ipa-inline.c (initial_insns, max_insns): Delete.
[gcc.git] / gcc / ChangeLog
1 2007-01-24 Jan Hubicka <jh@suse.cz>
2
3 * ipa-inline.c (initial_insns, max_insns): Delete.
4 (compute_max_insns): New function.
5 (cgraph_decide_inlining_of_small_function): Use it; take minimal amount
6 of insns as base for code growth.
7 (cgraph_decide_inlining): Make initial_insns local; do not compute
8 max_insns.
9 * params.def (PARAM_INLINE_UNIT_GROWTH): Set to 60.
10 * doc/invoke.texi (inline-unit-growth): Update docs.
11
12 2007-01-24 Jakub Jelinek <jakub@redhat.com>
13
14 * config/i386/i386.h (x86_cmpxchg16b): Remove const.
15 (TARGET_CMPXCHG16B): Define to x86_cmpxchg16b.
16 * config/i386/i386.c (x86_cmpxchg16b): Remove const.
17 (override_options): Add PTA_CX16 flag. Set x86_cmpxchg16b
18 for CPUs that have PTA_CX16 set.
19
20 PR middle-end/27416
21 * gimplify.c (omp_check_private): New function.
22 (gimplify_scan_omp_clauses): Use it for
23 firstprivate/lastprivate/reduction.
24
25 PR middle-end/30494
26 * gimplify.c (omp_add_variable): Don't call omp_notice_variable
27 on TYPE_SIZE_UNIT for GOVD_LOCAL VLAs.
28
29 PR middle-end/30421
30 * omp-low.c (lower_omp_for_lastprivate): Add dlist argument.
31 If lower_lastprivate_clauses emits some statements, append them
32 to dlist rather than body_p and to body_p append an initializer.
33 (lower_omp_for): Adjust caller.
34
35 2007-01-24 Steve Ellcey <sje@cup.hp.com>
36
37 * target.h (globalize_decl_name): New.
38 * target-def.h (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
39 * output.h (default_globalize_decl_name): New.
40 * varasm.c (asm_output_bss): Use globalize_decl_name instead of
41 globalize_label.
42 (globalize_decl): Ditto.
43 (default_globalize_decl_name): New.
44 * config/ia64/ia64.c (ia64_globalize_decl_name): New.
45 (ia64_handle_version_id_attribute): New.
46 (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
47 (ia64_asm_output_external): Use globalize_decl_name instead
48 of globalize_label.
49 * doc/extend.texi (version_id): New pragma.
50 * doc/tm.texi (ARGET_ASM_GLOBALIZE_DECL_NAME): New target hook.
51
52 2007-01-24 Andreas Krebbel <krebbel1@de.ibm.com>
53
54 * unwind-dw2-fde.c (get_cie_encoding): Replaced _Unwind_Word with
55 _uleb128_t and _Unwind_SWord with _sleb128_t.
56 * unwind-dw2.c (extract_cie_info, execute_stack_op, execute_cfa_program,
57 uw_frame_state_for, uw_update_context_1): Likewise.
58 * unwind-c.c (parse_lsda_header, PERSONALITY_FUNCTION): Likewise.
59 * unwind-pe.h (read_uleb128, read_sleb128,
60 read_encoded_value_with_base): Likewise.
61 * unwind-generic.h: Define _sleb128_t and _uleb128_t types.
62
63 2007-01-24 Richard Guenther <rguenther@suse.de>
64
65 * builtins.c (expand_builtin_cexpi): Get the fndecl
66 for cexp in the correct way.
67
68 2007-01-24 Jan Hubicka <jh@suse.cz>
69
70 * tree-ssa-dce.c (eliminate_unnecesary_stmts): Remove dead LHS of calls.
71
72 2007-01-24 Andreas Krebbel <krebbel1@de.ibm.com>
73
74 * c-cppbuiltin.c (builtin_define_type_sizeof): New function.
75 (c_cpp_builtins): New builtin macros: __SIZEOF_INT__, __SIZEOF_LONG__,
76 __SIZEOF_LONG_LONG__, __SIZEOF_SHORT__, __SIZEOF_POINTER__,
77 __SIZEOF_FLOAT__, __SIZEOF_DOUBLE__, __SIZEOF_LONG_DOUBLE__,
78 __SIZEOF_SIZE_T__, __SIZEOF_WCHAR_T__, __SIZEOF_WINT_T__ and
79 __SIZEOF_PTRDIFF_T__.
80 * doc/cpp.texi: Documentation for the new builtin macros added.
81
82 2007-01-24 Uros Bizjak <ubizjak@gmail.com>
83
84 * config/i386/i386.md (tanxf2, tan<mode>2, atan<mode>2, log<mode>2,
85 log10<mode>2, log2<mode>2, expxf2, exp10xf2, exp2xf2): Use op2
86 instead of operands[2] to avoid access past the end of array.
87
88 2007-01-24 Richard Sandiford <richard@codesourcery.com>
89
90 * reload1.c (emit_reload_insns): Pass the reload register
91 for a non-spill output reload through forget_old_reloads_1.
92
93 2007-01-23 Joseph Myers <joseph@codesourcery.com>
94
95 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Define
96 _SOFT_DOUBLE if doubles use software floating-point.
97 * config/rs6000/libgcc-ppc-glibc.ver: Export additional long
98 double functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
99 * config/rs6000/darwin-ldouble.c: Also compile functions for
100 hard-float without FPRs. Use fmsub function for all __NO_FPRS__
101 cases. Compile extra functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
102 * config/rs6000/linuxspe.h (SUBSUBTARGET_OVERRIDE_OPTIONS): Remove
103 commented-out long double override.
104 (CPP_LONGDOUBLE_DEFAULT_SPEC): Likewise.
105 * config/rs6000/eabispe.h: Likewise.
106 * config/rs6000/rs6000.c (rs6000_override_options): Don't override
107 long double for non-SPE.
108 (rs6000_handle_option): Likewise.
109 (invalid_e500_subreg): Disallow more subregs involding DImode,
110 DFmode, TImode or TFmode.
111 (rs6000_legitimate_offset_address_p): Check TFmode offsets for
112 E500 double.
113 (legitimate_lo_sum_address_p): Also check for TFmode for E500
114 double.
115 (rs6000_legitimize_address): Also handle TFmode for E500 double.
116 (rs6000_legitimize_reload_address): Also handle TFmode for E500
117 double.
118 (rs6000_legitimate_address): Also check for TFmode for E500
119 double.
120 (rs6000_emit_move): Use DFmode subregs of TFmode for E500 double.
121 (spe_build_register_parallel): Handle TFmode and TCmode.
122 (rs6000_spe_function_arg): Handle TFmode and TCmode for E500
123 double.
124 (function_arg): Handle TFmode and TCmode for E500 double.
125 (rs6000_init_libfuncs): Initialize extra libfuncs for soft double
126 in general.
127 (print_operand): Handle TFmode and TImode for %y.
128 (rs6000_generate_compare): Handle TFmode comparisons for E500
129 double.
130 (spe_func_has_64bit_regs_p): Check for TFmode for E500 double.
131 (rs6000_function_value): Handle TFmode and TCmode for E500 double.
132 (rs6000_libcall_value): Handle TFmode and TCmode for E500 double.
133 * config/rs6000/rs6000.h (CANNOT_CHANGE_MODE_CLASS): Check for
134 TFmode for E500 double.
135 * config/rs6000/rs6000.md (FP): Allow TF for E500 double.
136 (floatsidf2): Enable for E500 double.
137 (movtf_softfloat): Use rs6000_nonimmediate_operand.
138 (extenddftf2): Change to extenddftf2_fprs.
139 (extenddftf2): Call gen_spe_extenddftf2 or gen_extenddftf2_fprs
140 depending on TARGET_E500_DOUBLE.
141 (extendsftf2): Enable for E500 double.
142 (trunctfdf2): Enable for E500 double.
143 (trunctfsf2): Change to trunctfsf2_fprs.
144 (trunctfsf2): Call gen_spe_trunctfsf2 or gen_trunctfsf2_fprs
145 depending on TARGET_E500_DOUBLE.
146 (floatsitf2): Enable for E500 double.
147 (fix_trunctfsi2): Change to fix_trunctfsi2_fprs.
148 (fix_trunctfsi2): Call gen_spe_fix_trunctfsi2 or
149 gen_fix_trunctfsi2_fprs depending on TARGET_E500_DOUBLE.
150 (negtf2): Change to negtf2_internal.
151 (negtf2): New expander.
152 (abstf2): Enable for E500 double. Call gen_spe_abstf2_tst,
153 gen_spe_abstf2_cmp or gen_abstf2_internal depending on
154 TARGET_E500_DOUBLE and flag_unsafe_math_optimizations.
155 (movdi_internal32): Use rs6000_nonimmediate_operand.
156 (unnamed splitter): Likewise.
157 * config/rs6000/spe.md (CMPTFEQ_GPR, TSTTFEQ_GPR, CMPTFGT_GPR,
158 TSTTFGT_GPR, CMPTFLT_GPR, TSTTFLT_GPR): New unspecs.
159 (SPE64TF, DITI): New mode macros.
160 (frob_df_di): Change to frob_<SPE64:mode>_<DITI:mode>; allow more
161 modes.
162 (frob_tf_ti): New.
163 (frob_<mode>_di_2): New.
164 (frob_tf_di_8_2): New.
165 (frob_di_df): Change to frob_di_<mode>; allow more modes.
166 (frob_ti_tf): New.
167 (frob_di_df_2): Change to frob_<DITI:mode>_<SPE64:mode>_2; allow
168 more modes.
169 (frob_ti_<mode>_8_2): New.
170 (frob_ti_tf_2): New.
171 (mov_si<mode>_e500_subreg0, mov_si<mode>_e500_subreg0_2,
172 mov_si<mode>_e500_subreg4, mov_si<mode>_e500_subreg4_2): Allow
173 TFmode.
174 (mov_sitf_e500_subreg8, mov_sitf_e500_subreg8_2,
175 mov_sitf_e500_subreg12, mov_sitf_e500_subreg12_2): New.
176 (spe_trunctfdf2_internal1, spe_trunctfsf2, spe_extenddftf2,
177 spe_fix_trunctfsi2, spe_fix_trunctfsi2_internal,
178 spe_negtf2_internal, spe_abstf2_cmp, spe_abstf2_tst): New.
179 (cmptfeq_gpr, tsttfeq_gpr, cmptfgt_gpr, tsttfgt_gpr, cmptflt_gpr,
180 tsttflt_gp): New.
181
182 2007-01-23 Ian Lance Taylor <iant@google.com>
183
184 * Makefile.in (OBJS-common): Reformat, alphabetize, but put
185 insn-*.o first.
186 (OBJS-archive): Reformat, alphabetize.
187 (OBJS): Change out_object_file to OBJS-md.
188
189 2007-01-23 Uros Bizjak <ubizjak@gmail.com>
190
191 * config/i386/i386.md: Use REG_P, MEM_P, CONST_INT_P, LABEL_P,
192 JUMP_P and CALL_P predicates where applicable.
193 * config/i386/i386.c: Ditto.
194 * config/i386/i386.md: Ditto.
195 * config/i386/mmx.md: Ditto.
196 * config/i386/predicates.md: Ditto.
197
198 2007-01-22 Andreas Schwab <schwab@suse.de>
199
200 * config/m68k/m68k.h: Fix comment.
201
202 2007-01-22 Jan Hubicka <jh@suse.cz>
203
204 * passes.c (init_optimization_passes): Do not rerun
205 pass_early_warn_uninitialized.
206
207 2007-01-22 Richard Guenther <rguenther@suse.de>
208
209 PR tree-optimization/30038
210 * tree-ssa-math-opts.c (maybe_record_sincos): New static helper
211 function.
212 (execute_cse_sincos_1): Likewise.
213 (execute_cse_sincos): Likewise.
214 (gate_cse_sincos): Likewise.
215 (pass_cse_sincos): New pass CSEing sin() and cos() calls using
216 the cexpi() canonicalization of sincos().
217 * tree-pass.h (pass_cse_sincos): Declare.
218 * passes.c (init_optimization_passes): New pass pas_cse_sincos.
219
220 2007-01-21 Eric Botcazou <ebotcazou@libertysurf.fr>
221
222 PR rtl-optimization/29329
223 * combine.c (replaced_rhs_insn): Rename to i2mod.
224 (replaced_rhs_value): Rename to i2mod_new_rhs.
225 (i2mod_old_rhs): New global variable.
226 (combine_instructions): Adjust for above change. Save a copy of
227 the old RHS into i2mod_old_rhs when the contents of a REG_EQUAL
228 note are substituted in the second instruction.
229 (distribute_notes) <REG_DEAD>: Adjust for above change. Do not
230 ditch the note if it pertains to the second eliminated register
231 and this register is mentioned in i2mod_old_rhs.
232
233 Revert:
234 2006-09-12 Eric Botcazou <ebotcazou@libertysurf.fr>
235
236 * combine.c (distribute_notes) <REG_DEAD>: Do not consider SETs past
237 the insn to which the note was originally attached.
238
239 2007-01-21 Jan Hubicka <jh@suse.cz>
240
241 * ipa-inline.c (inlining_mode): Comment, move up.
242 (cgraph_decide_inlining_incrementally): Do not perform inlining
243 itself; fix handling of flattening of self recursive functions.
244 (cgraph_find_cycles): Remove.
245 (cgraph_flatten_node): Remove.
246 (cgraph_decide_inlining): Use incremental inliner to handle flattening.
247 (try_inline): New function.
248 (cgraph_early_inlining): Update call of
249 cgraph_decide_inlining_incrementally. Apply inlining here.
250 (apply_inline): Update call of cgraph_decide_inlining_incrementally.
251
252 2007-01-21 Dirk Mueller <dmueller@suse.de>
253
254 PR bootstrap/30511
255 * tree-vrp.c (check_array_bounds): do not warn
256 about ADDR_EXPR's of ARRAY_REF's which are immediately
257 used in binary expressions.
258
259 2007-01-21 Ira Rosen <irar@il.ibm.com>
260
261 * tree-vectorizer.h (struct _stmt_vec_info): Add new field
262 read_write_dep and macros for its access.
263 * tree-vectorizer.c (new_stmt_vec_info): Initialize the new field.
264 * tree-vect-analyze.c (vect_analyze_data_ref_dependence): Remove
265 argument, call vect_check_interleaving for every independent pair of
266 data-refs. Mark loads that access the same memory location as a store
267 in the loop.
268 (vect_check_dependences): Remove.
269 (vect_analyze_data_ref_dependences): Remove vect_check_dependences
270 call, fix the call to vect_analyze_data_ref_dependence.
271 (vect_analyze_data_ref_access): For statements that access the same
272 data-ref, check that they are not stores; for loads, check that there
273 is no store that access the same location.
274
275 2007-01-20 Roger Sayle <roger@eyesopen.com>
276 Joseph Myers <joseph@codesourcery.com>
277
278 * doc/invoke.texi (-fdump-rtl-combine): Fix under/overfull hbox.
279 (-fdump-rtl-stack): Likewise.
280 (-fno-signed-zeros): Use @minus{} for a minus sign. Correct typo.
281 (-mcheck-zero-division, -mcpu): Fix under/overfull hbox.
282 (-mpt-fixed): Use @minus{} for minus sign.
283 (Using Precompiled Headers): Fix under/overfull hbox.
284
285 2007-01-20 Jan Hubicka <jh@suse.cz>
286
287 * tree-flow.h (struct stmt_ann_d): Move references_memory to proper
288 place within annotation.
289
290 2007-01-20 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
291
292 * pa.c (output_move_double): Change array size of xoperands to 4.
293
294 2007-01-20 Jan Hubicka <jh@suse.cz>
295
296 * tree-tailcall.c (adjust_return_value): Do not use RESULT_DECL
297 as temporary.
298
299 2007-01-19 Ian Lance Taylor <iant@google.com>
300
301 * expmed.c (expand_divmod) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]: Cast
302 constant to unsigned HOST_WIDE_INT before negating.
303
304 2007-01-19 Ian Lance Taylor <iant@google.com>
305
306 * tree-ssa-operands.h (struct vuse_vec_d): Change num_vuse field
307 to unsigned.
308 (VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Use unsigned comparison.
309 (VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
310 (SET_VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Likewise.
311 (SET_VUSE_ELEMENT_VAR) [ENABLE_CHECKING]: Likewise.
312 (SET_VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
313 (realloc_vdef, realloc_vuse): Change second parameter to
314 unsigned.
315 (ssa_operand_iterator_d): Change vuse_index and mayuse_index
316 fields to unsigned.
317 * tree-ssa-operands.c (realloc_vop): Change num_elem parameter to
318 unsigned. Change x and lim locals to unsigned.
319 (realloc_vdef, realloc_vuse): Change num_elem parameter to
320 unsigned.
321 (finalize_ssa_vuse_ops): Change old_i local to unsigned.
322 (copy_virtual_operands): Change i and n locals to unsigned.
323
324 2007-01-19 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
325
326 PR middle-end/29335
327 * builtins.c (fold_builtin_1): Handle builtin fdim.
328
329 2007-01-20 Jan Hubicka <jh@suse.cz>
330
331 * tree-ssa.c (init_tree_ssa): Do not call init_alias_heapvars.
332 * tree-ssa-structalias.c (compute_points_to_sets): Do call
333 init_alias_heapvars.
334 (init_alias_heapvars): Initialize only when not already initialized.
335 (delete_alias_heapvars): Set heapvar_for_stmt to NULL.
336
337 2007-01-19 Roger Sayle <roger@eyesopen.com>
338
339 * common.opt (fsigned-zeros): New command line option.
340 * flags.h (HONOR_SIGNED_ZEROS): Control via flag_signed_zeros instead
341 of flag_unsafe_math_optimizations.
342 * opts.c (set_fast_math_flags): The -ffast-math command line option
343 implies -fno-signed-zeros.
344 (fast_math_flags_set_p): Likewise.
345
346 * doc/invoke.texi: Document new -fno-signed-zeros option, and update
347 the documentation of -ffast-math appropriately. Wrap long lines.
348
349 2007-01-19 Steve Ellcey <sje@cup.hp.com>
350
351 * system.h (ASM_MAKE_LABEL_LINKONCE): Poison.
352 * varasm.c (globalize_decl): Remove ASM_MAKE_LABEL_LINKONCE ifdef.
353
354 2007-01-19 Tomas Bily <tbily@suse.cz>
355
356 * cgraphunit.c (cgraph_finalize_function): Updating of pid
357 * tree-profile.c:
358 (tree_init_ic_make_global_vars): New function
359 (tree_init_edge_profiler): call of tree_init_ic_make_global_vars
360 (tree_gen_ic_profiler): New function
361 (tree_gen_ic_func_profiler): New function
362 (tree_profiling): Added calling of tree_gen_ic_func_profiler
363 (tree_profile_hooks): Added hook for indirec/virtual calls
364 * value-prof.c (tree_find_values_to_profile): New case for
365 indirect calls
366 (tree_values_to_profile): Call for determining indirect/virtual
367 counters
368 (tree_indirect_call_to_profile): New function
369 (tree_ic_transform): New function
370 (tree_ic): New function
371 (find_func_by_pid): New function
372 (init_pid_map): New function
373 (tree_value_profile_transformations): Added check for
374 indirect/virtual call transformation
375 * value-prof.h (enum hist_type): New counter type for
376 indirect/virtual calls
377 (profile_hooks): Added new hook for profiling indirect/virtual
378 calls
379 * profile.c (instrument_values): New case for indirect/virtual
380 call added
381 * gcov-io.h (GCOV_LAST_VALUE_COUNTER): Changed to 6
382 (GCOV_COUNTER_V_INDIR): New counter type
383 (GCOV_COUNTER_NAMES): New name of counter "indirect" added
384 (GCOV_MERGE_FUNCTIONS): New merge function for indirect/virtual
385 call added
386 * cgraph.c: Definition of cgraph_max_pid
387 (cgraph_create_node): Default init of pid attribute
388 * cgraph.h: Declaration of cgraph_max_pid
389 (struct cgraph_node): Added pid attribute
390 * libgcov.c (__gcov_indirect_call_profiler): New function
391 (__gcov_one_value_profiler_body): New function
392 (__gcov_one_value_profiler): Body was moved to
393 __gcov_one_value_profiler_body and calls it
394
395 2007-01-19 Basile Starynkevitch <basile@starynkevitch.net>
396
397 * doc/gty.texi (Options): Document the mark_hook option to GTY.
398 * gengtype.c (write_types_data, write_func_for_structure,
399 write_types, ggc_wtd, pch_wtd): Add skip_hooks to
400 write_types_data, ggc_wtd, pch_wtd for processing mark_hook.
401 (walk_type, write_func_for_structure): Generate the mark_hook if
402 needed.
403
404 2007-01-19 Jan Hubicka <jh@suse.cz>
405
406 * ipa-inline.c (cgraph_decide_inlining_incrementally): Instead of
407 'early' argument take inlining mode argument specifying whether to
408 inline for size/speeed or all functions; add support for flattening;
409 improve dumpting.
410 (cgraph_early_inlining): Update call of decide_inlining_incrementally.
411
412 2007-01-19 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
413
414 PR c++/17947
415 * toplev.c (warn_deprecated_use): Use %qD instead of %qs to print
416 the name of the declared identifier.
417
418 2007-01-19 Dirk Mueller <dmueller@suse.de>
419
420 * config/i386.h (CONDITIONAL_REGISTER_USAGE): Store
421 result of PIC_OFFSET_TABLE_REGNUM in temporary variable to avoid
422 duplicate evaluation.
423
424 2007-01-19 Uros Bizjak <ubizjak@gmail.com>
425
426 * config/i386/i386.md (acos<mode>2): Rename from acossf2 and acosdf2.
427 Macroize expander using X87MODEF12 mode macro. Extend operand 1
428 to XFMode, use acosxf2 and truncate result to requested mode.
429 Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
430 (asin<mode>2): Similarly, with asin expanders.
431 (*fscalexf4_i387): Rename from *fscalexf4.
432 (expNcorexf3): New expander.
433 (expxf2, exp10xf2, exp2xf2): Use expNcorexf3 expander.
434 (exp<mode>2): Rename from expsf2 and expdf2. Macroize expander using
435 X87MODEF12 mode macro. Extend operand 1 to XFMode, use expxf2 and
436 truncate result to requested mode. Use SSE_FLOAT_MODE_P to disable
437 patterns for SSE math.
438 (exp10<mode>2): Similarly, with exp10 expanders.
439 (exp2<mode>2): Similarly, with exp2 expanders.
440 (expm1<mode>2): Similarly, with expm1 expanders.
441 (ldexp<mode>3): Similarly, with ldexp expanders.
442 (log<mode>2, log10<mode>2, log2<mode>2, log1p<mode>2, logb<mode>2):
443 Use gen_truncxf<mode>2_i387_noop to truncate result.
444
445 2007-01-19 Richard Sandiford <richard@codesourcery.com>
446
447 * config/m68k/m68k.h (PREFERRED_STACK_BOUNDARY): Define to 32
448 for ColdFire targets.
449
450 2007-01-19 Nathan Sidwell <nathan@codesourcery.com>
451 Richard Sandiford <richard@codesourcery.com>
452
453 * config/m68k/m68k.h (M68K_STATIC_CHAIN_REG_NAME): New macro.
454 (INITIALIZE_TRAMPOLINE): Use STATIC_CHAIN_REGNUM.
455 (__transfer_from_trampoline): Use M68K_STATIC_CHAIN_REG_NAME.
456 * config/m68k/m68kelf.h (STATIC_CHAIN_REGNUM): Override.
457 (M68K_STATIC_CHAIN_REG_NAME): Likewise.
458 * config/m68k/netbsd-elf.h (M68K_STATIC_CHAIN_REG_NAME): Likewise.
459
460 2007-01-19 Richard Sandiford <richard@codesourcery.com>
461
462 * config/m68k/m68k.md (adddi_dilshr32): Rename to...
463 (*adddi_dilshr32): ...this. Fix formatting. Remove commented-out
464 non-canonical pattern. Restrict to !TARGET_COLDFIRE.
465 (*adddi_dilshr32_cf): New pattern.
466 (adddi3, subdi3): Remove first alternatives.
467
468 2007-01-19 Richard Sandiford <richard@codesourcery.com>
469
470 * config/m68k/m68k.c (notice_update_cc): If an SFmode move is
471 implemented using move.l, do not use its cc result for floating-point
472 comparisons.
473
474 2007-01-19 Richard Sandiford <richard@codesourcery.com>
475
476 * config/m68k/m68k.h (EXTRA_CONSTRAINT): Stop the 'T' constraint
477 from accepting 's' constraints if flag_pic.
478
479 2007-01-19 Richard Sandiford <richard@codesourcery.com>
480
481 * config/m68k/m68k.md (bordered, bunordered, buneq, bunge, bungt)
482 (bunle, bunlt, bltgt, bordered_rev, bunordered_rev, buneq_rev)
483 (bunge_rev, bungt_rev, bunle_rev, bunlt_rev, bltgt_rev): Change
484 condition from TARGET_68881 to TARGET_HARD_FLOAT.
485
486 2007-01-19 Sandra Loosemore <sandra@codesourcery.com>
487
488 * longlong.h (count_leading_zeros, COUNT_LEADING_ZEROS_0): Add
489 ColdFire alternatives.
490 * config/m68k/m68k.h (CLZ_DEFINED_VALUE_AT_ZERO): New macro.
491 * config/m68k/m68k.md (clzsi2): Define for ColdFire
492 architectures that support ff1 instruction.
493
494 2007-01-19 Richard Sandiford <richard@codesourcery.com>
495 Julian Brown <julian@codesourcery.com>
496
497 * config/m68k/m68k.h (CONST_OK_FOR_LETTER_P): Add an 'R' case.
498 * config/m68k/m68k.md (*movsi_cfv4): Fold into...
499 (*movsi_cf): ...here. Remove unnecessary 'R' from 'Rg'.
500 Add commentary.
501
502 2007-01-19 Richard Sandiford <richard@codesourcery.com>
503
504 * config/m68k/m68k-protos.h (valid_mov3q_const): Take a HOST_WIDE_INT
505 and return a bool.
506 (output_move_const_into_data_reg, output_move_simode_const): Delete.
507 * config/m68k/m68k.c (const_method, const_int_cost): Take a
508 HOST_WIDE_INT instead of an rtx.
509 (m68k_rtx_costs): Update call accordingly.
510 (output_move_const_into_data_reg): Likewise. Fix formatting.
511 (valid_mov3q_const): Take a HOST_WIDE_INT instead of an rtx.
512 Return a bool.
513 (output_move_simode_const): Update calls after above changes.
514 Rework to use automatic variables and predicates like MEM_P.
515 * config/m68k/m68k.md (pushexthisi_const): Update call to
516 valid_mov3q_const.
517
518 2007-01-19 Dirk Mueller <dmueller@suse.de>
519
520 * tree-ssa-alias.c (perform_var_substitution): Fix typo
521 in dump_flags test.
522
523 2007-01-19 Richard Guenther <rguenther@suse.de>
524
525 * builtins.c (expand_builtin_cexpi): Fall back to expanding
526 via cexp in case sincos is not available.
527
528 2007-01-19 Richard Guenther <rguenther@suse.de>
529
530 * doc/tm.texi (TARGET_HAS_SINCOS): Document new target macro.
531 * defaults.h (TARGET_HAS_SINCOS): Default to off.
532 * config/linux.h (TARGET_HAS_SINCOS): Set to on if we have glibc.
533 * config/alpha/linux.h (TARGET_HAS_SINCOS): Likewise.
534 * config/sparc/linux.h (TARGET_HAS_SINCOS): Likewise.
535 * config/sparc/linux64.h (TARGET_HAS_SINCOS): Likewise.
536 * config/rs6000/linux.h (TARGET_HAS_SINCOS): Likewise.
537 * config/rs6000/linux64.h (TARGET_HAS_SINCOS): Likewise.
538
539 2007-01-19 Uros Bizjak <ubizjak@gmail.com>
540
541 * config/i386/i386.md (*fpatanxf3_i387, fpatan_extend<mode>xf3_i387):
542 New insn patterns.
543 (atan2sf3_1, atan2df3_1, atan2xf3_1): Remove insn patterns.
544 (atan2xf3): Directly generate RTL pattern.
545 (atan2<mode>3): Rename from atan2sf3 and atan2df3 and macroize insn
546 patterns using X87MODEF12 mode macro. Use fpatan_extend<mode>xf3_i387
547 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
548 disable patterns for SSE math.
549 (atan<mode>2): Rename from atansf2 and atandf2 and macroize insn
550 patterns using X87MODEF12 mode macro. Use fpatan_extend<mode>xf3_i387
551 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
552 disable patterns for SSE math.
553
554 2007-01-19 Alexandre Oliva <aoliva@redhat.com>
555
556 * libgcc-std.ver: Fix typo in %inherit for GCC_4.3.0.
557
558 2007-01-18 Roger Sayle <roger@eyesopen.com>
559
560 * fold-const.c (fold_unary) <VIEW_CONVERT_EXPR>: Optimize away a
561 VIEW_CONVERT_EXPR to the same type as it's operand.
562
563 2007-01-18 David Edelsohn <edelsohn@gnu.org>
564
565 * config/rs6000/darwin-ldouble.c: Only build _SOFT_FLOAT if
566 configured for long double 128.
567
568 2007-01-18 Mike Stump <mrs@apple.com>
569
570 * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Fix build
571 error.
572
573 2007-01-18 Michael Meissner <michael.meissner@amd.com>
574
575 * i386.c (ix86_compute_frame_layout): Make fprintf's in #if 0 code
576 type correct.
577
578 2007-01-18 Jan Hubicka <jh@suse.cz>
579
580 * tree-ssa-operands.c (vop_free_bucket_size): Never return value
581 greater than NUM_VOP_FREE_BUCKETS.
582
583 2007-01-18 Daniel Berlin <dberlin@dberlin.org>
584
585 * tree-ssa-structalias.c: Update comments.
586 (ptabitmap_obstack): Removed.
587 (pta_obstack): New.
588 (oldpta_obstack): Ditto.
589 (stats): Add a few members.
590 (struct variable_info): Remove node, complex, address_taken, and
591 indirect_target members. Add oldsolution member.
592 (new_var_info): Do not initialize removed members.
593 (constraint_expr_type): Remove INCLUDES.
594 (constraint_graph): Add size, implicit_preds, rep,
595 indirect_cycles, eq_rep, label, direct_nodes, and complex members.
596 (FIRST_REF_NODE): New macro.
597 (LAST_REF_NODE): Ditto.
598 (FIRST_ADDR_NODE): Ditto.
599 (find): New function.
600 (unite): Ditto.
601 (dump_constraint): Do not handle INCLUDES.
602 (insert_into_complex): Do not insert duplicate constraints.
603 (condense_varmap_nodes): Renamed and rewritten into ...
604 (merge_node_constraints): This. Also fix bug in handling of
605 offseted copy constraints.
606 (clear_edges_for_node): No longer need to deal with preds at all,
607 or removing associated preds/succs.
608 (merge_graph_nodes): Deal with indirect_cycles.
609 Don't deal with predecessors.
610 (add_implicit_graph_edge): New function.
611 (add_pred_graph_edge): Ditto.
612 (add_graph_edge): Don't deal with predecessors.
613 (build_constraint_graph): Removed.
614 (build_pred_graph): New function.
615 (build_succ_graph): Ditto.
616 (struct scc_info): Removed in_component. Added roots, dfs, and
617 node_mapping. Remove visited_index, unification_queue.
618 (scc_visit): Deal with union-find we do now.
619 Deal with cycles with REF nodes.
620 (collapse_nodes): Renamed and rewritten to ...
621 (unify_nodes): This.
622 (process_unification_queue): Removed.
623 (topo_visit): Cleanup
624 (do_da_constraint): Use find.
625 (do_sd_constraint): Ditto.
626 (do_ds_constraint): Ditto.
627 (do_complex_constraint): Ditto.
628 (init_scc_info): Update for removed and added members.
629 (find_and_collapse_graph_cycles): Renamed and rewritten into ...
630 (find_indirect_cycles): This.
631 (equivalence_class): New variable.
632 (label_visit): New function.
633 (perform_variable_substitution): Rewritten.
634 (free_var_substitution_info): New function.
635 (find_equivalent_node): Ditto.
636 (move_complex_constraints): Ditto.
637 (eliminate_indirect_cycles): Ditto.
638 (solve_graph): Only propagate changed bits.
639 Use indirect cycle elimination.
640 Use find.
641 (tree_id_t): Rename to tree_vi_t, delete id member, add vi member.
642 (tree_id_eq): Renamed to ...
643 (tree_vi_eq): This. Update for member change
644 (insert_id_for_tree): Renamed and rewritten to ...
645 (insert_vi_for_tree): This.
646 (lookup_id_for_tree): Renamed and rewritten to ...
647 (lookup_vi_for_tree): This.
648 (get_id_for_tree): Renamed and rewritten to ...
649 (get_vi_for_tree): Ditto.
650 (get_constraint_exp_from_ssa_var): Update to use get_vi_for_tree.
651 (process_constraint): Don't handle INCLUDES.
652 Remove special ADDRESSOF case.
653 (find_func_aliases): Rewrite to use vi functions instead of id
654 ones.
655 (create_function_info_for): Ditto.
656 (create_variable_info_for): Ditto.
657 (intra_create_variable_infos): Ditto.
658 (merge_smts_into): Ditto.
659 (find_what_p_points_to): Ditto.
660 (init_base_vars): Ditto.
661 (init_alias_vars): Ditto.
662 (remove_preds_and_fake_succs): New function.
663 (dump_sa_points_to_info): Dump new stats.
664 (dump_solution_for_var): Use find.
665 (set_used_smts): Fix formatting.
666 (compute_points_to_sets): Updated for new functions.
667 (ipa_pta_execute): Ditto.
668
669 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
670 Richard Sandiford <richard@codesourcery.com>
671
672 * doc/tm.texi (TARGET_FUNCTION_VALUE): Expand documentation of
673 parallels.
674 * calls.c (expand_call): If the return value is a PARALLEL,
675 extract its first member.
676 * config/m68k/linux.h (FUNCTION_EXTRA_EPILOGUE): Remove.
677 * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
678 use FUNCTION_EXTRA_EPILOGUE.
679 (m68k_function_value): Return a PARALLEL if the return value
680 is of a pointer type.
681 * config/m68k/netbsd-elf.h (current_function_returns_pointer)
682 (FUNCTION_EXTRA_EPILOGUE): Remove.
683 * config/m68k/m68k.md (D0_REG): New constant.
684
685 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
686
687 * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
688 output a NOP for empty epilogues.
689
690 2007-01-18 Richard Sandiford <richard@codesourcery.com>
691
692 * config/m68k/m68k.c (m68k_use_return_insn): Update comments
693 before function. Extend register save check to include all
694 registers, not just integer ones.
695
696 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
697
698 * config/m68k/m68k-protos.h (use_return_insn): Rename to...
699 (m68k_use_return_insn): ...this.
700 * config/m68k/m68k.h (USE_RETURN_INSN): Delete.
701 * config/m68k/m68k.c (use_return_insn): Rename to...
702 (m68k_use_return_insn): ...this.
703 * config/m68k/m68k.md (return): Use m68k_use_return_insn instead
704 of USE_RETURN_INSN.
705
706 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
707
708 * config/m68k/fpgnulib.c (__truncdfsf2): Implement round to
709 nearest even, fix denormal rounding overflow.
710
711 2007-01-18 Richard Sandiford <richard@codesourcery.com>
712
713 * config/m68k/m68k.md (movsf_cf_hard): Use fsmove instead of
714 f%$smove and f%$move.
715 (movdf_cf_hard): Use fdmove for cases 0 and 3 and fmove for case 1.
716 (extendsfdf2_cf): Use fdmove instead of f%&move.
717 (truncdfsf2_cf): Use fsmove instead of f%$smove.
718 (add<mode>3_cf, sub<mode>3_cf): Use <FP:prec> instead of <FP:round>.
719
720 2007-01-18 Richard Sandiford <richard@codesourcery.com>
721
722 * config/m68k/m68k.md (movdf_cf_hard): Use output_move_double for
723 GPR<-GPR moves.
724
725 2007-01-18 Richard Sandiford <richard@codesourcery.com>
726
727 * real.h (real_format): Add a canonical_nan_lsbs_set field.
728 (coldfire_single_format): Declare.
729 (coldfire_double_format): Likewise.
730 * real.c (encode_ieee_single): Use canonical_nan_lsbs_set instead
731 of qnan_msb_set to determine the lower bits of a canonical
732 NaN significand.
733 (encode_ieee_double): Likewise.
734 (encode_ieee_quad): Likewise.
735 (ieee_single_format): Initialize canonical_nan_lsbs_set.
736 (mips_single_format): Likewise.
737 (ieee_double_format): Likewise.
738 (mips_double_format): Likewise.
739 (ieee_extended_motorola_format): Likewise.
740 (ieee_extended_intel_96_format): Likewise.
741 (ieee_extended_intel_128_format): Likewise.
742 (ieee_extended_intel_96_round_53_format): Likewise.
743 (ibm_extended_format): Likewise.
744 (mips_extended_format): Likewise.
745 (ieee_quad_format): Likewise.
746 (mips_quad_format): Likewise.
747 (vax_f_format): Likewise.
748 (vax_d_format): Likewise.
749 (vax_g_format): Likewise.
750 (i370_single_format): Likewise.
751 (i370_double_format): Likewise.
752 (decimal_single_format): Likewise.
753 (decimal_double_format): Likewise.
754 (decimal_quad_format): Likewise.
755 (c4x_single_format): Likewise.
756 (c4x_extended_format): Likewise.
757 (real_internal_format): Likewise.
758 (coldfire_single_format): New real_format.
759 (coldfire_double_format): Likewise.
760 * config/pdp11/pdp11.c (pdp11_f_format): Initialize
761 canonical_nan_lsbs_set.
762 (pdp11_d_format): Likewise.
763 * config/m68k/m68k.c (override_options): Override REAL_FORMAT_MODE
764 if TARGET_COLDFIRE_CPU.
765
766 2007-01-18 Richard Sandiford <richard@codesourcery.com>
767
768 * config/m68k/m68k-protos.h (m68k_output_pic_call): Delete.
769 (output_call, m68k_legitimize_call_address): Declare.
770 * config/m68k/m68k.h (EXTRA_CONSTRAINT): Remove unnecessary
771 parenthesees. Add support for a 'W' constraint.
772 (LEGITIMATE_PIC_OPERAND_P): Remove SYMBOL_REF_FLAG handling.
773 (PRINT_OPERAND_PUNCT_VALID_P): Remove comment about 'o'.
774 (m68k_symbolic_call, m68k_symbolic_jump): Declare.
775 * config/m68k/m68k.c (m68k_symbolic_call, m68k_symbolic_jump): New
776 variables.
777 (override_options): Initialize them. Do not set flag_no_function_cse
778 for TARGET_ID_SHARED_LIBRARY.
779 (m68k_output_pic_call): Delete.
780 (m68k_legitimize_call_address): New function.
781 (print_operand): Remove the %o prefix. Handle the %p prefix.
782 (output_call): New function.
783 (m68k_output_mi_thunk): Use m68k_symbolic_jump. Always load the
784 target address from the GOT if symbolic jumps are not allowed.
785 * config/m68k/m68k.md (call, general_operand): Do not set
786 SYMBOL_REF_FLAG. Use m68k_legitimize_call_address instead.
787 Merge separate flag_pic and !flag_pic define_insns into...
788 (*call, *call_value): ...these new patterns. Match the address
789 rather than the containing MEM and require it to be a call_operand.
790 Use output_call to generate the asm template.
791 * config/m68k/predicates.md (const_call_operand): New predicate.
792 (call_operand): Likewise.
793
794 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
795
796 * config/m68k/m68k.h (REGISTER_MOVE_COST): Simplify definition.
797 (STACK_GROWS_DOWNWARD): Define to 1.
798 (FUNCTION_VALUE, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P): Equivocate
799 comments, emphasizing that these values are only defaults.
800 * config/m68k/linux.h (LINK_SPEC): Fix formatting in #undef.
801 * config/m68k/m68k.c (const_method): Remove trailing whitespace.
802
803 2007-01-18 Richard Sandiford <richard@codesourcery.com>
804
805 * config/m68k/m68k.md (cmpsi): Remove outdated flag_pic handling.
806
807 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
808 Richard Sandiford <richard@codesourcery.com>
809
810 * config/m68k/m68k.h (DATA_REGNO_P, ADDRESS_REGNO_P, INT_REGNO_P)
811 (FP_REGNO_P): New macros.
812 (REGNO_OK_FOR_INDEX_P, REGNO_OK_FOR_BASE_P, REGNO_OK_FOR_DATA_P)
813 (REGNO_OK_FOR_FP_P, REG_OK_FOR_INDEX_P, REG_OK_FOR_BASE_P): Use them.
814 (EH_RETURN_STACKADJ_RTX): Use A0_REG.
815 * config/m68k/m68k.c (m68k_regno_mode_ok): Use the new REGNO macros.
816
817 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
818
819 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
820 (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-rtems*): Add t-floatlib
821 to $tmake_file.
822 * config/m68k/t-floatlib: New file, extracting common code from...
823 * config/m68k/t-m68kbare, config/m68k/t-m68kelf,
824 * config/m68k/t-uclinux: Here.
825 * config/m68k/fpgnulib.c: Do not compile extendeed precision
826 routines on ColdFire targets.
827
828 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
829
830 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
831 (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*)
832 (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-linux*)
833 (m68k-*-rtems*): Use tm_file rather than m68k/m68k.h and
834 explicitly set MOTOROLA to 1.
835 * config/m68k/m68k.h (MOTOROLA): Simplify definition accordingly.
836
837 2007-01-18 Richard Sandiford <richard@codesourcery.com>
838 Nathan Sidwell <nathan@codesourcery.com>
839
840 * config/m68k/m68k.h (PCC_STATIC_STRUCT_RETURN): Delete.
841 (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Add MOTOROLA cases.
842 * config/m68k/coff.h (REGISTER_PREFIX_MD): Delete.
843 * config/m68k/m68020-elf.h (LIB_SPEC): Delete.
844 * config/m68k/m68k-none.h (CC1_SPEC, CPP_SUBTARGET_SPEC): Delete.
845 * config/m68k/m68kelf.h (IMMEDIATE_PREFIX, REGISTER_PREFIX_MD)
846 (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Delete.
847 (NO_DOLLAR_IN_LABEL, PCC_STATIC_STRUCT_RETURN): Don't undefine.
848 (BSS_ASM_OP, ASM_OUTPUT_SKIP, ASM_OUTPUT_ASCII): Delete.
849 * config/m68k/m68kemb.h (PCC_STATIC_STRUCT_RETURN): Don't undefine.
850 * config/m68k/linux.h (SIZE_TYPE, PTRDIFF_TYPE, WCHAR_TYPE)
851 (WCHAR_TYPE_SIZE, TARGET_OBJFMT_CPP_BUILTINS): Delete.
852 (TARGET_OS_CPP_BUILTINS): Don't define mc68000 and mc68020 here.
853 (DBX_REGISTER_NUMBER): Delete.
854 * config/m68k/netbsd-elf.h (IMMEDIATE_PREFIX): Delete.
855 (PCC_STATIC_STRUCT_RETURN): Don't undefine.
856 * config/m68k/openbsd.h (PCC_STATIC_STRUCT_RETURN): Define.
857
858 2007-01-18 Richard Sandiford <richard@codesourcery.com>
859
860 * config.gcc (m68k-*-uclinux*): Add flat.h to $tm_file.
861 * config/flat.h: New file.
862 * crtstuff.c (USE_PT_GNU_EH_FRAME): Don't define if
863 OBJECT_FORMAT_FLAT.
864 * config/m68k/m68k.h (ASM_PREFERRED_EH_DATA_FORMAT): Do not use
865 indirect references for -msep-data or -mid-shared-library.
866 Do not use PC-relative code addresses either.
867 * config/m68k/m68k.c (override_options): Restrict -fPIC error
868 to -mpcrel.
869 * config/m68k/uclinux.h (STARTFILE_SPEC): Define. Use Scrt1.o
870 for shared libraries and crt1.o for executables. Use crti.o and
871 crtbegin.o.
872 (ENDFILE_SPEC): Use crtend.o and crtn.o.
873 (LIB_SPEC): Suppress -Rlibc.gdb if -static-libc is given.
874 Do not add -elf2flt or -shared-lib-id options here.
875 (LINK_SPEC): Define. Pass -elf2flt if no -elf2flt option is given.
876 Pass -shared-lib-id if -mid-shared-library, taking the library
877 identifier from -mshared-library-id if given, otherwise
878 defaulting to 0.
879 (EH_FRAME_IN_DATA_SECTION): Do not undefine.
880 (INIT_SECTION_ASM_OP, FINI_SECTION_ASM_OP): Likewise.
881 (TARGET_OS_CPP_BUILTINS): Define __GXX_MERGED_TYPEINFO_NAMES=0
882 and __GXX_TYPEINFO_EQUALITY_INLINE=0 if -mid-shared-library.
883 (DRIVER_SELF_SPECS): Map unadorned PIC options to -msep-data.
884 * config/m68k/t-uclinux (EXTRA_MULTILIB_PARTS): Add crtbegin.o
885 and crtend.o.
886 * config/m68k/lb1sf68.asm (PICCALL): Use an lea and pc-relative
887 jump sequence for ISA A and ISA A+.
888 (PICJUMP): Likewise.
889
890 2007-01-18 Richard Sandiford <richard@codesourcery.com>
891
892 * config.gcc (m68k-*-uclinux*): Base the port on the common
893 and m68k GNU/Linux files rather than on the generic ELF ones.
894 * config/m68k/uclinux.h (TARGET_VERSION): Override.
895 (TARGET_OS_CPP_BUILTINS): Use LINUX_TARGET_OS_CPP_BUILTINS.
896
897 2007-01-18 Julian Brown <julian@codesourcery.com>
898 Richard Sandiford <richard@codesourcery.com>
899
900 * config/m68k/m68k.h (LONG_DOUBLE_TYPE_SIZE): Make 64-bit on ColdFire.
901 (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
902 * config/m68k/netbsd-elf.h (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Undefine
903 before redefining.
904 * config/m68k/uclinux-oldabi.h (LONG_DOUBLE_TYPE_SIZE): Redefine to
905 80 unconditionally.
906 (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
907
908 2007-01-18 Richard Sandiford <richard@codesourcery.com>
909
910 * doc/install.texi: Document m68k-uclinuxoldabi.
911 * config.gcc (m68k-*-uclinuxoldabi): New configuration.
912 * config/m68k/uclinux-oldabi.h: New file, copied from
913 config/m68k/uclinux.h.
914
915 2007-01-18 Kaz Kojima <kkojima@rr.iij4u.or.jp>
916
917 * config/m32r/m32r-protos.h (m32r_expand_epilogue): Declare it.
918 * config/m32r/m32r.c (m32r_setup_incoming_varargs): Use gen_frame_mem.
919 (m32r_compute_frame_size): Use unsigned for regno.
920 (m32r_reload_lr): Use gen_frame_mem.
921 (pop): New.
922 (m32r_output_function_epilogue): Don't output the function epilogue
923 textually here.
924 (m32r_expand_epilogue): New.
925 (direct_return): Return false if the function has the interrupt
926 attribute.
927 (m32r_hard_regno_rename_ok): Remove code for the textual epilogue.
928 * config/m32r/m32r.md (epilogue): New expander.
929 (return_lr, return_rte): New insns.
930 (return): Make it expander.
931 (return_normal): New expander.
932
933 2007-01-18 Josh Conner <jconner@apple.com>
934
935 PR target/30485
936 * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Add
937 support for UNLE, UNLT, UNGE, and UNGT.
938
939 2007-01-18 Jan Hubicka <jh@suse.cz>
940
941 * tree-vrp.c (finalize_jump_threads): Do not call cleanup_cfg by hand.
942 * tree-tailcall (add_virtual_phis): Likewise.
943 (optimize_tail_call): Return TODOs.
944 (execute_tail_calls): Return TODOs.
945 * tree-ssa-ccp (execute_fold_all_builtins): Do cleanup_cfg via TODO.
946 * tree-cfgcleanup.c (cleanup_tree_cfg_loop): Return if something
947 changed.
948 * tree-ssa-forwprop.c (tree_ssa_forward_propagate_single_use_value):
949 Cleanup cfg using TODO.
950 * tree-flow.h (cleanup_tree_cfg_loop): Update prototype.
951 * passes.c (execute_function_todo): When cleanup did something, remove
952 unused locals.
953 * tree-cfg.c (pass_build_cfg): Add cleanup_cfg TODO.
954 (make_edges): Don't cleanup_cfg.
955
956 2007-01-18 Uros Bizjak <ubizjak@gmail.com>
957
958 * reg-stack.c (subst_stack_regs_pat) [UNSPEC_SINCOS_COS,
959 UNSPEC_XTRACT_FRACT]: Use generic code for instructions that
960 operate on the top of stack.
961 [UNSPEC_SINCOS_SIN, UNSPEC_XTRACT_EXP, UNSPEC_TAN]: Rewrite
962 register handling of instructions that output to the second
963 stack slot.
964 [UNSPEC_TAN_ONE, UNSPEC_TAN_TAN]: Remove.
965 (move_for_stack_reg): Special-case check for dead destination
966 stack slot for constant load of 1.0 inside UNSPEC_TAN.
967
968 * config/i386/i386.md (UNSPEC_TAN): New constant.
969 (UNSPEC_TAN_ONE, UNSPEC_TAN_TAN): Remove.
970 (fptanxf4_i387, fptan_extend<mode>xf4_i387): New patterns
971 to correctly model move of constant 1.0 to top stack slot.
972 (*tandf3_1, *tansf3_1, *tanxf3_1): Remove insn patterns.
973 (unnamed peephole2 pattern): Remove corresponding peephole2
974 pattern that optimizes tan insn and loading of constant 1.0.
975 (tanxf2): Use fptanxf4_i387.
976 (tan<mode>2): Rename from tansf2 and tandf2 and macroize insn
977 patterns using X87MODEF12 mode macro. Use fptan_extend<mode>xf4_i387
978 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
979 disable patterns for SSE math.
980 (sincos<mode>3): Use truncxf<mode>2_i387_noop for truncation.
981 (fyl2x_extend<mode>xf3_i387): Use X87MODEF12 for operand 1.
982
983 2007-01-18 Dirk Mueller <dmueller@suse.de>
984 Richard Guenther <rguenther@suse.de>
985
986 PR diagnostic/8268
987 * doc/invoke.texi (Warray-bounds): Document -Warray-bounds.
988 * common.opt (Warray-bounds): Add new warning option.
989 * c-opts.c (c_common_handle_option): Define -Warray-bounds
990 if -Wall is given.
991 * Makefile.in: make tree-vrp.o depend on toplev.h
992 * tree-vrp.c (vrp_finalize): Call check_array_refs if -Warray-bounds
993 is enabled.
994 (check_array_refs, check_array_bounds, check_array_ref): New.
995
996 2007-01-18 Jan Hubicka <jh@suse.cz>
997
998 * tree-ssa-ccp.c (ccp_finalize): Return if something changed.
999 (execute_ssa_ccp): Return flags conditionally.
1000 * tree-ssa-propagate.c (substitue_and_fold): Return if something was
1001 changed.
1002 * tree-ssa-propagate.h (substitute_and_fold): Update prototype.
1003
1004 2007-01-18 Steven Bosscher <steven@gcc.gnu.org>
1005
1006 * cfgcleanup.c (cleanup_cfg): Detect cfglayout mode and set
1007 the CLEANUP_CFGLAYOUT flag when in cfglayout mode.
1008
1009 * Makefile.c (GTFILES): Add cfglayout.h.
1010 * gengtype.c (open_base_files): Likewise.
1011 * cfglayout.c (cfg_layout_function_footer,
1012 cfg_layout_function_header) Reindent to make gengtype happy.
1013 * cfglayout.h (cfg_layout_function_footer,
1014 cfg_layout_function_header): Add GTY(()) marker.
1015
1016 * ifcvt.c (noce_try_sign_mask): Make sure INSN_B is non-null.
1017
1018 2007-01-18 Ben Elliston <bje@au.ibm.com>
1019
1020 * genautomata.c (write_automata): Include xstrerror output in the
1021 error message if writing the DFA description file fails.
1022
1023 2007-01-17 H.J. Lu <hongjiu.lu@intel.com>
1024
1025 * config/mips/mips-protos.h (mips_output_external): Make it
1026 return void.
1027 * config/mips/iris.h (TARGET_ASM_EXTERNAL_LIBCALL): Removed.
1028 * config/mips/mips.c (irix_output_external_libcall): Likewise.
1029 (extern_list): Likewise.
1030 (extern_head): Likewise.
1031 (TARGET_ASM_FILE_END): Likewise.
1032 (mips_file_end): Likewise.
1033 (mips_output_external): Rewritten.
1034
1035 2007-01-18 Ben Elliston <bje@au.ibm.com>
1036
1037 * genpreds.c (write_insn_preds_c): Only write out the function
1038 body for regclass_for_constraint if we have register constraints.
1039
1040 2007-01-17 Tom Tromey <tromey@redhat.com>
1041
1042 * doc/sourcebuild.texi (libgcj Tests): Use sourceware.org.
1043 * doc/install.texi (Testing): Use sourceware.org.
1044 (Binaries): Likewise.
1045 (Specific): Likewise.
1046 * doc/contrib.texi (Contributors): Use sourceware.org.
1047
1048 2007-01-17 Anatoly Sokolov <aesok@post.ru>
1049
1050 * config/avr/avr.h (AVR_HAVE_LPMX): New macro.
1051 (AVR_ENHANCED): Rename to ...
1052 (AVR_HAVE_MUL): ... new.
1053 (avr_enhanced_p): Rename to ...
1054 (avr_have_mul_p): ... new.
1055 (TARGET_CPU_CPP_BUILTINS): Use 'avr_have_mul_p' instead of
1056 'avr_enhanced_p' for "__AVR_ENHANCED__". Define "__AVR_HAVE_MUL__".
1057 * config/avr/avr.c (avr_enhanced_p): Rename to ...
1058 (avr_have_mul_p): ... new.
1059 (base_arch_s): Rename 'enhanced' to 'have_mul'.
1060 (avr_override_options): Use 'avr_have_mul_p' and 'have_mul' instead of
1061 'avr_enhanced_p' and 'enhanced'.
1062 (ashlhi3_out, ashrhi3_out, lshrhi3_out, avr_rtx_costs): Use
1063 AVR_HAVE_MUL instead of AVR_ENHANCED.
1064 * avr.md (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
1065 (mulqi3, *mulqi3_enh, *mulqi3_call, mulqihi3, umulqihi3, mulhi3,
1066 *mulhi3_enh, *mulhi3_call, mulsi3, *mulsi3_call): Use AVR_HAVE_MUL
1067 instead of AVR_ENHANCED.
1068 (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
1069 * libgcc.S: Use __AVR_HAVE_MUL__ instead of __AVR_ENHANCED__.
1070 (__tablejump__): Use __AVR_HAVE_LPMX__ instead of __AVR_ENHANCED__.
1071
1072 2007-01-17 Ian Lance Taylor <iant@google.com>
1073
1074 * vec.h (VEC_reserve_exact): Define.
1075 (vec_gc_p_reserve_exact): Declare.
1076 (vec_gc_o_reserve_exact): Declare.
1077 (vec_heap_p_reserve_exact): Declare.
1078 (vec_heap_o_reserve_exact): Declare.
1079 (VEC_OP (T,A,reserve_exact)): New static inline function, three
1080 versions.
1081 (VEC_OP (T,A,reserve)) [all versions]: Remove handling of
1082 negative parameter.
1083 (VEC_OP (T,A,alloc)) [all versions]: Call ...reserve_exact.
1084 (VEC_OP (T,A,copy)) [all versions]: Likewise.
1085 (VEC_OP (T,a,safe_grow)) [all versions]: Likewise.
1086 * vec.c (calculate_allocation): Add exact parameter. Change all
1087 callers.
1088 (vec_gc_o_reserve_1): New static function, from vec_gc_o_reserve.
1089 (vec_gc_p_reserve, vec_gc_o_reserve): Call vec_gc_o_reserve_1.
1090 (vec_gc_p_reserve_exact, vec_gc_o_reserve_exact): New functions.
1091 (vec_heap_o_reserve_1): New static function, from vec_heap_o_reserve.
1092 (vec_heap_p_reserve, vec_heap_o_reserve): Call vec_heap_o_reserve_1.
1093 (vec_heap_p_reserve_exact): New function.
1094 (vec_heap_o_reserve_exact): New function.
1095
1096 2007-01-17 Jan Hubicka <jh@suse.cz>
1097
1098 * ipa-type-escape.c (look_for_casts): Revamp using handled_component_p.
1099
1100 2007-01-17 Eric Christopher <echristo@apple.com>
1101
1102 * config.gcc: Support core2 processor.
1103
1104 2007-01-16 Jan Hubicka <jh@suse.cz>
1105
1106 * tree-ssanames.c (release_dead_ssa_names): Instead of ggc_freeing
1107 the names, just unlink the chain so we don't crash on dangling pointers
1108 to dead SSA names.
1109
1110 2007-01-16 Jan Hubicka <jh@suse.cz>
1111
1112 * cgraph.h (cgraph_decide_inlining_incrementally): Kill.
1113 * tree-pass.h: Reorder to make IPA passes appear toegher.
1114 (pass_early_inline, pass_inline_parameters, pass_apply_inline): Declare.
1115 * cgraphunit.c (cgraph_finalize_function): Do not compute inling
1116 parameters, do not call early inliner.
1117 * ipa-inline.c: Update comments. Include tree-flow.h
1118 (cgraph_decide_inlining): Do not compute inlining parameters.
1119 (cgraph_decide_inlining_incrementally): Return TODOs; assume to
1120 be called with function context set up.
1121 (pass_ipa_inline): Remove unreachable functions before pass.
1122 (cgraph_early_inlining): Simplify assuming to be called from the
1123 PM as local pass.
1124 (pass_early_inline): New pass.
1125 (cgraph_gate_ipa_early_inlining): New gate.
1126 (pass_ipa_early_inline): Turn into simple wrapper.
1127 (compute_inline_parameters): New function.
1128 (gate_inline_passes): New gate.
1129 (pass_inline_parameters): New pass.
1130 (apply_inline): Move here from tree-optimize.c
1131 (pass_apply_inline): New pass.
1132 * ipa.c (cgraph_remove_unreachable_nodes): Verify cgraph after
1133 transforming.
1134 * tree-inline.c (optimize_inline_calls): Return TODOs rather than
1135 doing them by hand.
1136 (tree_function_versioning): Do not allocate dummy struct function.
1137 * tree-inline.h (optimize_inline_calls): Update prototype.
1138 * tree-optimize.c (execute_fixup_cfg): Export.
1139 (pass_fixup_cfg): Remove
1140 (tree_rest_of_compilation): Do not apply inlines.
1141 * tree-flow.h (execute_fixup_cfg): Declare.
1142 * Makefile.in (gt-passes.c): New.
1143 * passes.c: Include gt-passes.h
1144 (init_optimization_passes): New passes.
1145 (nnodes, order): New static vars.
1146 (do_per_function_toporder): New function.
1147 (execute_one_pass): Dump current pass here.
1148 (execute_ipa_pass_list): Don't dump current pass here.
1149
1150 2007-01-16 Janis Johnson <janis187@us.ibm.com>
1151
1152 * config/dfp-bit.c (dfp_compare_op): Return separate value for NaN.
1153 (DFP_NE, DFP_LE, DFP_GE): Return false for NaN.
1154
1155 2007-01-16 David Edelsohn <edelsohn@gnu.org>
1156
1157 * config/rs6000/darwin-ldouble.c: Build file for SOFT_FLOAT.
1158 (strong_alias): Define.
1159 (__gcc_qmul): Provide non-FMA for soft-float.
1160 (__gcc_qdiv): Same.
1161 (__gcc_qneg): New.
1162 (__gcc_qeq): New.
1163 (__gcc_qle): New.
1164 (__gcc_qge): New.
1165 (__gcc_qunord): New.
1166 (__gcc_stoq): New.
1167 (__gcc_dtoq): New.
1168 (__gcc_qtos): New.
1169 (__gcc_qtod): New.
1170 (__gcc_qtoi): New.
1171 (__gcc_qtou): New.
1172 (__gcc_itoq): New.
1173 (__gcc_utoq): New.
1174 (fmsub): New.
1175 * config/rs6000/rs6000.c (rs6000_init_libfuncs): Initialize
1176 soft-float functions.
1177 * config/rs6000/libgcc-ppc-glibc.ver: Version soft-float symbols.
1178 * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Do not warn
1179 about long double soft float.
1180
1181 2007-01-16 Dorit Nuzman <dorit@il.ibm.com>
1182 Tehila Meyzels <tehila@il.ibm.com>
1183
1184 * tree-vectorizer.h (is_pattern_stmt_p): New.
1185 * tree-vect-analyze.c (vect_determine_vectorization_factor): Fix
1186 formatting (tabs instead of spaces). Cleanup and clarify setting
1187 of STMT_VINFO_VECTYPE. Call is_pattern_stmt_p.
1188 * tree-vect-transform.c (vect_get_vec_def_for_operand): Fix typo.
1189 (vectorizable_type_demotion): Check that types are integral.
1190 (vectorizable_type_promotion): Likewise.
1191 (vectorizable_store): Fix typo. Eliminate new-line at end of
1192 comments.
1193
1194 2007-01-16 Jan Hubicka <jh@suse.cz>
1195
1196 * tree-ssanames.c (release_dead_ssa_names): Remove invalidated
1197 cgraph edges too.
1198
1199 2007-01-15 Eric Christopher <echristo@apple.com>
1200
1201 * ifcvt.c: Include vec.h, vecprim.h.
1202 (check_cond_move_block): New argument regs.
1203 Reorganize. Add registers used to regs.
1204 (cond_move_process_if_block): Use regs set above as
1205 loop bounds.
1206
1207 2007-01-15 Eric Christopher <echristo@apple.com>
1208
1209 * config/darwin.h: Update copyright.
1210 (TARGET_OPTION_TRANSLATE_TABLE): Add umbrella.
1211 (LINK_COMMAND_SPEC): Add -u.
1212 (LINK_SPEC): Fix umbrella for above.
1213
1214 2007-01-15 Joseph S. Myers <joseph@codesourcery.com>
1215
1216 * config/soft-fp/op-common.h, config/soft-fp/op-4.h: Update from
1217 glibc CVS.
1218
1219 2007-01-15 Tom Tromey <tromey@redhat.com>
1220
1221 * doc/sourcebuild.texi (libgcj Tests): Don't mention jacks.
1222 * doc/install.texi (Testing): Don't mention jacks.
1223 (Configuration): Document --enable-java-maintainer-mode. Move
1224 --with-java-home to libgcj-specific section. Document
1225 --with-ecj-jar.
1226 (Prerequisites): Mention --enable-java-maintainer-mode, ecj1.
1227
1228 2007-01-15 Jan Hubicka <jh@suse.cz>
1229
1230 * tree-ssa-dce.c (DCE_TODOs): New.
1231 (propagate_necessity): Return if something changed.
1232 (eliminate_unnecessary_stmts): Likewise.
1233 (perform_tree_ssa_dce): Return TODO flags when needed.
1234 (pass_dce, pass_dce_loop, pass_cd_dce): Remove TODO flags.
1235
1236 2007-01-15 Uros Bizjak <ubizjak@gmail.com>
1237
1238 * config/i386/i386.md (fyl2xxf3_i387): Rename from fyl2x_xf3.
1239 (fyl2x_extend<mode>xf3_i387): New insn pattern.
1240 (log<mode>2): Rename from logsf2 and logdf2 and macroize insn
1241 patterns using X87MODEF12 mode macro. Extend operand 1
1242 to XFmode. Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
1243 (log10<mode>2): Ditto.
1244 (log2<mode>2): Ditto.
1245 (log1p<mode>2): Ditto.
1246 (logb<mode>2): Ditto.
1247 (fyl2xp1xf3_i387): Rename from fyl2xp1_xf3.
1248 (fyl2xp1_extend<mode>xf3_i387): New insn pattern.
1249 (*fxtractxf3_i387): Rename from *fxtractxf3.
1250 (fxtract_extend<mode>xf3_i387): New insn pattern.
1251 (ilogbsi2): Use match_dup 3, not match_operand:XF 3.
1252 * config/i386/i386.c (ix86_emit_i387_log1p): Use gen_fyl2xp1xf3_i387()
1253 and gen_fyl2xxf3_i387().
1254
1255 2007-01-14 Zdenek Dvorak <dvorakz@suse.cz>
1256
1257 * loop-unswitch.c (unswitch_loop): Do not call fix_loop_placement.
1258 * cfgloopmanip.c (fix_loop_placement): Made static. Use
1259 get_loop_exit_edges. Changed return type to bool.
1260 * cfgloop.h (fix_loop_placement): Declaration removed.
1261
1262 2007-01-14 Dorit Nuzman <dorit@il.ibm.com>
1263
1264 * param.h (MIN_VECT_LOOP_BOUND): New.
1265 * params.def (MIN_VECT_LOOP_BOUND): New.
1266 * tree-vectorizer.c (slpeel_tree_peel_loop_to_edge): Takes another
1267 argument - minimum threshold for number of iterations.
1268 * tree-vectorizer.h (slpeel_tree_peel_loop_to_edge): Add another
1269 argument to declaration.
1270 * tree-vect-analyze.c (vect_analyze_operations): Check value of
1271 MIN_VECT_LOOP_BOUND.
1272 * tree-vect-transform.c (vect_do_peeling_for_loop_bound): Call
1273 slpeel_tree_peel_loop_to_edge with additional argument.
1274 (vect_do_peeling_for_alignment): Likewise.
1275 * doc/invoke.texi (min-vect-loop-bound): Document new param option.
1276
1277 2007-01-14 Uros Bizjak <ubizjak@gmail.com>
1278
1279 PR target/30413
1280 * config/i386/i386.c (print_operand) ['z']: Output 'b' for
1281 operands of size 1.
1282
1283 2007-01-14 Jan Hubicka <jh@suse.cz>
1284
1285 * tree-dfa.c (remove_referenced_var): New function.
1286 * tree-ssa-live.c (remove_unused_locals): Walk referenced vars and
1287 prune referenced vars list too.
1288 * tree-flow.h (remove_referenced_var): Declare.
1289
1290 2007-01-14 Jan Hubicka <jh@suse.cz>
1291
1292 * tree-eh.c (add_stmt_to_eh_region_fn): Do not add call_exprs
1293 separately.
1294 (remove_stmt_from_eh_region_fn): Do not remove call_exprs.
1295 (verify_eh_throw_stmt_node, verify_eh_throw_table_statements): Kill.
1296 * except.h (verify_eh_throw_table_statements): Kill prototype.
1297 * cfgexpand.c (expand_gimple_basic_block): Propagate Eh regions
1298 into call exrepssions.
1299 * tree-optimize.c (execute_free_cfg_annotatiosn): Do not call
1300 eh trhow verifier.
1301 * tree-cfg.c: Include pointer-set.h.
1302 (verify_node_sharing): Work on pointer set.
1303 (verify_eh_throw_stmt_node): New.
1304 (verify_stmts): Use pointers sets, verify throw_stmt.
1305
1306 2007-01-13 Zdenek Dvorak <dvorakz@suse.cz>
1307
1308 * ipa-reference.c (analyze_function): Consider also addresses taken
1309 in phi nodes.
1310
1311 2007-01-12 Roger Sayle <roger@eyesopen.com>
1312
1313 * c-typeck.c (null_pointer_constant_p): Replace use of
1314 TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
1315 (build_c_cast): Likewise.
1316
1317 2007-01-12 Roger Sayle <roger@eyesopen.com>
1318
1319 * tree.h (force_fit_type_double): Remove unused final argument.
1320 * c-common.c (constant_expression_warning): Replace use of
1321 TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
1322 (convert_and_check): Likewise.
1323 (shorten_compare): Update call to force_fit_type_double.
1324 (c_common_truthvalue_conversion) <INTEGER_CST>: Use integer_zerop.
1325 * convert.c (convert_to_pointer): Update call to
1326 force_fit_type_double.
1327 * fold-const.c (force_fit_type_double): Remove overflowed_const
1328 argument.
1329 (int_const_binop, fold_convert_const_int_from_int,
1330 fold_convert_const_int_from_real, fold_div_compare,
1331 fold_sign_changed_comparison, fold_unary, fold_negate_const,
1332 fold_abs_const, fold_not_const): Remove the final argument from
1333 calls to force_fit_type_double.
1334
1335 2007-01-12 Andrew Pinski <andrew_pinski@playstation.sony.com>
1336
1337 * configure.ac: Set insn to "nop" for spu-*-* also.
1338 * configure: Regenerate.
1339
1340 2007-01-12 Olga Golovanevsky <olga@il.ibm.com>
1341
1342 * builtins.def : Add BUILT_IN_FREE.
1343
1344 2007-01-12 Jan Hubicka <jh@suse.cz>
1345
1346 PR tree-optimization/30443
1347 * tree-inline.c (tree_function_versioning): Do not optimize when
1348 cloning for inlining.
1349
1350 2007-01-12 Zdenek Dvorak <dvorakz@suse.cz>
1351
1352 * doc/loop.texi: Document recording of loop exits.
1353 * cfgloopmanip.c (loopify, duplicate_loop): Use alloc_loop.
1354 (update_single_exits_after_duplication,
1355 update_single_exit_for_duplicated_loop,
1356 update_single_exit_for_duplicated_loops): Removed.
1357 (duplicate_loop_to_header_edge): Do not call
1358 update_single_exits_after_duplication and
1359 update_single_exit_for_duplicated_loops.
1360 (loop_version): Do not update single_exit information.
1361 (fix_loop_structure): Use record_loop_exits instead of
1362 mark_single_exit_loops.
1363 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update
1364 the lists of loop exits.
1365 * cfghooks.c (redirect_edge_and_branch, redirect_edge_and_branch_force,
1366 split_edge, merge_blocks): Update the lists of loop exits.
1367 * modulo-sched.c (sms_schedule): Pass LOOPS_HAVE_RECORDED_EXITS to
1368 loop_optimizer_init.
1369 * loop-init.c (loop_optimizer_init): Call record_loop_exits instead
1370 of mark_single_exit_loops.
1371 (loop_optimizer_finalize): Call release_recorded_exits.
1372 * tree-ssa-loop.c (tree_loop_optimizer_init): Pass
1373 LOOPS_HAVE_RECORDED_EXITS to loop_optimizer_init.
1374 * tree-vectorizer.c (slpeel_tree_duplicate_loop_to_edge_cfg): Do not
1375 update single exit information.
1376 * lambda-code.c (perfect_nestify): Ditto.
1377 * cfgloop.c (flow_loop_free): Destroy the list of exits of the loop.
1378 (mark_single_exit_loops): Removed.
1379 (alloc_loop, loop_exit_hash, loop_exit_eq, loop_exit_free,
1380 get_exit_descriptions, rescan_loop_exit, record_loop_exits,
1381 dump_recorded_exit, dump_recorded_exits, release_recorded_exits): New
1382 functions.
1383 (get_loop_exit_edges, single_exit): Use recorded exit lists.
1384 (add_bb_to_loop, remove_bb_from_loops): Update the lists of loop exits.
1385 (verify_loop_structure): Verify consistency of the exit lists.
1386 (flow_loops_find): Use alloc_loop. Initialize exits hash.
1387 (set_single_exit): Removed.
1388 * cfgloop.h (struct loop_exit): New function.
1389 (struct loop): single_exit_ field replaced by exits field.
1390 (LOOPS_HAVE_MARKED_SINGLE_EXITS): Replaced by LOOPS_HAVE_RECORDED_EXITS.
1391 (struct loops): Added exits hash.
1392 (mark_single_exit_loops, set_single_exit): Declaration removed.
1393 (release_recorded_exits, record_loop_exits, rescan_loop_exit): Declare.
1394
1395 2007-01-12 Richard Sandiford <richard@codesourcery.com>
1396
1397 * doc/invoke.texi: Avoid use of @headitem.
1398
1399 2007-01-12 Richard Sandiford <richard@codesourcery.com>
1400
1401 * cse.c (cse_insn): Move HAVE_CC0 code after declarations.
1402
1403 2007-01-12 Richard Sandiford <richard@codesourcery.com>
1404
1405 * doc/install.texi: Fix m68k-*-* anchor and add m68k-*-* to the
1406 list of targets.
1407
1408 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
1409 Richard Sandiford <richard@codesourcery.com>
1410
1411 * doc/invoke.texi: Document -mno-bitfield, -mno-rtd and -mno-short.
1412 * config/m68k/m68k.opt: Resort options.
1413 (mbitfield, mrtd, mshort): Remove RejectNegative properties.
1414
1415 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
1416 Richard Sandiford <richard@codesourcery.com>
1417
1418 * doc/invoke.texi: Document the macros that are defined by
1419 m68k's -mtune and -mhard-float options.
1420 * config/m68k/m68k-protos.h (m68k_cpp_cpu_ident) Declare.
1421 (m68k_cpp_cpu_family): Likewise.
1422 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Add a full set
1423 of __ucfv*__ macros. Define __mcffpu__ if generating code for
1424 ColdFire FPUs. Define __mcf_cpu_* and __mcf_family_* macros.
1425 * config/m68k/m68k.c (m68k_cpp_cpu_ident): New function.
1426 (m68k_cpp_cpu_family): Likewise.
1427
1428 2007-01-12 Richard Sandiford <richard@codesourcery.com>
1429
1430 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Treat all mc68*
1431 macros besides mc68000 as tuning macros. Use a switch statement
1432 to set them and mcpu32.
1433
1434 2007-01-12 Julian Brown <julian@codesourcery.com>
1435
1436 * config/m68k/m68k.h: Use TARGET_68040 instead of TARGET_68040_ONLY.
1437 (TARGET_68040_ONLY): Rename to...
1438 (TARGET_68040): ...this.
1439 * config/m68k/m68k.c: Use TARGET_68040 instead of TARGET_68040_ONLY.
1440 * config/m68k/m68k.md: Likewise.
1441
1442 2007-01-12 Julian Brown <julian@codesourcery.com>
1443 Nathan Sidwell <nathan@codesourcery.com>
1444 Richard Sandiford <richard@codesourcery.com>
1445
1446 * config.gcc (m680[012]0-*-*, m68k*-*-*): Set m68k_cpu_ident to
1447 the -mcpu= argument associated with the --with-cpu setting.
1448 Define M68K_DEFAULT_TUNE to the default -mtune= option,
1449 if different from the one implied by the -mcpu setting.
1450 Accept --with-cpu=FOO if FOO is listed in m68k-devices.def,
1451 using mcpu=FOO as the default CPU option. Set target_cpu_default2.
1452 * doc/invoke.texi: Mention ColdFire in the introduction to the
1453 m68k options. Document the new -march, -mcpu, -mtune, -mdiv,
1454 -mno-div and -mhard-float options. Make -m68881 a synonym for
1455 -mhard-float. Document the previously-undocumented -m5206e,
1456 -m528x, -m5307 and -m5407 options. Tweak the existing option
1457 documentation for consistency.
1458 * doc/install.texi: Mention new --with-cpu arguments.
1459 * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Only use the
1460 default CPU if neither -mcpu nor -march are specified.
1461 (ASM_CPU_SPEC): Pass down -mcpu and -march options.
1462 (TARGET_CPU_CPP_BUILTINS): Set __mcfisa*__ macros from
1463 TARGET_ISA*. Set the legacy __mcf*__ cpu macros in the same way,
1464 using m68k_tune to decide between families that implement the
1465 same ISA. Use m68k_tune to set __mcfv4e__.
1466 (FL_BITFIELD, FL_68881, FL_COLDFIRE, FL_CF_HWDIV, FL_CF_MAC)
1467 (FL_CF_EMAC, FL_CF_EMAC_B, FL_CF_USP, FL_CF_FPU, FL_ISA_68000)
1468 (FL_ISA_68010, FL_ISA_68020, FL_ISA_68040, FL_ISA_A, FL_ISA_B)
1469 (FL_ISA_C, FL_ISA_MMU): New macros.
1470 (MASK_COLDFIRE): Delete.
1471 (TARGET_68010, TARGET_68020, TARGET_68040_ONLY, TARGET_COLDFIRE)
1472 (TARGET_ISAB): Redefine in terms of m68k_cpu_flags.
1473 (TARGET_68881, TARGET_COLDFIRE_FPU): Redefine in terms of m68k_fpu.
1474 (TARGET_HARD_FLOAT): Do not define here.
1475 (TARGET_ISAAPLUS, TARGET_ISAC): New macros.
1476 (TUNE_68000): New macro.
1477 (TUNE_68000_10): Redefine in terms of TUNE_68000 and TUNE_68010.
1478 (TUNE_68010, TUNE_68030, TUNE_68040, TUNE_68060, TUNE_CPU32)
1479 (TUNE_CFV2): Redefine in terms of m68k_tune.
1480 (uarch_type, target_device, fpu_type): New enums.
1481 (m68k_cpu, m68k_tune, m68k_fpu, m68k_cpu_flags): Declare.
1482 * config/m68k/m68k.c (TARGET_DEFAULT): Remove MASK_68881.
1483 (FL_FOR_isa_00, FL_FOR_isa_10, FL_FOR_isa_20, FL_FOR_isa_40)
1484 (FL_FOR_isa_cpu32, FL_FOR_isa_a, FL_FOR_isa_aplus, FL_FOR_isa_b)
1485 (FL_FOR_isa_c): New macros.
1486 (m68k_isa): New enum.
1487 (m68k_target_selection): New structure.
1488 (all_devices, all_isas, all_microarchs): New tables.
1489 (m68k_cpu_entry, m68k_arch_entry, m68k_tune_entry, m68k_cpu)
1490 (m68k_tune, m68k_fpu, m68k_cpu_flags): New variables.
1491 (MASK_ALL_CPU_BITS): Delete.
1492 (m68k_find_selection): New function.
1493 (m68k_handle_option): Handle -mcpu=, -march= and -mtune=.
1494 Map the legacy target options to a combination of the new ones.
1495 (override_options): Set m68k_cpu, m68k_tune, m68k_fpu and
1496 m68k_cpu_flags. Handle M68K_DEFAULT_TUNE. Use m68k_cpu_flags
1497 to derive default MASK_BITFIELD, MASK_CF_HWDIV and MASK_HARD_FLOAT
1498 settings.
1499 * config/m68k/m68k.opt (m5200, m5206e, m528x, m5307, m5407, mcfv4e)
1500 (m68010, m68020, m68020-40, m68020-60, m68030, m68040): Remove Mask
1501 properties.
1502 (m68881, msoft-float): Change mask from 68881 to HARD_FLOAT.
1503 (march=, mcpu=, mdiv, mhard-float, mtune=): New options.
1504 * config/m68k/m68k-devices.def: New file.
1505
1506 2007-01-12 Richard Sandiford <richard@codesourcery.com>
1507 Nathan Sidwell <nathan@codesourcery.com>
1508
1509 * config/m68k/m68k.h (ASM_CPU_SPEC, ASM_SPEC, EXTRA_SPECS)
1510 (SUBTARGET_EXTRA_SPECS): New macros.
1511 * config/m68k/linux.h (ASM_SPEC): Remove CPU flags;
1512 use %(asm_cpu_spec) instead.
1513 * config/m68k/m68k-none.h (ASM_SPEC): Likewise.
1514 * config/m68k/openbsd.h (ASM_SPEC): Likewise.
1515 * config/m68k/netbsd-elf.h (ASM_SPEC): Likewise.
1516 (EXTRA_SPECS): Rename to...
1517 (SUBTARGET_EXTRA_SPECS): ...this.
1518
1519 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
1520 Richard Sandiford <richard@codesourcery.com>
1521 Julian Brown <julian@codesourcery.com>
1522
1523 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
1524 (m68k-*-uclinux*, m68k-*-linux*, m68k-*-rtems*): Set default_m68k_cpu
1525 to the configuration's default CPU.
1526 (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*): Likewise.
1527 Remove default masks.
1528 (m680[012]0-*-*): Set the default with_cpu to the first part of
1529 the target name.
1530 (m68k*-*-*): Set the default with_cpu to m$default_m68k_cpu.
1531 (m68k*-*-linux): Extend the --with-cpu handling to...
1532 (m680[012]0-*-*, m68k*-*-*): ...these configurations. Allow m68000
1533 and m68010. Don't set target_cpu_default2.
1534 * doc/install.texi: Document --with-cpu for m68k.
1535 * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Define.
1536 * config/m68k/m68k-none.h (TARGET_CPU_DEFAULT, M68K_CPU_m68k)
1537 (M68K_CPU_m68000, M68K_CPU_m68010, M68K_CPU_m68020, M68K_CPU_m68030)
1538 (M68K_CPU_m68040, M68K_CPU_m68302, M68K_CPU_m68332, TARGET_DEFAULT)
1539 (ASM_CPU_DEFAULT_SPEC, CC1_CPU_DEFAULT_SPEC): Delete.
1540 (ASM_SPEC): Remove use of %(asm_cpu_default).
1541 (EXTRA_SPECS, SUBTARGET_EXTRA_SPECS, MULTILIB_DEFAULTS): Delete.
1542 * config/m68k/linux.h (TARGET_DEFAULT): Delete.
1543 (CPP_SPEC): Merge definitions. Do not handle __HAVE_68881__ here.
1544 * config/m68k/netbsd-elf.h (TARGET_OS_CPP_BUILTINS): Define
1545 __HAVE_FPU__ if TARGET_HARD_FLOAT.
1546 (TARGET_DEFAULT): Delete.
1547 (EXTRA_SPECS): Delete cpp_cpu_default_spec, cpp_cpu_spec,
1548 cpp_fpu_spec, asm_default_spec and netbsd_cpp_spec.
1549 (CPP_CPU_SPEC): Delete.
1550 (TARGET_VERSION): Merge definitions, using TARGET_68010 to pick
1551 the appropriate string.
1552 (CPP_CPU_DEFAULT_SPEC, ASM_DEFAULT_SPEC, CPP_FPU_SPEC): Delete.
1553 (CPP_SPEC): Define to NETBSD_CPP_SPEC.
1554 (ASM_SPEC): Don't use %(asm_default_spec).
1555 * config/m68k/m68k.c (TARGET_DEFAULT_TARGET_FLAGS): Remove
1556 TARGET_DEFAULT and add MASK_68881.
1557 * config/m68k/m68k.md: Remove mention of TARGET_DEFAULT from comments.
1558
1559 2007-01-12 Richard Sandiford <richard@codesourcery.com>
1560
1561 * config.gcc (m68010-*-netbsdelf*): Add MASK_68010.
1562 (m68k*-*-netbsdelf*, m68k*-*-openbsd*, m68k*-linux*): Add
1563 MASK_68010 alongside MASK_68020.
1564 * doc/invoke.texi: Document -m68010.
1565 * config/m68k/m68k.opt (m68010): New.
1566 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Define mc68010
1567 if TUNE_68010.
1568 (TUNE_68010): New macro.
1569 * config/m68k/m68k-none.h (M68K_CPU_m68k, M68K_CPU_m68010)
1570 (M68K_CPU_m68020, M68K_CPU_m68030, M68K_CPU_m68040)
1571 (M68K_CPU_m68332): Add MASK_68010.
1572 * config/m68k/linux.h (TARGET_DEFAULT): Add MASK_68010 to
1573 fallback definition.
1574 * config/m68k/netbsd-elf.h (CPP_CPU_SPEC): Remove now-redundant
1575 defines.
1576 * config/m68k/m68k.c (MASK_ALL_CPU_BITS): Add MASK_68010.
1577 (m68k_handle_option): Handle OPT_m68010. Add MASK_68010
1578 to all entries that use MASK_68020.
1579 (output_move_simode_const, output_move_himode, output_move_qimode)
1580 (output_move_stricthi, output_move_strictqi): Use TARGET_68010
1581 instead of TARGET_68020 to select clr behavior. Remove comment
1582 about there being no TARGET_68010.
1583 * config/m68k/m68k.md: Likewise throughout.
1584
1585 2007-01-12 Julian Brown <julian@codesourcery.com>
1586
1587 * config/m68k/m68k.h (TARGET_ISAB): New macro.
1588 * config/m68k/m68k.c: Use TARGET_ISAB rather than TARGET_CFV4.
1589 * config/m68k/m68k.md: Likewise.
1590
1591 2007-01-12 Julian Brown <julian@codesourcery.com>
1592
1593 * config/m68k/m68k.h (LEGITIMATE_INDEX_P, LEGITIMIZE_ADDRESS): Use
1594 TARGET_COLDFIRE_FPU instead of TARGET_CFV4E.
1595
1596 2007-01-12 Julian Brown <julian@codesourcery.com>
1597
1598 * config/m68k/m68k.h (TUNE_68040_60): New macro.
1599 * config/m68k/m68k.c (standard_68881_constant_p): Use it.
1600 * config/m68k/m68k.md: Likewise.
1601
1602 2007-01-12 Julian Brown <julian@codesourcery.com>
1603 Richard Sandiford <richard@codesourcery.com>
1604
1605 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Use TUNE_68030
1606 instead of TARGET_68030, TUNE_68040 instead of TARGET_68040,
1607 TUNE_68060 instead of TARGET_68060 and TUNE_CPU32 instead of
1608 TARGET_CPU32.
1609 (TARGET_CPU32): Rename to...
1610 (TUNE_CPU32): ...this.
1611 (TUNE_68000_10, TUNE_68030, TUNE_68040, TUNE_68060)
1612 (TUNE_CFV2): New macros.
1613 * config/m68k/netbsd-elf.h (LONG_DOUBLE_TYPE_SIZE): Simplify;
1614 remove conditions that are implied by TARGET_68020.
1615 * config/m68k/m68k.c (m68k_output_function_prologue): Use TUNE_68040
1616 instead of TARGET_68040 and TUNE_CPU32 instead of TARGET_CPU32.
1617 (m68k_output_function_epilogue): Likewise.
1618 (m68k_rtx_costs): Likewise. Use TUNE_68060 instead of TARGET_68060
1619 and TUNE_CFV2 instead of TARGET_5200. Use TUNE_68000_10 instead of
1620 "!TARGET_68020 && !TARGET_COLDFIRE" to choose between 68000 and
1621 non-68000 timings. Refactor multiplication and division costs.
1622 (output_addsi3): Use TUNE_68040 instead of TARGET_68040 and
1623 TUNE_CPU32 instead of TARGET_CPU32.
1624 (standard_68881_constant_p): Use TUNE_68040 instead of TARGET_68040
1625 and TUNE_68060 instead of TARGET_68060.
1626 * config/m68k/m68k.md: Use TUNE_68040 instead of TARGET_68040,
1627 TUNE_68060 instead of TARGET_68060, and TUNE_CPU32 instead of
1628 TARGET_CPU32.
1629 (movsi_const0): Use TUNE_68000_10 rather than "!TARGET_68020
1630 && !TARGET_COLDFIRE" to choose between moveq and clr.
1631 Likewise in the unnamed movsf pattern.
1632 (ashlsi_17_24, lshrsi_17_24): Guard with TUNE_68000_10 rather than
1633 "!TARGET_68020 && !TARGET_COLDFIRE". Likewise the unnamed
1634 ashiftrt pattern.
1635
1636 2007-01-12 Richard Sandiford <richard@codesourcery.com>
1637
1638 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Increase amount
1639 of tabbing before backslashes.
1640
1641 2007-01-11 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
1642
1643 * pa-linux.h (ASM_OUTPUT_INTERNAL_LABEL): Undefine.
1644 * pa.h (ASM_OUTPUT_LABEL): Output colon when using GAS.
1645 (ASM_OUTPUT_INTERNAL_LABEL): Define.
1646
1647 2007-01-11 Zdenek Dvorak <dvorakz@suse.cz>
1648
1649 * tree-ssa-loop-ivopts.c (extract_cond_operands): Split from
1650 find_interesting_uses_cond.
1651 (find_interesting_uses_cond): Use extract_cond_operands.
1652 (rewrite_use_compare): Use extract_cond_operands and
1653 force_gimple_operand_bsi. Do not call update_stmt.
1654 (determine_use_iv_cost_condition): Use extract_cond_operands.
1655 Return cheaper of using original bound and changing the exit bound.
1656
1657 2007-01-11 Zdenek Dvorak <dvorakz@suse.cz>
1658
1659 PR tree-optimization/29516
1660 * tree-ssa-address.c (tree_mem_ref_addr, add_to_parts,
1661 most_expensive_mult_to_index, addr_to_parts,
1662 create_mem_ref, maybe_fold_tmr): Make the type of
1663 fields of TARGET_MEM_REF sizetype.
1664 (move_fixed_address_to_symbol, move_pointer_to_base):
1665 New functions.
1666 * tree.def (TARGET_MEM_REF): Add comment on types of
1667 the operands.
1668
1669 2007-01-11 Joseph Myers <joseph@codesourcery.com>
1670
1671 * c-common.c (vector_types_convertible_p): Treat opaque types as
1672 always convertible if they have the same size, but not otherwise.
1673
1674 2007-01-11 Steven Bosscher <steven@gcc.gnu.org>
1675
1676 * ifcvt.c (struct noce_if_info): Add comments to the fields.
1677 Remove the b_unconditional field.
1678 (noce_try_sign_mask): Do not look at b_unconditional.
1679 (noce_process_if_block): Do not use merge_if_blocks. Update
1680 the CFG here. Do not set b_unconditional.
1681 (cond_move_process_if_block): Likewise.
1682 (find_cond_trap): Likewise.
1683 (check_cond_move_block): Require simple jump insns at the end
1684 of the basic block.
1685
1686 2007-01-11 Jan Hubicka <jh@suse.cz>
1687
1688 PR tree-optimization/1046
1689 * tree-tailcall.c (suitable_for_tail_call_opt_p): Use TREE_ADDRESSABLE
1690 when alias info is not ready.
1691 (pass_tail_recursion): Do not require aliasing.
1692 * tree-ssa-copyrename.c (pass_rename_ssa_cop): Likewise.
1693 * tree-ssa-ccp.c (pass_ccp, pass_fold_builtins): Likewise.
1694 * tree-ssa-copy.c (pass_copy_prop): Likewise.
1695 * tree-ssa-forwprop.c (pass_forwprop): Likewise.
1696 * tree-ssa-dce.c (pass_dce, pass_dce_loop, pass_cd_dce): Likewise.
1697 * passes.c (init_optimization_passes): Execute rename_ssa_copies,
1698 ccp, forwprop, copy_prop, merge_phi, copy_prop, dce and tail recursion
1699 before inlining.
1700 * tree-ssa-operands.c (add_virtual_operand, get_indirect_ref_operand):
1701 When aliasing is not build, mark statement as volatile.
1702
1703 2007-01-11 Tom Tromey <tromey@redhat.com>
1704
1705 PR preprocessor/15185, PR preprocessor/20989:
1706 * doc/cppopts.texi <-MT>: Update description of algorithm for
1707 computing default target.
1708 <-M, -MD>: Reword "basename" text.
1709
1710 2007-01-11 Roger Sayle <roger@eyesopen.com>
1711
1712 * builtins.c (expand_builtin_pow, expand_builtin_powi,
1713 fold_builtin_cabs, fold_builtin_sqrt, fold_builtin_trunc,
1714 fold_builtin_floor, fold_builtin_ceil, fold_builtin_round,
1715 fold_builtin_int_int_roundingfn, fold_builtin_bitop,
1716 fold_builtin_bswap, real_constp, fold_builtin_pow,
1717 fold_builtin_powi, fold_builtin_signbit, fold_builtin_copysign,
1718 do_mpfr_arg1, do_mpfr_arg2, do_mpfr_arg3, do_mpfr_sincos): Replace
1719 uses of the macro TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
1720 * convert.c (convert_to_pointer): Likewise.
1721 * expr.c (highest_pow2_factor, expand_expr_real_1): Likewise.
1722 * fold-const.c (force_fit_type, fold_negate_expr, int_const_binop,
1723 const_binop, fold_convert_const_int_from_int,
1724 fold_convert_const_int_from_real,
1725 fold_convert_const_real_from_real, sign_bit_p,
1726 optimize_minmax_comparison, extract_muldiv_1, fold_div_compare,
1727 fold_sign_changed_comparison, fold_unary, fold_comparison,
1728 fold_binary, multiple_of_p, tree_Expr_non_zero_p,
1729 fold_negate_const, fold_abs_const, fold_not_const): Likewise.
1730 * print-tree.c (print_node_brief, print_node): Likewise.
1731 * stor-layout.c (place_field, layout_type): Likewise.
1732 * tree-chrec.c (keep_cast): Likewise.
1733 * tree.c (build_vector, build_real, build_real_from_int_cst,
1734 build_complex): Likewise.
1735
1736 2007-01-11 Roger Sayle <roger@eyesopen.com>
1737
1738 * tree.h (TREE_CONSTANT_OVERFLOW): Obsolete. For the time being,
1739 treat TREE_CONSTANT_OVERFLOW as a synonym of TREE_OVERFLOW.
1740
1741 2007-01-11 Paolo Bonzini <bonzini@gnu.org>
1742
1743 * configure.ac (strict1_warn): Rename to strict_warn.
1744 (WERROR, --enable-werror, symlink hacks, stage1_cflags,
1745 cc_set_by_configure, quoted_cc_set_by_configure,
1746 stage_prefix_set_by_configure, quoted_stage_prefix_set_by_configure,
1747 all_boot_languages, all_stagestuff): Remove.
1748 (target_list): Remove bootstrap targets.
1749 * Makefile.in (quickstrap): Unconditionally make a synonym of all.
1750 (BOOT_LANGUAGES, STAGE1_CFLAGS, STAGE1_CHECKING,
1751 REMAKEFLAGS, FLAGS_TO_PASS, PREPEND_DOTDOT_TO_RELATIVE_PATHS,
1752 SUBDIR_FLAGS_TO_PASS, WERROR_FLAGS, STRICT2_WARN, LANG_STAGESTUFF,
1753 VOL_FILES, POSTSTAGE1_FLAGS_TO_PASS, STAGE2_FLAGS_TO_PASS,
1754 STAGEPROFILE_FLAGS_TO_PASS, STAGEFEEDBACK_FLAGS_TO_PASS, stage1_build,
1755 stage1_copy, stage2_build, stage2_copy, stageprofile_build,
1756 stageprofile_copy, stage3_build, stage3_copy, stagefeedback_build,
1757 stagefeedback_copy, stage4_build, clean_s1, clean_sw, bootstrap,
1758 bootstrap-lean, bootstrap2, bootstrap2-lean, bootstrap3,
1759 bootstrap3-lean, bootstrap4, bootstrap4-lean, unstage1, unstage2,
1760 unstage3, unstage4, unstageprofile, unstagefeedback, restage, restage2,
1761 restage3, restage4, restageprofile, restagefeedback, bubbleestrap,
1762 cleanstrap, unstrap, restrap, *compare, *compare3, *compare4,
1763 *compare-lean, *compare3-lean, *compare4-lean, stage1-start, stage1,
1764 stage2-start, stage2, stage3-start, stage3, stage4-start, stage4,
1765 stageprofile-start, stageprofile, stagefeedback-start, stagefeedback,
1766 risky-stage1, risky-stage2, risky-stage3, risky-stage4): Remove.
1767 (ORDINARY_FLAGS_TO_PASS): Rename to FLAGS_TO_PASS.
1768 (STAGECOPYSTUFF, STAGEMOVESTUFF): Consolidate into MOSTLYCLEANFILES.
1769 (mostlyclean): Adjust.
1770 (clean, distclean): Don't mention bootstrap stuff.
1771 * configure: Regenerate.
1772 * ada/config-lang.in, cp/config-lang.in, forttran/config-lang.in,
1773 java/config-lang.in, objc/config-lang.in, objcp/config-lang.in,
1774 treelang/config-lang.in (stagestuff): Remove.
1775 * doc/sourcebuild.texi (stage1, stage2, stage3, stage4,
1776 stageprofile, stagefeedback, stagestuff): Remove mention.
1777
1778 2007-01-11 Nick Clifton <nickc@redhat.com>
1779
1780 * config/mcore/predicates.md (mcore_general_movesrc_operand):
1781 Accept CONSTs.
1782 (mcore_general_movdst_operand): Do not accept CONST_INTs.
1783 (mcore_arith_K_S_operand): Run the test for the S constraint not
1784 the test for the M constraint.
1785 (mcore_addsub_operand): Do not accept integer values that are
1786 larger than 32 bits.
1787 * config/mcore/mcore.md: Remove unused constraints from split.
1788 (andsi3): Use HOST_WIDE_INT instead of int to hold an INTVAL.
1789 (addsi3): Likewise.
1790 (allocate_stack): Likewise.
1791 * config/mcore/mcore.c (mcore_print_operand): Restrict output of P
1792 operands to 32 bits.
1793 (mcore_const_costs): Use HOST_WIDE_INT instead of int to hold an
1794 INTVAL.
1795 (mcore_and_cost, mcore_modify_comparison, const_ok_for_mcore,
1796 mcore_const_ok_for_inline, mcore_const_trick_uses_not,
1797 try_constant_tricks, mcore_num_ones, mcore_num_zeros,
1798 mcore_output_bclri, mcore_output_andn, output_inline_const,
1799 mcore_output_move, mcore_output_movedouble): Likewise.
1800 (mcore_output_cmov): Use CONST_OK_FOR_M and CONST_OK_FOR_N.
1801 (output_inline_const): Likewise.
1802 (output_inline_const): Fix format strings used in sprintf
1803 statements.
1804 * config/mcore/mcore-protos.h: Update prototypes for changed
1805 functions in mcore.c.
1806 * config/mcore/mcore.h (CONST_OK_FOR_I): Cast values to
1807 HOST_WIDE_INT and not int.
1808 (CONST_OK_FOR_J, CONST_OK_FOR_K, CONST_OK_FOR_L, CONST_OK_FOR_M,
1809 CONST_OK_FOR_N): Likewise.
1810 (LEGITIMATE_CONSTANT_P): Also check CONSTANT_P.
1811 (GO_IF_LEGITIMATE_INDEX): Use HOST_WIDE_INT instead of int to hold
1812 an INTVAL.
1813
1814 2007-01-10 Jan Hubicka <jh@suse.cz>
1815
1816 * tree-vrp.c (remove_range_assertions): Release defs.
1817 * tree-ssa-loop-ivopts.c (rmeove_statement): Likewise.
1818 * tree-ssa-dom.c (remove_stmt_or_phi): Likewise.
1819
1820 2007-01-10 Paul Brook <paul@codesourcery.com>
1821
1822 * config/arm/arm.c (arm_rtx_costs_1): Handle mutiply-subtract.
1823 * config/arm/arm.md (mulsi3subsi): New insn.
1824
1825 2007-01-10 Zdenek Dvorak <dvorakz@suse.cz>
1826
1827 * tree-ssa-loop-manip.c (tree_unroll_loop): Make it a wrapper over ...
1828 (tree_transform_and_unroll_loop): New.
1829 * tree-flow.h (transform_callback, tree_transform_and_unroll_loop):
1830 Declare.
1831
1832 2007-01-10 Robert Kennedy <jimbob@google.com>
1833
1834 * fold-const.c (fold_comparison): Fold comparisons like (x *
1835 1000 < 0) to (x < 0).
1836
1837 2007-01-10 Ian Lance Taylor <iant@google.com>
1838
1839 * tree-pretty-print.c (dump_generic_node): Print parentheses when
1840 operands have the same priority.
1841
1842 2007-01-10 Tom Tromey <tromey@redhat.com>
1843
1844 * fold-const.c (fold_truthop): Don't check can_use_bit_fields_p.
1845 (fold_binary): Likewise.
1846 * langhooks.c (lhd_can_use_bit_fields_p): Removed.
1847 * langhooks-def.h (lhd_can_use_bit_fields_p): Removed.
1848 (LANG_HOOKS_CAN_USE_BIT_FIELDS_P): Removed.
1849 (LANG_HOOKS_INITIALIZER): Remove LANG_HOOKS_CAN_USE_BIT_FIELDS_P.
1850 * langhooks.h (struct lang_hooks): Removed field
1851 'can_use_bit_fields_p'.
1852
1853 2007-01-10 Ralf Corsépius <ralf.corsepius@rtems.org>
1854
1855 * config/bfin/t-bfin, config/bfin/t-bfin-elf: Remove GCC_CFLAGS.
1856
1857 2007-01-10 Razya Ladelsky <razya@il.ibm.com>
1858
1859 * function.c (get_last_funcdef_no): New function.
1860 * function.h (get_last_funcdef_no): Declare.
1861 * tree-inline.c (initialize_cfun): Add initialization.
1862 (tree_function_versioning): Cleanup.
1863
1864 2007-01-10 Jan Hubicka <jh@suse.cz>
1865
1866 * tree-inline.c (setup_one_parameter): Do not propagate into abnormal
1867 PHIs.
1868
1869 2007-01-10 Sa Liu <saliu@de.ibm.com>
1870 Ben Elliston <bje@au.ibm.com>
1871
1872 * spu.h (STACK_SAVE_AREA): Use VOIDmode for SAVE_FUNCTION, SImode
1873 for SAVE_NONLOCAL and Pmode for any other save level.
1874 * spu-protos.h (spu_restore_stack_block): Declare.
1875 * spu.md (save_stack_block): Remove.
1876 (restore_stack_block): Call spu_restore_stack_block.
1877 * spu.c (spu_restore_stack_block): New function.
1878 (spu_expand_epilogue): Remove old comment.
1879
1880 2007-01-09 Zdenek Dvorak <dvorakz@suse.cz>
1881
1882 PR tree-optimization/30322
1883 * tree-ssa-loop-ivopts.c (fold_affine_expr, iv_value): Removed.
1884 (cand_value_at): Return the value as aff_tree.
1885 (may_eliminate_iv): Convert the bound from aff_tree to tree.
1886 * tree-affine.c (aff_combination_add_cst, aff_combination_add_product,
1887 aff_combination_mult): New functions.
1888 (aff_combination_add): Use aff_combination_add_cst.
1889 (aff_combination_convert): Allow conversions to a wider type.
1890 (tree_to_aff_combination): Handle BIT_NOT_EXPR.
1891 * tree-affine.h (aff_combination_mult): Declare.
1892
1893 2007-01-09 Carlos O'Donell <carlos@codesourcery.com>
1894
1895 * doc/tm.texi: Update documentation to reflect reality of exec
1896 and start file search behaviours. Update copyright year.
1897 * doc/invoke.texi: Explain how GCC_EXEC_PREFIX is used to find
1898 header file directories.
1899
1900 2007-01-09 Uros Bizjak <ubizjak@gmail.com>
1901
1902 * config/i386/i386.md (*sinxf2): Rename to *sinxf2_i387.
1903 (*cosxf2): Rename to cosxf2_i387.
1904 (*sindf2, *sinsf2): Extend operand 1 to XFmode. Macroize patterns
1905 using X87MODEF12 mode macro. Rename patterns to
1906 *sin_extend<mode>xf2_i387. Use SSE_FLOAT_MODE_P to disable patterns
1907 for SSE math.
1908 (*cosdf2, *cossf2): Ditto.
1909 (sincosdf3, sincossf3): Ditto. Rewrite corresponding splitters
1910 to match extended input operands.
1911 (sincos<mode>3): New expander.
1912 (*sinextendsfdf2, *cosextendsfdf2, *sincosextendsfdf3): Remove
1913 insn patterns and corresponding splitters.
1914
1915 2007-01-09 Kaz Kojima <kkojima@gcc.gnu.org>
1916
1917 * config/sh/t-linux (TARGET_LIBGCC2_CFLAGS): Delete.
1918 (SHLIB_MAPFILES, SHLIB_LINK, SHLIB_INSTALL): Likewise.
1919
1920 2007-01-09 Nicolas Pitre <nico@cam.org>
1921
1922 PR target/30173
1923 * arm/ieee754-df.S (Lad_s): Also test the low word of X for zero.
1924
1925 2007-01-08 Geoffrey Keating <geoffk@apple.com>
1926
1927 * target.h (struct gcc_target): New field library_rtti_comdat.
1928 * target-def.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): New.
1929 (TARGET_CXX): Add TARGET_CXX_LIBRARY_RTTI_COMDAT.
1930 * doc/tm.texi (C++ ABI): Document TARGET_CXX_LIBRARY_RTTI_COMDAT.
1931 * config/darwin.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
1932
1933 2007-01-08 Geoffrey Keating <geoffk@apple.com>
1934
1935 * doc/invoke.texi (Optimize Options): Correct description of -O0.
1936
1937 2007-01-08 Richard Guenther <rguenther@suse.de>
1938
1939 * tree.h (force_fit_type_double): Export.
1940 (force_fit_type): Remove.
1941 * fold-const.c (force_fit_type_double): New function.
1942 (force_fit_type): Remove.
1943 (int_const_binop): Use it.
1944 (fold_convert_const_int_from_int): Likewise.
1945 (fold_convert_const_int_from_real): Likewise.
1946 (fold_div_compare): Likewise.
1947 (fold_sign_changed_comparison): Likewise.
1948 (fold_unary): Likewise.
1949 (fold_negate_const): Likewise.
1950 (fold_abs_const): Likewise.
1951 (fold_not_const): Likewise.
1952 * c-common.c (shorten_compare): Use force_fit_type_double.
1953 * convert.c (convert_to_pointer): Likewise.
1954
1955 2007-01-08 Richard Guenther <rguenther@suse.de>
1956
1957 * tree.h (build_int_cst_wide_type): Export.
1958 * tree.c (build_int_cst_wide_type): New function.
1959 (build_int_cst_wide): Fix comment.
1960 * builtins.c (fold_builtin_object_size): Use build_int_cst
1961 to build -1 or 0 of the correct type. Use fit_double_type
1962 to check for overflow.
1963 * fold-const.c (optimize_bit_field_compare): Use build_int_cst_type
1964 to build the mask.
1965 (decode_field_reference): Likewise.
1966 (all_ones_mask_p): Likewise.
1967 (native_interpret_int): Use build_int_cst_wide_type.
1968 (fold_binary): Use build_int_cst_type to build an all-ones
1969 value.
1970 * stor-layout.c (set_sizetype): Use build_int_cst_wide_type.
1971
1972 2007-01-08 Daniel Jacobowitz <dan@codesourcery.com>
1973
1974 * config/pa/t-pa64 (libgcc_stub.a): Use $(T).
1975
1976 2007-01-09 Ben Elliston <bje@au.ibm.com>
1977
1978 * genautomata.c (STATS_OPTION): New option.
1979 (stats_flag): New flag.
1980 (gen_automata_option): Handle it.
1981 (initiate_automaton_gen): Ditto.
1982 (write_automata): Output statistics only if stats_flag is
1983 set. Likewise, output time statistics only if time_flag is set.
1984 * doc/md.texi (Processor pipeline description): Document new flag.
1985
1986 2007-01-08 Richard Guenther <rguenther@suse.de>
1987
1988 * builtins.c (fold_builtin_int_roundingfn): Use fit_double_type.
1989 * tree.c (build_int_cst_type): Likewise.
1990 (size_in_bytes): Don't call force_fit_type on the result.
1991 (int_fits_type_p): Use fit_double_type.
1992 * fold-const.c (fit_double_type): New function.
1993 (force_fit_type): Use it.
1994 * tree.h (fit_double_type): Export.
1995
1996 2007-01-08 Jan Hubicka <jh@suse.cz>
1997
1998 * tree-vectorizer.c (gate_increase_alignment): Fix return type.
1999 * ipa.c (function_and_variable_visibility): Fix return type.
2000
2001 2007-01-08 Richard Guenther <rguenther@suse.de>
2002
2003 * tree-ssa-ccp.c (maybe_fold_offset_to_array_ref): Use type
2004 of offset to build the index.
2005 * tree-pretty-print.c (dump_generic_node): Don't build negated
2006 const just for printing.
2007 * c-pretty-print.c (pp_c_integer_constant): Likewise.
2008 * builtins.c (fold_builtin_int_roundingfn): Check if result
2009 fits the type by using force_fit_type and comparing the result.
2010 * predict.c (predict_loops): Use compare_tree_int for comparison.
2011 * tree.c (build_int_cst): Fall back to integer_type_node for
2012 NULL_TREE type.
2013 (build_int_cst_wide): Assert type is non-null.
2014
2015 2007-01-08 Roberto Costa <roberto.costa@st.com>
2016
2017 * tree-vrp.c (extract_range_from_cond_expr): New.
2018 (extract_range_from_expr): Handle COND_EXPR nodes used as expressions.
2019 * tree-ssa-ccp.c (get_maxval_strlen): Handle COND_EXPR nodes used
2020 as expressions.
2021 (fold_stmt): Bug fix, avoid infinite recursion when folding COND_EXPRs.
2022 * tree-ssa-forwprop.c (simplify_cond, forward_propagate_into_cond,
2023 tree_ssa_forward_propagate_single_use_vars): Handle COND_EXPR nodes
2024 used as expressions.
2025 * tree-object-size.c (cond_expr_object_size): New.
2026 (collect_object_sizes_for): Handle COND_EXPR nodes used as expressions.
2027
2028 2007-01-08 Jan Hubicka <jh@suse.cz>
2029
2030 * tree-ssa-forwprop.c (forward_propagate_into_cond,
2031 tree_ssa_forward_propagate_single_use_va): Release defs of propagated
2032 statement.
2033
2034 2007-01-08 Richard Guenther <rguenther@suse.de>
2035
2036 PR tree-optimization/23603
2037 * tree-vrp.c (set_value_range_to_truthvalue): New function.
2038 (extract_range_from_binary): Fall back to truthvalue instead of
2039 varying for TRUTH_*_EXPR.
2040 (extract_range_from_comparison): Fall back to truthvalue instead of
2041 varying.
2042 (vrp_visit_phi_node): Don't adjust new range bounds to +INF/-INF
2043 if all visited PHI values were constant.
2044
2045 2007-01-08 Jan Hubicka <jh@suse.cz>
2046
2047 * cgraphunit.c (cgraph_process_new_functions): Reset reachable flag.
2048 (cgraph_analyze_function): break out from ...
2049 (cgraph_finalize_compilation_unit): ... here.
2050 (cgraph_expand_function): Remove forgoten commented out line.
2051 (cgraph_optimize): Analyze functions.
2052
2053 2007-01-08 Jan Hubicka <jh@suse.cz>
2054
2055 * tree-pas.h (TODO_remove_function): New flag.
2056 (TODO_update*): Renumber.
2057 (pass_ipa_increase_alignment,
2058 pass_ipa_function_and_variable_visibility): New passes.
2059 * cgraphunit.c (cgraph_increase_alignment): Move to tree-vectorizer.c
2060 (cgraph_function_and_variable_visibility): Move to ipa.c
2061 (cgraph_optimize): Don't call cgraph_function_and_variable_visibility,
2062 cgraph_increase_alignment.
2063 * ipa-inline.c (cgraph_decide_inlining): Don't push timevar.
2064 (cgraph_decide_inlining_incrementally): Push TV_INTEGRATION before
2065 calling tree-inline.
2066 (cgraph_early_inlining): Do not call cgraph_remove_unreachable_nodes.
2067 (pass_ipa_inline, pass_early_ipa_inlining): Set TODO_remove_functions
2068 * tree-vectorizer.c (increase_alignment): Move here from cgraphunit.c
2069 (gate_increase_alignment): New function.
2070 (pass_ipa_increase_alignment): New pass.
2071 * ipa.c: Inline tree-pass.h and timevar.h
2072 (function_and_variable_visibility): Move here from cgraphunit.c
2073 * tree-optimize.c (pass_early_local_passes): Add TODO_remove_functions.
2074 * passes.c (init_optimization_passes): Add the two new passes.
2075 (execute_todo): Handle cgraph_remove_functions.
2076
2077 2007-01-08 Nick Clifton <nickc@redhat.com>
2078
2079 * config/frv/predicates.md (reg_or_0_operand): Accept
2080 CONST_DOUBLEs.
2081
2082 2007-01-08 Ralf Corsépius <ralf.corsepius@rtems.org>
2083
2084 * config/bfin/rtems.h, config/bfin/t-rtems: New.
2085 * config.gcc: Add bfin*-rtems*.
2086
2087 2007-01-08 Mark Shinwell <shinwell@codesourcery.com>
2088
2089 * c.opt: Add -flax-vector-conversions.
2090 * c-typeck.c (convert_for_assignment): Pass flag to
2091 vector_types_convertible_p to allow emission of note.
2092 (digest_init): Likewise.
2093 * c-opts.c: Handle -flax-vector-conversions.
2094 * c-common.c (flag_lax_vector_conversions): New.
2095 (vector_types_convertible_p): Unless -flax-vector conversions
2096 has been passed, disallow conversions between vectors with
2097 differing numbers of subparts and/or element types. If such
2098 a conversion is disallowed, possibly emit a note on the first
2099 occasion only to inform the user of -flax-vector-conversions.
2100 The new last argument specifies this.
2101 * c-common.h (flag_lax_vector_conversions): New.
2102 (vector_types_convertible_p): Add extra argument.
2103 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Use
2104 char_type_node for V*QI type vectors.
2105 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins):
2106 Update to satisfy new typechecking rules.
2107 * config/rs6000/altivec.h (vec_cmple): Use vec_cmpge, for both
2108 C and C++ variants.
2109 * doc/invoke.texi (C Dialect Options): Document
2110 -flax-vector-conversions.
2111
2112 2007-01-08 Mark Shinwell <shinwell@codesourcery.com>
2113
2114 PR tree-optimization/29877
2115 * tree-ssa-ter.c (is_replaceable_p): Deem assignments with
2116 a register variable on the RHS to not be replaceable.
2117
2118 2007-01-08 Chen Liqin <liqin@sunnorth.com.cn>
2119 * config/score/t-score-elf (MULTILIB_OPTIONS): Change.
2120 * config/score/predicates.md (const_uimm5, sr0_operand, const_simm12,
2121 const_simm15, const_pow2, const_npow2): Added.
2122 * config/score/misc.md (insv, extv, extzv, movmemsi,
2123 move_lbu_a/b, mov_lhu_a/b etc): Added and fix some bug.
2124 * config/score/score.c (score_address_cost, score_select_cc_mode):
2125 Added.
2126 Change CONST_OK_FOR_LETTER_P/EXTRA_CONSTRAINT define.
2127 Update score_rtx_costs for MACRO TARGET_RTX_COSTS.
2128 Update score_print_operand.
2129 * config/score/score.h (DATA_ALIGNMENT, SELECT_CC_MODE): Added.
2130 Adjust register allocate order and update some macro define.
2131 * config/score/score-mdaux.c (mdx_unaligned_load, mdx_unsigned_store,
2132 mdx_block_move_straight, mdx_block_move_loop_head,
2133 mdx_block_move_loop_body, mdx_block_move_loop_foot, mdx_block_move_loop,
2134 mdx_block_move): Added.
2135 (mdx_movsicc, mdp_select_add_imm, mdp_select, mds_zero_extract_andi,
2136 mdp_limm): Updated and fix some bug and typo.
2137 * config/score/score.md (movqi/hi/si, add/sub/zero/ext): Updated.
2138 (movsf, movdf, doloop_end): Added.
2139
2140 2007-01-08 Kazu Hirata <kazu@codesourcery.com>
2141
2142 * config/arm/arm.c, config/arm/arm.h, config/arm/arm.md,
2143 config/arm/thumb2.md: Fix comment typos.
2144 * doc/extend.texi: Fix a typo.
2145
2146 2007-01-07 Eric Christopher <echristo@apple.com>
2147
2148 * configure.ac: Check for __stack_chk_fail for darwin.
2149 * configure: Regenerate.
2150
2151 2007-01-07 Richard Guenther <rguenther@suse.de>
2152
2153 * tree-vrp.c (extract_range_from_assert): CSE calls to
2154 compare_values where possible.
2155 (extract_range_from_unary_expr): Likewise.
2156
2157 2007-01-07 Anatoly Sokolov <aesok@post.ru>
2158
2159 * config/avr/avr-protos.h (call_insn_operand): Delete prototype.
2160 * config/avr/avr.c (call_insn_operand): Delete function.
2161 * config/avr/avr.md (*pushqi, *pushhi, *pushsi, *pushsf): Use REG_SP
2162 instead of register number. Use predicates.
2163 * config/avr/predicates.md (const0_operand, reg_or_0_operand,
2164 call_insn_operand): Add.
2165
2166 2007-01-06 Jan Hubicka <jh@suse.cz>
2167
2168 * tree-pass.h (pass_build_cgraph_edges): Declare.
2169 * cgraphunit.c (record_refernece): Move to cgraphbuild.c
2170 (visited_nodes): Remove.
2171 (cgraph_create_edges): Move to cgraphbuild.c; rename to
2172 build_cgrpah_edges; make visited_nodes local.
2173 (cgraph_process_new_functions): DO not call initialize_inline_failed.
2174 (record_references_in_initializer): Move to cgraphbuild.c
2175 (initialize_inline_failed, rebuild_cgraph_edges,
2176 pass_rebuild_cgraph_edges): Move to cgraphbuild.c.
2177 (verify_cgraph_node): Make visited_nodes local.
2178 (cgraph_analyze_function): Do not call cgraph_create_edges and
2179 initialize_inline_failed.
2180 (cgraph_expand_function): Do not call cgraph_lower_function;
2181 assert that function is already lowered.
2182 * Makefile.in (cgraphbuild.o): New.
2183 * passes.c (init_optimization_passes): Add pass_build_cgraph_edges
2184 at the end of lowering passes.
2185
2186 2007-01-06 Steven Bosscher <steven@gcc.gnu.org>
2187
2188 * ifcvt.c (cond_move_convert_if_block): New function, code
2189 factored out from...
2190 (cond_move_process_if_block): ...here. Call the new function
2191 on the THEN and ELSE blocks.
2192 (merge_if_block): Do not copy global_live_at_end, merge_blocks
2193 already takes care of this.
2194
2195 2007-01-05 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
2196
2197 PR c/19978
2198 * tree.h (TREE_OVERFLOW_P): New.
2199 * c-typeck.c (parser_build_unary_op): Warn only if result
2200 overflowed and operands did not.
2201 (parser_build_binary_op): Likewise.
2202 (convert_for_assignment): Remove redundant overflow_warning.
2203 * c-common.c (overflow_warning): Don't check or set TREE_OVERFLOW.
2204
2205 2007-01-05 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
2206
2207 * c-typeck.c (store_init_value): Split over two lines to follow
2208 the GNU coding style.
2209
2210 2007-01-05 Benjamin Kosnik <bkoz@redhat.com>
2211
2212 * c-cppbuiltin.c (c_cpp_builtins): __GXX_EXPERIMENTAL_CPP0X__ to
2213 __GXX_EXPERIMENTAL_CXX0X__.
2214 * doc/cpp.texi: Same.
2215
2216 2007-01-05 Richard Guenther <rguenther@suse.de>
2217
2218 PR middle-end/27826
2219 * tree.c (get_narrower): Do not construct COMPONENT_REFs
2220 with mismatched types. Instead explicitly build a
2221 conversion NOP_EXPR.
2222
2223 2007-01-05 Ian Lance Taylor <iant@google.com>
2224
2225 * c-common.c (decl_with_nonnull_addr_p): New function.
2226 (c_common_truthvalue_conversion): Call it.
2227 * c-typeck.c (build_binary_op): Likewise.
2228 * c-common.h (decl_with_nonnull_addr_p): Declare.
2229
2230 2007-01-05 Jakub Jelinek <jakub@redhat.com>
2231
2232 PR c/30360
2233 * libgcc2.c (__divdc3): Compare c and d against 0.0 instead of
2234 denom against 0.0.
2235
2236 2007-01-05 Joel Brobecker <brobecker@adacore.com>
2237
2238 * doc/install.texi (Final install): Document the fact that
2239 the GNAT runtime should not be stripped.
2240
2241 2007-01-04 Jan Hubicka <jh@suse.cz>
2242
2243 * tree-inline.c (fold_marked_statements): Update operand caches
2244 and EH after folding
2245
2246 2007-01-04 Ian Lance Taylor <iant@google.com>
2247
2248 * c-common.c (check_function_nonnull): Whitespace fix.
2249
2250 2007-01-04 Jan Hubicka <jh@suse.cz>
2251
2252 * tree-optimize.c (execute_fixup_cfg): Correct previously mistakely
2253 comitted older version of patch.
2254 (pass_fixup_cfg): Add TODOs to verify flow and statements, dump
2255 function, celanup cfg and collect garbage.
2256
2257 2007-01-04 Mike Stump <mrs@apple.com>
2258
2259 * Makefile.in (mostlyclean): Don't remove libgcc anymore.
2260 (clean): Likewise.
2261
2262 2007-01-04 Eric Christopher <echristo@apple.com>
2263
2264 * libgcc2.c (__bswapsi2): Use SItype.
2265 (__bswapdi2): Use DItype.
2266 * libgcc2.h: Update for above.
2267
2268 2007-01-04 Paul Brook <paul@codesourcery.com>
2269
2270 * config/arm/arm.md (arm_mulsi3, thumb_mulsi3, mulsi3_compare0,
2271 mulsi_compare0_scratch, mulsi3addsi, mulsi3addsi_compare0,
2272 mulsi3addsi_compare0_scratch, mulsidi3adddi, mulsidi3,
2273 umulsidi3, umulsidi3adddi, smulsi3_highpart,
2274 umulsi3_highpart): Make conditional on !arm_arch6.
2275 (arm_mulsi3_v6, thumb_mulsi3_v6, mulsi3_compare0_v6,
2276 mulsi_compare0_scratch_v6, mulsi3addsi_v6, mulsi3addsi_compare0_v6,
2277 mulsi3addsi_compare0_scratch_v6, mulsidi3adddi_v6, mulsidi3_v6,
2278 umulsidi3_v6, umulsidi3adddi_v6, smulsi3_highpart_v6,
2279 umulsi3_highpart_v6): New insns.
2280
2281 2007-01-04 Roger Sayle <roger@eyesopen.com>
2282
2283 * fold-const.c (fold_convert): When casting an expression to void,
2284 fold_ignored_result may discover a GIMPLE_MODIFY_STMT which doesn't
2285 have a type. Instead of attempting to build a NOP_EXPR, return
2286 these "special" trees directly.
2287
2288 2007-01-04 Joseph Myers <joseph@codesourcery.com>
2289
2290 * config/rs6000/rs6000.c (rs6000_rtx_costs): Make adjustment for
2291 MULT inside MINUS as either argument. Use rs6000_cost->dmul -
2292 rs6000_cost->fp not 0 as adjustment for outer NEG.
2293
2294 2007-01-04 Jan Hubicka <jh@suse.cz>
2295
2296 * cgraph.c (cgraph_release_function_body): New function.
2297 (cgraph_remove_node): Use it.
2298 * cgraph.h (cgraph_release_function_body): Declare.
2299 * cgraphunit.c (cgraph_expand_function): Use it.
2300 * ipa.c (cgraph_remove_unreahchable_nodes): Use it.
2301 * tree-ssa.c (delete_tree_ssa): Allow to be called before aliasing
2302 is initialized and while compilation of other function is running.
2303 * tree-optimize.c (execute_free_cfg_annotations): Move code to clear
2304 statement CFG annotations from here to ...
2305 * tree-cfg.c (delete_tree_cfg_annotations): ... here.
2306
2307 2007-01-04 Zdenek Dvorak <dvorakz@suse.cz>
2308
2309 * cfgloop.h (enum li_flags): Make the constants powers of two.
2310
2311 2007-01-04 Jan Hubicka <jh@suse.cz>
2312
2313 * tree-inline.c (copy_bb): Insert new statements to statements_to_fold
2314 set.
2315 (fold_marked_statements): New function.
2316 (optimize_inline_calls, tree_function_versioning): Fold new statements.
2317 * tree-inline.h (copy_body_data): Add statements_to_fold.
2318
2319 2007-01-03 Daniel Jacobowitz <dan@codesourcery.com>
2320
2321 * config.gcc: Mention libgcc/config.host.
2322 * Makefile.in: Update comments mentioning libgcc.
2323 (LIBGCC, INSTALL_LIBGCC, GCC_PARTS, mklibgcc): Delete.
2324 (all.cross, start.encap, rest.encap, rest.cross): Update
2325 dependencies for libgcc move.
2326 (libgcc.mk, LIBGCC_DEPS, libgcov.a, libgcc.a, stmp-multilib)
2327 (clean-target, clean-target-libgcc): Delete.
2328 (srcdirify, GCC_EXTRA_PARTS): New macros.
2329 (libgcc-support, libgcc.mvars): New rules.
2330 (distclean): Remove mention of mklibgcc.
2331 (install): Don't reference INSTALL_LIBGCC.
2332 (install-common): Don't reference EXTRA_PARTS.
2333 (install-libgcc, install-multilib): Delete rules.
2334 * mklibgcc.in: Delete file.
2335 * doc/configfiles.texi: Don't mention mklibgcc.
2336
2337 * config/i386/t-darwin (SHLIB_VERPFX): Delete (moved to libgcc).
2338 * config/i386/t-darwin64 (SHLIB_VERPFX): Likewise.
2339 * config/rs6000/t-darwin (SHLIB_VERPFX): Likewise.
2340 * config/rs6000/t-ppccomm (TARGET_LIBGCC2_CFLAGS, SHLIB_MAPFILES)
2341 (mklibgcc, ldblspecs): Likewise.
2342
2343 * config/i386/t-nwld (libgcc.def, libc.def, libpcre.def)
2344 (posixpre.def): Use $(T).
2345 (SHLIB_EXT, SHLIB_NAME, SHLIB_SLIBDIR_QUAL, SHLIB_DEF, SHLIB_MAP)
2346 (SHLIB_SRC, SHLIB_INSTALL): Delete.
2347 (SHLIB_LINK): Make dummy.
2348 * config/t-slibgcc-darwin: Delete contents except for dummy SHLIB_LINK.
2349
2350 * config/frv/t-linux (EXTRA_MULTILIB_PARTS): Clear.
2351
2352 * config/alpha/t-crtfm: Use $(T) in rules for EXTRA_PARTS.
2353 * config/alpha/t-vms, config/alpha/t-vms64, config/fr30/t-fr30,
2354 config/i386/t-rtems-i386, config/ia64/t-ia64, config/rs6000/t-beos,
2355 config/rs6000/t-newas, config/sparc/t-elf: Likewise.
2356
2357 * configure.ac (all_outputs): Remove mklibgcc.
2358 * configure: Regenerated.
2359
2360 2007-01-03 Josh Conner <jconner@apple.com>
2361
2362 PR middle-end/29683
2363 * calls.c (compute_argument_addresses): Set stack and stack_slot
2364 for partial args, too.
2365 (store_one_arg): Use locate.size.constant for the size when
2366 generating a save_area.
2367
2368 2007-01-03 Robert Kennedy <jimbob@google.com>
2369
2370 * tree-cfg.c (tree_merge_blocks): Release SSA_NAME phi results
2371 whose definitions are deleted due to basic block merging.
2372
2373 2007-01-03 Paul Brook <paul@codesourcery.com>
2374
2375 PR target/16634
2376 * config/arm/arm.c (output_return_instruction): Pop PC in interrupt
2377 functions.
2378 (use_return_insn): Return 0 for Thumb interrupt functions.
2379 (print_multi_reg): Add rfe argument for IRQ returns.
2380 (arm_output_epilogue): Pop interrupt return address directly into PC.
2381 (arm_expand_prologue): Only adjust IRQ return address in Arm mode.
2382
2383 2007-01-03 Paul Brook <paul@codesourcery.com>
2384
2385 Merge from sourcerygxx-4_1.
2386 * config/arm/thumb2.md: New file.
2387 * config/arm/elf.h (JUMP_TABLES_IN_TEXT_SECTION): Return True for
2388 Thumb-2.
2389 * config/arm/coff.h (JUMP_TABLES_IN_TEXT_SECTION): Ditto.
2390 * config/arm/aout.h (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
2391 (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump tables.
2392 * config/arm/aof.h (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump
2393 tables.
2394 (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
2395 * config/arm/ieee754-df.S: Use macros for Thumb-2/Unified asm
2396 comptibility.
2397 * config/arm/ieee754-sf.S: Ditto.
2398 * config/arm/arm.c (thumb_base_register_rtx_p): Rename...
2399 (thumb1_base_register_rtx_p): ... to this.
2400 (thumb_index_register_rtx_p): Rename...
2401 (thumb1_index_register_rtx_p): ... to this.
2402 (thumb_output_function_prologue): Rename...
2403 (thumb1_output_function_prologue): ... to this.
2404 (thumb_legitimate_address_p): Rename...
2405 (thumb1_legitimate_address_p): ... to this.
2406 (thumb_rtx_costs): Rename...
2407 (thumb1_rtx_costs): ... to this.
2408 (thumb_compute_save_reg_mask): Rename...
2409 (thumb1_compute_save_reg_mask): ... to this.
2410 (thumb_final_prescan_insn): Rename...
2411 (thumb1_final_prescan_insn): ... to this.
2412 (thumb_expand_epilogue): Rename...
2413 (thumb1_expand_epilogue): ... to this.
2414 (arm_unwind_emit_stm): Rename...
2415 (arm_unwind_emit_sequence): ... to this.
2416 (thumb2_legitimate_index_p, thumb2_legitimate_address_p,
2417 thumb1_compute_save_reg_mask, arm_dwarf_handle_frame_unspec,
2418 thumb2_index_mul_operand, output_move_vfp, arm_shift_nmem,
2419 arm_save_coproc_regs, thumb_set_frame_pointer, arm_print_condition,
2420 thumb2_final_prescan_insn, thumb2_asm_output_opcode, arm_output_shift,
2421 thumb2_output_casesi): New functions.
2422 (TARGET_DWARF_HANDLE_FRAME_UNSPEC): Define.
2423 (FL_THUMB2, FL_NOTM, FL_DIV, FL_FOR_ARCH6T2, FL_FOR_ARCH7,
2424 FL_FOR_ARCH7A, FL_FOR_ARCH7R, FL_FOR_ARCH7M, ARM_LSL_NAME,
2425 THUMB2_WORK_REGS): Define.
2426 (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv, arm_condexec_count,
2427 arm_condexec_mask, arm_condexec_masklen)): New variables.
2428 (all_architectures): Add armv6t2, armv7, armv7a, armv7r and armv7m.
2429 (arm_override_options): Check new CPU capabilities.
2430 Set new architecture flag variables.
2431 (arm_isr_value): Handle v7m interrupt functions.
2432 (user_return_insn): Return 0 for v7m interrupt functions. Handle
2433 Thumb-2.
2434 (const_ok_for_arm): Handle Thumb-2 constants.
2435 (arm_gen_constant): Ditto. Use movw when available.
2436 (arm_function_ok_for_sibcall): Return false for v7m interrupt
2437 functions.
2438 (legitimize_pic_address, arm_call_tls_get_addr): Handle Thumb-2.
2439 (thumb_find_work_register, arm_load_pic_register,
2440 legitimize_tls_address, arm_address_cost, load_multiple_sequence,
2441 emit_ldm_seq, emit_stm_seq, arm_select_cc_mode, get_jump_table_size,
2442 print_multi_reg, output_mov_long_double_fpa_from_arm,
2443 output_mov_long_double_arm_from_fpa, output_mov_double_fpa_from_arm,
2444 output_mov_double_fpa_from_arm, output_move_double,
2445 arm_compute_save_reg_mask, arm_compute_save_reg0_reg12_mask,
2446 output_return_instruction, arm_output_function_prologue,
2447 arm_output_epilogue, arm_get_frame_offsets, arm_regno_class,
2448 arm_output_mi_thunk, thumb_set_return_address): Ditto.
2449 (arm_expand_prologue): Handle Thumb-2. Use arm_save_coproc_regs.
2450 (arm_coproc_mem_operand): Allow POST_INC/PRE_DEC.
2451 (arithmetic_instr, shift_op): Use arm_shift_nmem.
2452 (arm_print_operand): Use arm_print_condition. Handle '(', ')', '.',
2453 '!' and 'L'.
2454 (arm_final_prescan_insn): Use extract_constrain_insn_cached.
2455 (thumb_expand_prologue): Use thumb_set_frame_pointer.
2456 (arm_file_start): Output directive for unified syntax.
2457 (arm_unwind_emit_set): Handle stack alignment instruction.
2458 * config/arm/lib1funcs.asm: Remove default for __ARM_ARCH__.
2459 Add v6t2, v7, v7a, v7r and v7m.
2460 (RETLDM): Add Thumb-2 code.
2461 (do_it, shift1, do_push, do_pop, COND, THUMB_SYNTAX): New macros.
2462 * config/arm/arm.h (TARGET_CPU_CPP_BUILTINS): Define __thumb2__.
2463 (TARGET_THUMB1, TARGET_32BIT, TARGET_THUMB2, TARGET_DSP_MULTIPLY,
2464 TARGET_INT_SIMD, TARGET_UNIFIED_ASM, ARM_FT_STACKALIGN, IS_STACKALIGN,
2465 THUMB2_TRAMPOLINE_TEMPLATE, TRAMPOLINE_ADJUST_ADDRESS,
2466 ASM_OUTPUT_OPCODE, THUMB2_GO_IF_LEGITIMATE_ADDRESS,
2467 THUMB2_LEGITIMIZE_ADDRESS, CASE_VECTOR_PC_RELATIVE,
2468 CASE_VECTOR_SHORTEN_MODE, ADDR_VEC_ALIGN, ASM_OUTPUT_CASE_END,
2469 ADJUST_INSN_LENGTH): Define.
2470 (TARGET_REALLY_IWMMXT, TARGET_IWMMXT_ABI, CONDITIONAL_REGISTER_USAGE,
2471 STATIC_CHAIN_REGNUM, HARD_REGNO_NREGS, INDEX_REG_CLASS,
2472 BASE_REG_CLASS, MODE_BASE_REG_CLASS, SMALL_REGISTER_CLASSES,
2473 PREFERRED_RELOAD_CLASS, SECONDARY_OUTPUT_RELOAD_CLASS,
2474 SECONDARY_INPUT_RELOAD_CLASS, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P,
2475 TRAMPOLINE_SIZE, INITIALIZE_TRAMPOLINE, HAVE_PRE_INCREMENT,
2476 HAVE_POST_DECREMENT, HAVE_PRE_DECREMENT, HAVE_PRE_MODIFY_DISP,
2477 HAVE_POST_MODIFY_DISP, HAVE_PRE_MODIFY_REG, HAVE_POST_MODIFY_REG,
2478 REGNO_MODE_OK_FOR_BASE_P, LEGITIMATE_CONSTANT_P,
2479 REG_MODE_OK_FOR_BASE_P, REG_OK_FOR_INDEX_P, GO_IF_LEGITIMATE_ADDRESS,
2480 LEGITIMIZE_ADDRESS, THUMB2_LEGITIMIZE_ADDRESS,
2481 GO_IF_MODE_DEPENDENT_ADDRESS, MEMORY_MOVE_COST, BRANCH_COST,
2482 ASM_APP_OFF, ASM_OUTPUT_CASE_LABEL, ARM_DECLARE_FUNCTION_NAME,
2483 FINAL_PRESCAN_INSN, PRINT_OPERAND_PUNCT_VALID_P,
2484 PRINT_OPERAND_ADDRESS): Adjust for Thumb-2.
2485 (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv): New declarations.
2486 * config/arm/arm-cores.def: Add arm1156t2-s, cortex-a8, cortex-r4 and
2487 cortex-m3.
2488 * config/arm/arm-tune.md: Regenerate.
2489 * config/arm/arm-protos.h: Update prototypes.
2490 * config/arm/vfp.md: Enable patterns for Thumb-2.
2491 (arm_movsi_vfp): Add movw alternative. Use output_move_vfp.
2492 (arm_movdi_vfp, movsf_vfp, movdf_vfp): Use output_move_vfp.
2493 (thumb2_movsi_vfp, thumb2_movdi_vfp, thumb2_movsf_vfp,
2494 thumb2_movdf_vfp, thumb2_movsfcc_vfp, thumb2_movdfcc_vfp): New.
2495 * config/arm/libunwind.S: Add Thumb-2 code.
2496 * config/arm/constraints.md: Update include Thumb-2.
2497 * config/arm/ieee754-sf.S: Add Thumb-2/Unified asm support.
2498 * config/arm/ieee754-df.S: Ditto.
2499 * config/arm/bpabi.S: Ditto.
2500 * config/arm/t-arm (MD_INCLUDES): Add thumb2.md.
2501 * config/arm/predicates.md (low_register_operand,
2502 low_reg_or_int_operand, thumb_16bit_operator): New.
2503 (thumb_cmp_operand, thumb_cmpneg_operand): Rename...
2504 (thumb1_cmp_operand, thumb1_cmpneg_operand): ... to this.
2505 * config/arm/t-arm-elf: Add armv7 multilib.
2506 * config/arm/arm.md: Update patterns for Thumb-2 and Unified asm.
2507 Include thumb2.md.
2508 (UNSPEC_STACK_ALIGN, ce_count): New.
2509 (arm_incscc, arm_decscc, arm_umaxsi3, arm_uminsi3,
2510 arm_zero_extendsidi2, arm_zero_extendqidi2): New
2511 insns/expanders.
2512 * config/arm/fpa.md: Update patterns for Thumb-2 and Unified asm.
2513 (thumb2_movsf_fpa, thumb2_movdf_fpa, thumb2_movxf_fpa,
2514 thumb2_movsfcc_fpa, thumb2_movdfcc_fpa): New insns.
2515 * config/arm/cirrus.md: Update patterns for Thumb-2 and Unified asm.
2516 (cirrus_thumb2_movdi, cirrus_thumb2_movsi_insn,
2517 thumb2_cirrus_movsf_hard_insn, thumb2_cirrus_movdf_hard_insn): New
2518 insns.
2519 * doc/extend.texi: Document ARMv7-M interrupt functions.
2520 * doc/invoke.texi: Document Thumb-2 new cores+architectures.
2521
2522 2007-01-03 Jakub Jelinek <jakub@redhat.com>
2523
2524 * unwind-dw2.c (SIGNAL_FRAME_BIT, EXTENDED_CONTEXT_BIT): Define.
2525 (struct _Unwind_Context): Rename args_size to flags, remove
2526 signal_frame field, add a new args_size field and version field.
2527 (_Unwind_IsSignalFrame, _Unwind_SetSignalFrame,
2528 _Unwind_IsExtendedContext): New inline functions.
2529 (_Unwind_GetGR, _Unwind_SetGR, _Unwind_GetGRPtr, _Unwind_SetGRPtr):
2530 Assume by_value array is only present if _Unwind_IsExtendedContext.
2531 (_Unwind_GetIPInfo, execute_cfa_program, uw_frame_state_for): Use
2532 _Unwind_IsSignalFrame.
2533 (__frame_state_for): Initialize context.flags to EXTENDED_CONTEXT_BIT.
2534 (uw_update_context_1): Use _Unwind_SetSignalFrame.
2535 (uw_init_context_1): Initialize context->flags to
2536 EXTENDED_CONTEXT_BIT.
2537 * config/rs6000/linux-unwind.h (frob_update_context): Use
2538 _Unwind_SetSignalFrame.
2539
2540 2007-01-03 Andrew Pinski <andrew_pinski@playstation.sony.com>
2541
2542 PR middle-end/30353
2543 * gimplify.c (gimplify_modify_expr_complex_part): Move below
2544 tree_to_gimple_tuple. Call tree_to_gimple_tuple when we need
2545 the value.
2546
2547 2007-01-03 Kazu Hirata <kazu@codesourcery.com>
2548
2549 * config/i386/i386.h (NON_STACK_REG_P, REGNO_OK_FOR_SIREG_P,
2550 REGNO_OK_FOR_DIREG_P, REWRITE_ADDRESS, ASM_OPERAND_LETTER,
2551 RET, AT_SP): Remove.
2552 * config/i386/i386.md (*sse_prologue_save_insn): Use return
2553 instead of RET.
2554
2555 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
2556 * cfgbuild.c (find_basic_blocks): Likewise.
2557 * cfgrtl.c (rtl_create_basic_block): Likewise.
2558 * function.c (temp_slots_at_level): Likewise.
2559 * reg-stack.c (stack_regs_mentioned): Likewise.
2560 * regclass.c (allocate_reg_info): Likewise.
2561 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
2562 set_bb_for_stmt, move_block_to_fn): Likewise.
2563 * tree-complex.c (tree_lower_complex): Likewise.
2564 * vec.h (VEC_safe_grow_cleared): New.
2565
2566 * cgraphunit.c, tree-ssa-alias.c: Fix comment typos.
2567
2568 2007-01-03 Zdenek Dvorak <dvorakz@suse.cz>
2569
2570 * loop-unswitch.c (unswitch_loop): Pass probabilities to loopify.
2571 * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Pass probabilities
2572 to loop_version.
2573 * cfgloopmanip.c (scale_loop_frequencies): Export.
2574 (loopify): Scale the frequencies by prescribed coefficients.
2575 (set_zero_probability): New function.
2576 (duplicate_loop_to_header_edge): Improve updating of frequencies.
2577 (lv_adjust_loop_entry_edge, loop_version): Set probabilities
2578 and frequencies according to arguments.
2579 * tree-ssa-loop-manip.c (tree_unroll_loop): Set probabilities
2580 correctly.
2581 * cfg.c (scale_bbs_frequencies_int): Allow scaling the frequencies up.
2582 * modulo-sched.c (sms_schedule): Set probabilities for entering
2583 versioned loop correctly.
2584 * tree-vect-transform.c (vect_transform_loop): Ditto.
2585 * cfgloop.h (loopify, loop_version): Declaration changed.
2586 (scale_loop_frequencies): Declared.
2587
2588 2007-01-02 Jan Hubicka <jh@suse.cz>
2589
2590 * cgraph.c: Include tree-flow.h
2591 (cgraph_add_new-function): Handle IPA_SSA mode; execute
2592 early_local_passes.
2593 * cgraph.h (enum cgraph_state): Add CGRAPH_STATE_IPA_SSA.
2594 * tree-pass.h (pass_all_early_optimizations): Declare.
2595 * cgraphunit.c (cgraph_process_new_functions): Add IPA_SSA; execute
2596 early_local_passes.
2597 (cgraph_analyze_function): Do early_local_passes.
2598 * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
2599 Do not add referenced vars.
2600 * tree-optimize.c (gate_all_optimizations): Do not execute when not in
2601 SSA form.
2602 (gate_all_early_local_passes): New gate.
2603 (pass_early_local_passes): Use new gate.
2604 (execute_early_local_optimizations): New functions.
2605 (gate_all_early_optimizations): New gate.
2606 (pass_all_early_optimizations): New pass.
2607 (execute_free_datastructures): Free SSA only when initialized.
2608 (gate_init_datastructures): Init only when optimizing.
2609 (tree_lowering_passes): Do early local passes when called late.
2610 * tree-profile.c (do_tree_profiling): Don't profile functions added
2611 late.
2612 (do_early_tree_profiling, pass_early_tree_profile): Kill.
2613 * tree-cfg.c (update_modified_stmts): Do not update when operands are
2614 not active.
2615 * passes.c (init_optimizations_passes): Reorder so we go into SSA
2616 during early_local_passes.
2617 * Makefile.in (cgraph.o): Add dependency on tree-flow.h.
2618
2619
2620 2007-01-02 Carlos O'Donell <carlos@codesourcery.com>
2621
2622 * Makefile.in: Update copyright year.
2623
2624 2007-01-02 Carlos O'Donell <carlos@codesourcery.com>
2625
2626 * Makefile.in: Export GCC_EXEC_PREFIX before calling $(RUNTEST)
2627 in $(lang_checks) and check-consistency targets.
2628
2629 2007-01-02 Jan Hubicka <jh@suse.cz>
2630
2631 * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
2632 Do not add referenced vars.
2633 * tree-cfg.c (update_modified_stmts): Do not update when SSA operands
2634 are not active.
2635 * passes.c (init_optimization_passes): Put mudflap_2 after
2636 free_datastructures.
2637
2638 2007-01-02 Jan Hubicka <jh@suse.cz>
2639
2640 * tree-optimize (execute_fixup_cfg): Set after_inlining flag.
2641 Set NOTHROW flag on call statements proved to be nothrow.
2642 Update statement of local calls so new pure/const functions are
2643 updated. Update_ssa when in ssa form. Mark PHI nodes of nonlocal
2644 goto receivers.
2645 (tree_rest_of_compilation): Register hooks and initialize bitmap
2646 early. Do not set after_inlining flag.
2647
2648 2007-01-02 Steve Ellcey <sje@cup.hp.com>
2649
2650 * sbitmap.c (HOST_BITS_PER_LONG_LONG): Change to
2651 HOST_BITS_PER_LONGLONG
2652
2653 2007-01-02 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
2654
2655 PR c/19977
2656 * c-typeck.c (store_init_value): Don't emit pedantic overflow
2657 warning for non-static initializers.
2658
2659 2007-01-02 Steven Bosscher <steven@gcc.gnu.org>
2660
2661 * config/alpha/alpha.md, arm/arm.c, darwin.c, frv/frv.md,
2662 m32r/m32r.c, m32r/m32r.c, mn10300/mn10300.md, pa/pa.c,
2663 rs6000/rs6000.c, s390/s390.md, sh/sh.md, sparc/sparc.c:
2664 Always use set_unique_reg_note to add REG_EQUAL notes.
2665
2666 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
2667
2668 Revert:
2669 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
2670
2671 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
2672 * cfgbuild.c (find_basic_blocks): Likewise.
2673 * cfgrtl.c (rtl_create_basic_block): Likewise.
2674 * function.c (temp_slots_at_level): Likewise.
2675 * reg-stack.c (stack_regs_mentioned): Likewise.
2676 * regclass.c (allocate_reg_info): Likewise.
2677 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
2678 set_bb_for_stmt, move_block_to_fn): Likewise.
2679 * tree-complex.c (tree_lower_complex): Likewise.
2680 * vec.h (VEC_safe_grow_cleared): New.
2681
2682 2007-01-02 Ian Lance Taylor <iant@google.com>
2683
2684 * c-common.c (c_common_truthvalue_conversion): When warning about
2685 using an assignment as a truth value, set TREE_NO_WARNING.
2686
2687 2007-01-02 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
2688
2689 PR middle-end/7651
2690 * c.opt (Wold-style-declaration): New.
2691 * doc/invoke.texi (C-only Warning Options): New.
2692 (Wold-style-declaration): Document it.
2693 (Wextra): Enabled by -Wextra.
2694 * c-opts.c (c_common_post_options): Enabled by -Wextra.
2695 * c-decl.c (declspecs_add_scspec): Replace -Wextra with
2696 -Wold-style-declaration.
2697
2698 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
2699
2700 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
2701 * cfgbuild.c (find_basic_blocks): Likewise.
2702 * cfgrtl.c (rtl_create_basic_block): Likewise.
2703 * function.c (temp_slots_at_level): Likewise.
2704 * reg-stack.c (stack_regs_mentioned): Likewise.
2705 * regclass.c (allocate_reg_info): Likewise.
2706 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
2707 set_bb_for_stmt, move_block_to_fn): Likewise.
2708 * tree-complex.c (tree_lower_complex): Likewise.
2709 * vec.h (VEC_safe_grow_cleared): New.
2710
2711 2007-01-02 Douglas Gregor <doug.gregor@gmail.com>
2712
2713 * c-common.c(c_common_nodes_and_builtins): Since variants of
2714 void_type_node get built before it is given a name, we need to
2715 give those variants the name, too.
2716 (complete_array_type): We need to work with the canonical main
2717 type of the array, from which we will build the qualified version.
2718 * params.def (PARAM_VERIFY_CANONICAL_TYPES): New.
2719 * print-tree.c (print_node): Display canonical type information
2720 for each type.
2721 * stor-layout.c (layout_type): When we don't know the
2722 alignment of a type for which we're building an array, we end up
2723 guessing wrong, so make the type require structural equality.
2724 * tree.c (make_node_stat): When we build a new type, it is its
2725 own canonical type.
2726 (build_type_attribute_qual_variant): When building an attribute
2727 variant, its canonical type is the non-attribute variant. However,
2728 if the attributes are target-dependent and they differ, we need to
2729 use structural equality checks for this type.
2730 (build_qualified_type): A qualified type is not equivalent to its
2731 unqualified variant; set the canonical type appropriately.
2732 (build_distinct_type_copy): When building a distinct type from
2733 another type, the new type is its own canonical type.
2734 (build_variant_type_copy): When building a new type variant, we
2735 assume that it is equivalent to the original type.
2736 (build_pointer_type_for_mode): When building a pointer type, also
2737 build a canonical type pointer.
2738 (build_reference_type_for_mode): When building a reference type,
2739 also build a canonical type reference.
2740 (build_index_type): When we can't hash an index type (e.g.,
2741 because its maximum value is negative), the index type requires
2742 structural equality tests.
2743 (build_array_type): Build the canonical form of an array type.
2744 (build_function_type): Function types require structural equality,
2745 because they contain default arguments, attributes, etc.
2746 (build_method_type_directly): Ditto for method types.
2747 (build_offset_type): Build the canonical offset type.
2748 (build_complex_type): Build the canonical vector type.
2749 (make_vector_type): Build the canonical vector type.
2750 * tree.h (TYPE_CANONICAL): New.
2751 (TYPE_STRUCTURAL_EQUALITY_P): New.
2752 (SET_TYPE_STRUCTURAL_EQUALITY): New.
2753 (struct tree_type): Added "canonical" field.
2754 * params.h (VERIFY_CANONICAL_TYPES): New.
2755 * doc/c-tree.texi (TYPE_CANONICAL): Document.
2756 (TYPE_STRUCTURAL_EQUALITY_P): Document.
2757 (SET_TYPE_STRUCTURAL_EQUALITY): Document.
2758 * doc/invoke.texi (verify-canonical-types): Document --param
2759 parameter for verifying canonical types.
2760
2761 2007-01-02 Joseph Myers <joseph@codesourcery.com>
2762
2763 * config.gcc (powerpc-*-eabispe*, powerpc-*-eabisimaltivec*,
2764 powerpc-*-eabisim*, powerpc-*-eabialtivec*, powerpc-*-eabi*,
2765 powerpc-*-rtems*, powerpc-wrs-vxworks, powerpc-wrs-vxworksae,
2766 powerpcle-*-eabisim*, powerpcle-*-eabi*): Add rs6000/e500.h to
2767 tm_file.
2768 * config/rs6000/e500.h: New.
2769 * config/rs6000/eabi.h (TARGET_SPE_ABI, TARGET_SPE, TARGET_E500,
2770 TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE, TARGET_E500_DOUBLE):
2771 Remove.
2772 * config/rs6000/linuxspe.h (TARGET_SPE_ABI, TARGET_SPE,
2773 TARGET_E500, TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE,
2774 TARGET_E500_DOUBLE): Remove.
2775 * config/rs6000/vxworks.h (TARGET_SPE_ABI, TARGET_SPE,
2776 TARGET_E500, TARGET_ISEL, TARGET_FPRS): Remove.
2777 * config/rs6000/rs6000.h (CHECK_E500_OPTIONS): Define.
2778 * config/rs6000/rs6000.c (rs6000_override_options): Use
2779 CHECK_E500_OPTIONS.
2780
2781 2007-01-02 Joseph Myers <joseph@codesourcery.com>
2782
2783 * config/rs6000/rs6000.c (print_operand): Check (TARGET_SPE ||
2784 TARGET_E500_DOUBLE), not TARGET_E500, for %y.
2785 (rs6000_generate_compare, rs6000_emit_sCOND, output_cbranch,
2786 rs6000_emit_cmove): Don't check TARGET_E500.
2787 * config/rs6000/rs6000.md (bunordered, bordered, sunordered,
2788 sordered): Don't check TARGET_E500.
2789
2790 2007-01-01 Eric Christopher <echristo@apple.com>
2791
2792 * config/mips/mips.c (mips_regno_mode_ok_for_base_p): Use
2793 HARD_REGISTER_NUM_P.
2794
2795 2007-01-01 Roger Sayle <roger@eyesopen.com>
2796
2797 * fold-const.c (fold_binary) <EQ_EXPR>: Fold "(X^C1) eq/ne C2" into
2798 "X eq/ne (C1^C2)". Fold "(X^Z) eq/ne (Y^Z)" as "X eq/ne Y" when Z
2799 has no side-effects. Fold "(X^C1) eq/ne (Y^C2)" as "(X^(C1^C2))
2800 eq/ne Y".
2801
2802 2007-01-01 Mike Stump <mrs@apple.com>
2803
2804 * configure.ac: Remove support for building with Apple's gcc-3.1.
2805
2806 2007-01-02 Joseph Myers <joseph@codesourcery.com>
2807
2808 PR middle-end/30311
2809 * caller-save.c (add_stored_regs): Only handle SUBREGs if inner
2810 REG is a hard register. Do not modify REG before calling
2811 subreg_nregs.
2812 * rtlanal.c (subreg_get_info): Don't assert size of XMODE is a
2813 multiple of the size of YMODE for certain lowpart cases.
2814
2815 2007-01-01 Andrew Pinski <pinskia@gmail.com>
2816
2817 PR middle-end/30253
2818 * gimplify (voidify_wrapper_expr): Update for
2819 GIMPLE_MODIFY_STMT.
2820
2821 2007-01-01 Andreas Schwab <schwab@suse.de>
2822
2823 PR target/29166
2824 * config/ia64/ia64.c (ia64_compute_frame_size): Account space for
2825 save of BR0 in extra_spill_size instead of spill_size.
2826 (ia64_expand_prologue): Save BR0 outside of the gr/br/fr spill
2827 area.
2828 (ia64_expand_epilogue): Restore BR0 from its new location.
2829
2830 2007-01-01 Andrew Pinski <pinskia@gmail.com>
2831
2832 * gimplify.c (gimplify_init_constructor <case VECTOR_TYPE>):
2833 Use a temporary variable if the left hand side is not a gimple
2834 register.
2835
2836 2007-01-01 Andrew Pinski <pinskia@gmail.com>
2837
2838 * gimplify.c (gimplify_return_expr): Make the temporary variable
2839 for the return expression, a gimple register variable.
2840
2841 2007-01-01 Jan Hubicka <jh@suse.cz>
2842
2843 * emit-rtl.c (emit_copy_of_insn_after): Do not call copy_insn_1 for
2844 INSN_LIST.
2845
2846 2007-01-01 Mike Stump <mrs@apple.com>
2847
2848 * configure.ac (HAVE_GAS_LITERAL16): Add autoconf check for
2849 .literal16.
2850 * config/darwin.c (machopic_select_rtx_section): Use
2851 HAVE_GAS_LITERAL16.
2852 (darwin_mergeable_constant_section): Likewise.
2853 * configure: Regenerate.
2854 * config.in: Regenerate.
2855
2856 2007-01-01 Jan Hubicka <jh@suse.cz>
2857 Andrew Pinski <pinskia@gmail.com>
2858
2859 * cgraphunit.c (cgraph_optimize): Call cgraph_add_new_functions
2860 before starting IPA passes.