OSDN Git Service

* config/i386/i386.md (fixuns_trunc<SSEMODEF>si_1): New insn.
[pf3gnuchains/gcc-fork.git] / gcc / ChangeLog
1 2007-02-12  Richard Henderson  <rth@redhat.com>
2
3         * config/i386/i386.md (fixuns_trunc<SSEMODEF>si_1): New insn.
4         (fixuns_trunc<SSEMODEF>si2): Use it.
5         * config/i386/sse.md (vec_setv4sf_0): Export.
6         * config/i386/i386.c (ix86_build_const_vector): Export.
7         (ix86_split_convert_uns_si_sse): Rename from
8         ix86_expand_convert_uns_si_sse and rewrite as a splitter.
9         * config/i386/i386-protos.h: Update.
10
11 2007-02-13  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
12
13         PR c/29521
14         * c-typeck.c (c_finish_return): Improve warning message.
15
16 2007-02-12  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
17
18         * alias.c (find_symbolic_term): Delete unused function.
19         
20 2007-02-12  Uros Bizjak  <ubizjak@gmail.com>
21
22         * config/i386/i386.md (paritydi2, paritysi2): New expanders.
23         (paritydi2_cmp, paritydi2_cmp): New insn and split patterns.
24         (*parityhi2_cmp, *parityqi2_cmp): New insn patterns.
25
26 2007-02-12  Eric Botcazou  <ebotcazou@adacore.com>
27
28         * tree.h (DECL_IGNORED_P): Document further effect for FUNCTION_DECL.
29         * cgraphunit.c (cgraph_expand_function): If DECL_IGNORED_P is set on
30         the function, temporarily point the debug interface to the null one.
31
32 2007-02-12  Eric Botcazou  <ebotcazou@adacore.com>
33
34         * dwarf2out.c (round_up_to_align): New static function.
35         (field_byte_offset): Use it to round the offset.
36
37 2007-02-12  Richard Henderson  <rth@redhat.com>
38
39         * config/alpha/alpha.md (bswapsi2, bswapdi2): New.
40         (inswl_const): Export.
41
42 2007-02-12  Richard Henderson  <rth@redhat.com>
43
44         * calls.c (emit_library_call_value_1): If PROMOTE_MODE modifed the
45         result mode of the libcall, convert back to outmode.
46
47 2007-02-12  Roger Sayle  <roger@eyesopen.com>
48
49         * config/i386/i386.md (*bswapdi2_rex): Renamed from bswapdi2.
50         (bswapdi2): New define_expand to implement 32-bit implementation.
51
52 2007-02-12  Nick Clifton  <nickc@redhat.com>
53
54         * doc/invoke.texi (Overall Options): Document --help=.
55         * gcc.c (target_help_flag): Rename to print_subprocess_flag.
56         (cc1_options): Pass --help= on to cc1.
57         (display_help): Add description of --help=.
58         (process_command): Add code to handle --help=.  Allow translated
59         --help and --target-help switches to be passed on to compiler
60         sub-process.
61         (main): Remove unused if statement.
62         * opts.c (columns): Remove.
63         (LEFT_COLUMN): Define.
64         (wrap_help): Add columns argument.
65         (print_filtered_help): Change parameters to be an include bitmask,
66         an exclude bitmask, an any bitmask and the column width.  Move the
67         code to display the params list here.  Add code to display the
68         status of options rather than their descriptions if the quiet flag
69         is not active.
70         (print_specific_help): Change parameters to be an include bitmask,
71         an exclude bitmask and an any bitmask.  Move code to look up the
72         column width here.  Decide upon the title for an options listing.
73         (common_handle_options): Add code to handle --help=.  Adapt code
74         for --help and --target-help to use the revised form of the
75         print_specific_help function.
76         (print_help): Delete.
77         (print_param_help): Delete.
78         (print_switch): Delete.
79         * opts.h (cl_lang_count): Add prototype.
80         (CL_PARAMS, CL_WARNING, CL_OPTIMIZATION, CL_MIN_OPTION_CLASS,
81         CL_MAX_OPTION_CLASS): New defines.
82         * optc-gen.awk: Add construction of cl_lang_count.
83         * c.opt: Add Warning attribute to warning options and Optimization
84         attribute to optimization options.
85         * common.opt: Likewise.
86         Add --help=.
87         Add -fhelp and -ftarget-help as aliases for the transformed --help
88         and --target-help options.
89         * opt-functions.awk: Add code to handle Warning and Optimization
90         attributes.
91
92 2007-02-12  Richard Henderson  <rth@redhat.com>
93
94         * config/alpha/constraints.md: New file.
95         * config/alpha/alpha.c: Include tm-constrs.h.
96         (alpha_const_ok_for_letter_p, alpha_const_double_ok_for_letter_p,
97         alpha_extra_constraint): Remove.
98         (alpha_emit_conditional_branch): Use satisfies_constraint_*.
99         * config/alpha/alpha-protos.h: Update.
100         * config/alpha/alpha.h (REG_CLASS_FROM_LETTER): Remove.
101         (CONST_OK_FOR_LETTER_P, CONST_DOUBLE_OK_FOR_LETTER_P): Remove.
102         (EXTRA_CONSTRAINT): Remove.
103         * config/alpha/alpha.md: Include constraints.md.
104         (adddi splitter): Use satisfies_constraint_*.
105         * config/alpha/predicates.md (add_operand): Likewise.
106         (sext_add_operand, addition_operation): Likewise.
107
108 2007-02-12  Dorit Nuzman  <dorit@il.ibm.com>
109
110         PR tree-optimization/29145
111         * tree-data-ref.c (base_addr_differ_p): Make us more conservative
112         in our handling of restrict qualified pointers.
113
114 2007-02-12  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
115
116         PR middle-end/7651
117         * doc/invoke.texi (Wunused-value): Update description.
118         (Wextra): Delete item.
119         * opts.c (set_Wextra): Don't use the value of Wextra to set the
120         value of Wunused-value.
121         * c-typeck.c (c_process_expr_stmt): Don't check extra_warnings.
122         (c_finish_stmt_expr): Don't check extra_warnings.
123         (emit_side_effect_warnings): The caller is responsible to check
124         warn_unused_value.
125         
126 2007-02-11  Roger Sayle  <roger@eyesopen.com>
127             Matt Thomas  <matt@3am-software.com>
128
129         * simplify-rtx.c (simplify_relational_operation_1): Correct typo.
130
131 2007-02-11  Roger Sayle  <roger@eyesopen.com>
132
133         * simplify-rtx.c (simplify_relational_operation_1): Optimize
134         comparisons of POPCOUNT against zero.
135         (simplify_const_relational_operation): Likewise.
136
137 2007-02-11  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
138
139         * doc/invoke.texi (Wextra): Delete outdated paragraph.
140         
141 2007-02-11  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
142
143         * dwarf2out.c (root_type): Delete unused function.
144         
145 2007-02-11  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
146
147         * genattrtab.c (contained_in_p): Delete unused function.
148         (write_expr_attr_cache): Likewise.
149
150 2007-02-11  Jan Hubicka  <jh@suse.cz>
151
152         * ipa-inline.c (cgraph_edge_badness): Add "else" missing in the
153         previous patch.
154
155 2007-02-11  Steven Bosscher  <steven@gcc.gnu.org>
156
157         * fwprop.c (try_fwprop_subst): Use set_unique_reg_note
158         to add the REG_EQ* note.
159         * see.c (see_merge_one_use_extension): Likewise.
160         * local-alloc.c (update_equiv_regs): Likewise.  Also don't
161         turn REG_EQUAL notes into REG_EQUIV notes if the target
162         register may have more than one set.
163         * function.c (assign_parm_setup_reg): Use set_unique_reg_note.
164         * gcse.c (try_replace_reg): Likewise.
165         * alias.c (init_alias_analysis): Use find_reg_equal_equiv_note.
166         * calls.c (fixup_tail_calls): Likewise.  Abort if there is
167         more than one REG_EQUIV note.
168         * reload1.c (gen_reload): Use set_unique_reg_note.
169
170 2007-02-11  Uros Bizjak  <ubizjak@gmail.com>
171
172         * config/i386/i386.c (TARGET_VECTORIZE_BUILTIN_CONVERSION): Define.
173         (ix86_builtin_conversion): New function.
174
175 2007-02-06  Mark Mitchell  <mark@codesourcery.com>
176
177         PR target/29487
178         * tree.h (DECL_REPLACEABLE_P): New macro.
179         * except.c (set_nothrow_function_flags): Likewise.
180
181 2007-02-11  Tehila Meyzels  <tehila@il.ibm.com>
182             Ira Rosen  <irar@il.ibm.com>
183             Dorit Nuzman  <dorit@il.ibm.com>
184
185         * doc/tm.texi (TARGET_VECTORIZE_BUILTIN_CONVERSION): New target hook.
186         * targhooks.c (default_builtin_vectorized_conversion): New.
187         * targhooks.h (default_builtin_vectorized_function): New declaration.
188         * target.h (struct vectorize): Add builtin_conversion field.
189         * tree-vectorizer.h (type_conversion_vec_info_type): New enum
190         stmt_vec_info_type value.
191         (vectorizable_conversion): New declaration.
192         * tree-vect-analyze.c (vect_analyze_operations): Add
193         vectorizable_conversion call.
194         * target-def.h (TARGET_VECTORIZE_BUILTIN_CONVERSION): New.
195         * tree-vect-transform.c (vectorizable_conversion): New function.
196         (vect_transform_stmt): Add case for type_conversion_vec_info_type.
197         * tree-vect-generic.c (expand_vector_operations_1): Consider correct
198         mode.
199         * config/rs6000/rs6000.c (rs6000_builtin_conversion): New.
200         (TARGET_VECTORIZE_BUILTIN_CONVERSION): Defined.
201         (rs6000_expand_builtin): Add handling a case of ALTIVEC_BUILTIN_VCFUX
202         or ALTIVEC_BUILTIN_VCFSX.
203
204 2007-02-10  John David Anglin  <dave.anglin@nrc-cnrc.gc.ca>
205
206         PR target/30634
207         * pa.md (movdf): For 64-bit target, fail if operand 1 is a non-zero
208         CONST_DOUBLE and operand 0 is a hard register.
209         (movdi): For 64-bit target, remove code to force CONST_DOUBLE to
210         memory.  Fail if operand 1 is a non-zero CONST_INT and operand 0
211         is a hard floating-point register.
212
213 2007-02-10  Richard Henderson  <rth@redhat.com>
214             Jakub Jelinek  <jakub@redhat.com>
215             Alexandre Oliva  <aoliva@redhat.com>
216
217         * Makefile.in (libgcc-support, libgcc.mvars): Add emutls.c.
218         * builtin-types.def (BT_WORD): Make unsigned.
219         (BT_FN_VOID_PTR_WORD_WORD_PTR): New.
220         * builtins.def (BUILT_IN_EMUTLS_GET_ADDRESS): New.
221         (BUILT_IN_EMUTLS_REGISTER_COMMON): New.
222         * c-decl.c (grokdeclarator): Don't error if !have_tls.
223         * c-parser.c (c_parser_omp_threadprivate): Likewise.
224         * dwarf2out.c (loc_descriptor_from_tree_1): Don't do anything for
225         emulated tls.
226         * expr.c (emutls_var_address): New.
227         (expand_expr_real_1): Expand emulated tls.
228         (expand_expr_addr_expr_1): Likewise.
229         * libgcc-std.ver: Add __emutls_get_address, __emutls_register_common.
230         * output.h (emutls_finish): Declare.
231         * toplev.c (compile_file): Call it.
232         * tree-ssa-address.c (gen_addr_rtx): Check for const-ness of the
233         address before wrapping in CONST.
234         * varasm.c (emutls_htab, emutls_object_type): New.
235         (EMUTLS_VAR_PREFIX, EMUTLS_TMPL_PREFIX): New.
236         (get_emutls_object_name, get_emutls_object_type): New.
237         (get_emutls_init_templ_addr, emutls_decl): New.
238         (emutls_common_1, emutls_finish): New.
239         (assemble_variable): When emulating tls, swap decls; generate
240         constructor for the emutls objects.
241         (do_assemble_alias): When emulating tls, swap decl and target name.
242         (default_encode_section_info): Don't add SYMBOL_FLAG_TLS_SHIFT
243         for emulated tls.
244         * varpool.c (decide_is_variable_needed): Look at force_output.
245         Recurse for emulated tls.
246         (cgraph_varpool_remove_unreferenced_decls): Remove checks redundant
247         with decide_is_variable_needed.
248         * emutls.c: New file.
249         * config/sparc/sol2.h (ASM_DECLARE_OBJECT_NAME): Only emit
250         tls_object for real tls.
251
252 2007-02-10  Kaz Kojima  <kkojima@gcc.gnu.org>
253
254         PR rtl-optimization/29599
255         * reload1.c (eliminate_regs_in_insn): Take the destination
256         mode into account when computing the offset.
257
258 2007-02-09  Stuart Hastings  <stuart@apple.com>
259             Richard Henderson  <rth@redhat.com>
260
261         * gcc/config/i386/i386.h (TARGET_KEEPS_VECTOR_ALIGNED_STACK): New.
262         * gcc/config/i386/darwin.h: (TARGET_KEEPS_VECTOR_ALIGNED_STACK): New.
263         * gcc/config/i386/i386.md (fixuns_trunc<mode>si2, fixuns_truncsfhi2,
264         fixuns_truncdfhi2): New.
265         (fix_truncsfdi_sse): Call ix86_expand_convert_sign_didf_sse.
266         (floatunsdidf2): Call ix86_expand_convert_uns_didf_sse.
267         (floatunssisf2): Add call to ix86_expand_convert_uns_sisf_sse.
268         (floatunssidf2): Allow nonimmediate source.
269         * gcc/config/i386/sse.md (movdi_to_sse): New.
270         (vec_concatv2di): Drop '*'.
271         * gcc/config/i386/i386-protos.h (ix86_expand_convert_uns_si_sse,
272         ix86_expand_convert_uns_didf_sse, ix86_expand_convert_uns_sidf_sse,
273         ix86_expand_convert_uns_sisf_sse, ix86_expand_convert_sign_didf_sse):
274         New.
275         * gcc/config/i386/i386.c (ix86_expand_convert_uns_si_sse,
276         ix86_expand_convert_uns_didf_sse, ix86_expand_convert_uns_sidf_sse,
277         ix86_expand_convert_uns_sisf_sse, ix86_expand_convert_sign_didf_sse,
278         ix86_build_const_vector, ix86_expand_vector_init_one_nonzero): New.
279         (ix86_build_signbit_mask): Fix decl of v, refactor to call
280         ix86_build_const_vector.
281         (x86_emit_floatuns): Rewrite.
282
283 2007-02-10  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
284
285         * genautomata.c (longest_path_length): Delete unused function.
286         (struct state): Delete unused longest_path_length.
287         (UNDEFINED_LONGEST_PATH_LENGTH): Delete unused macro.
288         (get_free_state): Delete unused.
289         
290 2007-02-09  Jan Hubicka  <jh@suse.cz>
291
292         * params.def (PARAM_INLINE_UNIT_GROWTH): Set to 30.
293         * doc/invoke.texi (inline-unit-growth): Update default value.
294
295         * Makefile.in (passes.o, ipa-inline.o): Add dependencies.
296         * cgraphbuild.c (build_cgraph_edges): Compute frequencies.
297         (rebuild_cgraph_edges): Likewise.
298         * cgraph.c (cgraph_set_call_stmt): Add new argument frequency.
299         (dump_cgraph_node): Dump frequencies.
300         (cgraph_clone_edge): Add frequency scales.
301         (cgraph_clone_node): Add freuqnecy.
302         * cgraph.h (cgraph_edge): Add freuqnecy argument.
303         (CGRAPH_FREQ_BASE, CGRAPH_FREQ_MAX): New constants.
304         (cgraph_create_edge, cgraph_clone_edge, cgraph_clone_node): Update.
305         * tree-pass.h (TODO_rebuild_frequencies): New constant.
306         * cgraphunit.c (verify_cgraph_node): Verify frequencies.
307         (cgraph_copy_node_for_versioning): Update call of cgraph_clone_edge.
308         (save_inline_function_body): Likewise.
309         * ipa-inline.c: inluce rtl.h
310         (cgraph_clone_inlined_nods): Update call of cgraph_clone_node.
311         (cgraph_edge_badness): Use frequencies.
312         (cgraph_decide_recursive_inlining): Update clonning.
313         (cgraph_decide_inlining_of_small_function): Dump frequency.
314         * predict.c (estimate_bb_frequencies): Export.
315         * predict.h (estimate_bb_frequencies): Declare.
316         * tree-inline.c (copy_bb): Watch overflows.
317         (expand_call_inline): Update call of cgraph_create_edge.
318         (optimize_inline_calls): Use TODO flags to update frequnecies.
319         * passes.h: Include predict.h
320         (init_optimization_passes): Move profile ahead.
321         (execute_function_todo): Handle TODO_rebuild_frequencies.
322
323 2007-02-09  Roger Sayle  <roger@eyesopen.com>
324
325         * config/alpha/alpha.c (emit_insxl): Force the first operand of
326         the insbl or inswl pattern into a register.
327
328 2007-02-09  Roger Sayle  <roger@eyesopen.com>
329
330         * config/ia64/ia64.md (bswapdi2): New define_insn.
331
332 2007-02-09  Richard Henderson  <rth@redhat.com>
333
334         * config/i386/constraints.md (Ym): New constraint.
335         * config/i386/i386.md (movsi_1): Change Y2 to Yi constraints.
336         (movdi_1_rex64): Split sse and xmm general register moves from
337         memory move alternatives.  Use conditional register constraints.
338         (movsf_1, movdf_integer): Likewise.
339         (zero_extendsidi2_32, zero_extendsidi2_rex64): Likewise.
340         (movdf_integer_rex64): New.
341         (pushsf_rex64): Fix output constraints.
342         * config/i386/sse.md (sse2_loadld): Split rm alternative, use Yi.
343         (sse2_stored): Likewise.
344         (sse2_storeq_rex64): New.
345         * config/i386/i386.c (x86_inter_unit_moves): Enable for not
346         amd and not generic.
347         (ix86_secondary_memory_needed): Don't bypass TARGET_INTER_UNIT_MOVES
348         for optimize_size.  Remove SF/DFmode hack.
349
350 2007-02-09  Dwarakanath Rajagopal  <dwarak.rajagopal@amd.com>
351
352         * config/i386/driver-i386.c: Turn on -mtune=native for AMDFAM10.
353         (bit_SSE4a): New.
354
355 2007-02-09  Nathan Sidwell  <nathan@codesourcery.com>
356             Richard Sandiford  <richard@codesourcery.com>
357
358         * config.gcc (m68010-*-netbsdelf*, m68k*-*-netbsdelf*)
359         (m68k*-*-openbsd*, m68k-*-linux*): Set default_cf_cpu.
360         (m68k-*-aout*, m68k-*-coff*, m68k-*-uclinux*, m68k-*-rtems*): Add
361         m68k/t-mlib to tmake_file.
362         (m68020-*-elf*, m68k-*-elf*): Likewise.  Add t-m68kbare as well.
363         (m68k*-*-*): Use --with-arch to pick a default for --with-cpu.
364         (m680[012]0-*-*, m68k*-*-*): Add support for --with-arch.
365         Allow it to be cf or m68k.  Set m68k_arch_family.  If that
366         variable is not empty, add t-$m68k_arch_family to tmake_file.
367         Add t-mlibs to tmake_file.
368         * doc/install.texi: Document --with-arch=m68k and --with-arch=cf.
369         * config/m68k/t-cf: New file.
370         * config/m68k/t-m68k: Likewise.
371         * config/m68k/t-mlibs: Likewise.
372         * config/m68k/t-m68kbare (MULTILIB_OPTIONS, MULTILIB_DIRNAMES)
373         (MULTILIB_MATCHES, MULTILIB_EXCEPTIONS): Delete.
374         (M68K_MLIB_DIRNAMES, M68K_MLIB_OPTIONS): Define.
375         * config/m68k/t-m68kelf (MULTILIB_OPTIONS, MULTILIB_DIRNAMES)
376         (MULTILIB_MATCHES, MULTILIB_EXCEPTIONS, LIBGCC, INSTALL_LIBGCC):
377         Delete.
378         * config/m68k/t-openbsd (MULTILIB_OPTIONS, LIBGCC): Delete.
379         (INSTALL_LIBGCC): Delete.
380         (M68K_MLIB_DIRNAMES, M68K_MLIB_OPTIONS): Define.
381         * config/m68k/t-rtems (MULTILIB_OPTIONS, MULTILIB_DIRNAMES)
382         (MULTILIB_MATCHES, MULTILIB_EXCEPTIONS): Delete.
383         (M68K_MLIB_CPU): Define.
384         * config/m68k/t-uclinux (MULTILIB_OPTIONS, MULTILIB_DIRNAMES)
385         (MULTILIB_MATCHES, MULTILIB_EXCEPTIONS): Delete.
386         (M68K_MLIB_CPU, M68K_MLIB_OPTIONS, M68K_MLIB_DIRNAMES): Define.
387
388 2007-02-09  Zdenek Dvorak  <dvorakz@suse.cz>
389             Richard Guenther  <rguenther@suse.de>
390
391         PR middle-end/23361
392         * fold-const.c (fold_comparison): Handle obfuscated comparisons
393         against INT_MIN/INT_MAX.
394         * tree-ssa-loop-ivcanon.c (remove_empty_loop): Print to dump
395         file if a loop is removed.
396
397 2007-02-09  Joseph Myers  <joseph@codesourcery.com>
398
399         * calls.c (store_one_arg): Pass correct alignment to
400         emit_push_insn for non-BLKmode values.
401         * expr.c (emit_push_insn): If STRICT_ALIGNMENT, copy to an
402         unaligned stack slot via a suitably aligned slot.
403
404 2007-02-08  DJ Delorie  <dj@redhat.com>
405
406         * config/m32c/m32c.c (m32c_unpend_compare): Add default to silence
407         warnings.
408         (legal_subregs): Use unsigned char, make const.
409         (m32c_illegal_subreg_p): Use ARRAY_SIZE.  Delete unused variables.
410
411 2007-02-08  Paul Brook  <paul@codesourcery.com>
412
413         * config/arm/lib1funcs.asm (RETLDM): Pop directly into PC when no
414         special interworking needed.
415
416 2007-02-08  Harsha Jagasia  <harsha.jagasia@amd.com>
417
418         * config/i386/xmmintrin.h: Make inclusion of emmintrin.h
419         conditional to __SSE2__.
420         (Entries below should have been added to first ChangeLog
421         entry for amdfam10 dated 2007-02-05)
422         * config/i386/emmintrin.h: Generate #error if __SSE2__ is not
423         defined.
424         * config/i386/pmmintrin.h: Generate #error if __SSE3__ is not
425         defined.
426         * config/i386/tmmintrin.h: Generate #error if __SSSE3__ is not
427         defined.
428
429 2007-02-08  DJ Delorie  <dj@redhat.com>
430
431         * config/m32c/m32c-protos.h (m32c_illegal_subreg_p): New.
432         * config/m32c/m32c.c (legal_subregs): New.
433         (m32c_illegal_subreg_p): New.
434         * config/m32c/predicates.md (m32c_any_operand): Use it to reject
435         unsupported subregs of hard regs.
436
437 2007-02-08  Jan Hubicka  <jh@suse.cz>
438
439         * tree-cfg.c (bsi_replace): Shortcut when replacing the statement with
440         the same one; always update histograms.
441
442 2007-02-08  Diego Novillo  <dnovillo@redhat.com>
443
444         * passes.c (init_optimization_passes): Tidy comment.
445
446 2007-02-08  Roger Sayle  <roger@eyesopen.com>
447
448         * simplify-rtx.c (simplify_unary_operation_1) <POPCOUNT>: We can
449         strip zero_extend, bswap and rotates from POCOUNT's argument.
450         <PARITY>: Likewise, we can strip not, bswap, sign_extend,
451         zero_extend and rotates from PARITY's argument.
452         <BSWAP>: A byte-swap followed by a byte-swap is an identity.
453         (simplify_const_unary_operation) <BSWAP>: Evaluate the byte-swap
454         of an integer constant at compile-time.
455
456 2007-02-08  Diego Novillo  <dnovillo@redhat.com>
457
458         PR 30562
459         * tree-flow.h (struct var_ann_d): Remove field 'is_used'.
460         Update all users.
461         * tree-ssa-alias.c (compute_is_aliased): Remove.  Update all
462         users.
463         (init_alias_info):
464         * tree-ssa-live.c (remove_unused_locals): Do not remove
465         TREE_ADDRESSABLE variables.
466         * tree-ssa-structalias.c (compute_points_to_sets): Tidy.
467         * tree-ssa-operands.c (add_virtual_operand): Remove argument
468         FOR_CLOBBER.  Update all users.
469         If VAR has an associated alias set, add a virtual operand for
470         it if no alias is found to conflict with the memory reference.
471
472 2007-02-07  Jan Hubicka  <jh@suse.cz>
473             Robert Kidd <rkidd@crhc.uiuc.edu>
474
475         * value-prof.c (visit_hist, free_hist): Return 1 instead of 0.
476
477 2007-02-07  Ian Lance Taylor  <iant@google.com>
478
479         * lower-subreg.c (simple_move): Reject PARTIAL_INT modes.
480
481 2007-02-07  Roger Sayle  <roger@eyesopen.com>
482
483         * config/rs6000/rs6000.md (ctz<mode>2, ffs<mode>2, popcount<mode>2,
484         parity<mode>2, smulsi3_highpart, abstf2_internal, allocate_stack,
485         tablejumpdi, movsi_to_cr_one): Remove constraints from
486         define_expand's match_operands.
487
488 2007-02-07  Roger Sayle  <roger@eyesopen.com>
489
490         * global.c (compute_regsets): Move declatation of "i" inside of
491         #ifdef ELIMINABLE_REGS to avoid unused variable bootstrap failure.
492
493 2007-02-07  Jakub Jelinek  <jakub@redhat.com>
494
495         PR c++/30703
496         * gimplify.c (gimplify_scan_omp_clauses): Remove special casing
497         of INDIRECT_REF <RESULT_DECL>.
498
499         * config/i386/i386.c (override_options): Set PTA_SSSE3 for core2.
500
501 2007-02-06  J"orn Rennecke  <joern.rennecke@arc.com>
502             Kaz Kojima  <kkojima@gcc.gnu.org>
503
504         PR target/29746
505         * config/sh/sh.c (expand_cbranchdi4): Use scratch register
506         properly.
507         (sh_initialize_trampoline): Add parentheses to avoid a warning.
508
509 2007-02-06  Zdenek Dvorak <dvorakz@suse.cz>
510
511         * doc/loop.texi: Document possibility not to perform disambiguation
512         of loops with multiple latches.
513         * cfgloopmanip.c (alp_enum_p): Removed.
514         (add_loop): Handle subloops.  Use get_loop_body_with_size.
515         (create_preheader): Do not allow ENTRY_BLOCK_PTR to be preheader.
516         * cfghooks.c (redirect_edge_and_branch_force): Set dominator for
517         the new forwarder block.
518         (make_forwarder_block): Only call new_bb_cbk if it is not NULL.
519         Handle the case latch is NULL.
520         * tree-ssa-dom.c (tree_ssa_dominator_optimize): Avoid cfg modifications
521         when marking loop exits.
522         * ifcvt.c (if_convert): Ditto.  Mark loop exits even if cfg cannot
523         be modified.
524         * loop-init.c (loop_optimizer_init): Do not modify cfg.  Call
525         disambiguate_loops_with_multiple_latches.
526         * tree-cfgcleanup.c (cleanup_tree_cfg_loop): Calculate dominators
527         before fix_loop_structure.
528         * cfgloop.c: Include pointer-set.h and output.h.
529         (canonicalize_loop_headers, HEADER_BLOCK, LATCH_EDGE,
530         update_latch_info, mfb_keep_just, mfb_keep_nonlatch): Removed.
531         (get_loop_latch_edges, find_subloop_latch_edge_by_profile,
532         find_subloop_latch_edge_by_ivs, find_subloop_latch_edge,
533         mfb_redirect_edges_in_set, form_subloop, merge_latch_edges,
534         disambiguate_multiple_latches, get_loop_body_with_size,
535         disambiguate_loops_with_multiple_latches): New functions.
536         (flow_loop_dump): Dump multiple latch edges.
537         (flow_loop_nodes_find): Handle loops with multiple latches.
538         (flow_loops_find): Ditto. Do not call canonicalize_loop_headers.
539         (glb_enum_p): Modified.
540         (get_loop_body): Use get_loop_body_with_size.
541         * cfgloop.h (LOOPS_HAVE_RECORDED_EXITS): New flag.
542         (AVOID_CFG_MODIFICATIONS): New constant.
543         (disambiguate_loops_with_multiple_latches, add_loop,
544         get_loop_body_with_size): Declare.
545         * Makefile.in (cfgloop.o): Add pointer-set.h and output.h.
546
547 2007-02-06  Seongbae Park <seongbae.park@gmail.com>
548
549         PR inline-asm/28686
550         * global.c (compute_regsets): New function.
551         (global_alloc): Refactored ELIMINABLE_REGSET
552         and NO_GLOBAL_ALLOC_REGS computation out.
553         (rest_of_handle_global_alloc): Call compute_regsets()
554         for non-optimizing case.
555
556 2007-02-06  Richard Henderson  <rth@redhat.com>
557
558         * config/i386/constraints.md (Y2): Rename from Y.
559         (Yi): New constraint.
560         * config/i386/i386.md (movsi_1, movdi_2, pushdf_nointeger,
561         pushdf_integer, movdf_nointeger, movdf_integer, zero_extendsidi2_32,
562         zero_extendsidi2_rex64, truncxfdf2_mixed): Change Y constraints to Y2.
563         (extendsfdf2_mixed, extendsfdf2_sse, truncdfsf_fast_mixed,
564         truncdfsf_fast_sse, truncdfsf_mixed, fix_truncdfdi_sse,
565         fix_truncdfsi_sse, floatsidf2_mixed, floatsidf2_sse,
566         floatdidf2_mixed, floatdidf2_sse, absnegdf2_mixed, 
567         absnegdf2_sse, sse_setccdf, fop_df_comm_mixed, fop_df_comm_sse,
568         fop_df_1_mixed, fop_df_1_sse): Change Y constraints to x.
569         * config/i386/mmx.md (mov<MMXMODEI>_internal_rex64,
570         mov<MMXMODEI>_internal, movv2sf_internal_rex64, movv2sf_internal,
571         vec_extractv2si_1): Change Y constraints to Y2.
572         * config/i386/sse.md (vec_setv4sf_0, vec_concatv2df, vec_dupv4si,
573         vec_dupv2di, sse2_concatv2si, vec_concatv4si_1, vec_concatv2di):
574         Change Y constraints to Y2.
575         (sse2_loadld): Change Y constraints to x.
576
577 2007-02-06  Roger Sayle  <roger@eyesopen.com>
578
579         * config/rs6000/rs6000.md (popcount<mode>2): Rewrite.
580         (parity<mode>2): New define_expand using rs6000_emit_parity.
581         * config/rs6000/rs6000.c (rs6000_emit_popcount,
582         rs6000_emit_parity): New functions.
583         * config/rs6000/rs6000-protos.h (rs6000_emit_popcount,
584         rs6000_emit_parity): Prototype here.
585
586 2007-02-06  Ian Lance Taylor  <iant@google.com>
587
588         * lower-subreg.c (simple_move_operand): Reject CONST.
589         (resolve_clobber): Call validate_change rather than directly
590         assigning to XEXP (pat, 0).
591
592 2006-02-06  Paolo Bonzini  <bonzini@gnu.org>
593
594         * Makefile.in (tree-ssa-loop-ivopts.o): Add pointer-set.h dependency.
595         (tree-ssa-reassoc.o): Add pointer-set.h dependency.
596         (tree-cfg.o): Remove hashtab.h dependency.
597
598         * tree-ssa-loop-ivopts.c: Include pointer-set.h.
599         (struct ivopts_data): Change niters to pointer_map_t.
600         (struct nfe_cache_elt, nfe_hash, nfe_eq): Delete.
601         (niter_for_exit): Create pointer_map on demand.  Change for
602         pointer_map API.
603         (tree_ssa_iv_optimize_init): Initialize data->niters to NULL.
604         (free_loop_data): Destroy data->niters if created and reset field.
605         (tree_ssa_iv_optimize_finalize): Don't delete data->niters here.
606         (tree_ssa_iv_optimize_loop): Check for presence of stale data.
607
608         * tree-ssa-reassoc.c: Include pointer-set.h.
609         (bb_rank): Change to long *.
610         (operand_rank): Change to pointer_map_t.
611         (find_operand_rank): Return long, -1 if not found.  Declare as inline.
612         (insert_operand_rank): Accept long.
613         (operand_entry_hash, operand_entry_eq): Remove.
614         (get_rank): Return long.  Adjust for changes above.
615         (init_reassoc): Change rank type to long.  Adjust creation of bb_rank
616         and operand_rank.
617         (fini_reassoc): Delete operand_rank with pointer_map_destroy.
618
619         * tree-ssa-structalias.c (vi_for_tree): Change to pointer_map.
620         (struct tree_vi, tree_vi_t, tree_vi_hash, tree_vi_eq): Delete.
621         (insert_vi_for_tree): Rewrite for pointer_map API.  Assert argument
622         is not NULL.
623         (lookup_vi_for_tree): Rewrite for pointer_map API.  Return varinfo_t
624         directly since it cannot be NULL.
625         (get_vi_for_tree): Rewrite for pointer_map API.
626         (find_what_p_points_to): Adjust for change to lookup_vi_for_tree.
627         (init_alias_vars): Create vi_for_tree as pointer_map.
628         (delete_points_to_sets): Delete vi_for_tree using pointer_map_destroy.
629
630         * tree-cfg.c: Don't include hashtab.h.
631         (edge_to_cases): Declare as pointer_map.
632         (struct edge_to_cases_elt, edge_to_cases_hash, edge_to_cases_eq):
633         Delete.
634         (edge_to_cases_cleanup): Rewrite as pointer_map_traverse callback.
635         (start_recording_case_labels): Create edge_to_cases as pointer_map.
636         (end_recoding_case_labels): Cleanup edge_to_cases manually before
637         destroying it.
638         (record_switch_edge): Delete.
639         (get_cases_for_edge): Adjust for pointer_map API, inline
640         record_switch_edge (rewritten for new API), remove goto.
641
642 2006-02-06  Paolo Bonzini  <bonzini@gnu.org>
643
644         * Makefile.in (tree-nested.o): Add pointer-set.h dependency.
645         * tree-nested.c: Include pointer-set.h.
646         (var_map_elt, var_map_eq, var_map_hash): Delete.
647         (struct nesting_info): Remove GTY marker.  Change the two htab_t's
648         to pointer_map_t's.
649         (nesting_info_bitmap_obstack): New.
650         (lookup_field_for_decl): Adjust for pointer_map API.
651         (lookup_tramp_for_decl): Adjust for pointer_map API.
652         (get_nonlocal_debug_decl): Adjust for pointer_map API.
653         (get_local_debug_decl): Adjust for pointer_map API.
654         (convert_nl_goto_reference): Adjust for pointer_map API.
655         (convert_nl_goto_receiver): Adjust for pointer_map API.
656         (create_nesting_tree): Create outside GGC space.  Create bitmap on
657         the new obstack.  Create field_map and var_map as pointer_maps.
658         (free_nesting_tree): Adjust for changes to create_nesting_tree.
659         (root): Delete. 
660         (lower_nested_functions): Move root here, no need to NULL it.
661         Initialize and release the obstack.
662
663 2007-02-06  Paolo Bonzini  <bonzini@gnu.org>
664
665         * tree.c (tree_int_map_hash, tree_int_map_eq, tree_int_map_marked_p):
666         Remove prototypes and make them non-static.
667         (struct tree_int_map): Remove.
668         * tree.h (struct tree_int_map): Move here, turning TO into an
669         unsigned int.
670         (tree_int_map_hash, tree_int_map_eq, tree_int_map_marked_p): Declare.
671
672         * tree.h (TREE_COMPLEXITY): Remove.
673         (struct tree_exp): Remove complexity field.
674         * tree.c (build1_stat): Don't set it.
675
676 2007-02-06  Dorit Nuzman  <dorit@il.ibm.com>
677             Victor Kaplansky  <victork@il.ibm.com>
678
679         * tree-vectorizer.c (vect_is_simple_use): Support induction.
680         (vect_is_simple_reduction): Support reduction with induction as
681         one of the operands.
682         (vect_is_simple_iv_evolution): Fix formatting.
683         * tree-vect-analyze.c (vect_mark_stmts_to_be_vectorized): Fix 
684         formatting.  Don't mark induction phis for vectorization.
685         (vect_analyze_scalar_cycles): Analyze all inductions, then reductions.
686         * tree-vect-transform.c (get_initial_def_for_induction): New function.
687         (vect_get_vec_def_for_operand): Support induction.
688         (vect_get_vec_def_for_stmt_copy): Fix formatting and add check for
689         induction case.
690         (vectorizable_reduction): Support reduction with induction as one of 
691         the operands. 
692         (vectorizable_type_demotion): Use def-type of stmt argument rather
693         than dummy def-type.
694
695         * tree-ssa-loop.c (gate_scev_const_prop): Return the value of
696         flag_tree_scev_cprop.
697         * common.opt (tree-scev-cprop): New flag.
698
699         * tree-vect-transform.c (vect_create_destination_var): Use 'kind' in
700         call to vect_get_new_vect_var.
701
702 2007-02-06  Ira Rosen  <irar@il.ibm.com>
703
704         * tree-vect-patterns.c (vect_recog_widen_mult_pattern): Check that 
705         vectype is not NULL.
706         (vect_pattern_recog_1): Likewise.
707
708 2007-02-05  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
709
710         * fold-const.c (negate_expr_p): Handle CONJ_EXPR.
711         (fold_negate_expr): Likewise.
712
713 2007-02-05  Alexandre Oliva  <aoliva@redhat.com>
714
715         PR debug/30189
716         * dwarf2out.c (modified_type_die): Follow DECL_ORIGINAL_TYPE
717         even if cv-qualification is the same.
718
719 2007-02-05  Geoffrey Keating  <geoffk@apple.com>
720
721         * config/rs6000/darwin-tramp.asm (__trampoline_setup): Call
722         __enable_execute_stack on completion.
723
724 2007-02-05  Harsha Jagasia  <harsha.jagasia@amd.com>
725
726         * config/i386/athlon.md (athlon_fldxf_k8, athlon_fld_k8,
727         athlon_fstxf_k8, athlon_fst_k8, athlon_fist, athlon_fmov,
728         athlon_fadd_load, athlon_fadd_load_k8, athlon_fadd, athlon_fmul,
729         athlon_fmul_load, athlon_fmul_load_k8, athlon_fsgn,
730         athlon_fdiv_load, athlon_fdiv_load_k8, athlon_fdiv_k8,
731         athlon_fpspc_load, athlon_fpspc, athlon_fcmov_load,
732         athlon_fcmov_load_k8, athlon_fcmov_k8, athlon_fcomi_load_k8,
733         athlon_fcomi, athlon_fcom_load_k8, athlon_fcom): Added amdfam10.
734
735 2007-02-05  Harsha Jagasia  <harsha.jagasia@amd.com>
736
737         * config/i386/i386.md (x86_sahf_1, cmpfp_i_mixed, cmpfp_i_sse,
738         cmpfp_i_i387, cmpfp_iu_mixed, cmpfp_iu_sse, cmpfp_iu_387,
739         swapsi, swaphi_1, swapqi_1, swapdi_rex64, fix_truncsfdi_sse,
740         fix_truncdfdi_sse, fix_truncsfsi_sse, fix_truncdfsi_sse,
741         x86_fldcw_1, floatsisf2_mixed, floatsisf2_sse, floatdisf2_mixed,
742         floatdisf2_sse, floatsidf2_mixed, floatsidf2_sse,
743         floatdidf2_mixed, floatdidf2_sse, muldi3_1_rex64, mulsi3_1,
744         mulsi3_1_zext, mulhi3_1, mulqi3_1, umulqihi3_1, mulqihi3_insn,
745         umulditi3_insn, umulsidi3_insn, mulditi3_insn, mulsidi3_insn,
746         umuldi3_highpart_rex64, umulsi3_highpart_insn,
747         umulsi3_highpart_zext, smuldi3_highpart_rex64,
748         smulsi3_highpart_insn, smulsi3_highpart_zext, x86_64_shld,
749         x86_shld_1, x86_64_shrd, sqrtsf2_mixed, sqrtsf2_sse,
750         sqrtsf2_i387, sqrtdf2_mixed, sqrtdf2_sse, sqrtdf2_i387,
751         sqrtextendsfdf2_i387, sqrtxf2, sqrtextendsfxf2_i387,
752         sqrtextenddfxf2_i387): Added amdfam10_decode.
753         
754         * config/i386/athlon.md (athlon_idirect_amdfam10,
755         athlon_ivector_amdfam10, athlon_idirect_load_amdfam10,
756         athlon_ivector_load_amdfam10, athlon_idirect_both_amdfam10,
757         athlon_ivector_both_amdfam10, athlon_idirect_store_amdfam10,
758         athlon_ivector_store_amdfam10): New define_insn_reservation.
759         (athlon_idirect_loadmov, athlon_idirect_movstore): Added
760         amdfam10.
761
762 2007-02-05  Harsha Jagasia  <harsha.jagasia@amd.com>
763
764         * config/i386/athlon.md (athlon_call_amdfam10,
765         athlon_pop_amdfam10, athlon_lea_amdfam10): New
766         define_insn_reservation.
767         (athlon_branch, athlon_push, athlon_leave_k8, athlon_imul_k8,
768         athlon_imul_k8_DI, athlon_imul_mem_k8, athlon_imul_mem_k8_DI,
769         athlon_idiv, athlon_idiv_mem, athlon_str): Added amdfam10.
770
771 2007-02-05  Harsha Jagasia  <harsha.jagasia@amd.com>
772
773         * config/i386/athlon.md (athlon_sseld_amdfam10,
774         athlon_mmxld_amdfam10, athlon_ssest_amdfam10,
775         athlon_mmxssest_short_amdfam10): New define_insn_reservation.
776
777 2007-02-05  Harsha Jagasia  <harsha.jagasia@amd.com>
778
779         * config/i386/athlon.md (athlon_sseins_amdfam10): New
780         define_insn_reservation.
781         * config/i386/i386.md (sseins): Added sseins to define_attr type
782         and define_attr unit.
783         * config/i386/sse.md: Set type attribute to sseins for insertq
784         and insertqi.
785
786 2007-02-05  Harsha Jagasia  <harsha.jagasia@amd.com>
787
788         * config/i386/athlon.md (sselog_load_amdfam10, sselog_amdfam10,
789         ssecmpvector_load_amdfam10, ssecmpvector_amdfam10,
790         ssecomi_load_amdfam10, ssecomi_amdfam10,
791         sseaddvector_load_amdfam10, sseaddvector_amdfam10): New
792         define_insn_reservation.
793         (ssecmp_load_k8, ssecmp, sseadd_load_k8, seadd): Added amdfam10.
794
795 2007-02-05  Harsha Jagasia  <harsha.jagasia@amd.com>
796
797         * config/i386/athlon.md (cvtss2sd_load_amdfam10,
798         cvtss2sd_amdfam10, cvtps2pd_load_amdfam10, cvtps2pd_amdfam10,
799         cvtsi2sd_load_amdfam10, cvtsi2ss_load_amdfam10,
800         cvtsi2sd_amdfam10, cvtsi2ss_amdfam10, cvtsd2ss_load_amdfam10,
801         cvtsd2ss_amdfam10, cvtpd2ps_load_amdfam10, cvtpd2ps_amdfam10,
802         cvtsX2si_load_amdfam10, cvtsX2si_amdfam10): New 
803         define_insn_reservation.
804
805         * config/i386/sse.md (cvtsi2ss, cvtsi2ssq, cvtss2si,
806         cvtss2siq, cvttss2si, cvttss2siq, cvtsi2sd, cvtsi2sdq,
807         cvtsd2si, cvtsd2siq, cvttsd2si, cvttsd2siq,
808         cvtpd2dq, cvttpd2dq, cvtsd2ss, cvtss2sd,
809         cvtpd2ps, cvtps2pd): Added amdfam10_decode attribute.
810
811 2007-02-05  Harsha Jagasia  <harsha.jagasia@amd.com>
812
813         * config/i386/athlon.md (athlon_ssedivvector_amdfam10,
814         athlon_ssedivvector_load_amdfam10, athlon_ssemulvector_amdfam10,
815         athlon_ssemulvector_load_amdfam10): New define_insn_reservation.
816         (athlon_ssediv, athlon_ssediv_load_k8, athlon_ssemul,
817         athlon_ssemul_load_k8): Added amdfam10.
818
819 2007-02-05  Harsha Jagasia  <harsha.jagasia@amd.com>
820
821         * config/i386/i386.h (TARGET_SSE_UNALIGNED_MOVE_OPTIMAL): New macro.
822         (x86_sse_unaligned_move_optimal): New variable.
823         
824         * config/i386/i386.c (x86_sse_unaligned_move_optimal): Enable for  
825         m_AMDFAM10.
826         (ix86_expand_vector_move_misalign): Add code to generate movupd/movups
827         for unaligned vector SSE double/single precision loads for AMDFAM10.
828
829 2007-02-05  Harsha Jagasia  <harsha.jagasia@amd.com>
830
831         * config/i386/i386.h (TARGET_AMDFAM10): New macro.
832         (TARGET_CPU_CPP_BUILTINS): Add code for amdfam10.
833         Define TARGET_CPU_DEFAULT_amdfam10.
834         (TARGET_CPU_DEFAULT_NAMES): Add amdfam10.
835         (processor_type): Add PROCESSOR_AMDFAM10.       
836         
837         * config/i386/i386.md: Add amdfam10 as a new cpu attribute to match
838         processor_type in config/i386/i386.h.
839         Enable imul peepholes for TARGET_AMDFAM10.
840         
841         * config.gcc: Add support for --with-cpu option for amdfam10.
842         
843         * config/i386/i386.c (amdfam10_cost): New variable.
844         (m_AMDFAM10): New macro.
845         (m_ATHLON_K8_AMDFAM10): New macro.
846         (x86_use_leave, x86_push_memory, x86_movx, x86_unroll_strlen,
847         x86_cmove, x86_3dnow_a, x86_deep_branch, x86_use_simode_fiop,
848         x86_promote_QImode, x86_integer_DFmode_moves,
849         x86_partial_reg_dependency, x86_memory_mismatch_stall, 
850         x86_accumulate_outgoing_args, x86_arch_always_fancy_math_387,
851         x86_sse_partial_reg_dependency, x86_sse_typeless_stores,
852         x86_use_ffreep, x86_use_incdec, x86_four_jump_limit,
853         x86_schedule, x86_use_bt, x86_cmpxchg16b, x86_pad_returns):
854         Enable/disable for amdfam10.
855         (override_options): Add amdfam10_cost to processor_target_table.
856         Set up PROCESSOR_AMDFAM10 for amdfam10 entry in 
857         processor_alias_table.
858         (ix86_issue_rate): Add PROCESSOR_AMDFAM10.
859         (ix86_adjust_cost): Add code for amdfam10.
860
861 2007-02-05  Harsha Jagasia  <harsha.jagasia@amd.com>
862         
863         * config/i386/i386.opt: Add new Advanced Bit Manipulation (-mabm)
864         instruction set feature flag. Add new (-mpopcnt) flag for popcnt 
865         instruction. Add new SSE4A (-msse4a) instruction set feature flag.
866         * config/i386/i386.h: Add builtin definition for SSE4A.
867         * config/i386/i386.md: Add support for ABM instructions 
868         (popcnt and lzcnt).
869         * config/i386/sse.md: Add support for SSE4A instructions
870         (movntss, movntsd, extrq, insertq).
871         * config/i386/i386.c: Add support for ABM and SSE4A builtins.
872         Add -march=amdfam10 flag.
873         * config/i386/ammintrin.h: Add support for SSE4A intrinsics.
874         * doc/invoke.texi: Add documentation on flags for sse4a, abm, popcnt
875         and amdfam10.
876         * doc/extend.texi: Add documentation for SSE4A builtins.
877
878 2007-02-05  Bob Wilson  <bob.wilson@acm.org>
879
880         * config/xtensa/xtensa.c (constantpool_mem_p): Skip over SUBREGs.
881
882 2007-02-05  Richard Guenther  <rguenther@suse.de>
883
884         * tree-vectorizer.h (vectorizable_function): Add argument type
885         argument, change return type.
886         * tree-vect-patterns.c (vect_recog_pow_pattern): Adjust caller.
887         * tree-vect-transform.c (vectorizable_function): Handle extra
888         argument, return vectorized function decl.
889         (build_vectorized_function_call): Remove.
890         (vectorizable_call): Handle calls with result and argument types
891         differing.  Handle loop vectorization factor correctly.
892         * targhooks.c (default_builtin_vectorized_function): Adjust for
893         extra argument.
894         * targhooks.h (default_builtin_vectorized_function): Likewise.
895         * target.h (builtin_vectorized_function): Add argument type
896         argument.
897         * config/i386/i386.c (ix86_builtin_vectorized_function): Handle
898         extra argument, allow vectorizing of lrintf.
899         * doc/tm.texi (TARGET_VECTORIZE_BUILTIN_VECTORIZED_FUNCTION): Adjust
900         documentation of target hook.
901
902 2007-02-05  Hans-Peter Nilsson  <hp@axis.com>
903
904         PR target/30665
905         * config/cris/cris.md ("*andsi_movu", "*andsi_clear", "*andhi_movu")
906         ("*andhi_clear", andu (casesi+45)): For size-changed operand where
907         memory is allowed, require !side_effects_p, not just !MEM_VOLATILE_P.
908
909 2007-02-05  Roger Sayle  <roger@eyesopen.com>
910
911         * fold-const.c (fold_unary) <REAL_PART>: Test for availability of
912         BUILT_IN_COS before simplifying REAL_PART(CEXPI)) to COS.
913         <IMAG_PART>: Likewise, check for availability of BUILT_IN_SIN.
914         * builtins.c (fold_builtin_sincos): Check for TARGET_C99_FUNCTIONS
915         before canonicalizing sincos to cexpi.
916         (fold_builtin_cexp): Likewise, for canonicalizing cexp to cexpi.
917
918 2007-02-05  Roger Sayle  <roger@eyesopen.com>
919
920         * config/alpha/alpha.c (alpha_add_builtins): New Helper function.
921         Set TREE_READONLY and TREE_NOTHROW directly, not via attributes.
922         (alpha_init_builtins): Use alpha_add_builtins to process tables.
923
924 2007-02-05  Roger Sayle  <roger@eyesopen.com>
925
926         * mips-tfile.c (initialize_init_file): Correct endianness test.
927
928 2007-02-05  Kazu Hirata  <kazu@codesourcery.com>
929
930         * config/m68k/m68k.md (pushdi-1, pushdi, movsi+1): Don't use
931         the 'y' constraint.
932
933 2007-02-05  Richard Sandiford  <richard@codesourcery.com>
934
935         * dwarf2out.c (dwarf2out_frame_debug_expr): Record the register
936         saves in a PARALLEL before the register assignments.
937
938 2007-02-05  Richard Sandiford  <richard@codesourcery.com>
939
940         * doc/tm.texi (DWARF_ALT_FRAME_RETURN_COLUMN): Do not require
941         DWARF_FRAME_RETURN_COLUMN to be a general register.
942         * dwarf2out.c (init_return_column_size): New function, split from...
943         (expand_builtin_init_dwarf_reg_sizes): ...here.  Allow both
944         DWARF_FRAME_RETURN_COLUMN and DWARF_ALT_FRAME_RETURN_COLUMN
945         to be nongeneral registers.
946         * config/m68k/m68k.h (DWARF_FRAME_REGNUM): Only map FP and
947         integer registers.
948         (DWARF_FRAME_REGISTERS, DWARF_FRAME_RETURN_COLUMN): Define.
949         (DWARF_ALT_FRAME_RETURN_COLUMN): Define.
950
951 2007-02-04  Zdenek Dvorak <dvorakz@suse.cz>
952
953         * cfgcleanup.c (try_optimize_cfg): Avoid removing ENTRY_BLOCK_PTR.
954
955 2007-02-04  Zdenek Dvorak <dvorakz@suse.cz>
956
957         * cfgloopmanip.c (loop_delete_branch_edge): Removed.
958         (remove_path): Use can_remove_branch_p and remove_branch instead
959         of loop_delete_branch_edge.
960         * tree-ssa-loop-manip.c (scale_dominated_blocks_in_loop): New function.
961         (tree_transform_and_unroll_loop): Remove dead branches immediately.
962         Update profile using scale_dominated_blocks_in_loop.
963         * cfghooks.c (can_remove_branch_p, remove_branch): New functions.
964         * cfghooks.h (struct cfg_hooks): Add can_remove_branch_p.
965         (can_remove_branch_p, remove_branch): Declare.
966         * tree-cfg.c (tree_can_remove_branch_p): New function.
967         (tree_cfg_hooks): Add tree_can_remove_branch_p.
968         * cfgrtl.c (rtl_can_remove_branch_p): New function.
969         (rtl_cfg_hooks, cfg_layout_rtl_cfg_hook): Add rtl_can_remove_branch_p.
970
971 2007-02-05  Jan Hubicka  <jh@suse.cz>
972
973         PR middle-end/30696
974         * ipa-inline.c (cgraph_clone_inlined_nodes): When there are unanalyzed
975         nodes in cgraph, don't remove offline copy of the function.
976
977 2007-02-04  Jan Hubicka  <jh@suse.cz>
978
979         * tree-sra.c (sra_walk_expr): Add linebreaks. BITFIELD_REFs into
980         vectors might cause maybe_lookup_element_for_expr to be called
981         on non-sra-candidate.
982
983 2007-02-04  Kazu Hirata  <kazu@codesourcery.com>
984
985         * config/bfin/bfin-modes.def, config/bfin/bfin.c,
986         config/bfin/bfin.md, config/bfin/predicates.md: Follow
987         spelling conventions.
988
989 2007-02-04  Richard Guenther  <rguenther@suse.de>
990
991         PR middle-end/30636
992         * fold-const.c (try_move_mult_to_index): Make sure to not
993         overflow one dimension of a multi-dimensional array access.
994
995 2007-02-04  Jan Hubicka  <jh@suse.cz>
996
997         * passes.c (init_optimization_passes): Reindent.
998
999 2007-02-04  Jan Hubicka  <jh@suse.cz>
1000             Eric Botcazou  <ebotcazou@adacore.com>
1001
1002         * tree-optimize.c (has_abnormal_outgoing_edge_p): Move to...
1003         (execute_fixup_cfg): Break out the abnormal goto code.
1004         * tree-inline.c (has_abnormal_outgoing_edge_p): ...here.
1005         (make_nonlocal_label_edges): Move here from execute_fixup_cfg.
1006         (optimize_inline_calls): Call make_nonlocal_label_edges.
1007
1008 2007-02-04  Jan Hubicka  <jh@suse.cz>
1009
1010         * tree-ssa-copyrename.c (copy_rename_partition_coalesce): Return
1011         true when something was changed.
1012         (rename_ssa_copies): When something was changed, do
1013         TODO_remove_unused_locals.
1014         * tree-ssa-forwprop.c (tree_ssa_forward_propagate_single_use_value):
1015         add TODO_remove_unused_locals when instruction was removed.
1016
1017 2007-02-04  Jan Hubicka  <jh@suse.cz>
1018
1019         * ipa-inline.c (try_inline): Improve debug output; work on already
1020         inline edges too.
1021         (cgraph_decide_inlining_incrementally): Indent; improve debug output;
1022         call try_inline for already inlined edges too when flattening; 
1023         inline also functions that make callee growth but overall unit size
1024         reduce.
1025
1026 2007-02-04  Kazu Hirata  <kazu@codesourcery.com>
1027
1028         * config/m32c/bitops.md, config/m32c/jump.md,
1029         config/m32c/m32c.c, config/m32c/m32c.h, config/m32r/m32r.c,
1030         config/m32r/m32r.h, config/m32r/m32r.md,
1031         config/m32r/predicates.md, config/m68hc11/larith.asm,
1032         config/m68hc11/m68hc11.c, config/m68hc11/m68hc11.h,
1033         config/m68k/m68k.h, config/mcore/mcore.md, config/mips/4k.md,
1034         config/mips/mips-protos.h, config/mips/mips.c,
1035         config/mips/mips.h, config/mips/mips.md, config/mips/mips16.S,
1036         config/mn10300/mn10300.h, config/mn10300/predicates.md,
1037         config/mt/mt.c, config/mt/mt.h, config/mt/mt.md: Follow
1038         spelling conventions.
1039
1040         * config/v850/v850.c, config/v850/v850.h, config/v850/v850.md:
1041         Follow spelling conventions.
1042
1043 2007-02-03  Douglas Gregor  <doug.gregor@gmail.com>
1044
1045        * c-opts.c (c_common_post_options): If C++0x mode is enabled, don't
1046        warn about C++0x compatibility.
1047         
1048 2007-02-04  Kazu Hirata  <kazu@codesourcery.com>
1049
1050         * config/h8300/h8300.c, config/h8300/h8300.h,
1051         config/h8300/h8300.md: Follow spelling conventions.
1052
1053 2007-02-03  Uros Bizjak  <ubizjak@gmail.com>
1054
1055         PR middle-end/30667
1056         * combine.c (try_combine): Do not substitute source operand
1057         with constants wider than 2 * HOST_BITS_PER_WIDE_INT.
1058
1059 2007-02-03  Jan Hubicka  <jh@suse.cz>
1060
1061         PR gcov-profile/30650
1062         * value-prof.c (stringop_block_profile): Fix handling of size counter;
1063         do not divide by zero for never executed counters.
1064         (tree_find_values_to_profile): Fix counters.
1065         * gcov-ui.h (GCOV_COUNTER_AVERAGE, GCOV_COUNTER_IOR): Fix comments.
1066
1067 2007-02-03  Ian Lance Taylor  <iant@google.com>
1068
1069         * lower-subreg.c (simple_move_operand): New static function,
1070         broken out of simple_move.  Reject LABEL_REF, SYMBOL_REF, and HIGH
1071         operands.
1072         (simple_move): Call simple_move_operand.
1073         (find_decomposable_subregs): Add special handling of MEMs.
1074         (can_decompose_p): Rename from cannot_decompose_p.  Reverse
1075         meaning of return value.  If we see a hard register, test whether
1076         it can store a word_mode value.  Change all callers.
1077
1078 2007-02-03  John David Anglin  <dave.anglin@nrc-cnrc.gc.ca>
1079
1080         * pa.md (addvdi3, addvsi3, subvdi3, subvsi3, negvdi2, negvsi2): New
1081         ftrapv insns and expanders.
1082         (subdi3): Change define_expand operand 1 to arith11_operand, and
1083         operand 2 to reg_or_0_operand.  Change constraints of 64-bit insn
1084         pattern to handle reg_or_0 operands.  Revise 32-bit insn pattern to
1085         handle 11-bit constants and reg_or_0 operands in operands 1 and 2,
1086         respectively.
1087
1088         PR middle-end/30174
1089         * varasm.c (notice_global_symbol): Treat global objects as weak when
1090         flag_shlib is true.
1091
1092 2007-02-03  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
1093
1094         * emit-rtl.c (dconstpi): Delete.
1095         (dconstsqrt2): New.
1096         (init_emit_once): Delete dconstpi and init dconstsqrt2.
1097         * real.h (dconstpi): Delete.
1098         (dconstsqrt2): New.
1099         * builtins.c (fold_builtin_cabs): Use dconstsqrt2.
1100         (fold_builtin_hypot): Likewise.
1101
1102 2007-02-03  Tom Tromey  <tromey@redhat.com>
1103
1104         PR driver/30246
1105         * gcc.c (cpp_unique_options): Any of -ggdb3, -gstabs3,
1106         -gcoff3, -gxcoff3, -gvms3 implies -dD.
1107
1108 2007-02-03  Kazu Hirata  <kazu@codesourcery.com>
1109
1110         * c-decl.c, config/avr/avr.c, config/avr/avr.h,
1111         config/m68k/m68k.c, config/m68k/netbsd-elf.h,
1112         config/mn10300/mn10300.c, config/pdp11/pdp11.h,
1113         config/rs6000/cell.md, config/rs6000/darwin.h,
1114         config/sh/sh.md, config/sh/sh4-300.md, config/spu/spu.c,
1115         config/spu/spu.md, cselib.c, expr.c, haifa-sched.c, hwint.h,
1116         jump.c, reload.c, sched-deps.c, sched-int.h, tree-inline.c,
1117         tree-profile.c, tree-ssa-live.h, tree-vrp.c: Fix comment
1118         typos.  Follow spelling conventions.
1119         * doc/invoke.texi: Follow spelling conventions.
1120
1121 2007-02-03  Roger Sayle  <roger@eyesopen.com>
1122
1123         * simplify-rtx.c (simplify_relational_operation_1): Implement some
1124         canonicalization transformations that attempt to simplify integer
1125         constant comparisons to become comparisons against zero.
1126
1127 2007-02-03  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
1128
1129         * builtins.c (fold_builtin_cabs): Fold cabs(x+xi) into
1130         fabs(x)*sqrt(2).
1131         * fold-const.c (fold_binary): Fix comment typos.  Fold complex
1132         (x,0)-(0,y) into (x,-y).  Likewise (0,y)-(x,0) into (-x,y).
1133
1134 2007-02-02  Mike Stump  <mrs@apple.com>
1135
1136         * config/darwin9.h (DARWIN_LINKER_GENERATES_ISLANDS): Add.
1137         * config/rs6000/rs6000.c (DARWIN_GENERATE_ISLANDS): Add.
1138         (output_call): Use DARWIN_GENERATE_ISLANDS to decide when to
1139         generate a branch island.
1140
1141 2007-02-02  Bob Wilson  <bob.wilson@acm.org>
1142
1143         * config/xtensa/xtensa.c (smalloffset_mem_p): Use BASE_REG_P.
1144         (xtensa_legitimate_address_p): New.
1145         (xtensa_legitimize_address): New.
1146         (xtensa_output_addr_const_extra): New.
1147         * config/xtensa/xtensa.h (REG_OK_STRICT_FLAG): Define.
1148         (BASE_REG_P): New.
1149         (REG_OK_FOR_BASE_P): Use BASE_REG_P.
1150         (GO_IF_LEGITIMATE_ADDRESS): Move code to xtensa_legitimate_address_p.
1151         (LEGITIMIZE_ADDRESS): Move code to xtensa_legitimize_address.
1152         (OUTPUT_ADDR_CONST_EXTRA): Move code to xtensa_output_addr_const_extra.
1153         * config/xtensa/xtensa-protos.h (xtensa_legitimate_address_p): New.
1154         (xtensa_legitimize_address): New.
1155         (xtensa_output_addr_const_extra): New.
1156
1157 2007-02-02  Steve Ellcey  <sje@cup.hp.com>
1158
1159         * config/ia64/ia64.c (ia64_print_operand): Fix compare strings.
1160
1161 2007-02-02  Ian Lance Taylor  <iant@google.com>
1162
1163         * expmed.c (expand_divmod): Add comment.
1164
1165 2007-02-02  Kazu Hirata  <kazu@codesourcery.com>
1166
1167         * emit-rtl.c (renumber_insns): Remove.
1168         * flags.h: Remove the extern for flag_renumber_insns.
1169         * rtl.h: Remove the prototype for renumber_insns.
1170         * toplev.c (flag_renumber_insns): Remove.
1171
1172 2007-02-02  Hui-May Chang  <hm.chang@apple.com>
1173
1174         Revert for x86 darwin:
1175         2005-06-19  Uros Bizjak  <uros@kss-loka.si>
1176         
1177         * config/i386/i386.c (ix86_function_arg_regno_p): Put back the
1178         code before the following patch under TARGET_MACHO.
1179         (ix86_function_value_regno_p): Likewise.
1180
1181 2007-02-02  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
1182
1183         * fold-const.c (negate_expr_p, fold_negate_expr): Handle
1184         COMPLEX_EXPR.
1185
1186 2007-02-02  Maxim Kuvyrkov  <mkuvyrkov@ispras.ru>
1187
1188         PR target/29682
1189         * config/ia64/ia64.c (ia64_speculate_insn): Restrict to memory loads to
1190         general or fp registers.  Add comments.
1191         * config/ia64/ia64.md (reg_pred_prefix): Add comment.
1192
1193 2007-02-02  Paolo Bonzini  <bonzini@gnu.org>
1194
1195         * pointer-set.c (insert_aux): Only return insertion slot.
1196         (pointer_set_insert): Adjust.
1197         (pointer_set_traverse, struct pointer_map_t, pointer_map_create,
1198         pointer_map_destroy, pointer_map_insert, pointer_map_contains,
1199         pointer_map_traverse): New.
1200         * pointer-set.h (pointer_set_traverse, struct pointer_map_t,
1201         pointer_map_create,  pointer_map_destroy, pointer_map_insert,
1202         pointer_map_contains, pointer_map_traverse): Declare.
1203
1204 2007-02-02  Jakub Jelinek  <jakub@redhat.com>
1205
1206         PR middle-end/30473
1207         * builtins.c (fold_builtin_sprintf): Do not attempt to optimize
1208         sprintf (str, "%s").  Do not optimize sprintf (str, "nopercent", p++).
1209
1210 2007-02-02  Maxim Kuvyrkov  <mkuvyrkov@ispras.ru>
1211
1212         * sched-int.h (ds_to_dk, dk_to_ds): Declare functions.
1213         
1214         (struct _dep): New type.
1215         (dep_t): New typedef.
1216         (DEP_PRO, DEP_CON, DEP_KIND): New access macros.
1217         (DEP_STATUS): New access macro.  The macro with the same name was
1218         renamed to DEP_LINK_STATUS.
1219         (dep_init): Declare function
1220
1221         (struct _dep_link): New type.
1222         (dep_link_t): New typedef.
1223         (DEP_LINK_NODE, DEP_LINK_NEXT, DEP_LINK_PREV_NEXTP): New access macros.
1224         (DEP_LINK_DEP, DEP_LINK_PRO, DEP_LINK_CON, DEP_LINK_KIND): New macros.
1225         (DEP_LINK_STATUS): New macro.
1226         (debug_dep_links): New debug function.
1227
1228         (struct _deps_list): New type.
1229         (deps_list_t): New typedef.
1230         (DEPS_LIST_FIRST): New access macro.
1231         (FOR_EACH_DEP_LINK): New cycle macro.
1232         (create_deps_list, free_deps_list, delete_deps_list): Declare
1233         functions.
1234         (deps_list_empty_p, debug_deps_list, add_back_dep_to_deps_list): Ditto.
1235         (find_link_by_pro_in_deps_list, find_link_by_con_in_deps_list): Ditto.
1236         (copy_deps_list_change_con): Ditto.
1237
1238         (move_dep_link): Declare function.
1239
1240         (struct _dep_node): New type.
1241         (dep_node_t): New typedef.
1242         (DEP_NODE_BACK, DEP_NODE_DEP, DEP_NODE_FORW): New access macros.
1243
1244         (struct haifa_insn_data.back_deps): New field to hold backward
1245         dependencies of the insn.
1246         (struct haifa_insn_data.depend): Rename to forw_deps.  Change its type
1247         to deps_list_t.
1248         (struct haifa_insn_data.resolved_deps): Rename to resolved_back_deps.
1249         Change its type to deps_list_t.
1250         (INSN_BACK_DEPS): New access macro to use instead of LOG_LINKS.
1251         (INSN_DEPEND): Rename to INSN_FORW_DEPS.
1252         (RESOLVED_DEPS): Rename to INSN_RESOLVED_BACK_DEPS.
1253
1254         (INSN_COST): Move to haifa-sched.c.  Use insn_cost () instead.
1255         
1256         (DEP_STATUS): Rename to DEP_LINK_STATUS.  Fix typo in the comment.
1257
1258         (add_forw_dep, delete_back_forw_dep, insn_cost): Update declaration and
1259         all callers.
1260         (dep_cost): Declare.
1261         
1262         * sched-deps.c (CHECK): New macro to (en/dis)able sanity checks.
1263         (ds_to_dk, dk_to_ds): New functions.
1264         
1265         (init_dep_1): New static function.
1266         (init_dep): New function.
1267         (copy_dep): New static function.
1268         
1269         (dep_link_consistent_p, attach_dep_link, add_to_deps_list): New static
1270         functions.
1271         (detach_dep_link): New static function.
1272         (move_dep_link): New function.
1273         
1274         (dep_links_consistent_p, dump_dep_links): New static functions.
1275         (debug_dep_links): New debugging function.
1276         
1277         (deps_obstack, dl_obstack, dn_obstack): New static variables.
1278         
1279         (alloc_deps_list, init_deps_list): New static functions.
1280         (create_deps_list): New function.
1281         (clear_deps_list): New static function.
1282         (free_deps_list, delete_deps_list, deps_list_empty_p): New functions.
1283         (deps_list_consistent_p, dump_deps_list): New static functions.
1284         (debug_deps_list): New function.
1285         (add_back_dep_to_deps_list, find_link_by_pro_in_deps_list): New
1286         functions.
1287         (find_link_by_con_in_deps_list, copy_deps_list_change_con): Ditto.
1288
1289         (maybe_add_or_update_back_dep_1, add_or_update_back_dep_1): Update to
1290         use new scheduler dependencies lists.
1291         (add_back_dep, delete_all_dependences, fixup_sched_groups): Ditto.
1292         (sched_analyze): Ditto.  Initialize dependencies lists.
1293         (add_forw_dep, compute_forward_dependences): Update to use new
1294         scheduler dependencies lists.
1295         
1296         (init_dependency_caches): Init deps_obstack.
1297         (free_dependency_caches): Free deps_obstack.
1298         
1299         (adjust_add_sorted_back_dep, adjust_back_add_forw_dep): Update to use
1300         new scheduler dependencies lists.
1301         (delete_forw_dep, add_or_update_back_forw_dep): Ditto.
1302         (add_back_forw_dep, delete_back_forw_dep): Ditto.
1303
1304         * sched-rgn.c (set_spec_fed, find_conditional_protection, is_pfree):
1305         Update to use new scheduler dependencies lists.
1306         (is_conditionally_protected, is_prisky, add_branch_dependences): Ditto.
1307         (debug_dependencies): Ditto.
1308         (schedule_region): Update comments.
1309         
1310         * sched-ebb.c (earliest_block_with_similiar_load): Update to use new
1311         scheduler dependencies lists.
1312         (schedule_ebb): Update comments.
1313         
1314         * rtl.def (DEPS_LIST): Remove.
1315         
1316         * lists.c (unused_deps_list): Remove.
1317         (free_list): Update assertions.
1318         
1319         (alloc_DEPS_LIST, free_DEPS_LIST_list, free_DEPS_LIST_node): Remove.
1320         (remove_free_DEPS_LIST_elem, copy_DEPS_LIST_list): Ditto.
1321
1322         * rtl.h (free_DEPS_LIST_list, alloc_DEPS_LIST): Remove declarations.
1323         (remove_free_DEPS_LIST_elem, copy_DEPS_LIST_list): Ditto.
1324         
1325         * haifa-sched.c (comments): Update.
1326         (insn_cost1): Remove.  Inline the code into insn_cost ().
1327         (insn_cost): Update to use new scheduler dependencies lists.  Move
1328         processing of the dependency cost to dep_cost ().
1329         (dep_cost): New function.  Use it instead of insn_cost () when
1330         evaluating cost of the dependency.  Use compatible interface to
1331         interact with the target.
1332         (priority): Update to use new scheduler dependencies lists.
1333         (rank_for_schedule): Ditto.  Optimize heuristic that prefers the insn
1334         with greater number of insns that depend on the insn.
1335         (schedule_insn): Update to use new scheduler dependencies lists.  Add
1336         code to free backward dependencies lists.  Inline and optimize code
1337         from resolve_dep () - see PR28071.
1338         (ok_for_early_queue_removal): Update to use new scheduler dependencies
1339         lists.  Update call to targetm.sched.is_costly_dependence hook.
1340         
1341         (fix_inter_tick, try_ready, fix_tick_ready): Update to use new
1342         scheduler dependencies lists.
1343         
1344         (resolve_dep): Remove.  Move the logic to schedule_insn ().
1345         (init_h_i_d): Initialize dependencies lists.
1346         
1347         (process_insn_depend_be_in_spec): Rename to
1348         process_insn_forw_deps_be_in_spec.  Update to use new scheduler
1349         dependencies lists.
1350         (add_to_speculative_block, create_check_block_twin, fix_recovery_deps):
1351         Update to use new scheduler dependencies lists.
1352         (clear_priorities, calc_priorities, add_jump_dependencies): Ditto.
1353         
1354         * ddg.c (create_ddg_dependence, create_ddg_dep_no_link): Update to use
1355         new scheduler dependencies lists.
1356         (build_intra_loop_deps): Ditto.
1357         
1358         * target.h (struct _dep): Declare to use in
1359         gcc_target.sched.is_costly_dependence.
1360         (struct gcc_target.sched.adjust_cost): Fix typo.
1361         (struct gcc_target.sched.is_costly_dependence): Change signature to use
1362         single dep_t parameter instead of an equivalent triad.
1363         (struct gcc_target.sched.adjust_cost_2): Remove.
1364
1365         * target-def.h (TARGET_SCHED_ADJUST_COST_2): Remove.
1366
1367         * reg-notes.def (DEP_TRUE, DEP_OUTPUT, DEP_ANTI): Update comments.
1368
1369         * doc/tm.texi (TARGET_SCHED_IS_COSTLY_DEPENDENCE): Update
1370         documentation.
1371         (TARGET_SCHED_ADJUST_COST_2): Remove documentation.
1372
1373         * doc/rtl.texi (LOG_LINKS): Remove part about instruction scheduler.
1374         (REG_DEP_TRUE): Document.
1375         
1376         * config/ia64/ia64.c (ia64_adjust_cost_2): Rename to ia64_adjust_cost.
1377         Change signature to correspond to the targetm.sched.adjust_cost hook.  
1378         Update use in TARGET_SCHED_ADJUST_COST_2.
1379         (TARGET_SCHED_ADJUST_COST_2): Rename to TARGET_SCHED_ADJUST_COST.
1380         (ia64_dependencies_evaluation_hook, ia64_dfa_new_cycle): Update to use
1381         new scheduler dependencies lists.
1382         (ia64_gen_check): Ditto.
1383
1384         * config/mips/mips.c (vr4130_swap_insns_p): Update to use new scheduler
1385         dependencies lists.
1386         
1387         * config/rs6000/rs6000.c (rs6000_is_costly_dependence): Change
1388         signature to correspond to the targetm.sched.is_costly_dependence hook.
1389         (is_costly_group): Update to use new scheduler dependencies lists.
1390
1391         * config/spu/spu.c (spu_sched_adjust_cost): Use insn_cost () function
1392         instead of INSN_COST () macro.
1393
1394 2007-02-01  Ian Lance Taylor  <iant@google.com>
1395
1396         * lower-subreg.c (resolve_clobber): Handle a subreg of a concatn.
1397
1398 2007-02-01  Guy Martin  <gmsoft@gentoo.org>
1399
1400         * pa.md (tp_load): Correct mfctl instruction syntax.
1401
1402 2007-02-01  Geoffrey Keating  <geoffk@apple.com>
1403
1404         * config/rs6000/rs6000.c (rs6000_stack_info): Correct
1405         altivec_padding_size calculation on AIX.  Improve comment, add
1406         assert to verify that it's right.
1407
1408         * config/rs6000/darwin.h (MD_UNWIND_SUPPORT): Don't define for
1409         64-bit.
1410
1411 2007-2-01  Seongbae Park  <seongbae.park@gmail.com>
1412
1413         PR inline-asm/28686
1414         * global.c (global_alloc): Add mising initialization of
1415         ELIMINABLE_REGSET.
1416
1417 2007-02-01  Roger Sayle  <roger@eyesopen.com>
1418
1419         * alias.c (init_alias_analysis): Correct whitespace.
1420         * bb-reorder.c (fix_edges_for_rarely_executed_code,
1421         partition_hot_cold_basic_blocks): Likewise.
1422         * builtins.c (expand_builtin_printf, expand_builtin_fprintf,
1423         expand_builtin_sprintf, fold_builtin_carg, fold_builtin_sprintf,
1424         maybe_emit_sprintf_chk_warning, fold_builtin_sprintf_chk,
1425         fold_builtin_snprintf_chk, fold_builtin_printf,
1426         fold_builtin_fprintf, do_mpfr_ckconv, do_mpfr_arg1, do_mpfr_arg2,
1427         do_mpfr_arg3, do_mpfr_sincos): Likewise.
1428         * cfgcleanup.c (cleanup_cfg): Likewise.
1429         * cfgexpand.c (tree_expand_cfg): Likewise.
1430         * fold-const.c (fold_binary) <RDIV_EXPR>: Likewise.
1431         * function.c (get_next_funcdef_no): Likewise.
1432         * gengtype.c (main): Likewise.
1433         * genmodes.c (main): Likewise.
1434         * gcse.c (bypass_conditional_jumps, print_ldst_list): Likewise.
1435         * haifa-sched.c (schedule_block, extend_h_i_d): Likewise.
1436         * ifcvt.c (noce_emit_move_insn): Likewise.
1437         * modulo-sched.c (generate_prolog_epilog, sms_schedule_by_order):
1438         Likewise.
1439         * stor-layout.c (get_best_mode): Likewise.
1440         * tree-ssa-loop-niter.c (get_val_for): Likewise.
1441         * tree-ssa-structalias.c (get_varinfo, get_varinfo_fc,
1442         scc_visit, do_ds_constraint, do_complex_constraint, label_visit,
1443         perform_var_substitution, solve_graph): Likewise.
1444         * tree-vrp.c (vrp_finalize): Likewise.
1445
1446 2007-02-01  Ian Lance Taylor  <iant@google.com>
1447
1448         * lower-subreg.c (simplify_gen_subreg_concatn): If we ask for the
1449         high part of a paradoxical subreg, return a constant zero.
1450
1451 2007-02-01  Zdenek Dvorak <dvorakz@suse.cz>
1452
1453         * toplev.c (lang_dependent_init): Call init_set_costs.
1454         * loop-init.c (loop_optimizer_init): Do not call init_set_costs.
1455
1456 2007-02-01  Richard Guenther  <rguenther@suse.de>
1457
1458         PR middle-end/30656
1459         * fold-const.c (fold_negate_expr): Allow negating a
1460         constant if overflow does not change.
1461
1462 2007-02-01  Ralf Wildenhues  <Ralf.Wildenhues@gmx.de>
1463
1464         * doc/c-tree.texi (Expression trees): Improve markup.
1465         * doc/tm.texi (Register Classes, Addressing Modes)
1466         (Floating Point): Fix spacing after abbreviations.  Fix some
1467         typos.
1468
1469 2007-02-01  Ben Elliston  <bje@au.ibm.com>
1470
1471         * doc/invoke.texi: Replace "bugfix" with "bug fix" throughout.
1472         * doc/contrib.texi: Likewise.
1473         * doc/install.texi: Likewise.
1474
1475 2007-01-31  Richard Henderson  <rth@redhat.com>
1476             Ian Lance Taylor  <iant@google.com>
1477
1478         * lower-subreg.c: New file.
1479         * rtl.def (CONCATN): Define.
1480         * passes.c (init_optimization_passes): Add pass_lower_subreg and
1481         pass_lower_subreg2.
1482         * emit-rtl.c (update_reg_offset): New static function, broken out
1483         of gen_rtx_REG_offset.
1484         (gen_rtx_REG_offset): Call update_reg_offset.
1485         (gen_reg_rtx_offset): New function.
1486         * regclass.c: Revert patch of 2006-03-05, restoring
1487         reg_scan_update.
1488         (clear_reg_info_regno): New function.
1489         * dwarf2out.c (concatn_loc_descriptor): New static function.
1490         (loc_descriptor): Handle CONCATN.
1491         * common.opt (fsplit_wide_types): New option.
1492         * opts.c (decode_options): Set flag_split_wide_types when
1493         optimizing.
1494         * timevar.def (TV_LOWER_SUBREG): Define.
1495         * rtl.h (gen_reg_rtx_offset): Declare.
1496         (reg_scan_update): Declare.
1497         * regs.h (clear_reg_info_regno): Declare.
1498         * tree-pass.h (pass_lower_subreg): Declare.
1499         (pass_lower_subreg2): Declare.
1500         * doc/invoke.texi (Option Summary): List -fno-split-wide-types.
1501         (Optimize Options): Add -fsplit-wide-types to -O1 list.  Document
1502         -fsplit-wide-types.
1503         * doc/rtl.texi (Regs and Memory): Document concat and concatn.
1504         * Makefile.in (OBJS-common): Add lower-subreg.o.
1505         (lower-subreg.o): New target.
1506
1507 2007-01-31  Kazu Hirata  <kazu@codesourcery.com>
1508
1509         * config/sh/sh.h (HAVE_SECONDARY_RELOADS): Remove.
1510
1511 2007-01-31  Anatoly Sokolov <aesok@post.ru>
1512
1513         PR target/19087
1514         * config/avr/avr.c (DWARF2_ADDR_SIZE): Define.
1515
1516 2007-01-31  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
1517
1518         PR middle-end/29335
1519         * builtins.c (fold_builtin_sqrt): Use MPFR for constant args.
1520
1521 2007-01-31  Zdenek Dvorak <dvorakz@suse.cz>
1522
1523         * cfgloop.h: Include vec-prim.h.
1524         (enum li_flags): Remove LI_ONLY_OLD.
1525         (loop_iterator): Changed.
1526         (fel_next, fel_init): Iterate over loop tree.
1527         (FOR_EACH_LOOP_BREAK): New macro.
1528         * loop-unswitch.c (unswitch_loops): Do not pass LI_ONLY_OLD to
1529         FOR_EACH_LOOP.
1530         * tree-ssa-loop-unswitch.c (tree_ssa_unswitch_loops): Ditto.
1531         * modulo-sched.c (sms_schedule): Ditto.
1532         * tree-vectorizer.c (vectorize_loops): Ditto.
1533         * doc/loop.texi: Update information on loop numbering and behavior of
1534         FOR_EACH_LOOP wrto new loops.
1535         * tree-scalar-evolution.c (compute_overall_effect_of_inner_loop,
1536         add_to_evolution_1): Test nestedness of loops instead of comparing
1537         their numbers.
1538         * tree-chrec.c (chrec_fold_plus_poly_poly,
1539         chrec_fold_multiply_poly_poly, chrec_evaluate,
1540         hide_evolution_in_other_loops_than_loop, chrec_component_in_loop_num,
1541         reset_evolution_in_loop): Ditto.
1542         * Makefile.in (CFGLOOP_H): Add vecprim.h dependency.
1543
1544 2007-01-31  Dirk Mueller  <dmueller@suse.de>
1545
1546         * c-common.c (warn_about_parentheses): Separate warning about
1547         un-parenthized sequence of comparison operators from the one
1548         which is supposed to warn about x <= y <= z.
1549
1550 2007-01-31  Uros Bizjak  <ubizjak@gmail.com>
1551
1552         * optabs.h (enum optab_index): Add new OTI_isinf.
1553         (isinf_optab): Define corresponding macro.
1554         * optabs.c (init_optabs): Initialize isinf_optab.
1555         * genopinit.c (optabs): Implement isinf_optab using isinf?f2
1556         patterns.
1557         * builtins.c (mathfn_built_in): Handle BUILT_IN_ISINF{,F,L}.
1558         (expand_builtin_interclass_mathfn): Expand BUILT_IN_ISINF{,F,L}
1559         using isinf_optab.
1560         (expand_builtin): Expand BUILT_IN_ISINF{,F,L} using
1561         expand_builtin_interclass_mathfn.
1562         * reg_stack.c (subst_stack_regs_pat): Handle UNSPEC_FXAM.
1563         * config/i386/i386.md (UNSPEC_FXAM): New constant.
1564         (fxam<mode>2_i387): New insn pattern.
1565         (isinf<mode>2) New expander to implement isinf, isinff and isinfl
1566         built-in functions as x87 inline asm.
1567
1568 2007-01-31  Kazu Hirata  <kazu@codesourcery.com>
1569
1570         * gcc/config/arm/unwind-arm.h (_sleb128_t, _uleb128_t): New.
1571
1572 2007-01-30  Eric Christopher  <echristo@apple.com>
1573
1574         * config.gcc: Add geode.
1575
1576 2007-01-31  Kazu Hirata  <kazu@codesourcery.com>
1577
1578         * cgraphunit.c, config/arm/arm.c, config/m68k/m68k.c,
1579         ipa-inline.c, tree-profile.c, tree-ssa-live.c,
1580         tree-ssa-math-opts.c, tree-ssanames.c, tree-vect-analyze.c,
1581         value-prof.c: Fix comment typos.
1582
1583 2007-01-30  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
1584
1585         PR c++/24745
1586         * doc/invoke.texi (Wpointer-arith): Document warning.
1587
1588 2007-01-30  Janis Johnson  <janis187@us.ibm.com>
1589
1590         * doc/extend.texi (Decimal Floating Types): Remove decfloat.h from
1591         the list of discrepancies from the draft TR.
1592
1593 2007-01-30  Dirk Mueller  <dmueller@suse.de>
1594
1595         PR c++/30601
1596         * doc/invoke.texi (-Wreturn-type): Update description to
1597         match new behavior.
1598
1599 2007-01-30  Richard Sandiford  <richard@codesourcery.com>
1600
1601         * cfgrtl.c (try_redirect_by_replacing_jump): Check only_sets_cc0_p.
1602
1603 2007-01-30  Uros Bizjak  <ubizjak@gmail.com>
1604
1605         * builtins.c (expand_builtin_int_interclass_roundingfn): New function
1606         to handle optabs that operate on floating point input argument and
1607         output to integer output.
1608         (expand_builtin_mathfn) [BUILT_IN_ILOGB]: Move from here ...
1609         (expand_builtin_interclass_mathfn) [BUILT_IN_ILOGB]: ... to here.
1610         (expand_builtin): Expand BUILT_IN_ILOGB{,F,L} using
1611         expand_builtin_interclass_mathfn ().
1612         * config/i386/i386.md (fxtractxf3_i387): Rename from *fxtractxf3_i387.
1613         (ilogbsi2): Remove.
1614         (ilogbxf2, ilogb<mode>2): New expanders to implement ilogb, ilogbf and
1615         ilogbl built-in functions as x87 intrinsics.
1616
1617 2007-01-30  Richard Guenther  <rguenther@suse.de>
1618
1619         PR middle-end/27657
1620         * dwarf2out.c (reference_to_unused): Query varpool if the
1621         variable was output.
1622
1623 2007-01-30  Richard Guenther  <rguenther@suse.de>
1624
1625         PR middle-end/30313
1626         * passes.c (execute_one_pass): Reset in_gimple_form to not
1627         confuse non-unit-at-a-time mode.
1628
1629 2007-01-29  Roger Sayle  <roger@eyesopen.com>
1630             Richard Guenther  <rguenther@suse.de>
1631
1632         * fold-const.c (round_up): Make HIGH an unsigned HOST_WIDE_INT to
1633         avoid undefined behaviour on overflow.  Use force_fit_type_double
1634         to construct the constant with the specified TREE_OVERFLOW.
1635
1636 2007-01-29  Janis Johnson  <janis187@us.ibm.com>
1637
1638         * config/dfp-bit.c: Add parameterized support for fp exceptions.
1639         * config/dfp-bit.h: Ditto.
1640
1641 2007-01-29  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
1642
1643         * c-decl.c (pop_scope): Replace warnings with call to
1644         warn_for_unused_label.
1645         * c-common.h (warn_for_unused_label): Declare.
1646         * c-common.c (warn_for_unused_label): Define.
1647
1648 2007-01-29  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
1649
1650         * tree-optimize.c (update_inlined_to_pointers): Delete unused
1651         function.
1652
1653 2007-01-29  Janis Johnson  <janis187@us.ibm.com>
1654
1655         * Makefile.in (USER_H): Remove decfloat.h.
1656         * ginclude/decfloat.h: Delete, moving contents to ...
1657         * ginclude/float.h: Add support for decimal floating point,
1658         guarded by __STDC_WANT_DEC_FP__.
1659
1660 2007-01-29  Mike Stump  <mrs@apple.com>
1661
1662         * doc/gccint.texi (Top): Rename Loop Representation to Loop
1663         Analysis and Representation to resolve case insensitive conflict.
1664         * doc/loop.texi (Loop Analysis and Representation): Likewise.
1665
1666 2007-01-28  Daniel Berlin  <dberlin@dberlin.org>
1667
1668         * tree.h (struct tree_memory_tag): Add aliases member.
1669         (MTAG_ALIASES): New macro.
1670         * tree-ssa-alias.c (alias_bitmap_obstack): New variable.
1671         (add_may_alias): Remove pointer-set. Update for may_aliases being
1672         a bitmap.
1673         (mark_aliases_call_clobbered): Update for may_aliases being a
1674         bitmap.
1675         (compute_tag_properties): Ditto.
1676         (create_partition_for): Ditto.
1677         (compute_memory_partitions): Ditto.
1678         (dump_may_aliases_for): Ditto.
1679         (is_aliased_with): Ditto.
1680         (add_may_alias_for_new_tag): Ditto.
1681         (rewrite_alias_set_for): Rewrite for may_aliases being a bitmap.
1682         (compute_is_aliased): New function.
1683         (compute_may_aliases): Call compute_is_aliased).
1684         (init_alias_info): Initialize alias_bitmap_obstack.
1685         (union_alias_set_into): New function.
1686         (compute_flow_sensitive_aliasing): Use union_aliases_into.
1687         (have_common_aliases_p): Rewrite to take two bitmaps and use
1688         intersection.
1689         (compute_flow_insensitive_aliasing): Stop using pointer-sets.
1690         Update for bitmaps.
1691         (finalize_ref_all_pointers): Update for add_may_alias changes.
1692         (new_type_alias): Ditto.
1693         * tree-flow-inline.h (may_aliases): Return a bitmap.
1694         * tree-dfa.c (dump_variable): Check for MTAG_P'ness.
1695         * tree-ssa.c (verify_flow_insensitive_alias_info): Update for
1696         may_aliases being a bitmap.
1697         * tree-flow.h (struct var_ann_d): Remove may_aliases member.
1698         may_aliases now returns a bitmap.
1699         * tree-ssa-structalias.c (merge_smts_into): Update for may_aliases
1700         being a bitmap.
1701         * tree-ssa-operands.c (add_virtual_operand): Update for
1702         may_aliases being a bitmap.
1703
1704 2007-01-29  Daniel Berlin  <dberlin@dberlin.org>
1705
1706         PR tree-optimization/30630
1707         * tree-ssa-structalias.c (do_complex_constraint): Mark correct
1708         variable as changed.
1709
1710 2007-01-29  Simon Martin  <simartin@users.sourceforge.net>
1711
1712         PR c++/28266
1713         * gimplify.c (gimplify_target_expr): Make sure that the TARGET_EXPR is
1714         expanded only once even if an error occurs.
1715
1716 2007-01-29  Ben Elliston  <bje@au.ibm.com>
1717
1718         * gcov-io.h (__gcov_indirect_call_profiler): Declare.
1719         (__gcov_average_profiler): Likewise.
1720         (__gcov_ior_profiler): Likewise.
1721         (__gcov_merge_ior): Likewise.
1722
1723 2007-01-28  Jan Hubicka  <jh@suse.cz>
1724
1725         * builtins.c (expand_builtin_memset): Fix typo in my last patch.
1726         * value-prof.c (stringop_block_profile): Likewise.
1727
1728 2007-01-28  Jan Hubicka  <jh@suse.cz>
1729
1730         * expr.c (emit_block_move_via_movmem, emit_block_move_via_libcall): Add
1731         variant handling histograms; add wrapper.
1732         (clear_storage_via_libcall): Export.
1733         (emit_block_move_hints): Break out from ...; add histograms.
1734         (emit_block_move): ... this one.
1735         (clear_storage_hints): Break out from ...; add histograms.
1736         (clear_storage): ... this one.
1737         (set_storage_via_memset): Handle histogram.
1738         * expr.h (emit_block_move_via_libcall, emit_block_move_hints): Declare.
1739         (clear_storage_hints, clear_storage_via_libcall): Declare.
1740         (set_storage_via_setmem): Update prototype.
1741         * doc/md.texi (movmem, setmem): Document new arguments.
1742
1743         * value-prof.c (dump_histogram_value, tree_find_values_to_profile): Add
1744         new histograms.
1745         (stringop_block_profile): New global function.
1746         (tree_stringops_values_to_profile): Profile block size and alignment.
1747         * value-prof.h (enum hist_type): add HIST_TYPE_AVERAGE and
1748         HIST_TYPE_IOR.
1749         (struct profile_hooks): Add gen_average_profiler and gen_ior_profiler.
1750         (stringop_block_profile): Declare.
1751         * builtins.c: Include value-prof.h.
1752         (expand_builtin_memcpy, expand_builtin_memset): Pass block profile.
1753         * gcov-ui.h (GCOV_COUNTER_NAMES): Add new counter.
1754         (GCOV_COUNTER_AVERAGE, GCOV_COUNTER_IOR): New constants.
1755         (GCOV_COUNTERS, GCOV_LAST_VALUE_COUNTER): Update.
1756         * profile.c (instrument_values): Add new counters.
1757         * cfgexpand.c (expand_gimple_basic_block): Propagate histograms to
1758         calls.
1759         * tree-profile.c (tree_average_profiler_fn, tree_ior_profiler_fn): New.
1760         (tree_init_edge_profiler): Build new profilers.
1761         (tree_gen_average_profiler, tree_gen_ior_profiler): New.
1762         (pass_tree_profile): Add dump.
1763         (tree_profile_hooks): Update.
1764         * Makefile.in (LIBGCOV): Add new constants.
1765         * libgcov.c (__gcov_merge_ior, __gcov_average_profiler,
1766         __gcov_ior_profiler): New.
1767         * i386.md (movmem/setmem expanders): Add new optional arguments.
1768
1769 2007-01-28  David Edelsohn  <edelsohn@gnu.org>
1770
1771         * doc/md.texi (Standard Pattern Names): Document blockage pattern.
1772
1773 2007-01-28  Zdenek Dvorak <dvorakz@suse.cz>
1774
1775         * tree-ssa-loop-unswitch.c: Include tree-inline.h.
1776         (tree_unswitch_single_loop): Pass eni_size_weights to
1777         tree_num_loop_insns.
1778         * tree-ssa-loop-manip.c: Include tree-inline.h.
1779         (can_unroll_loop_p): Pass eni_size_weights to
1780         tree_num_loop_insns.
1781         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p):
1782         Pass eni_size_weights to estimate_num_insns.
1783         * tree.h (init_inline_once): Export.
1784         * toplev.c (backend_init): Call init_inline_once.
1785         * cgraphunit.c (cgraph_process_new_functions,
1786         cgraph_analyze_function): Pass eni_inlining_weights to
1787         estimate_num_insns.
1788         * ipa-inline.c (compute_inline_parameters): Ditto.
1789         * tree-ssa-loop-ivcanon.c (tree_num_loop_insns): Pass weights
1790         to estimate_num_insns.
1791         (try_unroll_loop_completely): Pass eni_size_weights to
1792         tree_num_loop_insns.
1793         * tree-eh.c (decide_copy_try_finally): Pass eni_size_weights
1794         ot estimate_num_insns.
1795         * tree-ssa-loop-prefetch.c: Include tree-inline.h.
1796         (loop_prefetch_arrays): Pass eni_time_weights to tree_num_loop_insns.
1797         * tree-inline.c (eni_inlining_weights, eni_size_weights,
1798         eni_time_weights): New variables.
1799         (init_inline_once): Initialize them.
1800         (struct eni_data): Mew.
1801         (estimate_num_insns_1, estimate_num_insns): Use weights.
1802         * tree-inline.h (struct eni_weights_d): New.
1803         (eni_inlining_weights, eni_size_weights, eni_time_weights): Declare.
1804         (estimate_num_insns): Declaration changed.
1805         * cfgloop.h (tree_num_loop_insns): Declaration changed.
1806         * Makefile.in (tree-ssa-loop-unswitch.o, tree-ssa-loop-prefetch.o,
1807         tree-ssa-loop-manip.o): Add TREE_INLINE_H dependency.
1808
1809 2007-01-28  Zdenek Dvorak <dvorakz@suse.cz>
1810
1811         * tree-data-ref.c (conflict_fn): Assert that the number of affine
1812         relations in the conflict function is valid.
1813
1814 2007-01-27  Ian Lance Taylor  <iant@google.com>
1815
1816         * common.opt: Add fstrict-overflow.
1817         * opts.c (decode_options): Set flag_strict_overflow if -O2.
1818         * flags.h (TYPE_OVERFLOW_WRAPS): Define.
1819         (TYPE_OVERFLOW_UNDEFINED): Define.
1820         (TYPE_OVERFLOW_TRAPS): Define.  This replaces TYPE_TRAP_SIGNED.
1821         Replace all uses.
1822         * tree.h (TYPE_TRAP_SIGNED): Don't define.
1823         * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_UNDEFINED.
1824         (fold_negate_expr): Likewise.
1825         (make_range): Likewise.
1826         (extract_muldiv_1): Likewise.
1827         (maybe_canonicalize_comparison): Likewise.
1828         (fold_comparison): Likewise.
1829         (fold_binary): Likewise.
1830         (tree_expr_nonnegative_p): Likewise.
1831         (tree_expr_nonzero_p): Likewise.
1832         * tree-vrp.c (compare_values): Likewise.
1833         (extract_range_from_binary_expr): Likewise.
1834         (extract_range_from_unary_expr): Likewise.
1835         * tree-ssa-loop-niter.c (infer_loop_bounds_from_signedness):
1836         Likewise.
1837         (nowrap_type_p): Likewise.
1838         * tree-scalar-evolution.c (simple_iv): Likewise.
1839         * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_WRAPS.
1840         (build_range_check): Likewise.
1841         (extract_muldiv_1): Likewise.
1842         (fold_comparison): Likewise.
1843         * tree-vrp.c (vrp_int_const_binop): Likewise.
1844         (extract_range_from_unary_expr): Likewise.
1845         * convert.c (convert_to_integer): Likewise.
1846         * fold-const.c (fold_negate_expr): Use TYPE_OVERFLOW_TRAPS.
1847         (fold_comparison): Likewise.
1848         (fold_binary): Likewise.
1849         * optabs.c (optab_for_tree_code): Likewise.
1850         * tree-vectorizer.c (vect_is_simple_reduction): Likewise.
1851         * simplify-rtx.c (simplify_const_relational_operation): Check
1852         flag_strict_overflow and flag_trapv.
1853         (simplify_const_relational_operation): Likewise.
1854         * doc/invoke.texi (Option Summary): Mention -fstrict-overflow.
1855         (Optimize Options): Add -fstrict-overflow to -O2 list.  Document
1856         -fstrict-overflow.
1857
1858 2007-01-27  Roger Sayle  <roger@eyesopen.com>
1859
1860         * tree.c (tree_fold_gcd): Delete.
1861         * tree.h (tree_fold_gcd): Remove prototype.
1862         * tree-data-ref.c (tree_fold_divides_p): Don't use tree_fold_gcd to
1863         test whether one constant integer is a multiple of another.  Instead
1864         call int_const_binop with TRUNC_MOD_EXPR and test for a zero result.
1865         * fold-const.c (multiple_of_p):  We've determined both TOP and
1866         BOTTOM are integer constants so we can call int_const_binop directly
1867         instead of the more generic const_binop.
1868
1869 2007-01-27  Roger Sayle  <roger@eyesopen.com>
1870
1871         * fold-const.c (size_binop): In the fast-paths for X+0, 0+X, X-0 and
1872         1*X check that the constant hasn't overflowed, to preserve the
1873         TREE_OVERFLOW bit.
1874         (round_up): Provide an efficient implementation when rouding-up an
1875         INTEGER_CST to a power-of-two.
1876
1877 2007-01-28  Ralf Wildenhues  <Ralf.Wildenhues@gmx.de>
1878
1879         * doc/sourcebuild.texi: Add comma for clarity.
1880         * doc/extend.texi: Fix some typos.
1881         * doc/passes.texi: Likewise.
1882         * doc/cppinternals.texi: Likewise.
1883         * doc/c-tree.texi: Likewise.
1884         * doc/tree-ssa.texi: Likewise.
1885         * doc/install.texi: Likewise.
1886
1887 2007-01-27  Jan Hubicka  <jh@suse.cz>
1888
1889         * tree-sra.c (sra_walk_function): Don't rely on aliases being build.
1890         (pass_sra): Do not require alias information.
1891         * passes.c (init_optimization_passes): Add SRA
1892
1893 2007-01-27  Steven Bosscher  <steven@gcc.gnu.org>
1894
1895         * tracer.c (rest_of_handle_tracer): We already cleaned
1896         up the CFG in tracer() so don't do it here again.
1897         * cfgcleanup.c (rest_of_handle_jump2): Don't repeat
1898         cleanup_cfg here, either.  And don't call renumber_insns.
1899
1900         * cfgrtl.c (rtl_verify_flow_info_1): Don't verify that BB_END
1901         and BB_HEAD are in the insn stream here.  Instead make sure
1902         that BB_INSN is valid on all insns.  Also, do check here that
1903         there are no pending branch predictions...
1904         (rtl_verify_flow_info): ...instead of doing it here.  Checks
1905         for BB_END and BB_HEAD moved from rtl_verify_flow_info_1 to
1906         here.
1907
1908 2007-01-26  Roger Sayle  <roger@eyesopen.com>
1909
1910         * config/i386/i386.c (ix86_swap_binary_operands_p): New helper
1911         function to simplify/factorize operand order canonicalization.
1912         (ix86_fixup_binary_operands): Reorganize using the above function.
1913         (ix86_binary_operator_ok): Likewise.
1914
1915 2007-01-27  Jakub Jelinek  <jakub@redhat.com>
1916
1917         * genattrtab.c (struct attr_value_list, insn_code_values): Move to
1918         file scope from optimize_attrs.
1919         (simplify_test_exp): If insn_code_values is not NULL, use it to speed
1920         up search.
1921         (optimize_attrs): Clear insn_code_values after freeing it.
1922
1923 2007-01-26  Zdenek Dvorak <dvorakz@suse.cz>
1924
1925         * tree-ssa-address.c (create_mem_ref): Remove ", bsi" from
1926         a parts.base assignment.
1927
1928 2007-01-26  Zdenek Dvorak <dvorakz@suse.cz>
1929
1930         * tree-data-ref.c (dump_subscript): Use dump_conflict_function.
1931         (compute_subscript_distance, initialize_data_dependence_relation,
1932         finalize_ddr_dependent, analyze_ziv_subscript,
1933         analyze_siv_subscript_cst_affine,
1934         compute_overlap_steps_for_affine_univar,
1935         compute_overlap_steps_for_affine_1_2, analyze_subscript_affine_affine,
1936         analyze_siv_subscript, analyze_miv_subscript,
1937         analyze_overlapping_iterations, subscript_dependence_tester_1,
1938         compute_self_dependence, free_dependence_relation): Work
1939         with affine_fn instead of chrecs.
1940         (dump_affine_function, dump_conflict_function, affine_function_equal_p,
1941         common_affine_function, affine_function_base,
1942         affine_function_constant_p, affine_fn_op, affine_fn_plus,
1943         affine_fn_minus, affine_fn_free, conflict_fn_not_known,
1944         conflict_fn_no_dependence, free_conflict_function, free_subscripts,
1945         conflict_fn, affine_fn_cst, affine_fn_univar): New functions.
1946         (all_chrecs_equal_p): Removed.
1947         * tree-data-ref.h (affine_fn, conflict_function): New types.
1948         (struct subscript): Change type of conflicting_iterations_in_a
1949         and conflicting_iterations_in_b.
1950
1951 2007-01-26  Steve Ellcey  <sje@cup.hp.com>
1952
1953         PR other/30182
1954         * config/pa/pa.h (TARGET_HPUX_11): New.
1955         * config/pa/pa-hpux11.h (TARGET_HPUX_11): New.
1956         * config/pa/pa.c (pa_init_builtins): Use TARGET_HPUX_11.
1957
1958 2007-01-26  Daniel Berlin  <dberlin@dberlin.org>
1959         Richard Guenther  <rguenther@suse.de>
1960
1961         * tree-ssa-structalias.c (solve_graph): Handle case
1962         we merged the variable to another.
1963
1964 2007-01-25  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
1965
1966         * builtins.c (fold_builtin_1): Treat ccos and ccosh as 'even'
1967         functions.
1968
1969         * fold-const.c (negate_mathfn_p): Treat casin, casinh, catan,
1970         catanh, cproj, csin, csinh, ctan and ctanh as 'odd' functions.
1971
1972 2007-01-25  DJ Delorie  <dj@redhat.com>
1973
1974         * config/m32c/m32c.c (m32c_cannot_change_mode_class): We don't
1975         allow changes to modes which don't fit in those registers.
1976
1977         * reload1.c (choose_reload_regs): Check for invalid subregs before
1978         computing their locations, not after.
1979
1980 2007-01-25  Geoffrey Keating  <geoffk@apple.com>
1981
1982         PR 25127
1983         * config/rs6000/rs6000.c (first_altivec_reg_to_save): On Darwin,
1984         save Altivec registers in an eh_return function.
1985         (compute_vrsave_mask): Likewise.
1986         (rs6000_stack_info): Correct AIX/Darwin stack alignment computation
1987         for saving Altivec registers.
1988         (rs6000_emit_prologue): Don't allocate stack twice in
1989         eh_return function.  Correct expected value of altivec_save_offset
1990         when using save_world.  Describe save of R0 to stack when using
1991         save_world.  Describe stack pointer adjustment when using
1992         save_world.  Remove duplicated eh_return parameter register saving.
1993         Update sp_offset variable after save_world.
1994         * config/rs6000/t-darwin (LIB2FUNCS_STATIC_EXTRA): Remove
1995         darwin-world.asm.
1996         (LIB2FUNCS_EXTRA): Add darwin-world.asm.
1997         * config/rs6000/darwin.h (SUBTARGET_OVERRIDE_OPTIONS): -m64
1998         implies Altivec.
1999
2000 2007-01-25  Steve Ellcey  <sje@cup.hp.com>
2001
2002         * config.gcc (ia64*-*-hpux*): Make posix threads the default.
2003
2004 2007-01-25  Steve Ellcey  <sje@cup.hp.com>
2005
2006         PR other/30182
2007         * config/pa/pa.c (pa_init_builtins): Set asm names for finite routines.
2008         * config/ia64/ia64.c (ia64_init_builtins):  Ditto.
2009
2010 2007-01-25  Richard Guenther  <rguenther@suse.de>
2011
2012         * doc/invoke.texi (-Wcoverage-mismatch): Document.
2013         * common.opt (-Wcoverage-mismatch): New warning option.
2014         * coverage.c (get_coverage_counts): Ignore coverage mismatch
2015         if -Wcoverage-mismatch is given.
2016
2017 2007-01-25  Razya Ladelsky  <razya@il.ibm.com>
2018
2019         * ipa-cp.c (ipcp_insert_stage, ipcp_driver): Support for SSA.
2020         (ipcp_driver): Change to static definition.
2021         Add dumping of the ifunctions.
2022         (constant_val_insert): Remove unused parameter. Support for SSA.
2023         (ipcp_propagate_const): Support for SSA.
2024         (ipcp_profile_bb_print): Print only analyzed nodes.
2025         (ipcp_replace_map_create): Remove support for Fortran constant
2026         for now.
2027         * ipa-prop.c (ipa_method_modify_stmt,
2028         ipa_callsite_compute_param): Support for SSA.
2029         * ipa-prop.h (ipcp_driver): Remove declaration.
2030         (IS_VALID_TREE_MAP_INDEX): Add define.
2031
2032 2007-01-24  Geoffrey Keating  <geoffk@apple.com>
2033
2034         * unwind-dw2.c (execute_stack_op): Handle DW_OP_swap.
2035
2036 2007-01-24  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
2037
2038         PR middle-end/30447
2039         * builtins.c (fold_builtin_cabs): Use MPFR to evaluate a
2040         constant argument to cabs and do it without checking for
2041         -funsafe-math-optimizations.
2042
2043 2007-01-24  Douglas Gregor  <dgregor@osl.iu.edu>
2044
2045         * c-common.h (RID_FIRST_CXX0X): New.
2046         (RID_LAST_CXX0X): New.
2047         * c-opts.c (c_common_handle_option): -Wc++0x-compat is triggered
2048         by -Wall.
2049         * c.opt (Wc++0x-compat): New.
2050         * doc/invoke.texi (-Wc++0x-compat): Document.
2051
2052 2007-01-24  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
2053
2054         * builtins.c (fold_builtin_carg): New.
2055         (fold_builtin_1): Use it.
2056
2057 2007-01-24  Jan Hubicka  <jh@suse.cz>
2058
2059         * ipa-inline.c (cgraph_decide_inlining): Initialize initial_insns.
2060
2061         * ipa-inline.c (initial_insns, max_insns): Delete.
2062         (compute_max_insns): New function.
2063         (cgraph_decide_inlining_of_small_function): Use it; take minimal amount
2064         of insns as base for code growth.
2065         (cgraph_decide_inlining): Make initial_insns local; do not compute
2066         max_insns.
2067         * params.def (PARAM_INLINE_UNIT_GROWTH): Set to 60.
2068         * doc/invoke.texi (inline-unit-growth): Update docs.
2069
2070 2007-01-24  Jakub Jelinek  <jakub@redhat.com>
2071
2072         * config/i386/i386.h (x86_cmpxchg16b): Remove const.
2073         (TARGET_CMPXCHG16B): Define to x86_cmpxchg16b.
2074         * config/i386/i386.c (x86_cmpxchg16b): Remove const.
2075         (override_options): Add PTA_CX16 flag.  Set x86_cmpxchg16b
2076         for CPUs that have PTA_CX16 set.
2077
2078         PR middle-end/27416
2079         * gimplify.c (omp_check_private): New function.
2080         (gimplify_scan_omp_clauses): Use it for
2081         firstprivate/lastprivate/reduction.
2082
2083         PR middle-end/30494
2084         * gimplify.c (omp_add_variable): Don't call omp_notice_variable
2085         on TYPE_SIZE_UNIT for GOVD_LOCAL VLAs.
2086
2087         PR middle-end/30421
2088         * omp-low.c (lower_omp_for_lastprivate): Add dlist argument.
2089         If lower_lastprivate_clauses emits some statements, append them
2090         to dlist rather than body_p and to body_p append an initializer.
2091         (lower_omp_for): Adjust caller.
2092
2093 2007-01-24  Steve Ellcey  <sje@cup.hp.com>
2094
2095         * target.h (globalize_decl_name): New.
2096         * target-def.h (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
2097         * output.h (default_globalize_decl_name): New.
2098         * varasm.c (asm_output_bss): Use globalize_decl_name instead of
2099         globalize_label.
2100         (globalize_decl): Ditto.
2101         (default_globalize_decl_name): New.
2102         * config/ia64/ia64.c (ia64_globalize_decl_name): New.
2103         (ia64_handle_version_id_attribute): New.
2104         (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
2105         (ia64_asm_output_external): Use globalize_decl_name instead
2106         of globalize_label.
2107         * doc/extend.texi (version_id): New pragma.
2108         * doc/tm.texi (ARGET_ASM_GLOBALIZE_DECL_NAME): New target hook.
2109
2110 2007-01-24  Andreas Krebbel  <krebbel1@de.ibm.com>
2111
2112         * unwind-dw2-fde.c (get_cie_encoding): Replaced _Unwind_Word with
2113         _uleb128_t and _Unwind_SWord with _sleb128_t.
2114         * unwind-dw2.c (extract_cie_info, execute_stack_op, execute_cfa_program,
2115         uw_frame_state_for, uw_update_context_1): Likewise.
2116         * unwind-c.c (parse_lsda_header, PERSONALITY_FUNCTION): Likewise.
2117         * unwind-pe.h (read_uleb128, read_sleb128,
2118         read_encoded_value_with_base): Likewise.
2119         * unwind-generic.h: Define _sleb128_t and _uleb128_t types.
2120
2121 2007-01-24  Richard Guenther  <rguenther@suse.de>
2122
2123         * builtins.c (expand_builtin_cexpi): Get the fndecl
2124         for cexp in the correct way.
2125
2126 2007-01-24  Jan Hubicka  <jh@suse.cz>
2127
2128         * tree-ssa-dce.c (eliminate_unnecesary_stmts): Remove dead LHS of calls.
2129
2130 2007-01-24  Andreas Krebbel  <krebbel1@de.ibm.com>
2131
2132         * c-cppbuiltin.c (builtin_define_type_sizeof): New function.
2133         (c_cpp_builtins): New builtin macros: __SIZEOF_INT__, __SIZEOF_LONG__,
2134         __SIZEOF_LONG_LONG__, __SIZEOF_SHORT__, __SIZEOF_POINTER__,
2135         __SIZEOF_FLOAT__, __SIZEOF_DOUBLE__, __SIZEOF_LONG_DOUBLE__,
2136         __SIZEOF_SIZE_T__, __SIZEOF_WCHAR_T__, __SIZEOF_WINT_T__ and
2137         __SIZEOF_PTRDIFF_T__.
2138         * doc/cpp.texi: Documentation for the new builtin macros added.
2139
2140 2007-01-24  Uros Bizjak  <ubizjak@gmail.com>
2141
2142         * config/i386/i386.md (tanxf2, tan<mode>2, atan<mode>2, log<mode>2,
2143         log10<mode>2, log2<mode>2, expxf2, exp10xf2, exp2xf2): Use op2
2144         instead of operands[2] to avoid access past the end of array.
2145
2146 2007-01-24  Richard Sandiford  <richard@codesourcery.com>
2147
2148         * reload1.c (emit_reload_insns): Pass the reload register
2149         for a non-spill output reload through forget_old_reloads_1.
2150
2151 2007-01-23  Joseph Myers  <joseph@codesourcery.com>
2152
2153         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Define
2154         _SOFT_DOUBLE if doubles use software floating-point.
2155         * config/rs6000/libgcc-ppc-glibc.ver: Export additional long
2156         double functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
2157         * config/rs6000/darwin-ldouble.c: Also compile functions for
2158         hard-float without FPRs.  Use fmsub function for all __NO_FPRS__
2159         cases.  Compile extra functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
2160         * config/rs6000/linuxspe.h (SUBSUBTARGET_OVERRIDE_OPTIONS): Remove
2161         commented-out long double override.
2162         (CPP_LONGDOUBLE_DEFAULT_SPEC): Likewise.
2163         * config/rs6000/eabispe.h: Likewise.
2164         * config/rs6000/rs6000.c (rs6000_override_options): Don't override
2165         long double for non-SPE.
2166         (rs6000_handle_option): Likewise.
2167         (invalid_e500_subreg): Disallow more subregs involding DImode,
2168         DFmode, TImode or TFmode.
2169         (rs6000_legitimate_offset_address_p): Check TFmode offsets for
2170         E500 double.
2171         (legitimate_lo_sum_address_p): Also check for TFmode for E500
2172         double.
2173         (rs6000_legitimize_address): Also handle TFmode for E500 double.
2174         (rs6000_legitimize_reload_address): Also handle TFmode for E500
2175         double.
2176         (rs6000_legitimate_address): Also check for TFmode for E500
2177         double.
2178         (rs6000_emit_move): Use DFmode subregs of TFmode for E500 double.
2179         (spe_build_register_parallel): Handle TFmode and TCmode.
2180         (rs6000_spe_function_arg): Handle TFmode and TCmode for E500
2181         double.
2182         (function_arg): Handle TFmode and TCmode for E500 double.
2183         (rs6000_init_libfuncs): Initialize extra libfuncs for soft double
2184         in general.
2185         (print_operand): Handle TFmode and TImode for %y.
2186         (rs6000_generate_compare): Handle TFmode comparisons for E500
2187         double.
2188         (spe_func_has_64bit_regs_p): Check for TFmode for E500 double.
2189         (rs6000_function_value): Handle TFmode and TCmode for E500 double.
2190         (rs6000_libcall_value): Handle TFmode and TCmode for E500 double.
2191         * config/rs6000/rs6000.h (CANNOT_CHANGE_MODE_CLASS): Check for
2192         TFmode for E500 double.
2193         * config/rs6000/rs6000.md (FP): Allow TF for E500 double.
2194         (floatsidf2): Enable for E500 double.
2195         (movtf_softfloat): Use rs6000_nonimmediate_operand.
2196         (extenddftf2): Change to extenddftf2_fprs.
2197         (extenddftf2): Call gen_spe_extenddftf2 or gen_extenddftf2_fprs
2198         depending on TARGET_E500_DOUBLE.
2199         (extendsftf2): Enable for E500 double.
2200         (trunctfdf2): Enable for E500 double.
2201         (trunctfsf2): Change to trunctfsf2_fprs.
2202         (trunctfsf2): Call gen_spe_trunctfsf2 or gen_trunctfsf2_fprs
2203         depending on TARGET_E500_DOUBLE.
2204         (floatsitf2): Enable for E500 double.
2205         (fix_trunctfsi2): Change to fix_trunctfsi2_fprs.
2206         (fix_trunctfsi2): Call gen_spe_fix_trunctfsi2 or
2207         gen_fix_trunctfsi2_fprs depending on TARGET_E500_DOUBLE.
2208         (negtf2): Change to negtf2_internal.
2209         (negtf2): New expander.
2210         (abstf2): Enable for E500 double.  Call gen_spe_abstf2_tst,
2211         gen_spe_abstf2_cmp or gen_abstf2_internal depending on
2212         TARGET_E500_DOUBLE and flag_unsafe_math_optimizations.
2213         (movdi_internal32): Use rs6000_nonimmediate_operand.
2214         (unnamed splitter): Likewise.
2215         * config/rs6000/spe.md (CMPTFEQ_GPR, TSTTFEQ_GPR, CMPTFGT_GPR,
2216         TSTTFGT_GPR, CMPTFLT_GPR, TSTTFLT_GPR): New unspecs.
2217         (SPE64TF, DITI): New mode macros.
2218         (frob_df_di): Change to frob_<SPE64:mode>_<DITI:mode>; allow more
2219         modes.
2220         (frob_tf_ti): New.
2221         (frob_<mode>_di_2): New.
2222         (frob_tf_di_8_2): New.
2223         (frob_di_df): Change to frob_di_<mode>; allow more modes.
2224         (frob_ti_tf): New.
2225         (frob_di_df_2): Change to frob_<DITI:mode>_<SPE64:mode>_2; allow
2226         more modes.
2227         (frob_ti_<mode>_8_2): New.
2228         (frob_ti_tf_2): New.
2229         (mov_si<mode>_e500_subreg0, mov_si<mode>_e500_subreg0_2,
2230         mov_si<mode>_e500_subreg4, mov_si<mode>_e500_subreg4_2): Allow
2231         TFmode.
2232         (mov_sitf_e500_subreg8, mov_sitf_e500_subreg8_2,
2233         mov_sitf_e500_subreg12, mov_sitf_e500_subreg12_2): New.
2234         (spe_trunctfdf2_internal1, spe_trunctfsf2, spe_extenddftf2,
2235         spe_fix_trunctfsi2, spe_fix_trunctfsi2_internal,
2236         spe_negtf2_internal, spe_abstf2_cmp, spe_abstf2_tst): New.
2237         (cmptfeq_gpr, tsttfeq_gpr, cmptfgt_gpr, tsttfgt_gpr, cmptflt_gpr,
2238         tsttflt_gp): New.
2239
2240 2007-01-23  Ian Lance Taylor  <iant@google.com>
2241
2242         * Makefile.in (OBJS-common): Reformat, alphabetize, but put
2243         insn-*.o first.
2244         (OBJS-archive): Reformat, alphabetize.
2245         (OBJS): Change out_object_file to OBJS-md.
2246
2247 2007-01-23  Uros Bizjak  <ubizjak@gmail.com>
2248
2249         * config/i386/i386.md: Use REG_P, MEM_P, CONST_INT_P, LABEL_P,
2250         JUMP_P and CALL_P predicates where applicable.
2251         * config/i386/i386.c: Ditto.
2252         * config/i386/i386.md: Ditto.
2253         * config/i386/mmx.md: Ditto.
2254         * config/i386/predicates.md: Ditto.
2255
2256 2007-01-22  Andreas Schwab  <schwab@suse.de>
2257
2258         * config/m68k/m68k.h: Fix comment.
2259
2260 2007-01-22  Jan Hubicka  <jh@suse.cz>
2261
2262         * passes.c (init_optimization_passes): Do not rerun
2263         pass_early_warn_uninitialized.
2264
2265 2007-01-22  Richard Guenther  <rguenther@suse.de>
2266
2267         PR tree-optimization/30038
2268         * tree-ssa-math-opts.c (maybe_record_sincos): New static helper
2269         function.
2270         (execute_cse_sincos_1): Likewise.
2271         (execute_cse_sincos): Likewise.
2272         (gate_cse_sincos): Likewise.
2273         (pass_cse_sincos): New pass CSEing sin() and cos() calls using
2274         the cexpi() canonicalization of sincos().
2275         * tree-pass.h (pass_cse_sincos): Declare.
2276         * passes.c (init_optimization_passes): New pass pas_cse_sincos.
2277
2278 2007-01-21  Eric Botcazou  <ebotcazou@libertysurf.fr>
2279
2280         PR rtl-optimization/29329
2281         * combine.c (replaced_rhs_insn): Rename to i2mod.
2282         (replaced_rhs_value): Rename to i2mod_new_rhs.
2283         (i2mod_old_rhs): New global variable.
2284         (combine_instructions): Adjust for above change.  Save a copy of
2285         the old RHS into i2mod_old_rhs when the contents of a REG_EQUAL
2286         note are substituted in the second instruction.
2287         (distribute_notes) <REG_DEAD>: Adjust for above change.  Do not
2288         ditch the note if it pertains to the second eliminated register
2289         and this register is mentioned in i2mod_old_rhs.
2290
2291         Revert:
2292         2006-09-12  Eric Botcazou  <ebotcazou@libertysurf.fr>
2293
2294         * combine.c (distribute_notes) <REG_DEAD>: Do not consider SETs past
2295         the insn to which the note was originally attached.
2296
2297 2007-01-21  Jan Hubicka  <jh@suse.cz>
2298
2299         * ipa-inline.c (inlining_mode): Comment, move up.
2300         (cgraph_decide_inlining_incrementally): Do not perform inlining
2301         itself; fix handling of flattening of self recursive functions.
2302         (cgraph_find_cycles): Remove.
2303         (cgraph_flatten_node): Remove.
2304         (cgraph_decide_inlining): Use incremental inliner to handle flattening.
2305         (try_inline): New function.
2306         (cgraph_early_inlining): Update call of
2307         cgraph_decide_inlining_incrementally.  Apply inlining here.
2308         (apply_inline): Update call of cgraph_decide_inlining_incrementally.
2309
2310 2007-01-21  Dirk Mueller  <dmueller@suse.de>
2311
2312         PR bootstrap/30511
2313         * tree-vrp.c (check_array_bounds): do not warn
2314         about ADDR_EXPR's of ARRAY_REF's which are immediately
2315         used in binary expressions.
2316
2317 2007-01-21  Ira Rosen  <irar@il.ibm.com>
2318
2319         * tree-vectorizer.h (struct _stmt_vec_info): Add new field
2320         read_write_dep and macros for its access.
2321         * tree-vectorizer.c (new_stmt_vec_info): Initialize the new field.
2322         * tree-vect-analyze.c (vect_analyze_data_ref_dependence): Remove
2323         argument, call vect_check_interleaving for every independent pair of
2324         data-refs. Mark loads that access the same memory location as a store
2325         in the loop.
2326         (vect_check_dependences): Remove.
2327         (vect_analyze_data_ref_dependences): Remove  vect_check_dependences
2328         call, fix the call to vect_analyze_data_ref_dependence.
2329         (vect_analyze_data_ref_access): For statements that access the same
2330         data-ref, check that they are not stores; for loads, check that there
2331         is no store that access the same location.
2332
2333 2007-01-20  Roger Sayle  <roger@eyesopen.com>
2334             Joseph Myers  <joseph@codesourcery.com>
2335
2336         * doc/invoke.texi (-fdump-rtl-combine): Fix under/overfull hbox.
2337         (-fdump-rtl-stack): Likewise.
2338         (-fno-signed-zeros): Use @minus{} for a minus sign.  Correct typo.
2339         (-mcheck-zero-division, -mcpu): Fix under/overfull hbox.
2340         (-mpt-fixed): Use @minus{} for minus sign.
2341         (Using Precompiled Headers): Fix under/overfull hbox.
2342
2343 2007-01-20  Jan Hubicka  <jh@suse.cz>
2344
2345         * tree-flow.h (struct stmt_ann_d): Move references_memory to proper
2346         place within annotation.
2347
2348 2007-01-20  John David Anglin  <dave.anglin@nrc-cnrc.gc.ca>
2349
2350         * pa.c (output_move_double): Change array size of xoperands to 4.
2351
2352 2007-01-20  Jan Hubicka  <jh@suse.cz>
2353
2354         * tree-tailcall.c (adjust_return_value): Do not use RESULT_DECL
2355         as temporary.
2356
2357 2007-01-19  Ian Lance Taylor  <iant@google.com>
2358
2359         * expmed.c (expand_divmod) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]: Cast
2360         constant to unsigned HOST_WIDE_INT before negating.
2361
2362 2007-01-19  Ian Lance Taylor  <iant@google.com>
2363
2364         * tree-ssa-operands.h (struct vuse_vec_d): Change num_vuse field
2365         to unsigned.
2366         (VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Use unsigned comparison.
2367         (VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
2368         (SET_VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Likewise.
2369         (SET_VUSE_ELEMENT_VAR) [ENABLE_CHECKING]: Likewise.
2370         (SET_VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
2371         (realloc_vdef, realloc_vuse): Change second parameter to
2372         unsigned.
2373         (ssa_operand_iterator_d): Change vuse_index and mayuse_index
2374         fields to unsigned.
2375         * tree-ssa-operands.c (realloc_vop): Change num_elem parameter to
2376         unsigned. Change x and lim locals to unsigned.
2377         (realloc_vdef, realloc_vuse): Change num_elem parameter to
2378         unsigned.
2379         (finalize_ssa_vuse_ops): Change old_i local to unsigned.
2380         (copy_virtual_operands): Change i and n locals to unsigned.
2381
2382 2007-01-19  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
2383
2384         PR middle-end/29335
2385         * builtins.c (fold_builtin_1): Handle builtin fdim.
2386
2387 2007-01-20  Jan Hubicka  <jh@suse.cz>
2388
2389         * tree-ssa.c (init_tree_ssa): Do not call init_alias_heapvars.
2390         * tree-ssa-structalias.c (compute_points_to_sets): Do call
2391         init_alias_heapvars.
2392         (init_alias_heapvars): Initialize only when not already initialized.
2393         (delete_alias_heapvars): Set heapvar_for_stmt to NULL.
2394
2395 2007-01-19  Roger Sayle  <roger@eyesopen.com>
2396
2397         * common.opt (fsigned-zeros): New command line option.
2398         * flags.h (HONOR_SIGNED_ZEROS): Control via flag_signed_zeros instead
2399         of flag_unsafe_math_optimizations.
2400         * opts.c (set_fast_math_flags): The -ffast-math command line option
2401         implies -fno-signed-zeros.
2402         (fast_math_flags_set_p): Likewise.
2403
2404         * doc/invoke.texi: Document new -fno-signed-zeros option, and update
2405         the documentation of -ffast-math appropriately.  Wrap long lines.
2406
2407 2007-01-19  Steve Ellcey  <sje@cup.hp.com>
2408
2409         * system.h (ASM_MAKE_LABEL_LINKONCE): Poison.
2410         * varasm.c (globalize_decl): Remove ASM_MAKE_LABEL_LINKONCE ifdef.
2411
2412 2007-01-19  Tomas Bily  <tbily@suse.cz>
2413
2414         * cgraphunit.c (cgraph_finalize_function): Updating of pid
2415         * tree-profile.c:
2416         (tree_init_ic_make_global_vars): New function
2417         (tree_init_edge_profiler): call of tree_init_ic_make_global_vars
2418         (tree_gen_ic_profiler): New function
2419         (tree_gen_ic_func_profiler): New function
2420         (tree_profiling): Added calling of tree_gen_ic_func_profiler
2421         (tree_profile_hooks): Added hook for indirec/virtual calls
2422         * value-prof.c (tree_find_values_to_profile): New case for
2423         indirect calls
2424         (tree_values_to_profile): Call for determining indirect/virtual
2425         counters
2426         (tree_indirect_call_to_profile): New function
2427         (tree_ic_transform): New function
2428         (tree_ic): New function
2429         (find_func_by_pid): New function
2430         (init_pid_map): New function
2431         (tree_value_profile_transformations): Added check for
2432         indirect/virtual call transformation
2433         * value-prof.h (enum hist_type): New counter type for
2434         indirect/virtual calls
2435         (profile_hooks): Added new hook for profiling indirect/virtual
2436         calls
2437         * profile.c (instrument_values): New case for indirect/virtual
2438         call added
2439         * gcov-io.h (GCOV_LAST_VALUE_COUNTER): Changed to 6
2440         (GCOV_COUNTER_V_INDIR): New counter type
2441         (GCOV_COUNTER_NAMES): New name of counter "indirect" added
2442         (GCOV_MERGE_FUNCTIONS): New merge function for indirect/virtual
2443         call added
2444         * cgraph.c: Definition of cgraph_max_pid
2445         (cgraph_create_node): Default init of pid attribute
2446         * cgraph.h: Declaration of cgraph_max_pid
2447         (struct cgraph_node): Added pid attribute
2448         * libgcov.c (__gcov_indirect_call_profiler): New function
2449         (__gcov_one_value_profiler_body): New function
2450         (__gcov_one_value_profiler): Body was moved to
2451         __gcov_one_value_profiler_body and calls it
2452
2453 2007-01-19  Basile Starynkevitch  <basile@starynkevitch.net>
2454
2455         * doc/gty.texi (Options): Document the mark_hook option to GTY.
2456         * gengtype.c (write_types_data, write_func_for_structure,
2457         write_types, ggc_wtd, pch_wtd): Add skip_hooks to
2458         write_types_data, ggc_wtd, pch_wtd for processing mark_hook.
2459         (walk_type, write_func_for_structure): Generate the mark_hook if
2460         needed.
2461
2462 2007-01-19  Jan Hubicka  <jh@suse.cz>
2463
2464         * ipa-inline.c (cgraph_decide_inlining_incrementally): Instead of
2465         'early' argument take inlining mode argument specifying whether to
2466         inline for size/speeed or all functions; add support for flattening;
2467         improve dumpting.
2468         (cgraph_early_inlining): Update call of decide_inlining_incrementally.
2469
2470 2007-01-19  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
2471
2472         PR c++/17947
2473         * toplev.c (warn_deprecated_use): Use %qD instead of %qs to print
2474         the name of the declared identifier.
2475
2476 2007-01-19  Dirk Mueller  <dmueller@suse.de>
2477
2478         * config/i386.h (CONDITIONAL_REGISTER_USAGE): Store
2479         result of PIC_OFFSET_TABLE_REGNUM in temporary variable to avoid
2480         duplicate evaluation.
2481
2482 2007-01-19  Uros Bizjak  <ubizjak@gmail.com>
2483
2484         * config/i386/i386.md (acos<mode>2): Rename from acossf2 and acosdf2.
2485         Macroize expander using X87MODEF12 mode macro.  Extend operand 1
2486         to XFMode, use acosxf2 and truncate result to requested mode.
2487         Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
2488         (asin<mode>2): Similarly, with asin expanders.
2489         (*fscalexf4_i387): Rename from *fscalexf4.
2490         (expNcorexf3): New expander.
2491         (expxf2, exp10xf2, exp2xf2): Use expNcorexf3 expander.
2492         (exp<mode>2): Rename from expsf2 and expdf2. Macroize expander using
2493         X87MODEF12 mode macro.  Extend operand 1 to XFMode, use expxf2 and
2494         truncate result to requested mode. Use SSE_FLOAT_MODE_P to disable
2495         patterns for SSE math.
2496         (exp10<mode>2): Similarly, with exp10 expanders.
2497         (exp2<mode>2): Similarly, with exp2 expanders.
2498         (expm1<mode>2): Similarly, with expm1 expanders.
2499         (ldexp<mode>3): Similarly, with ldexp expanders.
2500         (log<mode>2, log10<mode>2, log2<mode>2, log1p<mode>2, logb<mode>2):
2501         Use gen_truncxf<mode>2_i387_noop to truncate result.
2502
2503 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
2504
2505         * config/m68k/m68k.h (PREFERRED_STACK_BOUNDARY): Define to 32
2506         for ColdFire targets.
2507
2508 2007-01-19  Nathan Sidwell  <nathan@codesourcery.com>
2509             Richard Sandiford  <richard@codesourcery.com>
2510
2511         * config/m68k/m68k.h (M68K_STATIC_CHAIN_REG_NAME): New macro.
2512         (INITIALIZE_TRAMPOLINE): Use STATIC_CHAIN_REGNUM.
2513         (__transfer_from_trampoline): Use M68K_STATIC_CHAIN_REG_NAME.
2514         * config/m68k/m68kelf.h (STATIC_CHAIN_REGNUM): Override.
2515         (M68K_STATIC_CHAIN_REG_NAME): Likewise.
2516         * config/m68k/netbsd-elf.h (M68K_STATIC_CHAIN_REG_NAME): Likewise.
2517
2518 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
2519
2520         * config/m68k/m68k.md (adddi_dilshr32): Rename to...
2521         (*adddi_dilshr32): ...this.  Fix formatting.  Remove commented-out
2522         non-canonical pattern.  Restrict to !TARGET_COLDFIRE.
2523         (*adddi_dilshr32_cf): New pattern.
2524         (adddi3, subdi3): Remove first alternatives.
2525
2526 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
2527
2528         * config/m68k/m68k.c (notice_update_cc): If an SFmode move is
2529         implemented using move.l, do not use its cc result for floating-point
2530         comparisons.
2531
2532 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
2533
2534         * config/m68k/m68k.h (EXTRA_CONSTRAINT): Stop the 'T' constraint
2535         from accepting 's' constraints if flag_pic.
2536
2537 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
2538
2539         * config/m68k/m68k.md (bordered, bunordered, buneq, bunge, bungt)
2540         (bunle, bunlt, bltgt, bordered_rev, bunordered_rev, buneq_rev)
2541         (bunge_rev, bungt_rev, bunle_rev, bunlt_rev, bltgt_rev): Change
2542         condition from TARGET_68881 to TARGET_HARD_FLOAT.
2543
2544 2007-01-19  Sandra Loosemore  <sandra@codesourcery.com>
2545
2546         * longlong.h (count_leading_zeros, COUNT_LEADING_ZEROS_0): Add
2547         ColdFire alternatives.
2548         * config/m68k/m68k.h (CLZ_DEFINED_VALUE_AT_ZERO): New macro.
2549         * config/m68k/m68k.md (clzsi2):  Define for ColdFire
2550         architectures that support ff1 instruction.
2551
2552 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
2553             Julian Brown  <julian@codesourcery.com>
2554
2555         * config/m68k/m68k.h (CONST_OK_FOR_LETTER_P): Add an 'R' case.
2556         * config/m68k/m68k.md (*movsi_cfv4): Fold into...
2557         (*movsi_cf): ...here.  Remove unnecessary 'R' from 'Rg'.
2558         Add commentary.
2559
2560 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
2561
2562         * config/m68k/m68k-protos.h (valid_mov3q_const): Take a HOST_WIDE_INT
2563         and return a bool.
2564         (output_move_const_into_data_reg, output_move_simode_const): Delete.
2565         * config/m68k/m68k.c (const_method, const_int_cost): Take a
2566         HOST_WIDE_INT instead of an rtx.
2567         (m68k_rtx_costs): Update call accordingly.
2568         (output_move_const_into_data_reg): Likewise.  Fix formatting.
2569         (valid_mov3q_const): Take a HOST_WIDE_INT instead of an rtx.
2570         Return a bool.
2571         (output_move_simode_const): Update calls after above changes.
2572         Rework to use automatic variables and predicates like MEM_P.
2573         * config/m68k/m68k.md (pushexthisi_const): Update call to
2574         valid_mov3q_const.
2575
2576 2007-01-19  Dirk Mueller  <dmueller@suse.de>
2577
2578         * tree-ssa-alias.c (perform_var_substitution): Fix typo
2579         in dump_flags test.
2580
2581 2007-01-19  Richard Guenther  <rguenther@suse.de>
2582
2583         * builtins.c (expand_builtin_cexpi): Fall back to expanding
2584         via cexp in case sincos is not available.
2585
2586 2007-01-19  Richard Guenther  <rguenther@suse.de>
2587
2588         * doc/tm.texi (TARGET_HAS_SINCOS): Document new target macro.
2589         * defaults.h (TARGET_HAS_SINCOS): Default to off.
2590         * config/linux.h (TARGET_HAS_SINCOS): Set to on if we have glibc.
2591         * config/alpha/linux.h (TARGET_HAS_SINCOS): Likewise.
2592         * config/sparc/linux.h (TARGET_HAS_SINCOS): Likewise.
2593         * config/sparc/linux64.h (TARGET_HAS_SINCOS): Likewise.
2594         * config/rs6000/linux.h (TARGET_HAS_SINCOS): Likewise.
2595         * config/rs6000/linux64.h (TARGET_HAS_SINCOS): Likewise.
2596
2597 2007-01-19  Uros Bizjak  <ubizjak@gmail.com>
2598
2599         * config/i386/i386.md (*fpatanxf3_i387, fpatan_extend<mode>xf3_i387):
2600         New insn patterns.
2601         (atan2sf3_1, atan2df3_1, atan2xf3_1): Remove insn patterns.
2602         (atan2xf3): Directly generate RTL pattern.
2603         (atan2<mode>3): Rename from atan2sf3 and atan2df3 and macroize insn
2604         patterns using X87MODEF12 mode macro.  Use fpatan_extend<mode>xf3_i387
2605         and truncate result to requested mode.  Use SSE_FLOAT_MODE_P to
2606         disable patterns for SSE math.
2607         (atan<mode>2): Rename from atansf2 and atandf2 and macroize insn
2608         patterns using X87MODEF12 mode macro.  Use fpatan_extend<mode>xf3_i387
2609         and truncate result to requested mode.  Use SSE_FLOAT_MODE_P to
2610         disable patterns for SSE math.
2611
2612 2007-01-19  Alexandre Oliva  <aoliva@redhat.com>
2613
2614         * libgcc-std.ver: Fix typo in %inherit for GCC_4.3.0.
2615
2616 2007-01-18  Roger Sayle  <roger@eyesopen.com>
2617
2618         * fold-const.c (fold_unary) <VIEW_CONVERT_EXPR>: Optimize away a
2619         VIEW_CONVERT_EXPR to the same type as it's operand.
2620
2621 2007-01-18  David Edelsohn  <edelsohn@gnu.org>
2622
2623         * config/rs6000/darwin-ldouble.c: Only build _SOFT_FLOAT if
2624         configured for long double 128.
2625
2626 2007-01-18  Mike Stump  <mrs@apple.com>
2627
2628         * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Fix build
2629         error.
2630
2631 2007-01-18  Michael Meissner  <michael.meissner@amd.com>
2632
2633         * i386.c (ix86_compute_frame_layout): Make fprintf's in #if 0 code
2634         type correct.
2635
2636 2007-01-18  Jan Hubicka  <jh@suse.cz>
2637
2638         * tree-ssa-operands.c (vop_free_bucket_size): Never return value
2639         greater than NUM_VOP_FREE_BUCKETS.
2640
2641 2007-01-18  Daniel Berlin  <dberlin@dberlin.org>
2642
2643         * tree-ssa-structalias.c: Update comments.
2644         (ptabitmap_obstack): Removed.
2645         (pta_obstack): New.
2646         (oldpta_obstack): Ditto.
2647         (stats): Add a few members.
2648         (struct variable_info): Remove node, complex, address_taken, and
2649         indirect_target members. Add oldsolution member.
2650         (new_var_info): Do not initialize removed members.
2651         (constraint_expr_type): Remove INCLUDES.
2652         (constraint_graph): Add size, implicit_preds, rep,
2653         indirect_cycles, eq_rep, label, direct_nodes, and complex members.
2654         (FIRST_REF_NODE): New macro.
2655         (LAST_REF_NODE): Ditto.
2656         (FIRST_ADDR_NODE): Ditto.
2657         (find): New function.
2658         (unite): Ditto.
2659         (dump_constraint): Do not handle INCLUDES.
2660         (insert_into_complex): Do not insert duplicate constraints.
2661         (condense_varmap_nodes): Renamed and rewritten into ...
2662         (merge_node_constraints): This. Also fix bug in handling of
2663         offseted copy constraints.
2664         (clear_edges_for_node): No longer need to deal with preds at all,
2665         or removing associated preds/succs.
2666         (merge_graph_nodes): Deal with indirect_cycles.
2667         Don't deal with predecessors.
2668         (add_implicit_graph_edge): New function.
2669         (add_pred_graph_edge): Ditto.
2670         (add_graph_edge): Don't deal with predecessors.
2671         (build_constraint_graph): Removed.
2672         (build_pred_graph): New function.
2673         (build_succ_graph): Ditto.
2674         (struct scc_info): Removed in_component. Added roots, dfs, and
2675         node_mapping. Remove visited_index, unification_queue.
2676         (scc_visit): Deal with union-find we do now.
2677         Deal with cycles with REF nodes.
2678         (collapse_nodes): Renamed and rewritten to ...
2679         (unify_nodes): This.
2680         (process_unification_queue): Removed.
2681         (topo_visit): Cleanup
2682         (do_da_constraint): Use find.
2683         (do_sd_constraint): Ditto.
2684         (do_ds_constraint): Ditto.
2685         (do_complex_constraint): Ditto.
2686         (init_scc_info): Update for removed and added members.
2687         (find_and_collapse_graph_cycles): Renamed and rewritten into ...
2688         (find_indirect_cycles): This.
2689         (equivalence_class): New variable.
2690         (label_visit): New function.
2691         (perform_variable_substitution): Rewritten.
2692         (free_var_substitution_info): New function.
2693         (find_equivalent_node): Ditto.
2694         (move_complex_constraints): Ditto.
2695         (eliminate_indirect_cycles): Ditto.
2696         (solve_graph): Only propagate changed bits.
2697         Use indirect cycle elimination.
2698         Use find.
2699         (tree_id_t): Rename to tree_vi_t, delete id member, add vi member.
2700         (tree_id_eq): Renamed to ...
2701         (tree_vi_eq): This. Update for member change
2702         (insert_id_for_tree): Renamed and rewritten to ...
2703         (insert_vi_for_tree): This.
2704         (lookup_id_for_tree): Renamed and rewritten to ...
2705         (lookup_vi_for_tree): This.
2706         (get_id_for_tree): Renamed and rewritten to ...
2707         (get_vi_for_tree): Ditto.
2708         (get_constraint_exp_from_ssa_var): Update to use get_vi_for_tree.
2709         (process_constraint): Don't handle INCLUDES.
2710         Remove special ADDRESSOF case.
2711         (find_func_aliases): Rewrite to use vi functions instead of id
2712         ones.
2713         (create_function_info_for): Ditto.
2714         (create_variable_info_for): Ditto.
2715         (intra_create_variable_infos): Ditto.
2716         (merge_smts_into): Ditto.
2717         (find_what_p_points_to): Ditto.
2718         (init_base_vars): Ditto.
2719         (init_alias_vars): Ditto.
2720         (remove_preds_and_fake_succs): New function.
2721         (dump_sa_points_to_info): Dump new stats.
2722         (dump_solution_for_var): Use find.
2723         (set_used_smts): Fix formatting.
2724         (compute_points_to_sets): Updated for new functions.
2725         (ipa_pta_execute): Ditto.
2726
2727 2007-01-18  Kazu Hirata  <kazu@codesourcery.com>
2728             Richard Sandiford  <richard@codesourcery.com>
2729
2730         * doc/tm.texi (TARGET_FUNCTION_VALUE): Expand documentation of
2731         parallels.
2732         * calls.c (expand_call): If the return value is a PARALLEL,
2733         extract its first member.
2734         * config/m68k/linux.h (FUNCTION_EXTRA_EPILOGUE): Remove.
2735         * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
2736         use FUNCTION_EXTRA_EPILOGUE.
2737         (m68k_function_value): Return a PARALLEL if the return value
2738         is of a pointer type.
2739         * config/m68k/netbsd-elf.h (current_function_returns_pointer)
2740         (FUNCTION_EXTRA_EPILOGUE): Remove.
2741         * config/m68k/m68k.md (D0_REG): New constant.
2742
2743 2007-01-18  Kazu Hirata  <kazu@codesourcery.com>
2744
2745         * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
2746         output a NOP for empty epilogues.
2747
2748 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
2749
2750         * config/m68k/m68k.c (m68k_use_return_insn): Update comments
2751         before function.  Extend register save check to include all
2752         registers, not just integer ones.
2753
2754 2007-01-18  Kazu Hirata  <kazu@codesourcery.com>
2755
2756         * config/m68k/m68k-protos.h (use_return_insn): Rename to...
2757         (m68k_use_return_insn): ...this.
2758         * config/m68k/m68k.h (USE_RETURN_INSN): Delete.
2759         * config/m68k/m68k.c (use_return_insn): Rename to...
2760         (m68k_use_return_insn): ...this.
2761         * config/m68k/m68k.md (return): Use m68k_use_return_insn instead
2762         of USE_RETURN_INSN.
2763
2764 2007-01-18  Nathan Sidwell  <nathan@codesourcery.com>
2765
2766         * config/m68k/fpgnulib.c (__truncdfsf2): Implement round to
2767         nearest even, fix denormal rounding overflow.
2768
2769 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
2770
2771         * config/m68k/m68k.md (movsf_cf_hard): Use fsmove instead of
2772         f%$smove and f%$move.
2773         (movdf_cf_hard): Use fdmove for cases 0 and 3 and fmove for case 1.
2774         (extendsfdf2_cf): Use fdmove instead of f%&move.
2775         (truncdfsf2_cf): Use fsmove instead of f%$smove.
2776         (add<mode>3_cf, sub<mode>3_cf): Use <FP:prec> instead of <FP:round>.
2777
2778 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
2779
2780         * config/m68k/m68k.md (movdf_cf_hard): Use output_move_double for
2781         GPR<-GPR moves.
2782
2783 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
2784
2785         * real.h (real_format): Add a canonical_nan_lsbs_set field.
2786         (coldfire_single_format): Declare.
2787         (coldfire_double_format): Likewise.
2788         * real.c (encode_ieee_single): Use canonical_nan_lsbs_set instead
2789         of qnan_msb_set to determine the lower bits of a canonical
2790         NaN significand.
2791         (encode_ieee_double): Likewise.
2792         (encode_ieee_quad): Likewise.
2793         (ieee_single_format): Initialize canonical_nan_lsbs_set.
2794         (mips_single_format): Likewise.
2795         (ieee_double_format): Likewise.
2796         (mips_double_format): Likewise.
2797         (ieee_extended_motorola_format): Likewise.
2798         (ieee_extended_intel_96_format): Likewise.
2799         (ieee_extended_intel_128_format): Likewise.
2800         (ieee_extended_intel_96_round_53_format): Likewise.
2801         (ibm_extended_format): Likewise.
2802         (mips_extended_format): Likewise.
2803         (ieee_quad_format): Likewise.
2804         (mips_quad_format): Likewise.
2805         (vax_f_format): Likewise.
2806         (vax_d_format): Likewise.
2807         (vax_g_format): Likewise.
2808         (i370_single_format): Likewise.
2809         (i370_double_format): Likewise.
2810         (decimal_single_format): Likewise.
2811         (decimal_double_format): Likewise.
2812         (decimal_quad_format): Likewise.
2813         (c4x_single_format): Likewise.
2814         (c4x_extended_format): Likewise.
2815         (real_internal_format): Likewise.
2816         (coldfire_single_format): New real_format.
2817         (coldfire_double_format): Likewise.
2818         * config/pdp11/pdp11.c (pdp11_f_format): Initialize
2819         canonical_nan_lsbs_set.
2820         (pdp11_d_format): Likewise.
2821         * config/m68k/m68k.c (override_options): Override REAL_FORMAT_MODE
2822         if TARGET_COLDFIRE_CPU.
2823
2824 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
2825
2826         * config/m68k/m68k-protos.h (m68k_output_pic_call): Delete.
2827         (output_call, m68k_legitimize_call_address): Declare.
2828         * config/m68k/m68k.h (EXTRA_CONSTRAINT): Remove unnecessary
2829         parenthesees.  Add support for a 'W' constraint.
2830         (LEGITIMATE_PIC_OPERAND_P): Remove SYMBOL_REF_FLAG handling.
2831         (PRINT_OPERAND_PUNCT_VALID_P): Remove comment about 'o'.
2832         (m68k_symbolic_call, m68k_symbolic_jump): Declare.
2833         * config/m68k/m68k.c (m68k_symbolic_call, m68k_symbolic_jump): New
2834         variables.
2835         (override_options): Initialize them.  Do not set flag_no_function_cse
2836         for TARGET_ID_SHARED_LIBRARY.
2837         (m68k_output_pic_call): Delete.
2838         (m68k_legitimize_call_address): New function.
2839         (print_operand): Remove the %o prefix.  Handle the %p prefix.
2840         (output_call): New function.
2841         (m68k_output_mi_thunk): Use m68k_symbolic_jump.  Always load the
2842         target address from the GOT if symbolic jumps are not allowed.
2843         * config/m68k/m68k.md (call, general_operand): Do not set
2844         SYMBOL_REF_FLAG.  Use m68k_legitimize_call_address instead.
2845         Merge separate flag_pic and !flag_pic define_insns into...
2846         (*call, *call_value): ...these new patterns.  Match the address
2847         rather than the containing MEM and require it to be a call_operand.
2848         Use output_call to generate the asm template.
2849         * config/m68k/predicates.md (const_call_operand): New predicate.
2850         (call_operand): Likewise.
2851
2852 2007-01-18  Nathan Sidwell  <nathan@codesourcery.com>
2853
2854         * config/m68k/m68k.h (REGISTER_MOVE_COST): Simplify definition.
2855         (STACK_GROWS_DOWNWARD): Define to 1.
2856         (FUNCTION_VALUE, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P): Equivocate
2857         comments, emphasizing that these values are only defaults.
2858         * config/m68k/linux.h (LINK_SPEC): Fix formatting in #undef.
2859         * config/m68k/m68k.c (const_method): Remove trailing whitespace.
2860
2861 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
2862
2863         * config/m68k/m68k.md (cmpsi): Remove outdated flag_pic handling.
2864
2865 2007-01-18  Kazu Hirata  <kazu@codesourcery.com>
2866             Richard Sandiford  <richard@codesourcery.com>
2867
2868         * config/m68k/m68k.h (DATA_REGNO_P, ADDRESS_REGNO_P, INT_REGNO_P)
2869         (FP_REGNO_P): New macros.
2870         (REGNO_OK_FOR_INDEX_P, REGNO_OK_FOR_BASE_P, REGNO_OK_FOR_DATA_P)
2871         (REGNO_OK_FOR_FP_P, REG_OK_FOR_INDEX_P, REG_OK_FOR_BASE_P): Use them.
2872         (EH_RETURN_STACKADJ_RTX): Use A0_REG.
2873         * config/m68k/m68k.c (m68k_regno_mode_ok): Use the new REGNO macros.
2874
2875 2007-01-18  Nathan Sidwell  <nathan@codesourcery.com>
2876
2877         * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
2878         (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-rtems*): Add t-floatlib
2879         to $tmake_file.
2880         * config/m68k/t-floatlib: New file, extracting common code from...
2881         * config/m68k/t-m68kbare, config/m68k/t-m68kelf,
2882         * config/m68k/t-uclinux: Here.
2883         * config/m68k/fpgnulib.c: Do not compile extendeed precision
2884         routines on ColdFire targets.
2885
2886 2007-01-18  Nathan Sidwell  <nathan@codesourcery.com>
2887
2888         * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
2889         (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*)
2890         (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-linux*)
2891         (m68k-*-rtems*): Use tm_file rather than m68k/m68k.h and
2892         explicitly set MOTOROLA to 1.
2893         * config/m68k/m68k.h (MOTOROLA): Simplify definition accordingly.
2894
2895 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
2896             Nathan Sidwell  <nathan@codesourcery.com>
2897
2898         * config/m68k/m68k.h (PCC_STATIC_STRUCT_RETURN): Delete.
2899         (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Add MOTOROLA cases.
2900         * config/m68k/coff.h (REGISTER_PREFIX_MD): Delete.
2901         * config/m68k/m68020-elf.h (LIB_SPEC): Delete.
2902         * config/m68k/m68k-none.h (CC1_SPEC, CPP_SUBTARGET_SPEC): Delete.
2903         * config/m68k/m68kelf.h (IMMEDIATE_PREFIX, REGISTER_PREFIX_MD)
2904         (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Delete.
2905         (NO_DOLLAR_IN_LABEL, PCC_STATIC_STRUCT_RETURN): Don't undefine.
2906         (BSS_ASM_OP, ASM_OUTPUT_SKIP, ASM_OUTPUT_ASCII): Delete.
2907         * config/m68k/m68kemb.h (PCC_STATIC_STRUCT_RETURN): Don't undefine.
2908         * config/m68k/linux.h (SIZE_TYPE, PTRDIFF_TYPE, WCHAR_TYPE)
2909         (WCHAR_TYPE_SIZE, TARGET_OBJFMT_CPP_BUILTINS): Delete.
2910         (TARGET_OS_CPP_BUILTINS): Don't define mc68000 and mc68020 here.
2911         (DBX_REGISTER_NUMBER): Delete.
2912         * config/m68k/netbsd-elf.h (IMMEDIATE_PREFIX): Delete.
2913         (PCC_STATIC_STRUCT_RETURN): Don't undefine.
2914         * config/m68k/openbsd.h (PCC_STATIC_STRUCT_RETURN): Define.
2915
2916 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
2917
2918         * config.gcc (m68k-*-uclinux*): Add flat.h to $tm_file.
2919         * config/flat.h: New file.
2920         * crtstuff.c (USE_PT_GNU_EH_FRAME): Don't define if
2921         OBJECT_FORMAT_FLAT.
2922         * config/m68k/m68k.h (ASM_PREFERRED_EH_DATA_FORMAT): Do not use
2923         indirect references for -msep-data or -mid-shared-library.
2924         Do not use PC-relative code addresses either.
2925         * config/m68k/m68k.c (override_options): Restrict -fPIC error
2926         to -mpcrel.
2927         * config/m68k/uclinux.h (STARTFILE_SPEC): Define.  Use Scrt1.o
2928         for shared libraries and crt1.o for executables.  Use crti.o and
2929         crtbegin.o.
2930         (ENDFILE_SPEC): Use crtend.o and crtn.o.
2931         (LIB_SPEC): Suppress -Rlibc.gdb if -static-libc is given.
2932         Do not add -elf2flt or -shared-lib-id options here.
2933         (LINK_SPEC): Define.  Pass -elf2flt if no -elf2flt option is given.
2934         Pass -shared-lib-id if -mid-shared-library, taking the library
2935         identifier from -mshared-library-id if given, otherwise
2936         defaulting to 0.
2937         (EH_FRAME_IN_DATA_SECTION): Do not undefine.
2938         (INIT_SECTION_ASM_OP, FINI_SECTION_ASM_OP): Likewise.
2939         (TARGET_OS_CPP_BUILTINS): Define __GXX_MERGED_TYPEINFO_NAMES=0
2940         and __GXX_TYPEINFO_EQUALITY_INLINE=0 if -mid-shared-library.
2941         (DRIVER_SELF_SPECS): Map unadorned PIC options to -msep-data.
2942         * config/m68k/t-uclinux (EXTRA_MULTILIB_PARTS): Add crtbegin.o
2943         and crtend.o.
2944         * config/m68k/lb1sf68.asm (PICCALL): Use an lea and pc-relative
2945         jump sequence for ISA A and ISA A+.
2946         (PICJUMP): Likewise.
2947
2948 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
2949
2950         * config.gcc (m68k-*-uclinux*): Base the port on the common
2951         and m68k GNU/Linux files rather than on the generic ELF ones.
2952         * config/m68k/uclinux.h (TARGET_VERSION): Override.
2953         (TARGET_OS_CPP_BUILTINS): Use LINUX_TARGET_OS_CPP_BUILTINS.
2954
2955 2007-01-18  Julian Brown  <julian@codesourcery.com>
2956             Richard Sandiford  <richard@codesourcery.com>
2957
2958         * config/m68k/m68k.h (LONG_DOUBLE_TYPE_SIZE): Make 64-bit on ColdFire.
2959         (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
2960         * config/m68k/netbsd-elf.h (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Undefine
2961         before redefining.
2962         * config/m68k/uclinux-oldabi.h (LONG_DOUBLE_TYPE_SIZE): Redefine to
2963         80 unconditionally.
2964         (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
2965
2966 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
2967
2968         * doc/install.texi: Document m68k-uclinuxoldabi.
2969         * config.gcc (m68k-*-uclinuxoldabi): New configuration.
2970         * config/m68k/uclinux-oldabi.h: New file, copied from
2971         config/m68k/uclinux.h.
2972
2973 2007-01-18  Kaz Kojima  <kkojima@rr.iij4u.or.jp>
2974
2975         * config/m32r/m32r-protos.h (m32r_expand_epilogue): Declare it.
2976         * config/m32r/m32r.c (m32r_setup_incoming_varargs): Use gen_frame_mem.
2977         (m32r_compute_frame_size): Use unsigned for regno.
2978         (m32r_reload_lr): Use gen_frame_mem.
2979         (pop): New.
2980         (m32r_output_function_epilogue): Don't output the function epilogue
2981         textually here.
2982         (m32r_expand_epilogue): New.
2983         (direct_return): Return false if the function has the interrupt
2984         attribute.
2985         (m32r_hard_regno_rename_ok): Remove code for the textual epilogue.
2986         * config/m32r/m32r.md (epilogue): New expander.
2987         (return_lr, return_rte): New insns.
2988         (return): Make it expander.
2989         (return_normal): New expander.
2990
2991 2007-01-18  Josh Conner  <jconner@apple.com>
2992
2993         PR target/30485
2994         * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Add
2995         support for UNLE, UNLT, UNGE, and UNGT.
2996
2997 2007-01-18  Jan Hubicka  <jh@suse.cz>
2998
2999         * tree-vrp.c (finalize_jump_threads): Do not call cleanup_cfg by hand.
3000         * tree-tailcall (add_virtual_phis): Likewise.
3001         (optimize_tail_call): Return TODOs.
3002         (execute_tail_calls): Return TODOs.
3003         * tree-ssa-ccp (execute_fold_all_builtins): Do cleanup_cfg via TODO.
3004         * tree-cfgcleanup.c (cleanup_tree_cfg_loop): Return if something
3005         changed.
3006         * tree-ssa-forwprop.c (tree_ssa_forward_propagate_single_use_value):
3007         Cleanup cfg using TODO.
3008         * tree-flow.h (cleanup_tree_cfg_loop): Update prototype.
3009         * passes.c (execute_function_todo): When cleanup did something, remove
3010         unused locals.
3011         * tree-cfg.c (pass_build_cfg): Add cleanup_cfg TODO.
3012         (make_edges): Don't cleanup_cfg.
3013
3014 2007-01-18  Uros Bizjak  <ubizjak@gmail.com>
3015
3016         * reg-stack.c (subst_stack_regs_pat) [UNSPEC_SINCOS_COS,
3017         UNSPEC_XTRACT_FRACT]: Use generic code for instructions that
3018         operate on the top of stack.
3019         [UNSPEC_SINCOS_SIN, UNSPEC_XTRACT_EXP, UNSPEC_TAN]: Rewrite
3020         register handling of instructions that output to the second
3021         stack slot.
3022         [UNSPEC_TAN_ONE, UNSPEC_TAN_TAN]: Remove.
3023         (move_for_stack_reg): Special-case check for dead destination
3024         stack slot for constant load of 1.0 inside UNSPEC_TAN.
3025
3026         * config/i386/i386.md (UNSPEC_TAN): New constant.
3027         (UNSPEC_TAN_ONE, UNSPEC_TAN_TAN): Remove.
3028         (fptanxf4_i387, fptan_extend<mode>xf4_i387): New patterns
3029         to correctly model move of constant 1.0 to top stack slot.
3030         (*tandf3_1, *tansf3_1, *tanxf3_1): Remove insn patterns.
3031         (unnamed peephole2 pattern): Remove corresponding peephole2
3032         pattern that optimizes tan insn and loading of constant 1.0.
3033         (tanxf2): Use fptanxf4_i387.
3034         (tan<mode>2): Rename from tansf2 and tandf2 and macroize insn
3035         patterns using X87MODEF12 mode macro.  Use fptan_extend<mode>xf4_i387
3036         and truncate result to requested mode.  Use SSE_FLOAT_MODE_P to
3037         disable patterns for SSE math.
3038         (sincos<mode>3): Use truncxf<mode>2_i387_noop for truncation.
3039         (fyl2x_extend<mode>xf3_i387): Use X87MODEF12 for operand 1.
3040
3041 2007-01-18  Dirk Mueller  <dmueller@suse.de>
3042             Richard Guenther <rguenther@suse.de>
3043
3044         PR diagnostic/8268
3045         * doc/invoke.texi (Warray-bounds): Document -Warray-bounds.
3046         * common.opt (Warray-bounds): Add new warning option.
3047         * c-opts.c (c_common_handle_option): Define -Warray-bounds
3048         if -Wall is given.
3049         * Makefile.in: make tree-vrp.o depend on toplev.h
3050         * tree-vrp.c (vrp_finalize): Call check_array_refs if -Warray-bounds
3051         is enabled.
3052         (check_array_refs, check_array_bounds, check_array_ref): New.
3053
3054 2007-01-18  Jan Hubicka  <jh@suse.cz>
3055
3056         * tree-ssa-ccp.c (ccp_finalize): Return if something changed.
3057         (execute_ssa_ccp): Return flags conditionally.
3058         * tree-ssa-propagate.c (substitue_and_fold): Return if something was
3059         changed.
3060         * tree-ssa-propagate.h (substitute_and_fold): Update prototype.
3061
3062 2007-01-18  Steven Bosscher  <steven@gcc.gnu.org>
3063
3064         * cfgcleanup.c (cleanup_cfg): Detect cfglayout mode and set
3065         the CLEANUP_CFGLAYOUT flag when in cfglayout mode.
3066
3067         * Makefile.c (GTFILES): Add cfglayout.h.
3068         * gengtype.c (open_base_files): Likewise.
3069         * cfglayout.c (cfg_layout_function_footer,
3070         cfg_layout_function_header) Reindent to make gengtype happy.
3071         * cfglayout.h (cfg_layout_function_footer,
3072         cfg_layout_function_header): Add GTY(()) marker.
3073
3074         * ifcvt.c (noce_try_sign_mask): Make sure INSN_B is non-null.
3075
3076 2007-01-18  Ben Elliston  <bje@au.ibm.com>
3077
3078         * genautomata.c (write_automata): Include xstrerror output in the
3079         error message if writing the DFA description file fails.
3080
3081 2007-01-17  H.J. Lu  <hongjiu.lu@intel.com>
3082
3083         * config/mips/mips-protos.h (mips_output_external): Make it
3084         return void.
3085         * config/mips/iris.h (TARGET_ASM_EXTERNAL_LIBCALL): Removed.
3086         * config/mips/mips.c (irix_output_external_libcall): Likewise.
3087         (extern_list): Likewise.
3088         (extern_head): Likewise.
3089         (TARGET_ASM_FILE_END): Likewise.
3090         (mips_file_end): Likewise.
3091         (mips_output_external): Rewritten.
3092
3093 2007-01-18  Ben Elliston  <bje@au.ibm.com>
3094
3095         * genpreds.c (write_insn_preds_c): Only write out the function
3096         body for regclass_for_constraint if we have register constraints.
3097
3098 2007-01-17  Tom Tromey  <tromey@redhat.com>
3099
3100         * doc/sourcebuild.texi (libgcj Tests): Use sourceware.org.
3101         * doc/install.texi (Testing): Use sourceware.org.
3102         (Binaries): Likewise.
3103         (Specific): Likewise.
3104         * doc/contrib.texi (Contributors): Use sourceware.org.
3105
3106 2007-01-17  Anatoly Sokolov <aesok@post.ru>
3107
3108         * config/avr/avr.h (AVR_HAVE_LPMX): New macro.
3109         (AVR_ENHANCED): Rename to ...
3110         (AVR_HAVE_MUL): ... new.
3111         (avr_enhanced_p): Rename to ...
3112         (avr_have_mul_p): ... new.
3113         (TARGET_CPU_CPP_BUILTINS): Use 'avr_have_mul_p' instead of
3114         'avr_enhanced_p' for "__AVR_ENHANCED__". Define "__AVR_HAVE_MUL__".
3115         * config/avr/avr.c (avr_enhanced_p): Rename to ...
3116         (avr_have_mul_p): ... new.
3117         (base_arch_s): Rename 'enhanced' to 'have_mul'.
3118         (avr_override_options): Use 'avr_have_mul_p' and 'have_mul' instead of
3119         'avr_enhanced_p' and 'enhanced'.
3120         (ashlhi3_out, ashrhi3_out, lshrhi3_out, avr_rtx_costs): Use
3121         AVR_HAVE_MUL instead of AVR_ENHANCED.
3122         * avr.md (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
3123         (mulqi3, *mulqi3_enh, *mulqi3_call, mulqihi3, umulqihi3, mulhi3,
3124         *mulhi3_enh, *mulhi3_call, mulsi3, *mulsi3_call): Use AVR_HAVE_MUL
3125         instead of AVR_ENHANCED.
3126         (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
3127         * libgcc.S: Use __AVR_HAVE_MUL__ instead of __AVR_ENHANCED__.
3128         (__tablejump__): Use __AVR_HAVE_LPMX__ instead of __AVR_ENHANCED__.
3129
3130 2007-01-17  Ian Lance Taylor  <iant@google.com>
3131
3132         * vec.h (VEC_reserve_exact): Define.
3133         (vec_gc_p_reserve_exact): Declare.
3134         (vec_gc_o_reserve_exact): Declare.
3135         (vec_heap_p_reserve_exact): Declare.
3136         (vec_heap_o_reserve_exact): Declare.
3137         (VEC_OP (T,A,reserve_exact)): New static inline function, three
3138         versions.
3139         (VEC_OP (T,A,reserve)) [all versions]: Remove handling of
3140         negative parameter.
3141         (VEC_OP (T,A,alloc)) [all versions]: Call ...reserve_exact.
3142         (VEC_OP (T,A,copy)) [all versions]: Likewise.
3143         (VEC_OP (T,a,safe_grow)) [all versions]: Likewise.
3144         * vec.c (calculate_allocation): Add exact parameter.  Change all
3145         callers.
3146         (vec_gc_o_reserve_1): New static function, from vec_gc_o_reserve.
3147         (vec_gc_p_reserve, vec_gc_o_reserve): Call vec_gc_o_reserve_1.
3148         (vec_gc_p_reserve_exact, vec_gc_o_reserve_exact): New functions.
3149         (vec_heap_o_reserve_1): New static function, from vec_heap_o_reserve.
3150         (vec_heap_p_reserve, vec_heap_o_reserve): Call vec_heap_o_reserve_1.
3151         (vec_heap_p_reserve_exact): New function.
3152         (vec_heap_o_reserve_exact): New function.
3153
3154 2007-01-17  Jan Hubicka  <jh@suse.cz>
3155
3156         * ipa-type-escape.c (look_for_casts): Revamp using handled_component_p.
3157
3158 2007-01-17  Eric Christopher  <echristo@apple.com>
3159
3160         * config.gcc: Support core2 processor.
3161
3162 2007-01-16  Jan Hubicka  <jh@suse.cz>
3163
3164         * tree-ssanames.c (release_dead_ssa_names): Instead of ggc_freeing
3165         the names, just unlink the chain so we don't crash on dangling pointers
3166         to dead SSA names.
3167
3168 2007-01-16  Jan Hubicka  <jh@suse.cz>
3169
3170         * cgraph.h (cgraph_decide_inlining_incrementally): Kill.
3171         * tree-pass.h: Reorder to make IPA passes appear toegher.
3172         (pass_early_inline, pass_inline_parameters, pass_apply_inline): Declare.
3173         * cgraphunit.c (cgraph_finalize_function): Do not compute inling
3174         parameters, do not call early inliner.
3175         * ipa-inline.c: Update comments.  Include tree-flow.h
3176         (cgraph_decide_inlining): Do not compute inlining parameters.
3177         (cgraph_decide_inlining_incrementally): Return TODOs; assume to
3178         be called with function context set up.
3179         (pass_ipa_inline): Remove unreachable functions before pass.
3180         (cgraph_early_inlining): Simplify assuming to be called from the
3181         PM as local pass.
3182         (pass_early_inline): New pass.
3183         (cgraph_gate_ipa_early_inlining): New gate.
3184         (pass_ipa_early_inline): Turn into simple wrapper.
3185         (compute_inline_parameters): New function.
3186         (gate_inline_passes): New gate.
3187         (pass_inline_parameters): New pass.
3188         (apply_inline): Move here from tree-optimize.c
3189         (pass_apply_inline): New pass.
3190         * ipa.c (cgraph_remove_unreachable_nodes): Verify cgraph after
3191         transforming.
3192         * tree-inline.c (optimize_inline_calls): Return TODOs rather than
3193         doing them by hand.
3194         (tree_function_versioning): Do not allocate dummy struct function.
3195         * tree-inline.h (optimize_inline_calls): Update prototype.
3196         * tree-optimize.c (execute_fixup_cfg): Export.
3197         (pass_fixup_cfg): Remove
3198         (tree_rest_of_compilation): Do not apply inlines.
3199         * tree-flow.h (execute_fixup_cfg): Declare.
3200         * Makefile.in (gt-passes.c): New.
3201         * passes.c: Include gt-passes.h
3202         (init_optimization_passes): New passes.
3203         (nnodes, order): New static vars.
3204         (do_per_function_toporder): New function.
3205         (execute_one_pass): Dump current pass here.
3206         (execute_ipa_pass_list): Don't dump current pass here.
3207
3208 2007-01-16  Janis Johnson  <janis187@us.ibm.com>
3209
3210         * config/dfp-bit.c (dfp_compare_op): Return separate value for NaN.
3211         (DFP_NE, DFP_LE, DFP_GE): Return false for NaN.
3212
3213 2007-01-16  David Edelsohn  <edelsohn@gnu.org>
3214
3215         * config/rs6000/darwin-ldouble.c: Build file for SOFT_FLOAT.
3216         (strong_alias): Define.
3217         (__gcc_qmul): Provide non-FMA for soft-float.
3218         (__gcc_qdiv): Same.
3219         (__gcc_qneg): New.
3220         (__gcc_qeq): New.
3221         (__gcc_qle): New.
3222         (__gcc_qge): New.
3223         (__gcc_qunord): New.
3224         (__gcc_stoq): New.
3225         (__gcc_dtoq): New.
3226         (__gcc_qtos): New.
3227         (__gcc_qtod): New.
3228         (__gcc_qtoi): New.
3229         (__gcc_qtou): New.
3230         (__gcc_itoq): New.
3231         (__gcc_utoq): New.
3232         (fmsub): New.
3233         * config/rs6000/rs6000.c (rs6000_init_libfuncs): Initialize
3234         soft-float functions.
3235         * config/rs6000/libgcc-ppc-glibc.ver: Version soft-float symbols.
3236         * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Do not warn
3237         about long double soft float.
3238
3239 2007-01-16  Dorit Nuzman  <dorit@il.ibm.com>
3240             Tehila Meyzels  <tehila@il.ibm.com>
3241
3242         * tree-vectorizer.h (is_pattern_stmt_p): New.
3243         * tree-vect-analyze.c (vect_determine_vectorization_factor): Fix
3244         formatting (tabs instead of spaces). Cleanup and clarify setting
3245         of STMT_VINFO_VECTYPE. Call is_pattern_stmt_p.
3246         * tree-vect-transform.c (vect_get_vec_def_for_operand): Fix typo.
3247         (vectorizable_type_demotion): Check that types are integral.
3248         (vectorizable_type_promotion): Likewise.
3249         (vectorizable_store): Fix typo.  Eliminate new-line at end of
3250         comments.
3251
3252 2007-01-16  Jan Hubicka  <jh@suse.cz>
3253
3254         * tree-ssanames.c (release_dead_ssa_names): Remove invalidated
3255         cgraph edges too.
3256
3257 2007-01-15  Eric Christopher  <echristo@apple.com>
3258
3259         * ifcvt.c: Include vec.h, vecprim.h.
3260         (check_cond_move_block): New argument regs.
3261         Reorganize. Add registers used to regs.
3262         (cond_move_process_if_block): Use regs set above as
3263         loop bounds.
3264
3265 2007-01-15  Eric Christopher  <echristo@apple.com>
3266
3267         * config/darwin.h: Update copyright.
3268         (TARGET_OPTION_TRANSLATE_TABLE): Add umbrella.
3269         (LINK_COMMAND_SPEC): Add -u.
3270         (LINK_SPEC): Fix umbrella for above.
3271
3272 2007-01-15  Joseph S. Myers  <joseph@codesourcery.com>
3273
3274         * config/soft-fp/op-common.h, config/soft-fp/op-4.h: Update from
3275         glibc CVS.
3276
3277 2007-01-15  Tom Tromey  <tromey@redhat.com>
3278
3279         * doc/sourcebuild.texi (libgcj Tests): Don't mention jacks.
3280         * doc/install.texi (Testing): Don't mention jacks.
3281         (Configuration): Document --enable-java-maintainer-mode.  Move
3282         --with-java-home to libgcj-specific section.  Document
3283         --with-ecj-jar.
3284         (Prerequisites): Mention --enable-java-maintainer-mode, ecj1.
3285
3286 2007-01-15  Jan Hubicka  <jh@suse.cz>
3287
3288         * tree-ssa-dce.c (DCE_TODOs): New.
3289         (propagate_necessity): Return if something changed.
3290         (eliminate_unnecessary_stmts): Likewise.
3291         (perform_tree_ssa_dce): Return TODO flags when needed.
3292         (pass_dce, pass_dce_loop, pass_cd_dce): Remove TODO flags.
3293
3294 2007-01-15  Uros Bizjak  <ubizjak@gmail.com>
3295
3296         * config/i386/i386.md (fyl2xxf3_i387): Rename from fyl2x_xf3.
3297         (fyl2x_extend<mode>xf3_i387): New insn pattern.
3298         (log<mode>2): Rename from logsf2 and logdf2 and macroize insn
3299         patterns using X87MODEF12 mode macro.  Extend operand 1
3300         to XFmode. Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
3301         (log10<mode>2): Ditto.
3302         (log2<mode>2): Ditto.
3303         (log1p<mode>2): Ditto.
3304         (logb<mode>2): Ditto.
3305         (fyl2xp1xf3_i387): Rename from fyl2xp1_xf3.
3306         (fyl2xp1_extend<mode>xf3_i387): New insn pattern.
3307         (*fxtractxf3_i387): Rename from *fxtractxf3.
3308         (fxtract_extend<mode>xf3_i387): New insn pattern.
3309         (ilogbsi2): Use match_dup 3, not match_operand:XF 3.
3310         * config/i386/i386.c (ix86_emit_i387_log1p): Use gen_fyl2xp1xf3_i387()
3311         and gen_fyl2xxf3_i387().
3312
3313 2007-01-14  Zdenek Dvorak <dvorakz@suse.cz>
3314
3315         * loop-unswitch.c (unswitch_loop): Do not call fix_loop_placement.
3316         * cfgloopmanip.c (fix_loop_placement): Made static.  Use
3317         get_loop_exit_edges.  Changed return type to bool.
3318         * cfgloop.h (fix_loop_placement): Declaration removed.
3319
3320 2007-01-14  Dorit Nuzman  <dorit@il.ibm.com>
3321
3322         * param.h (MIN_VECT_LOOP_BOUND): New.
3323         * params.def (MIN_VECT_LOOP_BOUND): New.
3324         * tree-vectorizer.c (slpeel_tree_peel_loop_to_edge): Takes another
3325         argument - minimum threshold for number of iterations.
3326         * tree-vectorizer.h (slpeel_tree_peel_loop_to_edge): Add another
3327         argument to declaration.
3328         * tree-vect-analyze.c (vect_analyze_operations): Check value of
3329         MIN_VECT_LOOP_BOUND.
3330         * tree-vect-transform.c (vect_do_peeling_for_loop_bound): Call
3331         slpeel_tree_peel_loop_to_edge with additional argument.
3332         (vect_do_peeling_for_alignment): Likewise.
3333         * doc/invoke.texi (min-vect-loop-bound): Document new param option.
3334
3335 2007-01-14  Uros Bizjak  <ubizjak@gmail.com>
3336
3337         PR target/30413
3338         * config/i386/i386.c (print_operand) ['z']: Output 'b' for
3339         operands of size 1.
3340
3341 2007-01-14  Jan Hubicka  <jh@suse.cz>
3342
3343         * tree-dfa.c (remove_referenced_var): New function.
3344         * tree-ssa-live.c (remove_unused_locals): Walk referenced vars and
3345         prune referenced vars list too.
3346         * tree-flow.h (remove_referenced_var): Declare.
3347
3348 2007-01-14  Jan Hubicka  <jh@suse.cz>
3349
3350         * tree-eh.c (add_stmt_to_eh_region_fn): Do not add call_exprs
3351         separately.
3352         (remove_stmt_from_eh_region_fn): Do not remove call_exprs.
3353         (verify_eh_throw_stmt_node, verify_eh_throw_table_statements): Kill.
3354         * except.h (verify_eh_throw_table_statements): Kill prototype.
3355         * cfgexpand.c (expand_gimple_basic_block): Propagate Eh regions
3356         into call exrepssions.
3357         * tree-optimize.c (execute_free_cfg_annotatiosn): Do not call
3358         eh trhow verifier.
3359         * tree-cfg.c: Include pointer-set.h.
3360         (verify_node_sharing): Work on pointer set.
3361         (verify_eh_throw_stmt_node): New.
3362         (verify_stmts): Use pointers sets, verify throw_stmt.
3363
3364 2007-01-13  Zdenek Dvorak <dvorakz@suse.cz>
3365
3366         * ipa-reference.c (analyze_function): Consider also addresses taken
3367         in phi nodes.
3368
3369 2007-01-12  Roger Sayle  <roger@eyesopen.com>
3370
3371         * c-typeck.c (null_pointer_constant_p): Replace use of
3372         TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
3373         (build_c_cast): Likewise.
3374
3375 2007-01-12  Roger Sayle  <roger@eyesopen.com>
3376
3377         * tree.h (force_fit_type_double): Remove unused final argument.
3378         * c-common.c (constant_expression_warning): Replace use of
3379         TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
3380         (convert_and_check): Likewise.
3381         (shorten_compare): Update call to force_fit_type_double.
3382         (c_common_truthvalue_conversion) <INTEGER_CST>: Use integer_zerop.
3383         * convert.c (convert_to_pointer): Update call to
3384         force_fit_type_double.
3385         * fold-const.c (force_fit_type_double): Remove overflowed_const
3386         argument.
3387         (int_const_binop, fold_convert_const_int_from_int,
3388         fold_convert_const_int_from_real, fold_div_compare,
3389         fold_sign_changed_comparison, fold_unary, fold_negate_const,
3390         fold_abs_const, fold_not_const): Remove the final argument from
3391         calls to force_fit_type_double.
3392
3393 2007-01-12  Andrew Pinski  <andrew_pinski@playstation.sony.com>
3394
3395         * configure.ac: Set insn to "nop" for spu-*-* also.
3396         * configure: Regenerate.
3397
3398 2007-01-12  Olga Golovanevsky  <olga@il.ibm.com>
3399
3400         * builtins.def : Add BUILT_IN_FREE.
3401
3402 2007-01-12  Jan Hubicka  <jh@suse.cz>
3403
3404         PR tree-optimization/30443
3405         * tree-inline.c (tree_function_versioning): Do not optimize when
3406         cloning for inlining.
3407
3408 2007-01-12  Zdenek Dvorak <dvorakz@suse.cz>
3409
3410         * doc/loop.texi: Document recording of loop exits.
3411         * cfgloopmanip.c (loopify, duplicate_loop): Use alloc_loop.
3412         (update_single_exits_after_duplication,
3413         update_single_exit_for_duplicated_loop,
3414         update_single_exit_for_duplicated_loops): Removed.
3415         (duplicate_loop_to_header_edge): Do not call
3416         update_single_exits_after_duplication and
3417         update_single_exit_for_duplicated_loops.
3418         (loop_version): Do not update single_exit information.
3419         (fix_loop_structure): Use record_loop_exits instead of
3420         mark_single_exit_loops.
3421         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update
3422         the lists of loop exits.
3423         * cfghooks.c (redirect_edge_and_branch, redirect_edge_and_branch_force,
3424         split_edge, merge_blocks): Update the lists of loop exits.
3425         * modulo-sched.c (sms_schedule): Pass LOOPS_HAVE_RECORDED_EXITS to
3426         loop_optimizer_init.
3427         * loop-init.c (loop_optimizer_init): Call record_loop_exits instead
3428         of mark_single_exit_loops.
3429         (loop_optimizer_finalize): Call release_recorded_exits.
3430         * tree-ssa-loop.c (tree_loop_optimizer_init): Pass
3431         LOOPS_HAVE_RECORDED_EXITS to loop_optimizer_init.
3432         * tree-vectorizer.c (slpeel_tree_duplicate_loop_to_edge_cfg): Do not
3433         update single exit information.
3434         * lambda-code.c (perfect_nestify): Ditto.
3435         * cfgloop.c (flow_loop_free): Destroy the list of exits of the loop.
3436         (mark_single_exit_loops): Removed.
3437         (alloc_loop, loop_exit_hash, loop_exit_eq, loop_exit_free,
3438         get_exit_descriptions, rescan_loop_exit, record_loop_exits,
3439         dump_recorded_exit, dump_recorded_exits, release_recorded_exits): New
3440         functions.
3441         (get_loop_exit_edges, single_exit): Use recorded exit lists.
3442         (add_bb_to_loop, remove_bb_from_loops): Update the lists of loop exits.
3443         (verify_loop_structure): Verify consistency of the exit lists.
3444         (flow_loops_find): Use alloc_loop.  Initialize exits hash.
3445         (set_single_exit): Removed.
3446         * cfgloop.h (struct loop_exit): New function.
3447         (struct loop): single_exit_ field replaced by exits field.
3448         (LOOPS_HAVE_MARKED_SINGLE_EXITS): Replaced by LOOPS_HAVE_RECORDED_EXITS.
3449         (struct loops): Added exits hash.
3450         (mark_single_exit_loops, set_single_exit): Declaration removed.
3451         (release_recorded_exits, record_loop_exits, rescan_loop_exit): Declare.
3452
3453 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
3454
3455         * doc/invoke.texi: Avoid use of @headitem.
3456
3457 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
3458
3459         * cse.c (cse_insn): Move HAVE_CC0 code after declarations.
3460
3461 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
3462
3463         * doc/install.texi: Fix m68k-*-* anchor and add m68k-*-* to the
3464         list of targets.
3465
3466 2007-01-12  Nathan Sidwell  <nathan@codesourcery.com>
3467             Richard Sandiford  <richard@codesourcery.com>
3468
3469         * doc/invoke.texi: Document -mno-bitfield, -mno-rtd and -mno-short.
3470         * config/m68k/m68k.opt: Resort options.
3471         (mbitfield, mrtd, mshort): Remove RejectNegative properties.
3472
3473 2007-01-12  Nathan Sidwell  <nathan@codesourcery.com>
3474             Richard Sandiford  <richard@codesourcery.com>
3475
3476         * doc/invoke.texi: Document the macros that are defined by
3477         m68k's -mtune and -mhard-float options.
3478         * config/m68k/m68k-protos.h (m68k_cpp_cpu_ident) Declare.
3479         (m68k_cpp_cpu_family): Likewise.
3480         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Add a full set
3481         of __ucfv*__ macros.  Define __mcffpu__ if generating code for
3482         ColdFire FPUs.  Define __mcf_cpu_* and __mcf_family_* macros.
3483         * config/m68k/m68k.c (m68k_cpp_cpu_ident): New function.
3484         (m68k_cpp_cpu_family): Likewise.
3485
3486 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
3487
3488         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Treat all mc68*
3489         macros besides mc68000 as tuning macros.  Use a switch statement
3490         to set them and mcpu32.
3491
3492 2007-01-12  Julian Brown  <julian@codesourcery.com>
3493
3494         * config/m68k/m68k.h: Use TARGET_68040 instead of TARGET_68040_ONLY.
3495         (TARGET_68040_ONLY): Rename to...
3496         (TARGET_68040): ...this.
3497         * config/m68k/m68k.c: Use TARGET_68040 instead of TARGET_68040_ONLY.
3498         * config/m68k/m68k.md: Likewise.
3499
3500 2007-01-12  Julian Brown  <julian@codesourcery.com>
3501             Nathan Sidwell  <nathan@codesourcery.com>
3502             Richard Sandiford  <richard@codesourcery.com>
3503
3504         * config.gcc (m680[012]0-*-*, m68k*-*-*): Set m68k_cpu_ident to
3505         the -mcpu= argument associated with the --with-cpu setting.
3506         Define M68K_DEFAULT_TUNE to the default -mtune= option,
3507         if different from the one implied by the -mcpu setting.
3508         Accept --with-cpu=FOO if FOO is listed in m68k-devices.def,
3509         using mcpu=FOO as the default CPU option.  Set target_cpu_default2.
3510         * doc/invoke.texi: Mention ColdFire in the introduction to the
3511         m68k options.  Document the new -march, -mcpu, -mtune, -mdiv,
3512         -mno-div and -mhard-float options.  Make -m68881 a synonym for
3513         -mhard-float.  Document the previously-undocumented -m5206e,
3514         -m528x, -m5307 and -m5407 options.  Tweak the existing option
3515         documentation for consistency.
3516         * doc/install.texi: Mention new --with-cpu arguments.
3517         * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Only use the
3518         default CPU if neither -mcpu nor -march are specified.
3519         (ASM_CPU_SPEC): Pass down -mcpu and -march options.
3520         (TARGET_CPU_CPP_BUILTINS): Set __mcfisa*__ macros from
3521         TARGET_ISA*.  Set the legacy __mcf*__ cpu macros in the same way,
3522         using m68k_tune to decide between families that implement the
3523         same ISA.  Use m68k_tune to set __mcfv4e__.
3524         (FL_BITFIELD, FL_68881, FL_COLDFIRE, FL_CF_HWDIV, FL_CF_MAC)
3525         (FL_CF_EMAC, FL_CF_EMAC_B, FL_CF_USP, FL_CF_FPU, FL_ISA_68000)
3526         (FL_ISA_68010, FL_ISA_68020, FL_ISA_68040, FL_ISA_A, FL_ISA_B)
3527         (FL_ISA_C, FL_ISA_MMU): New macros.
3528         (MASK_COLDFIRE): Delete.
3529         (TARGET_68010, TARGET_68020, TARGET_68040_ONLY, TARGET_COLDFIRE)
3530         (TARGET_ISAB): Redefine in terms of m68k_cpu_flags.
3531         (TARGET_68881, TARGET_COLDFIRE_FPU): Redefine in terms of m68k_fpu.
3532         (TARGET_HARD_FLOAT): Do not define here.
3533         (TARGET_ISAAPLUS, TARGET_ISAC): New macros.
3534         (TUNE_68000): New macro.
3535         (TUNE_68000_10): Redefine in terms of TUNE_68000 and TUNE_68010.
3536         (TUNE_68010, TUNE_68030, TUNE_68040, TUNE_68060, TUNE_CPU32)
3537         (TUNE_CFV2): Redefine in terms of m68k_tune.
3538         (uarch_type, target_device, fpu_type): New enums.
3539         (m68k_cpu, m68k_tune, m68k_fpu, m68k_cpu_flags): Declare.
3540         * config/m68k/m68k.c (TARGET_DEFAULT): Remove MASK_68881.
3541         (FL_FOR_isa_00, FL_FOR_isa_10, FL_FOR_isa_20, FL_FOR_isa_40)
3542         (FL_FOR_isa_cpu32, FL_FOR_isa_a, FL_FOR_isa_aplus, FL_FOR_isa_b)
3543         (FL_FOR_isa_c): New macros.
3544         (m68k_isa): New enum.
3545         (m68k_target_selection): New structure.
3546         (all_devices, all_isas, all_microarchs): New tables.
3547         (m68k_cpu_entry, m68k_arch_entry, m68k_tune_entry, m68k_cpu)
3548         (m68k_tune, m68k_fpu, m68k_cpu_flags): New variables.
3549         (MASK_ALL_CPU_BITS): Delete.
3550         (m68k_find_selection): New function.
3551         (m68k_handle_option): Handle -mcpu=, -march= and -mtune=.
3552         Map the legacy target options to a combination of the new ones.
3553         (override_options): Set m68k_cpu, m68k_tune, m68k_fpu and
3554         m68k_cpu_flags.  Handle M68K_DEFAULT_TUNE.  Use m68k_cpu_flags
3555         to derive default MASK_BITFIELD, MASK_CF_HWDIV and MASK_HARD_FLOAT
3556         settings.
3557         * config/m68k/m68k.opt (m5200, m5206e, m528x, m5307, m5407, mcfv4e)
3558         (m68010, m68020, m68020-40, m68020-60, m68030, m68040): Remove Mask
3559         properties.
3560         (m68881, msoft-float): Change mask from 68881 to HARD_FLOAT.
3561         (march=, mcpu=, mdiv, mhard-float, mtune=): New options.
3562         * config/m68k/m68k-devices.def: New file.
3563
3564 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
3565             Nathan Sidwell  <nathan@codesourcery.com>
3566
3567         * config/m68k/m68k.h (ASM_CPU_SPEC, ASM_SPEC, EXTRA_SPECS)
3568         (SUBTARGET_EXTRA_SPECS): New macros.
3569         * config/m68k/linux.h (ASM_SPEC): Remove CPU flags;
3570         use %(asm_cpu_spec) instead.
3571         * config/m68k/m68k-none.h (ASM_SPEC): Likewise.
3572         * config/m68k/openbsd.h (ASM_SPEC): Likewise.
3573         * config/m68k/netbsd-elf.h (ASM_SPEC): Likewise.
3574         (EXTRA_SPECS): Rename to...
3575         (SUBTARGET_EXTRA_SPECS): ...this.
3576
3577 2007-01-12  Nathan Sidwell  <nathan@codesourcery.com>
3578             Richard Sandiford  <richard@codesourcery.com>
3579             Julian Brown  <julian@codesourcery.com>
3580
3581         * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
3582         (m68k-*-uclinux*, m68k-*-linux*, m68k-*-rtems*): Set default_m68k_cpu
3583         to the configuration's default CPU.
3584         (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*): Likewise.
3585         Remove default masks.
3586         (m680[012]0-*-*): Set the default with_cpu to the first part of
3587         the target name.
3588         (m68k*-*-*): Set the default with_cpu to m$default_m68k_cpu.
3589         (m68k*-*-linux): Extend the --with-cpu handling to...
3590         (m680[012]0-*-*, m68k*-*-*): ...these configurations.  Allow m68000
3591         and m68010.  Don't set target_cpu_default2.
3592         * doc/install.texi: Document --with-cpu for m68k.
3593         * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Define.
3594         * config/m68k/m68k-none.h (TARGET_CPU_DEFAULT, M68K_CPU_m68k)
3595         (M68K_CPU_m68000, M68K_CPU_m68010, M68K_CPU_m68020, M68K_CPU_m68030)
3596         (M68K_CPU_m68040, M68K_CPU_m68302, M68K_CPU_m68332, TARGET_DEFAULT)
3597         (ASM_CPU_DEFAULT_SPEC, CC1_CPU_DEFAULT_SPEC): Delete.
3598         (ASM_SPEC): Remove use of %(asm_cpu_default).
3599         (EXTRA_SPECS, SUBTARGET_EXTRA_SPECS, MULTILIB_DEFAULTS): Delete.
3600         * config/m68k/linux.h (TARGET_DEFAULT): Delete.
3601         (CPP_SPEC): Merge definitions.  Do not handle __HAVE_68881__ here.
3602         * config/m68k/netbsd-elf.h (TARGET_OS_CPP_BUILTINS): Define
3603         __HAVE_FPU__ if TARGET_HARD_FLOAT.
3604         (TARGET_DEFAULT): Delete.
3605         (EXTRA_SPECS): Delete cpp_cpu_default_spec, cpp_cpu_spec,
3606         cpp_fpu_spec, asm_default_spec and netbsd_cpp_spec.
3607         (CPP_CPU_SPEC): Delete.
3608         (TARGET_VERSION): Merge definitions, using TARGET_68010 to pick
3609         the appropriate string.
3610         (CPP_CPU_DEFAULT_SPEC, ASM_DEFAULT_SPEC, CPP_FPU_SPEC): Delete.
3611         (CPP_SPEC): Define to NETBSD_CPP_SPEC.
3612         (ASM_SPEC): Don't use %(asm_default_spec).
3613         * config/m68k/m68k.c (TARGET_DEFAULT_TARGET_FLAGS): Remove
3614         TARGET_DEFAULT and add MASK_68881.
3615         * config/m68k/m68k.md: Remove mention of TARGET_DEFAULT from comments.
3616
3617 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
3618
3619         * config.gcc (m68010-*-netbsdelf*): Add MASK_68010.
3620         (m68k*-*-netbsdelf*, m68k*-*-openbsd*, m68k*-linux*): Add
3621         MASK_68010 alongside MASK_68020.
3622         * doc/invoke.texi: Document -m68010.
3623         * config/m68k/m68k.opt (m68010): New.
3624         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Define mc68010
3625         if TUNE_68010.
3626         (TUNE_68010): New macro.
3627         * config/m68k/m68k-none.h (M68K_CPU_m68k, M68K_CPU_m68010)
3628         (M68K_CPU_m68020, M68K_CPU_m68030, M68K_CPU_m68040)
3629         (M68K_CPU_m68332): Add MASK_68010.
3630         * config/m68k/linux.h (TARGET_DEFAULT): Add MASK_68010 to
3631         fallback definition.
3632         * config/m68k/netbsd-elf.h (CPP_CPU_SPEC): Remove now-redundant
3633         defines.
3634         * config/m68k/m68k.c (MASK_ALL_CPU_BITS): Add MASK_68010.
3635         (m68k_handle_option): Handle OPT_m68010.  Add MASK_68010
3636         to all entries that use MASK_68020.
3637         (output_move_simode_const, output_move_himode, output_move_qimode)
3638         (output_move_stricthi, output_move_strictqi): Use TARGET_68010
3639         instead of TARGET_68020 to select clr behavior.  Remove comment
3640         about there being no TARGET_68010.
3641         * config/m68k/m68k.md: Likewise throughout.
3642
3643 2007-01-12  Julian Brown  <julian@codesourcery.com>
3644
3645         * config/m68k/m68k.h (TARGET_ISAB): New macro.
3646         * config/m68k/m68k.c: Use TARGET_ISAB rather than TARGET_CFV4.
3647         * config/m68k/m68k.md: Likewise.
3648
3649 2007-01-12  Julian Brown  <julian@codesourcery.com>
3650
3651         * config/m68k/m68k.h (LEGITIMATE_INDEX_P, LEGITIMIZE_ADDRESS): Use
3652         TARGET_COLDFIRE_FPU instead of TARGET_CFV4E.
3653
3654 2007-01-12  Julian Brown  <julian@codesourcery.com>
3655
3656         * config/m68k/m68k.h (TUNE_68040_60): New macro.
3657         * config/m68k/m68k.c (standard_68881_constant_p): Use it.
3658         * config/m68k/m68k.md: Likewise.
3659
3660 2007-01-12  Julian Brown  <julian@codesourcery.com>
3661             Richard Sandiford  <richard@codesourcery.com>
3662
3663         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Use TUNE_68030
3664         instead of TARGET_68030, TUNE_68040 instead of TARGET_68040,
3665         TUNE_68060 instead of TARGET_68060 and TUNE_CPU32 instead of
3666         TARGET_CPU32.
3667         (TARGET_CPU32): Rename to...
3668         (TUNE_CPU32): ...this.
3669         (TUNE_68000_10, TUNE_68030, TUNE_68040, TUNE_68060)
3670         (TUNE_CFV2): New macros.
3671         * config/m68k/netbsd-elf.h (LONG_DOUBLE_TYPE_SIZE): Simplify;
3672         remove conditions that are implied by TARGET_68020.
3673         * config/m68k/m68k.c (m68k_output_function_prologue): Use TUNE_68040
3674         instead of TARGET_68040 and TUNE_CPU32 instead of TARGET_CPU32.
3675         (m68k_output_function_epilogue): Likewise.
3676         (m68k_rtx_costs): Likewise.  Use TUNE_68060 instead of TARGET_68060
3677         and TUNE_CFV2 instead of TARGET_5200.  Use TUNE_68000_10 instead of
3678         "!TARGET_68020 && !TARGET_COLDFIRE" to choose between 68000 and
3679         non-68000 timings.  Refactor multiplication and division costs.
3680         (output_addsi3): Use TUNE_68040 instead of TARGET_68040 and
3681         TUNE_CPU32 instead of TARGET_CPU32.
3682         (standard_68881_constant_p): Use TUNE_68040 instead of TARGET_68040
3683         and TUNE_68060 instead of TARGET_68060.
3684         * config/m68k/m68k.md: Use TUNE_68040 instead of TARGET_68040,
3685         TUNE_68060 instead of TARGET_68060, and TUNE_CPU32 instead of
3686         TARGET_CPU32.
3687         (movsi_const0): Use TUNE_68000_10 rather than "!TARGET_68020
3688         && !TARGET_COLDFIRE" to choose between moveq and clr.
3689         Likewise in the unnamed movsf pattern.
3690         (ashlsi_17_24, lshrsi_17_24): Guard with TUNE_68000_10 rather than
3691         "!TARGET_68020 && !TARGET_COLDFIRE".  Likewise the unnamed
3692         ashiftrt pattern.
3693
3694 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
3695
3696         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Increase amount
3697         of tabbing before backslashes.
3698
3699 2007-01-11  John David Anglin  <dave.anglin@nrc-cnrc.gc.ca>
3700
3701         * pa-linux.h (ASM_OUTPUT_INTERNAL_LABEL): Undefine.
3702         * pa.h (ASM_OUTPUT_LABEL): Output colon when using GAS.
3703         (ASM_OUTPUT_INTERNAL_LABEL): Define.
3704
3705 2007-01-11  Zdenek Dvorak <dvorakz@suse.cz>
3706
3707         * tree-ssa-loop-ivopts.c (extract_cond_operands): Split from
3708         find_interesting_uses_cond.
3709         (find_interesting_uses_cond): Use extract_cond_operands.
3710         (rewrite_use_compare): Use extract_cond_operands and
3711         force_gimple_operand_bsi.  Do not call update_stmt.
3712         (determine_use_iv_cost_condition): Use extract_cond_operands.
3713         Return cheaper of using original bound and changing the exit bound.
3714
3715 2007-01-11  Zdenek Dvorak <dvorakz@suse.cz>
3716
3717         PR tree-optimization/29516
3718         * tree-ssa-address.c (tree_mem_ref_addr, add_to_parts,
3719         most_expensive_mult_to_index, addr_to_parts,
3720         create_mem_ref, maybe_fold_tmr): Make the type of
3721         fields of TARGET_MEM_REF sizetype.
3722         (move_fixed_address_to_symbol, move_pointer_to_base):
3723         New functions.
3724         * tree.def (TARGET_MEM_REF): Add comment on types of
3725         the operands.
3726
3727 2007-01-11  Joseph Myers  <joseph@codesourcery.com>
3728
3729         * c-common.c (vector_types_convertible_p): Treat opaque types as
3730         always convertible if they have the same size, but not otherwise.
3731
3732 2007-01-11  Steven Bosscher  <steven@gcc.gnu.org>
3733
3734         * ifcvt.c (struct noce_if_info): Add comments to the fields.
3735         Remove the b_unconditional field.
3736         (noce_try_sign_mask): Do not look at b_unconditional.
3737         (noce_process_if_block): Do not use merge_if_blocks.  Update
3738         the CFG here.  Do not set b_unconditional.
3739         (cond_move_process_if_block): Likewise.
3740         (find_cond_trap): Likewise.
3741         (check_cond_move_block): Require simple jump insns at the end
3742         of the basic block.
3743
3744 2007-01-11  Jan Hubicka  <jh@suse.cz>
3745
3746         PR tree-optimization/1046
3747         * tree-tailcall.c (suitable_for_tail_call_opt_p): Use TREE_ADDRESSABLE
3748         when alias info is not ready.
3749         (pass_tail_recursion): Do not require aliasing.
3750         * tree-ssa-copyrename.c (pass_rename_ssa_cop): Likewise.
3751         * tree-ssa-ccp.c (pass_ccp, pass_fold_builtins): Likewise.
3752         * tree-ssa-copy.c (pass_copy_prop): Likewise.
3753         * tree-ssa-forwprop.c (pass_forwprop): Likewise.
3754         * tree-ssa-dce.c (pass_dce, pass_dce_loop, pass_cd_dce): Likewise.
3755         * passes.c (init_optimization_passes): Execute rename_ssa_copies,
3756         ccp, forwprop, copy_prop, merge_phi, copy_prop, dce and tail recursion
3757         before inlining.
3758         * tree-ssa-operands.c (add_virtual_operand, get_indirect_ref_operand):
3759         When aliasing is not build, mark statement as volatile.
3760
3761 2007-01-11  Tom Tromey  <tromey@redhat.com>
3762
3763         PR preprocessor/15185, PR preprocessor/20989:
3764         * doc/cppopts.texi <-MT>: Update description of algorithm for
3765         computing default target.
3766         <-M, -MD>: Reword "basename" text.
3767
3768 2007-01-11  Roger Sayle  <roger@eyesopen.com>
3769
3770         * builtins.c (expand_builtin_pow, expand_builtin_powi,
3771         fold_builtin_cabs, fold_builtin_sqrt, fold_builtin_trunc,
3772         fold_builtin_floor, fold_builtin_ceil, fold_builtin_round,
3773         fold_builtin_int_int_roundingfn, fold_builtin_bitop,
3774         fold_builtin_bswap, real_constp, fold_builtin_pow,
3775         fold_builtin_powi, fold_builtin_signbit, fold_builtin_copysign,
3776         do_mpfr_arg1, do_mpfr_arg2, do_mpfr_arg3, do_mpfr_sincos): Replace
3777         uses of the macro TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
3778         * convert.c (convert_to_pointer): Likewise.
3779         * expr.c (highest_pow2_factor, expand_expr_real_1): Likewise.
3780         * fold-const.c (force_fit_type, fold_negate_expr, int_const_binop,
3781         const_binop, fold_convert_const_int_from_int,
3782         fold_convert_const_int_from_real,
3783         fold_convert_const_real_from_real, sign_bit_p,
3784         optimize_minmax_comparison, extract_muldiv_1, fold_div_compare,
3785         fold_sign_changed_comparison, fold_unary, fold_comparison,
3786         fold_binary, multiple_of_p, tree_Expr_non_zero_p,
3787         fold_negate_const, fold_abs_const, fold_not_const): Likewise.
3788         * print-tree.c (print_node_brief, print_node): Likewise.
3789         * stor-layout.c (place_field, layout_type): Likewise.
3790         * tree-chrec.c (keep_cast): Likewise.
3791         * tree.c (build_vector, build_real, build_real_from_int_cst,
3792         build_complex): Likewise.
3793
3794 2007-01-11  Roger Sayle  <roger@eyesopen.com>
3795
3796         * tree.h (TREE_CONSTANT_OVERFLOW): Obsolete.  For the time being,
3797         treat TREE_CONSTANT_OVERFLOW as a synonym of TREE_OVERFLOW.
3798
3799 2007-01-11  Paolo Bonzini  <bonzini@gnu.org>
3800
3801         * configure.ac (strict1_warn): Rename to strict_warn.
3802         (WERROR, --enable-werror, symlink hacks, stage1_cflags,
3803         cc_set_by_configure, quoted_cc_set_by_configure,
3804         stage_prefix_set_by_configure, quoted_stage_prefix_set_by_configure,
3805         all_boot_languages, all_stagestuff): Remove.
3806         (target_list): Remove bootstrap targets.
3807         * Makefile.in (quickstrap): Unconditionally make a synonym of all.
3808         (BOOT_LANGUAGES, STAGE1_CFLAGS, STAGE1_CHECKING,
3809         REMAKEFLAGS, FLAGS_TO_PASS, PREPEND_DOTDOT_TO_RELATIVE_PATHS,
3810         SUBDIR_FLAGS_TO_PASS, WERROR_FLAGS, STRICT2_WARN, LANG_STAGESTUFF,
3811         VOL_FILES, POSTSTAGE1_FLAGS_TO_PASS, STAGE2_FLAGS_TO_PASS,
3812         STAGEPROFILE_FLAGS_TO_PASS, STAGEFEEDBACK_FLAGS_TO_PASS, stage1_build,
3813         stage1_copy, stage2_build, stage2_copy, stageprofile_build,
3814         stageprofile_copy, stage3_build, stage3_copy, stagefeedback_build,
3815         stagefeedback_copy, stage4_build, clean_s1, clean_sw, bootstrap,
3816         bootstrap-lean, bootstrap2, bootstrap2-lean, bootstrap3,
3817         bootstrap3-lean, bootstrap4, bootstrap4-lean, unstage1, unstage2,
3818         unstage3, unstage4, unstageprofile, unstagefeedback, restage, restage2,
3819         restage3, restage4, restageprofile, restagefeedback, bubbleestrap,
3820         cleanstrap, unstrap, restrap, *compare, *compare3, *compare4,
3821         *compare-lean, *compare3-lean, *compare4-lean, stage1-start, stage1,
3822         stage2-start, stage2, stage3-start, stage3, stage4-start, stage4,
3823         stageprofile-start, stageprofile, stagefeedback-start, stagefeedback,
3824         risky-stage1, risky-stage2, risky-stage3, risky-stage4): Remove.
3825         (ORDINARY_FLAGS_TO_PASS): Rename to FLAGS_TO_PASS.
3826         (STAGECOPYSTUFF, STAGEMOVESTUFF): Consolidate into MOSTLYCLEANFILES.
3827         (mostlyclean): Adjust.
3828         (clean, distclean): Don't mention bootstrap stuff.
3829         * configure: Regenerate.
3830         * ada/config-lang.in, cp/config-lang.in, forttran/config-lang.in,
3831         java/config-lang.in, objc/config-lang.in, objcp/config-lang.in,
3832         treelang/config-lang.in (stagestuff): Remove.
3833         * doc/sourcebuild.texi (stage1, stage2, stage3, stage4,
3834         stageprofile, stagefeedback, stagestuff): Remove mention.
3835
3836 2007-01-11  Nick Clifton  <nickc@redhat.com>
3837
3838         * config/mcore/predicates.md (mcore_general_movesrc_operand):
3839         Accept CONSTs.
3840         (mcore_general_movdst_operand): Do not accept CONST_INTs.
3841         (mcore_arith_K_S_operand): Run the test for the S constraint not
3842         the test for the M constraint.
3843         (mcore_addsub_operand): Do not accept integer values that are
3844         larger than 32 bits.
3845         * config/mcore/mcore.md: Remove unused constraints from split.
3846         (andsi3): Use HOST_WIDE_INT instead of int to hold an INTVAL.
3847         (addsi3): Likewise.
3848         (allocate_stack): Likewise.
3849         * config/mcore/mcore.c (mcore_print_operand): Restrict output of P
3850         operands to 32 bits.
3851         (mcore_const_costs): Use HOST_WIDE_INT instead of int to hold an
3852         INTVAL.
3853         (mcore_and_cost, mcore_modify_comparison, const_ok_for_mcore,
3854         mcore_const_ok_for_inline, mcore_const_trick_uses_not,
3855         try_constant_tricks, mcore_num_ones, mcore_num_zeros,
3856         mcore_output_bclri, mcore_output_andn, output_inline_const,
3857         mcore_output_move, mcore_output_movedouble): Likewise.
3858         (mcore_output_cmov): Use CONST_OK_FOR_M and CONST_OK_FOR_N.
3859         (output_inline_const): Likewise.
3860         (output_inline_const): Fix format strings used in sprintf
3861         statements.
3862         * config/mcore/mcore-protos.h: Update prototypes for changed
3863         functions in mcore.c.
3864         * config/mcore/mcore.h (CONST_OK_FOR_I): Cast values to
3865         HOST_WIDE_INT and not int.
3866         (CONST_OK_FOR_J, CONST_OK_FOR_K, CONST_OK_FOR_L, CONST_OK_FOR_M,
3867         CONST_OK_FOR_N): Likewise.
3868         (LEGITIMATE_CONSTANT_P): Also check CONSTANT_P.
3869         (GO_IF_LEGITIMATE_INDEX): Use HOST_WIDE_INT instead of int to hold
3870         an INTVAL.
3871
3872 2007-01-10  Jan Hubicka  <jh@suse.cz>
3873
3874         * tree-vrp.c (remove_range_assertions): Release defs.
3875         * tree-ssa-loop-ivopts.c (rmeove_statement): Likewise.
3876         * tree-ssa-dom.c (remove_stmt_or_phi): Likewise.
3877
3878 2007-01-10  Paul Brook  <paul@codesourcery.com>
3879
3880         * config/arm/arm.c (arm_rtx_costs_1): Handle mutiply-subtract.
3881         * config/arm/arm.md (mulsi3subsi): New insn.
3882
3883 2007-01-10  Zdenek Dvorak <dvorakz@suse.cz>
3884
3885         * tree-ssa-loop-manip.c (tree_unroll_loop): Make it a wrapper over ...
3886         (tree_transform_and_unroll_loop): New.
3887         * tree-flow.h (transform_callback, tree_transform_and_unroll_loop):
3888         Declare.
3889
3890 2007-01-10  Robert Kennedy <jimbob@google.com>
3891
3892         * fold-const.c (fold_comparison): Fold comparisons like (x *
3893         1000 < 0) to (x < 0).
3894
3895 2007-01-10  Ian Lance Taylor  <iant@google.com>
3896
3897         * tree-pretty-print.c (dump_generic_node): Print parentheses when
3898         operands have the same priority.
3899
3900 2007-01-10  Tom Tromey  <tromey@redhat.com>
3901
3902         * fold-const.c (fold_truthop): Don't check can_use_bit_fields_p.
3903         (fold_binary): Likewise.
3904         * langhooks.c (lhd_can_use_bit_fields_p): Removed.
3905         * langhooks-def.h (lhd_can_use_bit_fields_p): Removed.
3906         (LANG_HOOKS_CAN_USE_BIT_FIELDS_P): Removed.
3907         (LANG_HOOKS_INITIALIZER): Remove LANG_HOOKS_CAN_USE_BIT_FIELDS_P.
3908         * langhooks.h (struct lang_hooks): Removed field
3909         'can_use_bit_fields_p'.
3910
3911 2007-01-10  Ralf Corsépius <ralf.corsepius@rtems.org>
3912
3913         * config/bfin/t-bfin, config/bfin/t-bfin-elf: Remove GCC_CFLAGS.
3914
3915 2007-01-10  Razya Ladelsky  <razya@il.ibm.com>
3916
3917         * function.c (get_last_funcdef_no): New function.
3918         * function.h (get_last_funcdef_no): Declare.
3919         * tree-inline.c (initialize_cfun): Add initialization.
3920         (tree_function_versioning): Cleanup.
3921
3922 2007-01-10  Jan Hubicka  <jh@suse.cz>
3923
3924         * tree-inline.c (setup_one_parameter): Do not propagate into abnormal
3925         PHIs.
3926
3927 2007-01-10  Sa Liu  <saliu@de.ibm.com>
3928             Ben Elliston  <bje@au.ibm.com>
3929
3930         * spu.h (STACK_SAVE_AREA): Use VOIDmode for SAVE_FUNCTION, SImode
3931         for SAVE_NONLOCAL and Pmode for any other save level.
3932         * spu-protos.h (spu_restore_stack_block): Declare.
3933         * spu.md (save_stack_block): Remove.
3934         (restore_stack_block): Call spu_restore_stack_block.
3935         * spu.c (spu_restore_stack_block): New function.
3936         (spu_expand_epilogue): Remove old comment.
3937
3938 2007-01-09  Zdenek Dvorak <dvorakz@suse.cz>
3939
3940         PR tree-optimization/30322
3941         * tree-ssa-loop-ivopts.c (fold_affine_expr, iv_value): Removed.
3942         (cand_value_at): Return the value as aff_tree.
3943         (may_eliminate_iv): Convert the bound from aff_tree to tree.
3944         * tree-affine.c (aff_combination_add_cst, aff_combination_add_product,
3945         aff_combination_mult): New functions.
3946         (aff_combination_add): Use aff_combination_add_cst.
3947         (aff_combination_convert): Allow conversions to a wider type.
3948         (tree_to_aff_combination): Handle BIT_NOT_EXPR.
3949         * tree-affine.h (aff_combination_mult): Declare.
3950
3951 2007-01-09  Carlos O'Donell  <carlos@codesourcery.com>
3952
3953         * doc/tm.texi: Update documentation to reflect reality of exec
3954         and start file search behaviours. Update copyright year.
3955         * doc/invoke.texi: Explain how GCC_EXEC_PREFIX is used to find
3956         header file directories.
3957
3958 2007-01-09  Uros Bizjak  <ubizjak@gmail.com>
3959
3960         * config/i386/i386.md (*sinxf2): Rename to *sinxf2_i387.
3961         (*cosxf2): Rename to cosxf2_i387.
3962         (*sindf2, *sinsf2): Extend operand 1 to XFmode.  Macroize patterns
3963         using X87MODEF12 mode macro. Rename patterns to
3964         *sin_extend<mode>xf2_i387.  Use SSE_FLOAT_MODE_P to disable patterns
3965         for SSE math.
3966         (*cosdf2, *cossf2): Ditto.
3967         (sincosdf3, sincossf3): Ditto.  Rewrite corresponding splitters
3968         to match extended input operands.
3969         (sincos<mode>3): New expander.
3970         (*sinextendsfdf2, *cosextendsfdf2, *sincosextendsfdf3): Remove
3971         insn patterns and corresponding splitters.
3972
3973 2007-01-09  Kaz Kojima  <kkojima@gcc.gnu.org>
3974
3975         * config/sh/t-linux (TARGET_LIBGCC2_CFLAGS): Delete.
3976         (SHLIB_MAPFILES, SHLIB_LINK, SHLIB_INSTALL): Likewise.
3977
3978 2007-01-09  Nicolas Pitre  <nico@cam.org>
3979
3980         PR target/30173
3981         * arm/ieee754-df.S (Lad_s): Also test the low word of X for zero.
3982
3983 2007-01-08  Geoffrey Keating  <geoffk@apple.com>
3984
3985         * target.h (struct gcc_target): New field library_rtti_comdat.
3986         * target-def.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): New.
3987         (TARGET_CXX): Add TARGET_CXX_LIBRARY_RTTI_COMDAT.
3988         * doc/tm.texi (C++ ABI): Document TARGET_CXX_LIBRARY_RTTI_COMDAT.
3989         * config/darwin.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
3990
3991 2007-01-08  Geoffrey Keating  <geoffk@apple.com>
3992
3993         * doc/invoke.texi (Optimize Options): Correct description of -O0.
3994
3995 2007-01-08  Richard Guenther  <rguenther@suse.de>
3996
3997         * tree.h (force_fit_type_double): Export.
3998         (force_fit_type): Remove.
3999         * fold-const.c (force_fit_type_double): New function.
4000         (force_fit_type): Remove.
4001         (int_const_binop): Use it.
4002         (fold_convert_const_int_from_int): Likewise.
4003         (fold_convert_const_int_from_real): Likewise.
4004         (fold_div_compare): Likewise.
4005         (fold_sign_changed_comparison): Likewise.
4006         (fold_unary): Likewise.
4007         (fold_negate_const): Likewise.
4008         (fold_abs_const): Likewise.
4009         (fold_not_const): Likewise.
4010         * c-common.c (shorten_compare): Use force_fit_type_double.
4011         * convert.c (convert_to_pointer): Likewise.
4012
4013 2007-01-08  Richard Guenther  <rguenther@suse.de>
4014
4015         * tree.h (build_int_cst_wide_type): Export.
4016         * tree.c (build_int_cst_wide_type): New function.
4017         (build_int_cst_wide): Fix comment.
4018         * builtins.c (fold_builtin_object_size): Use build_int_cst
4019         to build -1 or 0 of the correct type.  Use fit_double_type
4020         to check for overflow.
4021         * fold-const.c (optimize_bit_field_compare): Use build_int_cst_type
4022         to build the mask.
4023         (decode_field_reference): Likewise.
4024         (all_ones_mask_p): Likewise.
4025         (native_interpret_int): Use build_int_cst_wide_type.
4026         (fold_binary): Use build_int_cst_type to build an all-ones
4027         value.
4028         * stor-layout.c (set_sizetype): Use build_int_cst_wide_type.
4029
4030 2007-01-08  Daniel Jacobowitz  <dan@codesourcery.com>
4031
4032         * config/pa/t-pa64 (libgcc_stub.a): Use $(T).
4033
4034 2007-01-09  Ben Elliston  <bje@au.ibm.com>
4035
4036         * genautomata.c (STATS_OPTION): New option.
4037         (stats_flag): New flag.
4038         (gen_automata_option): Handle it.
4039         (initiate_automaton_gen): Ditto.
4040         (write_automata): Output statistics only if stats_flag is
4041         set. Likewise, output time statistics only if time_flag is set.
4042         * doc/md.texi (Processor pipeline description): Document new flag.
4043
4044 2007-01-08  Richard Guenther  <rguenther@suse.de>
4045
4046         * builtins.c (fold_builtin_int_roundingfn): Use fit_double_type.
4047         * tree.c (build_int_cst_type): Likewise.
4048         (size_in_bytes): Don't call force_fit_type on the result.
4049         (int_fits_type_p): Use fit_double_type.
4050         * fold-const.c (fit_double_type): New function.
4051         (force_fit_type): Use it.
4052         * tree.h (fit_double_type): Export.
4053
4054 2007-01-08  Jan Hubicka  <jh@suse.cz>
4055
4056         * tree-vectorizer.c (gate_increase_alignment): Fix return type.
4057         * ipa.c (function_and_variable_visibility): Fix return type.
4058
4059 2007-01-08  Richard Guenther  <rguenther@suse.de>
4060
4061         * tree-ssa-ccp.c (maybe_fold_offset_to_array_ref): Use type
4062         of offset to build the index.
4063         * tree-pretty-print.c (dump_generic_node): Don't build negated
4064         const just for printing.
4065         * c-pretty-print.c (pp_c_integer_constant): Likewise.
4066         * builtins.c (fold_builtin_int_roundingfn): Check if result
4067         fits the type by using force_fit_type and comparing the result.
4068         * predict.c (predict_loops): Use compare_tree_int for comparison.
4069         * tree.c (build_int_cst): Fall back to integer_type_node for
4070         NULL_TREE type.
4071         (build_int_cst_wide): Assert type is non-null.
4072
4073 2007-01-08  Roberto Costa  <roberto.costa@st.com>
4074
4075         * tree-vrp.c (extract_range_from_cond_expr): New.
4076         (extract_range_from_expr): Handle COND_EXPR nodes used as expressions.
4077         * tree-ssa-ccp.c (get_maxval_strlen): Handle COND_EXPR nodes used
4078         as expressions.
4079         (fold_stmt): Bug fix, avoid infinite recursion when folding COND_EXPRs.
4080         * tree-ssa-forwprop.c (simplify_cond, forward_propagate_into_cond,
4081         tree_ssa_forward_propagate_single_use_vars): Handle COND_EXPR nodes
4082         used as expressions.
4083         * tree-object-size.c (cond_expr_object_size): New.
4084         (collect_object_sizes_for): Handle COND_EXPR nodes used as expressions.
4085
4086 2007-01-08  Jan Hubicka  <jh@suse.cz>
4087
4088         * tree-ssa-forwprop.c (forward_propagate_into_cond,
4089         tree_ssa_forward_propagate_single_use_va): Release defs of propagated
4090         statement.
4091
4092 2007-01-08  Richard Guenther  <rguenther@suse.de>
4093
4094         PR tree-optimization/23603
4095         * tree-vrp.c (set_value_range_to_truthvalue): New function.
4096         (extract_range_from_binary): Fall back to truthvalue instead of
4097         varying for TRUTH_*_EXPR.
4098         (extract_range_from_comparison): Fall back to truthvalue instead of
4099         varying.
4100         (vrp_visit_phi_node): Don't adjust new range bounds to +INF/-INF
4101         if all visited PHI values were constant.
4102
4103 2007-01-08  Jan Hubicka  <jh@suse.cz>
4104
4105         * cgraphunit.c (cgraph_process_new_functions): Reset reachable flag.
4106         (cgraph_analyze_function): break out from ...
4107         (cgraph_finalize_compilation_unit): ... here.
4108         (cgraph_expand_function): Remove forgoten commented out line.
4109         (cgraph_optimize): Analyze functions.
4110
4111 2007-01-08  Jan Hubicka  <jh@suse.cz>
4112
4113         * tree-pas.h (TODO_remove_function): New flag.
4114         (TODO_update*): Renumber.
4115         (pass_ipa_increase_alignment,
4116         pass_ipa_function_and_variable_visibility): New passes.
4117         * cgraphunit.c (cgraph_increase_alignment): Move to tree-vectorizer.c
4118         (cgraph_function_and_variable_visibility): Move to ipa.c
4119         (cgraph_optimize): Don't call cgraph_function_and_variable_visibility,
4120         cgraph_increase_alignment.
4121         * ipa-inline.c (cgraph_decide_inlining): Don't push timevar.
4122         (cgraph_decide_inlining_incrementally): Push TV_INTEGRATION before
4123         calling tree-inline.
4124         (cgraph_early_inlining): Do not call cgraph_remove_unreachable_nodes.
4125         (pass_ipa_inline, pass_early_ipa_inlining): Set TODO_remove_functions
4126         * tree-vectorizer.c (increase_alignment): Move here from cgraphunit.c
4127         (gate_increase_alignment): New function.
4128         (pass_ipa_increase_alignment): New pass.
4129         * ipa.c: Inline tree-pass.h and timevar.h
4130         (function_and_variable_visibility): Move here from cgraphunit.c
4131         * tree-optimize.c (pass_early_local_passes): Add TODO_remove_functions.
4132         * passes.c (init_optimization_passes): Add the two new passes.
4133         (execute_todo): Handle cgraph_remove_functions.
4134
4135 2007-01-08  Nick Clifton  <nickc@redhat.com>
4136
4137         * config/frv/predicates.md (reg_or_0_operand): Accept
4138         CONST_DOUBLEs.
4139
4140 2007-01-08  Ralf Corsépius <ralf.corsepius@rtems.org>
4141
4142         * config/bfin/rtems.h, config/bfin/t-rtems: New.
4143         * config.gcc: Add bfin*-rtems*.
4144
4145 2007-01-08  Mark Shinwell  <shinwell@codesourcery.com>
4146
4147         * c.opt: Add -flax-vector-conversions.
4148         * c-typeck.c (convert_for_assignment): Pass flag to
4149         vector_types_convertible_p to allow emission of note.
4150         (digest_init): Likewise.
4151         * c-opts.c: Handle -flax-vector-conversions.
4152         * c-common.c (flag_lax_vector_conversions): New.
4153         (vector_types_convertible_p): Unless -flax-vector conversions
4154         has been passed, disallow conversions between vectors with
4155         differing numbers of subparts and/or element types.  If such
4156         a conversion is disallowed, possibly emit a note on the first
4157         occasion only to inform the user of -flax-vector-conversions.
4158         The new last argument specifies this.
4159         * c-common.h (flag_lax_vector_conversions): New.
4160         (vector_types_convertible_p): Add extra argument.
4161         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Use
4162         char_type_node for V*QI type vectors.
4163         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins):
4164         Update to satisfy new typechecking rules.
4165         * config/rs6000/altivec.h (vec_cmple): Use vec_cmpge, for both
4166         C and C++ variants.
4167         * doc/invoke.texi (C Dialect Options): Document
4168         -flax-vector-conversions.
4169
4170 2007-01-08  Mark Shinwell  <shinwell@codesourcery.com>
4171
4172         PR tree-optimization/29877
4173         * tree-ssa-ter.c (is_replaceable_p): Deem assignments with
4174         a register variable on the RHS to not be replaceable.
4175
4176 2007-01-08  Chen Liqin  <liqin@sunnorth.com.cn>
4177         * config/score/t-score-elf (MULTILIB_OPTIONS): Change.
4178         * config/score/predicates.md (const_uimm5, sr0_operand, const_simm12,
4179         const_simm15, const_pow2, const_npow2): Added.
4180         * config/score/misc.md (insv, extv, extzv, movmemsi,
4181         move_lbu_a/b, mov_lhu_a/b etc): Added and fix some bug.
4182         * config/score/score.c (score_address_cost, score_select_cc_mode):
4183         Added.
4184         Change CONST_OK_FOR_LETTER_P/EXTRA_CONSTRAINT define.
4185         Update score_rtx_costs for MACRO TARGET_RTX_COSTS.
4186         Update score_print_operand.
4187         * config/score/score.h (DATA_ALIGNMENT, SELECT_CC_MODE): Added.
4188         Adjust register allocate order and update some macro define.
4189         * config/score/score-mdaux.c (mdx_unaligned_load, mdx_unsigned_store,
4190         mdx_block_move_straight, mdx_block_move_loop_head,
4191         mdx_block_move_loop_body, mdx_block_move_loop_foot, mdx_block_move_loop,
4192         mdx_block_move): Added.
4193         (mdx_movsicc, mdp_select_add_imm, mdp_select, mds_zero_extract_andi,
4194         mdp_limm): Updated and fix some bug and typo.
4195         * config/score/score.md (movqi/hi/si, add/sub/zero/ext): Updated.
4196         (movsf, movdf, doloop_end): Added.
4197
4198 2007-01-08  Kazu Hirata  <kazu@codesourcery.com>
4199
4200         * config/arm/arm.c, config/arm/arm.h, config/arm/arm.md,
4201         config/arm/thumb2.md: Fix comment typos.
4202         * doc/extend.texi: Fix a typo.
4203
4204 2007-01-07  Eric Christopher  <echristo@apple.com>
4205
4206         * configure.ac: Check for __stack_chk_fail for darwin.
4207         * configure: Regenerate.
4208
4209 2007-01-07  Richard Guenther  <rguenther@suse.de>
4210
4211         * tree-vrp.c (extract_range_from_assert): CSE calls to
4212         compare_values where possible.
4213         (extract_range_from_unary_expr): Likewise.
4214
4215 2007-01-07  Anatoly Sokolov <aesok@post.ru>
4216
4217         * config/avr/avr-protos.h (call_insn_operand): Delete prototype.
4218         * config/avr/avr.c (call_insn_operand): Delete function.
4219         * config/avr/avr.md (*pushqi, *pushhi, *pushsi, *pushsf): Use REG_SP
4220         instead of register number. Use predicates.
4221         * config/avr/predicates.md (const0_operand, reg_or_0_operand,
4222         call_insn_operand): Add.
4223
4224 2007-01-06  Jan Hubicka  <jh@suse.cz>
4225
4226         * tree-pass.h (pass_build_cgraph_edges): Declare.
4227         * cgraphunit.c (record_refernece): Move to cgraphbuild.c
4228         (visited_nodes): Remove.
4229         (cgraph_create_edges): Move to cgraphbuild.c; rename to
4230         build_cgrpah_edges; make visited_nodes local.
4231         (cgraph_process_new_functions): DO not call initialize_inline_failed.
4232         (record_references_in_initializer): Move to cgraphbuild.c
4233         (initialize_inline_failed, rebuild_cgraph_edges,
4234         pass_rebuild_cgraph_edges): Move to cgraphbuild.c.
4235         (verify_cgraph_node): Make visited_nodes local.
4236         (cgraph_analyze_function): Do not call cgraph_create_edges and
4237         initialize_inline_failed.
4238         (cgraph_expand_function): Do not call cgraph_lower_function;
4239         assert that function is already lowered.
4240         * Makefile.in (cgraphbuild.o): New.
4241         * passes.c (init_optimization_passes): Add pass_build_cgraph_edges
4242         at the end of lowering passes.
4243
4244 2007-01-06  Steven Bosscher  <steven@gcc.gnu.org>
4245
4246         * ifcvt.c (cond_move_convert_if_block): New function, code
4247         factored out from...
4248         (cond_move_process_if_block): ...here.  Call the new function
4249         on the THEN and ELSE blocks.
4250         (merge_if_block): Do not copy global_live_at_end, merge_blocks
4251         already takes care of this.
4252
4253 2007-01-05  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
4254
4255         PR c/19978
4256         * tree.h (TREE_OVERFLOW_P): New.
4257         * c-typeck.c (parser_build_unary_op): Warn only if result
4258         overflowed and operands did not.
4259         (parser_build_binary_op): Likewise.
4260         (convert_for_assignment): Remove redundant overflow_warning.
4261         * c-common.c (overflow_warning): Don't check or set TREE_OVERFLOW.
4262
4263 2007-01-05  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
4264
4265         * c-typeck.c (store_init_value): Split over two lines to follow
4266         the GNU coding style.
4267
4268 2007-01-05  Benjamin Kosnik  <bkoz@redhat.com>
4269
4270         * c-cppbuiltin.c (c_cpp_builtins): __GXX_EXPERIMENTAL_CPP0X__ to
4271         __GXX_EXPERIMENTAL_CXX0X__.
4272         * doc/cpp.texi: Same.
4273
4274 2007-01-05  Richard Guenther  <rguenther@suse.de>
4275
4276         PR middle-end/27826
4277         * tree.c (get_narrower): Do not construct COMPONENT_REFs
4278         with mismatched types.  Instead explicitly build a
4279         conversion NOP_EXPR.
4280
4281 2007-01-05  Ian Lance Taylor  <iant@google.com>
4282
4283         * c-common.c (decl_with_nonnull_addr_p): New function.
4284         (c_common_truthvalue_conversion): Call it.
4285         * c-typeck.c (build_binary_op): Likewise.
4286         * c-common.h (decl_with_nonnull_addr_p): Declare.
4287
4288 2007-01-05  Jakub Jelinek  <jakub@redhat.com>
4289
4290         PR c/30360
4291         * libgcc2.c (__divdc3): Compare c and d against 0.0 instead of
4292         denom against 0.0.
4293
4294 2007-01-05  Joel Brobecker  <brobecker@adacore.com>
4295
4296         * doc/install.texi (Final install): Document the fact that
4297         the GNAT runtime should not be stripped.
4298
4299 2007-01-04  Jan Hubicka  <jh@suse.cz>
4300
4301         * tree-inline.c (fold_marked_statements): Update operand caches
4302         and EH after folding
4303
4304 2007-01-04  Ian Lance Taylor  <iant@google.com>
4305
4306         * c-common.c (check_function_nonnull): Whitespace fix.
4307
4308 2007-01-04  Jan Hubicka  <jh@suse.cz>
4309
4310         * tree-optimize.c (execute_fixup_cfg): Correct previously mistakely
4311         comitted older version of patch.
4312         (pass_fixup_cfg): Add TODOs to verify flow and statements, dump
4313         function, celanup cfg and collect garbage.
4314
4315 2007-01-04  Mike Stump  <mrs@apple.com>
4316
4317         * Makefile.in (mostlyclean): Don't remove libgcc anymore.
4318         (clean): Likewise.
4319
4320 2007-01-04  Eric Christopher  <echristo@apple.com>
4321
4322         * libgcc2.c (__bswapsi2): Use SItype.
4323         (__bswapdi2): Use DItype.
4324         * libgcc2.h: Update for above.
4325
4326 2007-01-04  Paul Brook  <paul@codesourcery.com>
4327
4328         * config/arm/arm.md (arm_mulsi3, thumb_mulsi3, mulsi3_compare0,
4329         mulsi_compare0_scratch, mulsi3addsi, mulsi3addsi_compare0,
4330         mulsi3addsi_compare0_scratch, mulsidi3adddi, mulsidi3,
4331         umulsidi3, umulsidi3adddi, smulsi3_highpart,
4332         umulsi3_highpart): Make conditional on !arm_arch6.
4333         (arm_mulsi3_v6, thumb_mulsi3_v6, mulsi3_compare0_v6,
4334         mulsi_compare0_scratch_v6, mulsi3addsi_v6, mulsi3addsi_compare0_v6,
4335         mulsi3addsi_compare0_scratch_v6, mulsidi3adddi_v6, mulsidi3_v6,
4336         umulsidi3_v6, umulsidi3adddi_v6, smulsi3_highpart_v6,
4337         umulsi3_highpart_v6): New insns.
4338
4339 2007-01-04  Roger Sayle  <roger@eyesopen.com>
4340
4341         * fold-const.c (fold_convert): When casting an expression to void,
4342         fold_ignored_result may discover a GIMPLE_MODIFY_STMT which doesn't
4343         have a type.  Instead of attempting to build a NOP_EXPR, return
4344         these "special" trees directly.
4345
4346 2007-01-04  Joseph Myers  <joseph@codesourcery.com>
4347
4348         * config/rs6000/rs6000.c (rs6000_rtx_costs): Make adjustment for
4349         MULT inside MINUS as either argument.  Use rs6000_cost->dmul -
4350         rs6000_cost->fp not 0 as adjustment for outer NEG.
4351
4352 2007-01-04  Jan Hubicka  <jh@suse.cz>
4353
4354         * cgraph.c (cgraph_release_function_body): New function.
4355         (cgraph_remove_node): Use it.
4356         * cgraph.h (cgraph_release_function_body): Declare.
4357         * cgraphunit.c (cgraph_expand_function): Use it.
4358         * ipa.c (cgraph_remove_unreahchable_nodes): Use it.
4359         * tree-ssa.c (delete_tree_ssa): Allow to be called before aliasing
4360         is initialized and while compilation of other function is running.
4361         * tree-optimize.c (execute_free_cfg_annotations): Move code to clear
4362         statement CFG annotations from here to ...
4363         * tree-cfg.c (delete_tree_cfg_annotations): ... here.
4364
4365 2007-01-04  Zdenek Dvorak <dvorakz@suse.cz>
4366
4367         * cfgloop.h (enum li_flags): Make the constants powers of two.
4368
4369 2007-01-04  Jan Hubicka  <jh@suse.cz>
4370
4371         * tree-inline.c (copy_bb): Insert new statements to statements_to_fold
4372         set.
4373         (fold_marked_statements): New function.
4374         (optimize_inline_calls, tree_function_versioning): Fold new statements.
4375         * tree-inline.h (copy_body_data): Add statements_to_fold.
4376
4377 2007-01-03  Daniel Jacobowitz  <dan@codesourcery.com>
4378
4379         * config.gcc: Mention libgcc/config.host.
4380         * Makefile.in: Update comments mentioning libgcc.
4381         (LIBGCC, INSTALL_LIBGCC, GCC_PARTS, mklibgcc): Delete.
4382         (all.cross, start.encap, rest.encap, rest.cross): Update
4383         dependencies for libgcc move.
4384         (libgcc.mk, LIBGCC_DEPS, libgcov.a, libgcc.a, stmp-multilib)
4385         (clean-target, clean-target-libgcc): Delete.
4386         (srcdirify, GCC_EXTRA_PARTS): New macros.
4387         (libgcc-support, libgcc.mvars): New rules.
4388         (distclean): Remove mention of mklibgcc.
4389         (install): Don't reference INSTALL_LIBGCC.
4390         (install-common): Don't reference EXTRA_PARTS.
4391         (install-libgcc, install-multilib): Delete rules.
4392         * mklibgcc.in: Delete file.
4393         * doc/configfiles.texi: Don't mention mklibgcc.
4394
4395         * config/i386/t-darwin (SHLIB_VERPFX): Delete (moved to libgcc).
4396         * config/i386/t-darwin64 (SHLIB_VERPFX): Likewise.
4397         * config/rs6000/t-darwin (SHLIB_VERPFX): Likewise.
4398         * config/rs6000/t-ppccomm (TARGET_LIBGCC2_CFLAGS, SHLIB_MAPFILES)
4399         (mklibgcc, ldblspecs): Likewise.
4400
4401         * config/i386/t-nwld (libgcc.def, libc.def, libpcre.def)
4402         (posixpre.def): Use $(T).
4403         (SHLIB_EXT, SHLIB_NAME, SHLIB_SLIBDIR_QUAL, SHLIB_DEF, SHLIB_MAP)
4404         (SHLIB_SRC, SHLIB_INSTALL): Delete.
4405         (SHLIB_LINK): Make dummy.
4406         * config/t-slibgcc-darwin: Delete contents except for dummy SHLIB_LINK.
4407
4408         * config/frv/t-linux (EXTRA_MULTILIB_PARTS): Clear.
4409
4410         * config/alpha/t-crtfm: Use $(T) in rules for EXTRA_PARTS.
4411         * config/alpha/t-vms, config/alpha/t-vms64, config/fr30/t-fr30,
4412         config/i386/t-rtems-i386, config/ia64/t-ia64, config/rs6000/t-beos,
4413         config/rs6000/t-newas, config/sparc/t-elf: Likewise.
4414
4415         * configure.ac (all_outputs): Remove mklibgcc.
4416         * configure: Regenerated.
4417
4418 2007-01-03  Josh Conner  <jconner@apple.com>
4419
4420         PR middle-end/29683
4421         * calls.c (compute_argument_addresses): Set stack and stack_slot
4422         for partial args, too.
4423         (store_one_arg): Use locate.size.constant for the size when
4424         generating a save_area.
4425
4426 2007-01-03  Robert Kennedy <jimbob@google.com>
4427
4428         * tree-cfg.c (tree_merge_blocks): Release SSA_NAME phi results
4429         whose definitions are deleted due to basic block merging.
4430
4431 2007-01-03  Paul Brook  <paul@codesourcery.com>
4432
4433         PR target/16634
4434         * config/arm/arm.c (output_return_instruction): Pop PC in interrupt
4435         functions.
4436         (use_return_insn): Return 0 for Thumb interrupt functions.
4437         (print_multi_reg): Add rfe argument for IRQ returns.
4438         (arm_output_epilogue): Pop interrupt return address directly into PC.
4439         (arm_expand_prologue): Only adjust IRQ return address in Arm mode.
4440
4441 2007-01-03  Paul Brook  <paul@codesourcery.com>
4442
4443         Merge from sourcerygxx-4_1.
4444         * config/arm/thumb2.md: New file.
4445         * config/arm/elf.h (JUMP_TABLES_IN_TEXT_SECTION): Return True for
4446         Thumb-2.
4447         * config/arm/coff.h (JUMP_TABLES_IN_TEXT_SECTION): Ditto.
4448         * config/arm/aout.h (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
4449         (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump tables.
4450         * config/arm/aof.h (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump
4451         tables.
4452         (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
4453         * config/arm/ieee754-df.S: Use macros for Thumb-2/Unified asm
4454         comptibility.
4455         * config/arm/ieee754-sf.S: Ditto.
4456         * config/arm/arm.c (thumb_base_register_rtx_p): Rename...
4457         (thumb1_base_register_rtx_p): ... to this.
4458         (thumb_index_register_rtx_p): Rename...
4459         (thumb1_index_register_rtx_p): ... to this.
4460         (thumb_output_function_prologue): Rename...
4461         (thumb1_output_function_prologue): ... to this.
4462         (thumb_legitimate_address_p): Rename...
4463         (thumb1_legitimate_address_p): ... to this.
4464         (thumb_rtx_costs): Rename...
4465         (thumb1_rtx_costs): ... to this.
4466         (thumb_compute_save_reg_mask): Rename...
4467         (thumb1_compute_save_reg_mask): ... to this.
4468         (thumb_final_prescan_insn): Rename...
4469         (thumb1_final_prescan_insn): ... to this.
4470         (thumb_expand_epilogue): Rename...
4471         (thumb1_expand_epilogue): ... to this.
4472         (arm_unwind_emit_stm): Rename...
4473         (arm_unwind_emit_sequence): ... to this.
4474         (thumb2_legitimate_index_p, thumb2_legitimate_address_p,
4475         thumb1_compute_save_reg_mask, arm_dwarf_handle_frame_unspec,
4476         thumb2_index_mul_operand, output_move_vfp, arm_shift_nmem,
4477         arm_save_coproc_regs, thumb_set_frame_pointer, arm_print_condition,
4478         thumb2_final_prescan_insn, thumb2_asm_output_opcode, arm_output_shift,
4479         thumb2_output_casesi): New functions.
4480         (TARGET_DWARF_HANDLE_FRAME_UNSPEC): Define.
4481         (FL_THUMB2, FL_NOTM, FL_DIV, FL_FOR_ARCH6T2, FL_FOR_ARCH7,
4482         FL_FOR_ARCH7A, FL_FOR_ARCH7R, FL_FOR_ARCH7M, ARM_LSL_NAME,
4483         THUMB2_WORK_REGS): Define.
4484         (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv, arm_condexec_count,
4485         arm_condexec_mask, arm_condexec_masklen)): New variables.
4486         (all_architectures): Add armv6t2, armv7, armv7a, armv7r and armv7m.
4487         (arm_override_options): Check new CPU capabilities.
4488         Set new architecture flag variables.
4489         (arm_isr_value): Handle v7m interrupt functions.
4490         (user_return_insn): Return 0 for v7m interrupt functions.  Handle
4491         Thumb-2.
4492         (const_ok_for_arm): Handle Thumb-2 constants.
4493         (arm_gen_constant): Ditto.  Use movw when available.
4494         (arm_function_ok_for_sibcall): Return false for v7m interrupt
4495         functions.
4496         (legitimize_pic_address, arm_call_tls_get_addr): Handle Thumb-2.
4497         (thumb_find_work_register, arm_load_pic_register,
4498         legitimize_tls_address, arm_address_cost, load_multiple_sequence,
4499         emit_ldm_seq, emit_stm_seq, arm_select_cc_mode, get_jump_table_size,
4500         print_multi_reg, output_mov_long_double_fpa_from_arm,
4501         output_mov_long_double_arm_from_fpa, output_mov_double_fpa_from_arm,
4502         output_mov_double_fpa_from_arm, output_move_double,
4503         arm_compute_save_reg_mask, arm_compute_save_reg0_reg12_mask,
4504         output_return_instruction, arm_output_function_prologue,
4505         arm_output_epilogue, arm_get_frame_offsets, arm_regno_class,
4506         arm_output_mi_thunk, thumb_set_return_address): Ditto.
4507         (arm_expand_prologue): Handle Thumb-2.  Use arm_save_coproc_regs.
4508         (arm_coproc_mem_operand): Allow POST_INC/PRE_DEC.
4509         (arithmetic_instr, shift_op): Use arm_shift_nmem.
4510         (arm_print_operand): Use arm_print_condition.  Handle '(', ')', '.',
4511         '!' and 'L'.
4512         (arm_final_prescan_insn): Use extract_constrain_insn_cached.
4513         (thumb_expand_prologue): Use thumb_set_frame_pointer.
4514         (arm_file_start): Output directive for unified syntax.
4515         (arm_unwind_emit_set): Handle stack alignment instruction.
4516         * config/arm/lib1funcs.asm: Remove default for __ARM_ARCH__.
4517         Add v6t2, v7, v7a, v7r and v7m.
4518         (RETLDM): Add Thumb-2 code.
4519         (do_it, shift1, do_push, do_pop, COND, THUMB_SYNTAX): New macros.
4520         * config/arm/arm.h (TARGET_CPU_CPP_BUILTINS): Define __thumb2__.
4521         (TARGET_THUMB1, TARGET_32BIT, TARGET_THUMB2, TARGET_DSP_MULTIPLY,
4522         TARGET_INT_SIMD, TARGET_UNIFIED_ASM, ARM_FT_STACKALIGN, IS_STACKALIGN,
4523         THUMB2_TRAMPOLINE_TEMPLATE, TRAMPOLINE_ADJUST_ADDRESS,
4524         ASM_OUTPUT_OPCODE, THUMB2_GO_IF_LEGITIMATE_ADDRESS,
4525         THUMB2_LEGITIMIZE_ADDRESS, CASE_VECTOR_PC_RELATIVE,
4526         CASE_VECTOR_SHORTEN_MODE, ADDR_VEC_ALIGN, ASM_OUTPUT_CASE_END,
4527         ADJUST_INSN_LENGTH): Define.
4528         (TARGET_REALLY_IWMMXT, TARGET_IWMMXT_ABI, CONDITIONAL_REGISTER_USAGE,
4529         STATIC_CHAIN_REGNUM, HARD_REGNO_NREGS, INDEX_REG_CLASS,
4530         BASE_REG_CLASS, MODE_BASE_REG_CLASS, SMALL_REGISTER_CLASSES,
4531         PREFERRED_RELOAD_CLASS, SECONDARY_OUTPUT_RELOAD_CLASS,
4532         SECONDARY_INPUT_RELOAD_CLASS, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P,
4533         TRAMPOLINE_SIZE, INITIALIZE_TRAMPOLINE, HAVE_PRE_INCREMENT,
4534         HAVE_POST_DECREMENT, HAVE_PRE_DECREMENT, HAVE_PRE_MODIFY_DISP,
4535         HAVE_POST_MODIFY_DISP, HAVE_PRE_MODIFY_REG, HAVE_POST_MODIFY_REG,
4536         REGNO_MODE_OK_FOR_BASE_P, LEGITIMATE_CONSTANT_P,
4537         REG_MODE_OK_FOR_BASE_P, REG_OK_FOR_INDEX_P, GO_IF_LEGITIMATE_ADDRESS,
4538         LEGITIMIZE_ADDRESS, THUMB2_LEGITIMIZE_ADDRESS,
4539         GO_IF_MODE_DEPENDENT_ADDRESS, MEMORY_MOVE_COST, BRANCH_COST,
4540         ASM_APP_OFF, ASM_OUTPUT_CASE_LABEL, ARM_DECLARE_FUNCTION_NAME,
4541         FINAL_PRESCAN_INSN, PRINT_OPERAND_PUNCT_VALID_P,
4542         PRINT_OPERAND_ADDRESS): Adjust for Thumb-2.
4543         (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv): New declarations.
4544         * config/arm/arm-cores.def: Add arm1156t2-s, cortex-a8, cortex-r4 and
4545         cortex-m3.
4546         * config/arm/arm-tune.md: Regenerate.
4547         * config/arm/arm-protos.h: Update prototypes.
4548         * config/arm/vfp.md: Enable patterns for Thumb-2.
4549         (arm_movsi_vfp): Add movw alternative.  Use output_move_vfp.
4550         (arm_movdi_vfp, movsf_vfp, movdf_vfp): Use output_move_vfp.
4551         (thumb2_movsi_vfp, thumb2_movdi_vfp, thumb2_movsf_vfp,
4552         thumb2_movdf_vfp, thumb2_movsfcc_vfp, thumb2_movdfcc_vfp): New.
4553         * config/arm/libunwind.S: Add Thumb-2 code.
4554         * config/arm/constraints.md: Update include Thumb-2.
4555         * config/arm/ieee754-sf.S: Add Thumb-2/Unified asm support.
4556         * config/arm/ieee754-df.S: Ditto.
4557         * config/arm/bpabi.S: Ditto.
4558         * config/arm/t-arm (MD_INCLUDES): Add thumb2.md.
4559         * config/arm/predicates.md (low_register_operand,
4560         low_reg_or_int_operand, thumb_16bit_operator): New.
4561         (thumb_cmp_operand, thumb_cmpneg_operand): Rename...
4562         (thumb1_cmp_operand, thumb1_cmpneg_operand): ... to this.
4563         * config/arm/t-arm-elf: Add armv7 multilib.
4564         * config/arm/arm.md: Update patterns for Thumb-2 and Unified asm.
4565         Include thumb2.md.
4566         (UNSPEC_STACK_ALIGN, ce_count): New.
4567         (arm_incscc, arm_decscc, arm_umaxsi3, arm_uminsi3,
4568         arm_zero_extendsidi2, arm_zero_extendqidi2): New
4569         insns/expanders.
4570         * config/arm/fpa.md: Update patterns for Thumb-2 and Unified asm.
4571         (thumb2_movsf_fpa, thumb2_movdf_fpa, thumb2_movxf_fpa,
4572         thumb2_movsfcc_fpa, thumb2_movdfcc_fpa): New insns.
4573         * config/arm/cirrus.md: Update patterns for Thumb-2 and Unified asm.
4574         (cirrus_thumb2_movdi, cirrus_thumb2_movsi_insn,
4575         thumb2_cirrus_movsf_hard_insn, thumb2_cirrus_movdf_hard_insn): New
4576         insns.
4577         * doc/extend.texi: Document ARMv7-M interrupt functions.
4578         * doc/invoke.texi: Document Thumb-2 new cores+architectures.
4579
4580 2007-01-03  Jakub Jelinek  <jakub@redhat.com>
4581
4582         * unwind-dw2.c (SIGNAL_FRAME_BIT, EXTENDED_CONTEXT_BIT): Define.
4583         (struct _Unwind_Context): Rename args_size to flags, remove
4584         signal_frame field, add a new args_size field and version field.
4585         (_Unwind_IsSignalFrame, _Unwind_SetSignalFrame,
4586         _Unwind_IsExtendedContext): New inline functions.
4587         (_Unwind_GetGR, _Unwind_SetGR, _Unwind_GetGRPtr, _Unwind_SetGRPtr):
4588         Assume by_value array is only present if _Unwind_IsExtendedContext.
4589         (_Unwind_GetIPInfo, execute_cfa_program, uw_frame_state_for): Use
4590         _Unwind_IsSignalFrame.
4591         (__frame_state_for): Initialize context.flags to EXTENDED_CONTEXT_BIT.
4592         (uw_update_context_1): Use _Unwind_SetSignalFrame.
4593         (uw_init_context_1): Initialize context->flags to
4594         EXTENDED_CONTEXT_BIT.
4595         * config/rs6000/linux-unwind.h (frob_update_context): Use
4596         _Unwind_SetSignalFrame.
4597
4598 2007-01-03  Andrew Pinski  <andrew_pinski@playstation.sony.com>
4599
4600         PR middle-end/30353
4601         * gimplify.c (gimplify_modify_expr_complex_part): Move below
4602         tree_to_gimple_tuple.  Call tree_to_gimple_tuple when we need
4603         the value.
4604
4605 2007-01-03  Kazu Hirata  <kazu@codesourcery.com>
4606
4607         * config/i386/i386.h (NON_STACK_REG_P, REGNO_OK_FOR_SIREG_P,
4608         REGNO_OK_FOR_DIREG_P, REWRITE_ADDRESS, ASM_OPERAND_LETTER,
4609         RET, AT_SP): Remove.
4610         * config/i386/i386.md (*sse_prologue_save_insn): Use return
4611         instead of RET.
4612
4613         * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
4614         * cfgbuild.c (find_basic_blocks): Likewise.
4615         * cfgrtl.c (rtl_create_basic_block): Likewise.
4616         * function.c (temp_slots_at_level): Likewise.
4617         * reg-stack.c (stack_regs_mentioned): Likewise.
4618         * regclass.c (allocate_reg_info): Likewise.
4619         * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
4620         set_bb_for_stmt, move_block_to_fn): Likewise.
4621         * tree-complex.c (tree_lower_complex): Likewise.
4622         * vec.h (VEC_safe_grow_cleared): New.
4623
4624         * cgraphunit.c, tree-ssa-alias.c: Fix comment typos.
4625
4626 2007-01-03  Zdenek Dvorak <dvorakz@suse.cz>
4627
4628         * loop-unswitch.c (unswitch_loop): Pass probabilities to loopify.
4629         * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Pass probabilities
4630         to loop_version.
4631         * cfgloopmanip.c (scale_loop_frequencies): Export.
4632         (loopify): Scale the frequencies by prescribed coefficients.
4633         (set_zero_probability): New function.
4634         (duplicate_loop_to_header_edge): Improve updating of frequencies.
4635         (lv_adjust_loop_entry_edge, loop_version): Set probabilities
4636         and frequencies according to arguments.
4637         * tree-ssa-loop-manip.c (tree_unroll_loop): Set probabilities
4638         correctly.
4639         * cfg.c (scale_bbs_frequencies_int): Allow scaling the frequencies up.
4640         * modulo-sched.c (sms_schedule): Set probabilities for entering
4641         versioned loop correctly.
4642         * tree-vect-transform.c (vect_transform_loop): Ditto.
4643         * cfgloop.h (loopify, loop_version): Declaration changed.
4644         (scale_loop_frequencies): Declared.
4645
4646 2007-01-02  Jan Hubicka  <jh@suse.cz>
4647
4648         * cgraph.c: Include tree-flow.h
4649         (cgraph_add_new-function): Handle IPA_SSA mode; execute
4650         early_local_passes.
4651         * cgraph.h (enum cgraph_state): Add CGRAPH_STATE_IPA_SSA.
4652         * tree-pass.h (pass_all_early_optimizations): Declare.
4653         * cgraphunit.c (cgraph_process_new_functions): Add IPA_SSA; execute
4654         early_local_passes.
4655         (cgraph_analyze_function): Do early_local_passes.
4656         * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
4657         Do not add referenced vars.
4658         * tree-optimize.c (gate_all_optimizations): Do not execute when not in
4659         SSA form.
4660         (gate_all_early_local_passes): New gate.
4661         (pass_early_local_passes): Use new gate.
4662         (execute_early_local_optimizations): New functions.
4663         (gate_all_early_optimizations): New gate.
4664         (pass_all_early_optimizations): New pass.
4665         (execute_free_datastructures): Free SSA only when initialized.
4666         (gate_init_datastructures): Init only when optimizing.
4667         (tree_lowering_passes): Do early local passes when called late.
4668         * tree-profile.c (do_tree_profiling): Don't profile functions added
4669         late.
4670         (do_early_tree_profiling, pass_early_tree_profile): Kill.
4671         * tree-cfg.c (update_modified_stmts): Do not update when operands are
4672         not active.
4673         * passes.c (init_optimizations_passes): Reorder so we go into SSA
4674         during early_local_passes.
4675         * Makefile.in (cgraph.o): Add dependency on tree-flow.h.
4676
4677
4678 2007-01-02  Carlos O'Donell  <carlos@codesourcery.com>
4679
4680         * Makefile.in: Update copyright year.
4681
4682 2007-01-02  Carlos O'Donell  <carlos@codesourcery.com>
4683
4684         * Makefile.in: Export GCC_EXEC_PREFIX before calling $(RUNTEST)
4685         in $(lang_checks) and check-consistency targets.
4686
4687 2007-01-02  Jan Hubicka  <jh@suse.cz>
4688
4689         * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
4690         Do not add referenced vars.
4691         * tree-cfg.c (update_modified_stmts): Do not update when SSA operands
4692         are not active.
4693         * passes.c (init_optimization_passes): Put mudflap_2 after
4694         free_datastructures.
4695
4696 2007-01-02  Jan Hubicka  <jh@suse.cz>
4697
4698         * tree-optimize (execute_fixup_cfg): Set after_inlining flag.
4699         Set NOTHROW flag on call statements proved to be nothrow.
4700         Update statement of local calls so new pure/const functions are
4701         updated. Update_ssa when in ssa form. Mark PHI nodes of nonlocal
4702         goto receivers.
4703         (tree_rest_of_compilation): Register hooks and initialize bitmap
4704         early. Do not set after_inlining flag.
4705
4706 2007-01-02  Steve Ellcey  <sje@cup.hp.com>
4707
4708         * sbitmap.c (HOST_BITS_PER_LONG_LONG):  Change to
4709         HOST_BITS_PER_LONGLONG
4710
4711 2007-01-02  Manuel Lopez-Ibanez <manu@gcc.gnu.org>
4712
4713         PR c/19977
4714         * c-typeck.c (store_init_value): Don't emit pedantic overflow
4715         warning for non-static initializers.
4716
4717 2007-01-02  Steven Bosscher  <steven@gcc.gnu.org>
4718
4719         * config/alpha/alpha.md, arm/arm.c, darwin.c, frv/frv.md,
4720         m32r/m32r.c, m32r/m32r.c, mn10300/mn10300.md, pa/pa.c,
4721         rs6000/rs6000.c, s390/s390.md, sh/sh.md, sparc/sparc.c:
4722         Always use set_unique_reg_note to add REG_EQUAL notes.
4723
4724 2007-01-02  Kazu Hirata  <kazu@codesourcery.com>
4725
4726         Revert:
4727         2007-01-02  Kazu Hirata  <kazu@codesourcery.com>
4728
4729         * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
4730         * cfgbuild.c (find_basic_blocks): Likewise.
4731         * cfgrtl.c (rtl_create_basic_block): Likewise.
4732         * function.c (temp_slots_at_level): Likewise.
4733         * reg-stack.c (stack_regs_mentioned): Likewise.
4734         * regclass.c (allocate_reg_info): Likewise.
4735         * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
4736         set_bb_for_stmt, move_block_to_fn): Likewise.
4737         * tree-complex.c (tree_lower_complex): Likewise.
4738         * vec.h (VEC_safe_grow_cleared): New.
4739
4740 2007-01-02  Ian Lance Taylor  <iant@google.com>
4741
4742         * c-common.c (c_common_truthvalue_conversion): When warning about
4743         using an assignment as a truth value, set TREE_NO_WARNING.
4744
4745 2007-01-02  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
4746
4747         PR middle-end/7651
4748         * c.opt (Wold-style-declaration): New.
4749         * doc/invoke.texi (C-only Warning Options): New.
4750         (Wold-style-declaration): Document it.
4751         (Wextra): Enabled by -Wextra.
4752         * c-opts.c (c_common_post_options): Enabled by -Wextra.
4753         * c-decl.c (declspecs_add_scspec): Replace -Wextra with
4754         -Wold-style-declaration.
4755
4756 2007-01-02  Kazu Hirata  <kazu@codesourcery.com>
4757
4758         * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
4759         * cfgbuild.c (find_basic_blocks): Likewise.
4760         * cfgrtl.c (rtl_create_basic_block): Likewise.
4761         * function.c (temp_slots_at_level): Likewise.
4762         * reg-stack.c (stack_regs_mentioned): Likewise.
4763         * regclass.c (allocate_reg_info): Likewise.
4764         * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
4765         set_bb_for_stmt, move_block_to_fn): Likewise.
4766         * tree-complex.c (tree_lower_complex): Likewise.
4767         * vec.h (VEC_safe_grow_cleared): New.
4768
4769 2007-01-02  Douglas Gregor  <doug.gregor@gmail.com>
4770
4771         * c-common.c (c_common_nodes_and_builtins): Since variants of
4772         void_type_node get built before it is given a name, we need to
4773         give those variants the name, too.
4774         (complete_array_type): We need to work with the canonical main
4775         type of the array, from which we will build the qualified version.
4776         * params.def (PARAM_VERIFY_CANONICAL_TYPES): New.
4777         * print-tree.c (print_node): Display canonical type information
4778         for each type.
4779         * stor-layout.c (layout_type): When we don't know the
4780         alignment of a type for which we're building an array, we end up
4781         guessing wrong, so make the type require structural equality.
4782         * tree.c (make_node_stat): When we build a new type, it is its
4783         own canonical type.
4784         (build_type_attribute_qual_variant): When building an attribute
4785         variant, its canonical type is the non-attribute variant. However,
4786         if the attributes are target-dependent and they differ, we need to
4787         use structural equality checks for this type.
4788         (build_qualified_type): A qualified type is not equivalent to its
4789         unqualified variant; set the canonical type appropriately.
4790         (build_distinct_type_copy): When building a distinct type from
4791         another type, the new type is its own canonical type.
4792         (build_variant_type_copy): When building a new type variant, we
4793         assume that it is equivalent to the original type.
4794         (build_pointer_type_for_mode): When building a pointer type, also
4795         build a canonical type pointer.
4796         (build_reference_type_for_mode): When building a reference type,
4797         also build a canonical type reference.
4798         (build_index_type): When we can't hash an index type (e.g.,
4799         because its maximum value is negative), the index type requires
4800         structural equality tests.
4801         (build_array_type): Build the canonical form of an array type.
4802         (build_function_type): Function types require structural equality,
4803         because they contain default arguments, attributes, etc.
4804         (build_method_type_directly): Ditto for method types.
4805         (build_offset_type): Build the canonical offset type.
4806         (build_complex_type): Build the canonical vector type.
4807         (make_vector_type): Build the canonical vector type.
4808         * tree.h (TYPE_CANONICAL): New.
4809         (TYPE_STRUCTURAL_EQUALITY_P): New.
4810         (SET_TYPE_STRUCTURAL_EQUALITY): New.
4811         (struct tree_type): Added "canonical" field.
4812         * params.h (VERIFY_CANONICAL_TYPES): New.
4813         * doc/c-tree.texi (TYPE_CANONICAL): Document.
4814         (TYPE_STRUCTURAL_EQUALITY_P): Document.
4815         (SET_TYPE_STRUCTURAL_EQUALITY): Document.
4816         * doc/invoke.texi (verify-canonical-types): Document --param
4817         parameter for verifying canonical types.
4818
4819 2007-01-02  Joseph Myers  <joseph@codesourcery.com>
4820
4821         * config.gcc (powerpc-*-eabispe*, powerpc-*-eabisimaltivec*,
4822         powerpc-*-eabisim*, powerpc-*-eabialtivec*, powerpc-*-eabi*,
4823         powerpc-*-rtems*, powerpc-wrs-vxworks, powerpc-wrs-vxworksae,
4824         powerpcle-*-eabisim*, powerpcle-*-eabi*): Add rs6000/e500.h to
4825         tm_file.
4826         * config/rs6000/e500.h: New.
4827         * config/rs6000/eabi.h (TARGET_SPE_ABI, TARGET_SPE, TARGET_E500,
4828         TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE, TARGET_E500_DOUBLE):
4829         Remove.
4830         * config/rs6000/linuxspe.h (TARGET_SPE_ABI, TARGET_SPE,
4831         TARGET_E500, TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE,
4832         TARGET_E500_DOUBLE): Remove.
4833         * config/rs6000/vxworks.h (TARGET_SPE_ABI, TARGET_SPE,
4834         TARGET_E500, TARGET_ISEL, TARGET_FPRS): Remove.
4835         * config/rs6000/rs6000.h (CHECK_E500_OPTIONS): Define.
4836         * config/rs6000/rs6000.c (rs6000_override_options): Use
4837         CHECK_E500_OPTIONS.
4838
4839 2007-01-02  Joseph Myers  <joseph@codesourcery.com>
4840
4841         * config/rs6000/rs6000.c (print_operand): Check (TARGET_SPE ||
4842         TARGET_E500_DOUBLE), not TARGET_E500, for %y.
4843         (rs6000_generate_compare, rs6000_emit_sCOND, output_cbranch,
4844         rs6000_emit_cmove): Don't check TARGET_E500.
4845         * config/rs6000/rs6000.md (bunordered, bordered, sunordered,
4846         sordered): Don't check TARGET_E500.
4847
4848 2007-01-01  Eric Christopher  <echristo@apple.com>
4849
4850         * config/mips/mips.c (mips_regno_mode_ok_for_base_p): Use
4851         HARD_REGISTER_NUM_P.
4852
4853 2007-01-01  Roger Sayle  <roger@eyesopen.com>
4854
4855         * fold-const.c (fold_binary) <EQ_EXPR>: Fold "(X^C1) eq/ne C2" into
4856         "X eq/ne (C1^C2)".  Fold "(X^Z) eq/ne (Y^Z)" as "X eq/ne Y" when Z
4857         has no side-effects.  Fold "(X^C1) eq/ne (Y^C2)" as "(X^(C1^C2))
4858         eq/ne Y".
4859
4860 2007-01-01  Mike Stump  <mrs@apple.com>
4861
4862         * configure.ac: Remove support for building with Apple's gcc-3.1.
4863
4864 2007-01-02  Joseph Myers  <joseph@codesourcery.com>
4865
4866         PR middle-end/30311
4867         * caller-save.c (add_stored_regs): Only handle SUBREGs if inner
4868         REG is a hard register.  Do not modify REG before calling
4869         subreg_nregs.
4870         * rtlanal.c (subreg_get_info): Don't assert size of XMODE is a
4871         multiple of the size of YMODE for certain lowpart cases.
4872
4873 2007-01-01  Andrew Pinski  <pinskia@gmail.com>
4874
4875         PR middle-end/30253
4876         * gimplify (voidify_wrapper_expr): Update for
4877         GIMPLE_MODIFY_STMT.
4878
4879 2007-01-01  Andreas Schwab  <schwab@suse.de>
4880
4881         PR target/29166
4882         * config/ia64/ia64.c (ia64_compute_frame_size): Account space for
4883         save of BR0 in extra_spill_size instead of spill_size.
4884         (ia64_expand_prologue): Save BR0 outside of the gr/br/fr spill
4885         area.
4886         (ia64_expand_epilogue): Restore BR0 from its new location.
4887
4888 2007-01-01  Andrew Pinski  <pinskia@gmail.com>
4889
4890         * gimplify.c (gimplify_init_constructor <case VECTOR_TYPE>):
4891         Use a temporary variable if the left hand side is not a gimple
4892         register.
4893
4894 2007-01-01  Andrew Pinski  <pinskia@gmail.com>
4895
4896         * gimplify.c (gimplify_return_expr): Make the temporary variable
4897         for the return expression, a gimple register variable.
4898
4899 2007-01-01  Jan Hubicka  <jh@suse.cz>
4900
4901         * emit-rtl.c (emit_copy_of_insn_after): Do not call copy_insn_1 for
4902         INSN_LIST.
4903
4904 2007-01-01  Mike Stump  <mrs@apple.com>
4905
4906         * configure.ac (HAVE_GAS_LITERAL16): Add autoconf check for
4907         .literal16.
4908         * config/darwin.c (machopic_select_rtx_section): Use
4909         HAVE_GAS_LITERAL16.
4910         (darwin_mergeable_constant_section): Likewise.
4911         * configure: Regenerate.
4912         * config.in: Regenerate.
4913
4914 2007-01-01  Jan Hubicka  <jh@suse.cz>
4915             Andrew Pinski  <pinskia@gmail.com>
4916
4917         * cgraphunit.c (cgraph_optimize): Call cgraph_add_new_functions
4918         before starting IPA passes.