Zdenek Dvorak <dvorakz@suse.cz>
1594 * tree-data-ref.c (conflict_fn): Assert that the number of affine
1595 relations in the conflict function is valid.
1597 2007-01-27 Ian Lance Taylor <iant@google.com>
1599 * common.opt: Add fstrict-overflow.
1600 * opts.c (decode_options): Set flag_strict_overflow if -O2.
1601 * flags.h (TYPE_OVERFLOW_WRAPS): Define.
1602 (TYPE_OVERFLOW_UNDEFINED): Define.
1603 (TYPE_OVERFLOW_TRAPS): Define. This replaces TYPE_TRAP_SIGNED.
1605 * tree.h (TYPE_TRAP_SIGNED): Don't define.
1606 * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_UNDEFINED.
1607 (fold_negate_expr): Likewise.
1608 (make_range): Likewise.
1609 (extract_muldiv_1): Likewise.
1610 (maybe_canonicalize_comparison): Likewise.
1611 (fold_comparison): Likewise.
1612 (fold_binary): Likewise.
1613 (tree_expr_nonnegative_p): Likewise.
1614 (tree_expr_nonzero_p): Likewise.
1615 * tree-vrp.c (compare_values): Likewise.
1616 (extract_range_from_binary_expr): Likewise.
1617 (extract_range_from_unary_expr): Likewise.
1618 * tree-ssa-loop-niter.c (infer_loop_bounds_from_signedness):
1620 (nowrap_type_p): Likewise.
1621 * tree-scalar-evolution.c (simple_iv): Likewise.
1622 * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_WRAPS.
1623 (build_range_check): Likewise.
1624 (extract_muldiv_1): Likewise.
1625 (fold_comparison): Likewise.
1626 * tree-vrp.c (vrp_int_const_binop): Likewise.
1627 (extract_range_from_unary_expr): Likewise.
1628 * convert.c (convert_to_integer): Likewise.
1629 * fold-const.c (fold_negate_expr): Use TYPE_OVERFLOW_TRAPS.
1630 (fold_comparison): Likewise.
1631 (fold_binary): Likewise.
1632 * optabs.c (optab_for_tree_code): Likewise.
1633 * tree-vectorizer.c (vect_is_simple_reduction): Likewise.
1634 * simplify-rtx.c (simplify_const_relational_operation): Check
1635 flag_strict_overflow and flag_trapv.
1636 (simplify_const_relational_operation): Likewise.
1637 * doc/invoke.texi (Option Summary): Mention -fstrict-overflow.
1638 (Optimize Options): Add -fstrict-overflow to -O2 list. Document
1641 2007-01-27 Roger Sayle <roger@eyesopen.com>
1643 * tree.c (tree_fold_gcd): Delete.
1644 * tree.h (tree_fold_gcd): Remove prototype.
1645 * tree-data-ref.c (tree_fold_divides_p): Don't use tree_fold_gcd to
1646 test whether one constant integer is a multiple of another. Instead
1647 call int_const_binop with TRUNC_MOD_EXPR and test for a zero result.
1648 * fold-const.c (multiple_of_p): We've determined both TOP and
1649 BOTTOM are integer constants so we can call int_const_binop directly
1650 instead of the more generic const_binop.
1652 2007-01-27 Roger Sayle <roger@eyesopen.com>
1654 * fold-const.c (size_binop): In the fast-paths for X+0, 0+X, X-0 and
1655 1*X check that the constant hasn't overflowed, to preserve the
1657 (round_up): Provide an efficient implementation when rouding-up an
1658 INTEGER_CST to a power-of-two.
1660 2007-01-28 Ralf Wildenhues <Ralf.Wildenhues@gmx.de>
1662 * doc/sourcebuild.texi: Add comma for clarity.
1663 * doc/extend.texi: Fix some typos.
1664 * doc/passes.texi: Likewise.
1665 * doc/cppinternals.texi: Likewise.
1666 * doc/c-tree.texi: Likewise.
1667 * doc/tree-ssa.texi: Likewise.
1668 * doc/install.texi: Likewise.
1670 2007-01-27 Jan Hubicka <jh@suse.cz>
1672 * tree-sra.c (sra_walk_function): Don't rely on aliases being build.
1673 (pass_sra): Do not require alias information.
1674 * passes.c (init_optimization_passes): Add SRA
1676 2007-01-27 Steven Bosscher <steven@gcc.gnu.org>
1678 * tracer.c (rest_of_handle_tracer): We already cleaned
1679 up the CFG in tracer() so don't do it here again.
1680 * cfgcleanup.c (rest_of_handle_jump2): Don't repeat
1681 cleanup_cfg here, either. And don't call renumber_insns.
1683 * cfgrtl.c (rtl_verify_flow_info_1): Don't verify that BB_END
1684 and BB_HEAD are in the insn stream here. Instead make sure
1685 that BB_INSN is valid on all insns. Also, do check here that
1686 there are no pending branch predictions...
1687 (rtl_verify_flow_info): ...instead of doing it here. Checks
1688 for BB_END and BB_HEAD moved from rtl_verify_flow_info_1 to
1691 2007-01-26 Roger Sayle <roger@eyesopen.com>
1693 * config/i386/i386.c (ix86_swap_binary_operands_p): New helper
1694 function to simplify/factorize operand order canonicalization.
1695 (ix86_fixup_binary_operands): Reorganize using the above function.
1696 (ix86_binary_operator_ok): Likewise.
1698 2007-01-27 Jakub Jelinek <jakub@redhat.com>
1700 * genattrtab.c (struct attr_value_list, insn_code_values): Move to
1701 file scope from optimize_attrs.
1702 (simplify_test_exp): If insn_code_values is not NULL, use it to speed
1704 (optimize_attrs): Clear insn_code_values after freeing it.
1706 2007-01-26 Zdenek Dvorak <dvorakz@suse.cz>
1708 * tree-ssa-address.c (create_mem_ref): Remove ", bsi" from
1709 a parts.base assignment.
1711 2007-01-26 Zdenek Dvorak <dvorakz@suse.cz>
1713 * tree-data-ref.c (dump_subscript): Use dump_conflict_function.
1714 (compute_subscript_distance, initialize_data_dependence_relation,
1715 finalize_ddr_dependent, analyze_ziv_subscript,
1716 analyze_siv_subscript_cst_affine,
1717 compute_overlap_steps_for_affine_univar,
1718 compute_overlap_steps_for_affine_1_2, analyze_subscript_affine_affine,
1719 analyze_siv_subscript, analyze_miv_subscript,
1720 analyze_overlapping_iterations, subscript_dependence_tester_1,
1721 compute_self_dependence, free_dependence_relation): Work
1722 with affine_fn instead of chrecs.
1723 (dump_affine_function, dump_conflict_function, affine_function_equal_p,
1724 common_affine_function, affine_function_base,
1725 affine_function_constant_p, affine_fn_op, affine_fn_plus,
1726 affine_fn_minus, affine_fn_free, conflict_fn_not_known,
1727 conflict_fn_no_dependence, free_conflict_function, free_subscripts,
1728 conflict_fn, affine_fn_cst, affine_fn_univar): New functions.
1729 (all_chrecs_equal_p): Removed.
1730 * tree-data-ref.h (affine_fn, conflict_function): New types.
1731 (struct subscript): Change type of conflicting_iterations_in_a
1732 and conflicting_iterations_in_b.
1734 2007-01-26 Steve Ellcey <sje@cup.hp.com>
1737 * config/pa/pa.h (TARGET_HPUX_11): New.
1738 * config/pa/pa-hpux11.h (TARGET_HPUX_11): New.
1739 * config/pa/pa.c (pa_init_builtins): Use TARGET_HPUX_11.
1741 2007-01-26 Daniel Berlin <dberlin@dberlin.org>
1742 Richard Guenther <rguenther@suse.de>
1744 * tree-ssa-structalias.c (solve_graph): Handle case
1745 we merged the variable to another.
1747 2007-01-25 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1749 * builtins.c (fold_builtin_1): Treat ccos and ccosh as 'even'
1752 * fold-const.c (negate_mathfn_p): Treat casin, casinh, catan,
1753 catanh, cproj, csin, csinh, ctan and ctanh as 'odd' functions.
1755 2007-01-25 DJ Delorie <dj@redhat.com>
1757 * config/m32c/m32c.c (m32c_cannot_change_mode_class): We don't
1758 allow changes to modes which don't fit in those registers.
1760 * reload1.c (choose_reload_regs): Check for invalid subregs before
1761 computing their locations, not after.
1763 2007-01-25 Geoffrey Keating <geoffk@apple.com>
1766 * config/rs6000/rs6000.c (first_altivec_reg_to_save): On Darwin,
1767 save Altivec registers in an eh_return function.
1768 (compute_vrsave_mask): Likewise.
1769 (rs6000_stack_info): Correct AIX/Darwin stack alignment computation
1770 for saving Altivec registers.
1771 (rs6000_emit_prologue): Don't allocate stack twice in
1772 eh_return function. Correct expected value of altivec_save_offset
1773 when using save_world. Describe save of R0 to stack when using
1774 save_world. Describe stack pointer adjustment when using
1775 save_world. Remove duplicated eh_return parameter register saving.
1776 Update sp_offset variable after save_world.
1777 * config/rs6000/t-darwin (LIB2FUNCS_STATIC_EXTRA): Remove
1779 (LIB2FUNCS_EXTRA): Add darwin-world.asm.
1780 * config/rs6000/darwin.h (SUBTARGET_OVERRIDE_OPTIONS): -m64
1783 2007-01-25 Steve Ellcey <sje@cup.hp.com>
1785 * config.gcc (ia64*-*-hpux*): Make posix threads the default.
1787 2007-01-25 Steve Ellcey <sje@cup.hp.com>
1790 * config/pa/pa.c (pa_init_builtins): Set asm names for finite routines.
1791 * config/ia64/ia64.c (ia64_init_builtins): Ditto.
1793 2007-01-25 Richard Guenther <rguenther@suse.de>
1795 * doc/invoke.texi (-Wcoverage-mismatch): Document.
1796 * common.opt (-Wcoverage-mismatch): New warning option.
1797 * coverage.c (get_coverage_counts): Ignore coverage mismatch
1798 if -Wcoverage-mismatch is given.
1800 2007-01-25 Razya Ladelsky <razya@il.ibm.com>
1802 * ipa-cp.c (ipcp_insert_stage, ipcp_driver): Support for SSA.
1803 (ipcp_driver): Change to static definition.
1804 Add dumping of the ifunctions.
1805 (constant_val_insert): Remove unused parameter. Support for SSA.
1806 (ipcp_propagate_const): Support for SSA.
1807 (ipcp_profile_bb_print): Print only analyzed nodes.
1808 (ipcp_replace_map_create): Remove support for Fortran constant
1810 * ipa-prop.c (ipa_method_modify_stmt,
1811 ipa_callsite_compute_param): Support for SSA.
1812 * ipa-prop.h (ipcp_driver): Remove declaration.
1813 (IS_VALID_TREE_MAP_INDEX): Add define.
1815 2007-01-24 Geoffrey Keating <geoffk@apple.com>
1817 * unwind-dw2.c (execute_stack_op): Handle DW_OP_swap.
1819 2007-01-24 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1822 * builtins.c (fold_builtin_cabs): Use MPFR to evaluate a
1823 constant argument to cabs and do it without checking for
1824 -funsafe-math-optimizations.
1826 2007-01-24 Douglas Gregor <dgregor@osl.iu.edu>
1828 * c-common.h (RID_FIRST_CXX0X): New.
1829 (RID_LAST_CXX0X): New.
1830 * c-opts.c (c_common_handle_option): -Wc++0x-compat is triggered
1832 * c.opt (Wc++0x-compat): New.
1833 * doc/invoke.texi (-Wc++0x-compat): Document.
1835 2007-01-24 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1837 * builtins.c (fold_builtin_carg): New.
1838 (fold_builtin_1): Use it.
1840 2007-01-24 Jan Hubicka <jh@suse.cz>
1842 * ipa-inline.c (cgraph_decide_inlining): Initialize initial_insns.
1844 * ipa-inline.c (initial_insns, max_insns): Delete.
1845 (compute_max_insns): New function.
1846 (cgraph_decide_inlining_of_small_function): Use it; take minimal amount
1847 of insns as base for code growth.
1848 (cgraph_decide_inlining): Make initial_insns local; do not compute
1850 * params.def (PARAM_INLINE_UNIT_GROWTH): Set to 60.
1851 * doc/invoke.texi (inline-unit-growth): Update docs.
1853 2007-01-24 Jakub Jelinek <jakub@redhat.com>
1855 * config/i386/i386.h (x86_cmpxchg16b): Remove const.
1856 (TARGET_CMPXCHG16B): Define to x86_cmpxchg16b.
1857 * config/i386/i386.c (x86_cmpxchg16b): Remove const.
1858 (override_options): Add PTA_CX16 flag. Set x86_cmpxchg16b
1859 for CPUs that have PTA_CX16 set.
1862 * gimplify.c (omp_check_private): New function.
1863 (gimplify_scan_omp_clauses): Use it for
1864 firstprivate/lastprivate/reduction.
1867 * gimplify.c (omp_add_variable): Don't call omp_notice_variable
1868 on TYPE_SIZE_UNIT for GOVD_LOCAL VLAs.
1871 * omp-low.c (lower_omp_for_lastprivate): Add dlist argument.
1872 If lower_lastprivate_clauses emits some statements, append them
1873 to dlist rather than body_p and to body_p append an initializer.
1874 (lower_omp_for): Adjust caller.
1876 2007-01-24 Steve Ellcey <sje@cup.hp.com>
1878 * target.h (globalize_decl_name): New.
1879 * target-def.h (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
1880 * output.h (default_globalize_decl_name): New.
1881 * varasm.c (asm_output_bss): Use globalize_decl_name instead of
1883 (globalize_decl): Ditto.
1884 (default_globalize_decl_name): New.
1885 * config/ia64/ia64.c (ia64_globalize_decl_name): New.
1886 (ia64_handle_version_id_attribute): New.
1887 (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
1888 (ia64_asm_output_external): Use globalize_decl_name instead
1890 * doc/extend.texi (version_id): New pragma.
1891 * doc/tm.texi (ARGET_ASM_GLOBALIZE_DECL_NAME): New target hook.
1893 2007-01-24 Andreas Krebbel <krebbel1@de.ibm.com>
1895 * unwind-dw2-fde.c (get_cie_encoding): Replaced _Unwind_Word with
1896 _uleb128_t and _Unwind_SWord with _sleb128_t.
1897 * unwind-dw2.c (extract_cie_info, execute_stack_op, execute_cfa_program,
1898 uw_frame_state_for, uw_update_context_1): Likewise.
1899 * unwind-c.c (parse_lsda_header, PERSONALITY_FUNCTION): Likewise.
1900 * unwind-pe.h (read_uleb128, read_sleb128,
1901 read_encoded_value_with_base): Likewise.
1902 * unwind-generic.h: Define _sleb128_t and _uleb128_t types.
1904 2007-01-24 Richard Guenther <rguenther@suse.de>
1906 * builtins.c (expand_builtin_cexpi): Get the fndecl
1907 for cexp in the correct way.
1909 2007-01-24 Jan Hubicka <jh@suse.cz>
1911 * tree-ssa-dce.c (eliminate_unnecesary_stmts): Remove dead LHS of calls.
1913 2007-01-24 Andreas Krebbel <krebbel1@de.ibm.com>
1915 * c-cppbuiltin.c (builtin_define_type_sizeof): New function.
1916 (c_cpp_builtins): New builtin macros: __SIZEOF_INT__, __SIZEOF_LONG__,
1917 __SIZEOF_LONG_LONG__, __SIZEOF_SHORT__, __SIZEOF_POINTER__,
1918 __SIZEOF_FLOAT__, __SIZEOF_DOUBLE__, __SIZEOF_LONG_DOUBLE__,
1919 __SIZEOF_SIZE_T__, __SIZEOF_WCHAR_T__, __SIZEOF_WINT_T__ and
1920 __SIZEOF_PTRDIFF_T__.
1921 * doc/cpp.texi: Documentation for the new builtin macros added.
1923 2007-01-24 Uros Bizjak <ubizjak@gmail.com>
1925 * config/i386/i386.md (tanxf2, tan<mode>2, atan<mode>2, log<mode>2,
1926 log10<mode>2, log2<mode>2, expxf2, exp10xf2, exp2xf2): Use op2
1927 instead of operands[2] to avoid access past the end of array.
1929 2007-01-24 Richard Sandiford <richard@codesourcery.com>
1931 * reload1.c (emit_reload_insns): Pass the reload register
1932 for a non-spill output reload through forget_old_reloads_1.
1934 2007-01-23 Joseph Myers <joseph@codesourcery.com>
1936 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Define
1937 _SOFT_DOUBLE if doubles use software floating-point.
1938 * config/rs6000/libgcc-ppc-glibc.ver: Export additional long
1939 double functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
1940 * config/rs6000/darwin-ldouble.c: Also compile functions for
1941 hard-float without FPRs. Use fmsub function for all __NO_FPRS__
1942 cases. Compile extra functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
1943 * config/rs6000/linuxspe.h (SUBSUBTARGET_OVERRIDE_OPTIONS): Remove
1944 commented-out long double override.
1945 (CPP_LONGDOUBLE_DEFAULT_SPEC): Likewise.
1946 * config/rs6000/eabispe.h: Likewise.
1947 * config/rs6000/rs6000.c (rs6000_override_options): Don't override
1948 long double for non-SPE.
1949 (rs6000_handle_option): Likewise.
1950 (invalid_e500_subreg): Disallow more subregs involding DImode,
1951 DFmode, TImode or TFmode.
1952 (rs6000_legitimate_offset_address_p): Check TFmode offsets for
1954 (legitimate_lo_sum_address_p): Also check for TFmode for E500
1956 (rs6000_legitimize_address): Also handle TFmode for E500 double.
1957 (rs6000_legitimize_reload_address): Also handle TFmode for E500
1959 (rs6000_legitimate_address): Also check for TFmode for E500
1961 (rs6000_emit_move): Use DFmode subregs of TFmode for E500 double.
1962 (spe_build_register_parallel): Handle TFmode and TCmode.
1963 (rs6000_spe_function_arg): Handle TFmode and TCmode for E500
1965 (function_arg): Handle TFmode and TCmode for E500 double.
1966 (rs6000_init_libfuncs): Initialize extra libfuncs for soft double
1968 (print_operand): Handle TFmode and TImode for %y.
1969 (rs6000_generate_compare): Handle TFmode comparisons for E500
1971 (spe_func_has_64bit_regs_p): Check for TFmode for E500 double.
1972 (rs6000_function_value): Handle TFmode and TCmode for E500 double.
1973 (rs6000_libcall_value): Handle TFmode and TCmode for E500 double.
1974 * config/rs6000/rs6000.h (CANNOT_CHANGE_MODE_CLASS): Check for
1975 TFmode for E500 double.
1976 * config/rs6000/rs6000.md (FP): Allow TF for E500 double.
1977 (floatsidf2): Enable for E500 double.
1978 (movtf_softfloat): Use rs6000_nonimmediate_operand.
1979 (extenddftf2): Change to extenddftf2_fprs.
1980 (extenddftf2): Call gen_spe_extenddftf2 or gen_extenddftf2_fprs
1981 depending on TARGET_E500_DOUBLE.
1982 (extendsftf2): Enable for E500 double.
1983 (trunctfdf2): Enable for E500 double.
1984 (trunctfsf2): Change to trunctfsf2_fprs.
1985 (trunctfsf2): Call gen_spe_trunctfsf2 or gen_trunctfsf2_fprs
1986 depending on TARGET_E500_DOUBLE.
1987 (floatsitf2): Enable for E500 double.
1988 (fix_trunctfsi2): Change to fix_trunctfsi2_fprs.
1989 (fix_trunctfsi2): Call gen_spe_fix_trunctfsi2 or
1990 gen_fix_trunctfsi2_fprs depending on TARGET_E500_DOUBLE.
1991 (negtf2): Change to negtf2_internal.
1992 (negtf2): New expander.
1993 (abstf2): Enable for E500 double. Call gen_spe_abstf2_tst,
1994 gen_spe_abstf2_cmp or gen_abstf2_internal depending on
1995 TARGET_E500_DOUBLE and flag_unsafe_math_optimizations.
1996 (movdi_internal32): Use rs6000_nonimmediate_operand.
1997 (unnamed splitter): Likewise.
1998 * config/rs6000/spe.md (CMPTFEQ_GPR, TSTTFEQ_GPR, CMPTFGT_GPR,
1999 TSTTFGT_GPR, CMPTFLT_GPR, TSTTFLT_GPR): New unspecs.
2000 (SPE64TF, DITI): New mode macros.
2001 (frob_df_di): Change to frob_<SPE64:mode>_<DITI:mode>; allow more
2004 (frob_<mode>_di_2): New.
2005 (frob_tf_di_8_2): New.
2006 (frob_di_df): Change to frob_di_<mode>; allow more modes.
2008 (frob_di_df_2): Change to frob_<DITI:mode>_<SPE64:mode>_2; allow
2010 (frob_ti_<mode>_8_2): New.
2011 (frob_ti_tf_2): New.
2012 (mov_si<mode>_e500_subreg0, mov_si<mode>_e500_subreg0_2,
2013 mov_si<mode>_e500_subreg4, mov_si<mode>_e500_subreg4_2): Allow
2015 (mov_sitf_e500_subreg8, mov_sitf_e500_subreg8_2,
2016 mov_sitf_e500_subreg12, mov_sitf_e500_subreg12_2): New.
2017 (spe_trunctfdf2_internal1, spe_trunctfsf2, spe_extenddftf2,
2018 spe_fix_trunctfsi2, spe_fix_trunctfsi2_internal,
2019 spe_negtf2_internal, spe_abstf2_cmp, spe_abstf2_tst): New.
2020 (cmptfeq_gpr, tsttfeq_gpr, cmptfgt_gpr, tsttfgt_gpr, cmptflt_gpr,
2023 2007-01-23 Ian Lance Taylor <iant@google.com>
2025 * Makefile.in (OBJS-common): Reformat, alphabetize, but put
2027 (OBJS-archive): Reformat, alphabetize.
2028 (OBJS): Change out_object_file to OBJS-md.
2030 2007-01-23 Uros Bizjak <ubizjak@gmail.com>
2032 * config/i386/i386.md: Use REG_P, MEM_P, CONST_INT_P, LABEL_P,
2033 JUMP_P and CALL_P predicates where applicable.
2034 * config/i386/i386.c: Ditto.
2035 * config/i386/i386.md: Ditto.
2036 * config/i386/mmx.md: Ditto.
2037 * config/i386/predicates.md: Ditto.
2039 2007-01-22 Andreas Schwab <schwab@suse.de>
2041 * config/m68k/m68k.h: Fix comment.
2043 2007-01-22 Jan Hubicka <jh@suse.cz>
2045 * passes.c (init_optimization_passes): Do not rerun
2046 pass_early_warn_uninitialized.
2048 2007-01-22 Richard Guenther <rguenther@suse.de>
2050 PR tree-optimization/30038
2051 * tree-ssa-math-opts.c (maybe_record_sincos): New static helper
2053 (execute_cse_sincos_1): Likewise.
2054 (execute_cse_sincos): Likewise.
2055 (gate_cse_sincos): Likewise.
2056 (pass_cse_sincos): New pass CSEing sin() and cos() calls using
2057 the cexpi() canonicalization of sincos().
2058 * tree-pass.h (pass_cse_sincos): Declare.
2059 * passes.c (init_optimization_passes): New pass pas_cse_sincos.
2061 2007-01-21 Eric Botcazou <ebotcazou@libertysurf.fr>
2063 PR rtl-optimization/29329
2064 * combine.c (replaced_rhs_insn): Rename to i2mod.
2065 (replaced_rhs_value): Rename to i2mod_new_rhs.
2066 (i2mod_old_rhs): New global variable.
2067 (combine_instructions): Adjust for above change. Save a copy of
2068 the old RHS into i2mod_old_rhs when the contents of a REG_EQUAL
2069 note are substituted in the second instruction.
2070 (distribute_notes) <REG_DEAD>: Adjust for above change. Do not
2071 ditch the note if it pertains to the second eliminated register
2072 and this register is mentioned in i2mod_old_rhs.
2075 2006-09-12 Eric Botcazou <ebotcazou@libertysurf.fr>
2077 * combine.c (distribute_notes) <REG_DEAD>: Do not consider SETs past
2078 the insn to which the note was originally attached.
2080 2007-01-21 Jan Hubicka <jh@suse.cz>
2082 * ipa-inline.c (inlining_mode): Comment, move up.
2083 (cgraph_decide_inlining_incrementally): Do not perform inlining
2084 itself; fix handling of flattening of self recursive functions.
2085 (cgraph_find_cycles): Remove.
2086 (cgraph_flatten_node): Remove.
2087 (cgraph_decide_inlining): Use incremental inliner to handle flattening.
2088 (try_inline): New function.
2089 (cgraph_early_inlining): Update call of
2090 cgraph_decide_inlining_incrementally. Apply inlining here.
2091 (apply_inline): Update call of cgraph_decide_inlining_incrementally.
2093 2007-01-21 Dirk Mueller <dmueller@suse.de>
2096 * tree-vrp.c (check_array_bounds): do not warn
2097 about ADDR_EXPR's of ARRAY_REF's which are immediately
2098 used in binary expressions.
2100 2007-01-21 Ira Rosen <irar@il.ibm.com>
2102 * tree-vectorizer.h (struct _stmt_vec_info): Add new field
2103 read_write_dep and macros for its access.
2104 * tree-vectorizer.c (new_stmt_vec_info): Initialize the new field.
2105 * tree-vect-analyze.c (vect_analyze_data_ref_dependence): Remove
2106 argument, call vect_check_interleaving for every independent pair of
2107 data-refs. Mark loads that access the same memory location as a store
2109 (vect_check_dependences): Remove.
2110 (vect_analyze_data_ref_dependences): Remove vect_check_dependences
2111 call, fix the call to vect_analyze_data_ref_dependence.
2112 (vect_analyze_data_ref_access): For statements that access the same
2113 data-ref, check that they are not stores; for loads, check that there
2114 is no store that access the same location.
2116 2007-01-20 Roger Sayle <roger@eyesopen.com>
2117 Joseph Myers <joseph@codesourcery.com>
2119 * doc/invoke.texi (-fdump-rtl-combine): Fix under/overfull hbox.
2120 (-fdump-rtl-stack): Likewise.
2121 (-fno-signed-zeros): Use @minus{} for a minus sign. Correct typo.
2122 (-mcheck-zero-division, -mcpu): Fix under/overfull hbox.
2123 (-mpt-fixed): Use @minus{} for minus sign.
2124 (Using Precompiled Headers): Fix under/overfull hbox.
2126 2007-01-20 Jan Hubicka <jh@suse.cz>
2128 * tree-flow.h (struct stmt_ann_d): Move references_memory to proper
2129 place within annotation.
2131 2007-01-20 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
2133 * pa.c (output_move_double): Change array size of xoperands to 4.
2135 2007-01-20 Jan Hubicka <jh@suse.cz>
2137 * tree-tailcall.c (adjust_return_value): Do not use RESULT_DECL
2140 2007-01-19 Ian Lance Taylor <iant@google.com>
2142 * expmed.c (expand_divmod) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]: Cast
2143 constant to unsigned HOST_WIDE_INT before negating.
2145 2007-01-19 Ian Lance Taylor <iant@google.com>
2147 * tree-ssa-operands.h (struct vuse_vec_d): Change num_vuse field
2149 (VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Use unsigned comparison.
2150 (VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
2151 (SET_VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Likewise.
2152 (SET_VUSE_ELEMENT_VAR) [ENABLE_CHECKING]: Likewise.
2153 (SET_VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
2154 (realloc_vdef, realloc_vuse): Change second parameter to
2156 (ssa_operand_iterator_d): Change vuse_index and mayuse_index
2158 * tree-ssa-operands.c (realloc_vop): Change num_elem parameter to
2159 unsigned. Change x and lim locals to unsigned.
2160 (realloc_vdef, realloc_vuse): Change num_elem parameter to
2162 (finalize_ssa_vuse_ops): Change old_i local to unsigned.
2163 (copy_virtual_operands): Change i and n locals to unsigned.
2165 2007-01-19 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
2168 * builtins.c (fold_builtin_1): Handle builtin fdim.
2170 2007-01-20 Jan Hubicka <jh@suse.cz>
2172 * tree-ssa.c (init_tree_ssa): Do not call init_alias_heapvars.
2173 * tree-ssa-structalias.c (compute_points_to_sets): Do call
2174 init_alias_heapvars.
2175 (init_alias_heapvars): Initialize only when not already initialized.
2176 (delete_alias_heapvars): Set heapvar_for_stmt to NULL.
2178 2007-01-19 Roger Sayle <roger@eyesopen.com>
2180 * common.opt (fsigned-zeros): New command line option.
2181 * flags.h (HONOR_SIGNED_ZEROS): Control via flag_signed_zeros instead
2182 of flag_unsafe_math_optimizations.
2183 * opts.c (set_fast_math_flags): The -ffast-math command line option
2184 implies -fno-signed-zeros.
2185 (fast_math_flags_set_p): Likewise.
2187 * doc/invoke.texi: Document new -fno-signed-zeros option, and update
2188 the documentation of -ffast-math appropriately. Wrap long lines.
2190 2007-01-19 Steve Ellcey <sje@cup.hp.com>
2192 * system.h (ASM_MAKE_LABEL_LINKONCE): Poison.
2193 * varasm.c (globalize_decl): Remove ASM_MAKE_LABEL_LINKONCE ifdef.
2195 2007-01-19 Tomas Bily <tbily@suse.cz>
2197 * cgraphunit.c (cgraph_finalize_function): Updating of pid
2199 (tree_init_ic_make_global_vars): New function
2200 (tree_init_edge_profiler): call of tree_init_ic_make_global_vars
2201 (tree_gen_ic_profiler): New function
2202 (tree_gen_ic_func_profiler): New function
2203 (tree_profiling): Added calling of tree_gen_ic_func_profiler
2204 (tree_profile_hooks): Added hook for indirec/virtual calls
2205 * value-prof.c (tree_find_values_to_profile): New case for
2207 (tree_values_to_profile): Call for determining indirect/virtual
2209 (tree_indirect_call_to_profile): New function
2210 (tree_ic_transform): New function
2211 (tree_ic): New function
2212 (find_func_by_pid): New function
2213 (init_pid_map): New function
2214 (tree_value_profile_transformations): Added check for
2215 indirect/virtual call transformation
2216 * value-prof.h (enum hist_type): New counter type for
2217 indirect/virtual calls
2218 (profile_hooks): Added new hook for profiling indirect/virtual
2220 * profile.c (instrument_values): New case for indirect/virtual
2222 * gcov-io.h (GCOV_LAST_VALUE_COUNTER): Changed to 6
2223 (GCOV_COUNTER_V_INDIR): New counter type
2224 (GCOV_COUNTER_NAMES): New name of counter "indirect" added
2225 (GCOV_MERGE_FUNCTIONS): New merge function for indirect/virtual
2227 * cgraph.c: Definition of cgraph_max_pid
2228 (cgraph_create_node): Default init of pid attribute
2229 * cgraph.h: Declaration of cgraph_max_pid
2230 (struct cgraph_node): Added pid attribute
2231 * libgcov.c (__gcov_indirect_call_profiler): New function
2232 (__gcov_one_value_profiler_body): New function
2233 (__gcov_one_value_profiler): Body was moved to
2234 __gcov_one_value_profiler_body and calls it
2236 2007-01-19 Basile Starynkevitch <basile@starynkevitch.net>
2238 * doc/gty.texi (Options): Document the mark_hook option to GTY.
2239 * gengtype.c (write_types_data, write_func_for_structure,
2240 write_types, ggc_wtd, pch_wtd): Add skip_hooks to
2241 write_types_data, ggc_wtd, pch_wtd for processing mark_hook.
2242 (walk_type, write_func_for_structure): Generate the mark_hook if
2245 2007-01-19 Jan Hubicka <jh@suse.cz>
2247 * ipa-inline.c (cgraph_decide_inlining_incrementally): Instead of
2248 'early' argument take inlining mode argument specifying whether to
2249 inline for size/speeed or all functions; add support for flattening;
2251 (cgraph_early_inlining): Update call of decide_inlining_incrementally.
2253 2007-01-19 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
2256 * toplev.c (warn_deprecated_use): Use %qD instead of %qs to print
2257 the name of the declared identifier.
2259 2007-01-19 Dirk Mueller <dmueller@suse.de>
2261 * config/i386.h (CONDITIONAL_REGISTER_USAGE): Store
2262 result of PIC_OFFSET_TABLE_REGNUM in temporary variable to avoid
2263 duplicate evaluation.
2265 2007-01-19 Uros Bizjak <ubizjak@gmail.com>
2267 * config/i386/i386.md (acos<mode>2): Rename from acossf2 and acosdf2.
2268 Macroize expander using X87MODEF12 mode macro. Extend operand 1
2269 to XFMode, use acosxf2 and truncate result to requested mode.
2270 Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
2271 (asin<mode>2): Similarly, with asin expanders.
2272 (*fscalexf4_i387): Rename from *fscalexf4.
2273 (expNcorexf3): New expander.
2274 (expxf2, exp10xf2, exp2xf2): Use expNcorexf3 expander.
2275 (exp<mode>2): Rename from expsf2 and expdf2. Macroize expander using
2276 X87MODEF12 mode macro. Extend operand 1 to XFMode, use expxf2 and
2277 truncate result to requested mode. Use SSE_FLOAT_MODE_P to disable
2278 patterns for SSE math.
2279 (exp10<mode>2): Similarly, with exp10 expanders.
2280 (exp2<mode>2): Similarly, with exp2 expanders.
2281 (expm1<mode>2): Similarly, with expm1 expanders.
2282 (ldexp<mode>3): Similarly, with ldexp expanders.
2283 (log<mode>2, log10<mode>2, log2<mode>2, log1p<mode>2, logb<mode>2):
2284 Use gen_truncxf<mode>2_i387_noop to truncate result.
2286 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2288 * config/m68k/m68k.h (PREFERRED_STACK_BOUNDARY): Define to 32
2289 for ColdFire targets.
2291 2007-01-19 Nathan Sidwell <nathan@codesourcery.com>
2292 Richard Sandiford <richard@codesourcery.com>
2294 * config/m68k/m68k.h (M68K_STATIC_CHAIN_REG_NAME): New macro.
2295 (INITIALIZE_TRAMPOLINE): Use STATIC_CHAIN_REGNUM.
2296 (__transfer_from_trampoline): Use M68K_STATIC_CHAIN_REG_NAME.
2297 * config/m68k/m68kelf.h (STATIC_CHAIN_REGNUM): Override.
2298 (M68K_STATIC_CHAIN_REG_NAME): Likewise.
2299 * config/m68k/netbsd-elf.h (M68K_STATIC_CHAIN_REG_NAME): Likewise.
2301 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2303 * config/m68k/m68k.md (adddi_dilshr32): Rename to...
2304 (*adddi_dilshr32): ...this. Fix formatting. Remove commented-out
2305 non-canonical pattern. Restrict to !TARGET_COLDFIRE.
2306 (*adddi_dilshr32_cf): New pattern.
2307 (adddi3, subdi3): Remove first alternatives.
2309 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2311 * config/m68k/m68k.c (notice_update_cc): If an SFmode move is
2312 implemented using move.l, do not use its cc result for floating-point
2315 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2317 * config/m68k/m68k.h (EXTRA_CONSTRAINT): Stop the 'T' constraint
2318 from accepting 's' constraints if flag_pic.
2320 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2322 * config/m68k/m68k.md (bordered, bunordered, buneq, bunge, bungt)
2323 (bunle, bunlt, bltgt, bordered_rev, bunordered_rev, buneq_rev)
2324 (bunge_rev, bungt_rev, bunle_rev, bunlt_rev, bltgt_rev): Change
2325 condition from TARGET_68881 to TARGET_HARD_FLOAT.
2327 2007-01-19 Sandra Loosemore <sandra@codesourcery.com>
2329 * longlong.h (count_leading_zeros, COUNT_LEADING_ZEROS_0): Add
2330 ColdFire alternatives.
2331 * config/m68k/m68k.h (CLZ_DEFINED_VALUE_AT_ZERO): New macro.
2332 * config/m68k/m68k.md (clzsi2): Define for ColdFire
2333 architectures that support ff1 instruction.
2335 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2336 Julian Brown <julian@codesourcery.com>
2338 * config/m68k/m68k.h (CONST_OK_FOR_LETTER_P): Add an 'R' case.
2339 * config/m68k/m68k.md (*movsi_cfv4): Fold into...
2340 (*movsi_cf): ...here. Remove unnecessary 'R' from 'Rg'.
2343 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2345 * config/m68k/m68k-protos.h (valid_mov3q_const): Take a HOST_WIDE_INT
2347 (output_move_const_into_data_reg, output_move_simode_const): Delete.
2348 * config/m68k/m68k.c (const_method, const_int_cost): Take a
2349 HOST_WIDE_INT instead of an rtx.
2350 (m68k_rtx_costs): Update call accordingly.
2351 (output_move_const_into_data_reg): Likewise. Fix formatting.
2352 (valid_mov3q_const): Take a HOST_WIDE_INT instead of an rtx.
2354 (output_move_simode_const): Update calls after above changes.
2355 Rework to use automatic variables and predicates like MEM_P.
2356 * config/m68k/m68k.md (pushexthisi_const): Update call to
2359 2007-01-19 Dirk Mueller <dmueller@suse.de>
2361 * tree-ssa-alias.c (perform_var_substitution): Fix typo
2364 2007-01-19 Richard Guenther <rguenther@suse.de>
2366 * builtins.c (expand_builtin_cexpi): Fall back to expanding
2367 via cexp in case sincos is not available.
2369 2007-01-19 Richard Guenther <rguenther@suse.de>
2371 * doc/tm.texi (TARGET_HAS_SINCOS): Document new target macro.
2372 * defaults.h (TARGET_HAS_SINCOS): Default to off.
2373 * config/linux.h (TARGET_HAS_SINCOS): Set to on if we have glibc.
2374 * config/alpha/linux.h (TARGET_HAS_SINCOS): Likewise.
2375 * config/sparc/linux.h (TARGET_HAS_SINCOS): Likewise.
2376 * config/sparc/linux64.h (TARGET_HAS_SINCOS): Likewise.
2377 * config/rs6000/linux.h (TARGET_HAS_SINCOS): Likewise.
2378 * config/rs6000/linux64.h (TARGET_HAS_SINCOS): Likewise.
2380 2007-01-19 Uros Bizjak <ubizjak@gmail.com>
2382 * config/i386/i386.md (*fpatanxf3_i387, fpatan_extend<mode>xf3_i387):
2384 (atan2sf3_1, atan2df3_1, atan2xf3_1): Remove insn patterns.
2385 (atan2xf3): Directly generate RTL pattern.
2386 (atan2<mode>3): Rename from atan2sf3 and atan2df3 and macroize insn
2387 patterns using X87MODEF12 mode macro. Use fpatan_extend<mode>xf3_i387
2388 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
2389 disable patterns for SSE math.
2390 (atan<mode>2): Rename from atansf2 and atandf2 and macroize insn
2391 patterns using X87MODEF12 mode macro. Use fpatan_extend<mode>xf3_i387
2392 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
2393 disable patterns for SSE math.
2395 2007-01-19 Alexandre Oliva <aoliva@redhat.com>
2397 * libgcc-std.ver: Fix typo in %inherit for GCC_4.3.0.
2399 2007-01-18 Roger Sayle <roger@eyesopen.com>
2401 * fold-const.c (fold_unary) <VIEW_CONVERT_EXPR>: Optimize away a
2402 VIEW_CONVERT_EXPR to the same type as it's operand.
2404 2007-01-18 David Edelsohn <edelsohn@gnu.org>
2406 * config/rs6000/darwin-ldouble.c: Only build _SOFT_FLOAT if
2407 configured for long double 128.
2409 2007-01-18 Mike Stump <mrs@apple.com>
2411 * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Fix build
2414 2007-01-18 Michael Meissner <michael.meissner@amd.com>
2416 * i386.c (ix86_compute_frame_layout): Make fprintf's in #if 0 code
2419 2007-01-18 Jan Hubicka <jh@suse.cz>
2421 * tree-ssa-operands.c (vop_free_bucket_size): Never return value
2422 greater than NUM_VOP_FREE_BUCKETS.
2424 2007-01-18 Daniel Berlin <dberlin@dberlin.org>
2426 * tree-ssa-structalias.c: Update comments.
2427 (ptabitmap_obstack): Removed.
2429 (oldpta_obstack): Ditto.
2430 (stats): Add a few members.
2431 (struct variable_info): Remove node, complex, address_taken, and
2432 indirect_target members. Add oldsolution member.
2433 (new_var_info): Do not initialize removed members.
2434 (constraint_expr_type): Remove INCLUDES.
2435 (constraint_graph): Add size, implicit_preds, rep,
2436 indirect_cycles, eq_rep, label, direct_nodes, and complex members.
2437 (FIRST_REF_NODE): New macro.
2438 (LAST_REF_NODE): Ditto.
2439 (FIRST_ADDR_NODE): Ditto.
2440 (find): New function.
2442 (dump_constraint): Do not handle INCLUDES.
2443 (insert_into_complex): Do not insert duplicate constraints.
2444 (condense_varmap_nodes): Renamed and rewritten into ...
2445 (merge_node_constraints): This. Also fix bug in handling of
2446 offseted copy constraints.
2447 (clear_edges_for_node): No longer need to deal with preds at all,
2448 or removing associated preds/succs.
2449 (merge_graph_nodes): Deal with indirect_cycles.
2450 Don't deal with predecessors.
2451 (add_implicit_graph_edge): New function.
2452 (add_pred_graph_edge): Ditto.
2453 (add_graph_edge): Don't deal with predecessors.
2454 (build_constraint_graph): Removed.
2455 (build_pred_graph): New function.
2456 (build_succ_graph): Ditto.
2457 (struct scc_info): Removed in_component. Added roots, dfs, and
2458 node_mapping. Remove visited_index, unification_queue.
2459 (scc_visit): Deal with union-find we do now.
2460 Deal with cycles with REF nodes.
2461 (collapse_nodes): Renamed and rewritten to ...
2462 (unify_nodes): This.
2463 (process_unification_queue): Removed.
2464 (topo_visit): Cleanup
2465 (do_da_constraint): Use find.
2466 (do_sd_constraint): Ditto.
2467 (do_ds_constraint): Ditto.
2468 (do_complex_constraint): Ditto.
2469 (init_scc_info): Update for removed and added members.
2470 (find_and_collapse_graph_cycles): Renamed and rewritten into ...
2471 (find_indirect_cycles): This.
2472 (equivalence_class): New variable.
2473 (label_visit): New function.
2474 (perform_variable_substitution): Rewritten.
2475 (free_var_substitution_info): New function.
2476 (find_equivalent_node): Ditto.
2477 (move_complex_constraints): Ditto.
2478 (eliminate_indirect_cycles): Ditto.
2479 (solve_graph): Only propagate changed bits.
2480 Use indirect cycle elimination.
2482 (tree_id_t): Rename to tree_vi_t, delete id member, add vi member.
2483 (tree_id_eq): Renamed to ...
2484 (tree_vi_eq): This. Update for member change
2485 (insert_id_for_tree): Renamed and rewritten to ...
2486 (insert_vi_for_tree): This.
2487 (lookup_id_for_tree): Renamed and rewritten to ...
2488 (lookup_vi_for_tree): This.
2489 (get_id_for_tree): Renamed and rewritten to ...
2490 (get_vi_for_tree): Ditto.
2491 (get_constraint_exp_from_ssa_var): Update to use get_vi_for_tree.
2492 (process_constraint): Don't handle INCLUDES.
2493 Remove special ADDRESSOF case.
2494 (find_func_aliases): Rewrite to use vi functions instead of id
2496 (create_function_info_for): Ditto.
2497 (create_variable_info_for): Ditto.
2498 (intra_create_variable_infos): Ditto.
2499 (merge_smts_into): Ditto.
2500 (find_what_p_points_to): Ditto.
2501 (init_base_vars): Ditto.
2502 (init_alias_vars): Ditto.
2503 (remove_preds_and_fake_succs): New function.
2504 (dump_sa_points_to_info): Dump new stats.
2505 (dump_solution_for_var): Use find.
2506 (set_used_smts): Fix formatting.
2507 (compute_points_to_sets): Updated for new functions.
2508 (ipa_pta_execute): Ditto.
2510 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
2511 Richard Sandiford <richard@codesourcery.com>
2513 * doc/tm.texi (TARGET_FUNCTION_VALUE): Expand documentation of
2515 * calls.c (expand_call): If the return value is a PARALLEL,
2516 extract its first member.
2517 * config/m68k/linux.h (FUNCTION_EXTRA_EPILOGUE): Remove.
2518 * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
2519 use FUNCTION_EXTRA_EPILOGUE.
2520 (m68k_function_value): Return a PARALLEL if the return value
2521 is of a pointer type.
2522 * config/m68k/netbsd-elf.h (current_function_returns_pointer)
2523 (FUNCTION_EXTRA_EPILOGUE): Remove.
2524 * config/m68k/m68k.md (D0_REG): New constant.
2526 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
2528 * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
2529 output a NOP for empty epilogues.
2531 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2533 * config/m68k/m68k.c (m68k_use_return_insn): Update comments
2534 before function. Extend register save check to include all
2535 registers, not just integer ones.
2537 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
2539 * config/m68k/m68k-protos.h (use_return_insn): Rename to...
2540 (m68k_use_return_insn): ...this.
2541 * config/m68k/m68k.h (USE_RETURN_INSN): Delete.
2542 * config/m68k/m68k.c (use_return_insn): Rename to...
2543 (m68k_use_return_insn): ...this.
2544 * config/m68k/m68k.md (return): Use m68k_use_return_insn instead
2547 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
2549 * config/m68k/fpgnulib.c (__truncdfsf2): Implement round to
2550 nearest even, fix denormal rounding overflow.
2552 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2554 * config/m68k/m68k.md (movsf_cf_hard): Use fsmove instead of
2555 f%$smove and f%$move.
2556 (movdf_cf_hard): Use fdmove for cases 0 and 3 and fmove for case 1.
2557 (extendsfdf2_cf): Use fdmove instead of f%&move.
2558 (truncdfsf2_cf): Use fsmove instead of f%$smove.
2559 (add<mode>3_cf, sub<mode>3_cf): Use <FP:prec> instead of <FP:round>.
2561 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2563 * config/m68k/m68k.md (movdf_cf_hard): Use output_move_double for
2566 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2568 * real.h (real_format): Add a canonical_nan_lsbs_set field.
2569 (coldfire_single_format): Declare.
2570 (coldfire_double_format): Likewise.
2571 * real.c (encode_ieee_single): Use canonical_nan_lsbs_set instead
2572 of qnan_msb_set to determine the lower bits of a canonical
2574 (encode_ieee_double): Likewise.
2575 (encode_ieee_quad): Likewise.
2576 (ieee_single_format): Initialize canonical_nan_lsbs_set.
2577 (mips_single_format): Likewise.
2578 (ieee_double_format): Likewise.
2579 (mips_double_format): Likewise.
2580 (ieee_extended_motorola_format): Likewise.
2581 (ieee_extended_intel_96_format): Likewise.
2582 (ieee_extended_intel_128_format): Likewise.
2583 (ieee_extended_intel_96_round_53_format): Likewise.
2584 (ibm_extended_format): Likewise.
2585 (mips_extended_format): Likewise.
2586 (ieee_quad_format): Likewise.
2587 (mips_quad_format): Likewise.
2588 (vax_f_format): Likewise.
2589 (vax_d_format): Likewise.
2590 (vax_g_format): Likewise.
2591 (i370_single_format): Likewise.
2592 (i370_double_format): Likewise.
2593 (decimal_single_format): Likewise.
2594 (decimal_double_format): Likewise.
2595 (decimal_quad_format): Likewise.
2596 (c4x_single_format): Likewise.
2597 (c4x_extended_format): Likewise.
2598 (real_internal_format): Likewise.
2599 (coldfire_single_format): New real_format.
2600 (coldfire_double_format): Likewise.
2601 * config/pdp11/pdp11.c (pdp11_f_format): Initialize
2602 canonical_nan_lsbs_set.
2603 (pdp11_d_format): Likewise.
2604 * config/m68k/m68k.c (override_options): Override REAL_FORMAT_MODE
2605 if TARGET_COLDFIRE_CPU.
2607 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2609 * config/m68k/m68k-protos.h (m68k_output_pic_call): Delete.
2610 (output_call, m68k_legitimize_call_address): Declare.
2611 * config/m68k/m68k.h (EXTRA_CONSTRAINT): Remove unnecessary
2612 parenthesees. Add support for a 'W' constraint.
2613 (LEGITIMATE_PIC_OPERAND_P): Remove SYMBOL_REF_FLAG handling.
2614 (PRINT_OPERAND_PUNCT_VALID_P): Remove comment about 'o'.
2615 (m68k_symbolic_call, m68k_symbolic_jump): Declare.
2616 * config/m68k/m68k.c (m68k_symbolic_call, m68k_symbolic_jump): New
2618 (override_options): Initialize them. Do not set flag_no_function_cse
2619 for TARGET_ID_SHARED_LIBRARY.
2620 (m68k_output_pic_call): Delete.
2621 (m68k_legitimize_call_address): New function.
2622 (print_operand): Remove the %o prefix. Handle the %p prefix.
2623 (output_call): New function.
2624 (m68k_output_mi_thunk): Use m68k_symbolic_jump. Always load the
2625 target address from the GOT if symbolic jumps are not allowed.
2626 * config/m68k/m68k.md (call, general_operand): Do not set
2627 SYMBOL_REF_FLAG. Use m68k_legitimize_call_address instead.
2628 Merge separate flag_pic and !flag_pic define_insns into...
2629 (*call, *call_value): ...these new patterns. Match the address
2630 rather than the containing MEM and require it to be a call_operand.
2631 Use output_call to generate the asm template.
2632 * config/m68k/predicates.md (const_call_operand): New predicate.
2633 (call_operand): Likewise.
2635 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
2637 * config/m68k/m68k.h (REGISTER_MOVE_COST): Simplify definition.
2638 (STACK_GROWS_DOWNWARD): Define to 1.
2639 (FUNCTION_VALUE, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P): Equivocate
2640 comments, emphasizing that these values are only defaults.
2641 * config/m68k/linux.h (LINK_SPEC): Fix formatting in #undef.
2642 * config/m68k/m68k.c (const_method): Remove trailing whitespace.
2644 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2646 * config/m68k/m68k.md (cmpsi): Remove outdated flag_pic handling.
2648 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
2649 Richard Sandiford <richard@codesourcery.com>
2651 * config/m68k/m68k.h (DATA_REGNO_P, ADDRESS_REGNO_P, INT_REGNO_P)
2652 (FP_REGNO_P): New macros.
2653 (REGNO_OK_FOR_INDEX_P, REGNO_OK_FOR_BASE_P, REGNO_OK_FOR_DATA_P)
2654 (REGNO_OK_FOR_FP_P, REG_OK_FOR_INDEX_P, REG_OK_FOR_BASE_P): Use them.
2655 (EH_RETURN_STACKADJ_RTX): Use A0_REG.
2656 * config/m68k/m68k.c (m68k_regno_mode_ok): Use the new REGNO macros.
2658 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
2660 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
2661 (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-rtems*): Add t-floatlib
2663 * config/m68k/t-floatlib: New file, extracting common code from...
2664 * config/m68k/t-m68kbare, config/m68k/t-m68kelf,
2665 * config/m68k/t-uclinux: Here.
2666 * config/m68k/fpgnulib.c: Do not compile extendeed precision
2667 routines on ColdFire targets.
2669 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
2671 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
2672 (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*)
2673 (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-linux*)
2674 (m68k-*-rtems*): Use tm_file rather than m68k/m68k.h and
2675 explicitly set MOTOROLA to 1.
2676 * config/m68k/m68k.h (MOTOROLA): Simplify definition accordingly.
2678 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2679 Nathan Sidwell <nathan@codesourcery.com>
2681 * config/m68k/m68k.h (PCC_STATIC_STRUCT_RETURN): Delete.
2682 (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Add MOTOROLA cases.
2683 * config/m68k/coff.h (REGISTER_PREFIX_MD): Delete.
2684 * config/m68k/m68020-elf.h (LIB_SPEC): Delete.
2685 * config/m68k/m68k-none.h (CC1_SPEC, CPP_SUBTARGET_SPEC): Delete.
2686 * config/m68k/m68kelf.h (IMMEDIATE_PREFIX, REGISTER_PREFIX_MD)
2687 (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Delete.
2688 (NO_DOLLAR_IN_LABEL, PCC_STATIC_STRUCT_RETURN): Don't undefine.
2689 (BSS_ASM_OP, ASM_OUTPUT_SKIP, ASM_OUTPUT_ASCII): Delete.
2690 * config/m68k/m68kemb.h (PCC_STATIC_STRUCT_RETURN): Don't undefine.
2691 * config/m68k/linux.h (SIZE_TYPE, PTRDIFF_TYPE, WCHAR_TYPE)
2692 (WCHAR_TYPE_SIZE, TARGET_OBJFMT_CPP_BUILTINS): Delete.
2693 (TARGET_OS_CPP_BUILTINS): Don't define mc68000 and mc68020 here.
2694 (DBX_REGISTER_NUMBER): Delete.
2695 * config/m68k/netbsd-elf.h (IMMEDIATE_PREFIX): Delete.
2696 (PCC_STATIC_STRUCT_RETURN): Don't undefine.
2697 * config/m68k/openbsd.h (PCC_STATIC_STRUCT_RETURN): Define.
2699 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2701 * config.gcc (m68k-*-uclinux*): Add flat.h to $tm_file.
2702 * config/flat.h: New file.
2703 * crtstuff.c (USE_PT_GNU_EH_FRAME): Don't define if
2705 * config/m68k/m68k.h (ASM_PREFERRED_EH_DATA_FORMAT): Do not use
2706 indirect references for -msep-data or -mid-shared-library.
2707 Do not use PC-relative code addresses either.
2708 * config/m68k/m68k.c (override_options): Restrict -fPIC error
2710 * config/m68k/uclinux.h (STARTFILE_SPEC): Define. Use Scrt1.o
2711 for shared libraries and crt1.o for executables. Use crti.o and
2713 (ENDFILE_SPEC): Use crtend.o and crtn.o.
2714 (LIB_SPEC): Suppress -Rlibc.gdb if -static-libc is given.
2715 Do not add -elf2flt or -shared-lib-id options here.
2716 (LINK_SPEC): Define. Pass -elf2flt if no -elf2flt option is given.
2717 Pass -shared-lib-id if -mid-shared-library, taking the library
2718 identifier from -mshared-library-id if given, otherwise
2720 (EH_FRAME_IN_DATA_SECTION): Do not undefine.
2721 (INIT_SECTION_ASM_OP, FINI_SECTION_ASM_OP): Likewise.
2722 (TARGET_OS_CPP_BUILTINS): Define __GXX_MERGED_TYPEINFO_NAMES=0
2723 and __GXX_TYPEINFO_EQUALITY_INLINE=0 if -mid-shared-library.
2724 (DRIVER_SELF_SPECS): Map unadorned PIC options to -msep-data.
2725 * config/m68k/t-uclinux (EXTRA_MULTILIB_PARTS): Add crtbegin.o
2727 * config/m68k/lb1sf68.asm (PICCALL): Use an lea and pc-relative
2728 jump sequence for ISA A and ISA A+.
2729 (PICJUMP): Likewise.
2731 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2733 * config.gcc (m68k-*-uclinux*): Base the port on the common
2734 and m68k GNU/Linux files rather than on the generic ELF ones.
2735 * config/m68k/uclinux.h (TARGET_VERSION): Override.
2736 (TARGET_OS_CPP_BUILTINS): Use LINUX_TARGET_OS_CPP_BUILTINS.
2738 2007-01-18 Julian Brown <julian@codesourcery.com>
2739 Richard Sandiford <richard@codesourcery.com>
2741 * config/m68k/m68k.h (LONG_DOUBLE_TYPE_SIZE): Make 64-bit on ColdFire.
2742 (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
2743 * config/m68k/netbsd-elf.h (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Undefine
2745 * config/m68k/uclinux-oldabi.h (LONG_DOUBLE_TYPE_SIZE): Redefine to
2747 (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
2749 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2751 * doc/install.texi: Document m68k-uclinuxoldabi.
2752 * config.gcc (m68k-*-uclinuxoldabi): New configuration.
2753 * config/m68k/uclinux-oldabi.h: New file, copied from
2754 config/m68k/uclinux.h.
2756 2007-01-18 Kaz Kojima <kkojima@rr.iij4u.or.jp>
2758 * config/m32r/m32r-protos.h (m32r_expand_epilogue): Declare it.
2759 * config/m32r/m32r.c (m32r_setup_incoming_varargs): Use gen_frame_mem.
2760 (m32r_compute_frame_size): Use unsigned for regno.
2761 (m32r_reload_lr): Use gen_frame_mem.
2763 (m32r_output_function_epilogue): Don't output the function epilogue
2765 (m32r_expand_epilogue): New.
2766 (direct_return): Return false if the function has the interrupt
2768 (m32r_hard_regno_rename_ok): Remove code for the textual epilogue.
2769 * config/m32r/m32r.md (epilogue): New expander.
2770 (return_lr, return_rte): New insns.
2771 (return): Make it expander.
2772 (return_normal): New expander.
2774 2007-01-18 Josh Conner <jconner@apple.com>
2777 * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Add
2778 support for UNLE, UNLT, UNGE, and UNGT.
2780 2007-01-18 Jan Hubicka <jh@suse.cz>
2782 * tree-vrp.c (finalize_jump_threads): Do not call cleanup_cfg by hand.
2783 * tree-tailcall (add_virtual_phis): Likewise.
2784 (optimize_tail_call): Return TODOs.
2785 (execute_tail_calls): Return TODOs.
2786 * tree-ssa-ccp (execute_fold_all_builtins): Do cleanup_cfg via TODO.
2787 * tree-cfgcleanup.c (cleanup_tree_cfg_loop): Return if something
2789 * tree-ssa-forwprop.c (tree_ssa_forward_propagate_single_use_value):
2790 Cleanup cfg using TODO.
2791 * tree-flow.h (cleanup_tree_cfg_loop): Update prototype.
2792 * passes.c (execute_function_todo): When cleanup did something, remove
2794 * tree-cfg.c (pass_build_cfg): Add cleanup_cfg TODO.
2795 (make_edges): Don't cleanup_cfg.
2797 2007-01-18 Uros Bizjak <ubizjak@gmail.com>
2799 * reg-stack.c (subst_stack_regs_pat) [UNSPEC_SINCOS_COS,
2800 UNSPEC_XTRACT_FRACT]: Use generic code for instructions that
2801 operate on the top of stack.
2802 [UNSPEC_SINCOS_SIN, UNSPEC_XTRACT_EXP, UNSPEC_TAN]: Rewrite
2803 register handling of instructions that output to the second
2805 [UNSPEC_TAN_ONE, UNSPEC_TAN_TAN]: Remove.
2806 (move_for_stack_reg): Special-case check for dead destination
2807 stack slot for constant load of 1.0 inside UNSPEC_TAN.
2809 * config/i386/i386.md (UNSPEC_TAN): New constant.
2810 (UNSPEC_TAN_ONE, UNSPEC_TAN_TAN): Remove.
2811 (fptanxf4_i387, fptan_extend<mode>xf4_i387): New patterns
2812 to correctly model move of constant 1.0 to top stack slot.
2813 (*tandf3_1, *tansf3_1, *tanxf3_1): Remove insn patterns.
2814 (unnamed peephole2 pattern): Remove corresponding peephole2
2815 pattern that optimizes tan insn and loading of constant 1.0.
2816 (tanxf2): Use fptanxf4_i387.
2817 (tan<mode>2): Rename from tansf2 and tandf2 and macroize insn
2818 patterns using X87MODEF12 mode macro. Use fptan_extend<mode>xf4_i387
2819 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
2820 disable patterns for SSE math.
2821 (sincos<mode>3): Use truncxf<mode>2_i387_noop for truncation.
2822 (fyl2x_extend<mode>xf3_i387): Use X87MODEF12 for operand 1.
2824 2007-01-18 Dirk Mueller <dmueller@suse.de>
2825 Richard Guenther <rguenther@suse.de>
2828 * doc/invoke.texi (Warray-bounds): Document -Warray-bounds.
2829 * common.opt (Warray-bounds): Add new warning option.
2830 * c-opts.c (c_common_handle_option): Define -Warray-bounds
2832 * Makefile.in: make tree-vrp.o depend on toplev.h
2833 * tree-vrp.c (vrp_finalize): Call check_array_refs if -Warray-bounds
2835 (check_array_refs, check_array_bounds, check_array_ref): New.
2837 2007-01-18 Jan Hubicka <jh@suse.cz>
2839 * tree-ssa-ccp.c (ccp_finalize): Return if something changed.
2840 (execute_ssa_ccp): Return flags conditionally.
2841 * tree-ssa-propagate.c (substitue_and_fold): Return if something was
2843 * tree-ssa-propagate.h (substitute_and_fold): Update prototype.
2845 2007-01-18 Steven Bosscher <steven@gcc.gnu.org>
2847 * cfgcleanup.c (cleanup_cfg): Detect cfglayout mode and set
2848 the CLEANUP_CFGLAYOUT flag when in cfglayout mode.
2850 * Makefile.c (GTFILES): Add cfglayout.h.
2851 * gengtype.c (open_base_files): Likewise.
2852 * cfglayout.c (cfg_layout_function_footer,
2853 cfg_layout_function_header) Reindent to make gengtype happy.
2854 * cfglayout.h (cfg_layout_function_footer,
2855 cfg_layout_function_header): Add GTY(()) marker.
2857 * ifcvt.c (noce_try_sign_mask): Make sure INSN_B is non-null.
2859 2007-01-18 Ben Elliston <bje@au.ibm.com>
2861 * genautomata.c (write_automata): Include xstrerror output in the
2862 error message if writing the DFA description file fails.
2864 2007-01-17 H.J. Lu <hongjiu.lu@intel.com>
2866 * config/mips/mips-protos.h (mips_output_external): Make it
2868 * config/mips/iris.h (TARGET_ASM_EXTERNAL_LIBCALL): Removed.
2869 * config/mips/mips.c (irix_output_external_libcall): Likewise.
2870 (extern_list): Likewise.
2871 (extern_head): Likewise.
2872 (TARGET_ASM_FILE_END): Likewise.
2873 (mips_file_end): Likewise.
2874 (mips_output_external): Rewritten.
2876 2007-01-18 Ben Elliston <bje@au.ibm.com>
2878 * genpreds.c (write_insn_preds_c): Only write out the function
2879 body for regclass_for_constraint if we have register constraints.
2881 2007-01-17 Tom Tromey <tromey@redhat.com>
2883 * doc/sourcebuild.texi (libgcj Tests): Use sourceware.org.
2884 * doc/install.texi (Testing): Use sourceware.org.
2885 (Binaries): Likewise.
2886 (Specific): Likewise.
2887 * doc/contrib.texi (Contributors): Use sourceware.org.
2889 2007-01-17 Anatoly Sokolov <aesok@post.ru>
2891 * config/avr/avr.h (AVR_HAVE_LPMX): New macro.
2892 (AVR_ENHANCED): Rename to ...
2893 (AVR_HAVE_MUL): ... new.
2894 (avr_enhanced_p): Rename to ...
2895 (avr_have_mul_p): ... new.
2896 (TARGET_CPU_CPP_BUILTINS): Use 'avr_have_mul_p' instead of
2897 'avr_enhanced_p' for "__AVR_ENHANCED__". Define "__AVR_HAVE_MUL__".
2898 * config/avr/avr.c (avr_enhanced_p): Rename to ...
2899 (avr_have_mul_p): ... new.
2900 (base_arch_s): Rename 'enhanced' to 'have_mul'.
2901 (avr_override_options): Use 'avr_have_mul_p' and 'have_mul' instead of
2902 'avr_enhanced_p' and 'enhanced'.
2903 (ashlhi3_out, ashrhi3_out, lshrhi3_out, avr_rtx_costs): Use
2904 AVR_HAVE_MUL instead of AVR_ENHANCED.
2905 * avr.md (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
2906 (mulqi3, *mulqi3_enh, *mulqi3_call, mulqihi3, umulqihi3, mulhi3,
2907 *mulhi3_enh, *mulhi3_call, mulsi3, *mulsi3_call): Use AVR_HAVE_MUL
2908 instead of AVR_ENHANCED.
2909 (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
2910 * libgcc.S: Use __AVR_HAVE_MUL__ instead of __AVR_ENHANCED__.
2911 (__tablejump__): Use __AVR_HAVE_LPMX__ instead of __AVR_ENHANCED__.
2913 2007-01-17 Ian Lance Taylor <iant@google.com>
2915 * vec.h (VEC_reserve_exact): Define.
2916 (vec_gc_p_reserve_exact): Declare.
2917 (vec_gc_o_reserve_exact): Declare.
2918 (vec_heap_p_reserve_exact): Declare.
2919 (vec_heap_o_reserve_exact): Declare.
2920 (VEC_OP (T,A,reserve_exact)): New static inline function, three
2922 (VEC_OP (T,A,reserve)) [all versions]: Remove handling of
2924 (VEC_OP (T,A,alloc)) [all versions]: Call ...reserve_exact.
2925 (VEC_OP (T,A,copy)) [all versions]: Likewise.
2926 (VEC_OP (T,a,safe_grow)) [all versions]: Likewise.
2927 * vec.c (calculate_allocation): Add exact parameter. Change all
2929 (vec_gc_o_reserve_1): New static function, from vec_gc_o_reserve.
2930 (vec_gc_p_reserve, vec_gc_o_reserve): Call vec_gc_o_reserve_1.
2931 (vec_gc_p_reserve_exact, vec_gc_o_reserve_exact): New functions.
2932 (vec_heap_o_reserve_1): New static function, from vec_heap_o_reserve.
2933 (vec_heap_p_reserve, vec_heap_o_reserve): Call vec_heap_o_reserve_1.
2934 (vec_heap_p_reserve_exact): New function.
2935 (vec_heap_o_reserve_exact): New function.
2937 2007-01-17 Jan Hubicka <jh@suse.cz>
2939 * ipa-type-escape.c (look_for_casts): Revamp using handled_component_p.
2941 2007-01-17 Eric Christopher <echristo@apple.com>
2943 * config.gcc: Support core2 processor.
2945 2007-01-16 Jan Hubicka <jh@suse.cz>
2947 * tree-ssanames.c (release_dead_ssa_names): Instead of ggc_freeing
2948 the names, just unlink the chain so we don't crash on dangling pointers
2951 2007-01-16 Jan Hubicka <jh@suse.cz>
2953 * cgraph.h (cgraph_decide_inlining_incrementally): Kill.
2954 * tree-pass.h: Reorder to make IPA passes appear toegher.
2955 (pass_early_inline, pass_inline_parameters, pass_apply_inline): Declare.
2956 * cgraphunit.c (cgraph_finalize_function): Do not compute inling
2957 parameters, do not call early inliner.
2958 * ipa-inline.c: Update comments. Include tree-flow.h
2959 (cgraph_decide_inlining): Do not compute inlining parameters.
2960 (cgraph_decide_inlining_incrementally): Return TODOs; assume to
2961 be called with function context set up.
2962 (pass_ipa_inline): Remove unreachable functions before pass.
2963 (cgraph_early_inlining): Simplify assuming to be called from the
2965 (pass_early_inline): New pass.
2966 (cgraph_gate_ipa_early_inlining): New gate.
2967 (pass_ipa_early_inline): Turn into simple wrapper.
2968 (compute_inline_parameters): New function.
2969 (gate_inline_passes): New gate.
2970 (pass_inline_parameters): New pass.
2971 (apply_inline): Move here from tree-optimize.c
2972 (pass_apply_inline): New pass.
2973 * ipa.c (cgraph_remove_unreachable_nodes): Verify cgraph after
2975 * tree-inline.c (optimize_inline_calls): Return TODOs rather than
2977 (tree_function_versioning): Do not allocate dummy struct function.
2978 * tree-inline.h (optimize_inline_calls): Update prototype.
2979 * tree-optimize.c (execute_fixup_cfg): Export.
2980 (pass_fixup_cfg): Remove
2981 (tree_rest_of_compilation): Do not apply inlines.
2982 * tree-flow.h (execute_fixup_cfg): Declare.
2983 * Makefile.in (gt-passes.c): New.
2984 * passes.c: Include gt-passes.h
2985 (init_optimization_passes): New passes.
2986 (nnodes, order): New static vars.
2987 (do_per_function_toporder): New function.
2988 (execute_one_pass): Dump current pass here.
2989 (execute_ipa_pass_list): Don't dump current pass here.
2991 2007-01-16 Janis Johnson <janis187@us.ibm.com>
2993 * config/dfp-bit.c (dfp_compare_op): Return separate value for NaN.
2994 (DFP_NE, DFP_LE, DFP_GE): Return false for NaN.
2996 2007-01-16 David Edelsohn <edelsohn@gnu.org>
2998 * config/rs6000/darwin-ldouble.c: Build file for SOFT_FLOAT.
2999 (strong_alias): Define.
3000 (__gcc_qmul): Provide non-FMA for soft-float.
3006 (__gcc_qunord): New.
3016 * config/rs6000/rs6000.c (rs6000_init_libfuncs): Initialize
3017 soft-float functions.
3018 * config/rs6000/libgcc-ppc-glibc.ver: Version soft-float symbols.
3019 * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Do not warn
3020 about long double soft float.
3022 2007-01-16 Dorit Nuzman <dorit@il.ibm.com>
3023 Tehila Meyzels <tehila@il.ibm.com>
3025 * tree-vectorizer.h (is_pattern_stmt_p): New.
3026 * tree-vect-analyze.c (vect_determine_vectorization_factor): Fix
3027 formatting (tabs instead of spaces). Cleanup and clarify setting
3028 of STMT_VINFO_VECTYPE. Call is_pattern_stmt_p.
3029 * tree-vect-transform.c (vect_get_vec_def_for_operand): Fix typo.
3030 (vectorizable_type_demotion): Check that types are integral.
3031 (vectorizable_type_promotion): Likewise.
3032 (vectorizable_store): Fix typo. Eliminate new-line at end of
3035 2007-01-16 Jan Hubicka <jh@suse.cz>
3037 * tree-ssanames.c (release_dead_ssa_names): Remove invalidated
3040 2007-01-15 Eric Christopher <echristo@apple.com>
3042 * ifcvt.c: Include vec.h, vecprim.h.
3043 (check_cond_move_block): New argument regs.
3044 Reorganize. Add registers used to regs.
3045 (cond_move_process_if_block): Use regs set above as
3048 2007-01-15 Eric Christopher <echristo@apple.com>
3050 * config/darwin.h: Update copyright.
3051 (TARGET_OPTION_TRANSLATE_TABLE): Add umbrella.
3052 (LINK_COMMAND_SPEC): Add -u.
3053 (LINK_SPEC): Fix umbrella for above.
3055 2007-01-15 Joseph S. Myers <joseph@codesourcery.com>
3057 * config/soft-fp/op-common.h, config/soft-fp/op-4.h: Update from
3060 2007-01-15 Tom Tromey <tromey@redhat.com>
3062 * doc/sourcebuild.texi (libgcj Tests): Don't mention jacks.
3063 * doc/install.texi (Testing): Don't mention jacks.
3064 (Configuration): Document --enable-java-maintainer-mode. Move
3065 --with-java-home to libgcj-specific section. Document
3067 (Prerequisites): Mention --enable-java-maintainer-mode, ecj1.
3069 2007-01-15 Jan Hubicka <jh@suse.cz>
3071 * tree-ssa-dce.c (DCE_TODOs): New.
3072 (propagate_necessity): Return if something changed.
3073 (eliminate_unnecessary_stmts): Likewise.
3074 (perform_tree_ssa_dce): Return TODO flags when needed.
3075 (pass_dce, pass_dce_loop, pass_cd_dce): Remove TODO flags.
3077 2007-01-15 Uros Bizjak <ubizjak@gmail.com>
3079 * config/i386/i386.md (fyl2xxf3_i387): Rename from fyl2x_xf3.
3080 (fyl2x_extend<mode>xf3_i387): New insn pattern.
3081 (log<mode>2): Rename from logsf2 and logdf2 and macroize insn
3082 patterns using X87MODEF12 mode macro. Extend operand 1
3083 to XFmode. Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
3084 (log10<mode>2): Ditto.
3085 (log2<mode>2): Ditto.
3086 (log1p<mode>2): Ditto.
3087 (logb<mode>2): Ditto.
3088 (fyl2xp1xf3_i387): Rename from fyl2xp1_xf3.
3089 (fyl2xp1_extend<mode>xf3_i387): New insn pattern.
3090 (*fxtractxf3_i387): Rename from *fxtractxf3.
3091 (fxtract_extend<mode>xf3_i387): New insn pattern.
3092 (ilogbsi2): Use match_dup 3, not match_operand:XF 3.
3093 * config/i386/i386.c (ix86_emit_i387_log1p): Use gen_fyl2xp1xf3_i387()
3094 and gen_fyl2xxf3_i387().
3096 2007-01-14 Zdenek Dvorak <dvorakz@suse.cz>
3098 * loop-unswitch.c (unswitch_loop): Do not call fix_loop_placement.
3099 * cfgloopmanip.c (fix_loop_placement): Made static. Use
3100 get_loop_exit_edges. Changed return type to bool.
3101 * cfgloop.h (fix_loop_placement): Declaration removed.
3103 2007-01-14 Dorit Nuzman <dorit@il.ibm.com>
3105 * param.h (MIN_VECT_LOOP_BOUND): New.
3106 * params.def (MIN_VECT_LOOP_BOUND): New.
3107 * tree-vectorizer.c (slpeel_tree_peel_loop_to_edge): Takes another
3108 argument - minimum threshold for number of iterations.
3109 * tree-vectorizer.h (slpeel_tree_peel_loop_to_edge): Add another
3110 argument to declaration.
3111 * tree-vect-analyze.c (vect_analyze_operations): Check value of
3112 MIN_VECT_LOOP_BOUND.
3113 * tree-vect-transform.c (vect_do_peeling_for_loop_bound): Call
3114 slpeel_tree_peel_loop_to_edge with additional argument.
3115 (vect_do_peeling_for_alignment): Likewise.
3116 * doc/invoke.texi (min-vect-loop-bound): Document new param option.
3118 2007-01-14 Uros Bizjak <ubizjak@gmail.com>
3121 * config/i386/i386.c (print_operand) ['z']: Output 'b' for
3124 2007-01-14 Jan Hubicka <jh@suse.cz>
3126 * tree-dfa.c (remove_referenced_var): New function.
3127 * tree-ssa-live.c (remove_unused_locals): Walk referenced vars and
3128 prune referenced vars list too.
3129 * tree-flow.h (remove_referenced_var): Declare.
3131 2007-01-14 Jan Hubicka <jh@suse.cz>
3133 * tree-eh.c (add_stmt_to_eh_region_fn): Do not add call_exprs
3135 (remove_stmt_from_eh_region_fn): Do not remove call_exprs.
3136 (verify_eh_throw_stmt_node, verify_eh_throw_table_statements): Kill.
3137 * except.h (verify_eh_throw_table_statements): Kill prototype.
3138 * cfgexpand.c (expand_gimple_basic_block): Propagate Eh regions
3139 into call exrepssions.
3140 * tree-optimize.c (execute_free_cfg_annotatiosn): Do not call
3142 * tree-cfg.c: Include pointer-set.h.
3143 (verify_node_sharing): Work on pointer set.
3144 (verify_eh_throw_stmt_node): New.
3145 (verify_stmts): Use pointers sets, verify throw_stmt.
3147 2007-01-13 Zdenek Dvorak <dvorakz@suse.cz>
3149 * ipa-reference.c (analyze_function): Consider also addresses taken
3152 2007-01-12 Roger Sayle <roger@eyesopen.com>
3154 * c-typeck.c (null_pointer_constant_p): Replace use of
3155 TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
3156 (build_c_cast): Likewise.
3158 2007-01-12 Roger Sayle <roger@eyesopen.com>
3160 * tree.h (force_fit_type_double): Remove unused final argument.
3161 * c-common.c (constant_expression_warning): Replace use of
3162 TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
3163 (convert_and_check): Likewise.
3164 (shorten_compare): Update call to force_fit_type_double.
3165 (c_common_truthvalue_conversion) <INTEGER_CST>: Use integer_zerop.
3166 * convert.c (convert_to_pointer): Update call to
3167 force_fit_type_double.
3168 * fold-const.c (force_fit_type_double): Remove overflowed_const
3170 (int_const_binop, fold_convert_const_int_from_int,
3171 fold_convert_const_int_from_real, fold_div_compare,
3172 fold_sign_changed_comparison, fold_unary, fold_negate_const,
3173 fold_abs_const, fold_not_const): Remove the final argument from
3174 calls to force_fit_type_double.
3176 2007-01-12 Andrew Pinski <andrew_pinski@playstation.sony.com>
3178 * configure.ac: Set insn to "nop" for spu-*-* also.
3179 * configure: Regenerate.
3181 2007-01-12 Olga Golovanevsky <olga@il.ibm.com>
3183 * builtins.def : Add BUILT_IN_FREE.
3185 2007-01-12 Jan Hubicka <jh@suse.cz>
3187 PR tree-optimization/30443
3188 * tree-inline.c (tree_function_versioning): Do not optimize when
3189 cloning for inlining.
3191 2007-01-12 Zdenek Dvorak <dvorakz@suse.cz>
3193 * doc/loop.texi: Document recording of loop exits.
3194 * cfgloopmanip.c (loopify, duplicate_loop): Use alloc_loop.
3195 (update_single_exits_after_duplication,
3196 update_single_exit_for_duplicated_loop,
3197 update_single_exit_for_duplicated_loops): Removed.
3198 (duplicate_loop_to_header_edge): Do not call
3199 update_single_exits_after_duplication and
3200 update_single_exit_for_duplicated_loops.
3201 (loop_version): Do not update single_exit information.
3202 (fix_loop_structure): Use record_loop_exits instead of
3203 mark_single_exit_loops.
3204 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update
3205 the lists of loop exits.
3206 * cfghooks.c (redirect_edge_and_branch, redirect_edge_and_branch_force,
3207 split_edge, merge_blocks): Update the lists of loop exits.
3208 * modulo-sched.c (sms_schedule): Pass LOOPS_HAVE_RECORDED_EXITS to
3209 loop_optimizer_init.
3210 * loop-init.c (loop_optimizer_init): Call record_loop_exits instead
3211 of mark_single_exit_loops.
3212 (loop_optimizer_finalize): Call release_recorded_exits.
3213 * tree-ssa-loop.c (tree_loop_optimizer_init): Pass
3214 LOOPS_HAVE_RECORDED_EXITS to loop_optimizer_init.
3215 * tree-vectorizer.c (slpeel_tree_duplicate_loop_to_edge_cfg): Do not
3216 update single exit information.
3217 * lambda-code.c (perfect_nestify): Ditto.
3218 * cfgloop.c (flow_loop_free): Destroy the list of exits of the loop.
3219 (mark_single_exit_loops): Removed.
3220 (alloc_loop, loop_exit_hash, loop_exit_eq, loop_exit_free,
3221 get_exit_descriptions, rescan_loop_exit, record_loop_exits,
3222 dump_recorded_exit, dump_recorded_exits, release_recorded_exits): New
3224 (get_loop_exit_edges, single_exit): Use recorded exit lists.
3225 (add_bb_to_loop, remove_bb_from_loops): Update the lists of loop exits.
3226 (verify_loop_structure): Verify consistency of the exit lists.
3227 (flow_loops_find): Use alloc_loop. Initialize exits hash.
3228 (set_single_exit): Removed.
3229 * cfgloop.h (struct loop_exit): New function.
3230 (struct loop): single_exit_ field replaced by exits field.
3231 (LOOPS_HAVE_MARKED_SINGLE_EXITS): Replaced by LOOPS_HAVE_RECORDED_EXITS.
3232 (struct loops): Added exits hash.
3233 (mark_single_exit_loops, set_single_exit): Declaration removed.
3234 (release_recorded_exits, record_loop_exits, rescan_loop_exit): Declare.
3236 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3238 * doc/invoke.texi: Avoid use of @headitem.
3240 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3242 * cse.c (cse_insn): Move HAVE_CC0 code after declarations.
3244 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3246 * doc/install.texi: Fix m68k-*-* anchor and add m68k-*-* to the
3249 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
3250 Richard Sandiford <richard@codesourcery.com>
3252 * doc/invoke.texi: Document -mno-bitfield, -mno-rtd and -mno-short.
3253 * config/m68k/m68k.opt: Resort options.
3254 (mbitfield, mrtd, mshort): Remove RejectNegative properties.
3256 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
3257 Richard Sandiford <richard@codesourcery.com>
3259 * doc/invoke.texi: Document the macros that are defined by
3260 m68k's -mtune and -mhard-float options.
3261 * config/m68k/m68k-protos.h (m68k_cpp_cpu_ident) Declare.
3262 (m68k_cpp_cpu_family): Likewise.
3263 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Add a full set
3264 of __ucfv*__ macros. Define __mcffpu__ if generating code for
3265 ColdFire FPUs. Define __mcf_cpu_* and __mcf_family_* macros.
3266 * config/m68k/m68k.c (m68k_cpp_cpu_ident): New function.
3267 (m68k_cpp_cpu_family): Likewise.
3269 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3271 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Treat all mc68*
3272 macros besides mc68000 as tuning macros. Use a switch statement
3273 to set them and mcpu32.
3275 2007-01-12 Julian Brown <julian@codesourcery.com>
3277 * config/m68k/m68k.h: Use TARGET_68040 instead of TARGET_68040_ONLY.
3278 (TARGET_68040_ONLY): Rename to...
3279 (TARGET_68040): ...this.
3280 * config/m68k/m68k.c: Use TARGET_68040 instead of TARGET_68040_ONLY.
3281 * config/m68k/m68k.md: Likewise.
3283 2007-01-12 Julian Brown <julian@codesourcery.com>
3284 Nathan Sidwell <nathan@codesourcery.com>
3285 Richard Sandiford <richard@codesourcery.com>
3287 * config.gcc (m680[012]0-*-*, m68k*-*-*): Set m68k_cpu_ident to
3288 the -mcpu= argument associated with the --with-cpu setting.
3289 Define M68K_DEFAULT_TUNE to the default -mtune= option,
3290 if different from the one implied by the -mcpu setting.
3291 Accept --with-cpu=FOO if FOO is listed in m68k-devices.def,
3292 using mcpu=FOO as the default CPU option. Set target_cpu_default2.
3293 * doc/invoke.texi: Mention ColdFire in the introduction to the
3294 m68k options. Document the new -march, -mcpu, -mtune, -mdiv,
3295 -mno-div and -mhard-float options. Make -m68881 a synonym for
3296 -mhard-float. Document the previously-undocumented -m5206e,
3297 -m528x, -m5307 and -m5407 options. Tweak the existing option
3298 documentation for consistency.
3299 * doc/install.texi: Mention new --with-cpu arguments.
3300 * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Only use the
3301 default CPU if neither -mcpu nor -march are specified.
3302 (ASM_CPU_SPEC): Pass down -mcpu and -march options.
3303 (TARGET_CPU_CPP_BUILTINS): Set __mcfisa*__ macros from
3304 TARGET_ISA*. Set the legacy __mcf*__ cpu macros in the same way,
3305 using m68k_tune to decide between families that implement the
3306 same ISA. Use m68k_tune to set __mcfv4e__.
3307 (FL_BITFIELD, FL_68881, FL_COLDFIRE, FL_CF_HWDIV, FL_CF_MAC)
3308 (FL_CF_EMAC, FL_CF_EMAC_B, FL_CF_USP, FL_CF_FPU, FL_ISA_68000)
3309 (FL_ISA_68010, FL_ISA_68020, FL_ISA_68040, FL_ISA_A, FL_ISA_B)
3310 (FL_ISA_C, FL_ISA_MMU): New macros.
3311 (MASK_COLDFIRE): Delete.
3312 (TARGET_68010, TARGET_68020, TARGET_68040_ONLY, TARGET_COLDFIRE)
3313 (TARGET_ISAB): Redefine in terms of m68k_cpu_flags.
3314 (TARGET_68881, TARGET_COLDFIRE_FPU): Redefine in terms of m68k_fpu.
3315 (TARGET_HARD_FLOAT): Do not define here.
3316 (TARGET_ISAAPLUS, TARGET_ISAC): New macros.
3317 (TUNE_68000): New macro.
3318 (TUNE_68000_10): Redefine in terms of TUNE_68000 and TUNE_68010.
3319 (TUNE_68010, TUNE_68030, TUNE_68040, TUNE_68060, TUNE_CPU32)
3320 (TUNE_CFV2): Redefine in terms of m68k_tune.
3321 (uarch_type, target_device, fpu_type): New enums.
3322 (m68k_cpu, m68k_tune, m68k_fpu, m68k_cpu_flags): Declare.
3323 * config/m68k/m68k.c (TARGET_DEFAULT): Remove MASK_68881.
3324 (FL_FOR_isa_00, FL_FOR_isa_10, FL_FOR_isa_20, FL_FOR_isa_40)
3325 (FL_FOR_isa_cpu32, FL_FOR_isa_a, FL_FOR_isa_aplus, FL_FOR_isa_b)
3326 (FL_FOR_isa_c): New macros.
3327 (m68k_isa): New enum.
3328 (m68k_target_selection): New structure.
3329 (all_devices, all_isas, all_microarchs): New tables.
3330 (m68k_cpu_entry, m68k_arch_entry, m68k_tune_entry, m68k_cpu)
3331 (m68k_tune, m68k_fpu, m68k_cpu_flags): New variables.
3332 (MASK_ALL_CPU_BITS): Delete.
3333 (m68k_find_selection): New function.
3334 (m68k_handle_option): Handle -mcpu=, -march= and -mtune=.
3335 Map the legacy target options to a combination of the new ones.
3336 (override_options): Set m68k_cpu, m68k_tune, m68k_fpu and
3337 m68k_cpu_flags. Handle M68K_DEFAULT_TUNE. Use m68k_cpu_flags
3338 to derive default MASK_BITFIELD, MASK_CF_HWDIV and MASK_HARD_FLOAT
3340 * config/m68k/m68k.opt (m5200, m5206e, m528x, m5307, m5407, mcfv4e)
3341 (m68010, m68020, m68020-40, m68020-60, m68030, m68040): Remove Mask
3343 (m68881, msoft-float): Change mask from 68881 to HARD_FLOAT.
3344 (march=, mcpu=, mdiv, mhard-float, mtune=): New options.
3345 * config/m68k/m68k-devices.def: New file.
3347 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3348 Nathan Sidwell <nathan@codesourcery.com>
3350 * config/m68k/m68k.h (ASM_CPU_SPEC, ASM_SPEC, EXTRA_SPECS)
3351 (SUBTARGET_EXTRA_SPECS): New macros.
3352 * config/m68k/linux.h (ASM_SPEC): Remove CPU flags;
3353 use %(asm_cpu_spec) instead.
3354 * config/m68k/m68k-none.h (ASM_SPEC): Likewise.
3355 * config/m68k/openbsd.h (ASM_SPEC): Likewise.
3356 * config/m68k/netbsd-elf.h (ASM_SPEC): Likewise.
3357 (EXTRA_SPECS): Rename to...
3358 (SUBTARGET_EXTRA_SPECS): ...this.
3360 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
3361 Richard Sandiford <richard@codesourcery.com>
3362 Julian Brown <julian@codesourcery.com>
3364 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
3365 (m68k-*-uclinux*, m68k-*-linux*, m68k-*-rtems*): Set default_m68k_cpu
3366 to the configuration's default CPU.
3367 (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*): Likewise.
3368 Remove default masks.
3369 (m680[012]0-*-*): Set the default with_cpu to the first part of
3371 (m68k*-*-*): Set the default with_cpu to m$default_m68k_cpu.
3372 (m68k*-*-linux): Extend the --with-cpu handling to...
3373 (m680[012]0-*-*, m68k*-*-*): ...these configurations. Allow m68000
3374 and m68010. Don't set target_cpu_default2.
3375 * doc/install.texi: Document --with-cpu for m68k.
3376 * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Define.
3377 * config/m68k/m68k-none.h (TARGET_CPU_DEFAULT, M68K_CPU_m68k)
3378 (M68K_CPU_m68000, M68K_CPU_m68010, M68K_CPU_m68020, M68K_CPU_m68030)
3379 (M68K_CPU_m68040, M68K_CPU_m68302, M68K_CPU_m68332, TARGET_DEFAULT)
3380 (ASM_CPU_DEFAULT_SPEC, CC1_CPU_DEFAULT_SPEC): Delete.
3381 (ASM_SPEC): Remove use of %(asm_cpu_default).
3382 (EXTRA_SPECS, SUBTARGET_EXTRA_SPECS, MULTILIB_DEFAULTS): Delete.
3383 * config/m68k/linux.h (TARGET_DEFAULT): Delete.
3384 (CPP_SPEC): Merge definitions. Do not handle __HAVE_68881__ here.
3385 * config/m68k/netbsd-elf.h (TARGET_OS_CPP_BUILTINS): Define
3386 __HAVE_FPU__ if TARGET_HARD_FLOAT.
3387 (TARGET_DEFAULT): Delete.
3388 (EXTRA_SPECS): Delete cpp_cpu_default_spec, cpp_cpu_spec,
3389 cpp_fpu_spec, asm_default_spec and netbsd_cpp_spec.
3390 (CPP_CPU_SPEC): Delete.
3391 (TARGET_VERSION): Merge definitions, using TARGET_68010 to pick
3392 the appropriate string.
3393 (CPP_CPU_DEFAULT_SPEC, ASM_DEFAULT_SPEC, CPP_FPU_SPEC): Delete.
3394 (CPP_SPEC): Define to NETBSD_CPP_SPEC.
3395 (ASM_SPEC): Don't use %(asm_default_spec).
3396 * config/m68k/m68k.c (TARGET_DEFAULT_TARGET_FLAGS): Remove
3397 TARGET_DEFAULT and add MASK_68881.
3398 * config/m68k/m68k.md: Remove mention of TARGET_DEFAULT from comments.
3400 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3402 * config.gcc (m68010-*-netbsdelf*): Add MASK_68010.
3403 (m68k*-*-netbsdelf*, m68k*-*-openbsd*, m68k*-linux*): Add
3404 MASK_68010 alongside MASK_68020.
3405 * doc/invoke.texi: Document -m68010.
3406 * config/m68k/m68k.opt (m68010): New.
3407 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Define mc68010
3409 (TUNE_68010): New macro.
3410 * config/m68k/m68k-none.h (M68K_CPU_m68k, M68K_CPU_m68010)
3411 (M68K_CPU_m68020, M68K_CPU_m68030, M68K_CPU_m68040)
3412 (M68K_CPU_m68332): Add MASK_68010.
3413 * config/m68k/linux.h (TARGET_DEFAULT): Add MASK_68010 to
3414 fallback definition.
3415 * config/m68k/netbsd-elf.h (CPP_CPU_SPEC): Remove now-redundant
3417 * config/m68k/m68k.c (MASK_ALL_CPU_BITS): Add MASK_68010.
3418 (m68k_handle_option): Handle OPT_m68010. Add MASK_68010
3419 to all entries that use MASK_68020.
3420 (output_move_simode_const, output_move_himode, output_move_qimode)
3421 (output_move_stricthi, output_move_strictqi): Use TARGET_68010
3422 instead of TARGET_68020 to select clr behavior. Remove comment
3423 about there being no TARGET_68010.
3424 * config/m68k/m68k.md: Likewise throughout.
3426 2007-01-12 Julian Brown <julian@codesourcery.com>
3428 * config/m68k/m68k.h (TARGET_ISAB): New macro.
3429 * config/m68k/m68k.c: Use TARGET_ISAB rather than TARGET_CFV4.
3430 * config/m68k/m68k.md: Likewise.
3432 2007-01-12 Julian Brown <julian@codesourcery.com>
3434 * config/m68k/m68k.h (LEGITIMATE_INDEX_P, LEGITIMIZE_ADDRESS): Use
3435 TARGET_COLDFIRE_FPU instead of TARGET_CFV4E.
3437 2007-01-12 Julian Brown <julian@codesourcery.com>
3439 * config/m68k/m68k.h (TUNE_68040_60): New macro.
3440 * config/m68k/m68k.c (standard_68881_constant_p): Use it.
3441 * config/m68k/m68k.md: Likewise.
3443 2007-01-12 Julian Brown <julian@codesourcery.com>
3444 Richard Sandiford <richard@codesourcery.com>
3446 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Use TUNE_68030
3447 instead of TARGET_68030, TUNE_68040 instead of TARGET_68040,
3448 TUNE_68060 instead of TARGET_68060 and TUNE_CPU32 instead of
3450 (TARGET_CPU32): Rename to...
3451 (TUNE_CPU32): ...this.
3452 (TUNE_68000_10, TUNE_68030, TUNE_68040, TUNE_68060)
3453 (TUNE_CFV2): New macros.
3454 * config/m68k/netbsd-elf.h (LONG_DOUBLE_TYPE_SIZE): Simplify;
3455 remove conditions that are implied by TARGET_68020.
3456 * config/m68k/m68k.c (m68k_output_function_prologue): Use TUNE_68040
3457 instead of TARGET_68040 and TUNE_CPU32 instead of TARGET_CPU32.
3458 (m68k_output_function_epilogue): Likewise.
3459 (m68k_rtx_costs): Likewise. Use TUNE_68060 instead of TARGET_68060
3460 and TUNE_CFV2 instead of TARGET_5200. Use TUNE_68000_10 instead of
3461 "!TARGET_68020 && !TARGET_COLDFIRE" to choose between 68000 and
3462 non-68000 timings. Refactor multiplication and division costs.
3463 (output_addsi3): Use TUNE_68040 instead of TARGET_68040 and
3464 TUNE_CPU32 instead of TARGET_CPU32.
3465 (standard_68881_constant_p): Use TUNE_68040 instead of TARGET_68040
3466 and TUNE_68060 instead of TARGET_68060.
3467 * config/m68k/m68k.md: Use TUNE_68040 instead of TARGET_68040,
3468 TUNE_68060 instead of TARGET_68060, and TUNE_CPU32 instead of
3470 (movsi_const0): Use TUNE_68000_10 rather than "!TARGET_68020
3471 && !TARGET_COLDFIRE" to choose between moveq and clr.
3472 Likewise in the unnamed movsf pattern.
3473 (ashlsi_17_24, lshrsi_17_24): Guard with TUNE_68000_10 rather than
3474 "!TARGET_68020 && !TARGET_COLDFIRE". Likewise the unnamed
3477 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3479 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Increase amount
3480 of tabbing before backslashes.
3482 2007-01-11 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
3484 * pa-linux.h (ASM_OUTPUT_INTERNAL_LABEL): Undefine.
3485 * pa.h (ASM_OUTPUT_LABEL): Output colon when using GAS.
3486 (ASM_OUTPUT_INTERNAL_LABEL): Define.
3488 2007-01-11 Zdenek Dvorak <dvorakz@suse.cz>
3490 * tree-ssa-loop-ivopts.c (extract_cond_operands): Split from
3491 find_interesting_uses_cond.
3492 (find_interesting_uses_cond): Use extract_cond_operands.
3493 (rewrite_use_compare): Use extract_cond_operands and
3494 force_gimple_operand_bsi. Do not call update_stmt.
3495 (determine_use_iv_cost_condition): Use extract_cond_operands.
3496 Return cheaper of using original bound and changing the exit bound.
3498 2007-01-11 Zdenek Dvorak <dvorakz@suse.cz>
3500 PR tree-optimization/29516
3501 * tree-ssa-address.c (tree_mem_ref_addr, add_to_parts,
3502 most_expensive_mult_to_index, addr_to_parts,
3503 create_mem_ref, maybe_fold_tmr): Make the type of
3504 fields of TARGET_MEM_REF sizetype.
3505 (move_fixed_address_to_symbol, move_pointer_to_base):
3507 * tree.def (TARGET_MEM_REF): Add comment on types of
3510 2007-01-11 Joseph Myers <joseph@codesourcery.com>
3512 * c-common.c (vector_types_convertible_p): Treat opaque types as
3513 always convertible if they have the same size, but not otherwise.
3515 2007-01-11 Steven Bosscher <steven@gcc.gnu.org>
3517 * ifcvt.c (struct noce_if_info): Add comments to the fields.
3518 Remove the b_unconditional field.
3519 (noce_try_sign_mask): Do not look at b_unconditional.
3520 (noce_process_if_block): Do not use merge_if_blocks. Update
3521 the CFG here. Do not set b_unconditional.
3522 (cond_move_process_if_block): Likewise.
3523 (find_cond_trap): Likewise.
3524 (check_cond_move_block): Require simple jump insns at the end
3527 2007-01-11 Jan Hubicka <jh@suse.cz>
3529 PR tree-optimization/1046
3530 * tree-tailcall.c (suitable_for_tail_call_opt_p): Use TREE_ADDRESSABLE
3531 when alias info is not ready.
3532 (pass_tail_recursion): Do not require aliasing.
3533 * tree-ssa-copyrename.c (pass_rename_ssa_cop): Likewise.
3534 * tree-ssa-ccp.c (pass_ccp, pass_fold_builtins): Likewise.
3535 * tree-ssa-copy.c (pass_copy_prop): Likewise.
3536 * tree-ssa-forwprop.c (pass_forwprop): Likewise.
3537 * tree-ssa-dce.c (pass_dce, pass_dce_loop, pass_cd_dce): Likewise.
3538 * passes.c (init_optimization_passes): Execute rename_ssa_copies,
3539 ccp, forwprop, copy_prop, merge_phi, copy_prop, dce and tail recursion
3541 * tree-ssa-operands.c (add_virtual_operand, get_indirect_ref_operand):
3542 When aliasing is not build, mark statement as volatile.
3544 2007-01-11 Tom Tromey <tromey@redhat.com>
3546 PR preprocessor/15185, PR preprocessor/20989:
3547 * doc/cppopts.texi <-MT>: Update description of algorithm for
3548 computing default target.
3549 <-M, -MD>: Reword "basename" text.
3551 2007-01-11 Roger Sayle <roger@eyesopen.com>
3553 * builtins.c (expand_builtin_pow, expand_builtin_powi,
3554 fold_builtin_cabs, fold_builtin_sqrt, fold_builtin_trunc,
3555 fold_builtin_floor, fold_builtin_ceil, fold_builtin_round,
3556 fold_builtin_int_int_roundingfn, fold_builtin_bitop,
3557 fold_builtin_bswap, real_constp, fold_builtin_pow,
3558 fold_builtin_powi, fold_builtin_signbit, fold_builtin_copysign,
3559 do_mpfr_arg1, do_mpfr_arg2, do_mpfr_arg3, do_mpfr_sincos): Replace
3560 uses of the macro TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
3561 * convert.c (convert_to_pointer): Likewise.
3562 * expr.c (highest_pow2_factor, expand_expr_real_1): Likewise.
3563 * fold-const.c (force_fit_type, fold_negate_expr, int_const_binop,
3564 const_binop, fold_convert_const_int_from_int,
3565 fold_convert_const_int_from_real,
3566 fold_convert_const_real_from_real, sign_bit_p,
3567 optimize_minmax_comparison, extract_muldiv_1, fold_div_compare,
3568 fold_sign_changed_comparison, fold_unary, fold_comparison,
3569 fold_binary, multiple_of_p, tree_Expr_non_zero_p,
3570 fold_negate_const, fold_abs_const, fold_not_const): Likewise.
3571 * print-tree.c (print_node_brief, print_node): Likewise.
3572 * stor-layout.c (place_field, layout_type): Likewise.
3573 * tree-chrec.c (keep_cast): Likewise.
3574 * tree.c (build_vector, build_real, build_real_from_int_cst,
3575 build_complex): Likewise.
3577 2007-01-11 Roger Sayle <roger@eyesopen.com>
3579 * tree.h (TREE_CONSTANT_OVERFLOW): Obsolete. For the time being,
3580 treat TREE_CONSTANT_OVERFLOW as a synonym of TREE_OVERFLOW.
3582 2007-01-11 Paolo Bonzini <bonzini@gnu.org>
3584 * configure.ac (strict1_warn): Rename to strict_warn.
3585 (WERROR, --enable-werror, symlink hacks, stage1_cflags,
3586 cc_set_by_configure, quoted_cc_set_by_configure,
3587 stage_prefix_set_by_configure, quoted_stage_prefix_set_by_configure,
3588 all_boot_languages, all_stagestuff): Remove.
3589 (target_list): Remove bootstrap targets.
3590 * Makefile.in (quickstrap): Unconditionally make a synonym of all.
3591 (BOOT_LANGUAGES, STAGE1_CFLAGS, STAGE1_CHECKING,
3592 REMAKEFLAGS, FLAGS_TO_PASS, PREPEND_DOTDOT_TO_RELATIVE_PATHS,
3593 SUBDIR_FLAGS_TO_PASS, WERROR_FLAGS, STRICT2_WARN, LANG_STAGESTUFF,
3594 VOL_FILES, POSTSTAGE1_FLAGS_TO_PASS, STAGE2_FLAGS_TO_PASS,
3595 STAGEPROFILE_FLAGS_TO_PASS, STAGEFEEDBACK_FLAGS_TO_PASS, stage1_build,
3596 stage1_copy, stage2_build, stage2_copy, stageprofile_build,
3597 stageprofile_copy, stage3_build, stage3_copy, stagefeedback_build,
3598 stagefeedback_copy, stage4_build, clean_s1, clean_sw, bootstrap,
3599 bootstrap-lean, bootstrap2, bootstrap2-lean, bootstrap3,
3600 bootstrap3-lean, bootstrap4, bootstrap4-lean, unstage1, unstage2,
3601 unstage3, unstage4, unstageprofile, unstagefeedback, restage, restage2,
3602 restage3, restage4, restageprofile, restagefeedback, bubbleestrap,
3603 cleanstrap, unstrap, restrap, *compare, *compare3, *compare4,
3604 *compare-lean, *compare3-lean, *compare4-lean, stage1-start, stage1,
3605 stage2-start, stage2, stage3-start, stage3, stage4-start, stage4,
3606 stageprofile-start, stageprofile, stagefeedback-start, stagefeedback,
3607 risky-stage1, risky-stage2, risky-stage3, risky-stage4): Remove.
3608 (ORDINARY_FLAGS_TO_PASS): Rename to FLAGS_TO_PASS.
3609 (STAGECOPYSTUFF, STAGEMOVESTUFF): Consolidate into MOSTLYCLEANFILES.
3610 (mostlyclean): Adjust.
3611 (clean, distclean): Don't mention bootstrap stuff.
3612 * configure: Regenerate.
3613 * ada/config-lang.in, cp/config-lang.in, forttran/config-lang.in,
3614 java/config-lang.in, objc/config-lang.in, objcp/config-lang.in,
3615 treelang/config-lang.in (stagestuff): Remove.
3616 * doc/sourcebuild.texi (stage1, stage2, stage3, stage4,
3617 stageprofile, stagefeedback, stagestuff): Remove mention.
3619 2007-01-11 Nick Clifton <nickc@redhat.com>
3621 * config/mcore/predicates.md (mcore_general_movesrc_operand):
3623 (mcore_general_movdst_operand): Do not accept CONST_INTs.
3624 (mcore_arith_K_S_operand): Run the test for the S constraint not
3625 the test for the M constraint.
3626 (mcore_addsub_operand): Do not accept integer values that are
3627 larger than 32 bits.
3628 * config/mcore/mcore.md: Remove unused constraints from split.
3629 (andsi3): Use HOST_WIDE_INT instead of int to hold an INTVAL.
3631 (allocate_stack): Likewise.
3632 * config/mcore/mcore.c (mcore_print_operand): Restrict output of P
3633 operands to 32 bits.
3634 (mcore_const_costs): Use HOST_WIDE_INT instead of int to hold an
3636 (mcore_and_cost, mcore_modify_comparison, const_ok_for_mcore,
3637 mcore_const_ok_for_inline, mcore_const_trick_uses_not,
3638 try_constant_tricks, mcore_num_ones, mcore_num_zeros,
3639 mcore_output_bclri, mcore_output_andn, output_inline_const,
3640 mcore_output_move, mcore_output_movedouble): Likewise.
3641 (mcore_output_cmov): Use CONST_OK_FOR_M and CONST_OK_FOR_N.
3642 (output_inline_const): Likewise.
3643 (output_inline_const): Fix format strings used in sprintf
3645 * config/mcore/mcore-protos.h: Update prototypes for changed
3646 functions in mcore.c.
3647 * config/mcore/mcore.h (CONST_OK_FOR_I): Cast values to
3648 HOST_WIDE_INT and not int.
3649 (CONST_OK_FOR_J, CONST_OK_FOR_K, CONST_OK_FOR_L, CONST_OK_FOR_M,
3650 CONST_OK_FOR_N): Likewise.
3651 (LEGITIMATE_CONSTANT_P): Also check CONSTANT_P.
3652 (GO_IF_LEGITIMATE_INDEX): Use HOST_WIDE_INT instead of int to hold
3655 2007-01-10 Jan Hubicka <jh@suse.cz>
3657 * tree-vrp.c (remove_range_assertions): Release defs.
3658 * tree-ssa-loop-ivopts.c (rmeove_statement): Likewise.
3659 * tree-ssa-dom.c (remove_stmt_or_phi): Likewise.
3661 2007-01-10 Paul Brook <paul@codesourcery.com>
3663 * config/arm/arm.c (arm_rtx_costs_1): Handle mutiply-subtract.
3664 * config/arm/arm.md (mulsi3subsi): New insn.
3666 2007-01-10 Zdenek Dvorak <dvorakz@suse.cz>
3668 * tree-ssa-loop-manip.c (tree_unroll_loop): Make it a wrapper over ...
3669 (tree_transform_and_unroll_loop): New.
3670 * tree-flow.h (transform_callback, tree_transform_and_unroll_loop):
3673 2007-01-10 Robert Kennedy <jimbob@google.com>
3675 * fold-const.c (fold_comparison): Fold comparisons like (x *
3676 1000 < 0) to (x < 0).
3678 2007-01-10 Ian Lance Taylor <iant@google.com>
3680 * tree-pretty-print.c (dump_generic_node): Print parentheses when
3681 operands have the same priority.
3683 2007-01-10 Tom Tromey <tromey@redhat.com>
3685 * fold-const.c (fold_truthop): Don't check can_use_bit_fields_p.
3686 (fold_binary): Likewise.
3687 * langhooks.c (lhd_can_use_bit_fields_p): Removed.
3688 * langhooks-def.h (lhd_can_use_bit_fields_p): Removed.
3689 (LANG_HOOKS_CAN_USE_BIT_FIELDS_P): Removed.
3690 (LANG_HOOKS_INITIALIZER): Remove LANG_HOOKS_CAN_USE_BIT_FIELDS_P.
3691 * langhooks.h (struct lang_hooks): Removed field
3692 'can_use_bit_fields_p'.
3694 2007-01-10 Ralf Corsépius <ralf.corsepius@rtems.org>
3696 * config/bfin/t-bfin, config/bfin/t-bfin-elf: Remove GCC_CFLAGS.
3698 2007-01-10 Razya Ladelsky <razya@il.ibm.com>
3700 * function.c (get_last_funcdef_no): New function.
3701 * function.h (get_last_funcdef_no): Declare.
3702 * tree-inline.c (initialize_cfun): Add initialization.
3703 (tree_function_versioning): Cleanup.
3705 2007-01-10 Jan Hubicka <jh@suse.cz>
3707 * tree-inline.c (setup_one_parameter): Do not propagate into abnormal
3710 2007-01-10 Sa Liu <saliu@de.ibm.com>
3711 Ben Elliston <bje@au.ibm.com>
3713 * spu.h (STACK_SAVE_AREA): Use VOIDmode for SAVE_FUNCTION, SImode
3714 for SAVE_NONLOCAL and Pmode for any other save level.
3715 * spu-protos.h (spu_restore_stack_block): Declare.
3716 * spu.md (save_stack_block): Remove.
3717 (restore_stack_block): Call spu_restore_stack_block.
3718 * spu.c (spu_restore_stack_block): New function.
3719 (spu_expand_epilogue): Remove old comment.
3721 2007-01-09 Zdenek Dvorak <dvorakz@suse.cz>
3723 PR tree-optimization/30322
3724 * tree-ssa-loop-ivopts.c (fold_affine_expr, iv_value): Removed.
3725 (cand_value_at): Return the value as aff_tree.
3726 (may_eliminate_iv): Convert the bound from aff_tree to tree.
3727 * tree-affine.c (aff_combination_add_cst, aff_combination_add_product,
3728 aff_combination_mult): New functions.
3729 (aff_combination_add): Use aff_combination_add_cst.
3730 (aff_combination_convert): Allow conversions to a wider type.
3731 (tree_to_aff_combination): Handle BIT_NOT_EXPR.
3732 * tree-affine.h (aff_combination_mult): Declare.
3734 2007-01-09 Carlos O'Donell <carlos@codesourcery.com>
3736 * doc/tm.texi: Update documentation to reflect reality of exec
3737 and start file search behaviours. Update copyright year.
3738 * doc/invoke.texi: Explain how GCC_EXEC_PREFIX is used to find
3739 header file directories.
3741 2007-01-09 Uros Bizjak <ubizjak@gmail.com>
3743 * config/i386/i386.md (*sinxf2): Rename to *sinxf2_i387.
3744 (*cosxf2): Rename to cosxf2_i387.
3745 (*sindf2, *sinsf2): Extend operand 1 to XFmode. Macroize patterns
3746 using X87MODEF12 mode macro. Rename patterns to
3747 *sin_extend<mode>xf2_i387. Use SSE_FLOAT_MODE_P to disable patterns
3749 (*cosdf2, *cossf2): Ditto.
3750 (sincosdf3, sincossf3): Ditto. Rewrite corresponding splitters
3751 to match extended input operands.
3752 (sincos<mode>3): New expander.
3753 (*sinextendsfdf2, *cosextendsfdf2, *sincosextendsfdf3): Remove
3754 insn patterns and corresponding splitters.
3756 2007-01-09 Kaz Kojima <kkojima@gcc.gnu.org>
3758 * config/sh/t-linux (TARGET_LIBGCC2_CFLAGS): Delete.
3759 (SHLIB_MAPFILES, SHLIB_LINK, SHLIB_INSTALL): Likewise.
3761 2007-01-09 Nicolas Pitre <nico@cam.org>
3764 * arm/ieee754-df.S (Lad_s): Also test the low word of X for zero.
3766 2007-01-08 Geoffrey Keating <geoffk@apple.com>
3768 * target.h (struct gcc_target): New field library_rtti_comdat.
3769 * target-def.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): New.
3770 (TARGET_CXX): Add TARGET_CXX_LIBRARY_RTTI_COMDAT.
3771 * doc/tm.texi (C++ ABI): Document TARGET_CXX_LIBRARY_RTTI_COMDAT.
3772 * config/darwin.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
3774 2007-01-08 Geoffrey Keating <geoffk@apple.com>
3776 * doc/invoke.texi (Optimize Options): Correct description of -O0.
3778 2007-01-08 Richard Guenther <rguenther@suse.de>
3780 * tree.h (force_fit_type_double): Export.
3781 (force_fit_type): Remove.
3782 * fold-const.c (force_fit_type_double): New function.
3783 (force_fit_type): Remove.
3784 (int_const_binop): Use it.
3785 (fold_convert_const_int_from_int): Likewise.
3786 (fold_convert_const_int_from_real): Likewise.
3787 (fold_div_compare): Likewise.
3788 (fold_sign_changed_comparison): Likewise.
3789 (fold_unary): Likewise.
3790 (fold_negate_const): Likewise.
3791 (fold_abs_const): Likewise.
3792 (fold_not_const): Likewise.
3793 * c-common.c (shorten_compare): Use force_fit_type_double.
3794 * convert.c (convert_to_pointer): Likewise.
3796 2007-01-08 Richard Guenther <rguenther@suse.de>
3798 * tree.h (build_int_cst_wide_type): Export.
3799 * tree.c (build_int_cst_wide_type): New function.
3800 (build_int_cst_wide): Fix comment.
3801 * builtins.c (fold_builtin_object_size): Use build_int_cst
3802 to build -1 or 0 of the correct type. Use fit_double_type
3803 to check for overflow.
3804 * fold-const.c (optimize_bit_field_compare): Use build_int_cst_type
3806 (decode_field_reference): Likewise.
3807 (all_ones_mask_p): Likewise.
3808 (native_interpret_int): Use build_int_cst_wide_type.
3809 (fold_binary): Use build_int_cst_type to build an all-ones
3811 * stor-layout.c (set_sizetype): Use build_int_cst_wide_type.
3813 2007-01-08 Daniel Jacobowitz <dan@codesourcery.com>
3815 * config/pa/t-pa64 (libgcc_stub.a): Use $(T).
3817 2007-01-09 Ben Elliston <bje@au.ibm.com>
3819 * genautomata.c (STATS_OPTION): New option.
3820 (stats_flag): New flag.
3821 (gen_automata_option): Handle it.
3822 (initiate_automaton_gen): Ditto.
3823 (write_automata): Output statistics only if stats_flag is
3824 set. Likewise, output time statistics only if time_flag is set.
3825 * doc/md.texi (Processor pipeline description): Document new flag.
3827 2007-01-08 Richard Guenther <rguenther@suse.de>
3829 * builtins.c (fold_builtin_int_roundingfn): Use fit_double_type.
3830 * tree.c (build_int_cst_type): Likewise.
3831 (size_in_bytes): Don't call force_fit_type on the result.
3832 (int_fits_type_p): Use fit_double_type.
3833 * fold-const.c (fit_double_type): New function.
3834 (force_fit_type): Use it.
3835 * tree.h (fit_double_type): Export.
3837 2007-01-08 Jan Hubicka <jh@suse.cz>
3839 * tree-vectorizer.c (gate_increase_alignment): Fix return type.
3840 * ipa.c (function_and_variable_visibility): Fix return type.
3842 2007-01-08 Richard Guenther <rguenther@suse.de>
3844 * tree-ssa-ccp.c (maybe_fold_offset_to_array_ref): Use type
3845 of offset to build the index.
3846 * tree-pretty-print.c (dump_generic_node): Don't build negated
3847 const just for printing.
3848 * c-pretty-print.c (pp_c_integer_constant): Likewise.
3849 * builtins.c (fold_builtin_int_roundingfn): Check if result
3850 fits the type by using force_fit_type and comparing the result.
3851 * predict.c (predict_loops): Use compare_tree_int for comparison.
3852 * tree.c (build_int_cst): Fall back to integer_type_node for
3854 (build_int_cst_wide): Assert type is non-null.
3856 2007-01-08 Roberto Costa <roberto.costa@st.com>
3858 * tree-vrp.c (extract_range_from_cond_expr): New.
3859 (extract_range_from_expr): Handle COND_EXPR nodes used as expressions.
3860 * tree-ssa-ccp.c (get_maxval_strlen): Handle COND_EXPR nodes used
3862 (fold_stmt): Bug fix, avoid infinite recursion when folding COND_EXPRs.
3863 * tree-ssa-forwprop.c (simplify_cond, forward_propagate_into_cond,
3864 tree_ssa_forward_propagate_single_use_vars): Handle COND_EXPR nodes
3865 used as expressions.
3866 * tree-object-size.c (cond_expr_object_size): New.
3867 (collect_object_sizes_for): Handle COND_EXPR nodes used as expressions.
3869 2007-01-08 Jan Hubicka <jh@suse.cz>
3871 * tree-ssa-forwprop.c (forward_propagate_into_cond,
3872 tree_ssa_forward_propagate_single_use_va): Release defs of propagated
3875 2007-01-08 Richard Guenther <rguenther@suse.de>
3877 PR tree-optimization/23603
3878 * tree-vrp.c (set_value_range_to_truthvalue): New function.
3879 (extract_range_from_binary): Fall back to truthvalue instead of
3880 varying for TRUTH_*_EXPR.
3881 (extract_range_from_comparison): Fall back to truthvalue instead of
3883 (vrp_visit_phi_node): Don't adjust new range bounds to +INF/-INF
3884 if all visited PHI values were constant.
3886 2007-01-08 Jan Hubicka <jh@suse.cz>
3888 * cgraphunit.c (cgraph_process_new_functions): Reset reachable flag.
3889 (cgraph_analyze_function): break out from ...
3890 (cgraph_finalize_compilation_unit): ... here.
3891 (cgraph_expand_function): Remove forgoten commented out line.
3892 (cgraph_optimize): Analyze functions.
3894 2007-01-08 Jan Hubicka <jh@suse.cz>
3896 * tree-pas.h (TODO_remove_function): New flag.
3897 (TODO_update*): Renumber.
3898 (pass_ipa_increase_alignment,
3899 pass_ipa_function_and_variable_visibility): New passes.
3900 * cgraphunit.c (cgraph_increase_alignment): Move to tree-vectorizer.c
3901 (cgraph_function_and_variable_visibility): Move to ipa.c
3902 (cgraph_optimize): Don't call cgraph_function_and_variable_visibility,
3903 cgraph_increase_alignment.
3904 * ipa-inline.c (cgraph_decide_inlining): Don't push timevar.
3905 (cgraph_decide_inlining_incrementally): Push TV_INTEGRATION before
3906 calling tree-inline.
3907 (cgraph_early_inlining): Do not call cgraph_remove_unreachable_nodes.
3908 (pass_ipa_inline, pass_early_ipa_inlining): Set TODO_remove_functions
3909 * tree-vectorizer.c (increase_alignment): Move here from cgraphunit.c
3910 (gate_increase_alignment): New function.
3911 (pass_ipa_increase_alignment): New pass.
3912 * ipa.c: Inline tree-pass.h and timevar.h
3913 (function_and_variable_visibility): Move here from cgraphunit.c
3914 * tree-optimize.c (pass_early_local_passes): Add TODO_remove_functions.
3915 * passes.c (init_optimization_passes): Add the two new passes.
3916 (execute_todo): Handle cgraph_remove_functions.
3918 2007-01-08 Nick Clifton <nickc@redhat.com>
3920 * config/frv/predicates.md (reg_or_0_operand): Accept
3923 2007-01-08 Ralf Corsépius <ralf.corsepius@rtems.org>
3925 * config/bfin/rtems.h, config/bfin/t-rtems: New.
3926 * config.gcc: Add bfin*-rtems*.
3928 2007-01-08 Mark Shinwell <shinwell@codesourcery.com>
3930 * c.opt: Add -flax-vector-conversions.
3931 * c-typeck.c (convert_for_assignment): Pass flag to
3932 vector_types_convertible_p to allow emission of note.
3933 (digest_init): Likewise.
3934 * c-opts.c: Handle -flax-vector-conversions.
3935 * c-common.c (flag_lax_vector_conversions): New.
3936 (vector_types_convertible_p): Unless -flax-vector conversions
3937 has been passed, disallow conversions between vectors with
3938 differing numbers of subparts and/or element types. If such
3939 a conversion is disallowed, possibly emit a note on the first
3940 occasion only to inform the user of -flax-vector-conversions.
3941 The new last argument specifies this.
3942 * c-common.h (flag_lax_vector_conversions): New.
3943 (vector_types_convertible_p): Add extra argument.
3944 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Use
3945 char_type_node for V*QI type vectors.
3946 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins):
3947 Update to satisfy new typechecking rules.
3948 * config/rs6000/altivec.h (vec_cmple): Use vec_cmpge, for both
3950 * doc/invoke.texi (C Dialect Options): Document
3951 -flax-vector-conversions.
3953 2007-01-08 Mark Shinwell <shinwell@codesourcery.com>
3955 PR tree-optimization/29877
3956 * tree-ssa-ter.c (is_replaceable_p): Deem assignments with
3957 a register variable on the RHS to not be replaceable.
3959 2007-01-08 Chen Liqin <liqin@sunnorth.com.cn>
3960 * config/score/t-score-elf (MULTILIB_OPTIONS): Change.
3961 * config/score/predicates.md (const_uimm5, sr0_operand, const_simm12,
3962 const_simm15, const_pow2, const_npow2): Added.
3963 * config/score/misc.md (insv, extv, extzv, movmemsi,
3964 move_lbu_a/b, mov_lhu_a/b etc): Added and fix some bug.
3965 * config/score/score.c (score_address_cost, score_select_cc_mode):
3967 Change CONST_OK_FOR_LETTER_P/EXTRA_CONSTRAINT define.
3968 Update score_rtx_costs for MACRO TARGET_RTX_COSTS.
3969 Update score_print_operand.
3970 * config/score/score.h (DATA_ALIGNMENT, SELECT_CC_MODE): Added.
3971 Adjust register allocate order and update some macro define.
3972 * config/score/score-mdaux.c (mdx_unaligned_load, mdx_unsigned_store,
3973 mdx_block_move_straight, mdx_block_move_loop_head,
3974 mdx_block_move_loop_body, mdx_block_move_loop_foot, mdx_block_move_loop,
3975 mdx_block_move): Added.
3976 (mdx_movsicc, mdp_select_add_imm, mdp_select, mds_zero_extract_andi,
3977 mdp_limm): Updated and fix some bug and typo.
3978 * config/score/score.md (movqi/hi/si, add/sub/zero/ext): Updated.
3979 (movsf, movdf, doloop_end): Added.
3981 2007-01-08 Kazu Hirata <kazu@codesourcery.com>
3983 * config/arm/arm.c, config/arm/arm.h, config/arm/arm.md,
3984 config/arm/thumb2.md: Fix comment typos.
3985 * doc/extend.texi: Fix a typo.
3987 2007-01-07 Eric Christopher <echristo@apple.com>
3989 * configure.ac: Check for __stack_chk_fail for darwin.
3990 * configure: Regenerate.
3992 2007-01-07 Richard Guenther <rguenther@suse.de>
3994 * tree-vrp.c (extract_range_from_assert): CSE calls to
3995 compare_values where possible.
3996 (extract_range_from_unary_expr): Likewise.
3998 2007-01-07 Anatoly Sokolov <aesok@post.ru>
4000 * config/avr/avr-protos.h (call_insn_operand): Delete prototype.
4001 * config/avr/avr.c (call_insn_operand): Delete function.
4002 * config/avr/avr.md (*pushqi, *pushhi, *pushsi, *pushsf): Use REG_SP
4003 instead of register number. Use predicates.
4004 * config/avr/predicates.md (const0_operand, reg_or_0_operand,
4005 call_insn_operand): Add.
4007 2007-01-06 Jan Hubicka <jh@suse.cz>
4009 * tree-pass.h (pass_build_cgraph_edges): Declare.
4010 * cgraphunit.c (record_refernece): Move to cgraphbuild.c
4011 (visited_nodes): Remove.
4012 (cgraph_create_edges): Move to cgraphbuild.c; rename to
4013 build_cgrpah_edges; make visited_nodes local.
4014 (cgraph_process_new_functions): DO not call initialize_inline_failed.
4015 (record_references_in_initializer): Move to cgraphbuild.c
4016 (initialize_inline_failed, rebuild_cgraph_edges,
4017 pass_rebuild_cgraph_edges): Move to cgraphbuild.c.
4018 (verify_cgraph_node): Make visited_nodes local.
4019 (cgraph_analyze_function): Do not call cgraph_create_edges and
4020 initialize_inline_failed.
4021 (cgraph_expand_function): Do not call cgraph_lower_function;
4022 assert that function is already lowered.
4023 * Makefile.in (cgraphbuild.o): New.
4024 * passes.c (init_optimization_passes): Add pass_build_cgraph_edges
4025 at the end of lowering passes.
4027 2007-01-06 Steven Bosscher <steven@gcc.gnu.org>
4029 * ifcvt.c (cond_move_convert_if_block): New function, code
4030 factored out from...
4031 (cond_move_process_if_block): ...here. Call the new function
4032 on the THEN and ELSE blocks.
4033 (merge_if_block): Do not copy global_live_at_end, merge_blocks
4034 already takes care of this.
4036 2007-01-05 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
4039 * tree.h (TREE_OVERFLOW_P): New.
4040 * c-typeck.c (parser_build_unary_op): Warn only if result
4041 overflowed and operands did not.
4042 (parser_build_binary_op): Likewise.
4043 (convert_for_assignment): Remove redundant overflow_warning.
4044 * c-common.c (overflow_warning): Don't check or set TREE_OVERFLOW.
4046 2007-01-05 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
4048 * c-typeck.c (store_init_value): Split over two lines to follow
4049 the GNU coding style.
4051 2007-01-05 Benjamin Kosnik <bkoz@redhat.com>
4053 * c-cppbuiltin.c (c_cpp_builtins): __GXX_EXPERIMENTAL_CPP0X__ to
4054 __GXX_EXPERIMENTAL_CXX0X__.
4055 * doc/cpp.texi: Same.
4057 2007-01-05 Richard Guenther <rguenther@suse.de>
4060 * tree.c (get_narrower): Do not construct COMPONENT_REFs
4061 with mismatched types. Instead explicitly build a
4062 conversion NOP_EXPR.
4064 2007-01-05 Ian Lance Taylor <iant@google.com>
4066 * c-common.c (decl_with_nonnull_addr_p): New function.
4067 (c_common_truthvalue_conversion): Call it.
4068 * c-typeck.c (build_binary_op): Likewise.
4069 * c-common.h (decl_with_nonnull_addr_p): Declare.
4071 2007-01-05 Jakub Jelinek <jakub@redhat.com>
4074 * libgcc2.c (__divdc3): Compare c and d against 0.0 instead of
4077 2007-01-05 Joel Brobecker <brobecker@adacore.com>
4079 * doc/install.texi (Final install): Document the fact that
4080 the GNAT runtime should not be stripped.
4082 2007-01-04 Jan Hubicka <jh@suse.cz>
4084 * tree-inline.c (fold_marked_statements): Update operand caches
4085 and EH after folding
4087 2007-01-04 Ian Lance Taylor <iant@google.com>
4089 * c-common.c (check_function_nonnull): Whitespace fix.
4091 2007-01-04 Jan Hubicka <jh@suse.cz>
4093 * tree-optimize.c (execute_fixup_cfg): Correct previously mistakely
4094 comitted older version of patch.
4095 (pass_fixup_cfg): Add TODOs to verify flow and statements, dump
4096 function, celanup cfg and collect garbage.
4098 2007-01-04 Mike Stump <mrs@apple.com>
4100 * Makefile.in (mostlyclean): Don't remove libgcc anymore.
4103 2007-01-04 Eric Christopher <echristo@apple.com>
4105 * libgcc2.c (__bswapsi2): Use SItype.
4106 (__bswapdi2): Use DItype.
4107 * libgcc2.h: Update for above.
4109 2007-01-04 Paul Brook <paul@codesourcery.com>
4111 * config/arm/arm.md (arm_mulsi3, thumb_mulsi3, mulsi3_compare0,
4112 mulsi_compare0_scratch, mulsi3addsi, mulsi3addsi_compare0,
4113 mulsi3addsi_compare0_scratch, mulsidi3adddi, mulsidi3,
4114 umulsidi3, umulsidi3adddi, smulsi3_highpart,
4115 umulsi3_highpart): Make conditional on !arm_arch6.
4116 (arm_mulsi3_v6, thumb_mulsi3_v6, mulsi3_compare0_v6,
4117 mulsi_compare0_scratch_v6, mulsi3addsi_v6, mulsi3addsi_compare0_v6,
4118 mulsi3addsi_compare0_scratch_v6, mulsidi3adddi_v6, mulsidi3_v6,
4119 umulsidi3_v6, umulsidi3adddi_v6, smulsi3_highpart_v6,
4120 umulsi3_highpart_v6): New insns.
4122 2007-01-04 Roger Sayle <roger@eyesopen.com>
4124 * fold-const.c (fold_convert): When casting an expression to void,
4125 fold_ignored_result may discover a GIMPLE_MODIFY_STMT which doesn't
4126 have a type. Instead of attempting to build a NOP_EXPR, return
4127 these "special" trees directly.
4129 2007-01-04 Joseph Myers <joseph@codesourcery.com>
4131 * config/rs6000/rs6000.c (rs6000_rtx_costs): Make adjustment for
4132 MULT inside MINUS as either argument. Use rs6000_cost->dmul -
4133 rs6000_cost->fp not 0 as adjustment for outer NEG.
4135 2007-01-04 Jan Hubicka <jh@suse.cz>
4137 * cgraph.c (cgraph_release_function_body): New function.
4138 (cgraph_remove_node): Use it.
4139 * cgraph.h (cgraph_release_function_body): Declare.
4140 * cgraphunit.c (cgraph_expand_function): Use it.
4141 * ipa.c (cgraph_remove_unreahchable_nodes): Use it.
4142 * tree-ssa.c (delete_tree_ssa): Allow to be called before aliasing
4143 is initialized and while compilation of other function is running.
4144 * tree-optimize.c (execute_free_cfg_annotations): Move code to clear
4145 statement CFG annotations from here to ...
4146 * tree-cfg.c (delete_tree_cfg_annotations): ... here.
4148 2007-01-04 Zdenek Dvorak <dvorakz@suse.cz>
4150 * cfgloop.h (enum li_flags): Make the constants powers of two.
4152 2007-01-04 Jan Hubicka <jh@suse.cz>
4154 * tree-inline.c (copy_bb): Insert new statements to statements_to_fold
4156 (fold_marked_statements): New function.
4157 (optimize_inline_calls, tree_function_versioning): Fold new statements.
4158 * tree-inline.h (copy_body_data): Add statements_to_fold.
4160 2007-01-03 Daniel Jacobowitz <dan@codesourcery.com>
4162 * config.gcc: Mention libgcc/config.host.
4163 * Makefile.in: Update comments mentioning libgcc.
4164 (LIBGCC, INSTALL_LIBGCC, GCC_PARTS, mklibgcc): Delete.
4165 (all.cross, start.encap, rest.encap, rest.cross): Update
4166 dependencies for libgcc move.
4167 (libgcc.mk, LIBGCC_DEPS, libgcov.a, libgcc.a, stmp-multilib)
4168 (clean-target, clean-target-libgcc): Delete.
4169 (srcdirify, GCC_EXTRA_PARTS): New macros.
4170 (libgcc-support, libgcc.mvars): New rules.
4171 (distclean): Remove mention of mklibgcc.
4172 (install): Don't reference INSTALL_LIBGCC.
4173 (install-common): Don't reference EXTRA_PARTS.
4174 (install-libgcc, install-multilib): Delete rules.
4175 * mklibgcc.in: Delete file.
4176 * doc/configfiles.texi: Don't mention mklibgcc.
4178 * config/i386/t-darwin (SHLIB_VERPFX): Delete (moved to libgcc).
4179 * config/i386/t-darwin64 (SHLIB_VERPFX): Likewise.
4180 * config/rs6000/t-darwin (SHLIB_VERPFX): Likewise.
4181 * config/rs6000/t-ppccomm (TARGET_LIBGCC2_CFLAGS, SHLIB_MAPFILES)
4182 (mklibgcc, ldblspecs): Likewise.
4184 * config/i386/t-nwld (libgcc.def, libc.def, libpcre.def)
4185 (posixpre.def): Use $(T).
4186 (SHLIB_EXT, SHLIB_NAME, SHLIB_SLIBDIR_QUAL, SHLIB_DEF, SHLIB_MAP)
4187 (SHLIB_SRC, SHLIB_INSTALL): Delete.
4188 (SHLIB_LINK): Make dummy.
4189 * config/t-slibgcc-darwin: Delete contents except for dummy SHLIB_LINK.
4191 * config/frv/t-linux (EXTRA_MULTILIB_PARTS): Clear.
4193 * config/alpha/t-crtfm: Use $(T) in rules for EXTRA_PARTS.
4194 * config/alpha/t-vms, config/alpha/t-vms64, config/fr30/t-fr30,
4195 config/i386/t-rtems-i386, config/ia64/t-ia64, config/rs6000/t-beos,
4196 config/rs6000/t-newas, config/sparc/t-elf: Likewise.
4198 * configure.ac (all_outputs): Remove mklibgcc.
4199 * configure: Regenerated.
4201 2007-01-03 Josh Conner <jconner@apple.com>
4204 * calls.c (compute_argument_addresses): Set stack and stack_slot
4205 for partial args, too.
4206 (store_one_arg): Use locate.size.constant for the size when
4207 generating a save_area.
4209 2007-01-03 Robert Kennedy <jimbob@google.com>
4211 * tree-cfg.c (tree_merge_blocks): Release SSA_NAME phi results
4212 whose definitions are deleted due to basic block merging.
4214 2007-01-03 Paul Brook <paul@codesourcery.com>
4217 * config/arm/arm.c (output_return_instruction): Pop PC in interrupt
4219 (use_return_insn): Return 0 for Thumb interrupt functions.
4220 (print_multi_reg): Add rfe argument for IRQ returns.
4221 (arm_output_epilogue): Pop interrupt return address directly into PC.
4222 (arm_expand_prologue): Only adjust IRQ return address in Arm mode.
4224 2007-01-03 Paul Brook <paul@codesourcery.com>
4226 Merge from sourcerygxx-4_1.
4227 * config/arm/thumb2.md: New file.
4228 * config/arm/elf.h (JUMP_TABLES_IN_TEXT_SECTION): Return True for
4230 * config/arm/coff.h (JUMP_TABLES_IN_TEXT_SECTION): Ditto.
4231 * config/arm/aout.h (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
4232 (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump tables.
4233 * config/arm/aof.h (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump
4235 (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
4236 * config/arm/ieee754-df.S: Use macros for Thumb-2/Unified asm
4238 * config/arm/ieee754-sf.S: Ditto.
4239 * config/arm/arm.c (thumb_base_register_rtx_p): Rename...
4240 (thumb1_base_register_rtx_p): ... to this.
4241 (thumb_index_register_rtx_p): Rename...
4242 (thumb1_index_register_rtx_p): ... to this.
4243 (thumb_output_function_prologue): Rename...
4244 (thumb1_output_function_prologue): ... to this.
4245 (thumb_legitimate_address_p): Rename...
4246 (thumb1_legitimate_address_p): ... to this.
4247 (thumb_rtx_costs): Rename...
4248 (thumb1_rtx_costs): ... to this.
4249 (thumb_compute_save_reg_mask): Rename...
4250 (thumb1_compute_save_reg_mask): ... to this.
4251 (thumb_final_prescan_insn): Rename...
4252 (thumb1_final_prescan_insn): ... to this.
4253 (thumb_expand_epilogue): Rename...
4254 (thumb1_expand_epilogue): ... to this.
4255 (arm_unwind_emit_stm): Rename...
4256 (arm_unwind_emit_sequence): ... to this.
4257 (thumb2_legitimate_index_p, thumb2_legitimate_address_p,
4258 thumb1_compute_save_reg_mask, arm_dwarf_handle_frame_unspec,
4259 thumb2_index_mul_operand, output_move_vfp, arm_shift_nmem,
4260 arm_save_coproc_regs, thumb_set_frame_pointer, arm_print_condition,
4261 thumb2_final_prescan_insn, thumb2_asm_output_opcode, arm_output_shift,
4262 thumb2_output_casesi): New functions.
4263 (TARGET_DWARF_HANDLE_FRAME_UNSPEC): Define.
4264 (FL_THUMB2, FL_NOTM, FL_DIV, FL_FOR_ARCH6T2, FL_FOR_ARCH7,
4265 FL_FOR_ARCH7A, FL_FOR_ARCH7R, FL_FOR_ARCH7M, ARM_LSL_NAME,
4266 THUMB2_WORK_REGS): Define.
4267 (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv, arm_condexec_count,
4268 arm_condexec_mask, arm_condexec_masklen)): New variables.
4269 (all_architectures): Add armv6t2, armv7, armv7a, armv7r and armv7m.
4270 (arm_override_options): Check new CPU capabilities.
4271 Set new architecture flag variables.
4272 (arm_isr_value): Handle v7m interrupt functions.
4273 (user_return_insn): Return 0 for v7m interrupt functions. Handle
4275 (const_ok_for_arm): Handle Thumb-2 constants.
4276 (arm_gen_constant): Ditto. Use movw when available.
4277 (arm_function_ok_for_sibcall): Return false for v7m interrupt
4279 (legitimize_pic_address, arm_call_tls_get_addr): Handle Thumb-2.
4280 (thumb_find_work_register, arm_load_pic_register,
4281 legitimize_tls_address, arm_address_cost, load_multiple_sequence,
4282 emit_ldm_seq, emit_stm_seq, arm_select_cc_mode, get_jump_table_size,
4283 print_multi_reg, output_mov_long_double_fpa_from_arm,
4284 output_mov_long_double_arm_from_fpa, output_mov_double_fpa_from_arm,
4285 output_mov_double_fpa_from_arm, output_move_double,
4286 arm_compute_save_reg_mask, arm_compute_save_reg0_reg12_mask,
4287 output_return_instruction, arm_output_function_prologue,
4288 arm_output_epilogue, arm_get_frame_offsets, arm_regno_class,
4289 arm_output_mi_thunk, thumb_set_return_address): Ditto.
4290 (arm_expand_prologue): Handle Thumb-2. Use arm_save_coproc_regs.
4291 (arm_coproc_mem_operand): Allow POST_INC/PRE_DEC.
4292 (arithmetic_instr, shift_op): Use arm_shift_nmem.
4293 (arm_print_operand): Use arm_print_condition. Handle '(', ')', '.',
4295 (arm_final_prescan_insn): Use extract_constrain_insn_cached.
4296 (thumb_expand_prologue): Use thumb_set_frame_pointer.
4297 (arm_file_start): Output directive for unified syntax.
4298 (arm_unwind_emit_set): Handle stack alignment instruction.
4299 * config/arm/lib1funcs.asm: Remove default for __ARM_ARCH__.
4300 Add v6t2, v7, v7a, v7r and v7m.
4301 (RETLDM): Add Thumb-2 code.
4302 (do_it, shift1, do_push, do_pop, COND, THUMB_SYNTAX): New macros.
4303 * config/arm/arm.h (TARGET_CPU_CPP_BUILTINS): Define __thumb2__.
4304 (TARGET_THUMB1, TARGET_32BIT, TARGET_THUMB2, TARGET_DSP_MULTIPLY,
4305 TARGET_INT_SIMD, TARGET_UNIFIED_ASM, ARM_FT_STACKALIGN, IS_STACKALIGN,
4306 THUMB2_TRAMPOLINE_TEMPLATE, TRAMPOLINE_ADJUST_ADDRESS,
4307 ASM_OUTPUT_OPCODE, THUMB2_GO_IF_LEGITIMATE_ADDRESS,
4308 THUMB2_LEGITIMIZE_ADDRESS, CASE_VECTOR_PC_RELATIVE,
4309 CASE_VECTOR_SHORTEN_MODE, ADDR_VEC_ALIGN, ASM_OUTPUT_CASE_END,
4310 ADJUST_INSN_LENGTH): Define.
4311 (TARGET_REALLY_IWMMXT, TARGET_IWMMXT_ABI, CONDITIONAL_REGISTER_USAGE,
4312 STATIC_CHAIN_REGNUM, HARD_REGNO_NREGS, INDEX_REG_CLASS,
4313 BASE_REG_CLASS, MODE_BASE_REG_CLASS, SMALL_REGISTER_CLASSES,
4314 PREFERRED_RELOAD_CLASS, SECONDARY_OUTPUT_RELOAD_CLASS,
4315 SECONDARY_INPUT_RELOAD_CLASS, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P,
4316 TRAMPOLINE_SIZE, INITIALIZE_TRAMPOLINE, HAVE_PRE_INCREMENT,
4317 HAVE_POST_DECREMENT, HAVE_PRE_DECREMENT, HAVE_PRE_MODIFY_DISP,
4318 HAVE_POST_MODIFY_DISP, HAVE_PRE_MODIFY_REG, HAVE_POST_MODIFY_REG,
4319 REGNO_MODE_OK_FOR_BASE_P, LEGITIMATE_CONSTANT_P,
4320 REG_MODE_OK_FOR_BASE_P, REG_OK_FOR_INDEX_P, GO_IF_LEGITIMATE_ADDRESS,
4321 LEGITIMIZE_ADDRESS, THUMB2_LEGITIMIZE_ADDRESS,
4322 GO_IF_MODE_DEPENDENT_ADDRESS, MEMORY_MOVE_COST, BRANCH_COST,
4323 ASM_APP_OFF, ASM_OUTPUT_CASE_LABEL, ARM_DECLARE_FUNCTION_NAME,
4324 FINAL_PRESCAN_INSN, PRINT_OPERAND_PUNCT_VALID_P,
4325 PRINT_OPERAND_ADDRESS): Adjust for Thumb-2.
4326 (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv): New declarations.
4327 * config/arm/arm-cores.def: Add arm1156t2-s, cortex-a8, cortex-r4 and
4329 * config/arm/arm-tune.md: Regenerate.
4330 * config/arm/arm-protos.h: Update prototypes.
4331 * config/arm/vfp.md: Enable patterns for Thumb-2.
4332 (arm_movsi_vfp): Add movw alternative. Use output_move_vfp.
4333 (arm_movdi_vfp, movsf_vfp, movdf_vfp): Use output_move_vfp.
4334 (thumb2_movsi_vfp, thumb2_movdi_vfp, thumb2_movsf_vfp,
4335 thumb2_movdf_vfp, thumb2_movsfcc_vfp, thumb2_movdfcc_vfp): New.
4336 * config/arm/libunwind.S: Add Thumb-2 code.
4337 * config/arm/constraints.md: Update include Thumb-2.
4338 * config/arm/ieee754-sf.S: Add Thumb-2/Unified asm support.
4339 * config/arm/ieee754-df.S: Ditto.
4340 * config/arm/bpabi.S: Ditto.
4341 * config/arm/t-arm (MD_INCLUDES): Add thumb2.md.
4342 * config/arm/predicates.md (low_register_operand,
4343 low_reg_or_int_operand, thumb_16bit_operator): New.
4344 (thumb_cmp_operand, thumb_cmpneg_operand): Rename...
4345 (thumb1_cmp_operand, thumb1_cmpneg_operand): ... to this.
4346 * config/arm/t-arm-elf: Add armv7 multilib.
4347 * config/arm/arm.md: Update patterns for Thumb-2 and Unified asm.
4349 (UNSPEC_STACK_ALIGN, ce_count): New.
4350 (arm_incscc, arm_decscc, arm_umaxsi3, arm_uminsi3,
4351 arm_zero_extendsidi2, arm_zero_extendqidi2): New
4353 * config/arm/fpa.md: Update patterns for Thumb-2 and Unified asm.
4354 (thumb2_movsf_fpa, thumb2_movdf_fpa, thumb2_movxf_fpa,
4355 thumb2_movsfcc_fpa, thumb2_movdfcc_fpa): New insns.
4356 * config/arm/cirrus.md: Update patterns for Thumb-2 and Unified asm.
4357 (cirrus_thumb2_movdi, cirrus_thumb2_movsi_insn,
4358 thumb2_cirrus_movsf_hard_insn, thumb2_cirrus_movdf_hard_insn): New
4360 * doc/extend.texi: Document ARMv7-M interrupt functions.
4361 * doc/invoke.texi: Document Thumb-2 new cores+architectures.
4363 2007-01-03 Jakub Jelinek <jakub@redhat.com>
4365 * unwind-dw2.c (SIGNAL_FRAME_BIT, EXTENDED_CONTEXT_BIT): Define.
4366 (struct _Unwind_Context): Rename args_size to flags, remove
4367 signal_frame field, add a new args_size field and version field.
4368 (_Unwind_IsSignalFrame, _Unwind_SetSignalFrame,
4369 _Unwind_IsExtendedContext): New inline functions.
4370 (_Unwind_GetGR, _Unwind_SetGR, _Unwind_GetGRPtr, _Unwind_SetGRPtr):
4371 Assume by_value array is only present if _Unwind_IsExtendedContext.
4372 (_Unwind_GetIPInfo, execute_cfa_program, uw_frame_state_for): Use
4373 _Unwind_IsSignalFrame.
4374 (__frame_state_for): Initialize context.flags to EXTENDED_CONTEXT_BIT.
4375 (uw_update_context_1): Use _Unwind_SetSignalFrame.
4376 (uw_init_context_1): Initialize context->flags to
4377 EXTENDED_CONTEXT_BIT.
4378 * config/rs6000/linux-unwind.h (frob_update_context): Use
4379 _Unwind_SetSignalFrame.
4381 2007-01-03 Andrew Pinski <andrew_pinski@playstation.sony.com>
4384 * gimplify.c (gimplify_modify_expr_complex_part): Move below
4385 tree_to_gimple_tuple. Call tree_to_gimple_tuple when we need
4388 2007-01-03 Kazu Hirata <kazu@codesourcery.com>
4390 * config/i386/i386.h (NON_STACK_REG_P, REGNO_OK_FOR_SIREG_P,
4391 REGNO_OK_FOR_DIREG_P, REWRITE_ADDRESS, ASM_OPERAND_LETTER,
4392 RET, AT_SP): Remove.
4393 * config/i386/i386.md (*sse_prologue_save_insn): Use return
4396 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
4397 * cfgbuild.c (find_basic_blocks): Likewise.
4398 * cfgrtl.c (rtl_create_basic_block): Likewise.
4399 * function.c (temp_slots_at_level): Likewise.
4400 * reg-stack.c (stack_regs_mentioned): Likewise.
4401 * regclass.c (allocate_reg_info): Likewise.
4402 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
4403 set_bb_for_stmt, move_block_to_fn): Likewise.
4404 * tree-complex.c (tree_lower_complex): Likewise.
4405 * vec.h (VEC_safe_grow_cleared): New.
4407 * cgraphunit.c, tree-ssa-alias.c: Fix comment typos.
4409 2007-01-03 Zdenek Dvorak <dvorakz@suse.cz>
4411 * loop-unswitch.c (unswitch_loop): Pass probabilities to loopify.
4412 * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Pass probabilities
4414 * cfgloopmanip.c (scale_loop_frequencies): Export.
4415 (loopify): Scale the frequencies by prescribed coefficients.
4416 (set_zero_probability): New function.
4417 (duplicate_loop_to_header_edge): Improve updating of frequencies.
4418 (lv_adjust_loop_entry_edge, loop_version): Set probabilities
4419 and frequencies according to arguments.
4420 * tree-ssa-loop-manip.c (tree_unroll_loop): Set probabilities
4422 * cfg.c (scale_bbs_frequencies_int): Allow scaling the frequencies up.
4423 * modulo-sched.c (sms_schedule): Set probabilities for entering
4424 versioned loop correctly.
4425 * tree-vect-transform.c (vect_transform_loop): Ditto.
4426 * cfgloop.h (loopify, loop_version): Declaration changed.
4427 (scale_loop_frequencies): Declared.
4429 2007-01-02 Jan Hubicka <jh@suse.cz>
4431 * cgraph.c: Include tree-flow.h
4432 (cgraph_add_new-function): Handle IPA_SSA mode; execute
4434 * cgraph.h (enum cgraph_state): Add CGRAPH_STATE_IPA_SSA.
4435 * tree-pass.h (pass_all_early_optimizations): Declare.
4436 * cgraphunit.c (cgraph_process_new_functions): Add IPA_SSA; execute
4438 (cgraph_analyze_function): Do early_local_passes.
4439 * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
4440 Do not add referenced vars.
4441 * tree-optimize.c (gate_all_optimizations): Do not execute when not in
4443 (gate_all_early_local_passes): New gate.
4444 (pass_early_local_passes): Use new gate.
4445 (execute_early_local_optimizations): New functions.
4446 (gate_all_early_optimizations): New gate.
4447 (pass_all_early_optimizations): New pass.
4448 (execute_free_datastructures): Free SSA only when initialized.
4449 (gate_init_datastructures): Init only when optimizing.
4450 (tree_lowering_passes): Do early local passes when called late.
4451 * tree-profile.c (do_tree_profiling): Don't profile functions added
4453 (do_early_tree_profiling, pass_early_tree_profile): Kill.
4454 * tree-cfg.c (update_modified_stmts): Do not update when operands are
4456 * passes.c (init_optimizations_passes): Reorder so we go into SSA
4457 during early_local_passes.
4458 * Makefile.in (cgraph.o): Add dependency on tree-flow.h.
4461 2007-01-02 Carlos O'Donell <carlos@codesourcery.com>
4463 * Makefile.in: Update copyright year.
4465 2007-01-02 Carlos O'Donell <carlos@codesourcery.com>
4467 * Makefile.in: Export GCC_EXEC_PREFIX before calling $(RUNTEST)
4468 in $(lang_checks) and check-consistency targets.
4470 2007-01-02 Jan Hubicka <jh@suse.cz>
4472 * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
4473 Do not add referenced vars.
4474 * tree-cfg.c (update_modified_stmts): Do not update when SSA operands
4476 * passes.c (init_optimization_passes): Put mudflap_2 after
4477 free_datastructures.
4479 2007-01-02 Jan Hubicka <jh@suse.cz>
4481 * tree-optimize (execute_fixup_cfg): Set after_inlining flag.
4482 Set NOTHROW flag on call statements proved to be nothrow.
4483 Update statement of local calls so new pure/const functions are
4484 updated. Update_ssa when in ssa form. Mark PHI nodes of nonlocal
4486 (tree_rest_of_compilation): Register hooks and initialize bitmap
4487 early. Do not set after_inlining flag.
4489 2007-01-02 Steve Ellcey <sje@cup.hp.com>
4491 * sbitmap.c (HOST_BITS_PER_LONG_LONG): Change to
4492 HOST_BITS_PER_LONGLONG
4494 2007-01-02 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
4497 * c-typeck.c (store_init_value): Don't emit pedantic overflow
4498 warning for non-static initializers.
4500 2007-01-02 Steven Bosscher <steven@gcc.gnu.org>
4502 * config/alpha/alpha.md, arm/arm.c, darwin.c, frv/frv.md,
4503 m32r/m32r.c, m32r/m32r.c, mn10300/mn10300.md, pa/pa.c,
4504 rs6000/rs6000.c, s390/s390.md, sh/sh.md, sparc/sparc.c:
4505 Always use set_unique_reg_note to add REG_EQUAL notes.
4507 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
4510 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
4512 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
4513 * cfgbuild.c (find_basic_blocks): Likewise.
4514 * cfgrtl.c (rtl_create_basic_block): Likewise.
4515 * function.c (temp_slots_at_level): Likewise.
4516 * reg-stack.c (stack_regs_mentioned): Likewise.
4517 * regclass.c (allocate_reg_info): Likewise.
4518 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
4519 set_bb_for_stmt, move_block_to_fn): Likewise.
4520 * tree-complex.c (tree_lower_complex): Likewise.
4521 * vec.h (VEC_safe_grow_cleared): New.
4523 2007-01-02 Ian Lance Taylor <iant@google.com>
4525 * c-common.c (c_common_truthvalue_conversion): When warning about
4526 using an assignment as a truth value, set TREE_NO_WARNING.
4528 2007-01-02 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
4531 * c.opt (Wold-style-declaration): New.
4532 * doc/invoke.texi (C-only Warning Options): New.
4533 (Wold-style-declaration): Document it.
4534 (Wextra): Enabled by -Wextra.
4535 * c-opts.c (c_common_post_options): Enabled by -Wextra.
4536 * c-decl.c (declspecs_add_scspec): Replace -Wextra with
4537 -Wold-style-declaration.
4539 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
4541 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
4542 * cfgbuild.c (find_basic_blocks): Likewise.
4543 * cfgrtl.c (rtl_create_basic_block): Likewise.
4544 * function.c (temp_slots_at_level): Likewise.
4545 * reg-stack.c (stack_regs_mentioned): Likewise.
4546 * regclass.c (allocate_reg_info): Likewise.
4547 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
4548 set_bb_for_stmt, move_block_to_fn): Likewise.
4549 * tree-complex.c (tree_lower_complex): Likewise.
4550 * vec.h (VEC_safe_grow_cleared): New.
4552 2007-01-02 Douglas Gregor <doug.gregor@gmail.com>
4554 * c-common.c (c_common_nodes_and_builtins): Since variants of
4555 void_type_node get built before it is given a name, we need to
4556 give those variants the name, too.
4557 (complete_array_type): We need to work with the canonical main
4558 type of the array, from which we will build the qualified version.
4559 * params.def (PARAM_VERIFY_CANONICAL_TYPES): New.
4560 * print-tree.c (print_node): Display canonical type information
4562 * stor-layout.c (layout_type): When we don't know the
4563 alignment of a type for which we're building an array, we end up
4564 guessing wrong, so make the type require structural equality.
4565 * tree.c (make_node_stat): When we build a new type, it is its
4567 (build_type_attribute_qual_variant): When building an attribute
4568 variant, its canonical type is the non-attribute variant. However,
4569 if the attributes are target-dependent and they differ, we need to
4570 use structural equality checks for this type.
4571 (build_qualified_type): A qualified type is not equivalent to its
4572 unqualified variant; set the canonical type appropriately.
4573 (build_distinct_type_copy): When building a distinct type from
4574 another type, the new type is its own canonical type.
4575 (build_variant_type_copy): When building a new type variant, we
4576 assume that it is equivalent to the original type.
4577 (build_pointer_type_for_mode): When building a pointer type, also
4578 build a canonical type pointer.
4579 (build_reference_type_for_mode): When building a reference type,
4580 also build a canonical type reference.
4581 (build_index_type): When we can't hash an index type (e.g.,
4582 because its maximum value is negative), the index type requires
4583 structural equality tests.
4584 (build_array_type): Build the canonical form of an array type.
4585 (build_function_type): Function types require structural equality,
4586 because they contain default arguments, attributes, etc.
4587 (build_method_type_directly): Ditto for method types.
4588 (build_offset_type): Build the canonical offset type.
4589 (build_complex_type): Build the canonical vector type.
4590 (make_vector_type): Build the canonical vector type.
4591 * tree.h (TYPE_CANONICAL): New.
4592 (TYPE_STRUCTURAL_EQUALITY_P): New.
4593 (SET_TYPE_STRUCTURAL_EQUALITY): New.
4594 (struct tree_type): Added "canonical" field.
4595 * params.h (VERIFY_CANONICAL_TYPES): New.
4596 * doc/c-tree.texi (TYPE_CANONICAL): Document.
4597 (TYPE_STRUCTURAL_EQUALITY_P): Document.
4598 (SET_TYPE_STRUCTURAL_EQUALITY): Document.
4599 * doc/invoke.texi (verify-canonical-types): Document --param
4600 parameter for verifying canonical types.
4602 2007-01-02 Joseph Myers <joseph@codesourcery.com>
4604 * config.gcc (powerpc-*-eabispe*, powerpc-*-eabisimaltivec*,
4605 powerpc-*-eabisim*, powerpc-*-eabialtivec*, powerpc-*-eabi*,
4606 powerpc-*-rtems*, powerpc-wrs-vxworks, powerpc-wrs-vxworksae,
4607 powerpcle-*-eabisim*, powerpcle-*-eabi*): Add rs6000/e500.h to
4609 * config/rs6000/e500.h: New.
4610 * config/rs6000/eabi.h (TARGET_SPE_ABI, TARGET_SPE, TARGET_E500,
4611 TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE, TARGET_E500_DOUBLE):
4613 * config/rs6000/linuxspe.h (TARGET_SPE_ABI, TARGET_SPE,
4614 TARGET_E500, TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE,
4615 TARGET_E500_DOUBLE): Remove.
4616 * config/rs6000/vxworks.h (TARGET_SPE_ABI, TARGET_SPE,
4617 TARGET_E500, TARGET_ISEL, TARGET_FPRS): Remove.
4618 * config/rs6000/rs6000.h (CHECK_E500_OPTIONS): Define.
4619 * config/rs6000/rs6000.c (rs6000_override_options): Use
4622 2007-01-02 Joseph Myers <joseph@codesourcery.com>
4624 * config/rs6000/rs6000.c (print_operand): Check (TARGET_SPE ||
4625 TARGET_E500_DOUBLE), not TARGET_E500, for %y.
4626 (rs6000_generate_compare, rs6000_emit_sCOND, output_cbranch,
4627 rs6000_emit_cmove): Don't check TARGET_E500.
4628 * config/rs6000/rs6000.md (bunordered, bordered, sunordered,
4629 sordered): Don't check TARGET_E500.
4631 2007-01-01 Eric Christopher <echristo@apple.com>
4633 * config/mips/mips.c (mips_regno_mode_ok_for_base_p): Use
4634 HARD_REGISTER_NUM_P.
4636 2007-01-01 Roger Sayle <roger@eyesopen.com>
4638 * fold-const.c (fold_binary) <EQ_EXPR>: Fold "(X^C1) eq/ne C2" into
4639 "X eq/ne (C1^C2)". Fold "(X^Z) eq/ne (Y^Z)" as "X eq/ne Y" when Z
4640 has no side-effects. Fold "(X^C1) eq/ne (Y^C2)" as "(X^(C1^C2))
4643 2007-01-01 Mike Stump <mrs@apple.com>
4645 * configure.ac: Remove support for building with Apple's gcc-3.1.
4647 2007-01-02 Joseph Myers <joseph@codesourcery.com>
4650 * caller-save.c (add_stored_regs): Only handle SUBREGs if inner
4651 REG is a hard register. Do not modify REG before calling
4653 * rtlanal.c (subreg_get_info): Don't assert size of XMODE is a
4654 multiple of the size of YMODE for certain lowpart cases.
4656 2007-01-01 Andrew Pinski <pinskia@gmail.com>
4659 * gimplify (voidify_wrapper_expr): Update for
4662 2007-01-01 Andreas Schwab <schwab@suse.de>
4665 * config/ia64/ia64.c (ia64_compute_frame_size): Account space for
4666 save of BR0 in extra_spill_size instead of spill_size.
4667 (ia64_expand_prologue): Save BR0 outside of the gr/br/fr spill
4669 (ia64_expand_epilogue): Restore BR0 from its new location.
4671 2007-01-01 Andrew Pinski <pinskia@gmail.com>
4673 * gimplify.c (gimplify_init_constructor <case VECTOR_TYPE>):
4674 Use a temporary variable if the left hand side is not a gimple
4677 2007-01-01 Andrew Pinski <pinskia@gmail.com>
4679 * gimplify.c (gimplify_return_expr): Make the temporary variable
4680 for the return expression, a gimple register variable.
4682 2007-01-01 Jan Hubicka <jh@suse.cz>
4684 * emit-rtl.c (emit_copy_of_insn_after): Do not call copy_insn_1 for
4687 2007-01-01 Mike Stump <mrs@apple.com>
4689 * configure.ac (HAVE_GAS_LITERAL16): Add autoconf check for
4691 * config/darwin.c (machopic_select_rtx_section): Use
4693 (darwin_mergeable_constant_section): Likewise.
4694 * configure: Regenerate.
4695 * config.in: Regenerate.
4697 2007-01-01 Jan Hubicka <jh@suse.cz>
4698 Andrew Pinski <pinskia@gmail.com>
4700 * cgraphunit.c (cgraph_optimize): Call cgraph_add_new_functions
4701 before starting IPA passes.