tree-ssa-address.c (create_mem_ref): Remove ", bsi" from a parts.base assignment.
[gcc.git] / gcc / ChangeLog
1 2007-01-26 Zdenek Dvorak <dvorakz@suse.cz>
2
3 * tree-ssa-address.c (create_mem_ref): Remove ", bsi" from
4 a parts.base assignment.
5
6 2007-01-26 Zdenek Dvorak <dvorakz@suse.cz>
7
8 * tree-data-ref.c (dump_subscript): Use dump_conflict_function.
9 (compute_subscript_distance, initialize_data_dependence_relation,
10 finalize_ddr_dependent, analyze_ziv_subscript,
11 analyze_siv_subscript_cst_affine,
12 compute_overlap_steps_for_affine_univar,
13 compute_overlap_steps_for_affine_1_2, analyze_subscript_affine_affine,
14 analyze_siv_subscript, analyze_miv_subscript,
15 analyze_overlapping_iterations, subscript_dependence_tester_1,
16 compute_self_dependence, free_dependence_relation): Work
17 with affine_fn instead of chrecs.
18 (dump_affine_function, dump_conflict_function, affine_function_equal_p,
19 common_affine_function, affine_function_base,
20 affine_function_constant_p, affine_fn_op, affine_fn_plus,
21 affine_fn_minus, affine_fn_free, conflict_fn_not_known,
22 conflict_fn_no_dependence, free_conflict_function, free_subscripts,
23 conflict_fn, affine_fn_cst, affine_fn_univar): New functions.
24 (all_chrecs_equal_p): Removed.
25 * tree-data-ref.h (affine_fn, conflict_function): New types.
26 (struct subscript): Change type of conflicting_iterations_in_a
27 and conflicting_iterations_in_b.
28
29 2007-01-26 Steve Ellcey <sje@cup.hp.com>
30
31 PR other/30182
32 * config/pa/pa.h (TARGET_HPUX_11): New.
33 * config/pa/pa-hpux11.h (TARGET_HPUX_11): New.
34 * config/pa/pa.c (pa_init_builtins): Use TARGET_HPUX_11.
35
36 2007-01-26 Daniel Berlin <dberlin@dberlin.org>
37 Richard Guenther <rguenther@suse.de>
38
39 * tree-ssa-structalias.c (solve_graph): Handle case
40 we merged the variable to another.
41
42 2007-01-25 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
43
44 * builtins.c (fold_builtin_1): Treat ccos and ccosh as 'even'
45 functions.
46
47 * fold-const.c (negate_mathfn_p): Treat casin, casinh, catan,
48 catanh, cproj, csin, csinh, ctan and ctanh as 'odd' functions.
49
50 2007-01-25 DJ Delorie <dj@redhat.com>
51
52 * config/m32c/m32c.c (m32c_cannot_change_mode_class): We don't
53 allow changes to modes which don't fit in those registers.
54
55 * reload1.c (choose_reload_regs): Check for invalid subregs before
56 computing their locations, not after.
57
58 2007-01-25 Geoffrey Keating <geoffk@apple.com>
59
60 PR 25127
61 * config/rs6000/rs6000.c (first_altivec_reg_to_save): On Darwin,
62 save Altivec registers in an eh_return function.
63 (compute_vrsave_mask): Likewise.
64 (rs6000_stack_info): Correct AIX/Darwin stack alignment computation
65 for saving Altivec registers.
66 (rs6000_emit_prologue): Don't allocate stack twice in
67 eh_return function. Correct expected value of altivec_save_offset
68 when using save_world. Describe save of R0 to stack when using
69 save_world. Describe stack pointer adjustment when using
70 save_world. Remove duplicated eh_return parameter register saving.
71 Update sp_offset variable after save_world.
72 * config/rs6000/t-darwin (LIB2FUNCS_STATIC_EXTRA): Remove
73 darwin-world.asm.
74 (LIB2FUNCS_EXTRA): Add darwin-world.asm.
75 * config/rs6000/darwin.h (SUBTARGET_OVERRIDE_OPTIONS): -m64
76 implies Altivec.
77
78 2007-01-25 Steve Ellcey <sje@cup.hp.com>
79
80 * config.gcc (ia64*-*-hpux*): Make posix threads the default.
81
82 2007-01-25 Steve Ellcey <sje@cup.hp.com>
83
84 PR other/30182
85 * config/pa/pa.c (pa_init_builtins): Set asm names for finite routines.
86 * config/ia64/ia64.c (ia64_init_builtins): Ditto.
87
88 2007-01-25 Richard Guenther <rguenther@suse.de>
89
90 * doc/invoke.texi (-Wcoverage-mismatch): Document.
91 * common.opt (-Wcoverage-mismatch): New warning option.
92 * coverage.c (get_coverage_counts): Ignore coverage mismatch
93 if -Wcoverage-mismatch is given.
94
95 2007-01-25 Razya Ladelsky <razya@il.ibm.com>
96
97 * ipa-cp.c (ipcp_insert_stage, ipcp_driver): Support for SSA.
98 (ipcp_driver): Change to static definition.
99 Add dumping of the ifunctions.
100 (constant_val_insert): Remove unused parameter. Support for SSA.
101 (ipcp_propagate_const): Support for SSA.
102 (ipcp_profile_bb_print): Print only analyzed nodes.
103 (ipcp_replace_map_create): Remove support for Fortran constant
104 for now.
105 * ipa-prop.c (ipa_method_modify_stmt,
106 ipa_callsite_compute_param): Support for SSA.
107 * ipa-prop.h (ipcp_driver): Remove declaration.
108 (IS_VALID_TREE_MAP_INDEX): Add define.
109
110 2007-01-24 Geoffrey Keating <geoffk@apple.com>
111
112 * unwind-dw2.c (execute_stack_op): Handle DW_OP_swap.
113
114 2007-01-24 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
115
116 PR middle-end/30447
117 * builtins.c (fold_builtin_cabs): Use MPFR to evaluate a
118 constant argument to cabs and do it without checking for
119 -funsafe-math-optimizations.
120
121 2007-01-24 Douglas Gregor <dgregor@osl.iu.edu>
122
123 * c-common.h (RID_FIRST_CXX0X): New.
124 (RID_LAST_CXX0X): New.
125 * c-opts.c (c_common_handle_option): -Wc++0x-compat is triggered
126 by -Wall.
127 * c.opt (Wc++0x-compat): New.
128 * doc/invoke.texi (-Wc++0x-compat): Document.
129
130 2007-01-24 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
131
132 * builtins.c (fold_builtin_carg): New.
133 (fold_builtin_1): Use it.
134
135 2007-01-24 Jan Hubicka <jh@suse.cz>
136
137 * ipa-inline.c (cgraph_decide_inlining): Initialize initial_insns.
138
139 * ipa-inline.c (initial_insns, max_insns): Delete.
140 (compute_max_insns): New function.
141 (cgraph_decide_inlining_of_small_function): Use it; take minimal amount
142 of insns as base for code growth.
143 (cgraph_decide_inlining): Make initial_insns local; do not compute
144 max_insns.
145 * params.def (PARAM_INLINE_UNIT_GROWTH): Set to 60.
146 * doc/invoke.texi (inline-unit-growth): Update docs.
147
148 2007-01-24 Jakub Jelinek <jakub@redhat.com>
149
150 * config/i386/i386.h (x86_cmpxchg16b): Remove const.
151 (TARGET_CMPXCHG16B): Define to x86_cmpxchg16b.
152 * config/i386/i386.c (x86_cmpxchg16b): Remove const.
153 (override_options): Add PTA_CX16 flag. Set x86_cmpxchg16b
154 for CPUs that have PTA_CX16 set.
155
156 PR middle-end/27416
157 * gimplify.c (omp_check_private): New function.
158 (gimplify_scan_omp_clauses): Use it for
159 firstprivate/lastprivate/reduction.
160
161 PR middle-end/30494
162 * gimplify.c (omp_add_variable): Don't call omp_notice_variable
163 on TYPE_SIZE_UNIT for GOVD_LOCAL VLAs.
164
165 PR middle-end/30421
166 * omp-low.c (lower_omp_for_lastprivate): Add dlist argument.
167 If lower_lastprivate_clauses emits some statements, append them
168 to dlist rather than body_p and to body_p append an initializer.
169 (lower_omp_for): Adjust caller.
170
171 2007-01-24 Steve Ellcey <sje@cup.hp.com>
172
173 * target.h (globalize_decl_name): New.
174 * target-def.h (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
175 * output.h (default_globalize_decl_name): New.
176 * varasm.c (asm_output_bss): Use globalize_decl_name instead of
177 globalize_label.
178 (globalize_decl): Ditto.
179 (default_globalize_decl_name): New.
180 * config/ia64/ia64.c (ia64_globalize_decl_name): New.
181 (ia64_handle_version_id_attribute): New.
182 (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
183 (ia64_asm_output_external): Use globalize_decl_name instead
184 of globalize_label.
185 * doc/extend.texi (version_id): New pragma.
186 * doc/tm.texi (ARGET_ASM_GLOBALIZE_DECL_NAME): New target hook.
187
188 2007-01-24 Andreas Krebbel <krebbel1@de.ibm.com>
189
190 * unwind-dw2-fde.c (get_cie_encoding): Replaced _Unwind_Word with
191 _uleb128_t and _Unwind_SWord with _sleb128_t.
192 * unwind-dw2.c (extract_cie_info, execute_stack_op, execute_cfa_program,
193 uw_frame_state_for, uw_update_context_1): Likewise.
194 * unwind-c.c (parse_lsda_header, PERSONALITY_FUNCTION): Likewise.
195 * unwind-pe.h (read_uleb128, read_sleb128,
196 read_encoded_value_with_base): Likewise.
197 * unwind-generic.h: Define _sleb128_t and _uleb128_t types.
198
199 2007-01-24 Richard Guenther <rguenther@suse.de>
200
201 * builtins.c (expand_builtin_cexpi): Get the fndecl
202 for cexp in the correct way.
203
204 2007-01-24 Jan Hubicka <jh@suse.cz>
205
206 * tree-ssa-dce.c (eliminate_unnecesary_stmts): Remove dead LHS of calls.
207
208 2007-01-24 Andreas Krebbel <krebbel1@de.ibm.com>
209
210 * c-cppbuiltin.c (builtin_define_type_sizeof): New function.
211 (c_cpp_builtins): New builtin macros: __SIZEOF_INT__, __SIZEOF_LONG__,
212 __SIZEOF_LONG_LONG__, __SIZEOF_SHORT__, __SIZEOF_POINTER__,
213 __SIZEOF_FLOAT__, __SIZEOF_DOUBLE__, __SIZEOF_LONG_DOUBLE__,
214 __SIZEOF_SIZE_T__, __SIZEOF_WCHAR_T__, __SIZEOF_WINT_T__ and
215 __SIZEOF_PTRDIFF_T__.
216 * doc/cpp.texi: Documentation for the new builtin macros added.
217
218 2007-01-24 Uros Bizjak <ubizjak@gmail.com>
219
220 * config/i386/i386.md (tanxf2, tan<mode>2, atan<mode>2, log<mode>2,
221 log10<mode>2, log2<mode>2, expxf2, exp10xf2, exp2xf2): Use op2
222 instead of operands[2] to avoid access past the end of array.
223
224 2007-01-24 Richard Sandiford <richard@codesourcery.com>
225
226 * reload1.c (emit_reload_insns): Pass the reload register
227 for a non-spill output reload through forget_old_reloads_1.
228
229 2007-01-23 Joseph Myers <joseph@codesourcery.com>
230
231 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Define
232 _SOFT_DOUBLE if doubles use software floating-point.
233 * config/rs6000/libgcc-ppc-glibc.ver: Export additional long
234 double functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
235 * config/rs6000/darwin-ldouble.c: Also compile functions for
236 hard-float without FPRs. Use fmsub function for all __NO_FPRS__
237 cases. Compile extra functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
238 * config/rs6000/linuxspe.h (SUBSUBTARGET_OVERRIDE_OPTIONS): Remove
239 commented-out long double override.
240 (CPP_LONGDOUBLE_DEFAULT_SPEC): Likewise.
241 * config/rs6000/eabispe.h: Likewise.
242 * config/rs6000/rs6000.c (rs6000_override_options): Don't override
243 long double for non-SPE.
244 (rs6000_handle_option): Likewise.
245 (invalid_e500_subreg): Disallow more subregs involding DImode,
246 DFmode, TImode or TFmode.
247 (rs6000_legitimate_offset_address_p): Check TFmode offsets for
248 E500 double.
249 (legitimate_lo_sum_address_p): Also check for TFmode for E500
250 double.
251 (rs6000_legitimize_address): Also handle TFmode for E500 double.
252 (rs6000_legitimize_reload_address): Also handle TFmode for E500
253 double.
254 (rs6000_legitimate_address): Also check for TFmode for E500
255 double.
256 (rs6000_emit_move): Use DFmode subregs of TFmode for E500 double.
257 (spe_build_register_parallel): Handle TFmode and TCmode.
258 (rs6000_spe_function_arg): Handle TFmode and TCmode for E500
259 double.
260 (function_arg): Handle TFmode and TCmode for E500 double.
261 (rs6000_init_libfuncs): Initialize extra libfuncs for soft double
262 in general.
263 (print_operand): Handle TFmode and TImode for %y.
264 (rs6000_generate_compare): Handle TFmode comparisons for E500
265 double.
266 (spe_func_has_64bit_regs_p): Check for TFmode for E500 double.
267 (rs6000_function_value): Handle TFmode and TCmode for E500 double.
268 (rs6000_libcall_value): Handle TFmode and TCmode for E500 double.
269 * config/rs6000/rs6000.h (CANNOT_CHANGE_MODE_CLASS): Check for
270 TFmode for E500 double.
271 * config/rs6000/rs6000.md (FP): Allow TF for E500 double.
272 (floatsidf2): Enable for E500 double.
273 (movtf_softfloat): Use rs6000_nonimmediate_operand.
274 (extenddftf2): Change to extenddftf2_fprs.
275 (extenddftf2): Call gen_spe_extenddftf2 or gen_extenddftf2_fprs
276 depending on TARGET_E500_DOUBLE.
277 (extendsftf2): Enable for E500 double.
278 (trunctfdf2): Enable for E500 double.
279 (trunctfsf2): Change to trunctfsf2_fprs.
280 (trunctfsf2): Call gen_spe_trunctfsf2 or gen_trunctfsf2_fprs
281 depending on TARGET_E500_DOUBLE.
282 (floatsitf2): Enable for E500 double.
283 (fix_trunctfsi2): Change to fix_trunctfsi2_fprs.
284 (fix_trunctfsi2): Call gen_spe_fix_trunctfsi2 or
285 gen_fix_trunctfsi2_fprs depending on TARGET_E500_DOUBLE.
286 (negtf2): Change to negtf2_internal.
287 (negtf2): New expander.
288 (abstf2): Enable for E500 double. Call gen_spe_abstf2_tst,
289 gen_spe_abstf2_cmp or gen_abstf2_internal depending on
290 TARGET_E500_DOUBLE and flag_unsafe_math_optimizations.
291 (movdi_internal32): Use rs6000_nonimmediate_operand.
292 (unnamed splitter): Likewise.
293 * config/rs6000/spe.md (CMPTFEQ_GPR, TSTTFEQ_GPR, CMPTFGT_GPR,
294 TSTTFGT_GPR, CMPTFLT_GPR, TSTTFLT_GPR): New unspecs.
295 (SPE64TF, DITI): New mode macros.
296 (frob_df_di): Change to frob_<SPE64:mode>_<DITI:mode>; allow more
297 modes.
298 (frob_tf_ti): New.
299 (frob_<mode>_di_2): New.
300 (frob_tf_di_8_2): New.
301 (frob_di_df): Change to frob_di_<mode>; allow more modes.
302 (frob_ti_tf): New.
303 (frob_di_df_2): Change to frob_<DITI:mode>_<SPE64:mode>_2; allow
304 more modes.
305 (frob_ti_<mode>_8_2): New.
306 (frob_ti_tf_2): New.
307 (mov_si<mode>_e500_subreg0, mov_si<mode>_e500_subreg0_2,
308 mov_si<mode>_e500_subreg4, mov_si<mode>_e500_subreg4_2): Allow
309 TFmode.
310 (mov_sitf_e500_subreg8, mov_sitf_e500_subreg8_2,
311 mov_sitf_e500_subreg12, mov_sitf_e500_subreg12_2): New.
312 (spe_trunctfdf2_internal1, spe_trunctfsf2, spe_extenddftf2,
313 spe_fix_trunctfsi2, spe_fix_trunctfsi2_internal,
314 spe_negtf2_internal, spe_abstf2_cmp, spe_abstf2_tst): New.
315 (cmptfeq_gpr, tsttfeq_gpr, cmptfgt_gpr, tsttfgt_gpr, cmptflt_gpr,
316 tsttflt_gp): New.
317
318 2007-01-23 Ian Lance Taylor <iant@google.com>
319
320 * Makefile.in (OBJS-common): Reformat, alphabetize, but put
321 insn-*.o first.
322 (OBJS-archive): Reformat, alphabetize.
323 (OBJS): Change out_object_file to OBJS-md.
324
325 2007-01-23 Uros Bizjak <ubizjak@gmail.com>
326
327 * config/i386/i386.md: Use REG_P, MEM_P, CONST_INT_P, LABEL_P,
328 JUMP_P and CALL_P predicates where applicable.
329 * config/i386/i386.c: Ditto.
330 * config/i386/i386.md: Ditto.
331 * config/i386/mmx.md: Ditto.
332 * config/i386/predicates.md: Ditto.
333
334 2007-01-22 Andreas Schwab <schwab@suse.de>
335
336 * config/m68k/m68k.h: Fix comment.
337
338 2007-01-22 Jan Hubicka <jh@suse.cz>
339
340 * passes.c (init_optimization_passes): Do not rerun
341 pass_early_warn_uninitialized.
342
343 2007-01-22 Richard Guenther <rguenther@suse.de>
344
345 PR tree-optimization/30038
346 * tree-ssa-math-opts.c (maybe_record_sincos): New static helper
347 function.
348 (execute_cse_sincos_1): Likewise.
349 (execute_cse_sincos): Likewise.
350 (gate_cse_sincos): Likewise.
351 (pass_cse_sincos): New pass CSEing sin() and cos() calls using
352 the cexpi() canonicalization of sincos().
353 * tree-pass.h (pass_cse_sincos): Declare.
354 * passes.c (init_optimization_passes): New pass pas_cse_sincos.
355
356 2007-01-21 Eric Botcazou <ebotcazou@libertysurf.fr>
357
358 PR rtl-optimization/29329
359 * combine.c (replaced_rhs_insn): Rename to i2mod.
360 (replaced_rhs_value): Rename to i2mod_new_rhs.
361 (i2mod_old_rhs): New global variable.
362 (combine_instructions): Adjust for above change. Save a copy of
363 the old RHS into i2mod_old_rhs when the contents of a REG_EQUAL
364 note are substituted in the second instruction.
365 (distribute_notes) <REG_DEAD>: Adjust for above change. Do not
366 ditch the note if it pertains to the second eliminated register
367 and this register is mentioned in i2mod_old_rhs.
368
369 Revert:
370 2006-09-12 Eric Botcazou <ebotcazou@libertysurf.fr>
371
372 * combine.c (distribute_notes) <REG_DEAD>: Do not consider SETs past
373 the insn to which the note was originally attached.
374
375 2007-01-21 Jan Hubicka <jh@suse.cz>
376
377 * ipa-inline.c (inlining_mode): Comment, move up.
378 (cgraph_decide_inlining_incrementally): Do not perform inlining
379 itself; fix handling of flattening of self recursive functions.
380 (cgraph_find_cycles): Remove.
381 (cgraph_flatten_node): Remove.
382 (cgraph_decide_inlining): Use incremental inliner to handle flattening.
383 (try_inline): New function.
384 (cgraph_early_inlining): Update call of
385 cgraph_decide_inlining_incrementally. Apply inlining here.
386 (apply_inline): Update call of cgraph_decide_inlining_incrementally.
387
388 2007-01-21 Dirk Mueller <dmueller@suse.de>
389
390 PR bootstrap/30511
391 * tree-vrp.c (check_array_bounds): do not warn
392 about ADDR_EXPR's of ARRAY_REF's which are immediately
393 used in binary expressions.
394
395 2007-01-21 Ira Rosen <irar@il.ibm.com>
396
397 * tree-vectorizer.h (struct _stmt_vec_info): Add new field
398 read_write_dep and macros for its access.
399 * tree-vectorizer.c (new_stmt_vec_info): Initialize the new field.
400 * tree-vect-analyze.c (vect_analyze_data_ref_dependence): Remove
401 argument, call vect_check_interleaving for every independent pair of
402 data-refs. Mark loads that access the same memory location as a store
403 in the loop.
404 (vect_check_dependences): Remove.
405 (vect_analyze_data_ref_dependences): Remove vect_check_dependences
406 call, fix the call to vect_analyze_data_ref_dependence.
407 (vect_analyze_data_ref_access): For statements that access the same
408 data-ref, check that they are not stores; for loads, check that there
409 is no store that access the same location.
410
411 2007-01-20 Roger Sayle <roger@eyesopen.com>
412 Joseph Myers <joseph@codesourcery.com>
413
414 * doc/invoke.texi (-fdump-rtl-combine): Fix under/overfull hbox.
415 (-fdump-rtl-stack): Likewise.
416 (-fno-signed-zeros): Use @minus{} for a minus sign. Correct typo.
417 (-mcheck-zero-division, -mcpu): Fix under/overfull hbox.
418 (-mpt-fixed): Use @minus{} for minus sign.
419 (Using Precompiled Headers): Fix under/overfull hbox.
420
421 2007-01-20 Jan Hubicka <jh@suse.cz>
422
423 * tree-flow.h (struct stmt_ann_d): Move references_memory to proper
424 place within annotation.
425
426 2007-01-20 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
427
428 * pa.c (output_move_double): Change array size of xoperands to 4.
429
430 2007-01-20 Jan Hubicka <jh@suse.cz>
431
432 * tree-tailcall.c (adjust_return_value): Do not use RESULT_DECL
433 as temporary.
434
435 2007-01-19 Ian Lance Taylor <iant@google.com>
436
437 * expmed.c (expand_divmod) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]: Cast
438 constant to unsigned HOST_WIDE_INT before negating.
439
440 2007-01-19 Ian Lance Taylor <iant@google.com>
441
442 * tree-ssa-operands.h (struct vuse_vec_d): Change num_vuse field
443 to unsigned.
444 (VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Use unsigned comparison.
445 (VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
446 (SET_VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Likewise.
447 (SET_VUSE_ELEMENT_VAR) [ENABLE_CHECKING]: Likewise.
448 (SET_VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
449 (realloc_vdef, realloc_vuse): Change second parameter to
450 unsigned.
451 (ssa_operand_iterator_d): Change vuse_index and mayuse_index
452 fields to unsigned.
453 * tree-ssa-operands.c (realloc_vop): Change num_elem parameter to
454 unsigned. Change x and lim locals to unsigned.
455 (realloc_vdef, realloc_vuse): Change num_elem parameter to
456 unsigned.
457 (finalize_ssa_vuse_ops): Change old_i local to unsigned.
458 (copy_virtual_operands): Change i and n locals to unsigned.
459
460 2007-01-19 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
461
462 PR middle-end/29335
463 * builtins.c (fold_builtin_1): Handle builtin fdim.
464
465 2007-01-20 Jan Hubicka <jh@suse.cz>
466
467 * tree-ssa.c (init_tree_ssa): Do not call init_alias_heapvars.
468 * tree-ssa-structalias.c (compute_points_to_sets): Do call
469 init_alias_heapvars.
470 (init_alias_heapvars): Initialize only when not already initialized.
471 (delete_alias_heapvars): Set heapvar_for_stmt to NULL.
472
473 2007-01-19 Roger Sayle <roger@eyesopen.com>
474
475 * common.opt (fsigned-zeros): New command line option.
476 * flags.h (HONOR_SIGNED_ZEROS): Control via flag_signed_zeros instead
477 of flag_unsafe_math_optimizations.
478 * opts.c (set_fast_math_flags): The -ffast-math command line option
479 implies -fno-signed-zeros.
480 (fast_math_flags_set_p): Likewise.
481
482 * doc/invoke.texi: Document new -fno-signed-zeros option, and update
483 the documentation of -ffast-math appropriately. Wrap long lines.
484
485 2007-01-19 Steve Ellcey <sje@cup.hp.com>
486
487 * system.h (ASM_MAKE_LABEL_LINKONCE): Poison.
488 * varasm.c (globalize_decl): Remove ASM_MAKE_LABEL_LINKONCE ifdef.
489
490 2007-01-19 Tomas Bily <tbily@suse.cz>
491
492 * cgraphunit.c (cgraph_finalize_function): Updating of pid
493 * tree-profile.c:
494 (tree_init_ic_make_global_vars): New function
495 (tree_init_edge_profiler): call of tree_init_ic_make_global_vars
496 (tree_gen_ic_profiler): New function
497 (tree_gen_ic_func_profiler): New function
498 (tree_profiling): Added calling of tree_gen_ic_func_profiler
499 (tree_profile_hooks): Added hook for indirec/virtual calls
500 * value-prof.c (tree_find_values_to_profile): New case for
501 indirect calls
502 (tree_values_to_profile): Call for determining indirect/virtual
503 counters
504 (tree_indirect_call_to_profile): New function
505 (tree_ic_transform): New function
506 (tree_ic): New function
507 (find_func_by_pid): New function
508 (init_pid_map): New function
509 (tree_value_profile_transformations): Added check for
510 indirect/virtual call transformation
511 * value-prof.h (enum hist_type): New counter type for
512 indirect/virtual calls
513 (profile_hooks): Added new hook for profiling indirect/virtual
514 calls
515 * profile.c (instrument_values): New case for indirect/virtual
516 call added
517 * gcov-io.h (GCOV_LAST_VALUE_COUNTER): Changed to 6
518 (GCOV_COUNTER_V_INDIR): New counter type
519 (GCOV_COUNTER_NAMES): New name of counter "indirect" added
520 (GCOV_MERGE_FUNCTIONS): New merge function for indirect/virtual
521 call added
522 * cgraph.c: Definition of cgraph_max_pid
523 (cgraph_create_node): Default init of pid attribute
524 * cgraph.h: Declaration of cgraph_max_pid
525 (struct cgraph_node): Added pid attribute
526 * libgcov.c (__gcov_indirect_call_profiler): New function
527 (__gcov_one_value_profiler_body): New function
528 (__gcov_one_value_profiler): Body was moved to
529 __gcov_one_value_profiler_body and calls it
530
531 2007-01-19 Basile Starynkevitch <basile@starynkevitch.net>
532
533 * doc/gty.texi (Options): Document the mark_hook option to GTY.
534 * gengtype.c (write_types_data, write_func_for_structure,
535 write_types, ggc_wtd, pch_wtd): Add skip_hooks to
536 write_types_data, ggc_wtd, pch_wtd for processing mark_hook.
537 (walk_type, write_func_for_structure): Generate the mark_hook if
538 needed.
539
540 2007-01-19 Jan Hubicka <jh@suse.cz>
541
542 * ipa-inline.c (cgraph_decide_inlining_incrementally): Instead of
543 'early' argument take inlining mode argument specifying whether to
544 inline for size/speeed or all functions; add support for flattening;
545 improve dumpting.
546 (cgraph_early_inlining): Update call of decide_inlining_incrementally.
547
548 2007-01-19 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
549
550 PR c++/17947
551 * toplev.c (warn_deprecated_use): Use %qD instead of %qs to print
552 the name of the declared identifier.
553
554 2007-01-19 Dirk Mueller <dmueller@suse.de>
555
556 * config/i386.h (CONDITIONAL_REGISTER_USAGE): Store
557 result of PIC_OFFSET_TABLE_REGNUM in temporary variable to avoid
558 duplicate evaluation.
559
560 2007-01-19 Uros Bizjak <ubizjak@gmail.com>
561
562 * config/i386/i386.md (acos<mode>2): Rename from acossf2 and acosdf2.
563 Macroize expander using X87MODEF12 mode macro. Extend operand 1
564 to XFMode, use acosxf2 and truncate result to requested mode.
565 Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
566 (asin<mode>2): Similarly, with asin expanders.
567 (*fscalexf4_i387): Rename from *fscalexf4.
568 (expNcorexf3): New expander.
569 (expxf2, exp10xf2, exp2xf2): Use expNcorexf3 expander.
570 (exp<mode>2): Rename from expsf2 and expdf2. Macroize expander using
571 X87MODEF12 mode macro. Extend operand 1 to XFMode, use expxf2 and
572 truncate result to requested mode. Use SSE_FLOAT_MODE_P to disable
573 patterns for SSE math.
574 (exp10<mode>2): Similarly, with exp10 expanders.
575 (exp2<mode>2): Similarly, with exp2 expanders.
576 (expm1<mode>2): Similarly, with expm1 expanders.
577 (ldexp<mode>3): Similarly, with ldexp expanders.
578 (log<mode>2, log10<mode>2, log2<mode>2, log1p<mode>2, logb<mode>2):
579 Use gen_truncxf<mode>2_i387_noop to truncate result.
580
581 2007-01-19 Richard Sandiford <richard@codesourcery.com>
582
583 * config/m68k/m68k.h (PREFERRED_STACK_BOUNDARY): Define to 32
584 for ColdFire targets.
585
586 2007-01-19 Nathan Sidwell <nathan@codesourcery.com>
587 Richard Sandiford <richard@codesourcery.com>
588
589 * config/m68k/m68k.h (M68K_STATIC_CHAIN_REG_NAME): New macro.
590 (INITIALIZE_TRAMPOLINE): Use STATIC_CHAIN_REGNUM.
591 (__transfer_from_trampoline): Use M68K_STATIC_CHAIN_REG_NAME.
592 * config/m68k/m68kelf.h (STATIC_CHAIN_REGNUM): Override.
593 (M68K_STATIC_CHAIN_REG_NAME): Likewise.
594 * config/m68k/netbsd-elf.h (M68K_STATIC_CHAIN_REG_NAME): Likewise.
595
596 2007-01-19 Richard Sandiford <richard@codesourcery.com>
597
598 * config/m68k/m68k.md (adddi_dilshr32): Rename to...
599 (*adddi_dilshr32): ...this. Fix formatting. Remove commented-out
600 non-canonical pattern. Restrict to !TARGET_COLDFIRE.
601 (*adddi_dilshr32_cf): New pattern.
602 (adddi3, subdi3): Remove first alternatives.
603
604 2007-01-19 Richard Sandiford <richard@codesourcery.com>
605
606 * config/m68k/m68k.c (notice_update_cc): If an SFmode move is
607 implemented using move.l, do not use its cc result for floating-point
608 comparisons.
609
610 2007-01-19 Richard Sandiford <richard@codesourcery.com>
611
612 * config/m68k/m68k.h (EXTRA_CONSTRAINT): Stop the 'T' constraint
613 from accepting 's' constraints if flag_pic.
614
615 2007-01-19 Richard Sandiford <richard@codesourcery.com>
616
617 * config/m68k/m68k.md (bordered, bunordered, buneq, bunge, bungt)
618 (bunle, bunlt, bltgt, bordered_rev, bunordered_rev, buneq_rev)
619 (bunge_rev, bungt_rev, bunle_rev, bunlt_rev, bltgt_rev): Change
620 condition from TARGET_68881 to TARGET_HARD_FLOAT.
621
622 2007-01-19 Sandra Loosemore <sandra@codesourcery.com>
623
624 * longlong.h (count_leading_zeros, COUNT_LEADING_ZEROS_0): Add
625 ColdFire alternatives.
626 * config/m68k/m68k.h (CLZ_DEFINED_VALUE_AT_ZERO): New macro.
627 * config/m68k/m68k.md (clzsi2): Define for ColdFire
628 architectures that support ff1 instruction.
629
630 2007-01-19 Richard Sandiford <richard@codesourcery.com>
631 Julian Brown <julian@codesourcery.com>
632
633 * config/m68k/m68k.h (CONST_OK_FOR_LETTER_P): Add an 'R' case.
634 * config/m68k/m68k.md (*movsi_cfv4): Fold into...
635 (*movsi_cf): ...here. Remove unnecessary 'R' from 'Rg'.
636 Add commentary.
637
638 2007-01-19 Richard Sandiford <richard@codesourcery.com>
639
640 * config/m68k/m68k-protos.h (valid_mov3q_const): Take a HOST_WIDE_INT
641 and return a bool.
642 (output_move_const_into_data_reg, output_move_simode_const): Delete.
643 * config/m68k/m68k.c (const_method, const_int_cost): Take a
644 HOST_WIDE_INT instead of an rtx.
645 (m68k_rtx_costs): Update call accordingly.
646 (output_move_const_into_data_reg): Likewise. Fix formatting.
647 (valid_mov3q_const): Take a HOST_WIDE_INT instead of an rtx.
648 Return a bool.
649 (output_move_simode_const): Update calls after above changes.
650 Rework to use automatic variables and predicates like MEM_P.
651 * config/m68k/m68k.md (pushexthisi_const): Update call to
652 valid_mov3q_const.
653
654 2007-01-19 Dirk Mueller <dmueller@suse.de>
655
656 * tree-ssa-alias.c (perform_var_substitution): Fix typo
657 in dump_flags test.
658
659 2007-01-19 Richard Guenther <rguenther@suse.de>
660
661 * builtins.c (expand_builtin_cexpi): Fall back to expanding
662 via cexp in case sincos is not available.
663
664 2007-01-19 Richard Guenther <rguenther@suse.de>
665
666 * doc/tm.texi (TARGET_HAS_SINCOS): Document new target macro.
667 * defaults.h (TARGET_HAS_SINCOS): Default to off.
668 * config/linux.h (TARGET_HAS_SINCOS): Set to on if we have glibc.
669 * config/alpha/linux.h (TARGET_HAS_SINCOS): Likewise.
670 * config/sparc/linux.h (TARGET_HAS_SINCOS): Likewise.
671 * config/sparc/linux64.h (TARGET_HAS_SINCOS): Likewise.
672 * config/rs6000/linux.h (TARGET_HAS_SINCOS): Likewise.
673 * config/rs6000/linux64.h (TARGET_HAS_SINCOS): Likewise.
674
675 2007-01-19 Uros Bizjak <ubizjak@gmail.com>
676
677 * config/i386/i386.md (*fpatanxf3_i387, fpatan_extend<mode>xf3_i387):
678 New insn patterns.
679 (atan2sf3_1, atan2df3_1, atan2xf3_1): Remove insn patterns.
680 (atan2xf3): Directly generate RTL pattern.
681 (atan2<mode>3): Rename from atan2sf3 and atan2df3 and macroize insn
682 patterns using X87MODEF12 mode macro. Use fpatan_extend<mode>xf3_i387
683 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
684 disable patterns for SSE math.
685 (atan<mode>2): Rename from atansf2 and atandf2 and macroize insn
686 patterns using X87MODEF12 mode macro. Use fpatan_extend<mode>xf3_i387
687 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
688 disable patterns for SSE math.
689
690 2007-01-19 Alexandre Oliva <aoliva@redhat.com>
691
692 * libgcc-std.ver: Fix typo in %inherit for GCC_4.3.0.
693
694 2007-01-18 Roger Sayle <roger@eyesopen.com>
695
696 * fold-const.c (fold_unary) <VIEW_CONVERT_EXPR>: Optimize away a
697 VIEW_CONVERT_EXPR to the same type as it's operand.
698
699 2007-01-18 David Edelsohn <edelsohn@gnu.org>
700
701 * config/rs6000/darwin-ldouble.c: Only build _SOFT_FLOAT if
702 configured for long double 128.
703
704 2007-01-18 Mike Stump <mrs@apple.com>
705
706 * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Fix build
707 error.
708
709 2007-01-18 Michael Meissner <michael.meissner@amd.com>
710
711 * i386.c (ix86_compute_frame_layout): Make fprintf's in #if 0 code
712 type correct.
713
714 2007-01-18 Jan Hubicka <jh@suse.cz>
715
716 * tree-ssa-operands.c (vop_free_bucket_size): Never return value
717 greater than NUM_VOP_FREE_BUCKETS.
718
719 2007-01-18 Daniel Berlin <dberlin@dberlin.org>
720
721 * tree-ssa-structalias.c: Update comments.
722 (ptabitmap_obstack): Removed.
723 (pta_obstack): New.
724 (oldpta_obstack): Ditto.
725 (stats): Add a few members.
726 (struct variable_info): Remove node, complex, address_taken, and
727 indirect_target members. Add oldsolution member.
728 (new_var_info): Do not initialize removed members.
729 (constraint_expr_type): Remove INCLUDES.
730 (constraint_graph): Add size, implicit_preds, rep,
731 indirect_cycles, eq_rep, label, direct_nodes, and complex members.
732 (FIRST_REF_NODE): New macro.
733 (LAST_REF_NODE): Ditto.
734 (FIRST_ADDR_NODE): Ditto.
735 (find): New function.
736 (unite): Ditto.
737 (dump_constraint): Do not handle INCLUDES.
738 (insert_into_complex): Do not insert duplicate constraints.
739 (condense_varmap_nodes): Renamed and rewritten into ...
740 (merge_node_constraints): This. Also fix bug in handling of
741 offseted copy constraints.
742 (clear_edges_for_node): No longer need to deal with preds at all,
743 or removing associated preds/succs.
744 (merge_graph_nodes): Deal with indirect_cycles.
745 Don't deal with predecessors.
746 (add_implicit_graph_edge): New function.
747 (add_pred_graph_edge): Ditto.
748 (add_graph_edge): Don't deal with predecessors.
749 (build_constraint_graph): Removed.
750 (build_pred_graph): New function.
751 (build_succ_graph): Ditto.
752 (struct scc_info): Removed in_component. Added roots, dfs, and
753 node_mapping. Remove visited_index, unification_queue.
754 (scc_visit): Deal with union-find we do now.
755 Deal with cycles with REF nodes.
756 (collapse_nodes): Renamed and rewritten to ...
757 (unify_nodes): This.
758 (process_unification_queue): Removed.
759 (topo_visit): Cleanup
760 (do_da_constraint): Use find.
761 (do_sd_constraint): Ditto.
762 (do_ds_constraint): Ditto.
763 (do_complex_constraint): Ditto.
764 (init_scc_info): Update for removed and added members.
765 (find_and_collapse_graph_cycles): Renamed and rewritten into ...
766 (find_indirect_cycles): This.
767 (equivalence_class): New variable.
768 (label_visit): New function.
769 (perform_variable_substitution): Rewritten.
770 (free_var_substitution_info): New function.
771 (find_equivalent_node): Ditto.
772 (move_complex_constraints): Ditto.
773 (eliminate_indirect_cycles): Ditto.
774 (solve_graph): Only propagate changed bits.
775 Use indirect cycle elimination.
776 Use find.
777 (tree_id_t): Rename to tree_vi_t, delete id member, add vi member.
778 (tree_id_eq): Renamed to ...
779 (tree_vi_eq): This. Update for member change
780 (insert_id_for_tree): Renamed and rewritten to ...
781 (insert_vi_for_tree): This.
782 (lookup_id_for_tree): Renamed and rewritten to ...
783 (lookup_vi_for_tree): This.
784 (get_id_for_tree): Renamed and rewritten to ...
785 (get_vi_for_tree): Ditto.
786 (get_constraint_exp_from_ssa_var): Update to use get_vi_for_tree.
787 (process_constraint): Don't handle INCLUDES.
788 Remove special ADDRESSOF case.
789 (find_func_aliases): Rewrite to use vi functions instead of id
790 ones.
791 (create_function_info_for): Ditto.
792 (create_variable_info_for): Ditto.
793 (intra_create_variable_infos): Ditto.
794 (merge_smts_into): Ditto.
795 (find_what_p_points_to): Ditto.
796 (init_base_vars): Ditto.
797 (init_alias_vars): Ditto.
798 (remove_preds_and_fake_succs): New function.
799 (dump_sa_points_to_info): Dump new stats.
800 (dump_solution_for_var): Use find.
801 (set_used_smts): Fix formatting.
802 (compute_points_to_sets): Updated for new functions.
803 (ipa_pta_execute): Ditto.
804
805 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
806 Richard Sandiford <richard@codesourcery.com>
807
808 * doc/tm.texi (TARGET_FUNCTION_VALUE): Expand documentation of
809 parallels.
810 * calls.c (expand_call): If the return value is a PARALLEL,
811 extract its first member.
812 * config/m68k/linux.h (FUNCTION_EXTRA_EPILOGUE): Remove.
813 * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
814 use FUNCTION_EXTRA_EPILOGUE.
815 (m68k_function_value): Return a PARALLEL if the return value
816 is of a pointer type.
817 * config/m68k/netbsd-elf.h (current_function_returns_pointer)
818 (FUNCTION_EXTRA_EPILOGUE): Remove.
819 * config/m68k/m68k.md (D0_REG): New constant.
820
821 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
822
823 * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
824 output a NOP for empty epilogues.
825
826 2007-01-18 Richard Sandiford <richard@codesourcery.com>
827
828 * config/m68k/m68k.c (m68k_use_return_insn): Update comments
829 before function. Extend register save check to include all
830 registers, not just integer ones.
831
832 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
833
834 * config/m68k/m68k-protos.h (use_return_insn): Rename to...
835 (m68k_use_return_insn): ...this.
836 * config/m68k/m68k.h (USE_RETURN_INSN): Delete.
837 * config/m68k/m68k.c (use_return_insn): Rename to...
838 (m68k_use_return_insn): ...this.
839 * config/m68k/m68k.md (return): Use m68k_use_return_insn instead
840 of USE_RETURN_INSN.
841
842 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
843
844 * config/m68k/fpgnulib.c (__truncdfsf2): Implement round to
845 nearest even, fix denormal rounding overflow.
846
847 2007-01-18 Richard Sandiford <richard@codesourcery.com>
848
849 * config/m68k/m68k.md (movsf_cf_hard): Use fsmove instead of
850 f%$smove and f%$move.
851 (movdf_cf_hard): Use fdmove for cases 0 and 3 and fmove for case 1.
852 (extendsfdf2_cf): Use fdmove instead of f%&move.
853 (truncdfsf2_cf): Use fsmove instead of f%$smove.
854 (add<mode>3_cf, sub<mode>3_cf): Use <FP:prec> instead of <FP:round>.
855
856 2007-01-18 Richard Sandiford <richard@codesourcery.com>
857
858 * config/m68k/m68k.md (movdf_cf_hard): Use output_move_double for
859 GPR<-GPR moves.
860
861 2007-01-18 Richard Sandiford <richard@codesourcery.com>
862
863 * real.h (real_format): Add a canonical_nan_lsbs_set field.
864 (coldfire_single_format): Declare.
865 (coldfire_double_format): Likewise.
866 * real.c (encode_ieee_single): Use canonical_nan_lsbs_set instead
867 of qnan_msb_set to determine the lower bits of a canonical
868 NaN significand.
869 (encode_ieee_double): Likewise.
870 (encode_ieee_quad): Likewise.
871 (ieee_single_format): Initialize canonical_nan_lsbs_set.
872 (mips_single_format): Likewise.
873 (ieee_double_format): Likewise.
874 (mips_double_format): Likewise.
875 (ieee_extended_motorola_format): Likewise.
876 (ieee_extended_intel_96_format): Likewise.
877 (ieee_extended_intel_128_format): Likewise.
878 (ieee_extended_intel_96_round_53_format): Likewise.
879 (ibm_extended_format): Likewise.
880 (mips_extended_format): Likewise.
881 (ieee_quad_format): Likewise.
882 (mips_quad_format): Likewise.
883 (vax_f_format): Likewise.
884 (vax_d_format): Likewise.
885 (vax_g_format): Likewise.
886 (i370_single_format): Likewise.
887 (i370_double_format): Likewise.
888 (decimal_single_format): Likewise.
889 (decimal_double_format): Likewise.
890 (decimal_quad_format): Likewise.
891 (c4x_single_format): Likewise.
892 (c4x_extended_format): Likewise.
893 (real_internal_format): Likewise.
894 (coldfire_single_format): New real_format.
895 (coldfire_double_format): Likewise.
896 * config/pdp11/pdp11.c (pdp11_f_format): Initialize
897 canonical_nan_lsbs_set.
898 (pdp11_d_format): Likewise.
899 * config/m68k/m68k.c (override_options): Override REAL_FORMAT_MODE
900 if TARGET_COLDFIRE_CPU.
901
902 2007-01-18 Richard Sandiford <richard@codesourcery.com>
903
904 * config/m68k/m68k-protos.h (m68k_output_pic_call): Delete.
905 (output_call, m68k_legitimize_call_address): Declare.
906 * config/m68k/m68k.h (EXTRA_CONSTRAINT): Remove unnecessary
907 parenthesees. Add support for a 'W' constraint.
908 (LEGITIMATE_PIC_OPERAND_P): Remove SYMBOL_REF_FLAG handling.
909 (PRINT_OPERAND_PUNCT_VALID_P): Remove comment about 'o'.
910 (m68k_symbolic_call, m68k_symbolic_jump): Declare.
911 * config/m68k/m68k.c (m68k_symbolic_call, m68k_symbolic_jump): New
912 variables.
913 (override_options): Initialize them. Do not set flag_no_function_cse
914 for TARGET_ID_SHARED_LIBRARY.
915 (m68k_output_pic_call): Delete.
916 (m68k_legitimize_call_address): New function.
917 (print_operand): Remove the %o prefix. Handle the %p prefix.
918 (output_call): New function.
919 (m68k_output_mi_thunk): Use m68k_symbolic_jump. Always load the
920 target address from the GOT if symbolic jumps are not allowed.
921 * config/m68k/m68k.md (call, general_operand): Do not set
922 SYMBOL_REF_FLAG. Use m68k_legitimize_call_address instead.
923 Merge separate flag_pic and !flag_pic define_insns into...
924 (*call, *call_value): ...these new patterns. Match the address
925 rather than the containing MEM and require it to be a call_operand.
926 Use output_call to generate the asm template.
927 * config/m68k/predicates.md (const_call_operand): New predicate.
928 (call_operand): Likewise.
929
930 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
931
932 * config/m68k/m68k.h (REGISTER_MOVE_COST): Simplify definition.
933 (STACK_GROWS_DOWNWARD): Define to 1.
934 (FUNCTION_VALUE, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P): Equivocate
935 comments, emphasizing that these values are only defaults.
936 * config/m68k/linux.h (LINK_SPEC): Fix formatting in #undef.
937 * config/m68k/m68k.c (const_method): Remove trailing whitespace.
938
939 2007-01-18 Richard Sandiford <richard@codesourcery.com>
940
941 * config/m68k/m68k.md (cmpsi): Remove outdated flag_pic handling.
942
943 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
944 Richard Sandiford <richard@codesourcery.com>
945
946 * config/m68k/m68k.h (DATA_REGNO_P, ADDRESS_REGNO_P, INT_REGNO_P)
947 (FP_REGNO_P): New macros.
948 (REGNO_OK_FOR_INDEX_P, REGNO_OK_FOR_BASE_P, REGNO_OK_FOR_DATA_P)
949 (REGNO_OK_FOR_FP_P, REG_OK_FOR_INDEX_P, REG_OK_FOR_BASE_P): Use them.
950 (EH_RETURN_STACKADJ_RTX): Use A0_REG.
951 * config/m68k/m68k.c (m68k_regno_mode_ok): Use the new REGNO macros.
952
953 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
954
955 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
956 (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-rtems*): Add t-floatlib
957 to $tmake_file.
958 * config/m68k/t-floatlib: New file, extracting common code from...
959 * config/m68k/t-m68kbare, config/m68k/t-m68kelf,
960 * config/m68k/t-uclinux: Here.
961 * config/m68k/fpgnulib.c: Do not compile extendeed precision
962 routines on ColdFire targets.
963
964 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
965
966 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
967 (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*)
968 (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-linux*)
969 (m68k-*-rtems*): Use tm_file rather than m68k/m68k.h and
970 explicitly set MOTOROLA to 1.
971 * config/m68k/m68k.h (MOTOROLA): Simplify definition accordingly.
972
973 2007-01-18 Richard Sandiford <richard@codesourcery.com>
974 Nathan Sidwell <nathan@codesourcery.com>
975
976 * config/m68k/m68k.h (PCC_STATIC_STRUCT_RETURN): Delete.
977 (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Add MOTOROLA cases.
978 * config/m68k/coff.h (REGISTER_PREFIX_MD): Delete.
979 * config/m68k/m68020-elf.h (LIB_SPEC): Delete.
980 * config/m68k/m68k-none.h (CC1_SPEC, CPP_SUBTARGET_SPEC): Delete.
981 * config/m68k/m68kelf.h (IMMEDIATE_PREFIX, REGISTER_PREFIX_MD)
982 (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Delete.
983 (NO_DOLLAR_IN_LABEL, PCC_STATIC_STRUCT_RETURN): Don't undefine.
984 (BSS_ASM_OP, ASM_OUTPUT_SKIP, ASM_OUTPUT_ASCII): Delete.
985 * config/m68k/m68kemb.h (PCC_STATIC_STRUCT_RETURN): Don't undefine.
986 * config/m68k/linux.h (SIZE_TYPE, PTRDIFF_TYPE, WCHAR_TYPE)
987 (WCHAR_TYPE_SIZE, TARGET_OBJFMT_CPP_BUILTINS): Delete.
988 (TARGET_OS_CPP_BUILTINS): Don't define mc68000 and mc68020 here.
989 (DBX_REGISTER_NUMBER): Delete.
990 * config/m68k/netbsd-elf.h (IMMEDIATE_PREFIX): Delete.
991 (PCC_STATIC_STRUCT_RETURN): Don't undefine.
992 * config/m68k/openbsd.h (PCC_STATIC_STRUCT_RETURN): Define.
993
994 2007-01-18 Richard Sandiford <richard@codesourcery.com>
995
996 * config.gcc (m68k-*-uclinux*): Add flat.h to $tm_file.
997 * config/flat.h: New file.
998 * crtstuff.c (USE_PT_GNU_EH_FRAME): Don't define if
999 OBJECT_FORMAT_FLAT.
1000 * config/m68k/m68k.h (ASM_PREFERRED_EH_DATA_FORMAT): Do not use
1001 indirect references for -msep-data or -mid-shared-library.
1002 Do not use PC-relative code addresses either.
1003 * config/m68k/m68k.c (override_options): Restrict -fPIC error
1004 to -mpcrel.
1005 * config/m68k/uclinux.h (STARTFILE_SPEC): Define. Use Scrt1.o
1006 for shared libraries and crt1.o for executables. Use crti.o and
1007 crtbegin.o.
1008 (ENDFILE_SPEC): Use crtend.o and crtn.o.
1009 (LIB_SPEC): Suppress -Rlibc.gdb if -static-libc is given.
1010 Do not add -elf2flt or -shared-lib-id options here.
1011 (LINK_SPEC): Define. Pass -elf2flt if no -elf2flt option is given.
1012 Pass -shared-lib-id if -mid-shared-library, taking the library
1013 identifier from -mshared-library-id if given, otherwise
1014 defaulting to 0.
1015 (EH_FRAME_IN_DATA_SECTION): Do not undefine.
1016 (INIT_SECTION_ASM_OP, FINI_SECTION_ASM_OP): Likewise.
1017 (TARGET_OS_CPP_BUILTINS): Define __GXX_MERGED_TYPEINFO_NAMES=0
1018 and __GXX_TYPEINFO_EQUALITY_INLINE=0 if -mid-shared-library.
1019 (DRIVER_SELF_SPECS): Map unadorned PIC options to -msep-data.
1020 * config/m68k/t-uclinux (EXTRA_MULTILIB_PARTS): Add crtbegin.o
1021 and crtend.o.
1022 * config/m68k/lb1sf68.asm (PICCALL): Use an lea and pc-relative
1023 jump sequence for ISA A and ISA A+.
1024 (PICJUMP): Likewise.
1025
1026 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1027
1028 * config.gcc (m68k-*-uclinux*): Base the port on the common
1029 and m68k GNU/Linux files rather than on the generic ELF ones.
1030 * config/m68k/uclinux.h (TARGET_VERSION): Override.
1031 (TARGET_OS_CPP_BUILTINS): Use LINUX_TARGET_OS_CPP_BUILTINS.
1032
1033 2007-01-18 Julian Brown <julian@codesourcery.com>
1034 Richard Sandiford <richard@codesourcery.com>
1035
1036 * config/m68k/m68k.h (LONG_DOUBLE_TYPE_SIZE): Make 64-bit on ColdFire.
1037 (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
1038 * config/m68k/netbsd-elf.h (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Undefine
1039 before redefining.
1040 * config/m68k/uclinux-oldabi.h (LONG_DOUBLE_TYPE_SIZE): Redefine to
1041 80 unconditionally.
1042 (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
1043
1044 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1045
1046 * doc/install.texi: Document m68k-uclinuxoldabi.
1047 * config.gcc (m68k-*-uclinuxoldabi): New configuration.
1048 * config/m68k/uclinux-oldabi.h: New file, copied from
1049 config/m68k/uclinux.h.
1050
1051 2007-01-18 Kaz Kojima <kkojima@rr.iij4u.or.jp>
1052
1053 * config/m32r/m32r-protos.h (m32r_expand_epilogue): Declare it.
1054 * config/m32r/m32r.c (m32r_setup_incoming_varargs): Use gen_frame_mem.
1055 (m32r_compute_frame_size): Use unsigned for regno.
1056 (m32r_reload_lr): Use gen_frame_mem.
1057 (pop): New.
1058 (m32r_output_function_epilogue): Don't output the function epilogue
1059 textually here.
1060 (m32r_expand_epilogue): New.
1061 (direct_return): Return false if the function has the interrupt
1062 attribute.
1063 (m32r_hard_regno_rename_ok): Remove code for the textual epilogue.
1064 * config/m32r/m32r.md (epilogue): New expander.
1065 (return_lr, return_rte): New insns.
1066 (return): Make it expander.
1067 (return_normal): New expander.
1068
1069 2007-01-18 Josh Conner <jconner@apple.com>
1070
1071 PR target/30485
1072 * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Add
1073 support for UNLE, UNLT, UNGE, and UNGT.
1074
1075 2007-01-18 Jan Hubicka <jh@suse.cz>
1076
1077 * tree-vrp.c (finalize_jump_threads): Do not call cleanup_cfg by hand.
1078 * tree-tailcall (add_virtual_phis): Likewise.
1079 (optimize_tail_call): Return TODOs.
1080 (execute_tail_calls): Return TODOs.
1081 * tree-ssa-ccp (execute_fold_all_builtins): Do cleanup_cfg via TODO.
1082 * tree-cfgcleanup.c (cleanup_tree_cfg_loop): Return if something
1083 changed.
1084 * tree-ssa-forwprop.c (tree_ssa_forward_propagate_single_use_value):
1085 Cleanup cfg using TODO.
1086 * tree-flow.h (cleanup_tree_cfg_loop): Update prototype.
1087 * passes.c (execute_function_todo): When cleanup did something, remove
1088 unused locals.
1089 * tree-cfg.c (pass_build_cfg): Add cleanup_cfg TODO.
1090 (make_edges): Don't cleanup_cfg.
1091
1092 2007-01-18 Uros Bizjak <ubizjak@gmail.com>
1093
1094 * reg-stack.c (subst_stack_regs_pat) [UNSPEC_SINCOS_COS,
1095 UNSPEC_XTRACT_FRACT]: Use generic code for instructions that
1096 operate on the top of stack.
1097 [UNSPEC_SINCOS_SIN, UNSPEC_XTRACT_EXP, UNSPEC_TAN]: Rewrite
1098 register handling of instructions that output to the second
1099 stack slot.
1100 [UNSPEC_TAN_ONE, UNSPEC_TAN_TAN]: Remove.
1101 (move_for_stack_reg): Special-case check for dead destination
1102 stack slot for constant load of 1.0 inside UNSPEC_TAN.
1103
1104 * config/i386/i386.md (UNSPEC_TAN): New constant.
1105 (UNSPEC_TAN_ONE, UNSPEC_TAN_TAN): Remove.
1106 (fptanxf4_i387, fptan_extend<mode>xf4_i387): New patterns
1107 to correctly model move of constant 1.0 to top stack slot.
1108 (*tandf3_1, *tansf3_1, *tanxf3_1): Remove insn patterns.
1109 (unnamed peephole2 pattern): Remove corresponding peephole2
1110 pattern that optimizes tan insn and loading of constant 1.0.
1111 (tanxf2): Use fptanxf4_i387.
1112 (tan<mode>2): Rename from tansf2 and tandf2 and macroize insn
1113 patterns using X87MODEF12 mode macro. Use fptan_extend<mode>xf4_i387
1114 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
1115 disable patterns for SSE math.
1116 (sincos<mode>3): Use truncxf<mode>2_i387_noop for truncation.
1117 (fyl2x_extend<mode>xf3_i387): Use X87MODEF12 for operand 1.
1118
1119 2007-01-18 Dirk Mueller <dmueller@suse.de>
1120 Richard Guenther <rguenther@suse.de>
1121
1122 PR diagnostic/8268
1123 * doc/invoke.texi (Warray-bounds): Document -Warray-bounds.
1124 * common.opt (Warray-bounds): Add new warning option.
1125 * c-opts.c (c_common_handle_option): Define -Warray-bounds
1126 if -Wall is given.
1127 * Makefile.in: make tree-vrp.o depend on toplev.h
1128 * tree-vrp.c (vrp_finalize): Call check_array_refs if -Warray-bounds
1129 is enabled.
1130 (check_array_refs, check_array_bounds, check_array_ref): New.
1131
1132 2007-01-18 Jan Hubicka <jh@suse.cz>
1133
1134 * tree-ssa-ccp.c (ccp_finalize): Return if something changed.
1135 (execute_ssa_ccp): Return flags conditionally.
1136 * tree-ssa-propagate.c (substitue_and_fold): Return if something was
1137 changed.
1138 * tree-ssa-propagate.h (substitute_and_fold): Update prototype.
1139
1140 2007-01-18 Steven Bosscher <steven@gcc.gnu.org>
1141
1142 * cfgcleanup.c (cleanup_cfg): Detect cfglayout mode and set
1143 the CLEANUP_CFGLAYOUT flag when in cfglayout mode.
1144
1145 * Makefile.c (GTFILES): Add cfglayout.h.
1146 * gengtype.c (open_base_files): Likewise.
1147 * cfglayout.c (cfg_layout_function_footer,
1148 cfg_layout_function_header) Reindent to make gengtype happy.
1149 * cfglayout.h (cfg_layout_function_footer,
1150 cfg_layout_function_header): Add GTY(()) marker.
1151
1152 * ifcvt.c (noce_try_sign_mask): Make sure INSN_B is non-null.
1153
1154 2007-01-18 Ben Elliston <bje@au.ibm.com>
1155
1156 * genautomata.c (write_automata): Include xstrerror output in the
1157 error message if writing the DFA description file fails.
1158
1159 2007-01-17 H.J. Lu <hongjiu.lu@intel.com>
1160
1161 * config/mips/mips-protos.h (mips_output_external): Make it
1162 return void.
1163 * config/mips/iris.h (TARGET_ASM_EXTERNAL_LIBCALL): Removed.
1164 * config/mips/mips.c (irix_output_external_libcall): Likewise.
1165 (extern_list): Likewise.
1166 (extern_head): Likewise.
1167 (TARGET_ASM_FILE_END): Likewise.
1168 (mips_file_end): Likewise.
1169 (mips_output_external): Rewritten.
1170
1171 2007-01-18 Ben Elliston <bje@au.ibm.com>
1172
1173 * genpreds.c (write_insn_preds_c): Only write out the function
1174 body for regclass_for_constraint if we have register constraints.
1175
1176 2007-01-17 Tom Tromey <tromey@redhat.com>
1177
1178 * doc/sourcebuild.texi (libgcj Tests): Use sourceware.org.
1179 * doc/install.texi (Testing): Use sourceware.org.
1180 (Binaries): Likewise.
1181 (Specific): Likewise.
1182 * doc/contrib.texi (Contributors): Use sourceware.org.
1183
1184 2007-01-17 Anatoly Sokolov <aesok@post.ru>
1185
1186 * config/avr/avr.h (AVR_HAVE_LPMX): New macro.
1187 (AVR_ENHANCED): Rename to ...
1188 (AVR_HAVE_MUL): ... new.
1189 (avr_enhanced_p): Rename to ...
1190 (avr_have_mul_p): ... new.
1191 (TARGET_CPU_CPP_BUILTINS): Use 'avr_have_mul_p' instead of
1192 'avr_enhanced_p' for "__AVR_ENHANCED__". Define "__AVR_HAVE_MUL__".
1193 * config/avr/avr.c (avr_enhanced_p): Rename to ...
1194 (avr_have_mul_p): ... new.
1195 (base_arch_s): Rename 'enhanced' to 'have_mul'.
1196 (avr_override_options): Use 'avr_have_mul_p' and 'have_mul' instead of
1197 'avr_enhanced_p' and 'enhanced'.
1198 (ashlhi3_out, ashrhi3_out, lshrhi3_out, avr_rtx_costs): Use
1199 AVR_HAVE_MUL instead of AVR_ENHANCED.
1200 * avr.md (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
1201 (mulqi3, *mulqi3_enh, *mulqi3_call, mulqihi3, umulqihi3, mulhi3,
1202 *mulhi3_enh, *mulhi3_call, mulsi3, *mulsi3_call): Use AVR_HAVE_MUL
1203 instead of AVR_ENHANCED.
1204 (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
1205 * libgcc.S: Use __AVR_HAVE_MUL__ instead of __AVR_ENHANCED__.
1206 (__tablejump__): Use __AVR_HAVE_LPMX__ instead of __AVR_ENHANCED__.
1207
1208 2007-01-17 Ian Lance Taylor <iant@google.com>
1209
1210 * vec.h (VEC_reserve_exact): Define.
1211 (vec_gc_p_reserve_exact): Declare.
1212 (vec_gc_o_reserve_exact): Declare.
1213 (vec_heap_p_reserve_exact): Declare.
1214 (vec_heap_o_reserve_exact): Declare.
1215 (VEC_OP (T,A,reserve_exact)): New static inline function, three
1216 versions.
1217 (VEC_OP (T,A,reserve)) [all versions]: Remove handling of
1218 negative parameter.
1219 (VEC_OP (T,A,alloc)) [all versions]: Call ...reserve_exact.
1220 (VEC_OP (T,A,copy)) [all versions]: Likewise.
1221 (VEC_OP (T,a,safe_grow)) [all versions]: Likewise.
1222 * vec.c (calculate_allocation): Add exact parameter. Change all
1223 callers.
1224 (vec_gc_o_reserve_1): New static function, from vec_gc_o_reserve.
1225 (vec_gc_p_reserve, vec_gc_o_reserve): Call vec_gc_o_reserve_1.
1226 (vec_gc_p_reserve_exact, vec_gc_o_reserve_exact): New functions.
1227 (vec_heap_o_reserve_1): New static function, from vec_heap_o_reserve.
1228 (vec_heap_p_reserve, vec_heap_o_reserve): Call vec_heap_o_reserve_1.
1229 (vec_heap_p_reserve_exact): New function.
1230 (vec_heap_o_reserve_exact): New function.
1231
1232 2007-01-17 Jan Hubicka <jh@suse.cz>
1233
1234 * ipa-type-escape.c (look_for_casts): Revamp using handled_component_p.
1235
1236 2007-01-17 Eric Christopher <echristo@apple.com>
1237
1238 * config.gcc: Support core2 processor.
1239
1240 2007-01-16 Jan Hubicka <jh@suse.cz>
1241
1242 * tree-ssanames.c (release_dead_ssa_names): Instead of ggc_freeing
1243 the names, just unlink the chain so we don't crash on dangling pointers
1244 to dead SSA names.
1245
1246 2007-01-16 Jan Hubicka <jh@suse.cz>
1247
1248 * cgraph.h (cgraph_decide_inlining_incrementally): Kill.
1249 * tree-pass.h: Reorder to make IPA passes appear toegher.
1250 (pass_early_inline, pass_inline_parameters, pass_apply_inline): Declare.
1251 * cgraphunit.c (cgraph_finalize_function): Do not compute inling
1252 parameters, do not call early inliner.
1253 * ipa-inline.c: Update comments. Include tree-flow.h
1254 (cgraph_decide_inlining): Do not compute inlining parameters.
1255 (cgraph_decide_inlining_incrementally): Return TODOs; assume to
1256 be called with function context set up.
1257 (pass_ipa_inline): Remove unreachable functions before pass.
1258 (cgraph_early_inlining): Simplify assuming to be called from the
1259 PM as local pass.
1260 (pass_early_inline): New pass.
1261 (cgraph_gate_ipa_early_inlining): New gate.
1262 (pass_ipa_early_inline): Turn into simple wrapper.
1263 (compute_inline_parameters): New function.
1264 (gate_inline_passes): New gate.
1265 (pass_inline_parameters): New pass.
1266 (apply_inline): Move here from tree-optimize.c
1267 (pass_apply_inline): New pass.
1268 * ipa.c (cgraph_remove_unreachable_nodes): Verify cgraph after
1269 transforming.
1270 * tree-inline.c (optimize_inline_calls): Return TODOs rather than
1271 doing them by hand.
1272 (tree_function_versioning): Do not allocate dummy struct function.
1273 * tree-inline.h (optimize_inline_calls): Update prototype.
1274 * tree-optimize.c (execute_fixup_cfg): Export.
1275 (pass_fixup_cfg): Remove
1276 (tree_rest_of_compilation): Do not apply inlines.
1277 * tree-flow.h (execute_fixup_cfg): Declare.
1278 * Makefile.in (gt-passes.c): New.
1279 * passes.c: Include gt-passes.h
1280 (init_optimization_passes): New passes.
1281 (nnodes, order): New static vars.
1282 (do_per_function_toporder): New function.
1283 (execute_one_pass): Dump current pass here.
1284 (execute_ipa_pass_list): Don't dump current pass here.
1285
1286 2007-01-16 Janis Johnson <janis187@us.ibm.com>
1287
1288 * config/dfp-bit.c (dfp_compare_op): Return separate value for NaN.
1289 (DFP_NE, DFP_LE, DFP_GE): Return false for NaN.
1290
1291 2007-01-16 David Edelsohn <edelsohn@gnu.org>
1292
1293 * config/rs6000/darwin-ldouble.c: Build file for SOFT_FLOAT.
1294 (strong_alias): Define.
1295 (__gcc_qmul): Provide non-FMA for soft-float.
1296 (__gcc_qdiv): Same.
1297 (__gcc_qneg): New.
1298 (__gcc_qeq): New.
1299 (__gcc_qle): New.
1300 (__gcc_qge): New.
1301 (__gcc_qunord): New.
1302 (__gcc_stoq): New.
1303 (__gcc_dtoq): New.
1304 (__gcc_qtos): New.
1305 (__gcc_qtod): New.
1306 (__gcc_qtoi): New.
1307 (__gcc_qtou): New.
1308 (__gcc_itoq): New.
1309 (__gcc_utoq): New.
1310 (fmsub): New.
1311 * config/rs6000/rs6000.c (rs6000_init_libfuncs): Initialize
1312 soft-float functions.
1313 * config/rs6000/libgcc-ppc-glibc.ver: Version soft-float symbols.
1314 * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Do not warn
1315 about long double soft float.
1316
1317 2007-01-16 Dorit Nuzman <dorit@il.ibm.com>
1318 Tehila Meyzels <tehila@il.ibm.com>
1319
1320 * tree-vectorizer.h (is_pattern_stmt_p): New.
1321 * tree-vect-analyze.c (vect_determine_vectorization_factor): Fix
1322 formatting (tabs instead of spaces). Cleanup and clarify setting
1323 of STMT_VINFO_VECTYPE. Call is_pattern_stmt_p.
1324 * tree-vect-transform.c (vect_get_vec_def_for_operand): Fix typo.
1325 (vectorizable_type_demotion): Check that types are integral.
1326 (vectorizable_type_promotion): Likewise.
1327 (vectorizable_store): Fix typo. Eliminate new-line at end of
1328 comments.
1329
1330 2007-01-16 Jan Hubicka <jh@suse.cz>
1331
1332 * tree-ssanames.c (release_dead_ssa_names): Remove invalidated
1333 cgraph edges too.
1334
1335 2007-01-15 Eric Christopher <echristo@apple.com>
1336
1337 * ifcvt.c: Include vec.h, vecprim.h.
1338 (check_cond_move_block): New argument regs.
1339 Reorganize. Add registers used to regs.
1340 (cond_move_process_if_block): Use regs set above as
1341 loop bounds.
1342
1343 2007-01-15 Eric Christopher <echristo@apple.com>
1344
1345 * config/darwin.h: Update copyright.
1346 (TARGET_OPTION_TRANSLATE_TABLE): Add umbrella.
1347 (LINK_COMMAND_SPEC): Add -u.
1348 (LINK_SPEC): Fix umbrella for above.
1349
1350 2007-01-15 Joseph S. Myers <joseph@codesourcery.com>
1351
1352 * config/soft-fp/op-common.h, config/soft-fp/op-4.h: Update from
1353 glibc CVS.
1354
1355 2007-01-15 Tom Tromey <tromey@redhat.com>
1356
1357 * doc/sourcebuild.texi (libgcj Tests): Don't mention jacks.
1358 * doc/install.texi (Testing): Don't mention jacks.
1359 (Configuration): Document --enable-java-maintainer-mode. Move
1360 --with-java-home to libgcj-specific section. Document
1361 --with-ecj-jar.
1362 (Prerequisites): Mention --enable-java-maintainer-mode, ecj1.
1363
1364 2007-01-15 Jan Hubicka <jh@suse.cz>
1365
1366 * tree-ssa-dce.c (DCE_TODOs): New.
1367 (propagate_necessity): Return if something changed.
1368 (eliminate_unnecessary_stmts): Likewise.
1369 (perform_tree_ssa_dce): Return TODO flags when needed.
1370 (pass_dce, pass_dce_loop, pass_cd_dce): Remove TODO flags.
1371
1372 2007-01-15 Uros Bizjak <ubizjak@gmail.com>
1373
1374 * config/i386/i386.md (fyl2xxf3_i387): Rename from fyl2x_xf3.
1375 (fyl2x_extend<mode>xf3_i387): New insn pattern.
1376 (log<mode>2): Rename from logsf2 and logdf2 and macroize insn
1377 patterns using X87MODEF12 mode macro. Extend operand 1
1378 to XFmode. Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
1379 (log10<mode>2): Ditto.
1380 (log2<mode>2): Ditto.
1381 (log1p<mode>2): Ditto.
1382 (logb<mode>2): Ditto.
1383 (fyl2xp1xf3_i387): Rename from fyl2xp1_xf3.
1384 (fyl2xp1_extend<mode>xf3_i387): New insn pattern.
1385 (*fxtractxf3_i387): Rename from *fxtractxf3.
1386 (fxtract_extend<mode>xf3_i387): New insn pattern.
1387 (ilogbsi2): Use match_dup 3, not match_operand:XF 3.
1388 * config/i386/i386.c (ix86_emit_i387_log1p): Use gen_fyl2xp1xf3_i387()
1389 and gen_fyl2xxf3_i387().
1390
1391 2007-01-14 Zdenek Dvorak <dvorakz@suse.cz>
1392
1393 * loop-unswitch.c (unswitch_loop): Do not call fix_loop_placement.
1394 * cfgloopmanip.c (fix_loop_placement): Made static. Use
1395 get_loop_exit_edges. Changed return type to bool.
1396 * cfgloop.h (fix_loop_placement): Declaration removed.
1397
1398 2007-01-14 Dorit Nuzman <dorit@il.ibm.com>
1399
1400 * param.h (MIN_VECT_LOOP_BOUND): New.
1401 * params.def (MIN_VECT_LOOP_BOUND): New.
1402 * tree-vectorizer.c (slpeel_tree_peel_loop_to_edge): Takes another
1403 argument - minimum threshold for number of iterations.
1404 * tree-vectorizer.h (slpeel_tree_peel_loop_to_edge): Add another
1405 argument to declaration.
1406 * tree-vect-analyze.c (vect_analyze_operations): Check value of
1407 MIN_VECT_LOOP_BOUND.
1408 * tree-vect-transform.c (vect_do_peeling_for_loop_bound): Call
1409 slpeel_tree_peel_loop_to_edge with additional argument.
1410 (vect_do_peeling_for_alignment): Likewise.
1411 * doc/invoke.texi (min-vect-loop-bound): Document new param option.
1412
1413 2007-01-14 Uros Bizjak <ubizjak@gmail.com>
1414
1415 PR target/30413
1416 * config/i386/i386.c (print_operand) ['z']: Output 'b' for
1417 operands of size 1.
1418
1419 2007-01-14 Jan Hubicka <jh@suse.cz>
1420
1421 * tree-dfa.c (remove_referenced_var): New function.
1422 * tree-ssa-live.c (remove_unused_locals): Walk referenced vars and
1423 prune referenced vars list too.
1424 * tree-flow.h (remove_referenced_var): Declare.
1425
1426 2007-01-14 Jan Hubicka <jh@suse.cz>
1427
1428 * tree-eh.c (add_stmt_to_eh_region_fn): Do not add call_exprs
1429 separately.
1430 (remove_stmt_from_eh_region_fn): Do not remove call_exprs.
1431 (verify_eh_throw_stmt_node, verify_eh_throw_table_statements): Kill.
1432 * except.h (verify_eh_throw_table_statements): Kill prototype.
1433 * cfgexpand.c (expand_gimple_basic_block): Propagate Eh regions
1434 into call exrepssions.
1435 * tree-optimize.c (execute_free_cfg_annotatiosn): Do not call
1436 eh trhow verifier.
1437 * tree-cfg.c: Include pointer-set.h.
1438 (verify_node_sharing): Work on pointer set.
1439 (verify_eh_throw_stmt_node): New.
1440 (verify_stmts): Use pointers sets, verify throw_stmt.
1441
1442 2007-01-13 Zdenek Dvorak <dvorakz@suse.cz>
1443
1444 * ipa-reference.c (analyze_function): Consider also addresses taken
1445 in phi nodes.
1446
1447 2007-01-12 Roger Sayle <roger@eyesopen.com>
1448
1449 * c-typeck.c (null_pointer_constant_p): Replace use of
1450 TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
1451 (build_c_cast): Likewise.
1452
1453 2007-01-12 Roger Sayle <roger@eyesopen.com>
1454
1455 * tree.h (force_fit_type_double): Remove unused final argument.
1456 * c-common.c (constant_expression_warning): Replace use of
1457 TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
1458 (convert_and_check): Likewise.
1459 (shorten_compare): Update call to force_fit_type_double.
1460 (c_common_truthvalue_conversion) <INTEGER_CST>: Use integer_zerop.
1461 * convert.c (convert_to_pointer): Update call to
1462 force_fit_type_double.
1463 * fold-const.c (force_fit_type_double): Remove overflowed_const
1464 argument.
1465 (int_const_binop, fold_convert_const_int_from_int,
1466 fold_convert_const_int_from_real, fold_div_compare,
1467 fold_sign_changed_comparison, fold_unary, fold_negate_const,
1468 fold_abs_const, fold_not_const): Remove the final argument from
1469 calls to force_fit_type_double.
1470
1471 2007-01-12 Andrew Pinski <andrew_pinski@playstation.sony.com>
1472
1473 * configure.ac: Set insn to "nop" for spu-*-* also.
1474 * configure: Regenerate.
1475
1476 2007-01-12 Olga Golovanevsky <olga@il.ibm.com>
1477
1478 * builtins.def : Add BUILT_IN_FREE.
1479
1480 2007-01-12 Jan Hubicka <jh@suse.cz>
1481
1482 PR tree-optimization/30443
1483 * tree-inline.c (tree_function_versioning): Do not optimize when
1484 cloning for inlining.
1485
1486 2007-01-12 Zdenek Dvorak <dvorakz@suse.cz>
1487
1488 * doc/loop.texi: Document recording of loop exits.
1489 * cfgloopmanip.c (loopify, duplicate_loop): Use alloc_loop.
1490 (update_single_exits_after_duplication,
1491 update_single_exit_for_duplicated_loop,
1492 update_single_exit_for_duplicated_loops): Removed.
1493 (duplicate_loop_to_header_edge): Do not call
1494 update_single_exits_after_duplication and
1495 update_single_exit_for_duplicated_loops.
1496 (loop_version): Do not update single_exit information.
1497 (fix_loop_structure): Use record_loop_exits instead of
1498 mark_single_exit_loops.
1499 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update
1500 the lists of loop exits.
1501 * cfghooks.c (redirect_edge_and_branch, redirect_edge_and_branch_force,
1502 split_edge, merge_blocks): Update the lists of loop exits.
1503 * modulo-sched.c (sms_schedule): Pass LOOPS_HAVE_RECORDED_EXITS to
1504 loop_optimizer_init.
1505 * loop-init.c (loop_optimizer_init): Call record_loop_exits instead
1506 of mark_single_exit_loops.
1507 (loop_optimizer_finalize): Call release_recorded_exits.
1508 * tree-ssa-loop.c (tree_loop_optimizer_init): Pass
1509 LOOPS_HAVE_RECORDED_EXITS to loop_optimizer_init.
1510 * tree-vectorizer.c (slpeel_tree_duplicate_loop_to_edge_cfg): Do not
1511 update single exit information.
1512 * lambda-code.c (perfect_nestify): Ditto.
1513 * cfgloop.c (flow_loop_free): Destroy the list of exits of the loop.
1514 (mark_single_exit_loops): Removed.
1515 (alloc_loop, loop_exit_hash, loop_exit_eq, loop_exit_free,
1516 get_exit_descriptions, rescan_loop_exit, record_loop_exits,
1517 dump_recorded_exit, dump_recorded_exits, release_recorded_exits): New
1518 functions.
1519 (get_loop_exit_edges, single_exit): Use recorded exit lists.
1520 (add_bb_to_loop, remove_bb_from_loops): Update the lists of loop exits.
1521 (verify_loop_structure): Verify consistency of the exit lists.
1522 (flow_loops_find): Use alloc_loop. Initialize exits hash.
1523 (set_single_exit): Removed.
1524 * cfgloop.h (struct loop_exit): New function.
1525 (struct loop): single_exit_ field replaced by exits field.
1526 (LOOPS_HAVE_MARKED_SINGLE_EXITS): Replaced by LOOPS_HAVE_RECORDED_EXITS.
1527 (struct loops): Added exits hash.
1528 (mark_single_exit_loops, set_single_exit): Declaration removed.
1529 (release_recorded_exits, record_loop_exits, rescan_loop_exit): Declare.
1530
1531 2007-01-12 Richard Sandiford <richard@codesourcery.com>
1532
1533 * doc/invoke.texi: Avoid use of @headitem.
1534
1535 2007-01-12 Richard Sandiford <richard@codesourcery.com>
1536
1537 * cse.c (cse_insn): Move HAVE_CC0 code after declarations.
1538
1539 2007-01-12 Richard Sandiford <richard@codesourcery.com>
1540
1541 * doc/install.texi: Fix m68k-*-* anchor and add m68k-*-* to the
1542 list of targets.
1543
1544 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
1545 Richard Sandiford <richard@codesourcery.com>
1546
1547 * doc/invoke.texi: Document -mno-bitfield, -mno-rtd and -mno-short.
1548 * config/m68k/m68k.opt: Resort options.
1549 (mbitfield, mrtd, mshort): Remove RejectNegative properties.
1550
1551 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
1552 Richard Sandiford <richard@codesourcery.com>
1553
1554 * doc/invoke.texi: Document the macros that are defined by
1555 m68k's -mtune and -mhard-float options.
1556 * config/m68k/m68k-protos.h (m68k_cpp_cpu_ident) Declare.
1557 (m68k_cpp_cpu_family): Likewise.
1558 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Add a full set
1559 of __ucfv*__ macros. Define __mcffpu__ if generating code for
1560 ColdFire FPUs. Define __mcf_cpu_* and __mcf_family_* macros.
1561 * config/m68k/m68k.c (m68k_cpp_cpu_ident): New function.
1562 (m68k_cpp_cpu_family): Likewise.
1563
1564 2007-01-12 Richard Sandiford <richard@codesourcery.com>
1565
1566 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Treat all mc68*
1567 macros besides mc68000 as tuning macros. Use a switch statement
1568 to set them and mcpu32.
1569
1570 2007-01-12 Julian Brown <julian@codesourcery.com>
1571
1572 * config/m68k/m68k.h: Use TARGET_68040 instead of TARGET_68040_ONLY.
1573 (TARGET_68040_ONLY): Rename to...
1574 (TARGET_68040): ...this.
1575 * config/m68k/m68k.c: Use TARGET_68040 instead of TARGET_68040_ONLY.
1576 * config/m68k/m68k.md: Likewise.
1577
1578 2007-01-12 Julian Brown <julian@codesourcery.com>
1579 Nathan Sidwell <nathan@codesourcery.com>
1580 Richard Sandiford <richard@codesourcery.com>
1581
1582 * config.gcc (m680[012]0-*-*, m68k*-*-*): Set m68k_cpu_ident to
1583 the -mcpu= argument associated with the --with-cpu setting.
1584 Define M68K_DEFAULT_TUNE to the default -mtune= option,
1585 if different from the one implied by the -mcpu setting.
1586 Accept --with-cpu=FOO if FOO is listed in m68k-devices.def,
1587 using mcpu=FOO as the default CPU option. Set target_cpu_default2.
1588 * doc/invoke.texi: Mention ColdFire in the introduction to the
1589 m68k options. Document the new -march, -mcpu, -mtune, -mdiv,
1590 -mno-div and -mhard-float options. Make -m68881 a synonym for
1591 -mhard-float. Document the previously-undocumented -m5206e,
1592 -m528x, -m5307 and -m5407 options. Tweak the existing option
1593 documentation for consistency.
1594 * doc/install.texi: Mention new --with-cpu arguments.
1595 * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Only use the
1596 default CPU if neither -mcpu nor -march are specified.
1597 (ASM_CPU_SPEC): Pass down -mcpu and -march options.
1598 (TARGET_CPU_CPP_BUILTINS): Set __mcfisa*__ macros from
1599 TARGET_ISA*. Set the legacy __mcf*__ cpu macros in the same way,
1600 using m68k_tune to decide between families that implement the
1601 same ISA. Use m68k_tune to set __mcfv4e__.
1602 (FL_BITFIELD, FL_68881, FL_COLDFIRE, FL_CF_HWDIV, FL_CF_MAC)
1603 (FL_CF_EMAC, FL_CF_EMAC_B, FL_CF_USP, FL_CF_FPU, FL_ISA_68000)
1604 (FL_ISA_68010, FL_ISA_68020, FL_ISA_68040, FL_ISA_A, FL_ISA_B)
1605 (FL_ISA_C, FL_ISA_MMU): New macros.
1606 (MASK_COLDFIRE): Delete.
1607 (TARGET_68010, TARGET_68020, TARGET_68040_ONLY, TARGET_COLDFIRE)
1608 (TARGET_ISAB): Redefine in terms of m68k_cpu_flags.
1609 (TARGET_68881, TARGET_COLDFIRE_FPU): Redefine in terms of m68k_fpu.
1610 (TARGET_HARD_FLOAT): Do not define here.
1611 (TARGET_ISAAPLUS, TARGET_ISAC): New macros.
1612 (TUNE_68000): New macro.
1613 (TUNE_68000_10): Redefine in terms of TUNE_68000 and TUNE_68010.
1614 (TUNE_68010, TUNE_68030, TUNE_68040, TUNE_68060, TUNE_CPU32)
1615 (TUNE_CFV2): Redefine in terms of m68k_tune.
1616 (uarch_type, target_device, fpu_type): New enums.
1617 (m68k_cpu, m68k_tune, m68k_fpu, m68k_cpu_flags): Declare.
1618 * config/m68k/m68k.c (TARGET_DEFAULT): Remove MASK_68881.
1619 (FL_FOR_isa_00, FL_FOR_isa_10, FL_FOR_isa_20, FL_FOR_isa_40)
1620 (FL_FOR_isa_cpu32, FL_FOR_isa_a, FL_FOR_isa_aplus, FL_FOR_isa_b)
1621 (FL_FOR_isa_c): New macros.
1622 (m68k_isa): New enum.
1623 (m68k_target_selection): New structure.
1624 (all_devices, all_isas, all_microarchs): New tables.
1625 (m68k_cpu_entry, m68k_arch_entry, m68k_tune_entry, m68k_cpu)
1626 (m68k_tune, m68k_fpu, m68k_cpu_flags): New variables.
1627 (MASK_ALL_CPU_BITS): Delete.
1628 (m68k_find_selection): New function.
1629 (m68k_handle_option): Handle -mcpu=, -march= and -mtune=.
1630 Map the legacy target options to a combination of the new ones.
1631 (override_options): Set m68k_cpu, m68k_tune, m68k_fpu and
1632 m68k_cpu_flags. Handle M68K_DEFAULT_TUNE. Use m68k_cpu_flags
1633 to derive default MASK_BITFIELD, MASK_CF_HWDIV and MASK_HARD_FLOAT
1634 settings.
1635 * config/m68k/m68k.opt (m5200, m5206e, m528x, m5307, m5407, mcfv4e)
1636 (m68010, m68020, m68020-40, m68020-60, m68030, m68040): Remove Mask
1637 properties.
1638 (m68881, msoft-float): Change mask from 68881 to HARD_FLOAT.
1639 (march=, mcpu=, mdiv, mhard-float, mtune=): New options.
1640 * config/m68k/m68k-devices.def: New file.
1641
1642 2007-01-12 Richard Sandiford <richard@codesourcery.com>
1643 Nathan Sidwell <nathan@codesourcery.com>
1644
1645 * config/m68k/m68k.h (ASM_CPU_SPEC, ASM_SPEC, EXTRA_SPECS)
1646 (SUBTARGET_EXTRA_SPECS): New macros.
1647 * config/m68k/linux.h (ASM_SPEC): Remove CPU flags;
1648 use %(asm_cpu_spec) instead.
1649 * config/m68k/m68k-none.h (ASM_SPEC): Likewise.
1650 * config/m68k/openbsd.h (ASM_SPEC): Likewise.
1651 * config/m68k/netbsd-elf.h (ASM_SPEC): Likewise.
1652 (EXTRA_SPECS): Rename to...
1653 (SUBTARGET_EXTRA_SPECS): ...this.
1654
1655 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
1656 Richard Sandiford <richard@codesourcery.com>
1657 Julian Brown <julian@codesourcery.com>
1658
1659 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
1660 (m68k-*-uclinux*, m68k-*-linux*, m68k-*-rtems*): Set default_m68k_cpu
1661 to the configuration's default CPU.
1662 (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*): Likewise.
1663 Remove default masks.
1664 (m680[012]0-*-*): Set the default with_cpu to the first part of
1665 the target name.
1666 (m68k*-*-*): Set the default with_cpu to m$default_m68k_cpu.
1667 (m68k*-*-linux): Extend the --with-cpu handling to...
1668 (m680[012]0-*-*, m68k*-*-*): ...these configurations. Allow m68000
1669 and m68010. Don't set target_cpu_default2.
1670 * doc/install.texi: Document --with-cpu for m68k.
1671 * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Define.
1672 * config/m68k/m68k-none.h (TARGET_CPU_DEFAULT, M68K_CPU_m68k)
1673 (M68K_CPU_m68000, M68K_CPU_m68010, M68K_CPU_m68020, M68K_CPU_m68030)
1674 (M68K_CPU_m68040, M68K_CPU_m68302, M68K_CPU_m68332, TARGET_DEFAULT)
1675 (ASM_CPU_DEFAULT_SPEC, CC1_CPU_DEFAULT_SPEC): Delete.
1676 (ASM_SPEC): Remove use of %(asm_cpu_default).
1677 (EXTRA_SPECS, SUBTARGET_EXTRA_SPECS, MULTILIB_DEFAULTS): Delete.
1678 * config/m68k/linux.h (TARGET_DEFAULT): Delete.
1679 (CPP_SPEC): Merge definitions. Do not handle __HAVE_68881__ here.
1680 * config/m68k/netbsd-elf.h (TARGET_OS_CPP_BUILTINS): Define
1681 __HAVE_FPU__ if TARGET_HARD_FLOAT.
1682 (TARGET_DEFAULT): Delete.
1683 (EXTRA_SPECS): Delete cpp_cpu_default_spec, cpp_cpu_spec,
1684 cpp_fpu_spec, asm_default_spec and netbsd_cpp_spec.
1685 (CPP_CPU_SPEC): Delete.
1686 (TARGET_VERSION): Merge definitions, using TARGET_68010 to pick
1687 the appropriate string.
1688 (CPP_CPU_DEFAULT_SPEC, ASM_DEFAULT_SPEC, CPP_FPU_SPEC): Delete.
1689 (CPP_SPEC): Define to NETBSD_CPP_SPEC.
1690 (ASM_SPEC): Don't use %(asm_default_spec).
1691 * config/m68k/m68k.c (TARGET_DEFAULT_TARGET_FLAGS): Remove
1692 TARGET_DEFAULT and add MASK_68881.
1693 * config/m68k/m68k.md: Remove mention of TARGET_DEFAULT from comments.
1694
1695 2007-01-12 Richard Sandiford <richard@codesourcery.com>
1696
1697 * config.gcc (m68010-*-netbsdelf*): Add MASK_68010.
1698 (m68k*-*-netbsdelf*, m68k*-*-openbsd*, m68k*-linux*): Add
1699 MASK_68010 alongside MASK_68020.
1700 * doc/invoke.texi: Document -m68010.
1701 * config/m68k/m68k.opt (m68010): New.
1702 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Define mc68010
1703 if TUNE_68010.
1704 (TUNE_68010): New macro.
1705 * config/m68k/m68k-none.h (M68K_CPU_m68k, M68K_CPU_m68010)
1706 (M68K_CPU_m68020, M68K_CPU_m68030, M68K_CPU_m68040)
1707 (M68K_CPU_m68332): Add MASK_68010.
1708 * config/m68k/linux.h (TARGET_DEFAULT): Add MASK_68010 to
1709 fallback definition.
1710 * config/m68k/netbsd-elf.h (CPP_CPU_SPEC): Remove now-redundant
1711 defines.
1712 * config/m68k/m68k.c (MASK_ALL_CPU_BITS): Add MASK_68010.
1713 (m68k_handle_option): Handle OPT_m68010. Add MASK_68010
1714 to all entries that use MASK_68020.
1715 (output_move_simode_const, output_move_himode, output_move_qimode)
1716 (output_move_stricthi, output_move_strictqi): Use TARGET_68010
1717 instead of TARGET_68020 to select clr behavior. Remove comment
1718 about there being no TARGET_68010.
1719 * config/m68k/m68k.md: Likewise throughout.
1720
1721 2007-01-12 Julian Brown <julian@codesourcery.com>
1722
1723 * config/m68k/m68k.h (TARGET_ISAB): New macro.
1724 * config/m68k/m68k.c: Use TARGET_ISAB rather than TARGET_CFV4.
1725 * config/m68k/m68k.md: Likewise.
1726
1727 2007-01-12 Julian Brown <julian@codesourcery.com>
1728
1729 * config/m68k/m68k.h (LEGITIMATE_INDEX_P, LEGITIMIZE_ADDRESS): Use
1730 TARGET_COLDFIRE_FPU instead of TARGET_CFV4E.
1731
1732 2007-01-12 Julian Brown <julian@codesourcery.com>
1733
1734 * config/m68k/m68k.h (TUNE_68040_60): New macro.
1735 * config/m68k/m68k.c (standard_68881_constant_p): Use it.
1736 * config/m68k/m68k.md: Likewise.
1737
1738 2007-01-12 Julian Brown <julian@codesourcery.com>
1739 Richard Sandiford <richard@codesourcery.com>
1740
1741 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Use TUNE_68030
1742 instead of TARGET_68030, TUNE_68040 instead of TARGET_68040,
1743 TUNE_68060 instead of TARGET_68060 and TUNE_CPU32 instead of
1744 TARGET_CPU32.
1745 (TARGET_CPU32): Rename to...
1746 (TUNE_CPU32): ...this.
1747 (TUNE_68000_10, TUNE_68030, TUNE_68040, TUNE_68060)
1748 (TUNE_CFV2): New macros.
1749 * config/m68k/netbsd-elf.h (LONG_DOUBLE_TYPE_SIZE): Simplify;
1750 remove conditions that are implied by TARGET_68020.
1751 * config/m68k/m68k.c (m68k_output_function_prologue): Use TUNE_68040
1752 instead of TARGET_68040 and TUNE_CPU32 instead of TARGET_CPU32.
1753 (m68k_output_function_epilogue): Likewise.
1754 (m68k_rtx_costs): Likewise. Use TUNE_68060 instead of TARGET_68060
1755 and TUNE_CFV2 instead of TARGET_5200. Use TUNE_68000_10 instead of
1756 "!TARGET_68020 && !TARGET_COLDFIRE" to choose between 68000 and
1757 non-68000 timings. Refactor multiplication and division costs.
1758 (output_addsi3): Use TUNE_68040 instead of TARGET_68040 and
1759 TUNE_CPU32 instead of TARGET_CPU32.
1760 (standard_68881_constant_p): Use TUNE_68040 instead of TARGET_68040
1761 and TUNE_68060 instead of TARGET_68060.
1762 * config/m68k/m68k.md: Use TUNE_68040 instead of TARGET_68040,
1763 TUNE_68060 instead of TARGET_68060, and TUNE_CPU32 instead of
1764 TARGET_CPU32.
1765 (movsi_const0): Use TUNE_68000_10 rather than "!TARGET_68020
1766 && !TARGET_COLDFIRE" to choose between moveq and clr.
1767 Likewise in the unnamed movsf pattern.
1768 (ashlsi_17_24, lshrsi_17_24): Guard with TUNE_68000_10 rather than
1769 "!TARGET_68020 && !TARGET_COLDFIRE". Likewise the unnamed
1770 ashiftrt pattern.
1771
1772 2007-01-12 Richard Sandiford <richard@codesourcery.com>
1773
1774 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Increase amount
1775 of tabbing before backslashes.
1776
1777 2007-01-11 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
1778
1779 * pa-linux.h (ASM_OUTPUT_INTERNAL_LABEL): Undefine.
1780 * pa.h (ASM_OUTPUT_LABEL): Output colon when using GAS.
1781 (ASM_OUTPUT_INTERNAL_LABEL): Define.
1782
1783 2007-01-11 Zdenek Dvorak <dvorakz@suse.cz>
1784
1785 * tree-ssa-loop-ivopts.c (extract_cond_operands): Split from
1786 find_interesting_uses_cond.
1787 (find_interesting_uses_cond): Use extract_cond_operands.
1788 (rewrite_use_compare): Use extract_cond_operands and
1789 force_gimple_operand_bsi. Do not call update_stmt.
1790 (determine_use_iv_cost_condition): Use extract_cond_operands.
1791 Return cheaper of using original bound and changing the exit bound.
1792
1793 2007-01-11 Zdenek Dvorak <dvorakz@suse.cz>
1794
1795 PR tree-optimization/29516
1796 * tree-ssa-address.c (tree_mem_ref_addr, add_to_parts,
1797 most_expensive_mult_to_index, addr_to_parts,
1798 create_mem_ref, maybe_fold_tmr): Make the type of
1799 fields of TARGET_MEM_REF sizetype.
1800 (move_fixed_address_to_symbol, move_pointer_to_base):
1801 New functions.
1802 * tree.def (TARGET_MEM_REF): Add comment on types of
1803 the operands.
1804
1805 2007-01-11 Joseph Myers <joseph@codesourcery.com>
1806
1807 * c-common.c (vector_types_convertible_p): Treat opaque types as
1808 always convertible if they have the same size, but not otherwise.
1809
1810 2007-01-11 Steven Bosscher <steven@gcc.gnu.org>
1811
1812 * ifcvt.c (struct noce_if_info): Add comments to the fields.
1813 Remove the b_unconditional field.
1814 (noce_try_sign_mask): Do not look at b_unconditional.
1815 (noce_process_if_block): Do not use merge_if_blocks. Update
1816 the CFG here. Do not set b_unconditional.
1817 (cond_move_process_if_block): Likewise.
1818 (find_cond_trap): Likewise.
1819 (check_cond_move_block): Require simple jump insns at the end
1820 of the basic block.
1821
1822 2007-01-11 Jan Hubicka <jh@suse.cz>
1823
1824 PR tree-optimization/1046
1825 * tree-tailcall.c (suitable_for_tail_call_opt_p): Use TREE_ADDRESSABLE
1826 when alias info is not ready.
1827 (pass_tail_recursion): Do not require aliasing.
1828 * tree-ssa-copyrename.c (pass_rename_ssa_cop): Likewise.
1829 * tree-ssa-ccp.c (pass_ccp, pass_fold_builtins): Likewise.
1830 * tree-ssa-copy.c (pass_copy_prop): Likewise.
1831 * tree-ssa-forwprop.c (pass_forwprop): Likewise.
1832 * tree-ssa-dce.c (pass_dce, pass_dce_loop, pass_cd_dce): Likewise.
1833 * passes.c (init_optimization_passes): Execute rename_ssa_copies,
1834 ccp, forwprop, copy_prop, merge_phi, copy_prop, dce and tail recursion
1835 before inlining.
1836 * tree-ssa-operands.c (add_virtual_operand, get_indirect_ref_operand):
1837 When aliasing is not build, mark statement as volatile.
1838
1839 2007-01-11 Tom Tromey <tromey@redhat.com>
1840
1841 PR preprocessor/15185, PR preprocessor/20989:
1842 * doc/cppopts.texi <-MT>: Update description of algorithm for
1843 computing default target.
1844 <-M, -MD>: Reword "basename" text.
1845
1846 2007-01-11 Roger Sayle <roger@eyesopen.com>
1847
1848 * builtins.c (expand_builtin_pow, expand_builtin_powi,
1849 fold_builtin_cabs, fold_builtin_sqrt, fold_builtin_trunc,
1850 fold_builtin_floor, fold_builtin_ceil, fold_builtin_round,
1851 fold_builtin_int_int_roundingfn, fold_builtin_bitop,
1852 fold_builtin_bswap, real_constp, fold_builtin_pow,
1853 fold_builtin_powi, fold_builtin_signbit, fold_builtin_copysign,
1854 do_mpfr_arg1, do_mpfr_arg2, do_mpfr_arg3, do_mpfr_sincos): Replace
1855 uses of the macro TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
1856 * convert.c (convert_to_pointer): Likewise.
1857 * expr.c (highest_pow2_factor, expand_expr_real_1): Likewise.
1858 * fold-const.c (force_fit_type, fold_negate_expr, int_const_binop,
1859 const_binop, fold_convert_const_int_from_int,
1860 fold_convert_const_int_from_real,
1861 fold_convert_const_real_from_real, sign_bit_p,
1862 optimize_minmax_comparison, extract_muldiv_1, fold_div_compare,
1863 fold_sign_changed_comparison, fold_unary, fold_comparison,
1864 fold_binary, multiple_of_p, tree_Expr_non_zero_p,
1865 fold_negate_const, fold_abs_const, fold_not_const): Likewise.
1866 * print-tree.c (print_node_brief, print_node): Likewise.
1867 * stor-layout.c (place_field, layout_type): Likewise.
1868 * tree-chrec.c (keep_cast): Likewise.
1869 * tree.c (build_vector, build_real, build_real_from_int_cst,
1870 build_complex): Likewise.
1871
1872 2007-01-11 Roger Sayle <roger@eyesopen.com>
1873
1874 * tree.h (TREE_CONSTANT_OVERFLOW): Obsolete. For the time being,
1875 treat TREE_CONSTANT_OVERFLOW as a synonym of TREE_OVERFLOW.
1876
1877 2007-01-11 Paolo Bonzini <bonzini@gnu.org>
1878
1879 * configure.ac (strict1_warn): Rename to strict_warn.
1880 (WERROR, --enable-werror, symlink hacks, stage1_cflags,
1881 cc_set_by_configure, quoted_cc_set_by_configure,
1882 stage_prefix_set_by_configure, quoted_stage_prefix_set_by_configure,
1883 all_boot_languages, all_stagestuff): Remove.
1884 (target_list): Remove bootstrap targets.
1885 * Makefile.in (quickstrap): Unconditionally make a synonym of all.
1886 (BOOT_LANGUAGES, STAGE1_CFLAGS, STAGE1_CHECKING,
1887 REMAKEFLAGS, FLAGS_TO_PASS, PREPEND_DOTDOT_TO_RELATIVE_PATHS,
1888 SUBDIR_FLAGS_TO_PASS, WERROR_FLAGS, STRICT2_WARN, LANG_STAGESTUFF,
1889 VOL_FILES, POSTSTAGE1_FLAGS_TO_PASS, STAGE2_FLAGS_TO_PASS,
1890 STAGEPROFILE_FLAGS_TO_PASS, STAGEFEEDBACK_FLAGS_TO_PASS, stage1_build,
1891 stage1_copy, stage2_build, stage2_copy, stageprofile_build,
1892 stageprofile_copy, stage3_build, stage3_copy, stagefeedback_build,
1893 stagefeedback_copy, stage4_build, clean_s1, clean_sw, bootstrap,
1894 bootstrap-lean, bootstrap2, bootstrap2-lean, bootstrap3,
1895 bootstrap3-lean, bootstrap4, bootstrap4-lean, unstage1, unstage2,
1896 unstage3, unstage4, unstageprofile, unstagefeedback, restage, restage2,
1897 restage3, restage4, restageprofile, restagefeedback, bubbleestrap,
1898 cleanstrap, unstrap, restrap, *compare, *compare3, *compare4,
1899 *compare-lean, *compare3-lean, *compare4-lean, stage1-start, stage1,
1900 stage2-start, stage2, stage3-start, stage3, stage4-start, stage4,
1901 stageprofile-start, stageprofile, stagefeedback-start, stagefeedback,
1902 risky-stage1, risky-stage2, risky-stage3, risky-stage4): Remove.
1903 (ORDINARY_FLAGS_TO_PASS): Rename to FLAGS_TO_PASS.
1904 (STAGECOPYSTUFF, STAGEMOVESTUFF): Consolidate into MOSTLYCLEANFILES.
1905 (mostlyclean): Adjust.
1906 (clean, distclean): Don't mention bootstrap stuff.
1907 * configure: Regenerate.
1908 * ada/config-lang.in, cp/config-lang.in, forttran/config-lang.in,
1909 java/config-lang.in, objc/config-lang.in, objcp/config-lang.in,
1910 treelang/config-lang.in (stagestuff): Remove.
1911 * doc/sourcebuild.texi (stage1, stage2, stage3, stage4,
1912 stageprofile, stagefeedback, stagestuff): Remove mention.
1913
1914 2007-01-11 Nick Clifton <nickc@redhat.com>
1915
1916 * config/mcore/predicates.md (mcore_general_movesrc_operand):
1917 Accept CONSTs.
1918 (mcore_general_movdst_operand): Do not accept CONST_INTs.
1919 (mcore_arith_K_S_operand): Run the test for the S constraint not
1920 the test for the M constraint.
1921 (mcore_addsub_operand): Do not accept integer values that are
1922 larger than 32 bits.
1923 * config/mcore/mcore.md: Remove unused constraints from split.
1924 (andsi3): Use HOST_WIDE_INT instead of int to hold an INTVAL.
1925 (addsi3): Likewise.
1926 (allocate_stack): Likewise.
1927 * config/mcore/mcore.c (mcore_print_operand): Restrict output of P
1928 operands to 32 bits.
1929 (mcore_const_costs): Use HOST_WIDE_INT instead of int to hold an
1930 INTVAL.
1931 (mcore_and_cost, mcore_modify_comparison, const_ok_for_mcore,
1932 mcore_const_ok_for_inline, mcore_const_trick_uses_not,
1933 try_constant_tricks, mcore_num_ones, mcore_num_zeros,
1934 mcore_output_bclri, mcore_output_andn, output_inline_const,
1935 mcore_output_move, mcore_output_movedouble): Likewise.
1936 (mcore_output_cmov): Use CONST_OK_FOR_M and CONST_OK_FOR_N.
1937 (output_inline_const): Likewise.
1938 (output_inline_const): Fix format strings used in sprintf
1939 statements.
1940 * config/mcore/mcore-protos.h: Update prototypes for changed
1941 functions in mcore.c.
1942 * config/mcore/mcore.h (CONST_OK_FOR_I): Cast values to
1943 HOST_WIDE_INT and not int.
1944 (CONST_OK_FOR_J, CONST_OK_FOR_K, CONST_OK_FOR_L, CONST_OK_FOR_M,
1945 CONST_OK_FOR_N): Likewise.
1946 (LEGITIMATE_CONSTANT_P): Also check CONSTANT_P.
1947 (GO_IF_LEGITIMATE_INDEX): Use HOST_WIDE_INT instead of int to hold
1948 an INTVAL.
1949
1950 2007-01-10 Jan Hubicka <jh@suse.cz>
1951
1952 * tree-vrp.c (remove_range_assertions): Release defs.
1953 * tree-ssa-loop-ivopts.c (rmeove_statement): Likewise.
1954 * tree-ssa-dom.c (remove_stmt_or_phi): Likewise.
1955
1956 2007-01-10 Paul Brook <paul@codesourcery.com>
1957
1958 * config/arm/arm.c (arm_rtx_costs_1): Handle mutiply-subtract.
1959 * config/arm/arm.md (mulsi3subsi): New insn.
1960
1961 2007-01-10 Zdenek Dvorak <dvorakz@suse.cz>
1962
1963 * tree-ssa-loop-manip.c (tree_unroll_loop): Make it a wrapper over ...
1964 (tree_transform_and_unroll_loop): New.
1965 * tree-flow.h (transform_callback, tree_transform_and_unroll_loop):
1966 Declare.
1967
1968 2007-01-10 Robert Kennedy <jimbob@google.com>
1969
1970 * fold-const.c (fold_comparison): Fold comparisons like (x *
1971 1000 < 0) to (x < 0).
1972
1973 2007-01-10 Ian Lance Taylor <iant@google.com>
1974
1975 * tree-pretty-print.c (dump_generic_node): Print parentheses when
1976 operands have the same priority.
1977
1978 2007-01-10 Tom Tromey <tromey@redhat.com>
1979
1980 * fold-const.c (fold_truthop): Don't check can_use_bit_fields_p.
1981 (fold_binary): Likewise.
1982 * langhooks.c (lhd_can_use_bit_fields_p): Removed.
1983 * langhooks-def.h (lhd_can_use_bit_fields_p): Removed.
1984 (LANG_HOOKS_CAN_USE_BIT_FIELDS_P): Removed.
1985 (LANG_HOOKS_INITIALIZER): Remove LANG_HOOKS_CAN_USE_BIT_FIELDS_P.
1986 * langhooks.h (struct lang_hooks): Removed field
1987 'can_use_bit_fields_p'.
1988
1989 2007-01-10 Ralf Corsépius <ralf.corsepius@rtems.org>
1990
1991 * config/bfin/t-bfin, config/bfin/t-bfin-elf: Remove GCC_CFLAGS.
1992
1993 2007-01-10 Razya Ladelsky <razya@il.ibm.com>
1994
1995 * function.c (get_last_funcdef_no): New function.
1996 * function.h (get_last_funcdef_no): Declare.
1997 * tree-inline.c (initialize_cfun): Add initialization.
1998 (tree_function_versioning): Cleanup.
1999
2000 2007-01-10 Jan Hubicka <jh@suse.cz>
2001
2002 * tree-inline.c (setup_one_parameter): Do not propagate into abnormal
2003 PHIs.
2004
2005 2007-01-10 Sa Liu <saliu@de.ibm.com>
2006 Ben Elliston <bje@au.ibm.com>
2007
2008 * spu.h (STACK_SAVE_AREA): Use VOIDmode for SAVE_FUNCTION, SImode
2009 for SAVE_NONLOCAL and Pmode for any other save level.
2010 * spu-protos.h (spu_restore_stack_block): Declare.
2011 * spu.md (save_stack_block): Remove.
2012 (restore_stack_block): Call spu_restore_stack_block.
2013 * spu.c (spu_restore_stack_block): New function.
2014 (spu_expand_epilogue): Remove old comment.
2015
2016 2007-01-09 Zdenek Dvorak <dvorakz@suse.cz>
2017
2018 PR tree-optimization/30322
2019 * tree-ssa-loop-ivopts.c (fold_affine_expr, iv_value): Removed.
2020 (cand_value_at): Return the value as aff_tree.
2021 (may_eliminate_iv): Convert the bound from aff_tree to tree.
2022 * tree-affine.c (aff_combination_add_cst, aff_combination_add_product,
2023 aff_combination_mult): New functions.
2024 (aff_combination_add): Use aff_combination_add_cst.
2025 (aff_combination_convert): Allow conversions to a wider type.
2026 (tree_to_aff_combination): Handle BIT_NOT_EXPR.
2027 * tree-affine.h (aff_combination_mult): Declare.
2028
2029 2007-01-09 Carlos O'Donell <carlos@codesourcery.com>
2030
2031 * doc/tm.texi: Update documentation to reflect reality of exec
2032 and start file search behaviours. Update copyright year.
2033 * doc/invoke.texi: Explain how GCC_EXEC_PREFIX is used to find
2034 header file directories.
2035
2036 2007-01-09 Uros Bizjak <ubizjak@gmail.com>
2037
2038 * config/i386/i386.md (*sinxf2): Rename to *sinxf2_i387.
2039 (*cosxf2): Rename to cosxf2_i387.
2040 (*sindf2, *sinsf2): Extend operand 1 to XFmode. Macroize patterns
2041 using X87MODEF12 mode macro. Rename patterns to
2042 *sin_extend<mode>xf2_i387. Use SSE_FLOAT_MODE_P to disable patterns
2043 for SSE math.
2044 (*cosdf2, *cossf2): Ditto.
2045 (sincosdf3, sincossf3): Ditto. Rewrite corresponding splitters
2046 to match extended input operands.
2047 (sincos<mode>3): New expander.
2048 (*sinextendsfdf2, *cosextendsfdf2, *sincosextendsfdf3): Remove
2049 insn patterns and corresponding splitters.
2050
2051 2007-01-09 Kaz Kojima <kkojima@gcc.gnu.org>
2052
2053 * config/sh/t-linux (TARGET_LIBGCC2_CFLAGS): Delete.
2054 (SHLIB_MAPFILES, SHLIB_LINK, SHLIB_INSTALL): Likewise.
2055
2056 2007-01-09 Nicolas Pitre <nico@cam.org>
2057
2058 PR target/30173
2059 * arm/ieee754-df.S (Lad_s): Also test the low word of X for zero.
2060
2061 2007-01-08 Geoffrey Keating <geoffk@apple.com>
2062
2063 * target.h (struct gcc_target): New field library_rtti_comdat.
2064 * target-def.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): New.
2065 (TARGET_CXX): Add TARGET_CXX_LIBRARY_RTTI_COMDAT.
2066 * doc/tm.texi (C++ ABI): Document TARGET_CXX_LIBRARY_RTTI_COMDAT.
2067 * config/darwin.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
2068
2069 2007-01-08 Geoffrey Keating <geoffk@apple.com>
2070
2071 * doc/invoke.texi (Optimize Options): Correct description of -O0.
2072
2073 2007-01-08 Richard Guenther <rguenther@suse.de>
2074
2075 * tree.h (force_fit_type_double): Export.
2076 (force_fit_type): Remove.
2077 * fold-const.c (force_fit_type_double): New function.
2078 (force_fit_type): Remove.
2079 (int_const_binop): Use it.
2080 (fold_convert_const_int_from_int): Likewise.
2081 (fold_convert_const_int_from_real): Likewise.
2082 (fold_div_compare): Likewise.
2083 (fold_sign_changed_comparison): Likewise.
2084 (fold_unary): Likewise.
2085 (fold_negate_const): Likewise.
2086 (fold_abs_const): Likewise.
2087 (fold_not_const): Likewise.
2088 * c-common.c (shorten_compare): Use force_fit_type_double.
2089 * convert.c (convert_to_pointer): Likewise.
2090
2091 2007-01-08 Richard Guenther <rguenther@suse.de>
2092
2093 * tree.h (build_int_cst_wide_type): Export.
2094 * tree.c (build_int_cst_wide_type): New function.
2095 (build_int_cst_wide): Fix comment.
2096 * builtins.c (fold_builtin_object_size): Use build_int_cst
2097 to build -1 or 0 of the correct type. Use fit_double_type
2098 to check for overflow.
2099 * fold-const.c (optimize_bit_field_compare): Use build_int_cst_type
2100 to build the mask.
2101 (decode_field_reference): Likewise.
2102 (all_ones_mask_p): Likewise.
2103 (native_interpret_int): Use build_int_cst_wide_type.
2104 (fold_binary): Use build_int_cst_type to build an all-ones
2105 value.
2106 * stor-layout.c (set_sizetype): Use build_int_cst_wide_type.
2107
2108 2007-01-08 Daniel Jacobowitz <dan@codesourcery.com>
2109
2110 * config/pa/t-pa64 (libgcc_stub.a): Use $(T).
2111
2112 2007-01-09 Ben Elliston <bje@au.ibm.com>
2113
2114 * genautomata.c (STATS_OPTION): New option.
2115 (stats_flag): New flag.
2116 (gen_automata_option): Handle it.
2117 (initiate_automaton_gen): Ditto.
2118 (write_automata): Output statistics only if stats_flag is
2119 set. Likewise, output time statistics only if time_flag is set.
2120 * doc/md.texi (Processor pipeline description): Document new flag.
2121
2122 2007-01-08 Richard Guenther <rguenther@suse.de>
2123
2124 * builtins.c (fold_builtin_int_roundingfn): Use fit_double_type.
2125 * tree.c (build_int_cst_type): Likewise.
2126 (size_in_bytes): Don't call force_fit_type on the result.
2127 (int_fits_type_p): Use fit_double_type.
2128 * fold-const.c (fit_double_type): New function.
2129 (force_fit_type): Use it.
2130 * tree.h (fit_double_type): Export.
2131
2132 2007-01-08 Jan Hubicka <jh@suse.cz>
2133
2134 * tree-vectorizer.c (gate_increase_alignment): Fix return type.
2135 * ipa.c (function_and_variable_visibility): Fix return type.
2136
2137 2007-01-08 Richard Guenther <rguenther@suse.de>
2138
2139 * tree-ssa-ccp.c (maybe_fold_offset_to_array_ref): Use type
2140 of offset to build the index.
2141 * tree-pretty-print.c (dump_generic_node): Don't build negated
2142 const just for printing.
2143 * c-pretty-print.c (pp_c_integer_constant): Likewise.
2144 * builtins.c (fold_builtin_int_roundingfn): Check if result
2145 fits the type by using force_fit_type and comparing the result.
2146 * predict.c (predict_loops): Use compare_tree_int for comparison.
2147 * tree.c (build_int_cst): Fall back to integer_type_node for
2148 NULL_TREE type.
2149 (build_int_cst_wide): Assert type is non-null.
2150
2151 2007-01-08 Roberto Costa <roberto.costa@st.com>
2152
2153 * tree-vrp.c (extract_range_from_cond_expr): New.
2154 (extract_range_from_expr): Handle COND_EXPR nodes used as expressions.
2155 * tree-ssa-ccp.c (get_maxval_strlen): Handle COND_EXPR nodes used
2156 as expressions.
2157 (fold_stmt): Bug fix, avoid infinite recursion when folding COND_EXPRs.
2158 * tree-ssa-forwprop.c (simplify_cond, forward_propagate_into_cond,
2159 tree_ssa_forward_propagate_single_use_vars): Handle COND_EXPR nodes
2160 used as expressions.
2161 * tree-object-size.c (cond_expr_object_size): New.
2162 (collect_object_sizes_for): Handle COND_EXPR nodes used as expressions.
2163
2164 2007-01-08 Jan Hubicka <jh@suse.cz>
2165
2166 * tree-ssa-forwprop.c (forward_propagate_into_cond,
2167 tree_ssa_forward_propagate_single_use_va): Release defs of propagated
2168 statement.
2169
2170 2007-01-08 Richard Guenther <rguenther@suse.de>
2171
2172 PR tree-optimization/23603
2173 * tree-vrp.c (set_value_range_to_truthvalue): New function.
2174 (extract_range_from_binary): Fall back to truthvalue instead of
2175 varying for TRUTH_*_EXPR.
2176 (extract_range_from_comparison): Fall back to truthvalue instead of
2177 varying.
2178 (vrp_visit_phi_node): Don't adjust new range bounds to +INF/-INF
2179 if all visited PHI values were constant.
2180
2181 2007-01-08 Jan Hubicka <jh@suse.cz>
2182
2183 * cgraphunit.c (cgraph_process_new_functions): Reset reachable flag.
2184 (cgraph_analyze_function): break out from ...
2185 (cgraph_finalize_compilation_unit): ... here.
2186 (cgraph_expand_function): Remove forgoten commented out line.
2187 (cgraph_optimize): Analyze functions.
2188
2189 2007-01-08 Jan Hubicka <jh@suse.cz>
2190
2191 * tree-pas.h (TODO_remove_function): New flag.
2192 (TODO_update*): Renumber.
2193 (pass_ipa_increase_alignment,
2194 pass_ipa_function_and_variable_visibility): New passes.
2195 * cgraphunit.c (cgraph_increase_alignment): Move to tree-vectorizer.c
2196 (cgraph_function_and_variable_visibility): Move to ipa.c
2197 (cgraph_optimize): Don't call cgraph_function_and_variable_visibility,
2198 cgraph_increase_alignment.
2199 * ipa-inline.c (cgraph_decide_inlining): Don't push timevar.
2200 (cgraph_decide_inlining_incrementally): Push TV_INTEGRATION before
2201 calling tree-inline.
2202 (cgraph_early_inlining): Do not call cgraph_remove_unreachable_nodes.
2203 (pass_ipa_inline, pass_early_ipa_inlining): Set TODO_remove_functions
2204 * tree-vectorizer.c (increase_alignment): Move here from cgraphunit.c
2205 (gate_increase_alignment): New function.
2206 (pass_ipa_increase_alignment): New pass.
2207 * ipa.c: Inline tree-pass.h and timevar.h
2208 (function_and_variable_visibility): Move here from cgraphunit.c
2209 * tree-optimize.c (pass_early_local_passes): Add TODO_remove_functions.
2210 * passes.c (init_optimization_passes): Add the two new passes.
2211 (execute_todo): Handle cgraph_remove_functions.
2212
2213 2007-01-08 Nick Clifton <nickc@redhat.com>
2214
2215 * config/frv/predicates.md (reg_or_0_operand): Accept
2216 CONST_DOUBLEs.
2217
2218 2007-01-08 Ralf Corsépius <ralf.corsepius@rtems.org>
2219
2220 * config/bfin/rtems.h, config/bfin/t-rtems: New.
2221 * config.gcc: Add bfin*-rtems*.
2222
2223 2007-01-08 Mark Shinwell <shinwell@codesourcery.com>
2224
2225 * c.opt: Add -flax-vector-conversions.
2226 * c-typeck.c (convert_for_assignment): Pass flag to
2227 vector_types_convertible_p to allow emission of note.
2228 (digest_init): Likewise.
2229 * c-opts.c: Handle -flax-vector-conversions.
2230 * c-common.c (flag_lax_vector_conversions): New.
2231 (vector_types_convertible_p): Unless -flax-vector conversions
2232 has been passed, disallow conversions between vectors with
2233 differing numbers of subparts and/or element types. If such
2234 a conversion is disallowed, possibly emit a note on the first
2235 occasion only to inform the user of -flax-vector-conversions.
2236 The new last argument specifies this.
2237 * c-common.h (flag_lax_vector_conversions): New.
2238 (vector_types_convertible_p): Add extra argument.
2239 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Use
2240 char_type_node for V*QI type vectors.
2241 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins):
2242 Update to satisfy new typechecking rules.
2243 * config/rs6000/altivec.h (vec_cmple): Use vec_cmpge, for both
2244 C and C++ variants.
2245 * doc/invoke.texi (C Dialect Options): Document
2246 -flax-vector-conversions.
2247
2248 2007-01-08 Mark Shinwell <shinwell@codesourcery.com>
2249
2250 PR tree-optimization/29877
2251 * tree-ssa-ter.c (is_replaceable_p): Deem assignments with
2252 a register variable on the RHS to not be replaceable.
2253
2254 2007-01-08 Chen Liqin <liqin@sunnorth.com.cn>
2255 * config/score/t-score-elf (MULTILIB_OPTIONS): Change.
2256 * config/score/predicates.md (const_uimm5, sr0_operand, const_simm12,
2257 const_simm15, const_pow2, const_npow2): Added.
2258 * config/score/misc.md (insv, extv, extzv, movmemsi,
2259 move_lbu_a/b, mov_lhu_a/b etc): Added and fix some bug.
2260 * config/score/score.c (score_address_cost, score_select_cc_mode):
2261 Added.
2262 Change CONST_OK_FOR_LETTER_P/EXTRA_CONSTRAINT define.
2263 Update score_rtx_costs for MACRO TARGET_RTX_COSTS.
2264 Update score_print_operand.
2265 * config/score/score.h (DATA_ALIGNMENT, SELECT_CC_MODE): Added.
2266 Adjust register allocate order and update some macro define.
2267 * config/score/score-mdaux.c (mdx_unaligned_load, mdx_unsigned_store,
2268 mdx_block_move_straight, mdx_block_move_loop_head,
2269 mdx_block_move_loop_body, mdx_block_move_loop_foot, mdx_block_move_loop,
2270 mdx_block_move): Added.
2271 (mdx_movsicc, mdp_select_add_imm, mdp_select, mds_zero_extract_andi,
2272 mdp_limm): Updated and fix some bug and typo.
2273 * config/score/score.md (movqi/hi/si, add/sub/zero/ext): Updated.
2274 (movsf, movdf, doloop_end): Added.
2275
2276 2007-01-08 Kazu Hirata <kazu@codesourcery.com>
2277
2278 * config/arm/arm.c, config/arm/arm.h, config/arm/arm.md,
2279 config/arm/thumb2.md: Fix comment typos.
2280 * doc/extend.texi: Fix a typo.
2281
2282 2007-01-07 Eric Christopher <echristo@apple.com>
2283
2284 * configure.ac: Check for __stack_chk_fail for darwin.
2285 * configure: Regenerate.
2286
2287 2007-01-07 Richard Guenther <rguenther@suse.de>
2288
2289 * tree-vrp.c (extract_range_from_assert): CSE calls to
2290 compare_values where possible.
2291 (extract_range_from_unary_expr): Likewise.
2292
2293 2007-01-07 Anatoly Sokolov <aesok@post.ru>
2294
2295 * config/avr/avr-protos.h (call_insn_operand): Delete prototype.
2296 * config/avr/avr.c (call_insn_operand): Delete function.
2297 * config/avr/avr.md (*pushqi, *pushhi, *pushsi, *pushsf): Use REG_SP
2298 instead of register number. Use predicates.
2299 * config/avr/predicates.md (const0_operand, reg_or_0_operand,
2300 call_insn_operand): Add.
2301
2302 2007-01-06 Jan Hubicka <jh@suse.cz>
2303
2304 * tree-pass.h (pass_build_cgraph_edges): Declare.
2305 * cgraphunit.c (record_refernece): Move to cgraphbuild.c
2306 (visited_nodes): Remove.
2307 (cgraph_create_edges): Move to cgraphbuild.c; rename to
2308 build_cgrpah_edges; make visited_nodes local.
2309 (cgraph_process_new_functions): DO not call initialize_inline_failed.
2310 (record_references_in_initializer): Move to cgraphbuild.c
2311 (initialize_inline_failed, rebuild_cgraph_edges,
2312 pass_rebuild_cgraph_edges): Move to cgraphbuild.c.
2313 (verify_cgraph_node): Make visited_nodes local.
2314 (cgraph_analyze_function): Do not call cgraph_create_edges and
2315 initialize_inline_failed.
2316 (cgraph_expand_function): Do not call cgraph_lower_function;
2317 assert that function is already lowered.
2318 * Makefile.in (cgraphbuild.o): New.
2319 * passes.c (init_optimization_passes): Add pass_build_cgraph_edges
2320 at the end of lowering passes.
2321
2322 2007-01-06 Steven Bosscher <steven@gcc.gnu.org>
2323
2324 * ifcvt.c (cond_move_convert_if_block): New function, code
2325 factored out from...
2326 (cond_move_process_if_block): ...here. Call the new function
2327 on the THEN and ELSE blocks.
2328 (merge_if_block): Do not copy global_live_at_end, merge_blocks
2329 already takes care of this.
2330
2331 2007-01-05 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
2332
2333 PR c/19978
2334 * tree.h (TREE_OVERFLOW_P): New.
2335 * c-typeck.c (parser_build_unary_op): Warn only if result
2336 overflowed and operands did not.
2337 (parser_build_binary_op): Likewise.
2338 (convert_for_assignment): Remove redundant overflow_warning.
2339 * c-common.c (overflow_warning): Don't check or set TREE_OVERFLOW.
2340
2341 2007-01-05 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
2342
2343 * c-typeck.c (store_init_value): Split over two lines to follow
2344 the GNU coding style.
2345
2346 2007-01-05 Benjamin Kosnik <bkoz@redhat.com>
2347
2348 * c-cppbuiltin.c (c_cpp_builtins): __GXX_EXPERIMENTAL_CPP0X__ to
2349 __GXX_EXPERIMENTAL_CXX0X__.
2350 * doc/cpp.texi: Same.
2351
2352 2007-01-05 Richard Guenther <rguenther@suse.de>
2353
2354 PR middle-end/27826
2355 * tree.c (get_narrower): Do not construct COMPONENT_REFs
2356 with mismatched types. Instead explicitly build a
2357 conversion NOP_EXPR.
2358
2359 2007-01-05 Ian Lance Taylor <iant@google.com>
2360
2361 * c-common.c (decl_with_nonnull_addr_p): New function.
2362 (c_common_truthvalue_conversion): Call it.
2363 * c-typeck.c (build_binary_op): Likewise.
2364 * c-common.h (decl_with_nonnull_addr_p): Declare.
2365
2366 2007-01-05 Jakub Jelinek <jakub@redhat.com>
2367
2368 PR c/30360
2369 * libgcc2.c (__divdc3): Compare c and d against 0.0 instead of
2370 denom against 0.0.
2371
2372 2007-01-05 Joel Brobecker <brobecker@adacore.com>
2373
2374 * doc/install.texi (Final install): Document the fact that
2375 the GNAT runtime should not be stripped.
2376
2377 2007-01-04 Jan Hubicka <jh@suse.cz>
2378
2379 * tree-inline.c (fold_marked_statements): Update operand caches
2380 and EH after folding
2381
2382 2007-01-04 Ian Lance Taylor <iant@google.com>
2383
2384 * c-common.c (check_function_nonnull): Whitespace fix.
2385
2386 2007-01-04 Jan Hubicka <jh@suse.cz>
2387
2388 * tree-optimize.c (execute_fixup_cfg): Correct previously mistakely
2389 comitted older version of patch.
2390 (pass_fixup_cfg): Add TODOs to verify flow and statements, dump
2391 function, celanup cfg and collect garbage.
2392
2393 2007-01-04 Mike Stump <mrs@apple.com>
2394
2395 * Makefile.in (mostlyclean): Don't remove libgcc anymore.
2396 (clean): Likewise.
2397
2398 2007-01-04 Eric Christopher <echristo@apple.com>
2399
2400 * libgcc2.c (__bswapsi2): Use SItype.
2401 (__bswapdi2): Use DItype.
2402 * libgcc2.h: Update for above.
2403
2404 2007-01-04 Paul Brook <paul@codesourcery.com>
2405
2406 * config/arm/arm.md (arm_mulsi3, thumb_mulsi3, mulsi3_compare0,
2407 mulsi_compare0_scratch, mulsi3addsi, mulsi3addsi_compare0,
2408 mulsi3addsi_compare0_scratch, mulsidi3adddi, mulsidi3,
2409 umulsidi3, umulsidi3adddi, smulsi3_highpart,
2410 umulsi3_highpart): Make conditional on !arm_arch6.
2411 (arm_mulsi3_v6, thumb_mulsi3_v6, mulsi3_compare0_v6,
2412 mulsi_compare0_scratch_v6, mulsi3addsi_v6, mulsi3addsi_compare0_v6,
2413 mulsi3addsi_compare0_scratch_v6, mulsidi3adddi_v6, mulsidi3_v6,
2414 umulsidi3_v6, umulsidi3adddi_v6, smulsi3_highpart_v6,
2415 umulsi3_highpart_v6): New insns.
2416
2417 2007-01-04 Roger Sayle <roger@eyesopen.com>
2418
2419 * fold-const.c (fold_convert): When casting an expression to void,
2420 fold_ignored_result may discover a GIMPLE_MODIFY_STMT which doesn't
2421 have a type. Instead of attempting to build a NOP_EXPR, return
2422 these "special" trees directly.
2423
2424 2007-01-04 Joseph Myers <joseph@codesourcery.com>
2425
2426 * config/rs6000/rs6000.c (rs6000_rtx_costs): Make adjustment for
2427 MULT inside MINUS as either argument. Use rs6000_cost->dmul -
2428 rs6000_cost->fp not 0 as adjustment for outer NEG.
2429
2430 2007-01-04 Jan Hubicka <jh@suse.cz>
2431
2432 * cgraph.c (cgraph_release_function_body): New function.
2433 (cgraph_remove_node): Use it.
2434 * cgraph.h (cgraph_release_function_body): Declare.
2435 * cgraphunit.c (cgraph_expand_function): Use it.
2436 * ipa.c (cgraph_remove_unreahchable_nodes): Use it.
2437 * tree-ssa.c (delete_tree_ssa): Allow to be called before aliasing
2438 is initialized and while compilation of other function is running.
2439 * tree-optimize.c (execute_free_cfg_annotations): Move code to clear
2440 statement CFG annotations from here to ...
2441 * tree-cfg.c (delete_tree_cfg_annotations): ... here.
2442
2443 2007-01-04 Zdenek Dvorak <dvorakz@suse.cz>
2444
2445 * cfgloop.h (enum li_flags): Make the constants powers of two.
2446
2447 2007-01-04 Jan Hubicka <jh@suse.cz>
2448
2449 * tree-inline.c (copy_bb): Insert new statements to statements_to_fold
2450 set.
2451 (fold_marked_statements): New function.
2452 (optimize_inline_calls, tree_function_versioning): Fold new statements.
2453 * tree-inline.h (copy_body_data): Add statements_to_fold.
2454
2455 2007-01-03 Daniel Jacobowitz <dan@codesourcery.com>
2456
2457 * config.gcc: Mention libgcc/config.host.
2458 * Makefile.in: Update comments mentioning libgcc.
2459 (LIBGCC, INSTALL_LIBGCC, GCC_PARTS, mklibgcc): Delete.
2460 (all.cross, start.encap, rest.encap, rest.cross): Update
2461 dependencies for libgcc move.
2462 (libgcc.mk, LIBGCC_DEPS, libgcov.a, libgcc.a, stmp-multilib)
2463 (clean-target, clean-target-libgcc): Delete.
2464 (srcdirify, GCC_EXTRA_PARTS): New macros.
2465 (libgcc-support, libgcc.mvars): New rules.
2466 (distclean): Remove mention of mklibgcc.
2467 (install): Don't reference INSTALL_LIBGCC.
2468 (install-common): Don't reference EXTRA_PARTS.
2469 (install-libgcc, install-multilib): Delete rules.
2470 * mklibgcc.in: Delete file.
2471 * doc/configfiles.texi: Don't mention mklibgcc.
2472
2473 * config/i386/t-darwin (SHLIB_VERPFX): Delete (moved to libgcc).
2474 * config/i386/t-darwin64 (SHLIB_VERPFX): Likewise.
2475 * config/rs6000/t-darwin (SHLIB_VERPFX): Likewise.
2476 * config/rs6000/t-ppccomm (TARGET_LIBGCC2_CFLAGS, SHLIB_MAPFILES)
2477 (mklibgcc, ldblspecs): Likewise.
2478
2479 * config/i386/t-nwld (libgcc.def, libc.def, libpcre.def)
2480 (posixpre.def): Use $(T).
2481 (SHLIB_EXT, SHLIB_NAME, SHLIB_SLIBDIR_QUAL, SHLIB_DEF, SHLIB_MAP)
2482 (SHLIB_SRC, SHLIB_INSTALL): Delete.
2483 (SHLIB_LINK): Make dummy.
2484 * config/t-slibgcc-darwin: Delete contents except for dummy SHLIB_LINK.
2485
2486 * config/frv/t-linux (EXTRA_MULTILIB_PARTS): Clear.
2487
2488 * config/alpha/t-crtfm: Use $(T) in rules for EXTRA_PARTS.
2489 * config/alpha/t-vms, config/alpha/t-vms64, config/fr30/t-fr30,
2490 config/i386/t-rtems-i386, config/ia64/t-ia64, config/rs6000/t-beos,
2491 config/rs6000/t-newas, config/sparc/t-elf: Likewise.
2492
2493 * configure.ac (all_outputs): Remove mklibgcc.
2494 * configure: Regenerated.
2495
2496 2007-01-03 Josh Conner <jconner@apple.com>
2497
2498 PR middle-end/29683
2499 * calls.c (compute_argument_addresses): Set stack and stack_slot
2500 for partial args, too.
2501 (store_one_arg): Use locate.size.constant for the size when
2502 generating a save_area.
2503
2504 2007-01-03 Robert Kennedy <jimbob@google.com>
2505
2506 * tree-cfg.c (tree_merge_blocks): Release SSA_NAME phi results
2507 whose definitions are deleted due to basic block merging.
2508
2509 2007-01-03 Paul Brook <paul@codesourcery.com>
2510
2511 PR target/16634
2512 * config/arm/arm.c (output_return_instruction): Pop PC in interrupt
2513 functions.
2514 (use_return_insn): Return 0 for Thumb interrupt functions.
2515 (print_multi_reg): Add rfe argument for IRQ returns.
2516 (arm_output_epilogue): Pop interrupt return address directly into PC.
2517 (arm_expand_prologue): Only adjust IRQ return address in Arm mode.
2518
2519 2007-01-03 Paul Brook <paul@codesourcery.com>
2520
2521 Merge from sourcerygxx-4_1.
2522 * config/arm/thumb2.md: New file.
2523 * config/arm/elf.h (JUMP_TABLES_IN_TEXT_SECTION): Return True for
2524 Thumb-2.
2525 * config/arm/coff.h (JUMP_TABLES_IN_TEXT_SECTION): Ditto.
2526 * config/arm/aout.h (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
2527 (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump tables.
2528 * config/arm/aof.h (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump
2529 tables.
2530 (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
2531 * config/arm/ieee754-df.S: Use macros for Thumb-2/Unified asm
2532 comptibility.
2533 * config/arm/ieee754-sf.S: Ditto.
2534 * config/arm/arm.c (thumb_base_register_rtx_p): Rename...
2535 (thumb1_base_register_rtx_p): ... to this.
2536 (thumb_index_register_rtx_p): Rename...
2537 (thumb1_index_register_rtx_p): ... to this.
2538 (thumb_output_function_prologue): Rename...
2539 (thumb1_output_function_prologue): ... to this.
2540 (thumb_legitimate_address_p): Rename...
2541 (thumb1_legitimate_address_p): ... to this.
2542 (thumb_rtx_costs): Rename...
2543 (thumb1_rtx_costs): ... to this.
2544 (thumb_compute_save_reg_mask): Rename...
2545 (thumb1_compute_save_reg_mask): ... to this.
2546 (thumb_final_prescan_insn): Rename...
2547 (thumb1_final_prescan_insn): ... to this.
2548 (thumb_expand_epilogue): Rename...
2549 (thumb1_expand_epilogue): ... to this.
2550 (arm_unwind_emit_stm): Rename...
2551 (arm_unwind_emit_sequence): ... to this.
2552 (thumb2_legitimate_index_p, thumb2_legitimate_address_p,
2553 thumb1_compute_save_reg_mask, arm_dwarf_handle_frame_unspec,
2554 thumb2_index_mul_operand, output_move_vfp, arm_shift_nmem,
2555 arm_save_coproc_regs, thumb_set_frame_pointer, arm_print_condition,
2556 thumb2_final_prescan_insn, thumb2_asm_output_opcode, arm_output_shift,
2557 thumb2_output_casesi): New functions.
2558 (TARGET_DWARF_HANDLE_FRAME_UNSPEC): Define.
2559 (FL_THUMB2, FL_NOTM, FL_DIV, FL_FOR_ARCH6T2, FL_FOR_ARCH7,
2560 FL_FOR_ARCH7A, FL_FOR_ARCH7R, FL_FOR_ARCH7M, ARM_LSL_NAME,
2561 THUMB2_WORK_REGS): Define.
2562 (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv, arm_condexec_count,
2563 arm_condexec_mask, arm_condexec_masklen)): New variables.
2564 (all_architectures): Add armv6t2, armv7, armv7a, armv7r and armv7m.
2565 (arm_override_options): Check new CPU capabilities.
2566 Set new architecture flag variables.
2567 (arm_isr_value): Handle v7m interrupt functions.
2568 (user_return_insn): Return 0 for v7m interrupt functions. Handle
2569 Thumb-2.
2570 (const_ok_for_arm): Handle Thumb-2 constants.
2571 (arm_gen_constant): Ditto. Use movw when available.
2572 (arm_function_ok_for_sibcall): Return false for v7m interrupt
2573 functions.
2574 (legitimize_pic_address, arm_call_tls_get_addr): Handle Thumb-2.
2575 (thumb_find_work_register, arm_load_pic_register,
2576 legitimize_tls_address, arm_address_cost, load_multiple_sequence,
2577 emit_ldm_seq, emit_stm_seq, arm_select_cc_mode, get_jump_table_size,
2578 print_multi_reg, output_mov_long_double_fpa_from_arm,
2579 output_mov_long_double_arm_from_fpa, output_mov_double_fpa_from_arm,
2580 output_mov_double_fpa_from_arm, output_move_double,
2581 arm_compute_save_reg_mask, arm_compute_save_reg0_reg12_mask,
2582 output_return_instruction, arm_output_function_prologue,
2583 arm_output_epilogue, arm_get_frame_offsets, arm_regno_class,
2584 arm_output_mi_thunk, thumb_set_return_address): Ditto.
2585 (arm_expand_prologue): Handle Thumb-2. Use arm_save_coproc_regs.
2586 (arm_coproc_mem_operand): Allow POST_INC/PRE_DEC.
2587 (arithmetic_instr, shift_op): Use arm_shift_nmem.
2588 (arm_print_operand): Use arm_print_condition. Handle '(', ')', '.',
2589 '!' and 'L'.
2590 (arm_final_prescan_insn): Use extract_constrain_insn_cached.
2591 (thumb_expand_prologue): Use thumb_set_frame_pointer.
2592 (arm_file_start): Output directive for unified syntax.
2593 (arm_unwind_emit_set): Handle stack alignment instruction.
2594 * config/arm/lib1funcs.asm: Remove default for __ARM_ARCH__.
2595 Add v6t2, v7, v7a, v7r and v7m.
2596 (RETLDM): Add Thumb-2 code.
2597 (do_it, shift1, do_push, do_pop, COND, THUMB_SYNTAX): New macros.
2598 * config/arm/arm.h (TARGET_CPU_CPP_BUILTINS): Define __thumb2__.
2599 (TARGET_THUMB1, TARGET_32BIT, TARGET_THUMB2, TARGET_DSP_MULTIPLY,
2600 TARGET_INT_SIMD, TARGET_UNIFIED_ASM, ARM_FT_STACKALIGN, IS_STACKALIGN,
2601 THUMB2_TRAMPOLINE_TEMPLATE, TRAMPOLINE_ADJUST_ADDRESS,
2602 ASM_OUTPUT_OPCODE, THUMB2_GO_IF_LEGITIMATE_ADDRESS,
2603 THUMB2_LEGITIMIZE_ADDRESS, CASE_VECTOR_PC_RELATIVE,
2604 CASE_VECTOR_SHORTEN_MODE, ADDR_VEC_ALIGN, ASM_OUTPUT_CASE_END,
2605 ADJUST_INSN_LENGTH): Define.
2606 (TARGET_REALLY_IWMMXT, TARGET_IWMMXT_ABI, CONDITIONAL_REGISTER_USAGE,
2607 STATIC_CHAIN_REGNUM, HARD_REGNO_NREGS, INDEX_REG_CLASS,
2608 BASE_REG_CLASS, MODE_BASE_REG_CLASS, SMALL_REGISTER_CLASSES,
2609 PREFERRED_RELOAD_CLASS, SECONDARY_OUTPUT_RELOAD_CLASS,
2610 SECONDARY_INPUT_RELOAD_CLASS, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P,
2611 TRAMPOLINE_SIZE, INITIALIZE_TRAMPOLINE, HAVE_PRE_INCREMENT,
2612 HAVE_POST_DECREMENT, HAVE_PRE_DECREMENT, HAVE_PRE_MODIFY_DISP,
2613 HAVE_POST_MODIFY_DISP, HAVE_PRE_MODIFY_REG, HAVE_POST_MODIFY_REG,
2614 REGNO_MODE_OK_FOR_BASE_P, LEGITIMATE_CONSTANT_P,
2615 REG_MODE_OK_FOR_BASE_P, REG_OK_FOR_INDEX_P, GO_IF_LEGITIMATE_ADDRESS,
2616 LEGITIMIZE_ADDRESS, THUMB2_LEGITIMIZE_ADDRESS,
2617 GO_IF_MODE_DEPENDENT_ADDRESS, MEMORY_MOVE_COST, BRANCH_COST,
2618 ASM_APP_OFF, ASM_OUTPUT_CASE_LABEL, ARM_DECLARE_FUNCTION_NAME,
2619 FINAL_PRESCAN_INSN, PRINT_OPERAND_PUNCT_VALID_P,
2620 PRINT_OPERAND_ADDRESS): Adjust for Thumb-2.
2621 (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv): New declarations.
2622 * config/arm/arm-cores.def: Add arm1156t2-s, cortex-a8, cortex-r4 and
2623 cortex-m3.
2624 * config/arm/arm-tune.md: Regenerate.
2625 * config/arm/arm-protos.h: Update prototypes.
2626 * config/arm/vfp.md: Enable patterns for Thumb-2.
2627 (arm_movsi_vfp): Add movw alternative. Use output_move_vfp.
2628 (arm_movdi_vfp, movsf_vfp, movdf_vfp): Use output_move_vfp.
2629 (thumb2_movsi_vfp, thumb2_movdi_vfp, thumb2_movsf_vfp,
2630 thumb2_movdf_vfp, thumb2_movsfcc_vfp, thumb2_movdfcc_vfp): New.
2631 * config/arm/libunwind.S: Add Thumb-2 code.
2632 * config/arm/constraints.md: Update include Thumb-2.
2633 * config/arm/ieee754-sf.S: Add Thumb-2/Unified asm support.
2634 * config/arm/ieee754-df.S: Ditto.
2635 * config/arm/bpabi.S: Ditto.
2636 * config/arm/t-arm (MD_INCLUDES): Add thumb2.md.
2637 * config/arm/predicates.md (low_register_operand,
2638 low_reg_or_int_operand, thumb_16bit_operator): New.
2639 (thumb_cmp_operand, thumb_cmpneg_operand): Rename...
2640 (thumb1_cmp_operand, thumb1_cmpneg_operand): ... to this.
2641 * config/arm/t-arm-elf: Add armv7 multilib.
2642 * config/arm/arm.md: Update patterns for Thumb-2 and Unified asm.
2643 Include thumb2.md.
2644 (UNSPEC_STACK_ALIGN, ce_count): New.
2645 (arm_incscc, arm_decscc, arm_umaxsi3, arm_uminsi3,
2646 arm_zero_extendsidi2, arm_zero_extendqidi2): New
2647 insns/expanders.
2648 * config/arm/fpa.md: Update patterns for Thumb-2 and Unified asm.
2649 (thumb2_movsf_fpa, thumb2_movdf_fpa, thumb2_movxf_fpa,
2650 thumb2_movsfcc_fpa, thumb2_movdfcc_fpa): New insns.
2651 * config/arm/cirrus.md: Update patterns for Thumb-2 and Unified asm.
2652 (cirrus_thumb2_movdi, cirrus_thumb2_movsi_insn,
2653 thumb2_cirrus_movsf_hard_insn, thumb2_cirrus_movdf_hard_insn): New
2654 insns.
2655 * doc/extend.texi: Document ARMv7-M interrupt functions.
2656 * doc/invoke.texi: Document Thumb-2 new cores+architectures.
2657
2658 2007-01-03 Jakub Jelinek <jakub@redhat.com>
2659
2660 * unwind-dw2.c (SIGNAL_FRAME_BIT, EXTENDED_CONTEXT_BIT): Define.
2661 (struct _Unwind_Context): Rename args_size to flags, remove
2662 signal_frame field, add a new args_size field and version field.
2663 (_Unwind_IsSignalFrame, _Unwind_SetSignalFrame,
2664 _Unwind_IsExtendedContext): New inline functions.
2665 (_Unwind_GetGR, _Unwind_SetGR, _Unwind_GetGRPtr, _Unwind_SetGRPtr):
2666 Assume by_value array is only present if _Unwind_IsExtendedContext.
2667 (_Unwind_GetIPInfo, execute_cfa_program, uw_frame_state_for): Use
2668 _Unwind_IsSignalFrame.
2669 (__frame_state_for): Initialize context.flags to EXTENDED_CONTEXT_BIT.
2670 (uw_update_context_1): Use _Unwind_SetSignalFrame.
2671 (uw_init_context_1): Initialize context->flags to
2672 EXTENDED_CONTEXT_BIT.
2673 * config/rs6000/linux-unwind.h (frob_update_context): Use
2674 _Unwind_SetSignalFrame.
2675
2676 2007-01-03 Andrew Pinski <andrew_pinski@playstation.sony.com>
2677
2678 PR middle-end/30353
2679 * gimplify.c (gimplify_modify_expr_complex_part): Move below
2680 tree_to_gimple_tuple. Call tree_to_gimple_tuple when we need
2681 the value.
2682
2683 2007-01-03 Kazu Hirata <kazu@codesourcery.com>
2684
2685 * config/i386/i386.h (NON_STACK_REG_P, REGNO_OK_FOR_SIREG_P,
2686 REGNO_OK_FOR_DIREG_P, REWRITE_ADDRESS, ASM_OPERAND_LETTER,
2687 RET, AT_SP): Remove.
2688 * config/i386/i386.md (*sse_prologue_save_insn): Use return
2689 instead of RET.
2690
2691 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
2692 * cfgbuild.c (find_basic_blocks): Likewise.
2693 * cfgrtl.c (rtl_create_basic_block): Likewise.
2694 * function.c (temp_slots_at_level): Likewise.
2695 * reg-stack.c (stack_regs_mentioned): Likewise.
2696 * regclass.c (allocate_reg_info): Likewise.
2697 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
2698 set_bb_for_stmt, move_block_to_fn): Likewise.
2699 * tree-complex.c (tree_lower_complex): Likewise.
2700 * vec.h (VEC_safe_grow_cleared): New.
2701
2702 * cgraphunit.c, tree-ssa-alias.c: Fix comment typos.
2703
2704 2007-01-03 Zdenek Dvorak <dvorakz@suse.cz>
2705
2706 * loop-unswitch.c (unswitch_loop): Pass probabilities to loopify.
2707 * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Pass probabilities
2708 to loop_version.
2709 * cfgloopmanip.c (scale_loop_frequencies): Export.
2710 (loopify): Scale the frequencies by prescribed coefficients.
2711 (set_zero_probability): New function.
2712 (duplicate_loop_to_header_edge): Improve updating of frequencies.
2713 (lv_adjust_loop_entry_edge, loop_version): Set probabilities
2714 and frequencies according to arguments.
2715 * tree-ssa-loop-manip.c (tree_unroll_loop): Set probabilities
2716 correctly.
2717 * cfg.c (scale_bbs_frequencies_int): Allow scaling the frequencies up.
2718 * modulo-sched.c (sms_schedule): Set probabilities for entering
2719 versioned loop correctly.
2720 * tree-vect-transform.c (vect_transform_loop): Ditto.
2721 * cfgloop.h (loopify, loop_version): Declaration changed.
2722 (scale_loop_frequencies): Declared.
2723
2724 2007-01-02 Jan Hubicka <jh@suse.cz>
2725
2726 * cgraph.c: Include tree-flow.h
2727 (cgraph_add_new-function): Handle IPA_SSA mode; execute
2728 early_local_passes.
2729 * cgraph.h (enum cgraph_state): Add CGRAPH_STATE_IPA_SSA.
2730 * tree-pass.h (pass_all_early_optimizations): Declare.
2731 * cgraphunit.c (cgraph_process_new_functions): Add IPA_SSA; execute
2732 early_local_passes.
2733 (cgraph_analyze_function): Do early_local_passes.
2734 * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
2735 Do not add referenced vars.
2736 * tree-optimize.c (gate_all_optimizations): Do not execute when not in
2737 SSA form.
2738 (gate_all_early_local_passes): New gate.
2739 (pass_early_local_passes): Use new gate.
2740 (execute_early_local_optimizations): New functions.
2741 (gate_all_early_optimizations): New gate.
2742 (pass_all_early_optimizations): New pass.
2743 (execute_free_datastructures): Free SSA only when initialized.
2744 (gate_init_datastructures): Init only when optimizing.
2745 (tree_lowering_passes): Do early local passes when called late.
2746 * tree-profile.c (do_tree_profiling): Don't profile functions added
2747 late.
2748 (do_early_tree_profiling, pass_early_tree_profile): Kill.
2749 * tree-cfg.c (update_modified_stmts): Do not update when operands are
2750 not active.
2751 * passes.c (init_optimizations_passes): Reorder so we go into SSA
2752 during early_local_passes.
2753 * Makefile.in (cgraph.o): Add dependency on tree-flow.h.
2754
2755
2756 2007-01-02 Carlos O'Donell <carlos@codesourcery.com>
2757
2758 * Makefile.in: Update copyright year.
2759
2760 2007-01-02 Carlos O'Donell <carlos@codesourcery.com>
2761
2762 * Makefile.in: Export GCC_EXEC_PREFIX before calling $(RUNTEST)
2763 in $(lang_checks) and check-consistency targets.
2764
2765 2007-01-02 Jan Hubicka <jh@suse.cz>
2766
2767 * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
2768 Do not add referenced vars.
2769 * tree-cfg.c (update_modified_stmts): Do not update when SSA operands
2770 are not active.
2771 * passes.c (init_optimization_passes): Put mudflap_2 after
2772 free_datastructures.
2773
2774 2007-01-02 Jan Hubicka <jh@suse.cz>
2775
2776 * tree-optimize (execute_fixup_cfg): Set after_inlining flag.
2777 Set NOTHROW flag on call statements proved to be nothrow.
2778 Update statement of local calls so new pure/const functions are
2779 updated. Update_ssa when in ssa form. Mark PHI nodes of nonlocal
2780 goto receivers.
2781 (tree_rest_of_compilation): Register hooks and initialize bitmap
2782 early. Do not set after_inlining flag.
2783
2784 2007-01-02 Steve Ellcey <sje@cup.hp.com>
2785
2786 * sbitmap.c (HOST_BITS_PER_LONG_LONG): Change to
2787 HOST_BITS_PER_LONGLONG
2788
2789 2007-01-02 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
2790
2791 PR c/19977
2792 * c-typeck.c (store_init_value): Don't emit pedantic overflow
2793 warning for non-static initializers.
2794
2795 2007-01-02 Steven Bosscher <steven@gcc.gnu.org>
2796
2797 * config/alpha/alpha.md, arm/arm.c, darwin.c, frv/frv.md,
2798 m32r/m32r.c, m32r/m32r.c, mn10300/mn10300.md, pa/pa.c,
2799 rs6000/rs6000.c, s390/s390.md, sh/sh.md, sparc/sparc.c:
2800 Always use set_unique_reg_note to add REG_EQUAL notes.
2801
2802 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
2803
2804 Revert:
2805 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
2806
2807 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
2808 * cfgbuild.c (find_basic_blocks): Likewise.
2809 * cfgrtl.c (rtl_create_basic_block): Likewise.
2810 * function.c (temp_slots_at_level): Likewise.
2811 * reg-stack.c (stack_regs_mentioned): Likewise.
2812 * regclass.c (allocate_reg_info): Likewise.
2813 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
2814 set_bb_for_stmt, move_block_to_fn): Likewise.
2815 * tree-complex.c (tree_lower_complex): Likewise.
2816 * vec.h (VEC_safe_grow_cleared): New.
2817
2818 2007-01-02 Ian Lance Taylor <iant@google.com>
2819
2820 * c-common.c (c_common_truthvalue_conversion): When warning about
2821 using an assignment as a truth value, set TREE_NO_WARNING.
2822
2823 2007-01-02 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
2824
2825 PR middle-end/7651
2826 * c.opt (Wold-style-declaration): New.
2827 * doc/invoke.texi (C-only Warning Options): New.
2828 (Wold-style-declaration): Document it.
2829 (Wextra): Enabled by -Wextra.
2830 * c-opts.c (c_common_post_options): Enabled by -Wextra.
2831 * c-decl.c (declspecs_add_scspec): Replace -Wextra with
2832 -Wold-style-declaration.
2833
2834 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
2835
2836 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
2837 * cfgbuild.c (find_basic_blocks): Likewise.
2838 * cfgrtl.c (rtl_create_basic_block): Likewise.
2839 * function.c (temp_slots_at_level): Likewise.
2840 * reg-stack.c (stack_regs_mentioned): Likewise.
2841 * regclass.c (allocate_reg_info): Likewise.
2842 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
2843 set_bb_for_stmt, move_block_to_fn): Likewise.
2844 * tree-complex.c (tree_lower_complex): Likewise.
2845 * vec.h (VEC_safe_grow_cleared): New.
2846
2847 2007-01-02 Douglas Gregor <doug.gregor@gmail.com>
2848
2849 * c-common.c(c_common_nodes_and_builtins): Since variants of
2850 void_type_node get built before it is given a name, we need to
2851 give those variants the name, too.
2852 (complete_array_type): We need to work with the canonical main
2853 type of the array, from which we will build the qualified version.
2854 * params.def (PARAM_VERIFY_CANONICAL_TYPES): New.
2855 * print-tree.c (print_node): Display canonical type information
2856 for each type.
2857 * stor-layout.c (layout_type): When we don't know the
2858 alignment of a type for which we're building an array, we end up
2859 guessing wrong, so make the type require structural equality.
2860 * tree.c (make_node_stat): When we build a new type, it is its
2861 own canonical type.
2862 (build_type_attribute_qual_variant): When building an attribute
2863 variant, its canonical type is the non-attribute variant. However,
2864 if the attributes are target-dependent and they differ, we need to
2865 use structural equality checks for this type.
2866 (build_qualified_type): A qualified type is not equivalent to its
2867 unqualified variant; set the canonical type appropriately.
2868 (build_distinct_type_copy): When building a distinct type from
2869 another type, the new type is its own canonical type.
2870 (build_variant_type_copy): When building a new type variant, we
2871 assume that it is equivalent to the original type.
2872 (build_pointer_type_for_mode): When building a pointer type, also
2873 build a canonical type pointer.
2874 (build_reference_type_for_mode): When building a reference type,
2875 also build a canonical type reference.
2876 (build_index_type): When we can't hash an index type (e.g.,
2877 because its maximum value is negative), the index type requires
2878 structural equality tests.
2879 (build_array_type): Build the canonical form of an array type.
2880 (build_function_type): Function types require structural equality,
2881 because they contain default arguments, attributes, etc.
2882 (build_method_type_directly): Ditto for method types.
2883 (build_offset_type): Build the canonical offset type.
2884 (build_complex_type): Build the canonical vector type.
2885 (make_vector_type): Build the canonical vector type.
2886 * tree.h (TYPE_CANONICAL): New.
2887 (TYPE_STRUCTURAL_EQUALITY_P): New.
2888 (SET_TYPE_STRUCTURAL_EQUALITY): New.
2889 (struct tree_type): Added "canonical" field.
2890 * params.h (VERIFY_CANONICAL_TYPES): New.
2891 * doc/c-tree.texi (TYPE_CANONICAL): Document.
2892 (TYPE_STRUCTURAL_EQUALITY_P): Document.
2893 (SET_TYPE_STRUCTURAL_EQUALITY): Document.
2894 * doc/invoke.texi (verify-canonical-types): Document --param
2895 parameter for verifying canonical types.
2896
2897 2007-01-02 Joseph Myers <joseph@codesourcery.com>
2898
2899 * config.gcc (powerpc-*-eabispe*, powerpc-*-eabisimaltivec*,
2900 powerpc-*-eabisim*, powerpc-*-eabialtivec*, powerpc-*-eabi*,
2901 powerpc-*-rtems*, powerpc-wrs-vxworks, powerpc-wrs-vxworksae,
2902 powerpcle-*-eabisim*, powerpcle-*-eabi*): Add rs6000/e500.h to
2903 tm_file.
2904 * config/rs6000/e500.h: New.
2905 * config/rs6000/eabi.h (TARGET_SPE_ABI, TARGET_SPE, TARGET_E500,
2906 TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE, TARGET_E500_DOUBLE):
2907 Remove.
2908 * config/rs6000/linuxspe.h (TARGET_SPE_ABI, TARGET_SPE,
2909 TARGET_E500, TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE,
2910 TARGET_E500_DOUBLE): Remove.
2911 * config/rs6000/vxworks.h (TARGET_SPE_ABI, TARGET_SPE,
2912 TARGET_E500, TARGET_ISEL, TARGET_FPRS): Remove.
2913 * config/rs6000/rs6000.h (CHECK_E500_OPTIONS): Define.
2914 * config/rs6000/rs6000.c (rs6000_override_options): Use
2915 CHECK_E500_OPTIONS.
2916
2917 2007-01-02 Joseph Myers <joseph@codesourcery.com>
2918
2919 * config/rs6000/rs6000.c (print_operand): Check (TARGET_SPE ||
2920 TARGET_E500_DOUBLE), not TARGET_E500, for %y.
2921 (rs6000_generate_compare, rs6000_emit_sCOND, output_cbranch,
2922 rs6000_emit_cmove): Don't check TARGET_E500.
2923 * config/rs6000/rs6000.md (bunordered, bordered, sunordered,
2924 sordered): Don't check TARGET_E500.
2925
2926 2007-01-01 Eric Christopher <echristo@apple.com>
2927
2928 * config/mips/mips.c (mips_regno_mode_ok_for_base_p): Use
2929 HARD_REGISTER_NUM_P.
2930
2931 2007-01-01 Roger Sayle <roger@eyesopen.com>
2932
2933 * fold-const.c (fold_binary) <EQ_EXPR>: Fold "(X^C1) eq/ne C2" into
2934 "X eq/ne (C1^C2)". Fold "(X^Z) eq/ne (Y^Z)" as "X eq/ne Y" when Z
2935 has no side-effects. Fold "(X^C1) eq/ne (Y^C2)" as "(X^(C1^C2))
2936 eq/ne Y".
2937
2938 2007-01-01 Mike Stump <mrs@apple.com>
2939
2940 * configure.ac: Remove support for building with Apple's gcc-3.1.
2941
2942 2007-01-02 Joseph Myers <joseph@codesourcery.com>
2943
2944 PR middle-end/30311
2945 * caller-save.c (add_stored_regs): Only handle SUBREGs if inner
2946 REG is a hard register. Do not modify REG before calling
2947 subreg_nregs.
2948 * rtlanal.c (subreg_get_info): Don't assert size of XMODE is a
2949 multiple of the size of YMODE for certain lowpart cases.
2950
2951 2007-01-01 Andrew Pinski <pinskia@gmail.com>
2952
2953 PR middle-end/30253
2954 * gimplify (voidify_wrapper_expr): Update for
2955 GIMPLE_MODIFY_STMT.
2956
2957 2007-01-01 Andreas Schwab <schwab@suse.de>
2958
2959 PR target/29166
2960 * config/ia64/ia64.c (ia64_compute_frame_size): Account space for
2961 save of BR0 in extra_spill_size instead of spill_size.
2962 (ia64_expand_prologue): Save BR0 outside of the gr/br/fr spill
2963 area.
2964 (ia64_expand_epilogue): Restore BR0 from its new location.
2965
2966 2007-01-01 Andrew Pinski <pinskia@gmail.com>
2967
2968 * gimplify.c (gimplify_init_constructor <case VECTOR_TYPE>):
2969 Use a temporary variable if the left hand side is not a gimple
2970 register.
2971
2972 2007-01-01 Andrew Pinski <pinskia@gmail.com>
2973
2974 * gimplify.c (gimplify_return_expr): Make the temporary variable
2975 for the return expression, a gimple register variable.
2976
2977 2007-01-01 Jan Hubicka <jh@suse.cz>
2978
2979 * emit-rtl.c (emit_copy_of_insn_after): Do not call copy_insn_1 for
2980 INSN_LIST.
2981
2982 2007-01-01 Mike Stump <mrs@apple.com>
2983
2984 * configure.ac (HAVE_GAS_LITERAL16): Add autoconf check for
2985 .literal16.
2986 * config/darwin.c (machopic_select_rtx_section): Use
2987 HAVE_GAS_LITERAL16.
2988 (darwin_mergeable_constant_section): Likewise.
2989 * configure: Regenerate.
2990 * config.in: Regenerate.
2991
2992 2007-01-01 Jan Hubicka <jh@suse.cz>
2993 Andrew Pinski <pinskia@gmail.com>
2994
2995 * cgraphunit.c (cgraph_optimize): Call cgraph_add_new_functions
2996 before starting IPA passes.