OSDN Git Service

de0752707b2d05a2856605edf0f2fff8816b4935
[pf3gnuchains/gcc-fork.git] / gcc / ChangeLog
1 2007-01-21  Ira Rosen  <irar@il.ibm.com>
2
3         * tree-vectorizer.h (struct _stmt_vec_info): Add new field 
4         read_write_dep and macros for its access.
5         * tree-vectorizer.c (new_stmt_vec_info): Initialize the new field.
6         * tree-vect-analyze.c (vect_analyze_data_ref_dependence): Remove 
7         argument, call vect_check_interleaving for every independent pair of
8         data-refs. Mark loads that access the same memory location as a store
9         in the loop.
10         (vect_check_dependences): Remove.
11         (vect_analyze_data_ref_dependences): Remove  vect_check_dependences 
12         call, fix the call to vect_analyze_data_ref_dependence.
13         (vect_analyze_data_ref_access): For statements that access the same 
14         data-ref, check that they are not stores; for loads, check that there
15         is no store that access the same location.
16
17 2007-01-20  Roger Sayle  <roger@eyesopen.com>
18             Joseph Myers  <joseph@codesourcery.com>
19
20         * doc/invoke.texi (-fdump-rtl-combine): Fix under/overfull hbox.
21         (-fdump-rtl-stack): Likewise.
22         (-fno-signed-zeros): Use @minus{} for a minus sign.  Correct typo.
23         (-mcheck-zero-division, -mcpu): Fix under/overfull hbox.
24         (-mpt-fixed): Use @minus{} for minus sign.
25         (Using Precompiled Headers): Fix under/overfull hbox.
26
27 2007-01-20  Jan Hubicka  <jh@suse.cz>
28
29         * tree-flow.h (struct stmt_ann_d): Move references_memory to proper
30         place within annotation.
31
32 2007-01-20  John David Anglin  <dave.anglin@nrc-cnrc.gc.ca>
33
34         * pa.c (output_move_double): Change array size of xoperands to 4.
35
36 2007-01-20  Jan Hubicka  <jh@suse.cz>
37
38         * tree-tailcall.c (adjust_return_value): Do not use RESULT_DECL
39         as temporary.
40
41 2007-01-19  Ian Lance Taylor  <iant@google.com>
42
43         * expmed.c (expand_divmod) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]: Cast
44         constant to unsigned HOST_WIDE_INT before negating.
45
46 2007-01-19  Ian Lance Taylor  <iant@google.com>
47
48         * tree-ssa-operands.h (struct vuse_vec_d): Change num_vuse field
49         to unsigned.
50         (VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Use unsigned comparison.
51         (VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
52         (SET_VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Likewise.
53         (SET_VUSE_ELEMENT_VAR) [ENABLE_CHECKING]: Likewise.
54         (SET_VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
55         (realloc_vdef, realloc_vuse): Change second parameter to
56         unsigned.
57         (ssa_operand_iterator_d): Change vuse_index and mayuse_index
58         fields to unsigned.
59         * tree-ssa-operands.c (realloc_vop): Change num_elem parameter to
60         unsigned. Change x and lim locals to unsigned.
61         (realloc_vdef, realloc_vuse): Change num_elem parameter to
62         unsigned.
63         (finalize_ssa_vuse_ops): Change old_i local to unsigned.
64         (copy_virtual_operands): Change i and n locals to unsigned.
65
66 2007-01-19  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
67
68         PR middle-end/29335
69         * builtins.c (fold_builtin_1): Handle builtin fdim.
70
71 2007-01-20  Jan Hubicka  <jh@suse.cz>
72
73         * tree-ssa.c (init_tree_ssa): Do not call init_alias_heapvars.
74         * tree-ssa-structalias.c (compute_points_to_sets): Do call
75         init_alias_heapvars.
76         (init_alias_heapvars): Initialize only when not already initialized.
77         (delete_alias_heapvars): Set heapvar_for_stmt to NULL.
78
79 2007-01-19  Roger Sayle  <roger@eyesopen.com>
80
81         * common.opt (fsigned-zeros): New command line option.
82         * flags.h (HONOR_SIGNED_ZEROS): Control via flag_signed_zeros instead
83         of flag_unsafe_math_optimizations.
84         * opts.c (set_fast_math_flags): The -ffast-math command line option
85         implies -fno-signed-zeros.
86         (fast_math_flags_set_p): Likewise.
87
88         * doc/invoke.texi: Document new -fno-signed-zeros option, and update
89         the documentation of -ffast-math appropriately.  Wrap long lines.
90
91 2007-01-19  Steve Ellcey  <sje@cup.hp.com>
92
93         * system.h (ASM_MAKE_LABEL_LINKONCE): Poison.
94         * varasm.c (globalize_decl): Remove ASM_MAKE_LABEL_LINKONCE ifdef.
95
96 2007-01-19  Tomas Bily  <tbily@suse.cz>
97
98         * cgraphunit.c (cgraph_finalize_function): Updating of pid
99         * tree-profile.c:
100         (tree_init_ic_make_global_vars): New function
101         (tree_init_edge_profiler): call of tree_init_ic_make_global_vars
102         (tree_gen_ic_profiler): New function
103         (tree_gen_ic_func_profiler): New function
104         (tree_profiling): Added calling of tree_gen_ic_func_profiler
105         (tree_profile_hooks): Added hook for indirec/virtual calls
106         * value-prof.c (tree_find_values_to_profile): New case for
107         indirect calls
108         (tree_values_to_profile): Call for determining indirect/virtual
109         counters
110         (tree_indirect_call_to_profile): New function
111         (tree_ic_transform): New function
112         (tree_ic): New function
113         (find_func_by_pid): New function
114         (init_pid_map): New function
115         (tree_value_profile_transformations): Added check for
116         indirect/virtual call transformation
117         * value-prof.h (enum hist_type): New counter type for
118         indirect/virtual calls
119         (profile_hooks): Added new hook for profiling indirect/virtual
120         calls
121         * profile.c (instrument_values): New case for indirect/virtual
122         call added
123         * gcov-io.h (GCOV_LAST_VALUE_COUNTER): Changed to 6
124         (GCOV_COUNTER_V_INDIR): New counter type
125         (GCOV_COUNTER_NAMES): New name of counter "indirect" added
126         (GCOV_MERGE_FUNCTIONS): New merge function for indirect/virtual
127         call added
128         * cgraph.c: Definition of cgraph_max_pid
129         (cgraph_create_node): Default init of pid attribute
130         * cgraph.h: Declaration of cgraph_max_pid
131         (struct cgraph_node): Added pid attribute
132         * libgcov.c (__gcov_indirect_call_profiler): New function
133         (__gcov_one_value_profiler_body): New function
134         (__gcov_one_value_profiler): Body was moved to
135         __gcov_one_value_profiler_body and calls it
136
137 2007-01-19  Basile Starynkevitch  <basile@starynkevitch.net>
138         
139         * doc/gty.texi (Options): Document the mark_hook option to GTY.
140         * gengtype.c (write_types_data, write_func_for_structure,
141         write_types, ggc_wtd, pch_wtd): Add skip_hooks to
142         write_types_data, ggc_wtd, pch_wtd for processing mark_hook.
143         (walk_type, write_func_for_structure): Generate the mark_hook if
144         needed.
145
146 2007-01-19  Jan Hubicka  <jh@suse.cz>
147
148         * ipa-inline.c (cgraph_decide_inlining_incrementally): Instead of 'early' argument
149         take inlining mode argument specifying whether to inline for size/speeed or all
150         functions; add support for flattening; improve dumpting.
151         (cgraph_early_inlining): Update call of decide_inlining_incrementally.
152
153 2007-01-19  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
154
155         PR c++/17947
156         * toplev.c (warn_deprecated_use): Use %qD instead of %qs to print
157         the name of the declared identifier.
158                 
159 2007-01-19  Dirk Mueller  <dmueller@suse.de>
160
161         * config/i386.h (CONDITIONAL_REGISTER_USAGE): Store
162         result of PIC_OFFSET_TABLE_REGNUM in temporary variable to avoid
163         duplicate evaluation.
164
165 2007-01-19  Uros Bizjak  <ubizjak@gmail.com>
166
167         * config/i386/i386.md (acos<mode>2): Rename from acossf2 and acosdf2.
168         Macroize expander using X87MODEF12 mode macro.  Extend operand 1
169         to XFMode, use acosxf2 and truncate result to requested mode.
170         Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
171         (asin<mode>2): Similarly, with asin expanders.
172         (*fscalexf4_i387): Rename from *fscalexf4.
173         (expNcorexf3): New expander.
174         (expxf2, exp10xf2, exp2xf2): Use expNcorexf3 expander.
175         (exp<mode>2): Rename from expsf2 and expdf2. Macroize expander using
176         X87MODEF12 mode macro.  Extend operand 1 to XFMode, use expxf2 and
177         truncate result to requested mode. Use SSE_FLOAT_MODE_P to disable
178         patterns for SSE math.
179         (exp10<mode>2): Similarly, with exp10 expanders.
180         (exp2<mode>2): Similarly, with exp2 expanders.
181         (expm1<mode>2): Similarly, with expm1 expanders.
182         (ldexp<mode>3): Similarly, with ldexp expanders.
183         (log<mode>2, log10<mode>2, log2<mode>2, log1p<mode>2, logb<mode>2):
184         Use gen_truncxf<mode>2_i387_noop to truncate result.
185
186 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
187
188         * config/m68k/m68k.h (PREFERRED_STACK_BOUNDARY): Define to 32
189         for ColdFire targets.
190
191 2007-01-19  Nathan Sidwell  <nathan@codesourcery.com>
192             Richard Sandiford  <richard@codesourcery.com>
193
194         * config/m68k/m68k.h (M68K_STATIC_CHAIN_REG_NAME): New macro.
195         (INITIALIZE_TRAMPOLINE): Use STATIC_CHAIN_REGNUM.
196         (__transfer_from_trampoline): Use M68K_STATIC_CHAIN_REG_NAME.
197         * config/m68k/m68kelf.h (STATIC_CHAIN_REGNUM): Override.
198         (M68K_STATIC_CHAIN_REG_NAME): Likewise.
199         * config/m68k/netbsd-elf.h (M68K_STATIC_CHAIN_REG_NAME): Likewise.
200
201 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
202
203         * config/m68k/m68k.md (adddi_dilshr32): Rename to...
204         (*adddi_dilshr32): ...this.  Fix formatting.  Remove commented-out
205         non-canonical pattern.  Restrict to !TARGET_COLDFIRE.
206         (*adddi_dilshr32_cf): New pattern.
207         (adddi3, subdi3): Remove first alternatives.
208
209 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
210
211         * config/m68k/m68k.c (notice_update_cc): If an SFmode move is
212         implemented using move.l, do not use its cc result for floating-point
213         comparisons.
214
215 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
216
217         * config/m68k/m68k.h (EXTRA_CONSTRAINT): Stop the 'T' constraint
218         from accepting 's' constraints if flag_pic.
219
220 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
221
222         * config/m68k/m68k.md (bordered, bunordered, buneq, bunge, bungt)
223         (bunle, bunlt, bltgt, bordered_rev, bunordered_rev, buneq_rev)
224         (bunge_rev, bungt_rev, bunle_rev, bunlt_rev, bltgt_rev): Change
225         condition from TARGET_68881 to TARGET_HARD_FLOAT.
226
227 2007-01-19  Sandra Loosemore  <sandra@codesourcery.com>
228
229         * longlong.h (count_leading_zeros, COUNT_LEADING_ZEROS_0): Add
230         ColdFire alternatives.
231         * config/m68k/m68k.h (CLZ_DEFINED_VALUE_AT_ZERO): New macro.
232         * config/m68k/m68k.md (clzsi2):  Define for ColdFire
233         architectures that support ff1 instruction.
234
235 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
236             Julian Brown  <julian@codesourcery.com>
237
238         * config/m68k/m68k.h (CONST_OK_FOR_LETTER_P): Add an 'R' case.
239         * config/m68k/m68k.md (*movsi_cfv4): Fold into...
240         (*movsi_cf): ...here.  Remove unnecessary 'R' from 'Rg'.
241         Add commentary.
242
243 2007-01-19  Richard Sandiford  <richard@codesourcery.com>
244
245         * config/m68k/m68k-protos.h (valid_mov3q_const): Take a HOST_WIDE_INT
246         and return a bool.
247         (output_move_const_into_data_reg, output_move_simode_const): Delete.
248         * config/m68k/m68k.c (const_method, const_int_cost): Take a
249         HOST_WIDE_INT instead of an rtx.
250         (m68k_rtx_costs): Update call accordingly.
251         (output_move_const_into_data_reg): Likewise.  Fix formatting.
252         (valid_mov3q_const): Take a HOST_WIDE_INT instead of an rtx.
253         Return a bool.
254         (output_move_simode_const): Update calls after above changes.
255         Rework to use automatic variables and predicates like MEM_P.
256         * config/m68k/m68k.md (pushexthisi_const): Update call to
257         valid_mov3q_const.
258
259 2007-01-19  Dirk Mueller  <dmueller@suse.de>
260
261         * tree-ssa-alias.c (perform_var_substitution): Fix typo
262         in dump_flags test.
263
264 2007-01-19  Richard Guenther  <rguenther@suse.de>
265
266         * builtins.c (expand_builtin_cexpi): Fall back to expanding
267         via cexp in case sincos is not available.
268
269 2007-01-19  Richard Guenther  <rguenther@suse.de>
270
271         * doc/tm.texi (TARGET_HAS_SINCOS): Document new target macro.
272         * defaults.h (TARGET_HAS_SINCOS): Default to off.
273         * config/linux.h (TARGET_HAS_SINCOS): Set to on if we have glibc.
274         * config/alpha/linux.h (TARGET_HAS_SINCOS): Likewise.
275         * config/sparc/linux.h (TARGET_HAS_SINCOS): Likewise.
276         * config/sparc/linux64.h (TARGET_HAS_SINCOS): Likewise.
277         * config/rs6000/linux.h (TARGET_HAS_SINCOS): Likewise.
278         * config/rs6000/linux64.h (TARGET_HAS_SINCOS): Likewise.
279
280 2007-01-19  Uros Bizjak  <ubizjak@gmail.com>
281
282         * config/i386/i386.md (*fpatanxf3_i387, fpatan_extend<mode>xf3_i387):
283         New insn patterns.
284         (atan2sf3_1, atan2df3_1, atan2xf3_1): Remove insn patterns.
285         (atan2xf3): Directly generate RTL pattern.
286         (atan2<mode>3): Rename from atan2sf3 and atan2df3 and macroize insn
287         patterns using X87MODEF12 mode macro.  Use fpatan_extend<mode>xf3_i387
288         and truncate result to requested mode.  Use SSE_FLOAT_MODE_P to
289         disable patterns for SSE math.
290         (atan<mode>2): Rename from atansf2 and atandf2 and macroize insn
291         patterns using X87MODEF12 mode macro.  Use fpatan_extend<mode>xf3_i387
292         and truncate result to requested mode.  Use SSE_FLOAT_MODE_P to
293         disable patterns for SSE math.
294
295 2007-01-19  Alexandre Oliva  <aoliva@redhat.com>
296
297         * libgcc-std.ver: Fix typo in %inherit for GCC_4.3.0.
298
299 2007-01-18  Roger Sayle  <roger@eyesopen.com>
300
301         * fold-const.c (fold_unary) <VIEW_CONVERT_EXPR>: Optimize away a
302         VIEW_CONVERT_EXPR to the same type as it's operand.
303
304 2007-01-18  David Edelsohn  <edelsohn@gnu.org>
305
306         * config/rs6000/darwin-ldouble.c: Only build _SOFT_FLOAT if
307         configured for long double 128.
308
309 2007-01-18  Mike Stump  <mrs@apple.com>
310
311         * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Fix build
312         error.
313
314 2007-01-18  Michael Meissner  <michael.meissner@amd.com>
315
316         * i386.c (ix86_compute_frame_layout): Make fprintf's in #if 0 code
317         type correct.
318
319 2007-01-18  Jan Hubicka  <jh@suse.cz>
320
321         * tree-ssa-operands.c (vop_free_bucket_size): Never return value
322         greater than NUM_VOP_FREE_BUCKETS.
323
324 2007-01-18  Daniel Berlin  <dberlin@dberlin.org>
325
326         * tree-ssa-structalias.c: Update comments.
327         (ptabitmap_obstack): Removed.
328         (pta_obstack): New.
329         (oldpta_obstack): Ditto.
330         (stats): Add a few members.
331         (struct variable_info): Remove node, complex, address_taken, and
332         indirect_target members. Add oldsolution member.
333         (new_var_info): Do not initialize removed members.
334         (constraint_expr_type): Remove INCLUDES.
335         (constraint_graph): Add size, implicit_preds, rep,
336         indirect_cycles, eq_rep, label, direct_nodes, and complex members.
337         (FIRST_REF_NODE): New macro.
338         (LAST_REF_NODE): Ditto.
339         (FIRST_ADDR_NODE): Ditto.
340         (find): New function.
341         (unite): Ditto.
342         (dump_constraint): Do not handle INCLUDES.
343         (insert_into_complex): Do not insert duplicate constraints.
344         (condense_varmap_nodes): Renamed and rewritten into ...
345         (merge_node_constraints): This. Also fix bug in handling of
346         offseted copy constraints.
347         (clear_edges_for_node): No longer need to deal with preds at all,
348         or removing associated preds/succs. 
349         (merge_graph_nodes): Deal with indirect_cycles.
350         Don't deal with predecessors.
351         (add_implicit_graph_edge): New function.
352         (add_pred_graph_edge): Ditto.
353         (add_graph_edge): Don't deal with predecessors.
354         (build_constraint_graph): Removed.
355         (build_pred_graph): New function.
356         (build_succ_graph): Ditto.
357         (struct scc_info): Removed in_component. Added roots, dfs, and
358         node_mapping. Remove visited_index, unification_queue.
359         (scc_visit): Deal with union-find we do now.
360         Deal with cycles with REF nodes.
361         (collapse_nodes): Renamed and rewritten to ...
362         (unify_nodes): This.
363         (process_unification_queue): Removed.
364         (topo_visit): Cleanup
365         (do_da_constraint): Use find.
366         (do_sd_constraint): Ditto.
367         (do_ds_constraint): Ditto.
368         (do_complex_constraint): Ditto.
369         (init_scc_info): Update for removed and added members.
370         (find_and_collapse_graph_cycles): Renamed and rewritten into ...
371         (find_indirect_cycles): This.
372         (equivalence_class): New variable.
373         (label_visit): New function.
374         (perform_variable_substitution): Rewritten.
375         (free_var_substitution_info): New function.
376         (find_equivalent_node): Ditto.
377         (move_complex_constraints): Ditto.
378         (eliminate_indirect_cycles): Ditto.
379         (solve_graph): Only propagate changed bits.
380         Use indirect cycle elimination.
381         Use find.
382         (tree_id_t): Rename to tree_vi_t, delete id member, add vi member.
383         (tree_id_eq): Renamed to ...
384         (tree_vi_eq): This. Update for member change
385         (insert_id_for_tree): Renamed and rewritten to ...
386         (insert_vi_for_tree): This.
387         (lookup_id_for_tree): Renamed and rewritten to ...
388         (lookup_vi_for_tree): This.
389         (get_id_for_tree): Renamed and rewritten to ...
390         (get_vi_for_tree): Ditto.
391         (get_constraint_exp_from_ssa_var): Update to use get_vi_for_tree.
392         (process_constraint): Don't handle INCLUDES.
393         Remove special ADDRESSOF case.
394         (find_func_aliases): Rewrite to use vi functions instead of id
395         ones.
396         (create_function_info_for): Ditto.
397         (create_variable_info_for): Ditto.
398         (intra_create_variable_infos): Ditto.
399         (merge_smts_into): Ditto.
400         (find_what_p_points_to): Ditto.
401         (init_base_vars): Ditto.
402         (init_alias_vars): Ditto.
403         (remove_preds_and_fake_succs): New function.
404         (dump_sa_points_to_info): Dump new stats.
405         (dump_solution_for_var): Use find.
406         (set_used_smts): Fix formatting.
407         (compute_points_to_sets): Updated for new functions.
408         (ipa_pta_execute): Ditto.
409
410 2007-01-18  Kazu Hirata  <kazu@codesourcery.com>
411             Richard Sandiford  <richard@codesourcery.com>
412
413         * doc/tm.texi (TARGET_FUNCTION_VALUE): Expand documentation of
414         parallels.
415         * calls.c (expand_call): If the return value is a PARALLEL,
416         extract its first member.
417         * config/m68k/linux.h (FUNCTION_EXTRA_EPILOGUE): Remove.
418         * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
419         use FUNCTION_EXTRA_EPILOGUE.
420         (m68k_function_value): Return a PARALLEL if the return value
421         is of a pointer type.
422         * config/m68k/netbsd-elf.h (current_function_returns_pointer)
423         (FUNCTION_EXTRA_EPILOGUE): Remove.
424         * config/m68k/m68k.md (D0_REG): New constant.
425
426 2007-01-18  Kazu Hirata  <kazu@codesourcery.com>
427
428         * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
429         output a NOP for empty epilogues.
430
431 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
432
433         * config/m68k/m68k.c (m68k_use_return_insn): Update comments
434         before function.  Extend register save check to include all
435         registers, not just integer ones.
436
437 2007-01-18  Kazu Hirata  <kazu@codesourcery.com>
438
439         * config/m68k/m68k-protos.h (use_return_insn): Rename to...
440         (m68k_use_return_insn): ...this.
441         * config/m68k/m68k.h (USE_RETURN_INSN): Delete.
442         * config/m68k/m68k.c (use_return_insn): Rename to...
443         (m68k_use_return_insn): ...this.
444         * config/m68k/m68k.md (return): Use m68k_use_return_insn instead
445         of USE_RETURN_INSN.
446
447 2007-01-18  Nathan Sidwell  <nathan@codesourcery.com>
448
449         * config/m68k/fpgnulib.c (__truncdfsf2): Implement round to
450         nearest even, fix denormal rounding overflow.
451
452 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
453
454         * config/m68k/m68k.md (movsf_cf_hard): Use fsmove instead of
455         f%$smove and f%$move.
456         (movdf_cf_hard): Use fdmove for cases 0 and 3 and fmove for case 1.
457         (extendsfdf2_cf): Use fdmove instead of f%&move.
458         (truncdfsf2_cf): Use fsmove instead of f%$smove.
459         (add<mode>3_cf, sub<mode>3_cf): Use <FP:prec> instead of <FP:round>.
460
461 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
462
463         * config/m68k/m68k.md (movdf_cf_hard): Use output_move_double for
464         GPR<-GPR moves.
465
466 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
467
468         * real.h (real_format): Add a canonical_nan_lsbs_set field.
469         (coldfire_single_format): Declare.
470         (coldfire_double_format): Likewise.
471         * real.c (encode_ieee_single): Use canonical_nan_lsbs_set instead
472         of qnan_msb_set to determine the lower bits of a canonical
473         NaN significand.
474         (encode_ieee_double): Likewise.
475         (encode_ieee_quad): Likewise.
476         (ieee_single_format): Initialize canonical_nan_lsbs_set.
477         (mips_single_format): Likewise.
478         (ieee_double_format): Likewise.
479         (mips_double_format): Likewise.
480         (ieee_extended_motorola_format): Likewise.
481         (ieee_extended_intel_96_format): Likewise.
482         (ieee_extended_intel_128_format): Likewise.
483         (ieee_extended_intel_96_round_53_format): Likewise.
484         (ibm_extended_format): Likewise.
485         (mips_extended_format): Likewise.
486         (ieee_quad_format): Likewise.
487         (mips_quad_format): Likewise.
488         (vax_f_format): Likewise.
489         (vax_d_format): Likewise.
490         (vax_g_format): Likewise.
491         (i370_single_format): Likewise.
492         (i370_double_format): Likewise.
493         (decimal_single_format): Likewise.
494         (decimal_double_format): Likewise.
495         (decimal_quad_format): Likewise.
496         (c4x_single_format): Likewise.
497         (c4x_extended_format): Likewise.
498         (real_internal_format): Likewise.
499         (coldfire_single_format): New real_format.
500         (coldfire_double_format): Likewise.
501         * config/pdp11/pdp11.c (pdp11_f_format): Initialize
502         canonical_nan_lsbs_set.
503         (pdp11_d_format): Likewise.
504         * config/m68k/m68k.c (override_options): Override REAL_FORMAT_MODE
505         if TARGET_COLDFIRE_CPU.
506
507 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
508
509         * config/m68k/m68k-protos.h (m68k_output_pic_call): Delete.
510         (output_call, m68k_legitimize_call_address): Declare.
511         * config/m68k/m68k.h (EXTRA_CONSTRAINT): Remove unnecessary
512         parenthesees.  Add support for a 'W' constraint.
513         (LEGITIMATE_PIC_OPERAND_P): Remove SYMBOL_REF_FLAG handling.
514         (PRINT_OPERAND_PUNCT_VALID_P): Remove comment about 'o'.
515         (m68k_symbolic_call, m68k_symbolic_jump): Declare.
516         * config/m68k/m68k.c (m68k_symbolic_call, m68k_symbolic_jump): New
517         variables.
518         (override_options): Initialize them.  Do not set flag_no_function_cse
519         for TARGET_ID_SHARED_LIBRARY.
520         (m68k_output_pic_call): Delete.
521         (m68k_legitimize_call_address): New function.
522         (print_operand): Remove the %o prefix.  Handle the %p prefix.
523         (output_call): New function.
524         (m68k_output_mi_thunk): Use m68k_symbolic_jump.  Always load the
525         target address from the GOT if symbolic jumps are not allowed.
526         * config/m68k/m68k.md (call, general_operand): Do not set
527         SYMBOL_REF_FLAG.  Use m68k_legitimize_call_address instead.
528         Merge separate flag_pic and !flag_pic define_insns into...
529         (*call, *call_value): ...these new patterns.  Match the address
530         rather than the containing MEM and require it to be a call_operand.
531         Use output_call to generate the asm template.
532         * config/m68k/predicates.md (const_call_operand): New predicate.
533         (call_operand): Likewise.
534
535 2007-01-18  Nathan Sidwell  <nathan@codesourcery.com>
536
537         * config/m68k/m68k.h (REGISTER_MOVE_COST): Simplify definition.
538         (STACK_GROWS_DOWNWARD): Define to 1.
539         (FUNCTION_VALUE, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P): Equivocate
540         comments, emphasizing that these values are only defaults.
541         * config/m68k/linux.h (LINK_SPEC): Fix formatting in #undef.
542         * config/m68k/m68k.c (const_method): Remove trailing whitespace.
543
544 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
545
546         * config/m68k/m68k.md (cmpsi): Remove outdated flag_pic handling.
547
548 2007-01-18  Kazu Hirata  <kazu@codesourcery.com>
549             Richard Sandiford  <richard@codesourcery.com>
550
551         * config/m68k/m68k.h (DATA_REGNO_P, ADDRESS_REGNO_P, INT_REGNO_P)
552         (FP_REGNO_P): New macros.
553         (REGNO_OK_FOR_INDEX_P, REGNO_OK_FOR_BASE_P, REGNO_OK_FOR_DATA_P)
554         (REGNO_OK_FOR_FP_P, REG_OK_FOR_INDEX_P, REG_OK_FOR_BASE_P): Use them.
555         (EH_RETURN_STACKADJ_RTX): Use A0_REG.
556         * config/m68k/m68k.c (m68k_regno_mode_ok): Use the new REGNO macros.
557
558 2007-01-18  Nathan Sidwell  <nathan@codesourcery.com>
559
560         * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
561         (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-rtems*): Add t-floatlib
562         to $tmake_file.
563         * config/m68k/t-floatlib: New file, extracting common code from...
564         * config/m68k/t-m68kbare, config/m68k/t-m68kelf,
565         * config/m68k/t-uclinux: Here.
566         * config/m68k/fpgnulib.c: Do not compile extendeed precision
567         routines on ColdFire targets.
568
569 2007-01-18  Nathan Sidwell  <nathan@codesourcery.com>
570
571         * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
572         (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*)
573         (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-linux*)
574         (m68k-*-rtems*): Use tm_file rather than m68k/m68k.h and
575         explicitly set MOTOROLA to 1.
576         * config/m68k/m68k.h (MOTOROLA): Simplify definition accordingly.
577
578 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
579             Nathan Sidwell  <nathan@codesourcery.com>
580
581         * config/m68k/m68k.h (PCC_STATIC_STRUCT_RETURN): Delete.
582         (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Add MOTOROLA cases.
583         * config/m68k/coff.h (REGISTER_PREFIX_MD): Delete.
584         * config/m68k/m68020-elf.h (LIB_SPEC): Delete.
585         * config/m68k/m68k-none.h (CC1_SPEC, CPP_SUBTARGET_SPEC): Delete.
586         * config/m68k/m68kelf.h (IMMEDIATE_PREFIX, REGISTER_PREFIX_MD)
587         (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Delete.
588         (NO_DOLLAR_IN_LABEL, PCC_STATIC_STRUCT_RETURN): Don't undefine.
589         (BSS_ASM_OP, ASM_OUTPUT_SKIP, ASM_OUTPUT_ASCII): Delete.
590         * config/m68k/m68kemb.h (PCC_STATIC_STRUCT_RETURN): Don't undefine.
591         * config/m68k/linux.h (SIZE_TYPE, PTRDIFF_TYPE, WCHAR_TYPE)
592         (WCHAR_TYPE_SIZE, TARGET_OBJFMT_CPP_BUILTINS): Delete.
593         (TARGET_OS_CPP_BUILTINS): Don't define mc68000 and mc68020 here.
594         (DBX_REGISTER_NUMBER): Delete.
595         * config/m68k/netbsd-elf.h (IMMEDIATE_PREFIX): Delete.
596         (PCC_STATIC_STRUCT_RETURN): Don't undefine.
597         * config/m68k/openbsd.h (PCC_STATIC_STRUCT_RETURN): Define.
598
599 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
600
601         * config.gcc (m68k-*-uclinux*): Add flat.h to $tm_file.
602         * config/flat.h: New file.
603         * crtstuff.c (USE_PT_GNU_EH_FRAME): Don't define if
604         OBJECT_FORMAT_FLAT.
605         * config/m68k/m68k.h (ASM_PREFERRED_EH_DATA_FORMAT): Do not use
606         indirect references for -msep-data or -mid-shared-library.
607         Do not use PC-relative code addresses either.
608         * config/m68k/m68k.c (override_options): Restrict -fPIC error
609         to -mpcrel.
610         * config/m68k/uclinux.h (STARTFILE_SPEC): Define.  Use Scrt1.o
611         for shared libraries and crt1.o for executables.  Use crti.o and
612         crtbegin.o.
613         (ENDFILE_SPEC): Use crtend.o and crtn.o.
614         (LIB_SPEC): Suppress -Rlibc.gdb if -static-libc is given.
615         Do not add -elf2flt or -shared-lib-id options here.
616         (LINK_SPEC): Define.  Pass -elf2flt if no -elf2flt option is given.
617         Pass -shared-lib-id if -mid-shared-library, taking the library
618         identifier from -mshared-library-id if given, otherwise
619         defaulting to 0.
620         (EH_FRAME_IN_DATA_SECTION): Do not undefine.
621         (INIT_SECTION_ASM_OP, FINI_SECTION_ASM_OP): Likewise.
622         (TARGET_OS_CPP_BUILTINS): Define __GXX_MERGED_TYPEINFO_NAMES=0
623         and __GXX_TYPEINFO_EQUALITY_INLINE=0 if -mid-shared-library.
624         (DRIVER_SELF_SPECS): Map unadorned PIC options to -msep-data.
625         * config/m68k/t-uclinux (EXTRA_MULTILIB_PARTS): Add crtbegin.o
626         and crtend.o.
627         * config/m68k/lb1sf68.asm (PICCALL): Use an lea and pc-relative
628         jump sequence for ISA A and ISA A+.
629         (PICJUMP): Likewise.
630
631 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
632
633         * config.gcc (m68k-*-uclinux*): Base the port on the common
634         and m68k GNU/Linux files rather than on the generic ELF ones.
635         * config/m68k/uclinux.h (TARGET_VERSION): Override.
636         (TARGET_OS_CPP_BUILTINS): Use LINUX_TARGET_OS_CPP_BUILTINS.
637
638 2007-01-18  Julian Brown  <julian@codesourcery.com>
639             Richard Sandiford  <richard@codesourcery.com>
640
641         * config/m68k/m68k.h (LONG_DOUBLE_TYPE_SIZE): Make 64-bit on ColdFire.
642         (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
643         * config/m68k/netbsd-elf.h (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Undefine
644         before redefining.
645         * config/m68k/uclinux-oldabi.h (LONG_DOUBLE_TYPE_SIZE): Redefine to
646         80 unconditionally.
647         (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
648
649 2007-01-18  Richard Sandiford  <richard@codesourcery.com>
650
651         * doc/install.texi: Document m68k-uclinuxoldabi.
652         * config.gcc (m68k-*-uclinuxoldabi): New configuration.
653         * config/m68k/uclinux-oldabi.h: New file, copied from
654         config/m68k/uclinux.h.
655
656 2007-01-18  Kaz Kojima  <kkojima@rr.iij4u.or.jp>
657
658         * config/m32r/m32r-protos.h (m32r_expand_epilogue): Declare it.
659         * config/m32r/m32r.c (m32r_setup_incoming_varargs): Use gen_frame_mem.
660         (m32r_compute_frame_size): Use unsigned for regno.
661         (m32r_reload_lr): Use gen_frame_mem.
662         (pop): New.
663         (m32r_output_function_epilogue): Don't output the function epilogue
664         textually here.
665         (m32r_expand_epilogue): New.
666         (direct_return): Return false if the function has the interrupt
667         attribute.
668         (m32r_hard_regno_rename_ok): Remove code for the textual epilogue.
669         * config/m32r/m32r.md (epilogue): New expander.
670         (return_lr, return_rte): New insns.
671         (return): Make it expander.
672         (return_normal): New expander.
673
674 2007-01-18  Josh Conner  <jconner@apple.com>
675
676         PR target/30485
677         * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Add
678         support for UNLE, UNLT, UNGE, and UNGT.
679
680 2007-01-18  Jan Hubicka  <jh@suse.cz>
681
682         * tree-vrp.c (finalize_jump_threads): Do not call cleanup_cfg by hand.
683         * tree-tailcall (add_virtual_phis): Likewise.
684         (optimize_tail_call): Return TODOs.
685         (execute_tail_calls): Return TODOs.
686         * tree-ssa-ccp (execute_fold_all_builtins): Do cleanup_cfg via TODO.
687         * tree-cfgcleanup.c (cleanup_tree_cfg_loop): Return if something
688         changed.
689         * tree-ssa-forwprop.c (tree_ssa_forward_propagate_single_use_value):
690         Cleanup cfg using TODO.
691         * tree-flow.h (cleanup_tree_cfg_loop): Update prototype.
692         * passes.c (execute_function_todo): When cleanup did something, remove
693         unused locals.
694         * tree-cfg.c (pass_build_cfg): Add cleanup_cfg TODO.
695         (make_edges): Don't cleanup_cfg.
696
697 2007-01-18  Uros Bizjak  <ubizjak@gmail.com>
698
699         * reg-stack.c (subst_stack_regs_pat) [UNSPEC_SINCOS_COS,
700         UNSPEC_XTRACT_FRACT]: Use generic code for instructions that
701         operate on the top of stack.
702         [UNSPEC_SINCOS_SIN, UNSPEC_XTRACT_EXP, UNSPEC_TAN]: Rewrite
703         register handling of instructions that output to the second
704         stack slot.
705         [UNSPEC_TAN_ONE, UNSPEC_TAN_TAN]: Remove.
706         (move_for_stack_reg): Special-case check for dead destination
707         stack slot for constant load of 1.0 inside UNSPEC_TAN.
708
709         * config/i386/i386.md (UNSPEC_TAN): New constant.
710         (UNSPEC_TAN_ONE, UNSPEC_TAN_TAN): Remove.
711         (fptanxf4_i387, fptan_extend<mode>xf4_i387): New patterns
712         to correctly model move of constant 1.0 to top stack slot.
713         (*tandf3_1, *tansf3_1, *tanxf3_1): Remove insn patterns.
714         (unnamed peephole2 pattern): Remove corresponding peephole2
715         pattern that optimizes tan insn and loading of constant 1.0.
716         (tanxf2): Use fptanxf4_i387.
717         (tan<mode>2): Rename from tansf2 and tandf2 and macroize insn
718         patterns using X87MODEF12 mode macro.  Use fptan_extend<mode>xf4_i387
719         and truncate result to requested mode.  Use SSE_FLOAT_MODE_P to
720         disable patterns for SSE math.
721         (sincos<mode>3): Use truncxf<mode>2_i387_noop for truncation.
722         (fyl2x_extend<mode>xf3_i387): Use X87MODEF12 for operand 1.
723
724 2007-01-18  Dirk Mueller  <dmueller@suse.de>
725             Richard Guenther <rguenther@suse.de>
726
727         PR diagnostic/8268
728         * doc/invoke.texi (Warray-bounds): Document -Warray-bounds.
729         * common.opt (Warray-bounds): Add new warning option.
730         * c-opts.c (c_common_handle_option): Define -Warray-bounds
731         if -Wall is given.
732         * Makefile.in: make tree-vrp.o depend on toplev.h
733         * tree-vrp.c (vrp_finalize): Call check_array_refs if -Warray-bounds
734         is enabled.
735         (check_array_refs, check_array_bounds, check_array_ref): New.
736
737 2007-01-18  Jan Hubicka  <jh@suse.cz>
738
739         * tree-ssa-ccp.c (ccp_finalize): Return if something changed.
740         (execute_ssa_ccp): Return flags conditionally.
741         * tree-ssa-propagate.c (substitue_and_fold): Return if something was
742         changed.
743         * tree-ssa-propagate.h (substitute_and_fold): Update prototype.
744
745 2007-01-18  Steven Bosscher  <steven@gcc.gnu.org>
746
747         * cfgcleanup.c (cleanup_cfg): Detect cfglayout mode and set
748         the CLEANUP_CFGLAYOUT flag when in cfglayout mode.
749
750         * Makefile.c (GTFILES): Add cfglayout.h.
751         * gengtype.c (open_base_files): Likewise.
752         * cfglayout.c (cfg_layout_function_footer,
753         cfg_layout_function_header) Reindent to make gengtype happy.
754         * cfglayout.h (cfg_layout_function_footer,
755         cfg_layout_function_header): Add GTY(()) marker.
756
757         * ifcvt.c (noce_try_sign_mask): Make sure INSN_B is non-null.
758
759 2007-01-18  Ben Elliston  <bje@au.ibm.com>
760
761         * genautomata.c (write_automata): Include xstrerror output in the
762         error message if writing the DFA description file fails.
763
764 2007-01-17  H.J. Lu  <hongjiu.lu@intel.com>
765
766         * config/mips/mips-protos.h (mips_output_external): Make it
767         return void.
768         * config/mips/iris.h (TARGET_ASM_EXTERNAL_LIBCALL): Removed.
769         * config/mips/mips.c (irix_output_external_libcall): Likewise.
770         (extern_list): Likewise.
771         (extern_head): Likewise.
772         (TARGET_ASM_FILE_END): Likewise.
773         (mips_file_end): Likewise.
774         (mips_output_external): Rewritten.
775
776 2007-01-18  Ben Elliston  <bje@au.ibm.com>
777
778         * genpreds.c (write_insn_preds_c): Only write out the function
779         body for regclass_for_constraint if we have register constraints.
780
781 2007-01-17  Tom Tromey  <tromey@redhat.com>
782
783         * doc/sourcebuild.texi (libgcj Tests): Use sourceware.org.
784         * doc/install.texi (Testing): Use sourceware.org.
785         (Binaries): Likewise.
786         (Specific): Likewise.
787         * doc/contrib.texi (Contributors): Use sourceware.org.
788
789 2007-01-17  Anatoly Sokolov <aesok@post.ru>
790
791         * config/avr/avr.h (AVR_HAVE_LPMX): New macro.
792         (AVR_ENHANCED): Rename to ...
793         (AVR_HAVE_MUL): ... new.
794         (avr_enhanced_p): Rename to ...
795         (avr_have_mul_p): ... new.
796         (TARGET_CPU_CPP_BUILTINS): Use 'avr_have_mul_p' instead of 
797         'avr_enhanced_p' for "__AVR_ENHANCED__". Define "__AVR_HAVE_MUL__".
798         * config/avr/avr.c (avr_enhanced_p): Rename to ...
799         (avr_have_mul_p): ... new.
800         (base_arch_s): Rename 'enhanced' to 'have_mul'.
801         (avr_override_options): Use 'avr_have_mul_p' and 'have_mul' instead of
802         'avr_enhanced_p' and 'enhanced'.
803         (ashlhi3_out, ashrhi3_out, lshrhi3_out, avr_rtx_costs): Use 
804         AVR_HAVE_MUL instead of AVR_ENHANCED.
805         * avr.md (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
806         (mulqi3, *mulqi3_enh, *mulqi3_call, mulqihi3, umulqihi3, mulhi3, 
807         *mulhi3_enh, *mulhi3_call, mulsi3, *mulsi3_call): Use AVR_HAVE_MUL 
808         instead of AVR_ENHANCED.
809         (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
810         * libgcc.S: Use __AVR_HAVE_MUL__ instead of __AVR_ENHANCED__.
811         (__tablejump__): Use __AVR_HAVE_LPMX__ instead of __AVR_ENHANCED__.
812
813 2007-01-17  Ian Lance Taylor  <iant@google.com>
814
815         * vec.h (VEC_reserve_exact): Define.
816         (vec_gc_p_reserve_exact): Declare.
817         (vec_gc_o_reserve_exact): Declare.
818         (vec_heap_p_reserve_exact): Declare.
819         (vec_heap_o_reserve_exact): Declare.
820         (VEC_OP (T,A,reserve_exact)): New static inline function, three
821         versions.
822         (VEC_OP (T,A,reserve)) [all versions]: Remove handling of
823         negative parameter.
824         (VEC_OP (T,A,alloc)) [all versions]: Call ...reserve_exact.
825         (VEC_OP (T,A,copy)) [all versions]: Likewise.
826         (VEC_OP (T,a,safe_grow)) [all versions]: Likewise.
827         * vec.c (calculate_allocation): Add exact parameter.  Change all
828         callers.
829         (vec_gc_o_reserve_1): New static function, from vec_gc_o_reserve.
830         (vec_gc_p_reserve, vec_gc_o_reserve): Call vec_gc_o_reserve_1.
831         (vec_gc_p_reserve_exact, vec_gc_o_reserve_exact): New functions.
832         (vec_heap_o_reserve_1): New static function, from vec_heap_o_reserve.
833         (vec_heap_p_reserve, vec_heap_o_reserve): Call vec_heap_o_reserve_1.
834         (vec_heap_p_reserve_exact): New function.
835         (vec_heap_o_reserve_exact): New function.
836
837 2007-01-17  Jan Hubicka  <jh@suse.cz>
838
839         * ipa-type-escape.c (look_for_casts): Revamp using handled_component_p.
840
841 2007-01-17  Eric Christopher  <echristo@apple.com>
842
843         * config.gcc: Support core2 processor.
844
845 2007-01-16  Jan Hubicka  <jh@suse.cz>
846
847         * tree-ssanames.c (release_dead_ssa_names): Instead of ggc_freeing
848         the names, just unlink the chain so we don't crash on dangling pointers
849         to dead SSA names.
850
851 2007-01-16  Jan Hubicka  <jh@suse.cz>
852
853         * cgraph.h (cgraph_decide_inlining_incrementally): Kill.
854         * tree-pass.h: Reorder to make IPA passes appear toegher.
855         (pass_early_inline, pass_inline_parameters, pass_apply_inline): Declare.
856         * cgraphunit.c (cgraph_finalize_function): Do not compute inling
857         parameters, do not call early inliner.
858         * ipa-inline.c: Update comments.  Include tree-flow.h
859         (cgraph_decide_inlining): Do not compute inlining parameters.
860         (cgraph_decide_inlining_incrementally): Return TODOs; assume to
861         be called with function context set up.
862         (pass_ipa_inline): Remove unreachable functions before pass.
863         (cgraph_early_inlining): Simplify assuming to be called from the
864         PM as local pass.
865         (pass_early_inline): New pass.
866         (cgraph_gate_ipa_early_inlining): New gate.
867         (pass_ipa_early_inline): Turn into simple wrapper.
868         (compute_inline_parameters): New function.
869         (gate_inline_passes): New gate.
870         (pass_inline_parameters): New pass.
871         (apply_inline): Move here from tree-optimize.c
872         (pass_apply_inline): New pass.
873         * ipa.c (cgraph_remove_unreachable_nodes): Verify cgraph after
874         transforming.
875         * tree-inline.c (optimize_inline_calls): Return TODOs rather than
876         doing them by hand.
877         (tree_function_versioning): Do not allocate dummy struct function.
878         * tree-inline.h (optimize_inline_calls): Update prototype.
879         * tree-optimize.c (execute_fixup_cfg): Export.
880         (pass_fixup_cfg): Remove
881         (tree_rest_of_compilation): Do not apply inlines.
882         * tree-flow.h (execute_fixup_cfg): Declare.
883         * Makefile.in (gt-passes.c): New.
884         * passes.c: Include gt-passes.h
885         (init_optimization_passes): New passes.
886         (nnodes, order): New static vars.
887         (do_per_function_toporder): New function.
888         (execute_one_pass): Dump current pass here.
889         (execute_ipa_pass_list): Don't dump current pass here.
890
891 2007-01-16  Janis Johnson  <janis187@us.ibm.com>
892
893         * config/dfp-bit.c (dfp_compare_op): Return separate value for NaN.
894         (DFP_NE, DFP_LE, DFP_GE): Return false for NaN.
895
896 2007-01-16  David Edelsohn  <edelsohn@gnu.org>
897
898         * config/rs6000/darwin-ldouble.c: Build file for SOFT_FLOAT.
899         (strong_alias): Define.
900         (__gcc_qmul): Provide non-FMA for soft-float.
901         (__gcc_qdiv): Same.
902         (__gcc_qneg): New.
903         (__gcc_qeq): New.
904         (__gcc_qle): New.
905         (__gcc_qge): New.
906         (__gcc_qunord): New.
907         (__gcc_stoq): New.
908         (__gcc_dtoq): New.
909         (__gcc_qtos): New.
910         (__gcc_qtod): New.
911         (__gcc_qtoi): New.
912         (__gcc_qtou): New.
913         (__gcc_itoq): New.
914         (__gcc_utoq): New.
915         (fmsub): New.
916         * config/rs6000/rs6000.c (rs6000_init_libfuncs): Initialize
917         soft-float functions.
918         * config/rs6000/libgcc-ppc-glibc.ver: Version soft-float symbols.
919         * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Do not warn
920         about long double soft float.
921
922 2007-01-16  Dorit Nuzman  <dorit@il.ibm.com>
923             Tehila Meyzels  <tehila@il.ibm.com>
924
925         * tree-vectorizer.h (is_pattern_stmt_p): New.
926         * tree-vect-analyze.c (vect_determine_vectorization_factor): Fix
927         formatting (tabs instead of spaces). Cleanup and clarify setting
928         of STMT_VINFO_VECTYPE. Call is_pattern_stmt_p.
929         * tree-vect-transform.c (vect_get_vec_def_for_operand): Fix typo.
930         (vectorizable_type_demotion): Check that types are integral.
931         (vectorizable_type_promotion): Likewise.
932         (vectorizable_store): Fix typo.  Eliminate new-line at end of
933         comments.
934
935 2007-01-16  Jan Hubicka  <jh@suse.cz>
936
937         * tree-ssanames.c (release_dead_ssa_names): Remove invalidated
938         cgraph edges too.
939
940 2007-01-15  Eric Christopher  <echristo@apple.com>
941
942         * ifcvt.c: Include vec.h, vecprim.h.
943         (check_cond_move_block): New argument regs.
944         Reorganize. Add registers used to regs.
945         (cond_move_process_if_block): Use regs set above as
946         loop bounds.
947
948 2007-01-15  Eric Christopher  <echristo@apple.com>
949
950         * config/darwin.h: Update copyright.
951         (TARGET_OPTION_TRANSLATE_TABLE): Add umbrella.
952         (LINK_COMMAND_SPEC): Add -u.
953         (LINK_SPEC): Fix umbrella for above.
954
955 2007-01-15  Joseph S. Myers  <joseph@codesourcery.com>
956
957         * config/soft-fp/op-common.h, config/soft-fp/op-4.h: Update from
958         glibc CVS.
959
960 2007-01-15  Tom Tromey  <tromey@redhat.com>
961
962         * doc/sourcebuild.texi (libgcj Tests): Don't mention jacks.
963         * doc/install.texi (Testing): Don't mention jacks.
964         (Configuration): Document --enable-java-maintainer-mode.  Move
965         --with-java-home to libgcj-specific section.  Document
966         --with-ecj-jar.
967         (Prerequisites): Mention --enable-java-maintainer-mode, ecj1.
968
969 2007-01-15  Jan Hubicka  <jh@suse.cz>
970
971         * tree-ssa-dce.c (DCE_TODOs): New.
972         (propagate_necessity): Return if something changed.
973         (eliminate_unnecessary_stmts): Likewise.
974         (perform_tree_ssa_dce): Return TODO flags when needed.
975         (pass_dce, pass_dce_loop, pass_cd_dce): Remove TODO flags.
976
977 2007-01-15  Uros Bizjak  <ubizjak@gmail.com>
978
979         * config/i386/i386.md (fyl2xxf3_i387): Rename from fyl2x_xf3.
980         (fyl2x_extend<mode>xf3_i387): New insn pattern.
981         (log<mode>2): Rename from logsf2 and logdf2 and macroize insn
982         patterns using X87MODEF12 mode macro.  Extend operand 1
983         to XFmode. Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
984         (log10<mode>2): Ditto.
985         (log2<mode>2): Ditto.
986         (log1p<mode>2): Ditto.
987         (logb<mode>2): Ditto.
988         (fyl2xp1xf3_i387): Rename from fyl2xp1_xf3.
989         (fyl2xp1_extend<mode>xf3_i387): New insn pattern.
990         (*fxtractxf3_i387): Rename from *fxtractxf3.
991         (fxtract_extend<mode>xf3_i387): New insn pattern.
992         (ilogbsi2): Use match_dup 3, not match_operand:XF 3.
993         * config/i386/i386.c (ix86_emit_i387_log1p): Use gen_fyl2xp1xf3_i387()
994         and gen_fyl2xxf3_i387().
995
996 2007-01-14  Zdenek Dvorak <dvorakz@suse.cz>
997
998         * loop-unswitch.c (unswitch_loop): Do not call fix_loop_placement.
999         * cfgloopmanip.c (fix_loop_placement): Made static.  Use
1000         get_loop_exit_edges.  Changed return type to bool.
1001         * cfgloop.h (fix_loop_placement): Declaration removed.
1002
1003 2007-01-14  Dorit Nuzman  <dorit@il.ibm.com>
1004
1005         * param.h (MIN_VECT_LOOP_BOUND): New.
1006         * params.def (MIN_VECT_LOOP_BOUND): New.
1007         * tree-vectorizer.c (slpeel_tree_peel_loop_to_edge): Takes another
1008         argument - minimum threshold for number of iterations.
1009         * tree-vectorizer.h (slpeel_tree_peel_loop_to_edge): Add another
1010         argument to declaration.
1011         * tree-vect-analyze.c (vect_analyze_operations): Check value of
1012         MIN_VECT_LOOP_BOUND.
1013         * tree-vect-transform.c (vect_do_peeling_for_loop_bound): Call
1014         slpeel_tree_peel_loop_to_edge with additional argument.
1015         (vect_do_peeling_for_alignment): Likewise.
1016         * doc/invoke.texi (min-vect-loop-bound): Document new param option.
1017
1018 2007-01-14  Uros Bizjak  <ubizjak@gmail.com>
1019
1020         PR target/30413
1021         * config/i386/i386.c (print_operand) ['z']: Output 'b' for
1022         operands of size 1.
1023
1024 2007-01-14  Jan Hubicka  <jh@suse.cz>
1025
1026         * tree-dfa.c (remove_referenced_var): New function.
1027         * tree-ssa-live.c (remove_unused_locals): Walk referenced vars and
1028         prune referenced vars list too.
1029         * tree-flow.h (remove_referenced_var): Declare.
1030
1031 2007-01-14  Jan Hubicka  <jh@suse.cz>
1032
1033         * tree-eh.c (add_stmt_to_eh_region_fn): Do not add call_exprs
1034         separately.
1035         (remove_stmt_from_eh_region_fn): Do not remove call_exprs.
1036         (verify_eh_throw_stmt_node, verify_eh_throw_table_statements): Kill.
1037         * except.h (verify_eh_throw_table_statements): Kill prototype.
1038         * cfgexpand.c (expand_gimple_basic_block): Propagate Eh regions
1039         into call exrepssions.
1040         * tree-optimize.c (execute_free_cfg_annotatiosn): Do not call
1041         eh trhow verifier.
1042         * tree-cfg.c: Include pointer-set.h.
1043         (verify_node_sharing): Work on pointer set.
1044         (verify_eh_throw_stmt_node): New.
1045         (verify_stmts): Use pointers sets, verify throw_stmt.
1046
1047 2007-01-13  Zdenek Dvorak <dvorakz@suse.cz>
1048
1049         * ipa-reference.c (analyze_function): Consider also addresses taken
1050         in phi nodes.
1051
1052 2007-01-12  Roger Sayle  <roger@eyesopen.com>
1053
1054         * c-typeck.c (null_pointer_constant_p): Replace use of
1055         TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
1056         (build_c_cast): Likewise.
1057
1058 2007-01-12  Roger Sayle  <roger@eyesopen.com>
1059
1060         * tree.h (force_fit_type_double): Remove unused final argument.
1061         * c-common.c (constant_expression_warning): Replace use of
1062         TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
1063         (convert_and_check): Likewise.
1064         (shorten_compare): Update call to force_fit_type_double.
1065         (c_common_truthvalue_conversion) <INTEGER_CST>: Use integer_zerop.
1066         * convert.c (convert_to_pointer): Update call to
1067         force_fit_type_double.
1068         * fold-const.c (force_fit_type_double): Remove overflowed_const
1069         argument.
1070         (int_const_binop, fold_convert_const_int_from_int,
1071         fold_convert_const_int_from_real, fold_div_compare,
1072         fold_sign_changed_comparison, fold_unary, fold_negate_const,
1073         fold_abs_const, fold_not_const): Remove the final argument from
1074         calls to force_fit_type_double.
1075
1076 2007-01-12  Andrew Pinski  <andrew_pinski@playstation.sony.com>
1077
1078         * configure.ac: Set insn to "nop" for spu-*-* also.
1079         * configure: Regenerate.
1080
1081 2007-01-12  Olga Golovanevsky  <olga@il.ibm.com>
1082
1083         * builtins.def : Add BUILT_IN_FREE.
1084
1085 2007-01-12  Jan Hubicka  <jh@suse.cz>
1086
1087         PR tree-optimization/30443
1088         * tree-inline.c (tree_function_versioning): Do not optimize when
1089         cloning for inlining.
1090
1091 2007-01-12  Zdenek Dvorak <dvorakz@suse.cz>
1092
1093         * doc/loop.texi: Document recording of loop exits.
1094         * cfgloopmanip.c (loopify, duplicate_loop): Use alloc_loop.
1095         (update_single_exits_after_duplication,
1096         update_single_exit_for_duplicated_loop,
1097         update_single_exit_for_duplicated_loops): Removed.
1098         (duplicate_loop_to_header_edge): Do not call
1099         update_single_exits_after_duplication and
1100         update_single_exit_for_duplicated_loops.
1101         (loop_version): Do not update single_exit information.
1102         (fix_loop_structure): Use record_loop_exits instead of
1103         mark_single_exit_loops.
1104         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update
1105         the lists of loop exits.
1106         * cfghooks.c (redirect_edge_and_branch, redirect_edge_and_branch_force,
1107         split_edge, merge_blocks): Update the lists of loop exits.
1108         * modulo-sched.c (sms_schedule): Pass LOOPS_HAVE_RECORDED_EXITS to
1109         loop_optimizer_init.
1110         * loop-init.c (loop_optimizer_init): Call record_loop_exits instead
1111         of mark_single_exit_loops.
1112         (loop_optimizer_finalize): Call release_recorded_exits.
1113         * tree-ssa-loop.c (tree_loop_optimizer_init): Pass
1114         LOOPS_HAVE_RECORDED_EXITS to loop_optimizer_init.
1115         * tree-vectorizer.c (slpeel_tree_duplicate_loop_to_edge_cfg): Do not
1116         update single exit information.
1117         * lambda-code.c (perfect_nestify): Ditto.
1118         * cfgloop.c (flow_loop_free): Destroy the list of exits of the loop.
1119         (mark_single_exit_loops): Removed.
1120         (alloc_loop, loop_exit_hash, loop_exit_eq, loop_exit_free,
1121         get_exit_descriptions, rescan_loop_exit, record_loop_exits,
1122         dump_recorded_exit, dump_recorded_exits, release_recorded_exits): New
1123         functions.
1124         (get_loop_exit_edges, single_exit): Use recorded exit lists.
1125         (add_bb_to_loop, remove_bb_from_loops): Update the lists of loop exits.
1126         (verify_loop_structure): Verify consistency of the exit lists.
1127         (flow_loops_find): Use alloc_loop.  Initialize exits hash.
1128         (set_single_exit): Removed.
1129         * cfgloop.h (struct loop_exit): New function.
1130         (struct loop): single_exit_ field replaced by exits field.
1131         (LOOPS_HAVE_MARKED_SINGLE_EXITS): Replaced by LOOPS_HAVE_RECORDED_EXITS.
1132         (struct loops): Added exits hash.
1133         (mark_single_exit_loops, set_single_exit): Declaration removed.
1134         (release_recorded_exits, record_loop_exits, rescan_loop_exit): Declare.
1135
1136 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
1137
1138         * doc/invoke.texi: Avoid use of @headitem.
1139
1140 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
1141
1142         * cse.c (cse_insn): Move HAVE_CC0 code after declarations.
1143
1144 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
1145
1146         * doc/install.texi: Fix m68k-*-* anchor and add m68k-*-* to the
1147         list of targets.
1148
1149 2007-01-12  Nathan Sidwell  <nathan@codesourcery.com>
1150             Richard Sandiford  <richard@codesourcery.com>
1151
1152         * doc/invoke.texi: Document -mno-bitfield, -mno-rtd and -mno-short.
1153         * config/m68k/m68k.opt: Resort options.
1154         (mbitfield, mrtd, mshort): Remove RejectNegative properties.
1155
1156 2007-01-12  Nathan Sidwell  <nathan@codesourcery.com>
1157             Richard Sandiford  <richard@codesourcery.com>
1158
1159         * doc/invoke.texi: Document the macros that are defined by
1160         m68k's -mtune and -mhard-float options.
1161         * config/m68k/m68k-protos.h (m68k_cpp_cpu_ident) Declare.
1162         (m68k_cpp_cpu_family): Likewise.
1163         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Add a full set
1164         of __ucfv*__ macros.  Define __mcffpu__ if generating code for
1165         ColdFire FPUs.  Define __mcf_cpu_* and __mcf_family_* macros.
1166         * config/m68k/m68k.c (m68k_cpp_cpu_ident): New function.
1167         (m68k_cpp_cpu_family): Likewise.
1168
1169 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
1170
1171         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Treat all mc68*
1172         macros besides mc68000 as tuning macros.  Use a switch statement
1173         to set them and mcpu32.
1174
1175 2007-01-12  Julian Brown  <julian@codesourcery.com>
1176
1177         * config/m68k/m68k.h: Use TARGET_68040 instead of TARGET_68040_ONLY.
1178         (TARGET_68040_ONLY): Rename to...
1179         (TARGET_68040): ...this.
1180         * config/m68k/m68k.c: Use TARGET_68040 instead of TARGET_68040_ONLY.
1181         * config/m68k/m68k.md: Likewise.
1182
1183 2007-01-12  Julian Brown  <julian@codesourcery.com>
1184             Nathan Sidwell  <nathan@codesourcery.com>
1185             Richard Sandiford  <richard@codesourcery.com>
1186
1187         * config.gcc (m680[012]0-*-*, m68k*-*-*): Set m68k_cpu_ident to
1188         the -mcpu= argument associated with the --with-cpu setting.
1189         Define M68K_DEFAULT_TUNE to the default -mtune= option,
1190         if different from the one implied by the -mcpu setting.
1191         Accept --with-cpu=FOO if FOO is listed in m68k-devices.def,
1192         using mcpu=FOO as the default CPU option.  Set target_cpu_default2.
1193         * doc/invoke.texi: Mention ColdFire in the introduction to the
1194         m68k options.  Document the new -march, -mcpu, -mtune, -mdiv,
1195         -mno-div and -mhard-float options.  Make -m68881 a synonym for
1196         -mhard-float.  Document the previously-undocumented -m5206e,
1197         -m528x, -m5307 and -m5407 options.  Tweak the existing option
1198         documentation for consistency.
1199         * doc/install.texi: Mention new --with-cpu arguments.
1200         * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Only use the
1201         default CPU if neither -mcpu nor -march are specified.
1202         (ASM_CPU_SPEC): Pass down -mcpu and -march options.
1203         (TARGET_CPU_CPP_BUILTINS): Set __mcfisa*__ macros from
1204         TARGET_ISA*.  Set the legacy __mcf*__ cpu macros in the same way,
1205         using m68k_tune to decide between families that implement the
1206         same ISA.  Use m68k_tune to set __mcfv4e__.
1207         (FL_BITFIELD, FL_68881, FL_COLDFIRE, FL_CF_HWDIV, FL_CF_MAC)
1208         (FL_CF_EMAC, FL_CF_EMAC_B, FL_CF_USP, FL_CF_FPU, FL_ISA_68000)
1209         (FL_ISA_68010, FL_ISA_68020, FL_ISA_68040, FL_ISA_A, FL_ISA_B)
1210         (FL_ISA_C, FL_ISA_MMU): New macros.
1211         (MASK_COLDFIRE): Delete.
1212         (TARGET_68010, TARGET_68020, TARGET_68040_ONLY, TARGET_COLDFIRE)
1213         (TARGET_ISAB): Redefine in terms of m68k_cpu_flags.
1214         (TARGET_68881, TARGET_COLDFIRE_FPU): Redefine in terms of m68k_fpu.
1215         (TARGET_HARD_FLOAT): Do not define here.
1216         (TARGET_ISAAPLUS, TARGET_ISAC): New macros.
1217         (TUNE_68000): New macro.
1218         (TUNE_68000_10): Redefine in terms of TUNE_68000 and TUNE_68010.
1219         (TUNE_68010, TUNE_68030, TUNE_68040, TUNE_68060, TUNE_CPU32)
1220         (TUNE_CFV2): Redefine in terms of m68k_tune.
1221         (uarch_type, target_device, fpu_type): New enums.
1222         (m68k_cpu, m68k_tune, m68k_fpu, m68k_cpu_flags): Declare.
1223         * config/m68k/m68k.c (TARGET_DEFAULT): Remove MASK_68881.
1224         (FL_FOR_isa_00, FL_FOR_isa_10, FL_FOR_isa_20, FL_FOR_isa_40)
1225         (FL_FOR_isa_cpu32, FL_FOR_isa_a, FL_FOR_isa_aplus, FL_FOR_isa_b)
1226         (FL_FOR_isa_c): New macros.
1227         (m68k_isa): New enum.
1228         (m68k_target_selection): New structure.
1229         (all_devices, all_isas, all_microarchs): New tables.
1230         (m68k_cpu_entry, m68k_arch_entry, m68k_tune_entry, m68k_cpu)
1231         (m68k_tune, m68k_fpu, m68k_cpu_flags): New variables.
1232         (MASK_ALL_CPU_BITS): Delete.
1233         (m68k_find_selection): New function.
1234         (m68k_handle_option): Handle -mcpu=, -march= and -mtune=.
1235         Map the legacy target options to a combination of the new ones.
1236         (override_options): Set m68k_cpu, m68k_tune, m68k_fpu and
1237         m68k_cpu_flags.  Handle M68K_DEFAULT_TUNE.  Use m68k_cpu_flags
1238         to derive default MASK_BITFIELD, MASK_CF_HWDIV and MASK_HARD_FLOAT
1239         settings.
1240         * config/m68k/m68k.opt (m5200, m5206e, m528x, m5307, m5407, mcfv4e)
1241         (m68010, m68020, m68020-40, m68020-60, m68030, m68040): Remove Mask
1242         properties.
1243         (m68881, msoft-float): Change mask from 68881 to HARD_FLOAT.
1244         (march=, mcpu=, mdiv, mhard-float, mtune=): New options.
1245         * config/m68k/m68k-devices.def: New file.
1246
1247 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
1248             Nathan Sidwell  <nathan@codesourcery.com>
1249
1250         * config/m68k/m68k.h (ASM_CPU_SPEC, ASM_SPEC, EXTRA_SPECS)
1251         (SUBTARGET_EXTRA_SPECS): New macros.
1252         * config/m68k/linux.h (ASM_SPEC): Remove CPU flags;
1253         use %(asm_cpu_spec) instead.
1254         * config/m68k/m68k-none.h (ASM_SPEC): Likewise.
1255         * config/m68k/openbsd.h (ASM_SPEC): Likewise.
1256         * config/m68k/netbsd-elf.h (ASM_SPEC): Likewise.
1257         (EXTRA_SPECS): Rename to...
1258         (SUBTARGET_EXTRA_SPECS): ...this.
1259
1260 2007-01-12  Nathan Sidwell  <nathan@codesourcery.com>
1261             Richard Sandiford  <richard@codesourcery.com>
1262             Julian Brown  <julian@codesourcery.com>
1263
1264         * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
1265         (m68k-*-uclinux*, m68k-*-linux*, m68k-*-rtems*): Set default_m68k_cpu
1266         to the configuration's default CPU.
1267         (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*): Likewise.
1268         Remove default masks.
1269         (m680[012]0-*-*): Set the default with_cpu to the first part of
1270         the target name.
1271         (m68k*-*-*): Set the default with_cpu to m$default_m68k_cpu.
1272         (m68k*-*-linux): Extend the --with-cpu handling to...
1273         (m680[012]0-*-*, m68k*-*-*): ...these configurations.  Allow m68000
1274         and m68010.  Don't set target_cpu_default2.
1275         * doc/install.texi: Document --with-cpu for m68k.
1276         * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Define.
1277         * config/m68k/m68k-none.h (TARGET_CPU_DEFAULT, M68K_CPU_m68k)
1278         (M68K_CPU_m68000, M68K_CPU_m68010, M68K_CPU_m68020, M68K_CPU_m68030)
1279         (M68K_CPU_m68040, M68K_CPU_m68302, M68K_CPU_m68332, TARGET_DEFAULT)
1280         (ASM_CPU_DEFAULT_SPEC, CC1_CPU_DEFAULT_SPEC): Delete.
1281         (ASM_SPEC): Remove use of %(asm_cpu_default).
1282         (EXTRA_SPECS, SUBTARGET_EXTRA_SPECS, MULTILIB_DEFAULTS): Delete.
1283         * config/m68k/linux.h (TARGET_DEFAULT): Delete.
1284         (CPP_SPEC): Merge definitions.  Do not handle __HAVE_68881__ here.
1285         * config/m68k/netbsd-elf.h (TARGET_OS_CPP_BUILTINS): Define
1286         __HAVE_FPU__ if TARGET_HARD_FLOAT.
1287         (TARGET_DEFAULT): Delete.
1288         (EXTRA_SPECS): Delete cpp_cpu_default_spec, cpp_cpu_spec,
1289         cpp_fpu_spec, asm_default_spec and netbsd_cpp_spec.
1290         (CPP_CPU_SPEC): Delete.
1291         (TARGET_VERSION): Merge definitions, using TARGET_68010 to pick
1292         the appropriate string.
1293         (CPP_CPU_DEFAULT_SPEC, ASM_DEFAULT_SPEC, CPP_FPU_SPEC): Delete.
1294         (CPP_SPEC): Define to NETBSD_CPP_SPEC.
1295         (ASM_SPEC): Don't use %(asm_default_spec).
1296         * config/m68k/m68k.c (TARGET_DEFAULT_TARGET_FLAGS): Remove
1297         TARGET_DEFAULT and add MASK_68881.
1298         * config/m68k/m68k.md: Remove mention of TARGET_DEFAULT from comments.
1299
1300 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
1301
1302         * config.gcc (m68010-*-netbsdelf*): Add MASK_68010.
1303         (m68k*-*-netbsdelf*, m68k*-*-openbsd*, m68k*-linux*): Add
1304         MASK_68010 alongside MASK_68020.
1305         * doc/invoke.texi: Document -m68010.
1306         * config/m68k/m68k.opt (m68010): New.
1307         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Define mc68010
1308         if TUNE_68010.
1309         (TUNE_68010): New macro.
1310         * config/m68k/m68k-none.h (M68K_CPU_m68k, M68K_CPU_m68010)
1311         (M68K_CPU_m68020, M68K_CPU_m68030, M68K_CPU_m68040)
1312         (M68K_CPU_m68332): Add MASK_68010.
1313         * config/m68k/linux.h (TARGET_DEFAULT): Add MASK_68010 to
1314         fallback definition.
1315         * config/m68k/netbsd-elf.h (CPP_CPU_SPEC): Remove now-redundant
1316         defines.
1317         * config/m68k/m68k.c (MASK_ALL_CPU_BITS): Add MASK_68010.
1318         (m68k_handle_option): Handle OPT_m68010.  Add MASK_68010
1319         to all entries that use MASK_68020.
1320         (output_move_simode_const, output_move_himode, output_move_qimode)
1321         (output_move_stricthi, output_move_strictqi): Use TARGET_68010
1322         instead of TARGET_68020 to select clr behavior.  Remove comment
1323         about there being no TARGET_68010.
1324         * config/m68k/m68k.md: Likewise throughout.
1325
1326 2007-01-12  Julian Brown  <julian@codesourcery.com>
1327
1328         * config/m68k/m68k.h (TARGET_ISAB): New macro.
1329         * config/m68k/m68k.c: Use TARGET_ISAB rather than TARGET_CFV4.
1330         * config/m68k/m68k.md: Likewise.
1331
1332 2007-01-12  Julian Brown  <julian@codesourcery.com>
1333
1334         * config/m68k/m68k.h (LEGITIMATE_INDEX_P, LEGITIMIZE_ADDRESS): Use
1335         TARGET_COLDFIRE_FPU instead of TARGET_CFV4E.
1336
1337 2007-01-12  Julian Brown  <julian@codesourcery.com>
1338
1339         * config/m68k/m68k.h (TUNE_68040_60): New macro.
1340         * config/m68k/m68k.c (standard_68881_constant_p): Use it.
1341         * config/m68k/m68k.md: Likewise.
1342
1343 2007-01-12  Julian Brown  <julian@codesourcery.com>
1344             Richard Sandiford  <richard@codesourcery.com>
1345
1346         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Use TUNE_68030
1347         instead of TARGET_68030, TUNE_68040 instead of TARGET_68040,
1348         TUNE_68060 instead of TARGET_68060 and TUNE_CPU32 instead of
1349         TARGET_CPU32.
1350         (TARGET_CPU32): Rename to...
1351         (TUNE_CPU32): ...this.
1352         (TUNE_68000_10, TUNE_68030, TUNE_68040, TUNE_68060)
1353         (TUNE_CFV2): New macros.
1354         * config/m68k/netbsd-elf.h (LONG_DOUBLE_TYPE_SIZE): Simplify;
1355         remove conditions that are implied by TARGET_68020.
1356         * config/m68k/m68k.c (m68k_output_function_prologue): Use TUNE_68040
1357         instead of TARGET_68040 and TUNE_CPU32 instead of TARGET_CPU32.
1358         (m68k_output_function_epilogue): Likewise.
1359         (m68k_rtx_costs): Likewise.  Use TUNE_68060 instead of TARGET_68060
1360         and TUNE_CFV2 instead of TARGET_5200.  Use TUNE_68000_10 instead of
1361         "!TARGET_68020 && !TARGET_COLDFIRE" to choose between 68000 and
1362         non-68000 timings.  Refactor multiplication and division costs.
1363         (output_addsi3): Use TUNE_68040 instead of TARGET_68040 and
1364         TUNE_CPU32 instead of TARGET_CPU32.
1365         (standard_68881_constant_p): Use TUNE_68040 instead of TARGET_68040
1366         and TUNE_68060 instead of TARGET_68060.
1367         * config/m68k/m68k.md: Use TUNE_68040 instead of TARGET_68040,
1368         TUNE_68060 instead of TARGET_68060, and TUNE_CPU32 instead of
1369         TARGET_CPU32.
1370         (movsi_const0): Use TUNE_68000_10 rather than "!TARGET_68020
1371         && !TARGET_COLDFIRE" to choose between moveq and clr.
1372         Likewise in the unnamed movsf pattern.
1373         (ashlsi_17_24, lshrsi_17_24): Guard with TUNE_68000_10 rather than
1374         "!TARGET_68020 && !TARGET_COLDFIRE".  Likewise the unnamed
1375         ashiftrt pattern.
1376
1377 2007-01-12  Richard Sandiford  <richard@codesourcery.com>
1378
1379         * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Increase amount
1380         of tabbing before backslashes.
1381
1382 2007-01-11  John David Anglin  <dave.anglin@nrc-cnrc.gc.ca>
1383
1384         * pa-linux.h (ASM_OUTPUT_INTERNAL_LABEL): Undefine.
1385         * pa.h (ASM_OUTPUT_LABEL): Output colon when using GAS.
1386         (ASM_OUTPUT_INTERNAL_LABEL): Define.
1387
1388 2007-01-11  Zdenek Dvorak <dvorakz@suse.cz>
1389
1390         * tree-ssa-loop-ivopts.c (extract_cond_operands): Split from
1391         find_interesting_uses_cond.
1392         (find_interesting_uses_cond): Use extract_cond_operands.
1393         (rewrite_use_compare): Use extract_cond_operands and
1394         force_gimple_operand_bsi.  Do not call update_stmt.
1395         (determine_use_iv_cost_condition): Use extract_cond_operands.
1396         Return cheaper of using original bound and changing the exit bound.
1397
1398 2007-01-11  Zdenek Dvorak <dvorakz@suse.cz>
1399
1400         PR tree-optimization/29516
1401         * tree-ssa-address.c (tree_mem_ref_addr, add_to_parts,
1402         most_expensive_mult_to_index, addr_to_parts,
1403         create_mem_ref, maybe_fold_tmr): Make the type of
1404         fields of TARGET_MEM_REF sizetype.
1405         (move_fixed_address_to_symbol, move_pointer_to_base):
1406         New functions.
1407         * tree.def (TARGET_MEM_REF): Add comment on types of
1408         the operands.
1409
1410 2007-01-11  Joseph Myers  <joseph@codesourcery.com>
1411
1412         * c-common.c (vector_types_convertible_p): Treat opaque types as
1413         always convertible if they have the same size, but not otherwise.
1414
1415 2007-01-11  Steven Bosscher  <steven@gcc.gnu.org>
1416
1417         * ifcvt.c (struct noce_if_info): Add comments to the fields.
1418         Remove the b_unconditional field.
1419         (noce_try_sign_mask): Do not look at b_unconditional.
1420         (noce_process_if_block): Do not use merge_if_blocks.  Update
1421         the CFG here.  Do not set b_unconditional.
1422         (cond_move_process_if_block): Likewise.
1423         (find_cond_trap): Likewise.
1424         (check_cond_move_block): Require simple jump insns at the end
1425         of the basic block.
1426
1427 2007-01-11  Jan Hubicka  <jh@suse.cz>
1428
1429         PR tree-optimization/1046
1430         * tree-tailcall.c (suitable_for_tail_call_opt_p): Use TREE_ADDRESSABLE
1431         when alias info is not ready.
1432         (pass_tail_recursion): Do not require aliasing.
1433         * tree-ssa-copyrename.c (pass_rename_ssa_cop): Likewise.
1434         * tree-ssa-ccp.c (pass_ccp, pass_fold_builtins): Likewise.
1435         * tree-ssa-copy.c (pass_copy_prop): Likewise.
1436         * tree-ssa-forwprop.c (pass_forwprop): Likewise.
1437         * tree-ssa-dce.c (pass_dce, pass_dce_loop, pass_cd_dce): Likewise.
1438         * passes.c (init_optimization_passes): Execute rename_ssa_copies,
1439         ccp, forwprop, copy_prop, merge_phi, copy_prop, dce and tail recursion
1440         before inlining.
1441         * tree-ssa-operands.c (add_virtual_operand, get_indirect_ref_operand):
1442         When aliasing is not build, mark statement as volatile.
1443
1444 2007-01-11  Tom Tromey  <tromey@redhat.com>
1445
1446         PR preprocessor/15185, PR preprocessor/20989:
1447         * doc/cppopts.texi <-MT>: Update description of algorithm for
1448         computing default target.
1449         <-M, -MD>: Reword "basename" text.
1450
1451 2007-01-11  Roger Sayle  <roger@eyesopen.com>
1452
1453         * builtins.c (expand_builtin_pow, expand_builtin_powi,
1454         fold_builtin_cabs, fold_builtin_sqrt, fold_builtin_trunc,
1455         fold_builtin_floor, fold_builtin_ceil, fold_builtin_round,
1456         fold_builtin_int_int_roundingfn, fold_builtin_bitop,
1457         fold_builtin_bswap, real_constp, fold_builtin_pow,
1458         fold_builtin_powi, fold_builtin_signbit, fold_builtin_copysign,
1459         do_mpfr_arg1, do_mpfr_arg2, do_mpfr_arg3, do_mpfr_sincos): Replace
1460         uses of the macro TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
1461         * convert.c (convert_to_pointer): Likewise.
1462         * expr.c (highest_pow2_factor, expand_expr_real_1): Likewise.
1463         * fold-const.c (force_fit_type, fold_negate_expr, int_const_binop,
1464         const_binop, fold_convert_const_int_from_int,
1465         fold_convert_const_int_from_real,
1466         fold_convert_const_real_from_real, sign_bit_p,
1467         optimize_minmax_comparison, extract_muldiv_1, fold_div_compare,
1468         fold_sign_changed_comparison, fold_unary, fold_comparison,
1469         fold_binary, multiple_of_p, tree_Expr_non_zero_p,
1470         fold_negate_const, fold_abs_const, fold_not_const): Likewise.
1471         * print-tree.c (print_node_brief, print_node): Likewise.
1472         * stor-layout.c (place_field, layout_type): Likewise.
1473         * tree-chrec.c (keep_cast): Likewise.
1474         * tree.c (build_vector, build_real, build_real_from_int_cst,
1475         build_complex): Likewise.
1476
1477 2007-01-11  Roger Sayle  <roger@eyesopen.com>
1478
1479         * tree.h (TREE_CONSTANT_OVERFLOW): Obsolete.  For the time being,
1480         treat TREE_CONSTANT_OVERFLOW as a synonym of TREE_OVERFLOW.
1481
1482 2007-01-11  Paolo Bonzini  <bonzini@gnu.org>
1483
1484         * configure.ac (strict1_warn): Rename to strict_warn.
1485         (WERROR, --enable-werror, symlink hacks, stage1_cflags,
1486         cc_set_by_configure, quoted_cc_set_by_configure,
1487         stage_prefix_set_by_configure, quoted_stage_prefix_set_by_configure,
1488         all_boot_languages, all_stagestuff): Remove.
1489         (target_list): Remove bootstrap targets.
1490         * Makefile.in (quickstrap): Unconditionally make a synonym of all.
1491         (BOOT_LANGUAGES, STAGE1_CFLAGS, STAGE1_CHECKING,
1492         REMAKEFLAGS, FLAGS_TO_PASS, PREPEND_DOTDOT_TO_RELATIVE_PATHS,
1493         SUBDIR_FLAGS_TO_PASS, WERROR_FLAGS, STRICT2_WARN, LANG_STAGESTUFF,
1494         VOL_FILES, POSTSTAGE1_FLAGS_TO_PASS, STAGE2_FLAGS_TO_PASS,
1495         STAGEPROFILE_FLAGS_TO_PASS, STAGEFEEDBACK_FLAGS_TO_PASS, stage1_build,
1496         stage1_copy, stage2_build, stage2_copy, stageprofile_build,
1497         stageprofile_copy, stage3_build, stage3_copy, stagefeedback_build,
1498         stagefeedback_copy, stage4_build, clean_s1, clean_sw, bootstrap,
1499         bootstrap-lean, bootstrap2, bootstrap2-lean, bootstrap3,
1500         bootstrap3-lean, bootstrap4, bootstrap4-lean, unstage1, unstage2,
1501         unstage3, unstage4, unstageprofile, unstagefeedback, restage, restage2,
1502         restage3, restage4, restageprofile, restagefeedback, bubbleestrap,
1503         cleanstrap, unstrap, restrap, *compare, *compare3, *compare4,
1504         *compare-lean, *compare3-lean, *compare4-lean, stage1-start, stage1,
1505         stage2-start, stage2, stage3-start, stage3, stage4-start, stage4,
1506         stageprofile-start, stageprofile, stagefeedback-start, stagefeedback,
1507         risky-stage1, risky-stage2, risky-stage3, risky-stage4): Remove.
1508         (ORDINARY_FLAGS_TO_PASS): Rename to FLAGS_TO_PASS.
1509         (STAGECOPYSTUFF, STAGEMOVESTUFF): Consolidate into MOSTLYCLEANFILES.
1510         (mostlyclean): Adjust.
1511         (clean, distclean): Don't mention bootstrap stuff.
1512         * configure: Regenerate.
1513         * ada/config-lang.in, cp/config-lang.in, forttran/config-lang.in,
1514         java/config-lang.in, objc/config-lang.in, objcp/config-lang.in,
1515         treelang/config-lang.in (stagestuff): Remove.
1516         * doc/sourcebuild.texi (stage1, stage2, stage3, stage4,
1517         stageprofile, stagefeedback, stagestuff): Remove mention.
1518
1519 2007-01-11  Nick Clifton  <nickc@redhat.com>
1520
1521         * config/mcore/predicates.md (mcore_general_movesrc_operand):
1522         Accept CONSTs.
1523         (mcore_general_movdst_operand): Do not accept CONST_INTs.
1524         (mcore_arith_K_S_operand): Run the test for the S constraint not
1525         the test for the M constraint.
1526         (mcore_addsub_operand): Do not accept integer values that are
1527         larger than 32 bits.
1528         * config/mcore/mcore.md: Remove unused constraints from split.
1529         (andsi3): Use HOST_WIDE_INT instead of int to hold an INTVAL.
1530         (addsi3): Likewise.
1531         (allocate_stack): Likewise.
1532         * config/mcore/mcore.c (mcore_print_operand): Restrict output of P
1533         operands to 32 bits.
1534         (mcore_const_costs): Use HOST_WIDE_INT instead of int to hold an
1535         INTVAL.
1536         (mcore_and_cost, mcore_modify_comparison, const_ok_for_mcore,
1537         mcore_const_ok_for_inline, mcore_const_trick_uses_not,
1538         try_constant_tricks, mcore_num_ones, mcore_num_zeros,
1539         mcore_output_bclri, mcore_output_andn, output_inline_const,
1540         mcore_output_move, mcore_output_movedouble): Likewise.
1541         (mcore_output_cmov): Use CONST_OK_FOR_M and CONST_OK_FOR_N.
1542         (output_inline_const): Likewise.
1543         (output_inline_const): Fix format strings used in sprintf
1544         statements.
1545         * config/mcore/mcore-protos.h: Update prototypes for changed
1546         functions in mcore.c.
1547         * config/mcore/mcore.h (CONST_OK_FOR_I): Cast values to
1548         HOST_WIDE_INT and not int.
1549         (CONST_OK_FOR_J, CONST_OK_FOR_K, CONST_OK_FOR_L, CONST_OK_FOR_M,
1550         CONST_OK_FOR_N): Likewise.
1551         (LEGITIMATE_CONSTANT_P): Also check CONSTANT_P.
1552         (GO_IF_LEGITIMATE_INDEX): Use HOST_WIDE_INT instead of int to hold
1553         an INTVAL.
1554
1555 2007-01-10  Jan Hubicka  <jh@suse.cz>
1556
1557         * tree-vrp.c (remove_range_assertions): Release defs.
1558         * tree-ssa-loop-ivopts.c (rmeove_statement): Likewise.
1559         * tree-ssa-dom.c (remove_stmt_or_phi): Likewise.
1560
1561 2007-01-10  Paul Brook  <paul@codesourcery.com>
1562
1563         * config/arm/arm.c (arm_rtx_costs_1): Handle mutiply-subtract.
1564         * config/arm/arm.md (mulsi3subsi): New insn.
1565
1566 2007-01-10  Zdenek Dvorak <dvorakz@suse.cz>
1567
1568         * tree-ssa-loop-manip.c (tree_unroll_loop): Make it a wrapper over ...
1569         (tree_transform_and_unroll_loop): New.
1570         * tree-flow.h (transform_callback, tree_transform_and_unroll_loop):
1571         Declare.
1572
1573 2007-01-10  Robert Kennedy <jimbob@google.com>
1574
1575         * fold-const.c (fold_comparison): Fold comparisons like (x *
1576         1000 < 0) to (x < 0).
1577
1578 2007-01-10  Ian Lance Taylor  <iant@google.com>
1579
1580         * tree-pretty-print.c (dump_generic_node): Print parentheses when
1581         operands have the same priority.
1582
1583 2007-01-10  Tom Tromey  <tromey@redhat.com>
1584
1585         * fold-const.c (fold_truthop): Don't check can_use_bit_fields_p.
1586         (fold_binary): Likewise.
1587         * langhooks.c (lhd_can_use_bit_fields_p): Removed.
1588         * langhooks-def.h (lhd_can_use_bit_fields_p): Removed.
1589         (LANG_HOOKS_CAN_USE_BIT_FIELDS_P): Removed.
1590         (LANG_HOOKS_INITIALIZER): Remove LANG_HOOKS_CAN_USE_BIT_FIELDS_P.
1591         * langhooks.h (struct lang_hooks): Removed field
1592         'can_use_bit_fields_p'.
1593
1594 2007-01-10  Ralf Cors├ępius <ralf.corsepius@rtems.org>
1595
1596         * config/bfin/t-bfin, config/bfin/t-bfin-elf: Remove GCC_CFLAGS.
1597
1598 2007-01-10  Razya Ladelsky  <razya@il.ibm.com>
1599
1600         * function.c (get_last_funcdef_no): New function.
1601         * function.h (get_last_funcdef_no): Declare.
1602         * tree-inline.c (initialize_cfun): Add initialization.
1603         (tree_function_versioning): Cleanup.
1604
1605 2007-01-10  Jan Hubicka  <jh@suse.cz>
1606
1607         * tree-inline.c (setup_one_parameter): Do not propagate into abnormal
1608         PHIs.
1609
1610 2007-01-10  Sa Liu  <saliu@de.ibm.com>
1611             Ben Elliston  <bje@au.ibm.com>
1612
1613         * spu.h (STACK_SAVE_AREA): Use VOIDmode for SAVE_FUNCTION, SImode
1614         for SAVE_NONLOCAL and Pmode for any other save level.
1615         * spu-protos.h (spu_restore_stack_block): Declare.
1616         * spu.md (save_stack_block): Remove.
1617         (restore_stack_block): Call spu_restore_stack_block.
1618         * spu.c (spu_restore_stack_block): New function.
1619         (spu_expand_epilogue): Remove old comment.
1620
1621 2007-01-09  Zdenek Dvorak <dvorakz@suse.cz>
1622
1623         PR tree-optimization/30322
1624         * tree-ssa-loop-ivopts.c (fold_affine_expr, iv_value): Removed.
1625         (cand_value_at): Return the value as aff_tree.
1626         (may_eliminate_iv): Convert the bound from aff_tree to tree.
1627         * tree-affine.c (aff_combination_add_cst, aff_combination_add_product,
1628         aff_combination_mult): New functions.
1629         (aff_combination_add): Use aff_combination_add_cst.
1630         (aff_combination_convert): Allow conversions to a wider type.
1631         (tree_to_aff_combination): Handle BIT_NOT_EXPR.
1632         * tree-affine.h (aff_combination_mult): Declare.
1633
1634 2007-01-09  Carlos O'Donell  <carlos@codesourcery.com>
1635
1636         * doc/tm.texi: Update documentation to reflect reality of exec
1637         and start file search behaviours. Update copyright year.
1638         * doc/invoke.texi: Explain how GCC_EXEC_PREFIX is used to find
1639         header file directories.
1640
1641 2007-01-09  Uros Bizjak  <ubizjak@gmail.com>
1642
1643         * config/i386/i386.md (*sinxf2): Rename to *sinxf2_i387.
1644         (*cosxf2): Rename to cosxf2_i387.
1645         (*sindf2, *sinsf2): Extend operand 1 to XFmode.  Macroize patterns
1646         using X87MODEF12 mode macro. Rename patterns to
1647         *sin_extend<mode>xf2_i387.  Use SSE_FLOAT_MODE_P to disable patterns
1648         for SSE math.
1649         (*cosdf2, *cossf2): Ditto.
1650         (sincosdf3, sincossf3): Ditto.  Rewrite corresponding splitters
1651         to match extended input operands.
1652         (sincos<mode>3): New expander.
1653         (*sinextendsfdf2, *cosextendsfdf2, *sincosextendsfdf3): Remove
1654         insn patterns and corresponding splitters.
1655
1656 2007-01-09  Kaz Kojima  <kkojima@gcc.gnu.org>
1657
1658         * config/sh/t-linux (TARGET_LIBGCC2_CFLAGS): Delete.
1659         (SHLIB_MAPFILES, SHLIB_LINK, SHLIB_INSTALL): Likewise.
1660
1661 2007-01-09  Nicolas Pitre  <nico@cam.org>
1662
1663         PR target/30173
1664         * arm/ieee754-df.S (Lad_s): Also test the low word of X for zero.
1665
1666 2007-01-08  Geoffrey Keating  <geoffk@apple.com>
1667
1668         * target.h (struct gcc_target): New field library_rtti_comdat.
1669         * target-def.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): New.
1670         (TARGET_CXX): Add TARGET_CXX_LIBRARY_RTTI_COMDAT.
1671         * doc/tm.texi (C++ ABI): Document TARGET_CXX_LIBRARY_RTTI_COMDAT.
1672         * config/darwin.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
1673
1674 2007-01-08  Geoffrey Keating  <geoffk@apple.com>
1675
1676         * doc/invoke.texi (Optimize Options): Correct description of -O0.
1677
1678 2007-01-08  Richard Guenther  <rguenther@suse.de>
1679
1680         * tree.h (force_fit_type_double): Export.
1681         (force_fit_type): Remove.
1682         * fold-const.c (force_fit_type_double): New function.
1683         (force_fit_type): Remove.
1684         (int_const_binop): Use it.
1685         (fold_convert_const_int_from_int): Likewise.
1686         (fold_convert_const_int_from_real): Likewise.
1687         (fold_div_compare): Likewise.
1688         (fold_sign_changed_comparison): Likewise.
1689         (fold_unary): Likewise.
1690         (fold_negate_const): Likewise.
1691         (fold_abs_const): Likewise.
1692         (fold_not_const): Likewise.
1693         * c-common.c (shorten_compare): Use force_fit_type_double.
1694         * convert.c (convert_to_pointer): Likewise.
1695
1696 2007-01-08  Richard Guenther  <rguenther@suse.de>
1697
1698         * tree.h (build_int_cst_wide_type): Export.
1699         * tree.c (build_int_cst_wide_type): New function.
1700         (build_int_cst_wide): Fix comment.
1701         * builtins.c (fold_builtin_object_size): Use build_int_cst
1702         to build -1 or 0 of the correct type.  Use fit_double_type
1703         to check for overflow.
1704         * fold-const.c (optimize_bit_field_compare): Use build_int_cst_type
1705         to build the mask.
1706         (decode_field_reference): Likewise.
1707         (all_ones_mask_p): Likewise.
1708         (native_interpret_int): Use build_int_cst_wide_type.
1709         (fold_binary): Use build_int_cst_type to build an all-ones
1710         value.
1711         * stor-layout.c (set_sizetype): Use build_int_cst_wide_type.
1712
1713 2007-01-08  Daniel Jacobowitz  <dan@codesourcery.com>
1714
1715         * config/pa/t-pa64 (libgcc_stub.a): Use $(T).
1716
1717 2007-01-09  Ben Elliston  <bje@au.ibm.com>
1718
1719         * genautomata.c (STATS_OPTION): New option.
1720         (stats_flag): New flag.
1721         (gen_automata_option): Handle it.
1722         (initiate_automaton_gen): Ditto.
1723         (write_automata): Output statistics only if stats_flag is
1724         set. Likewise, output time statistics only if time_flag is set.
1725         * doc/md.texi (Processor pipeline description): Document new flag.
1726
1727 2007-01-08  Richard Guenther  <rguenther@suse.de>
1728
1729         * builtins.c (fold_builtin_int_roundingfn): Use fit_double_type.
1730         * tree.c (build_int_cst_type): Likewise.
1731         (size_in_bytes): Don't call force_fit_type on the result.
1732         (int_fits_type_p): Use fit_double_type.
1733         * fold-const.c (fit_double_type): New function.
1734         (force_fit_type): Use it.
1735         * tree.h (fit_double_type): Export.
1736
1737 2007-01-08  Jan Hubicka  <jh@suse.cz>
1738
1739         * tree-vectorizer.c (gate_increase_alignment): Fix return type.
1740         * ipa.c (function_and_variable_visibility): Fix return type.
1741
1742 2007-01-08  Richard Guenther  <rguenther@suse.de>
1743
1744         * tree-ssa-ccp.c (maybe_fold_offset_to_array_ref): Use type
1745         of offset to build the index.
1746         * tree-pretty-print.c (dump_generic_node): Don't build negated
1747         const just for printing.
1748         * c-pretty-print.c (pp_c_integer_constant): Likewise.
1749         * builtins.c (fold_builtin_int_roundingfn): Check if result
1750         fits the type by using force_fit_type and comparing the result.
1751         * predict.c (predict_loops): Use compare_tree_int for comparison.
1752         * tree.c (build_int_cst): Fall back to integer_type_node for
1753         NULL_TREE type.
1754         (build_int_cst_wide): Assert type is non-null.
1755
1756 2007-01-08  Roberto Costa  <roberto.costa@st.com>
1757
1758         * tree-vrp.c (extract_range_from_cond_expr): New.
1759         (extract_range_from_expr): Handle COND_EXPR nodes used as expressions.
1760         * tree-ssa-ccp.c (get_maxval_strlen): Handle COND_EXPR nodes used
1761         as expressions.
1762         (fold_stmt): Bug fix, avoid infinite recursion when folding COND_EXPRs.
1763         * tree-ssa-forwprop.c (simplify_cond, forward_propagate_into_cond,
1764         tree_ssa_forward_propagate_single_use_vars): Handle COND_EXPR nodes
1765         used as expressions.
1766         * tree-object-size.c (cond_expr_object_size): New.
1767         (collect_object_sizes_for): Handle COND_EXPR nodes used as expressions.
1768
1769 2007-01-08  Jan Hubicka  <jh@suse.cz>
1770
1771         * tree-ssa-forwprop.c (forward_propagate_into_cond,
1772         tree_ssa_forward_propagate_single_use_va): Release defs of propagated
1773         statement.
1774
1775 2007-01-08  Richard Guenther  <rguenther@suse.de>
1776
1777         PR tree-optimization/23603
1778         * tree-vrp.c (set_value_range_to_truthvalue): New function.
1779         (extract_range_from_binary): Fall back to truthvalue instead of
1780         varying for TRUTH_*_EXPR.
1781         (extract_range_from_comparison): Fall back to truthvalue instead of
1782         varying.
1783         (vrp_visit_phi_node): Don't adjust new range bounds to +INF/-INF
1784         if all visited PHI values were constant.
1785
1786 2007-01-08  Jan Hubicka  <jh@suse.cz>
1787
1788         * cgraphunit.c (cgraph_process_new_functions): Reset reachable flag.
1789         (cgraph_analyze_function): break out from ...
1790         (cgraph_finalize_compilation_unit): ... here.
1791         (cgraph_expand_function): Remove forgoten commented out line.
1792         (cgraph_optimize): Analyze functions.
1793
1794 2007-01-08  Jan Hubicka  <jh@suse.cz>
1795
1796         * tree-pas.h (TODO_remove_function): New flag.
1797         (TODO_update*): Renumber.
1798         (pass_ipa_increase_alignment,
1799         pass_ipa_function_and_variable_visibility): New passes.
1800         * cgraphunit.c (cgraph_increase_alignment): Move to tree-vectorizer.c
1801         (cgraph_function_and_variable_visibility): Move to ipa.c
1802         (cgraph_optimize): Don't call cgraph_function_and_variable_visibility,
1803         cgraph_increase_alignment.
1804         * ipa-inline.c (cgraph_decide_inlining): Don't push timevar.
1805         (cgraph_decide_inlining_incrementally): Push TV_INTEGRATION before
1806         calling tree-inline.
1807         (cgraph_early_inlining): Do not call cgraph_remove_unreachable_nodes.
1808         (pass_ipa_inline, pass_early_ipa_inlining): Set TODO_remove_functions
1809         * tree-vectorizer.c (increase_alignment): Move here from cgraphunit.c
1810         (gate_increase_alignment): New function.
1811         (pass_ipa_increase_alignment): New pass.
1812         * ipa.c: Inline tree-pass.h and timevar.h
1813         (function_and_variable_visibility): Move here from cgraphunit.c
1814         * tree-optimize.c (pass_early_local_passes): Add TODO_remove_functions.
1815         * passes.c (init_optimization_passes): Add the two new passes.
1816         (execute_todo): Handle cgraph_remove_functions.
1817
1818 2007-01-08  Nick Clifton  <nickc@redhat.com>
1819
1820         * config/frv/predicates.md (reg_or_0_operand): Accept
1821         CONST_DOUBLEs.
1822
1823 2007-01-08  Ralf Cors├ępius <ralf.corsepius@rtems.org>
1824
1825         * config/bfin/rtems.h, config/bfin/t-rtems: New.
1826         * config.gcc: Add bfin*-rtems*.
1827
1828 2007-01-08  Mark Shinwell  <shinwell@codesourcery.com>
1829
1830         * c.opt: Add -flax-vector-conversions.
1831         * c-typeck.c (convert_for_assignment): Pass flag to
1832         vector_types_convertible_p to allow emission of note.
1833         (digest_init): Likewise.
1834         * c-opts.c: Handle -flax-vector-conversions.
1835         * c-common.c (flag_lax_vector_conversions): New.
1836         (vector_types_convertible_p): Unless -flax-vector conversions
1837         has been passed, disallow conversions between vectors with
1838         differing numbers of subparts and/or element types.  If such
1839         a conversion is disallowed, possibly emit a note on the first
1840         occasion only to inform the user of -flax-vector-conversions.
1841         The new last argument specifies this.
1842         * c-common.h (flag_lax_vector_conversions): New.
1843         (vector_types_convertible_p): Add extra argument.
1844         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Use
1845         char_type_node for V*QI type vectors.
1846         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins):
1847         Update to satisfy new typechecking rules.
1848         * config/rs6000/altivec.h (vec_cmple): Use vec_cmpge, for both
1849         C and C++ variants.
1850         * doc/invoke.texi (C Dialect Options): Document
1851         -flax-vector-conversions.
1852
1853 2007-01-08  Mark Shinwell  <shinwell@codesourcery.com>
1854
1855         PR tree-optimization/29877
1856         * tree-ssa-ter.c (is_replaceable_p): Deem assignments with
1857         a register variable on the RHS to not be replaceable.
1858
1859 2007-01-08  Chen Liqin  <liqin@sunnorth.com.cn>
1860         * config/score/t-score-elf (MULTILIB_OPTIONS): Change.
1861         * config/score/predicates.md (const_uimm5, sr0_operand, const_simm12,
1862         const_simm15, const_pow2, const_npow2): Added.
1863         * config/score/misc.md (insv, extv, extzv, movmemsi,
1864         move_lbu_a/b, mov_lhu_a/b etc): Added and fix some bug.
1865         * config/score/score.c (score_address_cost, score_select_cc_mode):
1866         Added.
1867         Change CONST_OK_FOR_LETTER_P/EXTRA_CONSTRAINT define.
1868         Update score_rtx_costs for MACRO TARGET_RTX_COSTS.
1869         Update score_print_operand.
1870         * config/score/score.h (DATA_ALIGNMENT, SELECT_CC_MODE): Added.
1871         Adjust register allocate order and update some macro define.
1872         * config/score/score-mdaux.c (mdx_unaligned_load, mdx_unsigned_store,
1873         mdx_block_move_straight, mdx_block_move_loop_head,
1874         mdx_block_move_loop_body, mdx_block_move_loop_foot, mdx_block_move_loop,
1875         mdx_block_move): Added.
1876         (mdx_movsicc, mdp_select_add_imm, mdp_select, mds_zero_extract_andi,
1877         mdp_limm): Updated and fix some bug and typo.
1878         * config/score/score.md (movqi/hi/si, add/sub/zero/ext): Updated.
1879         (movsf, movdf, doloop_end): Added.
1880
1881 2007-01-08  Kazu Hirata  <kazu@codesourcery.com>
1882
1883         * config/arm/arm.c, config/arm/arm.h, config/arm/arm.md,
1884         config/arm/thumb2.md: Fix comment typos.
1885         * doc/extend.texi: Fix a typo.
1886
1887 2007-01-07  Eric Christopher  <echristo@apple.com>
1888
1889         * configure.ac: Check for __stack_chk_fail for darwin.
1890         * configure: Regenerate.
1891
1892 2007-01-07  Richard Guenther  <rguenther@suse.de>
1893
1894         * tree-vrp.c (extract_range_from_assert): CSE calls to
1895         compare_values where possible.
1896         (extract_range_from_unary_expr): Likewise.
1897
1898 2007-01-07  Anatoly Sokolov <aesok@post.ru>
1899
1900         * config/avr/avr-protos.h (call_insn_operand): Delete prototype.
1901         * config/avr/avr.c (call_insn_operand): Delete function.
1902         * config/avr/avr.md (*pushqi, *pushhi, *pushsi, *pushsf): Use REG_SP
1903         instead of register number. Use predicates.
1904         * config/avr/predicates.md (const0_operand, reg_or_0_operand,
1905         call_insn_operand): Add.
1906
1907 2007-01-06  Jan Hubicka  <jh@suse.cz>
1908
1909         * tree-pass.h (pass_build_cgraph_edges): Declare.
1910         * cgraphunit.c (record_refernece): Move to cgraphbuild.c
1911         (visited_nodes): Remove.
1912         (cgraph_create_edges): Move to cgraphbuild.c; rename to
1913         build_cgrpah_edges; make visited_nodes local.
1914         (cgraph_process_new_functions): DO not call initialize_inline_failed.
1915         (record_references_in_initializer): Move to cgraphbuild.c
1916         (initialize_inline_failed, rebuild_cgraph_edges,
1917         pass_rebuild_cgraph_edges): Move to cgraphbuild.c.
1918         (verify_cgraph_node): Make visited_nodes local.
1919         (cgraph_analyze_function): Do not call cgraph_create_edges and
1920         initialize_inline_failed.
1921         (cgraph_expand_function): Do not call cgraph_lower_function;
1922         assert that function is already lowered.
1923         * Makefile.in (cgraphbuild.o): New.
1924         * passes.c (init_optimization_passes): Add pass_build_cgraph_edges
1925         at the end of lowering passes.
1926
1927 2007-01-06  Steven Bosscher  <steven@gcc.gnu.org>
1928
1929         * ifcvt.c (cond_move_convert_if_block): New function, code
1930         factored out from...
1931         (cond_move_process_if_block): ...here.  Call the new function
1932         on the THEN and ELSE blocks.
1933         (merge_if_block): Do not copy global_live_at_end, merge_blocks
1934         already takes care of this.
1935
1936 2007-01-05  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
1937
1938         PR c/19978
1939         * tree.h (TREE_OVERFLOW_P): New.
1940         * c-typeck.c (parser_build_unary_op): Warn only if result
1941         overflowed and operands did not.
1942         (parser_build_binary_op): Likewise.
1943         (convert_for_assignment): Remove redundant overflow_warning.
1944         * c-common.c (overflow_warning): Don't check or set TREE_OVERFLOW.
1945
1946 2007-01-05  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
1947
1948         * c-typeck.c (store_init_value): Split over two lines to follow
1949         the GNU coding style.
1950
1951 2007-01-05  Benjamin Kosnik  <bkoz@redhat.com>
1952
1953         * c-cppbuiltin.c (c_cpp_builtins): __GXX_EXPERIMENTAL_CPP0X__ to
1954         __GXX_EXPERIMENTAL_CXX0X__.
1955         * doc/cpp.texi: Same.
1956
1957 2007-01-05  Richard Guenther  <rguenther@suse.de>
1958
1959         PR middle-end/27826
1960         * tree.c (get_narrower): Do not construct COMPONENT_REFs
1961         with mismatched types.  Instead explicitly build a
1962         conversion NOP_EXPR.
1963
1964 2007-01-05  Ian Lance Taylor  <iant@google.com>
1965
1966         * c-common.c (decl_with_nonnull_addr_p): New function.
1967         (c_common_truthvalue_conversion): Call it.
1968         * c-typeck.c (build_binary_op): Likewise.
1969         * c-common.h (decl_with_nonnull_addr_p): Declare.
1970
1971 2007-01-05  Jakub Jelinek  <jakub@redhat.com>
1972
1973         PR c/30360
1974         * libgcc2.c (__divdc3): Compare c and d against 0.0 instead of
1975         denom against 0.0.
1976
1977 2007-01-05  Joel Brobecker  <brobecker@adacore.com>
1978
1979         * doc/install.texi (Final install): Document the fact that
1980         the GNAT runtime should not be stripped.
1981
1982 2007-01-04  Jan Hubicka  <jh@suse.cz>
1983
1984         * tree-inline.c (fold_marked_statements): Update operand caches
1985         and EH after folding
1986
1987 2007-01-04  Ian Lance Taylor  <iant@google.com>
1988
1989         * c-common.c (check_function_nonnull): Whitespace fix.
1990
1991 2007-01-04  Jan Hubicka  <jh@suse.cz>
1992
1993         * tree-optimize.c (execute_fixup_cfg): Correct previously mistakely
1994         comitted older version of patch.
1995         (pass_fixup_cfg): Add TODOs to verify flow and statements, dump
1996         function, celanup cfg and collect garbage.
1997
1998 2007-01-04  Mike Stump  <mrs@apple.com>
1999
2000         * Makefile.in (mostlyclean): Don't remove libgcc anymore.
2001         (clean): Likewise.
2002
2003 2007-01-04  Eric Christopher  <echristo@apple.com>
2004
2005         * libgcc2.c (__bswapsi2): Use SItype.
2006         (__bswapdi2): Use DItype.
2007         * libgcc2.h: Update for above.
2008
2009 2007-01-04  Paul Brook  <paul@codesourcery.com>
2010
2011         * config/arm/arm.md (arm_mulsi3, thumb_mulsi3, mulsi3_compare0,
2012         mulsi_compare0_scratch, mulsi3addsi, mulsi3addsi_compare0,
2013         mulsi3addsi_compare0_scratch, mulsidi3adddi, mulsidi3,
2014         umulsidi3, umulsidi3adddi, smulsi3_highpart,
2015         umulsi3_highpart): Make conditional on !arm_arch6.
2016         (arm_mulsi3_v6, thumb_mulsi3_v6, mulsi3_compare0_v6,
2017         mulsi_compare0_scratch_v6, mulsi3addsi_v6, mulsi3addsi_compare0_v6,
2018         mulsi3addsi_compare0_scratch_v6, mulsidi3adddi_v6, mulsidi3_v6,
2019         umulsidi3_v6, umulsidi3adddi_v6, smulsi3_highpart_v6,
2020         umulsi3_highpart_v6): New insns.
2021
2022 2007-01-04  Roger Sayle  <roger@eyesopen.com>
2023
2024         * fold-const.c (fold_convert): When casting an expression to void,
2025         fold_ignored_result may discover a GIMPLE_MODIFY_STMT which doesn't
2026         have a type.  Instead of attempting to build a NOP_EXPR, return
2027         these "special" trees directly.
2028
2029 2007-01-04  Joseph Myers  <joseph@codesourcery.com>
2030
2031         * config/rs6000/rs6000.c (rs6000_rtx_costs): Make adjustment for
2032         MULT inside MINUS as either argument.  Use rs6000_cost->dmul -
2033         rs6000_cost->fp not 0 as adjustment for outer NEG.
2034
2035 2007-01-04  Jan Hubicka  <jh@suse.cz>
2036
2037         * cgraph.c (cgraph_release_function_body): New function.
2038         (cgraph_remove_node): Use it.
2039         * cgraph.h (cgraph_release_function_body): Declare.
2040         * cgraphunit.c (cgraph_expand_function): Use it.
2041         * ipa.c (cgraph_remove_unreahchable_nodes): Use it.
2042         * tree-ssa.c (delete_tree_ssa): Allow to be called before aliasing
2043         is initialized and while compilation of other function is running.
2044         * tree-optimize.c (execute_free_cfg_annotations): Move code to clear
2045         statement CFG annotations from here to ...
2046         * tree-cfg.c (delete_tree_cfg_annotations): ... here.
2047
2048 2007-01-04  Zdenek Dvorak <dvorakz@suse.cz>
2049
2050         * cfgloop.h (enum li_flags): Make the constants powers of two.
2051
2052 2007-01-04  Jan Hubicka  <jh@suse.cz>
2053
2054         * tree-inline.c (copy_bb): Insert new statements to statements_to_fold
2055         set.
2056         (fold_marked_statements): New function.
2057         (optimize_inline_calls, tree_function_versioning): Fold new statements.
2058         * tree-inline.h (copy_body_data): Add statements_to_fold.
2059
2060 2007-01-03  Daniel Jacobowitz  <dan@codesourcery.com>
2061
2062         * config.gcc: Mention libgcc/config.host.
2063         * Makefile.in: Update comments mentioning libgcc.
2064         (LIBGCC, INSTALL_LIBGCC, GCC_PARTS, mklibgcc): Delete.
2065         (all.cross, start.encap, rest.encap, rest.cross): Update
2066         dependencies for libgcc move.
2067         (libgcc.mk, LIBGCC_DEPS, libgcov.a, libgcc.a, stmp-multilib)
2068         (clean-target, clean-target-libgcc): Delete.
2069         (srcdirify, GCC_EXTRA_PARTS): New macros.
2070         (libgcc-support, libgcc.mvars): New rules.
2071         (distclean): Remove mention of mklibgcc.
2072         (install): Don't reference INSTALL_LIBGCC.
2073         (install-common): Don't reference EXTRA_PARTS.
2074         (install-libgcc, install-multilib): Delete rules.
2075         * mklibgcc.in: Delete file.
2076         * doc/configfiles.texi: Don't mention mklibgcc.
2077
2078         * config/i386/t-darwin (SHLIB_VERPFX): Delete (moved to libgcc).
2079         * config/i386/t-darwin64 (SHLIB_VERPFX): Likewise.
2080         * config/rs6000/t-darwin (SHLIB_VERPFX): Likewise.
2081         * config/rs6000/t-ppccomm (TARGET_LIBGCC2_CFLAGS, SHLIB_MAPFILES)
2082         (mklibgcc, ldblspecs): Likewise.
2083
2084         * config/i386/t-nwld (libgcc.def, libc.def, libpcre.def)
2085         (posixpre.def): Use $(T).
2086         (SHLIB_EXT, SHLIB_NAME, SHLIB_SLIBDIR_QUAL, SHLIB_DEF, SHLIB_MAP)
2087         (SHLIB_SRC, SHLIB_INSTALL): Delete.
2088         (SHLIB_LINK): Make dummy.
2089         * config/t-slibgcc-darwin: Delete contents except for dummy SHLIB_LINK.
2090
2091         * config/frv/t-linux (EXTRA_MULTILIB_PARTS): Clear.
2092
2093         * config/alpha/t-crtfm: Use $(T) in rules for EXTRA_PARTS.
2094         * config/alpha/t-vms, config/alpha/t-vms64, config/fr30/t-fr30,
2095         config/i386/t-rtems-i386, config/ia64/t-ia64, config/rs6000/t-beos,
2096         config/rs6000/t-newas, config/sparc/t-elf: Likewise.
2097
2098         * configure.ac (all_outputs): Remove mklibgcc.
2099         * configure: Regenerated.
2100
2101 2007-01-03  Josh Conner  <jconner@apple.com>
2102
2103         PR middle-end/29683
2104         * calls.c (compute_argument_addresses): Set stack and stack_slot
2105         for partial args, too.
2106         (store_one_arg): Use locate.size.constant for the size when
2107         generating a save_area.
2108
2109 2007-01-03  Robert Kennedy <jimbob@google.com>
2110
2111         * tree-cfg.c (tree_merge_blocks): Release SSA_NAME phi results
2112         whose definitions are deleted due to basic block merging.
2113
2114 2007-01-03  Paul Brook  <paul@codesourcery.com>
2115
2116         PR target/16634
2117         * config/arm/arm.c (output_return_instruction): Pop PC in interrupt
2118         functions.
2119         (use_return_insn): Return 0 for Thumb interrupt functions.
2120         (print_multi_reg): Add rfe argument for IRQ returns.
2121         (arm_output_epilogue): Pop interrupt return address directly into PC.
2122         (arm_expand_prologue): Only adjust IRQ return address in Arm mode.
2123
2124 2007-01-03  Paul Brook  <paul@codesourcery.com>
2125
2126         Merge from sourcerygxx-4_1.
2127         * config/arm/thumb2.md: New file.
2128         * config/arm/elf.h (JUMP_TABLES_IN_TEXT_SECTION): Return True for
2129         Thumb-2.
2130         * config/arm/coff.h (JUMP_TABLES_IN_TEXT_SECTION): Ditto.
2131         * config/arm/aout.h (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
2132         (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump tables.
2133         * config/arm/aof.h (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump
2134         tables.
2135         (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
2136         * config/arm/ieee754-df.S: Use macros for Thumb-2/Unified asm
2137         comptibility.
2138         * config/arm/ieee754-sf.S: Ditto.
2139         * config/arm/arm.c (thumb_base_register_rtx_p): Rename...
2140         (thumb1_base_register_rtx_p): ... to this.
2141         (thumb_index_register_rtx_p): Rename...
2142         (thumb1_index_register_rtx_p): ... to this.
2143         (thumb_output_function_prologue): Rename...
2144         (thumb1_output_function_prologue): ... to this.
2145         (thumb_legitimate_address_p): Rename...
2146         (thumb1_legitimate_address_p): ... to this.
2147         (thumb_rtx_costs): Rename...
2148         (thumb1_rtx_costs): ... to this.
2149         (thumb_compute_save_reg_mask): Rename...
2150         (thumb1_compute_save_reg_mask): ... to this.
2151         (thumb_final_prescan_insn): Rename...
2152         (thumb1_final_prescan_insn): ... to this.
2153         (thumb_expand_epilogue): Rename...
2154         (thumb1_expand_epilogue): ... to this.
2155         (arm_unwind_emit_stm): Rename...
2156         (arm_unwind_emit_sequence): ... to this.
2157         (thumb2_legitimate_index_p, thumb2_legitimate_address_p,
2158         thumb1_compute_save_reg_mask, arm_dwarf_handle_frame_unspec,
2159         thumb2_index_mul_operand, output_move_vfp, arm_shift_nmem,
2160         arm_save_coproc_regs, thumb_set_frame_pointer, arm_print_condition,
2161         thumb2_final_prescan_insn, thumb2_asm_output_opcode, arm_output_shift,
2162         thumb2_output_casesi): New functions.
2163         (TARGET_DWARF_HANDLE_FRAME_UNSPEC): Define.
2164         (FL_THUMB2, FL_NOTM, FL_DIV, FL_FOR_ARCH6T2, FL_FOR_ARCH7,
2165         FL_FOR_ARCH7A, FL_FOR_ARCH7R, FL_FOR_ARCH7M, ARM_LSL_NAME,
2166         THUMB2_WORK_REGS): Define.
2167         (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv, arm_condexec_count,
2168         arm_condexec_mask, arm_condexec_masklen)): New variables.
2169         (all_architectures): Add armv6t2, armv7, armv7a, armv7r and armv7m.
2170         (arm_override_options): Check new CPU capabilities.
2171         Set new architecture flag variables.
2172         (arm_isr_value): Handle v7m interrupt functions.
2173         (user_return_insn): Return 0 for v7m interrupt functions.  Handle
2174         Thumb-2.
2175         (const_ok_for_arm): Handle Thumb-2 constants.
2176         (arm_gen_constant): Ditto.  Use movw when available.
2177         (arm_function_ok_for_sibcall): Return false for v7m interrupt
2178         functions.
2179         (legitimize_pic_address, arm_call_tls_get_addr): Handle Thumb-2.
2180         (thumb_find_work_register, arm_load_pic_register,
2181         legitimize_tls_address, arm_address_cost, load_multiple_sequence,
2182         emit_ldm_seq, emit_stm_seq, arm_select_cc_mode, get_jump_table_size,
2183         print_multi_reg, output_mov_long_double_fpa_from_arm,
2184         output_mov_long_double_arm_from_fpa, output_mov_double_fpa_from_arm,
2185         output_mov_double_fpa_from_arm, output_move_double,
2186         arm_compute_save_reg_mask, arm_compute_save_reg0_reg12_mask,
2187         output_return_instruction, arm_output_function_prologue,
2188         arm_output_epilogue, arm_get_frame_offsets, arm_regno_class,
2189         arm_output_mi_thunk, thumb_set_return_address): Ditto.
2190         (arm_expand_prologue): Handle Thumb-2.  Use arm_save_coproc_regs.
2191         (arm_coproc_mem_operand): Allow POST_INC/PRE_DEC.
2192         (arithmetic_instr, shift_op): Use arm_shift_nmem.
2193         (arm_print_operand): Use arm_print_condition.  Handle '(', ')', '.',
2194         '!' and 'L'.
2195         (arm_final_prescan_insn): Use extract_constrain_insn_cached.
2196         (thumb_expand_prologue): Use thumb_set_frame_pointer.
2197         (arm_file_start): Output directive for unified syntax.
2198         (arm_unwind_emit_set): Handle stack alignment instruction.
2199         * config/arm/lib1funcs.asm: Remove default for __ARM_ARCH__.
2200         Add v6t2, v7, v7a, v7r and v7m.
2201         (RETLDM): Add Thumb-2 code.
2202         (do_it, shift1, do_push, do_pop, COND, THUMB_SYNTAX): New macros.
2203         * config/arm/arm.h (TARGET_CPU_CPP_BUILTINS): Define __thumb2__.
2204         (TARGET_THUMB1, TARGET_32BIT, TARGET_THUMB2, TARGET_DSP_MULTIPLY,
2205         TARGET_INT_SIMD, TARGET_UNIFIED_ASM, ARM_FT_STACKALIGN, IS_STACKALIGN,
2206         THUMB2_TRAMPOLINE_TEMPLATE, TRAMPOLINE_ADJUST_ADDRESS,
2207         ASM_OUTPUT_OPCODE, THUMB2_GO_IF_LEGITIMATE_ADDRESS,
2208         THUMB2_LEGITIMIZE_ADDRESS, CASE_VECTOR_PC_RELATIVE,
2209         CASE_VECTOR_SHORTEN_MODE, ADDR_VEC_ALIGN, ASM_OUTPUT_CASE_END,
2210         ADJUST_INSN_LENGTH): Define.
2211         (TARGET_REALLY_IWMMXT, TARGET_IWMMXT_ABI, CONDITIONAL_REGISTER_USAGE,
2212         STATIC_CHAIN_REGNUM, HARD_REGNO_NREGS, INDEX_REG_CLASS,
2213         BASE_REG_CLASS, MODE_BASE_REG_CLASS, SMALL_REGISTER_CLASSES,
2214         PREFERRED_RELOAD_CLASS, SECONDARY_OUTPUT_RELOAD_CLASS,
2215         SECONDARY_INPUT_RELOAD_CLASS, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P,
2216         TRAMPOLINE_SIZE, INITIALIZE_TRAMPOLINE, HAVE_PRE_INCREMENT,
2217         HAVE_POST_DECREMENT, HAVE_PRE_DECREMENT, HAVE_PRE_MODIFY_DISP,
2218         HAVE_POST_MODIFY_DISP, HAVE_PRE_MODIFY_REG, HAVE_POST_MODIFY_REG,
2219         REGNO_MODE_OK_FOR_BASE_P, LEGITIMATE_CONSTANT_P,
2220         REG_MODE_OK_FOR_BASE_P, REG_OK_FOR_INDEX_P, GO_IF_LEGITIMATE_ADDRESS,
2221         LEGITIMIZE_ADDRESS, THUMB2_LEGITIMIZE_ADDRESS,
2222         GO_IF_MODE_DEPENDENT_ADDRESS, MEMORY_MOVE_COST, BRANCH_COST,
2223         ASM_APP_OFF, ASM_OUTPUT_CASE_LABEL, ARM_DECLARE_FUNCTION_NAME,
2224         FINAL_PRESCAN_INSN, PRINT_OPERAND_PUNCT_VALID_P,
2225         PRINT_OPERAND_ADDRESS): Adjust for Thumb-2.
2226         (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv): New declarations.
2227         * config/arm/arm-cores.def: Add arm1156t2-s, cortex-a8, cortex-r4 and
2228         cortex-m3.
2229         * config/arm/arm-tune.md: Regenerate.
2230         * config/arm/arm-protos.h: Update prototypes.
2231         * config/arm/vfp.md: Enable patterns for Thumb-2.
2232         (arm_movsi_vfp): Add movw alternative.  Use output_move_vfp.
2233         (arm_movdi_vfp, movsf_vfp, movdf_vfp): Use output_move_vfp.
2234         (thumb2_movsi_vfp, thumb2_movdi_vfp, thumb2_movsf_vfp,
2235         thumb2_movdf_vfp, thumb2_movsfcc_vfp, thumb2_movdfcc_vfp): New.
2236         * config/arm/libunwind.S: Add Thumb-2 code.
2237         * config/arm/constraints.md: Update include Thumb-2.
2238         * config/arm/ieee754-sf.S: Add Thumb-2/Unified asm support.
2239         * config/arm/ieee754-df.S: Ditto.
2240         * config/arm/bpabi.S: Ditto.
2241         * config/arm/t-arm (MD_INCLUDES): Add thumb2.md.
2242         * config/arm/predicates.md (low_register_operand,
2243         low_reg_or_int_operand, thumb_16bit_operator): New.
2244         (thumb_cmp_operand, thumb_cmpneg_operand): Rename...
2245         (thumb1_cmp_operand, thumb1_cmpneg_operand): ... to this.
2246         * config/arm/t-arm-elf: Add armv7 multilib.
2247         * config/arm/arm.md: Update patterns for Thumb-2 and Unified asm.
2248         Include thumb2.md.
2249         (UNSPEC_STACK_ALIGN, ce_count): New.
2250         (arm_incscc, arm_decscc, arm_umaxsi3, arm_uminsi3,
2251         arm_zero_extendsidi2, arm_zero_extendqidi2): New
2252         insns/expanders.
2253         * config/arm/fpa.md: Update patterns for Thumb-2 and Unified asm.
2254         (thumb2_movsf_fpa, thumb2_movdf_fpa, thumb2_movxf_fpa,
2255         thumb2_movsfcc_fpa, thumb2_movdfcc_fpa): New insns.
2256         * config/arm/cirrus.md: Update patterns for Thumb-2 and Unified asm.
2257         (cirrus_thumb2_movdi, cirrus_thumb2_movsi_insn,
2258         thumb2_cirrus_movsf_hard_insn, thumb2_cirrus_movdf_hard_insn): New
2259         insns.
2260         * doc/extend.texi: Document ARMv7-M interrupt functions.
2261         * doc/invoke.texi: Document Thumb-2 new cores+architectures.
2262
2263 2007-01-03  Jakub Jelinek  <jakub@redhat.com>
2264
2265         * unwind-dw2.c (SIGNAL_FRAME_BIT, EXTENDED_CONTEXT_BIT): Define.
2266         (struct _Unwind_Context): Rename args_size to flags, remove
2267         signal_frame field, add a new args_size field and version field.
2268         (_Unwind_IsSignalFrame, _Unwind_SetSignalFrame,
2269         _Unwind_IsExtendedContext): New inline functions.
2270         (_Unwind_GetGR, _Unwind_SetGR, _Unwind_GetGRPtr, _Unwind_SetGRPtr):
2271         Assume by_value array is only present if _Unwind_IsExtendedContext.
2272         (_Unwind_GetIPInfo, execute_cfa_program, uw_frame_state_for): Use
2273         _Unwind_IsSignalFrame.
2274         (__frame_state_for): Initialize context.flags to EXTENDED_CONTEXT_BIT.
2275         (uw_update_context_1): Use _Unwind_SetSignalFrame.
2276         (uw_init_context_1): Initialize context->flags to
2277         EXTENDED_CONTEXT_BIT.
2278         * config/rs6000/linux-unwind.h (frob_update_context): Use
2279         _Unwind_SetSignalFrame.
2280
2281 2007-01-03  Andrew Pinski  <andrew_pinski@playstation.sony.com>
2282
2283         PR middle-end/30353
2284         * gimplify.c (gimplify_modify_expr_complex_part): Move below
2285         tree_to_gimple_tuple.  Call tree_to_gimple_tuple when we need
2286         the value.
2287
2288 2007-01-03  Kazu Hirata  <kazu@codesourcery.com>
2289
2290         * config/i386/i386.h (NON_STACK_REG_P, REGNO_OK_FOR_SIREG_P,
2291         REGNO_OK_FOR_DIREG_P, REWRITE_ADDRESS, ASM_OPERAND_LETTER,
2292         RET, AT_SP): Remove.
2293         * config/i386/i386.md (*sse_prologue_save_insn): Use return
2294         instead of RET.
2295
2296         * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
2297         * cfgbuild.c (find_basic_blocks): Likewise.
2298         * cfgrtl.c (rtl_create_basic_block): Likewise.
2299         * function.c (temp_slots_at_level): Likewise.
2300         * reg-stack.c (stack_regs_mentioned): Likewise.
2301         * regclass.c (allocate_reg_info): Likewise.
2302         * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
2303         set_bb_for_stmt, move_block_to_fn): Likewise.
2304         * tree-complex.c (tree_lower_complex): Likewise.
2305         * vec.h (VEC_safe_grow_cleared): New.
2306
2307         * cgraphunit.c, tree-ssa-alias.c: Fix comment typos.
2308
2309 2007-01-03  Zdenek Dvorak <dvorakz@suse.cz>
2310
2311         * loop-unswitch.c (unswitch_loop): Pass probabilities to loopify.
2312         * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Pass probabilities
2313         to loop_version.
2314         * cfgloopmanip.c (scale_loop_frequencies): Export.
2315         (loopify): Scale the frequencies by prescribed coefficients.
2316         (set_zero_probability): New function.
2317         (duplicate_loop_to_header_edge): Improve updating of frequencies.
2318         (lv_adjust_loop_entry_edge, loop_version): Set probabilities
2319         and frequencies according to arguments.
2320         * tree-ssa-loop-manip.c (tree_unroll_loop): Set probabilities
2321         correctly.
2322         * cfg.c (scale_bbs_frequencies_int): Allow scaling the frequencies up.
2323         * modulo-sched.c (sms_schedule): Set probabilities for entering
2324         versioned loop correctly.
2325         * tree-vect-transform.c (vect_transform_loop): Ditto.
2326         * cfgloop.h (loopify, loop_version): Declaration changed.
2327         (scale_loop_frequencies): Declared.
2328
2329 2007-01-02  Jan Hubicka  <jh@suse.cz>
2330
2331         * cgraph.c: Include tree-flow.h
2332         (cgraph_add_new-function): Handle IPA_SSA mode; execute
2333         early_local_passes.
2334         * cgraph.h (enum cgraph_state): Add CGRAPH_STATE_IPA_SSA.
2335         * tree-pass.h (pass_all_early_optimizations): Declare.
2336         * cgraphunit.c (cgraph_process_new_functions): Add IPA_SSA; execute
2337         early_local_passes.
2338         (cgraph_analyze_function): Do early_local_passes.
2339         * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
2340         Do not add referenced vars.
2341         * tree-optimize.c (gate_all_optimizations): Do not execute when not in
2342         SSA form.
2343         (gate_all_early_local_passes): New gate.
2344         (pass_early_local_passes): Use new gate.
2345         (execute_early_local_optimizations): New functions.
2346         (gate_all_early_optimizations): New gate.
2347         (pass_all_early_optimizations): New pass.
2348         (execute_free_datastructures): Free SSA only when initialized.
2349         (gate_init_datastructures): Init only when optimizing.
2350         (tree_lowering_passes): Do early local passes when called late.
2351         * tree-profile.c (do_tree_profiling): Don't profile functions added
2352         late.
2353         (do_early_tree_profiling, pass_early_tree_profile): Kill.
2354         * tree-cfg.c (update_modified_stmts): Do not update when operands are
2355         not active.
2356         * passes.c (init_optimizations_passes): Reorder so we go into SSA
2357         during early_local_passes.
2358         * Makefile.in (cgraph.o): Add dependency on tree-flow.h.
2359
2360
2361 2007-01-02  Carlos O'Donell  <carlos@codesourcery.com>
2362
2363         * Makefile.in: Update copyright year.
2364
2365 2007-01-02  Carlos O'Donell  <carlos@codesourcery.com>
2366
2367         * Makefile.in: Export GCC_EXEC_PREFIX before calling $(RUNTEST)
2368         in $(lang_checks) and check-consistency targets.
2369
2370 2007-01-02  Jan Hubicka  <jh@suse.cz>
2371
2372         * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
2373         Do not add referenced vars.
2374         * tree-cfg.c (update_modified_stmts): Do not update when SSA operands
2375         are not active.
2376         * passes.c (init_optimization_passes): Put mudflap_2 after
2377         free_datastructures.
2378
2379 2007-01-02  Jan Hubicka  <jh@suse.cz>
2380
2381         * tree-optimize (execute_fixup_cfg): Set after_inlining flag.
2382         Set NOTHROW flag on call statements proved to be nothrow.
2383         Update statement of local calls so new pure/const functions are
2384         updated. Update_ssa when in ssa form. Mark PHI nodes of nonlocal
2385         goto receivers.
2386         (tree_rest_of_compilation): Register hooks and initialize bitmap
2387         early. Do not set after_inlining flag.
2388
2389 2007-01-02  Steve Ellcey  <sje@cup.hp.com>
2390
2391         * sbitmap.c (HOST_BITS_PER_LONG_LONG):  Change to
2392         HOST_BITS_PER_LONGLONG
2393
2394 2007-01-02  Manuel Lopez-Ibanez <manu@gcc.gnu.org>
2395
2396         PR c/19977
2397         * c-typeck.c (store_init_value): Don't emit pedantic overflow
2398         warning for non-static initializers.
2399
2400 2007-01-02  Steven Bosscher  <steven@gcc.gnu.org>
2401
2402         * config/alpha/alpha.md, arm/arm.c, darwin.c, frv/frv.md,
2403         m32r/m32r.c, m32r/m32r.c, mn10300/mn10300.md, pa/pa.c,
2404         rs6000/rs6000.c, s390/s390.md, sh/sh.md, sparc/sparc.c:
2405         Always use set_unique_reg_note to add REG_EQUAL notes.
2406
2407 2007-01-02  Kazu Hirata  <kazu@codesourcery.com>
2408
2409         Revert:
2410         2007-01-02  Kazu Hirata  <kazu@codesourcery.com>
2411
2412         * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
2413         * cfgbuild.c (find_basic_blocks): Likewise.
2414         * cfgrtl.c (rtl_create_basic_block): Likewise.
2415         * function.c (temp_slots_at_level): Likewise.
2416         * reg-stack.c (stack_regs_mentioned): Likewise.
2417         * regclass.c (allocate_reg_info): Likewise.
2418         * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
2419         set_bb_for_stmt, move_block_to_fn): Likewise.
2420         * tree-complex.c (tree_lower_complex): Likewise.
2421         * vec.h (VEC_safe_grow_cleared): New.
2422
2423 2007-01-02  Ian Lance Taylor  <iant@google.com>
2424
2425         * c-common.c (c_common_truthvalue_conversion): When warning about
2426         using an assignment as a truth value, set TREE_NO_WARNING.
2427
2428 2007-01-02  Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
2429
2430         PR middle-end/7651
2431         * c.opt (Wold-style-declaration): New.
2432         * doc/invoke.texi (C-only Warning Options): New.
2433         (Wold-style-declaration): Document it.
2434         (Wextra): Enabled by -Wextra.
2435         * c-opts.c (c_common_post_options): Enabled by -Wextra.
2436         * c-decl.c (declspecs_add_scspec): Replace -Wextra with
2437         -Wold-style-declaration.
2438
2439 2007-01-02  Kazu Hirata  <kazu@codesourcery.com>
2440
2441         * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
2442         * cfgbuild.c (find_basic_blocks): Likewise.
2443         * cfgrtl.c (rtl_create_basic_block): Likewise.
2444         * function.c (temp_slots_at_level): Likewise.
2445         * reg-stack.c (stack_regs_mentioned): Likewise.
2446         * regclass.c (allocate_reg_info): Likewise.
2447         * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
2448         set_bb_for_stmt, move_block_to_fn): Likewise.
2449         * tree-complex.c (tree_lower_complex): Likewise.
2450         * vec.h (VEC_safe_grow_cleared): New.
2451
2452 2007-01-02  Douglas Gregor  <doug.gregor@gmail.com>
2453
2454         * c-common.c(c_common_nodes_and_builtins): Since variants of
2455         void_type_node get built before it is given a name, we need to
2456         give those variants the name, too.
2457         (complete_array_type): We need to work with the canonical main
2458         type of the array, from which we will build the qualified version.
2459         * params.def (PARAM_VERIFY_CANONICAL_TYPES): New.
2460         * print-tree.c (print_node): Display canonical type information
2461         for each type.
2462         * stor-layout.c (layout_type): When we don't know the
2463         alignment of a type for which we're building an array, we end up
2464         guessing wrong, so make the type require structural equality.
2465         * tree.c (make_node_stat): When we build a new type, it is its
2466         own canonical type.
2467         (build_type_attribute_qual_variant): When building an attribute
2468         variant, its canonical type is the non-attribute variant. However,
2469         if the attributes are target-dependent and they differ, we need to
2470         use structural equality checks for this type.
2471         (build_qualified_type): A qualified type is not equivalent to its
2472         unqualified variant; set the canonical type appropriately.
2473         (build_distinct_type_copy): When building a distinct type from
2474         another type, the new type is its own canonical type.
2475         (build_variant_type_copy): When building a new type variant, we
2476         assume that it is equivalent to the original type.
2477         (build_pointer_type_for_mode): When building a pointer type, also
2478         build a canonical type pointer.
2479         (build_reference_type_for_mode): When building a reference type,
2480         also build a canonical type reference.
2481         (build_index_type): When we can't hash an index type (e.g.,
2482         because its maximum value is negative), the index type requires
2483         structural equality tests.
2484         (build_array_type): Build the canonical form of an array type.
2485         (build_function_type): Function types require structural equality,
2486         because they contain default arguments, attributes, etc.
2487         (build_method_type_directly): Ditto for method types.
2488         (build_offset_type): Build the canonical offset type.
2489         (build_complex_type): Build the canonical vector type.
2490         (make_vector_type): Build the canonical vector type.
2491         * tree.h (TYPE_CANONICAL): New.
2492         (TYPE_STRUCTURAL_EQUALITY_P): New.
2493         (SET_TYPE_STRUCTURAL_EQUALITY): New.
2494         (struct tree_type): Added "canonical" field.
2495         * params.h (VERIFY_CANONICAL_TYPES): New.
2496         * doc/c-tree.texi (TYPE_CANONICAL): Document.
2497         (TYPE_STRUCTURAL_EQUALITY_P): Document.
2498         (SET_TYPE_STRUCTURAL_EQUALITY): Document.
2499         * doc/invoke.texi (verify-canonical-types): Document --param
2500         parameter for verifying canonical types.
2501
2502 2007-01-02  Joseph Myers  <joseph@codesourcery.com>
2503
2504         * config.gcc (powerpc-*-eabispe*, powerpc-*-eabisimaltivec*,
2505         powerpc-*-eabisim*, powerpc-*-eabialtivec*, powerpc-*-eabi*,
2506         powerpc-*-rtems*, powerpc-wrs-vxworks, powerpc-wrs-vxworksae,
2507         powerpcle-*-eabisim*, powerpcle-*-eabi*): Add rs6000/e500.h to
2508         tm_file.
2509         * config/rs6000/e500.h: New.
2510         * config/rs6000/eabi.h (TARGET_SPE_ABI, TARGET_SPE, TARGET_E500,
2511         TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE, TARGET_E500_DOUBLE):
2512         Remove.
2513         * config/rs6000/linuxspe.h (TARGET_SPE_ABI, TARGET_SPE,
2514         TARGET_E500, TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE,
2515         TARGET_E500_DOUBLE): Remove.
2516         * config/rs6000/vxworks.h (TARGET_SPE_ABI, TARGET_SPE,
2517         TARGET_E500, TARGET_ISEL, TARGET_FPRS): Remove.
2518         * config/rs6000/rs6000.h (CHECK_E500_OPTIONS): Define.
2519         * config/rs6000/rs6000.c (rs6000_override_options): Use
2520         CHECK_E500_OPTIONS.
2521
2522 2007-01-02  Joseph Myers  <joseph@codesourcery.com>
2523
2524         * config/rs6000/rs6000.c (print_operand): Check (TARGET_SPE ||
2525         TARGET_E500_DOUBLE), not TARGET_E500, for %y.
2526         (rs6000_generate_compare, rs6000_emit_sCOND, output_cbranch,
2527         rs6000_emit_cmove): Don't check TARGET_E500.
2528         * config/rs6000/rs6000.md (bunordered, bordered, sunordered,
2529         sordered): Don't check TARGET_E500.
2530
2531 2007-01-01  Eric Christopher  <echristo@apple.com>
2532
2533         * config/mips/mips.c (mips_regno_mode_ok_for_base_p): Use
2534         HARD_REGISTER_NUM_P.
2535
2536 2007-01-01  Roger Sayle  <roger@eyesopen.com>
2537
2538         * fold-const.c (fold_binary) <EQ_EXPR>: Fold "(X^C1) eq/ne C2" into
2539         "X eq/ne (C1^C2)".  Fold "(X^Z) eq/ne (Y^Z)" as "X eq/ne Y" when Z
2540         has no side-effects.  Fold "(X^C1) eq/ne (Y^C2)" as "(X^(C1^C2))
2541         eq/ne Y".
2542
2543 2007-01-01  Mike Stump  <mrs@apple.com>
2544
2545         * configure.ac: Remove support for building with Apple's gcc-3.1.
2546
2547 2007-01-02  Joseph Myers  <joseph@codesourcery.com>
2548
2549         PR middle-end/30311
2550         * caller-save.c (add_stored_regs): Only handle SUBREGs if inner
2551         REG is a hard register.  Do not modify REG before calling
2552         subreg_nregs.
2553         * rtlanal.c (subreg_get_info): Don't assert size of XMODE is a
2554         multiple of the size of YMODE for certain lowpart cases.
2555
2556 2007-01-01  Andrew Pinski  <pinskia@gmail.com>
2557
2558         PR middle-end/30253
2559         * gimplify (voidify_wrapper_expr): Update for
2560         GIMPLE_MODIFY_STMT.
2561
2562 2007-01-01  Andreas Schwab  <schwab@suse.de>
2563
2564         PR target/29166
2565         * config/ia64/ia64.c (ia64_compute_frame_size): Account space for
2566         save of BR0 in extra_spill_size instead of spill_size.
2567         (ia64_expand_prologue): Save BR0 outside of the gr/br/fr spill
2568         area.
2569         (ia64_expand_epilogue): Restore BR0 from its new location.
2570
2571 2007-01-01  Andrew Pinski  <pinskia@gmail.com>
2572
2573         * gimplify.c (gimplify_init_constructor <case VECTOR_TYPE>):
2574         Use a temporary variable if the left hand side is not a gimple
2575         register.
2576
2577 2007-01-01  Andrew Pinski  <pinskia@gmail.com>
2578
2579         * gimplify.c (gimplify_return_expr): Make the temporary variable
2580         for the return expression, a gimple register variable.
2581
2582 2007-01-01  Jan Hubicka  <jh@suse.cz>
2583
2584         * emit-rtl.c (emit_copy_of_insn_after): Do not call copy_insn_1 for
2585         INSN_LIST.
2586
2587 2007-01-01  Mike Stump  <mrs@apple.com>
2588
2589         * configure.ac (HAVE_GAS_LITERAL16): Add autoconf check for
2590         .literal16.
2591         * config/darwin.c (machopic_select_rtx_section): Use
2592         HAVE_GAS_LITERAL16.
2593         (darwin_mergeable_constant_section): Likewise.
2594         * configure: Regenerate.
2595         * config.in: Regenerate.
2596
2597 2007-01-01  Jan Hubicka  <jh@suse.cz>
2598             Andrew Pinski  <pinskia@gmail.com>
2599
2600         * cgraphunit.c (cgraph_optimize): Call cgraph_add_new_functions
2601         before starting IPA passes.