| blob | 55a40469404e3b1a796aa0501d5321e4f915b437 |
1 2007-01-27 Jakub Jelinek <jakub@redhat.com>
3 * genattrtab.c (struct attr_value_list, insn_code_values): Move to
4 file scope from optimize_attrs.
5 (simplify_test_exp): If insn_code_values is not NULL, use it to speed
6 up search.
7 (optimize_attrs): Clear insn_code_values after freeing it.
9 2007-01-26 Zdenek Dvorak <dvorakz@suse.cz>
11 * tree-ssa-address.c (create_mem_ref): Remove ", bsi" from
12 a parts.base assignment.
14 2007-01-26 Zdenek Dvorak <dvorakz@suse.cz>
16 * tree-data-ref.c (dump_subscript): Use dump_conflict_function.
17 (compute_subscript_distance, initialize_data_dependence_relation,
18 finalize_ddr_dependent, analyze_ziv_subscript,
19 analyze_siv_subscript_cst_affine,
20 compute_overlap_steps_for_affine_univar,
21 compute_overlap_steps_for_affine_1_2, analyze_subscript_affine_affine,
22 analyze_siv_subscript, analyze_miv_subscript,
23 analyze_overlapping_iterations, subscript_dependence_tester_1,
24 compute_self_dependence, free_dependence_relation): Work
25 with affine_fn instead of chrecs.
26 (dump_affine_function, dump_conflict_function, affine_function_equal_p,
27 common_affine_function, affine_function_base,
28 affine_function_constant_p, affine_fn_op, affine_fn_plus,
29 affine_fn_minus, affine_fn_free, conflict_fn_not_known,
30 conflict_fn_no_dependence, free_conflict_function, free_subscripts,
31 conflict_fn, affine_fn_cst, affine_fn_univar): New functions.
32 (all_chrecs_equal_p): Removed.
33 * tree-data-ref.h (affine_fn, conflict_function): New types.
34 (struct subscript): Change type of conflicting_iterations_in_a
35 and conflicting_iterations_in_b.
37 2007-01-26 Steve Ellcey <sje@cup.hp.com>
39 PR other/30182
40 * config/pa/pa.h (TARGET_HPUX_11): New.
41 * config/pa/pa-hpux11.h (TARGET_HPUX_11): New.
42 * config/pa/pa.c (pa_init_builtins): Use TARGET_HPUX_11.
44 2007-01-26 Daniel Berlin <dberlin@dberlin.org>
45 Richard Guenther <rguenther@suse.de>
47 * tree-ssa-structalias.c (solve_graph): Handle case
48 we merged the variable to another.
50 2007-01-25 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
52 * builtins.c (fold_builtin_1): Treat ccos and ccosh as 'even'
53 functions.
55 * fold-const.c (negate_mathfn_p): Treat casin, casinh, catan,
56 catanh, cproj, csin, csinh, ctan and ctanh as 'odd' functions.
58 2007-01-25 DJ Delorie <dj@redhat.com>
60 * config/m32c/m32c.c (m32c_cannot_change_mode_class): We don't
61 allow changes to modes which don't fit in those registers.
63 * reload1.c (choose_reload_regs): Check for invalid subregs before
64 computing their locations, not after.
66 2007-01-25 Geoffrey Keating <geoffk@apple.com>
68 PR 25127
69 * config/rs6000/rs6000.c (first_altivec_reg_to_save): On Darwin,
70 save Altivec registers in an eh_return function.
71 (compute_vrsave_mask): Likewise.
72 (rs6000_stack_info): Correct AIX/Darwin stack alignment computation
73 for saving Altivec registers.
74 (rs6000_emit_prologue): Don't allocate stack twice in
75 eh_return function. Correct expected value of altivec_save_offset
76 when using save_world. Describe save of R0 to stack when using
77 save_world. Describe stack pointer adjustment when using
78 save_world. Remove duplicated eh_return parameter register saving.
79 Update sp_offset variable after save_world.
80 * config/rs6000/t-darwin (LIB2FUNCS_STATIC_EXTRA): Remove
81 darwin-world.asm.
82 (LIB2FUNCS_EXTRA): Add darwin-world.asm.
83 * config/rs6000/darwin.h (SUBTARGET_OVERRIDE_OPTIONS): -m64
84 implies Altivec.
86 2007-01-25 Steve Ellcey <sje@cup.hp.com>
88 * config.gcc (ia64*-*-hpux*): Make posix threads the default.
90 2007-01-25 Steve Ellcey <sje@cup.hp.com>
92 PR other/30182
93 * config/pa/pa.c (pa_init_builtins): Set asm names for finite routines.
94 * config/ia64/ia64.c (ia64_init_builtins): Ditto.
96 2007-01-25 Richard Guenther <rguenther@suse.de>
98 * doc/invoke.texi (-Wcoverage-mismatch): Document.
99 * common.opt (-Wcoverage-mismatch): New warning option.
100 * coverage.c (get_coverage_counts): Ignore coverage mismatch
101 if -Wcoverage-mismatch is given.
103 2007-01-25 Razya Ladelsky <razya@il.ibm.com>
105 * ipa-cp.c (ipcp_insert_stage, ipcp_driver): Support for SSA.
106 (ipcp_driver): Change to static definition.
107 Add dumping of the ifunctions.
108 (constant_val_insert): Remove unused parameter. Support for SSA.
109 (ipcp_propagate_const): Support for SSA.
110 (ipcp_profile_bb_print): Print only analyzed nodes.
111 (ipcp_replace_map_create): Remove support for Fortran constant
112 for now.
113 * ipa-prop.c (ipa_method_modify_stmt,
114 ipa_callsite_compute_param): Support for SSA.
115 * ipa-prop.h (ipcp_driver): Remove declaration.
116 (IS_VALID_TREE_MAP_INDEX): Add define.
118 2007-01-24 Geoffrey Keating <geoffk@apple.com>
120 * unwind-dw2.c (execute_stack_op): Handle DW_OP_swap.
122 2007-01-24 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
124 PR middle-end/30447
125 * builtins.c (fold_builtin_cabs): Use MPFR to evaluate a
126 constant argument to cabs and do it without checking for
127 -funsafe-math-optimizations.
129 2007-01-24 Douglas Gregor <dgregor@osl.iu.edu>
131 * c-common.h (RID_FIRST_CXX0X): New.
132 (RID_LAST_CXX0X): New.
133 * c-opts.c (c_common_handle_option): -Wc++0x-compat is triggered
134 by -Wall.
135 * c.opt (Wc++0x-compat): New.
136 * doc/invoke.texi (-Wc++0x-compat): Document.
138 2007-01-24 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
140 * builtins.c (fold_builtin_carg): New.
141 (fold_builtin_1): Use it.
143 2007-01-24 Jan Hubicka <jh@suse.cz>
145 * ipa-inline.c (cgraph_decide_inlining): Initialize initial_insns.
147 * ipa-inline.c (initial_insns, max_insns): Delete.
148 (compute_max_insns): New function.
149 (cgraph_decide_inlining_of_small_function): Use it; take minimal amount
150 of insns as base for code growth.
151 (cgraph_decide_inlining): Make initial_insns local; do not compute
152 max_insns.
153 * params.def (PARAM_INLINE_UNIT_GROWTH): Set to 60.
154 * doc/invoke.texi (inline-unit-growth): Update docs.
156 2007-01-24 Jakub Jelinek <jakub@redhat.com>
158 * config/i386/i386.h (x86_cmpxchg16b): Remove const.
159 (TARGET_CMPXCHG16B): Define to x86_cmpxchg16b.
160 * config/i386/i386.c (x86_cmpxchg16b): Remove const.
161 (override_options): Add PTA_CX16 flag. Set x86_cmpxchg16b
162 for CPUs that have PTA_CX16 set.
164 PR middle-end/27416
165 * gimplify.c (omp_check_private): New function.
166 (gimplify_scan_omp_clauses): Use it for
167 firstprivate/lastprivate/reduction.
169 PR middle-end/30494
170 * gimplify.c (omp_add_variable): Don't call omp_notice_variable
171 on TYPE_SIZE_UNIT for GOVD_LOCAL VLAs.
173 PR middle-end/30421
174 * omp-low.c (lower_omp_for_lastprivate): Add dlist argument.
175 If lower_lastprivate_clauses emits some statements, append them
176 to dlist rather than body_p and to body_p append an initializer.
177 (lower_omp_for): Adjust caller.
179 2007-01-24 Steve Ellcey <sje@cup.hp.com>
181 * target.h (globalize_decl_name): New.
182 * target-def.h (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
183 * output.h (default_globalize_decl_name): New.
184 * varasm.c (asm_output_bss): Use globalize_decl_name instead of
185 globalize_label.
186 (globalize_decl): Ditto.
187 (default_globalize_decl_name): New.
188 * config/ia64/ia64.c (ia64_globalize_decl_name): New.
189 (ia64_handle_version_id_attribute): New.
190 (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
191 (ia64_asm_output_external): Use globalize_decl_name instead
192 of globalize_label.
193 * doc/extend.texi (version_id): New pragma.
194 * doc/tm.texi (ARGET_ASM_GLOBALIZE_DECL_NAME): New target hook.
196 2007-01-24 Andreas Krebbel <krebbel1@de.ibm.com>
198 * unwind-dw2-fde.c (get_cie_encoding): Replaced _Unwind_Word with
199 _uleb128_t and _Unwind_SWord with _sleb128_t.
200 * unwind-dw2.c (extract_cie_info, execute_stack_op, execute_cfa_program,
201 uw_frame_state_for, uw_update_context_1): Likewise.
202 * unwind-c.c (parse_lsda_header, PERSONALITY_FUNCTION): Likewise.
203 * unwind-pe.h (read_uleb128, read_sleb128,
204 read_encoded_value_with_base): Likewise.
205 * unwind-generic.h: Define _sleb128_t and _uleb128_t types.
207 2007-01-24 Richard Guenther <rguenther@suse.de>
209 * builtins.c (expand_builtin_cexpi): Get the fndecl
210 for cexp in the correct way.
212 2007-01-24 Jan Hubicka <jh@suse.cz>
214 * tree-ssa-dce.c (eliminate_unnecesary_stmts): Remove dead LHS of calls.
216 2007-01-24 Andreas Krebbel <krebbel1@de.ibm.com>
218 * c-cppbuiltin.c (builtin_define_type_sizeof): New function.
219 (c_cpp_builtins): New builtin macros: __SIZEOF_INT__, __SIZEOF_LONG__,
220 __SIZEOF_LONG_LONG__, __SIZEOF_SHORT__, __SIZEOF_POINTER__,
221 __SIZEOF_FLOAT__, __SIZEOF_DOUBLE__, __SIZEOF_LONG_DOUBLE__,
222 __SIZEOF_SIZE_T__, __SIZEOF_WCHAR_T__, __SIZEOF_WINT_T__ and
223 __SIZEOF_PTRDIFF_T__.
224 * doc/cpp.texi: Documentation for the new builtin macros added.
226 2007-01-24 Uros Bizjak <ubizjak@gmail.com>
228 * config/i386/i386.md (tanxf2, tan<mode>2, atan<mode>2, log<mode>2,
229 log10<mode>2, log2<mode>2, expxf2, exp10xf2, exp2xf2): Use op2
230 instead of operands[2] to avoid access past the end of array.
232 2007-01-24 Richard Sandiford <richard@codesourcery.com>
234 * reload1.c (emit_reload_insns): Pass the reload register
235 for a non-spill output reload through forget_old_reloads_1.
237 2007-01-23 Joseph Myers <joseph@codesourcery.com>
239 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Define
240 _SOFT_DOUBLE if doubles use software floating-point.
241 * config/rs6000/libgcc-ppc-glibc.ver: Export additional long
242 double functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
243 * config/rs6000/darwin-ldouble.c: Also compile functions for
244 hard-float without FPRs. Use fmsub function for all __NO_FPRS__
245 cases. Compile extra functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
246 * config/rs6000/linuxspe.h (SUBSUBTARGET_OVERRIDE_OPTIONS): Remove
247 commented-out long double override.
248 (CPP_LONGDOUBLE_DEFAULT_SPEC): Likewise.
249 * config/rs6000/eabispe.h: Likewise.
250 * config/rs6000/rs6000.c (rs6000_override_options): Don't override
251 long double for non-SPE.
252 (rs6000_handle_option): Likewise.
253 (invalid_e500_subreg): Disallow more subregs involding DImode,
254 DFmode, TImode or TFmode.
255 (rs6000_legitimate_offset_address_p): Check TFmode offsets for
256 E500 double.
257 (legitimate_lo_sum_address_p): Also check for TFmode for E500
258 double.
259 (rs6000_legitimize_address): Also handle TFmode for E500 double.
260 (rs6000_legitimize_reload_address): Also handle TFmode for E500
261 double.
262 (rs6000_legitimate_address): Also check for TFmode for E500
263 double.
264 (rs6000_emit_move): Use DFmode subregs of TFmode for E500 double.
265 (spe_build_register_parallel): Handle TFmode and TCmode.
266 (rs6000_spe_function_arg): Handle TFmode and TCmode for E500
267 double.
268 (function_arg): Handle TFmode and TCmode for E500 double.
269 (rs6000_init_libfuncs): Initialize extra libfuncs for soft double
270 in general.
271 (print_operand): Handle TFmode and TImode for %y.
272 (rs6000_generate_compare): Handle TFmode comparisons for E500
273 double.
274 (spe_func_has_64bit_regs_p): Check for TFmode for E500 double.
275 (rs6000_function_value): Handle TFmode and TCmode for E500 double.
276 (rs6000_libcall_value): Handle TFmode and TCmode for E500 double.
277 * config/rs6000/rs6000.h (CANNOT_CHANGE_MODE_CLASS): Check for
278 TFmode for E500 double.
279 * config/rs6000/rs6000.md (FP): Allow TF for E500 double.
280 (floatsidf2): Enable for E500 double.
281 (movtf_softfloat): Use rs6000_nonimmediate_operand.
282 (extenddftf2): Change to extenddftf2_fprs.
283 (extenddftf2): Call gen_spe_extenddftf2 or gen_extenddftf2_fprs
284 depending on TARGET_E500_DOUBLE.
285 (extendsftf2): Enable for E500 double.
286 (trunctfdf2): Enable for E500 double.
287 (trunctfsf2): Change to trunctfsf2_fprs.
288 (trunctfsf2): Call gen_spe_trunctfsf2 or gen_trunctfsf2_fprs
289 depending on TARGET_E500_DOUBLE.
290 (floatsitf2): Enable for E500 double.
291 (fix_trunctfsi2): Change to fix_trunctfsi2_fprs.
292 (fix_trunctfsi2): Call gen_spe_fix_trunctfsi2 or
293 gen_fix_trunctfsi2_fprs depending on TARGET_E500_DOUBLE.
294 (negtf2): Change to negtf2_internal.
295 (negtf2): New expander.
296 (abstf2): Enable for E500 double. Call gen_spe_abstf2_tst,
297 gen_spe_abstf2_cmp or gen_abstf2_internal depending on
298 TARGET_E500_DOUBLE and flag_unsafe_math_optimizations.
299 (movdi_internal32): Use rs6000_nonimmediate_operand.
300 (unnamed splitter): Likewise.
301 * config/rs6000/spe.md (CMPTFEQ_GPR, TSTTFEQ_GPR, CMPTFGT_GPR,
302 TSTTFGT_GPR, CMPTFLT_GPR, TSTTFLT_GPR): New unspecs.
303 (SPE64TF, DITI): New mode macros.
304 (frob_df_di): Change to frob_<SPE64:mode>_<DITI:mode>; allow more
305 modes.
306 (frob_tf_ti): New.
307 (frob_<mode>_di_2): New.
308 (frob_tf_di_8_2): New.
309 (frob_di_df): Change to frob_di_<mode>; allow more modes.
310 (frob_ti_tf): New.
311 (frob_di_df_2): Change to frob_<DITI:mode>_<SPE64:mode>_2; allow
312 more modes.
313 (frob_ti_<mode>_8_2): New.
314 (frob_ti_tf_2): New.
315 (mov_si<mode>_e500_subreg0, mov_si<mode>_e500_subreg0_2,
316 mov_si<mode>_e500_subreg4, mov_si<mode>_e500_subreg4_2): Allow
317 TFmode.
318 (mov_sitf_e500_subreg8, mov_sitf_e500_subreg8_2,
319 mov_sitf_e500_subreg12, mov_sitf_e500_subreg12_2): New.
320 (spe_trunctfdf2_internal1, spe_trunctfsf2, spe_extenddftf2,
321 spe_fix_trunctfsi2, spe_fix_trunctfsi2_internal,
322 spe_negtf2_internal, spe_abstf2_cmp, spe_abstf2_tst): New.
323 (cmptfeq_gpr, tsttfeq_gpr, cmptfgt_gpr, tsttfgt_gpr, cmptflt_gpr,
324 tsttflt_gp): New.
326 2007-01-23 Ian Lance Taylor <iant@google.com>
328 * Makefile.in (OBJS-common): Reformat, alphabetize, but put
329 insn-*.o first.
330 (OBJS-archive): Reformat, alphabetize.
331 (OBJS): Change out_object_file to OBJS-md.
333 2007-01-23 Uros Bizjak <ubizjak@gmail.com>
335 * config/i386/i386.md: Use REG_P, MEM_P, CONST_INT_P, LABEL_P,
336 JUMP_P and CALL_P predicates where applicable.
337 * config/i386/i386.c: Ditto.
338 * config/i386/i386.md: Ditto.
339 * config/i386/mmx.md: Ditto.
340 * config/i386/predicates.md: Ditto.
342 2007-01-22 Andreas Schwab <schwab@suse.de>
344 * config/m68k/m68k.h: Fix comment.
346 2007-01-22 Jan Hubicka <jh@suse.cz>
348 * passes.c (init_optimization_passes): Do not rerun
349 pass_early_warn_uninitialized.
351 2007-01-22 Richard Guenther <rguenther@suse.de>
353 PR tree-optimization/30038
354 * tree-ssa-math-opts.c (maybe_record_sincos): New static helper
355 function.
356 (execute_cse_sincos_1): Likewise.
357 (execute_cse_sincos): Likewise.
358 (gate_cse_sincos): Likewise.
359 (pass_cse_sincos): New pass CSEing sin() and cos() calls using
360 the cexpi() canonicalization of sincos().
361 * tree-pass.h (pass_cse_sincos): Declare.
362 * passes.c (init_optimization_passes): New pass pas_cse_sincos.
364 2007-01-21 Eric Botcazou <ebotcazou@libertysurf.fr>
366 PR rtl-optimization/29329
367 * combine.c (replaced_rhs_insn): Rename to i2mod.
368 (replaced_rhs_value): Rename to i2mod_new_rhs.
369 (i2mod_old_rhs): New global variable.
370 (combine_instructions): Adjust for above change. Save a copy of
371 the old RHS into i2mod_old_rhs when the contents of a REG_EQUAL
372 note are substituted in the second instruction.
373 (distribute_notes) <REG_DEAD>: Adjust for above change. Do not
374 ditch the note if it pertains to the second eliminated register
375 and this register is mentioned in i2mod_old_rhs.
377 Revert:
378 2006-09-12 Eric Botcazou <ebotcazou@libertysurf.fr>
380 * combine.c (distribute_notes) <REG_DEAD>: Do not consider SETs past
381 the insn to which the note was originally attached.
383 2007-01-21 Jan Hubicka <jh@suse.cz>
385 * ipa-inline.c (inlining_mode): Comment, move up.
386 (cgraph_decide_inlining_incrementally): Do not perform inlining
387 itself; fix handling of flattening of self recursive functions.
388 (cgraph_find_cycles): Remove.
389 (cgraph_flatten_node): Remove.
390 (cgraph_decide_inlining): Use incremental inliner to handle flattening.
391 (try_inline): New function.
392 (cgraph_early_inlining): Update call of
393 cgraph_decide_inlining_incrementally. Apply inlining here.
394 (apply_inline): Update call of cgraph_decide_inlining_incrementally.
396 2007-01-21 Dirk Mueller <dmueller@suse.de>
398 PR bootstrap/30511
399 * tree-vrp.c (check_array_bounds): do not warn
400 about ADDR_EXPR's of ARRAY_REF's which are immediately
401 used in binary expressions.
403 2007-01-21 Ira Rosen <irar@il.ibm.com>
405 * tree-vectorizer.h (struct _stmt_vec_info): Add new field
406 read_write_dep and macros for its access.
407 * tree-vectorizer.c (new_stmt_vec_info): Initialize the new field.
408 * tree-vect-analyze.c (vect_analyze_data_ref_dependence): Remove
409 argument, call vect_check_interleaving for every independent pair of
410 data-refs. Mark loads that access the same memory location as a store
411 in the loop.
412 (vect_check_dependences): Remove.
413 (vect_analyze_data_ref_dependences): Remove vect_check_dependences
414 call, fix the call to vect_analyze_data_ref_dependence.
415 (vect_analyze_data_ref_access): For statements that access the same
416 data-ref, check that they are not stores; for loads, check that there
417 is no store that access the same location.
419 2007-01-20 Roger Sayle <roger@eyesopen.com>
420 Joseph Myers <joseph@codesourcery.com>
422 * doc/invoke.texi (-fdump-rtl-combine): Fix under/overfull hbox.
423 (-fdump-rtl-stack): Likewise.
424 (-fno-signed-zeros): Use @minus{} for a minus sign. Correct typo.
425 (-mcheck-zero-division, -mcpu): Fix under/overfull hbox.
426 (-mpt-fixed): Use @minus{} for minus sign.
427 (Using Precompiled Headers): Fix under/overfull hbox.
429 2007-01-20 Jan Hubicka <jh@suse.cz>
431 * tree-flow.h (struct stmt_ann_d): Move references_memory to proper
432 place within annotation.
434 2007-01-20 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
436 * pa.c (output_move_double): Change array size of xoperands to 4.
438 2007-01-20 Jan Hubicka <jh@suse.cz>
440 * tree-tailcall.c (adjust_return_value): Do not use RESULT_DECL
441 as temporary.
443 2007-01-19 Ian Lance Taylor <iant@google.com>
445 * expmed.c (expand_divmod) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]: Cast
446 constant to unsigned HOST_WIDE_INT before negating.
448 2007-01-19 Ian Lance Taylor <iant@google.com>
450 * tree-ssa-operands.h (struct vuse_vec_d): Change num_vuse field
451 to unsigned.
452 (VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Use unsigned comparison.
453 (VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
454 (SET_VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Likewise.
455 (SET_VUSE_ELEMENT_VAR) [ENABLE_CHECKING]: Likewise.
456 (SET_VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
457 (realloc_vdef, realloc_vuse): Change second parameter to
458 unsigned.
459 (ssa_operand_iterator_d): Change vuse_index and mayuse_index
460 fields to unsigned.
461 * tree-ssa-operands.c (realloc_vop): Change num_elem parameter to
462 unsigned. Change x and lim locals to unsigned.
463 (realloc_vdef, realloc_vuse): Change num_elem parameter to
464 unsigned.
465 (finalize_ssa_vuse_ops): Change old_i local to unsigned.
466 (copy_virtual_operands): Change i and n locals to unsigned.
468 2007-01-19 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
470 PR middle-end/29335
471 * builtins.c (fold_builtin_1): Handle builtin fdim.
473 2007-01-20 Jan Hubicka <jh@suse.cz>
475 * tree-ssa.c (init_tree_ssa): Do not call init_alias_heapvars.
476 * tree-ssa-structalias.c (compute_points_to_sets): Do call
477 init_alias_heapvars.
478 (init_alias_heapvars): Initialize only when not already initialized.
479 (delete_alias_heapvars): Set heapvar_for_stmt to NULL.
481 2007-01-19 Roger Sayle <roger@eyesopen.com>
483 * common.opt (fsigned-zeros): New command line option.
484 * flags.h (HONOR_SIGNED_ZEROS): Control via flag_signed_zeros instead
485 of flag_unsafe_math_optimizations.
486 * opts.c (set_fast_math_flags): The -ffast-math command line option
487 implies -fno-signed-zeros.
488 (fast_math_flags_set_p): Likewise.
490 * doc/invoke.texi: Document new -fno-signed-zeros option, and update
491 the documentation of -ffast-math appropriately. Wrap long lines.
493 2007-01-19 Steve Ellcey <sje@cup.hp.com>
495 * system.h (ASM_MAKE_LABEL_LINKONCE): Poison.
496 * varasm.c (globalize_decl): Remove ASM_MAKE_LABEL_LINKONCE ifdef.
498 2007-01-19 Tomas Bily <tbily@suse.cz>
500 * cgraphunit.c (cgraph_finalize_function): Updating of pid
501 * tree-profile.c:
502 (tree_init_ic_make_global_vars): New function
503 (tree_init_edge_profiler): call of tree_init_ic_make_global_vars
504 (tree_gen_ic_profiler): New function
505 (tree_gen_ic_func_profiler): New function
506 (tree_profiling): Added calling of tree_gen_ic_func_profiler
507 (tree_profile_hooks): Added hook for indirec/virtual calls
508 * value-prof.c (tree_find_values_to_profile): New case for
509 indirect calls
510 (tree_values_to_profile): Call for determining indirect/virtual
511 counters
512 (tree_indirect_call_to_profile): New function
513 (tree_ic_transform): New function
514 (tree_ic): New function
515 (find_func_by_pid): New function
516 (init_pid_map): New function
517 (tree_value_profile_transformations): Added check for
518 indirect/virtual call transformation
519 * value-prof.h (enum hist_type): New counter type for
520 indirect/virtual calls
521 (profile_hooks): Added new hook for profiling indirect/virtual
522 calls
523 * profile.c (instrument_values): New case for indirect/virtual
524 call added
525 * gcov-io.h (GCOV_LAST_VALUE_COUNTER): Changed to 6
526 (GCOV_COUNTER_V_INDIR): New counter type
527 (GCOV_COUNTER_NAMES): New name of counter "indirect" added
528 (GCOV_MERGE_FUNCTIONS): New merge function for indirect/virtual
529 call added
530 * cgraph.c: Definition of cgraph_max_pid
531 (cgraph_create_node): Default init of pid attribute
532 * cgraph.h: Declaration of cgraph_max_pid
533 (struct cgraph_node): Added pid attribute
534 * libgcov.c (__gcov_indirect_call_profiler): New function
535 (__gcov_one_value_profiler_body): New function
536 (__gcov_one_value_profiler): Body was moved to
537 __gcov_one_value_profiler_body and calls it
539 2007-01-19 Basile Starynkevitch <basile@starynkevitch.net>
541 * doc/gty.texi (Options): Document the mark_hook option to GTY.
542 * gengtype.c (write_types_data, write_func_for_structure,
543 write_types, ggc_wtd, pch_wtd): Add skip_hooks to
544 write_types_data, ggc_wtd, pch_wtd for processing mark_hook.
545 (walk_type, write_func_for_structure): Generate the mark_hook if
546 needed.
548 2007-01-19 Jan Hubicka <jh@suse.cz>
550 * ipa-inline.c (cgraph_decide_inlining_incrementally): Instead of
551 'early' argument take inlining mode argument specifying whether to
552 inline for size/speeed or all functions; add support for flattening;
553 improve dumpting.
554 (cgraph_early_inlining): Update call of decide_inlining_incrementally.
556 2007-01-19 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
558 PR c++/17947
559 * toplev.c (warn_deprecated_use): Use %qD instead of %qs to print
560 the name of the declared identifier.
562 2007-01-19 Dirk Mueller <dmueller@suse.de>
564 * config/i386.h (CONDITIONAL_REGISTER_USAGE): Store
565 result of PIC_OFFSET_TABLE_REGNUM in temporary variable to avoid
566 duplicate evaluation.
568 2007-01-19 Uros Bizjak <ubizjak@gmail.com>
570 * config/i386/i386.md (acos<mode>2): Rename from acossf2 and acosdf2.
571 Macroize expander using X87MODEF12 mode macro. Extend operand 1
572 to XFMode, use acosxf2 and truncate result to requested mode.
573 Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
574 (asin<mode>2): Similarly, with asin expanders.
575 (*fscalexf4_i387): Rename from *fscalexf4.
576 (expNcorexf3): New expander.
577 (expxf2, exp10xf2, exp2xf2): Use expNcorexf3 expander.
578 (exp<mode>2): Rename from expsf2 and expdf2. Macroize expander using
579 X87MODEF12 mode macro. Extend operand 1 to XFMode, use expxf2 and
580 truncate result to requested mode. Use SSE_FLOAT_MODE_P to disable
581 patterns for SSE math.
582 (exp10<mode>2): Similarly, with exp10 expanders.
583 (exp2<mode>2): Similarly, with exp2 expanders.
584 (expm1<mode>2): Similarly, with expm1 expanders.
585 (ldexp<mode>3): Similarly, with ldexp expanders.
586 (log<mode>2, log10<mode>2, log2<mode>2, log1p<mode>2, logb<mode>2):
587 Use gen_truncxf<mode>2_i387_noop to truncate result.
589 2007-01-19 Richard Sandiford <richard@codesourcery.com>
591 * config/m68k/m68k.h (PREFERRED_STACK_BOUNDARY): Define to 32
592 for ColdFire targets.
594 2007-01-19 Nathan Sidwell <nathan@codesourcery.com>
595 Richard Sandiford <richard@codesourcery.com>
597 * config/m68k/m68k.h (M68K_STATIC_CHAIN_REG_NAME): New macro.
598 (INITIALIZE_TRAMPOLINE): Use STATIC_CHAIN_REGNUM.
599 (__transfer_from_trampoline): Use M68K_STATIC_CHAIN_REG_NAME.
600 * config/m68k/m68kelf.h (STATIC_CHAIN_REGNUM): Override.
601 (M68K_STATIC_CHAIN_REG_NAME): Likewise.
602 * config/m68k/netbsd-elf.h (M68K_STATIC_CHAIN_REG_NAME): Likewise.
604 2007-01-19 Richard Sandiford <richard@codesourcery.com>
606 * config/m68k/m68k.md (adddi_dilshr32): Rename to...
607 (*adddi_dilshr32): ...this. Fix formatting. Remove commented-out
608 non-canonical pattern. Restrict to !TARGET_COLDFIRE.
609 (*adddi_dilshr32_cf): New pattern.
610 (adddi3, subdi3): Remove first alternatives.
612 2007-01-19 Richard Sandiford <richard@codesourcery.com>
614 * config/m68k/m68k.c (notice_update_cc): If an SFmode move is
615 implemented using move.l, do not use its cc result for floating-point
616 comparisons.
618 2007-01-19 Richard Sandiford <richard@codesourcery.com>
620 * config/m68k/m68k.h (EXTRA_CONSTRAINT): Stop the 'T' constraint
621 from accepting 's' constraints if flag_pic.
623 2007-01-19 Richard Sandiford <richard@codesourcery.com>
625 * config/m68k/m68k.md (bordered, bunordered, buneq, bunge, bungt)
626 (bunle, bunlt, bltgt, bordered_rev, bunordered_rev, buneq_rev)
627 (bunge_rev, bungt_rev, bunle_rev, bunlt_rev, bltgt_rev): Change
628 condition from TARGET_68881 to TARGET_HARD_FLOAT.
630 2007-01-19 Sandra Loosemore <sandra@codesourcery.com>
632 * longlong.h (count_leading_zeros, COUNT_LEADING_ZEROS_0): Add
633 ColdFire alternatives.
634 * config/m68k/m68k.h (CLZ_DEFINED_VALUE_AT_ZERO): New macro.
635 * config/m68k/m68k.md (clzsi2): Define for ColdFire
636 architectures that support ff1 instruction.
638 2007-01-19 Richard Sandiford <richard@codesourcery.com>
639 Julian Brown <julian@codesourcery.com>
641 * config/m68k/m68k.h (CONST_OK_FOR_LETTER_P): Add an 'R' case.
642 * config/m68k/m68k.md (*movsi_cfv4): Fold into...
643 (*movsi_cf): ...here. Remove unnecessary 'R' from 'Rg'.
644 Add commentary.
646 2007-01-19 Richard Sandiford <richard@codesourcery.com>
648 * config/m68k/m68k-protos.h (valid_mov3q_const): Take a HOST_WIDE_INT
649 and return a bool.
650 (output_move_const_into_data_reg, output_move_simode_const): Delete.
651 * config/m68k/m68k.c (const_method, const_int_cost): Take a
652 HOST_WIDE_INT instead of an rtx.
653 (m68k_rtx_costs): Update call accordingly.
654 (output_move_const_into_data_reg): Likewise. Fix formatting.
655 (valid_mov3q_const): Take a HOST_WIDE_INT instead of an rtx.
656 Return a bool.
657 (output_move_simode_const): Update calls after above changes.
658 Rework to use automatic variables and predicates like MEM_P.
659 * config/m68k/m68k.md (pushexthisi_const): Update call to
660 valid_mov3q_const.
662 2007-01-19 Dirk Mueller <dmueller@suse.de>
664 * tree-ssa-alias.c (perform_var_substitution): Fix typo
665 in dump_flags test.
667 2007-01-19 Richard Guenther <rguenther@suse.de>
669 * builtins.c (expand_builtin_cexpi): Fall back to expanding
670 via cexp in case sincos is not available.
672 2007-01-19 Richard Guenther <rguenther@suse.de>
674 * doc/tm.texi (TARGET_HAS_SINCOS): Document new target macro.
675 * defaults.h (TARGET_HAS_SINCOS): Default to off.
676 * config/linux.h (TARGET_HAS_SINCOS): Set to on if we have glibc.
677 * config/alpha/linux.h (TARGET_HAS_SINCOS): Likewise.
678 * config/sparc/linux.h (TARGET_HAS_SINCOS): Likewise.
679 * config/sparc/linux64.h (TARGET_HAS_SINCOS): Likewise.
680 * config/rs6000/linux.h (TARGET_HAS_SINCOS): Likewise.
681 * config/rs6000/linux64.h (TARGET_HAS_SINCOS): Likewise.
683 2007-01-19 Uros Bizjak <ubizjak@gmail.com>
685 * config/i386/i386.md (*fpatanxf3_i387, fpatan_extend<mode>xf3_i387):
686 New insn patterns.
687 (atan2sf3_1, atan2df3_1, atan2xf3_1): Remove insn patterns.
688 (atan2xf3): Directly generate RTL pattern.
689 (atan2<mode>3): Rename from atan2sf3 and atan2df3 and macroize insn
690 patterns using X87MODEF12 mode macro. Use fpatan_extend<mode>xf3_i387
691 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
692 disable patterns for SSE math.
693 (atan<mode>2): Rename from atansf2 and atandf2 and macroize insn
694 patterns using X87MODEF12 mode macro. Use fpatan_extend<mode>xf3_i387
695 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
696 disable patterns for SSE math.
698 2007-01-19 Alexandre Oliva <aoliva@redhat.com>
700 * libgcc-std.ver: Fix typo in %inherit for GCC_4.3.0.
702 2007-01-18 Roger Sayle <roger@eyesopen.com>
704 * fold-const.c (fold_unary) <VIEW_CONVERT_EXPR>: Optimize away a
705 VIEW_CONVERT_EXPR to the same type as it's operand.
707 2007-01-18 David Edelsohn <edelsohn@gnu.org>
709 * config/rs6000/darwin-ldouble.c: Only build _SOFT_FLOAT if
710 configured for long double 128.
712 2007-01-18 Mike Stump <mrs@apple.com>
714 * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Fix build
715 error.
717 2007-01-18 Michael Meissner <michael.meissner@amd.com>
719 * i386.c (ix86_compute_frame_layout): Make fprintf's in #if 0 code
720 type correct.
722 2007-01-18 Jan Hubicka <jh@suse.cz>
724 * tree-ssa-operands.c (vop_free_bucket_size): Never return value
725 greater than NUM_VOP_FREE_BUCKETS.
727 2007-01-18 Daniel Berlin <dberlin@dberlin.org>
729 * tree-ssa-structalias.c: Update comments.
730 (ptabitmap_obstack): Removed.
731 (pta_obstack): New.
732 (oldpta_obstack): Ditto.
733 (stats): Add a few members.
734 (struct variable_info): Remove node, complex, address_taken, and
735 indirect_target members. Add oldsolution member.
736 (new_var_info): Do not initialize removed members.
737 (constraint_expr_type): Remove INCLUDES.
738 (constraint_graph): Add size, implicit_preds, rep,
739 indirect_cycles, eq_rep, label, direct_nodes, and complex members.
740 (FIRST_REF_NODE): New macro.
741 (LAST_REF_NODE): Ditto.
742 (FIRST_ADDR_NODE): Ditto.
743 (find): New function.
744 (unite): Ditto.
745 (dump_constraint): Do not handle INCLUDES.
746 (insert_into_complex): Do not insert duplicate constraints.
747 (condense_varmap_nodes): Renamed and rewritten into ...
748 (merge_node_constraints): This. Also fix bug in handling of
749 offseted copy constraints.
750 (clear_edges_for_node): No longer need to deal with preds at all,
751 or removing associated preds/succs.
752 (merge_graph_nodes): Deal with indirect_cycles.
753 Don't deal with predecessors.
754 (add_implicit_graph_edge): New function.
755 (add_pred_graph_edge): Ditto.
756 (add_graph_edge): Don't deal with predecessors.
757 (build_constraint_graph): Removed.
758 (build_pred_graph): New function.
759 (build_succ_graph): Ditto.
760 (struct scc_info): Removed in_component. Added roots, dfs, and
761 node_mapping. Remove visited_index, unification_queue.
762 (scc_visit): Deal with union-find we do now.
763 Deal with cycles with REF nodes.
764 (collapse_nodes): Renamed and rewritten to ...
765 (unify_nodes): This.
766 (process_unification_queue): Removed.
767 (topo_visit): Cleanup
768 (do_da_constraint): Use find.
769 (do_sd_constraint): Ditto.
770 (do_ds_constraint): Ditto.
771 (do_complex_constraint): Ditto.
772 (init_scc_info): Update for removed and added members.
773 (find_and_collapse_graph_cycles): Renamed and rewritten into ...
774 (find_indirect_cycles): This.
775 (equivalence_class): New variable.
776 (label_visit): New function.
777 (perform_variable_substitution): Rewritten.
778 (free_var_substitution_info): New function.
779 (find_equivalent_node): Ditto.
780 (move_complex_constraints): Ditto.
781 (eliminate_indirect_cycles): Ditto.
782 (solve_graph): Only propagate changed bits.
783 Use indirect cycle elimination.
784 Use find.
785 (tree_id_t): Rename to tree_vi_t, delete id member, add vi member.
786 (tree_id_eq): Renamed to ...
787 (tree_vi_eq): This. Update for member change
788 (insert_id_for_tree): Renamed and rewritten to ...
789 (insert_vi_for_tree): This.
790 (lookup_id_for_tree): Renamed and rewritten to ...
791 (lookup_vi_for_tree): This.
792 (get_id_for_tree): Renamed and rewritten to ...
793 (get_vi_for_tree): Ditto.
794 (get_constraint_exp_from_ssa_var): Update to use get_vi_for_tree.
795 (process_constraint): Don't handle INCLUDES.
796 Remove special ADDRESSOF case.
797 (find_func_aliases): Rewrite to use vi functions instead of id
798 ones.
799 (create_function_info_for): Ditto.
800 (create_variable_info_for): Ditto.
801 (intra_create_variable_infos): Ditto.
802 (merge_smts_into): Ditto.
803 (find_what_p_points_to): Ditto.
804 (init_base_vars): Ditto.
805 (init_alias_vars): Ditto.
806 (remove_preds_and_fake_succs): New function.
807 (dump_sa_points_to_info): Dump new stats.
808 (dump_solution_for_var): Use find.
809 (set_used_smts): Fix formatting.
810 (compute_points_to_sets): Updated for new functions.
811 (ipa_pta_execute): Ditto.
813 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
814 Richard Sandiford <richard@codesourcery.com>
816 * doc/tm.texi (TARGET_FUNCTION_VALUE): Expand documentation of
817 parallels.
818 * calls.c (expand_call): If the return value is a PARALLEL,
819 extract its first member.
820 * config/m68k/linux.h (FUNCTION_EXTRA_EPILOGUE): Remove.
821 * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
822 use FUNCTION_EXTRA_EPILOGUE.
823 (m68k_function_value): Return a PARALLEL if the return value
824 is of a pointer type.
825 * config/m68k/netbsd-elf.h (current_function_returns_pointer)
826 (FUNCTION_EXTRA_EPILOGUE): Remove.
827 * config/m68k/m68k.md (D0_REG): New constant.
829 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
831 * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
832 output a NOP for empty epilogues.
834 2007-01-18 Richard Sandiford <richard@codesourcery.com>
836 * config/m68k/m68k.c (m68k_use_return_insn): Update comments
837 before function. Extend register save check to include all
838 registers, not just integer ones.
840 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
842 * config/m68k/m68k-protos.h (use_return_insn): Rename to...
843 (m68k_use_return_insn): ...this.
844 * config/m68k/m68k.h (USE_RETURN_INSN): Delete.
845 * config/m68k/m68k.c (use_return_insn): Rename to...
846 (m68k_use_return_insn): ...this.
847 * config/m68k/m68k.md (return): Use m68k_use_return_insn instead
848 of USE_RETURN_INSN.
850 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
852 * config/m68k/fpgnulib.c (__truncdfsf2): Implement round to
853 nearest even, fix denormal rounding overflow.
855 2007-01-18 Richard Sandiford <richard@codesourcery.com>
857 * config/m68k/m68k.md (movsf_cf_hard): Use fsmove instead of
858 f%$smove and f%$move.
859 (movdf_cf_hard): Use fdmove for cases 0 and 3 and fmove for case 1.
860 (extendsfdf2_cf): Use fdmove instead of f%&move.
861 (truncdfsf2_cf): Use fsmove instead of f%$smove.
862 (add<mode>3_cf, sub<mode>3_cf): Use <FP:prec> instead of <FP:round>.
864 2007-01-18 Richard Sandiford <richard@codesourcery.com>
866 * config/m68k/m68k.md (movdf_cf_hard): Use output_move_double for
867 GPR<-GPR moves.
869 2007-01-18 Richard Sandiford <richard@codesourcery.com>
871 * real.h (real_format): Add a canonical_nan_lsbs_set field.
872 (coldfire_single_format): Declare.
873 (coldfire_double_format): Likewise.
874 * real.c (encode_ieee_single): Use canonical_nan_lsbs_set instead
875 of qnan_msb_set to determine the lower bits of a canonical
876 NaN significand.
877 (encode_ieee_double): Likewise.
878 (encode_ieee_quad): Likewise.
879 (ieee_single_format): Initialize canonical_nan_lsbs_set.
880 (mips_single_format): Likewise.
881 (ieee_double_format): Likewise.
882 (mips_double_format): Likewise.
883 (ieee_extended_motorola_format): Likewise.
884 (ieee_extended_intel_96_format): Likewise.
885 (ieee_extended_intel_128_format): Likewise.
886 (ieee_extended_intel_96_round_53_format): Likewise.
887 (ibm_extended_format): Likewise.
888 (mips_extended_format): Likewise.
889 (ieee_quad_format): Likewise.
890 (mips_quad_format): Likewise.
891 (vax_f_format): Likewise.
892 (vax_d_format): Likewise.
893 (vax_g_format): Likewise.
894 (i370_single_format): Likewise.
895 (i370_double_format): Likewise.
896 (decimal_single_format): Likewise.
897 (decimal_double_format): Likewise.
898 (decimal_quad_format): Likewise.
899 (c4x_single_format): Likewise.
900 (c4x_extended_format): Likewise.
901 (real_internal_format): Likewise.
902 (coldfire_single_format): New real_format.
903 (coldfire_double_format): Likewise.
904 * config/pdp11/pdp11.c (pdp11_f_format): Initialize
905 canonical_nan_lsbs_set.
906 (pdp11_d_format): Likewise.
907 * config/m68k/m68k.c (override_options): Override REAL_FORMAT_MODE
908 if TARGET_COLDFIRE_CPU.
910 2007-01-18 Richard Sandiford <richard@codesourcery.com>
912 * config/m68k/m68k-protos.h (m68k_output_pic_call): Delete.
913 (output_call, m68k_legitimize_call_address): Declare.
914 * config/m68k/m68k.h (EXTRA_CONSTRAINT): Remove unnecessary
915 parenthesees. Add support for a 'W' constraint.
916 (LEGITIMATE_PIC_OPERAND_P): Remove SYMBOL_REF_FLAG handling.
917 (PRINT_OPERAND_PUNCT_VALID_P): Remove comment about 'o'.
918 (m68k_symbolic_call, m68k_symbolic_jump): Declare.
919 * config/m68k/m68k.c (m68k_symbolic_call, m68k_symbolic_jump): New
920 variables.
921 (override_options): Initialize them. Do not set flag_no_function_cse
922 for TARGET_ID_SHARED_LIBRARY.
923 (m68k_output_pic_call): Delete.
924 (m68k_legitimize_call_address): New function.
925 (print_operand): Remove the %o prefix. Handle the %p prefix.
926 (output_call): New function.
927 (m68k_output_mi_thunk): Use m68k_symbolic_jump. Always load the
928 target address from the GOT if symbolic jumps are not allowed.
929 * config/m68k/m68k.md (call, general_operand): Do not set
930 SYMBOL_REF_FLAG. Use m68k_legitimize_call_address instead.
931 Merge separate flag_pic and !flag_pic define_insns into...
932 (*call, *call_value): ...these new patterns. Match the address
933 rather than the containing MEM and require it to be a call_operand.
934 Use output_call to generate the asm template.
935 * config/m68k/predicates.md (const_call_operand): New predicate.
936 (call_operand): Likewise.
938 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
940 * config/m68k/m68k.h (REGISTER_MOVE_COST): Simplify definition.
941 (STACK_GROWS_DOWNWARD): Define to 1.
942 (FUNCTION_VALUE, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P): Equivocate
943 comments, emphasizing that these values are only defaults.
944 * config/m68k/linux.h (LINK_SPEC): Fix formatting in #undef.
945 * config/m68k/m68k.c (const_method): Remove trailing whitespace.
947 2007-01-18 Richard Sandiford <richard@codesourcery.com>
949 * config/m68k/m68k.md (cmpsi): Remove outdated flag_pic handling.
951 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
952 Richard Sandiford <richard@codesourcery.com>
954 * config/m68k/m68k.h (DATA_REGNO_P, ADDRESS_REGNO_P, INT_REGNO_P)
955 (FP_REGNO_P): New macros.
956 (REGNO_OK_FOR_INDEX_P, REGNO_OK_FOR_BASE_P, REGNO_OK_FOR_DATA_P)
957 (REGNO_OK_FOR_FP_P, REG_OK_FOR_INDEX_P, REG_OK_FOR_BASE_P): Use them.
958 (EH_RETURN_STACKADJ_RTX): Use A0_REG.
959 * config/m68k/m68k.c (m68k_regno_mode_ok): Use the new REGNO macros.
961 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
963 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
964 (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-rtems*): Add t-floatlib
965 to $tmake_file.
966 * config/m68k/t-floatlib: New file, extracting common code from...
967 * config/m68k/t-m68kbare, config/m68k/t-m68kelf,
968 * config/m68k/t-uclinux: Here.
969 * config/m68k/fpgnulib.c: Do not compile extendeed precision
970 routines on ColdFire targets.
972 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
974 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
975 (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*)
976 (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-linux*)
977 (m68k-*-rtems*): Use tm_file rather than m68k/m68k.h and
978 explicitly set MOTOROLA to 1.
979 * config/m68k/m68k.h (MOTOROLA): Simplify definition accordingly.
981 2007-01-18 Richard Sandiford <richard@codesourcery.com>
982 Nathan Sidwell <nathan@codesourcery.com>
984 * config/m68k/m68k.h (PCC_STATIC_STRUCT_RETURN): Delete.
985 (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Add MOTOROLA cases.
986 * config/m68k/coff.h (REGISTER_PREFIX_MD): Delete.
987 * config/m68k/m68020-elf.h (LIB_SPEC): Delete.
988 * config/m68k/m68k-none.h (CC1_SPEC, CPP_SUBTARGET_SPEC): Delete.
989 * config/m68k/m68kelf.h (IMMEDIATE_PREFIX, REGISTER_PREFIX_MD)
990 (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Delete.
991 (NO_DOLLAR_IN_LABEL, PCC_STATIC_STRUCT_RETURN): Don't undefine.
992 (BSS_ASM_OP, ASM_OUTPUT_SKIP, ASM_OUTPUT_ASCII): Delete.
993 * config/m68k/m68kemb.h (PCC_STATIC_STRUCT_RETURN): Don't undefine.
994 * config/m68k/linux.h (SIZE_TYPE, PTRDIFF_TYPE, WCHAR_TYPE)
995 (WCHAR_TYPE_SIZE, TARGET_OBJFMT_CPP_BUILTINS): Delete.
996 (TARGET_OS_CPP_BUILTINS): Don't define mc68000 and mc68020 here.
997 (DBX_REGISTER_NUMBER): Delete.
998 * config/m68k/netbsd-elf.h (IMMEDIATE_PREFIX): Delete.
999 (PCC_STATIC_STRUCT_RETURN): Don't undefine.
1000 * config/m68k/openbsd.h (PCC_STATIC_STRUCT_RETURN): Define.
1002 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1004 * config.gcc (m68k-*-uclinux*): Add flat.h to $tm_file.
1005 * config/flat.h: New file.
1006 * crtstuff.c (USE_PT_GNU_EH_FRAME): Don't define if
1007 OBJECT_FORMAT_FLAT.
1008 * config/m68k/m68k.h (ASM_PREFERRED_EH_DATA_FORMAT): Do not use
1009 indirect references for -msep-data or -mid-shared-library.
1010 Do not use PC-relative code addresses either.
1011 * config/m68k/m68k.c (override_options): Restrict -fPIC error
1012 to -mpcrel.
1013 * config/m68k/uclinux.h (STARTFILE_SPEC): Define. Use Scrt1.o
1014 for shared libraries and crt1.o for executables. Use crti.o and
1015 crtbegin.o.
1016 (ENDFILE_SPEC): Use crtend.o and crtn.o.
1017 (LIB_SPEC): Suppress -Rlibc.gdb if -static-libc is given.
1018 Do not add -elf2flt or -shared-lib-id options here.
1019 (LINK_SPEC): Define. Pass -elf2flt if no -elf2flt option is given.
1020 Pass -shared-lib-id if -mid-shared-library, taking the library
1021 identifier from -mshared-library-id if given, otherwise
1022 defaulting to 0.
1023 (EH_FRAME_IN_DATA_SECTION): Do not undefine.
1024 (INIT_SECTION_ASM_OP, FINI_SECTION_ASM_OP): Likewise.
1025 (TARGET_OS_CPP_BUILTINS): Define __GXX_MERGED_TYPEINFO_NAMES=0
1026 and __GXX_TYPEINFO_EQUALITY_INLINE=0 if -mid-shared-library.
1027 (DRIVER_SELF_SPECS): Map unadorned PIC options to -msep-data.
1028 * config/m68k/t-uclinux (EXTRA_MULTILIB_PARTS): Add crtbegin.o
1029 and crtend.o.
1030 * config/m68k/lb1sf68.asm (PICCALL): Use an lea and pc-relative
1031 jump sequence for ISA A and ISA A+.
1032 (PICJUMP): Likewise.
1034 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1036 * config.gcc (m68k-*-uclinux*): Base the port on the common
1037 and m68k GNU/Linux files rather than on the generic ELF ones.
1038 * config/m68k/uclinux.h (TARGET_VERSION): Override.
1039 (TARGET_OS_CPP_BUILTINS): Use LINUX_TARGET_OS_CPP_BUILTINS.
1041 2007-01-18 Julian Brown <julian@codesourcery.com>
1042 Richard Sandiford <richard@codesourcery.com>
1044 * config/m68k/m68k.h (LONG_DOUBLE_TYPE_SIZE): Make 64-bit on ColdFire.
1045 (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
1046 * config/m68k/netbsd-elf.h (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Undefine
1047 before redefining.
1048 * config/m68k/uclinux-oldabi.h (LONG_DOUBLE_TYPE_SIZE): Redefine to
1049 80 unconditionally.
1050 (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
1052 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1054 * doc/install.texi: Document m68k-uclinuxoldabi.
1055 * config.gcc (m68k-*-uclinuxoldabi): New configuration.
1056 * config/m68k/uclinux-oldabi.h: New file, copied from
1057 config/m68k/uclinux.h.
1059 2007-01-18 Kaz Kojima <kkojima@rr.iij4u.or.jp>
1061 * config/m32r/m32r-protos.h (m32r_expand_epilogue): Declare it.
1062 * config/m32r/m32r.c (m32r_setup_incoming_varargs): Use gen_frame_mem.
1063 (m32r_compute_frame_size): Use unsigned for regno.
1064 (m32r_reload_lr): Use gen_frame_mem.
1065 (pop): New.
1066 (m32r_output_function_epilogue): Don't output the function epilogue
1067 textually here.
1068 (m32r_expand_epilogue): New.
1069 (direct_return): Return false if the function has the interrupt
1070 attribute.
1071 (m32r_hard_regno_rename_ok): Remove code for the textual epilogue.
1072 * config/m32r/m32r.md (epilogue): New expander.
1073 (return_lr, return_rte): New insns.
1074 (return): Make it expander.
1075 (return_normal): New expander.
1077 2007-01-18 Josh Conner <jconner@apple.com>
1079 PR target/30485
1080 * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Add
1081 support for UNLE, UNLT, UNGE, and UNGT.
1083 2007-01-18 Jan Hubicka <jh@suse.cz>
1085 * tree-vrp.c (finalize_jump_threads): Do not call cleanup_cfg by hand.
1086 * tree-tailcall (add_virtual_phis): Likewise.
1087 (optimize_tail_call): Return TODOs.
1088 (execute_tail_calls): Return TODOs.
1089 * tree-ssa-ccp (execute_fold_all_builtins): Do cleanup_cfg via TODO.
1090 * tree-cfgcleanup.c (cleanup_tree_cfg_loop): Return if something
1091 changed.
1092 * tree-ssa-forwprop.c (tree_ssa_forward_propagate_single_use_value):
1093 Cleanup cfg using TODO.
1094 * tree-flow.h (cleanup_tree_cfg_loop): Update prototype.
1095 * passes.c (execute_function_todo): When cleanup did something, remove
1096 unused locals.
1097 * tree-cfg.c (pass_build_cfg): Add cleanup_cfg TODO.
1098 (make_edges): Don't cleanup_cfg.
1100 2007-01-18 Uros Bizjak <ubizjak@gmail.com>
1102 * reg-stack.c (subst_stack_regs_pat) [UNSPEC_SINCOS_COS,
1103 UNSPEC_XTRACT_FRACT]: Use generic code for instructions that
1104 operate on the top of stack.
1105 [UNSPEC_SINCOS_SIN, UNSPEC_XTRACT_EXP, UNSPEC_TAN]: Rewrite
1106 register handling of instructions that output to the second
1107 stack slot.
1108 [UNSPEC_TAN_ONE, UNSPEC_TAN_TAN]: Remove.
1109 (move_for_stack_reg): Special-case check for dead destination
1110 stack slot for constant load of 1.0 inside UNSPEC_TAN.
1112 * config/i386/i386.md (UNSPEC_TAN): New constant.
1113 (UNSPEC_TAN_ONE, UNSPEC_TAN_TAN): Remove.
1114 (fptanxf4_i387, fptan_extend<mode>xf4_i387): New patterns
1115 to correctly model move of constant 1.0 to top stack slot.
1116 (*tandf3_1, *tansf3_1, *tanxf3_1): Remove insn patterns.
1117 (unnamed peephole2 pattern): Remove corresponding peephole2
1118 pattern that optimizes tan insn and loading of constant 1.0.
1119 (tanxf2): Use fptanxf4_i387.
1120 (tan<mode>2): Rename from tansf2 and tandf2 and macroize insn
1121 patterns using X87MODEF12 mode macro. Use fptan_extend<mode>xf4_i387
1122 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
1123 disable patterns for SSE math.
1124 (sincos<mode>3): Use truncxf<mode>2_i387_noop for truncation.
1125 (fyl2x_extend<mode>xf3_i387): Use X87MODEF12 for operand 1.
1127 2007-01-18 Dirk Mueller <dmueller@suse.de>
1128 Richard Guenther <rguenther@suse.de>
1130 PR diagnostic/8268
1131 * doc/invoke.texi (Warray-bounds): Document -Warray-bounds.
1132 * common.opt (Warray-bounds): Add new warning option.
1133 * c-opts.c (c_common_handle_option): Define -Warray-bounds
1134 if -Wall is given.
1135 * Makefile.in: make tree-vrp.o depend on toplev.h
1136 * tree-vrp.c (vrp_finalize): Call check_array_refs if -Warray-bounds
1137 is enabled.
1138 (check_array_refs, check_array_bounds, check_array_ref): New.
1140 2007-01-18 Jan Hubicka <jh@suse.cz>
1142 * tree-ssa-ccp.c (ccp_finalize): Return if something changed.
1143 (execute_ssa_ccp): Return flags conditionally.
1144 * tree-ssa-propagate.c (substitue_and_fold): Return if something was
1145 changed.
1146 * tree-ssa-propagate.h (substitute_and_fold): Update prototype.
1148 2007-01-18 Steven Bosscher <steven@gcc.gnu.org>
1150 * cfgcleanup.c (cleanup_cfg): Detect cfglayout mode and set
1151 the CLEANUP_CFGLAYOUT flag when in cfglayout mode.
1153 * Makefile.c (GTFILES): Add cfglayout.h.
1154 * gengtype.c (open_base_files): Likewise.
1155 * cfglayout.c (cfg_layout_function_footer,
1156 cfg_layout_function_header) Reindent to make gengtype happy.
1157 * cfglayout.h (cfg_layout_function_footer,
1158 cfg_layout_function_header): Add GTY(()) marker.
1160 * ifcvt.c (noce_try_sign_mask): Make sure INSN_B is non-null.
1162 2007-01-18 Ben Elliston <bje@au.ibm.com>
1164 * genautomata.c (write_automata): Include xstrerror output in the
1165 error message if writing the DFA description file fails.
1167 2007-01-17 H.J. Lu <hongjiu.lu@intel.com>
1169 * config/mips/mips-protos.h (mips_output_external): Make it
1170 return void.
1171 * config/mips/iris.h (TARGET_ASM_EXTERNAL_LIBCALL): Removed.
1172 * config/mips/mips.c (irix_output_external_libcall): Likewise.
1173 (extern_list): Likewise.
1174 (extern_head): Likewise.
1175 (TARGET_ASM_FILE_END): Likewise.
1176 (mips_file_end): Likewise.
1177 (mips_output_external): Rewritten.
1179 2007-01-18 Ben Elliston <bje@au.ibm.com>
1181 * genpreds.c (write_insn_preds_c): Only write out the function
1182 body for regclass_for_constraint if we have register constraints.
1184 2007-01-17 Tom Tromey <tromey@redhat.com>
1186 * doc/sourcebuild.texi (libgcj Tests): Use sourceware.org.
1187 * doc/install.texi (Testing): Use sourceware.org.
1188 (Binaries): Likewise.
1189 (Specific): Likewise.
1190 * doc/contrib.texi (Contributors): Use sourceware.org.
1192 2007-01-17 Anatoly Sokolov <aesok@post.ru>
1194 * config/avr/avr.h (AVR_HAVE_LPMX): New macro.
1195 (AVR_ENHANCED): Rename to ...
1196 (AVR_HAVE_MUL): ... new.
1197 (avr_enhanced_p): Rename to ...
1198 (avr_have_mul_p): ... new.
1199 (TARGET_CPU_CPP_BUILTINS): Use 'avr_have_mul_p' instead of
1200 'avr_enhanced_p' for "__AVR_ENHANCED__". Define "__AVR_HAVE_MUL__".
1201 * config/avr/avr.c (avr_enhanced_p): Rename to ...
1202 (avr_have_mul_p): ... new.
1203 (base_arch_s): Rename 'enhanced' to 'have_mul'.
1204 (avr_override_options): Use 'avr_have_mul_p' and 'have_mul' instead of
1205 'avr_enhanced_p' and 'enhanced'.
1206 (ashlhi3_out, ashrhi3_out, lshrhi3_out, avr_rtx_costs): Use
1207 AVR_HAVE_MUL instead of AVR_ENHANCED.
1208 * avr.md (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
1209 (mulqi3, *mulqi3_enh, *mulqi3_call, mulqihi3, umulqihi3, mulhi3,
1210 *mulhi3_enh, *mulhi3_call, mulsi3, *mulsi3_call): Use AVR_HAVE_MUL
1211 instead of AVR_ENHANCED.
1212 (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
1213 * libgcc.S: Use __AVR_HAVE_MUL__ instead of __AVR_ENHANCED__.
1214 (__tablejump__): Use __AVR_HAVE_LPMX__ instead of __AVR_ENHANCED__.
1216 2007-01-17 Ian Lance Taylor <iant@google.com>
1218 * vec.h (VEC_reserve_exact): Define.
1219 (vec_gc_p_reserve_exact): Declare.
1220 (vec_gc_o_reserve_exact): Declare.
1221 (vec_heap_p_reserve_exact): Declare.
1222 (vec_heap_o_reserve_exact): Declare.
1223 (VEC_OP (T,A,reserve_exact)): New static inline function, three
1224 versions.
1225 (VEC_OP (T,A,reserve)) [all versions]: Remove handling of
1226 negative parameter.
1227 (VEC_OP (T,A,alloc)) [all versions]: Call ...reserve_exact.
1228 (VEC_OP (T,A,copy)) [all versions]: Likewise.
1229 (VEC_OP (T,a,safe_grow)) [all versions]: Likewise.
1230 * vec.c (calculate_allocation): Add exact parameter. Change all
1231 callers.
1232 (vec_gc_o_reserve_1): New static function, from vec_gc_o_reserve.
1233 (vec_gc_p_reserve, vec_gc_o_reserve): Call vec_gc_o_reserve_1.
1234 (vec_gc_p_reserve_exact, vec_gc_o_reserve_exact): New functions.
1235 (vec_heap_o_reserve_1): New static function, from vec_heap_o_reserve.
1236 (vec_heap_p_reserve, vec_heap_o_reserve): Call vec_heap_o_reserve_1.
1237 (vec_heap_p_reserve_exact): New function.
1238 (vec_heap_o_reserve_exact): New function.
1240 2007-01-17 Jan Hubicka <jh@suse.cz>
1242 * ipa-type-escape.c (look_for_casts): Revamp using handled_component_p.
1244 2007-01-17 Eric Christopher <echristo@apple.com>
1246 * config.gcc: Support core2 processor.
1248 2007-01-16 Jan Hubicka <jh@suse.cz>
1250 * tree-ssanames.c (release_dead_ssa_names): Instead of ggc_freeing
1251 the names, just unlink the chain so we don't crash on dangling pointers
1252 to dead SSA names.
1254 2007-01-16 Jan Hubicka <jh@suse.cz>
1256 * cgraph.h (cgraph_decide_inlining_incrementally): Kill.
1257 * tree-pass.h: Reorder to make IPA passes appear toegher.
1258 (pass_early_inline, pass_inline_parameters, pass_apply_inline): Declare.
1259 * cgraphunit.c (cgraph_finalize_function): Do not compute inling
1260 parameters, do not call early inliner.
1261 * ipa-inline.c: Update comments. Include tree-flow.h
1262 (cgraph_decide_inlining): Do not compute inlining parameters.
1263 (cgraph_decide_inlining_incrementally): Return TODOs; assume to
1264 be called with function context set up.
1265 (pass_ipa_inline): Remove unreachable functions before pass.
1266 (cgraph_early_inlining): Simplify assuming to be called from the
1267 PM as local pass.
1268 (pass_early_inline): New pass.
1269 (cgraph_gate_ipa_early_inlining): New gate.
1270 (pass_ipa_early_inline): Turn into simple wrapper.
1271 (compute_inline_parameters): New function.
1272 (gate_inline_passes): New gate.
1273 (pass_inline_parameters): New pass.
1274 (apply_inline): Move here from tree-optimize.c
1275 (pass_apply_inline): New pass.
1276 * ipa.c (cgraph_remove_unreachable_nodes): Verify cgraph after
1277 transforming.
1278 * tree-inline.c (optimize_inline_calls): Return TODOs rather than
1279 doing them by hand.
1280 (tree_function_versioning): Do not allocate dummy struct function.
1281 * tree-inline.h (optimize_inline_calls): Update prototype.
1282 * tree-optimize.c (execute_fixup_cfg): Export.
1283 (pass_fixup_cfg): Remove
1284 (tree_rest_of_compilation): Do not apply inlines.
1285 * tree-flow.h (execute_fixup_cfg): Declare.
1286 * Makefile.in (gt-passes.c): New.
1287 * passes.c: Include gt-passes.h
1288 (init_optimization_passes): New passes.
1289 (nnodes, order): New static vars.
1290 (do_per_function_toporder): New function.
1291 (execute_one_pass): Dump current pass here.
1292 (execute_ipa_pass_list): Don't dump current pass here.
1294 2007-01-16 Janis Johnson <janis187@us.ibm.com>
1296 * config/dfp-bit.c (dfp_compare_op): Return separate value for NaN.
1297 (DFP_NE, DFP_LE, DFP_GE): Return false for NaN.
1299 2007-01-16 David Edelsohn <edelsohn@gnu.org>
1301 * config/rs6000/darwin-ldouble.c: Build file for SOFT_FLOAT.
1302 (strong_alias): Define.
1303 (__gcc_qmul): Provide non-FMA for soft-float.
1304 (__gcc_qdiv): Same.
1305 (__gcc_qneg): New.
1306 (__gcc_qeq): New.
1307 (__gcc_qle): New.
1308 (__gcc_qge): New.
1309 (__gcc_qunord): New.
1310 (__gcc_stoq): New.
1311 (__gcc_dtoq): New.
1312 (__gcc_qtos): New.
1313 (__gcc_qtod): New.
1314 (__gcc_qtoi): New.
1315 (__gcc_qtou): New.
1316 (__gcc_itoq): New.
1317 (__gcc_utoq): New.
1318 (fmsub): New.
1319 * config/rs6000/rs6000.c (rs6000_init_libfuncs): Initialize
1320 soft-float functions.
1321 * config/rs6000/libgcc-ppc-glibc.ver: Version soft-float symbols.
1322 * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Do not warn
1323 about long double soft float.
1325 2007-01-16 Dorit Nuzman <dorit@il.ibm.com>
1326 Tehila Meyzels <tehila@il.ibm.com>
1328 * tree-vectorizer.h (is_pattern_stmt_p): New.
1329 * tree-vect-analyze.c (vect_determine_vectorization_factor): Fix
1330 formatting (tabs instead of spaces). Cleanup and clarify setting
1331 of STMT_VINFO_VECTYPE. Call is_pattern_stmt_p.
1332 * tree-vect-transform.c (vect_get_vec_def_for_operand): Fix typo.
1333 (vectorizable_type_demotion): Check that types are integral.
1334 (vectorizable_type_promotion): Likewise.
1335 (vectorizable_store): Fix typo. Eliminate new-line at end of
1336 comments.
1338 2007-01-16 Jan Hubicka <jh@suse.cz>
1340 * tree-ssanames.c (release_dead_ssa_names): Remove invalidated
1341 cgraph edges too.
1343 2007-01-15 Eric Christopher <echristo@apple.com>
1345 * ifcvt.c: Include vec.h, vecprim.h.
1346 (check_cond_move_block): New argument regs.
1347 Reorganize. Add registers used to regs.
1348 (cond_move_process_if_block): Use regs set above as
1349 loop bounds.
1351 2007-01-15 Eric Christopher <echristo@apple.com>
1353 * config/darwin.h: Update copyright.
1354 (TARGET_OPTION_TRANSLATE_TABLE): Add umbrella.
1355 (LINK_COMMAND_SPEC): Add -u.
1356 (LINK_SPEC): Fix umbrella for above.
1358 2007-01-15 Joseph S. Myers <joseph@codesourcery.com>
1360 * config/soft-fp/op-common.h, config/soft-fp/op-4.h: Update from
1361 glibc CVS.
1363 2007-01-15 Tom Tromey <tromey@redhat.com>
1365 * doc/sourcebuild.texi (libgcj Tests): Don't mention jacks.
1366 * doc/install.texi (Testing): Don't mention jacks.
1367 (Configuration): Document --enable-java-maintainer-mode. Move
1368 --with-java-home to libgcj-specific section. Document
1369 --with-ecj-jar.
1370 (Prerequisites): Mention --enable-java-maintainer-mode, ecj1.
1372 2007-01-15 Jan Hubicka <jh@suse.cz>
1374 * tree-ssa-dce.c (DCE_TODOs): New.
1375 (propagate_necessity): Return if something changed.
1376 (eliminate_unnecessary_stmts): Likewise.
1377 (perform_tree_ssa_dce): Return TODO flags when needed.
1378 (pass_dce, pass_dce_loop, pass_cd_dce): Remove TODO flags.
1380 2007-01-15 Uros Bizjak <ubizjak@gmail.com>
1382 * config/i386/i386.md (fyl2xxf3_i387): Rename from fyl2x_xf3.
1383 (fyl2x_extend<mode>xf3_i387): New insn pattern.
1384 (log<mode>2): Rename from logsf2 and logdf2 and macroize insn
1385 patterns using X87MODEF12 mode macro. Extend operand 1
1386 to XFmode. Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
1387 (log10<mode>2): Ditto.
1388 (log2<mode>2): Ditto.
1389 (log1p<mode>2): Ditto.
1390 (logb<mode>2): Ditto.
1391 (fyl2xp1xf3_i387): Rename from fyl2xp1_xf3.
1392 (fyl2xp1_extend<mode>xf3_i387): New insn pattern.
1393 (*fxtractxf3_i387): Rename from *fxtractxf3.
1394 (fxtract_extend<mode>xf3_i387): New insn pattern.
1395 (ilogbsi2): Use match_dup 3, not match_operand:XF 3.
1396 * config/i386/i386.c (ix86_emit_i387_log1p): Use gen_fyl2xp1xf3_i387()
1397 and gen_fyl2xxf3_i387().
1399 2007-01-14 Zdenek Dvorak <dvorakz@suse.cz>
1401 * loop-unswitch.c (unswitch_loop): Do not call fix_loop_placement.
1402 * cfgloopmanip.c (fix_loop_placement): Made static. Use
1403 get_loop_exit_edges. Changed return type to bool.
1404 * cfgloop.h (fix_loop_placement): Declaration removed.
1406 2007-01-14 Dorit Nuzman <dorit@il.ibm.com>
1408 * param.h (MIN_VECT_LOOP_BOUND): New.
1409 * params.def (MIN_VECT_LOOP_BOUND): New.
1410 * tree-vectorizer.c (slpeel_tree_peel_loop_to_edge): Takes another
1411 argument - minimum threshold for number of iterations.
1412 * tree-vectorizer.h (slpeel_tree_peel_loop_to_edge): Add another
1413 argument to declaration.
1414 * tree-vect-analyze.c (vect_analyze_operations): Check value of
1415 MIN_VECT_LOOP_BOUND.
1416 * tree-vect-transform.c (vect_do_peeling_for_loop_bound): Call
1417 slpeel_tree_peel_loop_to_edge with additional argument.
1418 (vect_do_peeling_for_alignment): Likewise.
1419 * doc/invoke.texi (min-vect-loop-bound): Document new param option.
1421 2007-01-14 Uros Bizjak <ubizjak@gmail.com>
1423 PR target/30413
1424 * config/i386/i386.c (print_operand) ['z']: Output 'b' for
1425 operands of size 1.
1427 2007-01-14 Jan Hubicka <jh@suse.cz>
1429 * tree-dfa.c (remove_referenced_var): New function.
1430 * tree-ssa-live.c (remove_unused_locals): Walk referenced vars and
1431 prune referenced vars list too.
1432 * tree-flow.h (remove_referenced_var): Declare.
1434 2007-01-14 Jan Hubicka <jh@suse.cz>
1436 * tree-eh.c (add_stmt_to_eh_region_fn): Do not add call_exprs
1437 separately.
1438 (remove_stmt_from_eh_region_fn): Do not remove call_exprs.
1439 (verify_eh_throw_stmt_node, verify_eh_throw_table_statements): Kill.
1440 * except.h (verify_eh_throw_table_statements): Kill prototype.
1441 * cfgexpand.c (expand_gimple_basic_block): Propagate Eh regions
1442 into call exrepssions.
1443 * tree-optimize.c (execute_free_cfg_annotatiosn): Do not call
1444 eh trhow verifier.
1445 * tree-cfg.c: Include pointer-set.h.
1446 (verify_node_sharing): Work on pointer set.
1447 (verify_eh_throw_stmt_node): New.
1448 (verify_stmts): Use pointers sets, verify throw_stmt.
1450 2007-01-13 Zdenek Dvorak <dvorakz@suse.cz>
1452 * ipa-reference.c (analyze_function): Consider also addresses taken
1453 in phi nodes.
1455 2007-01-12 Roger Sayle <roger@eyesopen.com>
1457 * c-typeck.c (null_pointer_constant_p): Replace use of
1458 TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
1459 (build_c_cast): Likewise.
1461 2007-01-12 Roger Sayle <roger@eyesopen.com>
1463 * tree.h (force_fit_type_double): Remove unused final argument.
1464 * c-common.c (constant_expression_warning): Replace use of
1465 TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
1466 (convert_and_check): Likewise.
1467 (shorten_compare): Update call to force_fit_type_double.
1468 (c_common_truthvalue_conversion) <INTEGER_CST>: Use integer_zerop.
1469 * convert.c (convert_to_pointer): Update call to
1470 force_fit_type_double.
1471 * fold-const.c (force_fit_type_double): Remove overflowed_const
1472 argument.
1473 (int_const_binop, fold_convert_const_int_from_int,
1474 fold_convert_const_int_from_real, fold_div_compare,
1475 fold_sign_changed_comparison, fold_unary, fold_negate_const,
1476 fold_abs_const, fold_not_const): Remove the final argument from
1477 calls to force_fit_type_double.
1479 2007-01-12 Andrew Pinski <andrew_pinski@playstation.sony.com>
1481 * configure.ac: Set insn to "nop" for spu-*-* also.
1482 * configure: Regenerate.
1484 2007-01-12 Olga Golovanevsky <olga@il.ibm.com>
1486 * builtins.def : Add BUILT_IN_FREE.
1488 2007-01-12 Jan Hubicka <jh@suse.cz>
1490 PR tree-optimization/30443
1491 * tree-inline.c (tree_function_versioning): Do not optimize when
1492 cloning for inlining.
1494 2007-01-12 Zdenek Dvorak <dvorakz@suse.cz>
1496 * doc/loop.texi: Document recording of loop exits.
1497 * cfgloopmanip.c (loopify, duplicate_loop): Use alloc_loop.
1498 (update_single_exits_after_duplication,
1499 update_single_exit_for_duplicated_loop,
1500 update_single_exit_for_duplicated_loops): Removed.
1501 (duplicate_loop_to_header_edge): Do not call
1502 update_single_exits_after_duplication and
1503 update_single_exit_for_duplicated_loops.
1504 (loop_version): Do not update single_exit information.
1505 (fix_loop_structure): Use record_loop_exits instead of
1506 mark_single_exit_loops.
1507 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update
1508 the lists of loop exits.
1509 * cfghooks.c (redirect_edge_and_branch, redirect_edge_and_branch_force,
1510 split_edge, merge_blocks): Update the lists of loop exits.
1511 * modulo-sched.c (sms_schedule): Pass LOOPS_HAVE_RECORDED_EXITS to
1512 loop_optimizer_init.
1513 * loop-init.c (loop_optimizer_init): Call record_loop_exits instead
1514 of mark_single_exit_loops.
1515 (loop_optimizer_finalize): Call release_recorded_exits.
1516 * tree-ssa-loop.c (tree_loop_optimizer_init): Pass
1517 LOOPS_HAVE_RECORDED_EXITS to loop_optimizer_init.
1518 * tree-vectorizer.c (slpeel_tree_duplicate_loop_to_edge_cfg): Do not
1519 update single exit information.
1520 * lambda-code.c (perfect_nestify): Ditto.
1521 * cfgloop.c (flow_loop_free): Destroy the list of exits of the loop.
1522 (mark_single_exit_loops): Removed.
1523 (alloc_loop, loop_exit_hash, loop_exit_eq, loop_exit_free,
1524 get_exit_descriptions, rescan_loop_exit, record_loop_exits,
1525 dump_recorded_exit, dump_recorded_exits, release_recorded_exits): New
1526 functions.
1527 (get_loop_exit_edges, single_exit): Use recorded exit lists.
1528 (add_bb_to_loop, remove_bb_from_loops): Update the lists of loop exits.
1529 (verify_loop_structure): Verify consistency of the exit lists.
1530 (flow_loops_find): Use alloc_loop. Initialize exits hash.
1531 (set_single_exit): Removed.
1532 * cfgloop.h (struct loop_exit): New function.
1533 (struct loop): single_exit_ field replaced by exits field.
1534 (LOOPS_HAVE_MARKED_SINGLE_EXITS): Replaced by LOOPS_HAVE_RECORDED_EXITS.
1535 (struct loops): Added exits hash.
1536 (mark_single_exit_loops, set_single_exit): Declaration removed.
1537 (release_recorded_exits, record_loop_exits, rescan_loop_exit): Declare.
1539 2007-01-12 Richard Sandiford <richard@codesourcery.com>
1541 * doc/invoke.texi: Avoid use of @headitem.
1543 2007-01-12 Richard Sandiford <richard@codesourcery.com>
1545 * cse.c (cse_insn): Move HAVE_CC0 code after declarations.
1547 2007-01-12 Richard Sandiford <richard@codesourcery.com>
1549 * doc/install.texi: Fix m68k-*-* anchor and add m68k-*-* to the
1550 list of targets.
1552 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
1553 Richard Sandiford <richard@codesourcery.com>
1555 * doc/invoke.texi: Document -mno-bitfield, -mno-rtd and -mno-short.
1556 * config/m68k/m68k.opt: Resort options.
1557 (mbitfield, mrtd, mshort): Remove RejectNegative properties.
1559 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
1560 Richard Sandiford <richard@codesourcery.com>
1562 * doc/invoke.texi: Document the macros that are defined by
1563 m68k's -mtune and -mhard-float options.
1564 * config/m68k/m68k-protos.h (m68k_cpp_cpu_ident) Declare.
1565 (m68k_cpp_cpu_family): Likewise.
1566 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Add a full set
1567 of __ucfv*__ macros. Define __mcffpu__ if generating code for
1568 ColdFire FPUs. Define __mcf_cpu_* and __mcf_family_* macros.
1569 * config/m68k/m68k.c (m68k_cpp_cpu_ident): New function.
1570 (m68k_cpp_cpu_family): Likewise.
1572 2007-01-12 Richard Sandiford <richard@codesourcery.com>
1574 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Treat all mc68*
1575 macros besides mc68000 as tuning macros. Use a switch statement
1576 to set them and mcpu32.
1578 2007-01-12 Julian Brown <julian@codesourcery.com>
1580 * config/m68k/m68k.h: Use TARGET_68040 instead of TARGET_68040_ONLY.
1581 (TARGET_68040_ONLY): Rename to...
1582 (TARGET_68040): ...this.
1583 * config/m68k/m68k.c: Use TARGET_68040 instead of TARGET_68040_ONLY.
1584 * config/m68k/m68k.md: Likewise.
1586 2007-01-12 Julian Brown <julian@codesourcery.com>
1587 Nathan Sidwell <nathan@codesourcery.com>
1588 Richard Sandiford <richard@codesourcery.com>
1590 * config.gcc (m680[012]0-*-*, m68k*-*-*): Set m68k_cpu_ident to
1591 the -mcpu= argument associated with the --with-cpu setting.
1592 Define M68K_DEFAULT_TUNE to the default -mtune= option,
1593 if different from the one implied by the -mcpu setting.
1594 Accept --with-cpu=FOO if FOO is listed in m68k-devices.def,
1595 using mcpu=FOO as the default CPU option. Set target_cpu_default2.
1596 * doc/invoke.texi: Mention ColdFire in the introduction to the
1597 m68k options. Document the new -march, -mcpu, -mtune, -mdiv,
1598 -mno-div and -mhard-float options. Make -m68881 a synonym for
1599 -mhard-float. Document the previously-undocumented -m5206e,
1600 -m528x, -m5307 and -m5407 options. Tweak the existing option
1601 documentation for consistency.
1602 * doc/install.texi: Mention new --with-cpu arguments.
1603 * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Only use the
1604 default CPU if neither -mcpu nor -march are specified.
1605 (ASM_CPU_SPEC): Pass down -mcpu and -march options.
1606 (TARGET_CPU_CPP_BUILTINS): Set __mcfisa*__ macros from
1607 TARGET_ISA*. Set the legacy __mcf*__ cpu macros in the same way,
1608 using m68k_tune to decide between families that implement the
1609 same ISA. Use m68k_tune to set __mcfv4e__.
1610 (FL_BITFIELD, FL_68881, FL_COLDFIRE, FL_CF_HWDIV, FL_CF_MAC)
1611 (FL_CF_EMAC, FL_CF_EMAC_B, FL_CF_USP, FL_CF_FPU, FL_ISA_68000)
1612 (FL_ISA_68010, FL_ISA_68020, FL_ISA_68040, FL_ISA_A, FL_ISA_B)
1613 (FL_ISA_C, FL_ISA_MMU): New macros.
1614 (MASK_COLDFIRE): Delete.
1615 (TARGET_68010, TARGET_68020, TARGET_68040_ONLY, TARGET_COLDFIRE)
1616 (TARGET_ISAB): Redefine in terms of m68k_cpu_flags.
1617 (TARGET_68881, TARGET_COLDFIRE_FPU): Redefine in terms of m68k_fpu.
1618 (TARGET_HARD_FLOAT): Do not define here.
1619 (TARGET_ISAAPLUS, TARGET_ISAC): New macros.
1620 (TUNE_68000): New macro.
1621 (TUNE_68000_10): Redefine in terms of TUNE_68000 and TUNE_68010.
1622 (TUNE_68010, TUNE_68030, TUNE_68040, TUNE_68060, TUNE_CPU32)
1623 (TUNE_CFV2): Redefine in terms of m68k_tune.
1624 (uarch_type, target_device, fpu_type): New enums.
1625 (m68k_cpu, m68k_tune, m68k_fpu, m68k_cpu_flags): Declare.
1626 * config/m68k/m68k.c (TARGET_DEFAULT): Remove MASK_68881.
1627 (FL_FOR_isa_00, FL_FOR_isa_10, FL_FOR_isa_20, FL_FOR_isa_40)
1628 (FL_FOR_isa_cpu32, FL_FOR_isa_a, FL_FOR_isa_aplus, FL_FOR_isa_b)
1629 (FL_FOR_isa_c): New macros.
1630 (m68k_isa): New enum.
1631 (m68k_target_selection): New structure.
1632 (all_devices, all_isas, all_microarchs): New tables.
1633 (m68k_cpu_entry, m68k_arch_entry, m68k_tune_entry, m68k_cpu)
1634 (m68k_tune, m68k_fpu, m68k_cpu_flags): New variables.
1635 (MASK_ALL_CPU_BITS): Delete.
1636 (m68k_find_selection): New function.
1637 (m68k_handle_option): Handle -mcpu=, -march= and -mtune=.
1638 Map the legacy target options to a combination of the new ones.
1639 (override_options): Set m68k_cpu, m68k_tune, m68k_fpu and
1640 m68k_cpu_flags. Handle M68K_DEFAULT_TUNE. Use m68k_cpu_flags
1641 to derive default MASK_BITFIELD, MASK_CF_HWDIV and MASK_HARD_FLOAT
1642 settings.
1643 * config/m68k/m68k.opt (m5200, m5206e, m528x, m5307, m5407, mcfv4e)
1644 (m68010, m68020, m68020-40, m68020-60, m68030, m68040): Remove Mask
1645 properties.
1646 (m68881, msoft-float): Change mask from 68881 to HARD_FLOAT.
1647 (march=, mcpu=, mdiv, mhard-float, mtune=): New options.
1648 * config/m68k/m68k-devices.def: New file.
1650 2007-01-12 Richard Sandiford <richard@codesourcery.com>
1651 Nathan Sidwell <nathan@codesourcery.com>
1653 * config/m68k/m68k.h (ASM_CPU_SPEC, ASM_SPEC, EXTRA_SPECS)
1654 (SUBTARGET_EXTRA_SPECS): New macros.
1655 * config/m68k/linux.h (ASM_SPEC): Remove CPU flags;
1656 use %(asm_cpu_spec) instead.
1657 * config/m68k/m68k-none.h (ASM_SPEC): Likewise.
1658 * config/m68k/openbsd.h (ASM_SPEC): Likewise.
1659 * config/m68k/netbsd-elf.h (ASM_SPEC): Likewise.
1660 (EXTRA_SPECS): Rename to...
1661 (SUBTARGET_EXTRA_SPECS): ...this.
1663 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
1664 Richard Sandiford <richard@codesourcery.com>
1665 Julian Brown <julian@codesourcery.com>
1667 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
1668 (m68k-*-uclinux*, m68k-*-linux*, m68k-*-rtems*): Set default_m68k_cpu
1669 to the configuration's default CPU.
1670 (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*): Likewise.
1671 Remove default masks.
1672 (m680[012]0-*-*): Set the default with_cpu to the first part of
1673 the target name.
1674 (m68k*-*-*): Set the default with_cpu to m$default_m68k_cpu.
1675 (m68k*-*-linux): Extend the --with-cpu handling to...
1676 (m680[012]0-*-*, m68k*-*-*): ...these configurations. Allow m68000
1677 and m68010. Don't set target_cpu_default2.
1678 * doc/install.texi: Document --with-cpu for m68k.
1679 * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Define.
1680 * config/m68k/m68k-none.h (TARGET_CPU_DEFAULT, M68K_CPU_m68k)
1681 (M68K_CPU_m68000, M68K_CPU_m68010, M68K_CPU_m68020, M68K_CPU_m68030)
1682 (M68K_CPU_m68040, M68K_CPU_m68302, M68K_CPU_m68332, TARGET_DEFAULT)
1683 (ASM_CPU_DEFAULT_SPEC, CC1_CPU_DEFAULT_SPEC): Delete.
1684 (ASM_SPEC): Remove use of %(asm_cpu_default).
1685 (EXTRA_SPECS, SUBTARGET_EXTRA_SPECS, MULTILIB_DEFAULTS): Delete.
1686 * config/m68k/linux.h (TARGET_DEFAULT): Delete.
1687 (CPP_SPEC): Merge definitions. Do not handle __HAVE_68881__ here.
1688 * config/m68k/netbsd-elf.h (TARGET_OS_CPP_BUILTINS): Define
1689 __HAVE_FPU__ if TARGET_HARD_FLOAT.
1690 (TARGET_DEFAULT): Delete.
1691 (EXTRA_SPECS): Delete cpp_cpu_default_spec, cpp_cpu_spec,
1692 cpp_fpu_spec, asm_default_spec and netbsd_cpp_spec.
1693 (CPP_CPU_SPEC): Delete.
1694 (TARGET_VERSION): Merge definitions, using TARGET_68010 to pick
1695 the appropriate string.
1696 (CPP_CPU_DEFAULT_SPEC, ASM_DEFAULT_SPEC, CPP_FPU_SPEC): Delete.
1697 (CPP_SPEC): Define to NETBSD_CPP_SPEC.
1698 (ASM_SPEC): Don't use %(asm_default_spec).
1699 * config/m68k/m68k.c (TARGET_DEFAULT_TARGET_FLAGS): Remove
1700 TARGET_DEFAULT and add MASK_68881.
1701 * config/m68k/m68k.md: Remove mention of TARGET_DEFAULT from comments.
1703 2007-01-12 Richard Sandiford <richard@codesourcery.com>
1705 * config.gcc (m68010-*-netbsdelf*): Add MASK_68010.
1706 (m68k*-*-netbsdelf*, m68k*-*-openbsd*, m68k*-linux*): Add
1707 MASK_68010 alongside MASK_68020.
1708 * doc/invoke.texi: Document -m68010.
1709 * config/m68k/m68k.opt (m68010): New.
1710 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Define mc68010
1711 if TUNE_68010.
1712 (TUNE_68010): New macro.
1713 * config/m68k/m68k-none.h (M68K_CPU_m68k, M68K_CPU_m68010)
1714 (M68K_CPU_m68020, M68K_CPU_m68030, M68K_CPU_m68040)
1715 (M68K_CPU_m68332): Add MASK_68010.
1716 * config/m68k/linux.h (TARGET_DEFAULT): Add MASK_68010 to
1717 fallback definition.
1718 * config/m68k/netbsd-elf.h (CPP_CPU_SPEC): Remove now-redundant
1719 defines.
1720 * config/m68k/m68k.c (MASK_ALL_CPU_BITS): Add MASK_68010.
1721 (m68k_handle_option): Handle OPT_m68010. Add MASK_68010
1722 to all entries that use MASK_68020.
1723 (output_move_simode_const, output_move_himode, output_move_qimode)
1724 (output_move_stricthi, output_move_strictqi): Use TARGET_68010
1725 instead of TARGET_68020 to select clr behavior. Remove comment
1726 about there being no TARGET_68010.
1727 * config/m68k/m68k.md: Likewise throughout.
1729 2007-01-12 Julian Brown <julian@codesourcery.com>
1731 * config/m68k/m68k.h (TARGET_ISAB): New macro.
1732 * config/m68k/m68k.c: Use TARGET_ISAB rather than TARGET_CFV4.
1733 * config/m68k/m68k.md: Likewise.
1735 2007-01-12 Julian Brown <julian@codesourcery.com>
1737 * config/m68k/m68k.h (LEGITIMATE_INDEX_P, LEGITIMIZE_ADDRESS): Use
1738 TARGET_COLDFIRE_FPU instead of TARGET_CFV4E.
1740 2007-01-12 Julian Brown <julian@codesourcery.com>
1742 * config/m68k/m68k.h (TUNE_68040_60): New macro.
1743 * config/m68k/m68k.c (standard_68881_constant_p): Use it.
1744 * config/m68k/m68k.md: Likewise.
1746 2007-01-12 Julian Brown <julian@codesourcery.com>
1747 Richard Sandiford <richard@codesourcery.com>
1749 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Use TUNE_68030
1750 instead of TARGET_68030, TUNE_68040 instead of TARGET_68040,
1751 TUNE_68060 instead of TARGET_68060 and TUNE_CPU32 instead of
1752 TARGET_CPU32.
1753 (TARGET_CPU32): Rename to...
1754 (TUNE_CPU32): ...this.
1755 (TUNE_68000_10, TUNE_68030, TUNE_68040, TUNE_68060)
1756 (TUNE_CFV2): New macros.
1757 * config/m68k/netbsd-elf.h (LONG_DOUBLE_TYPE_SIZE): Simplify;
1758 remove conditions that are implied by TARGET_68020.
1759 * config/m68k/m68k.c (m68k_output_function_prologue): Use TUNE_68040
1760 instead of TARGET_68040 and TUNE_CPU32 instead of TARGET_CPU32.
1761 (m68k_output_function_epilogue): Likewise.
1762 (m68k_rtx_costs): Likewise. Use TUNE_68060 instead of TARGET_68060
1763 and TUNE_CFV2 instead of TARGET_5200. Use TUNE_68000_10 instead of
1764 "!TARGET_68020 && !TARGET_COLDFIRE" to choose between 68000 and
1765 non-68000 timings. Refactor multiplication and division costs.
1766 (output_addsi3): Use TUNE_68040 instead of TARGET_68040 and
1767 TUNE_CPU32 instead of TARGET_CPU32.
1768 (standard_68881_constant_p): Use TUNE_68040 instead of TARGET_68040
1769 and TUNE_68060 instead of TARGET_68060.
1770 * config/m68k/m68k.md: Use TUNE_68040 instead of TARGET_68040,
1771 TUNE_68060 instead of TARGET_68060, and TUNE_CPU32 instead of
1772 TARGET_CPU32.
1773 (movsi_const0): Use TUNE_68000_10 rather than "!TARGET_68020
1774 && !TARGET_COLDFIRE" to choose between moveq and clr.
1775 Likewise in the unnamed movsf pattern.
1776 (ashlsi_17_24, lshrsi_17_24): Guard with TUNE_68000_10 rather than
1777 "!TARGET_68020 && !TARGET_COLDFIRE". Likewise the unnamed
1778 ashiftrt pattern.
1780 2007-01-12 Richard Sandiford <richard@codesourcery.com>
1782 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Increase amount
1783 of tabbing before backslashes.
1785 2007-01-11 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
1787 * pa-linux.h (ASM_OUTPUT_INTERNAL_LABEL): Undefine.
1788 * pa.h (ASM_OUTPUT_LABEL): Output colon when using GAS.
1789 (ASM_OUTPUT_INTERNAL_LABEL): Define.
1791 2007-01-11 Zdenek Dvorak <dvorakz@suse.cz>
1793 * tree-ssa-loop-ivopts.c (extract_cond_operands): Split from
1794 find_interesting_uses_cond.
1795 (find_interesting_uses_cond): Use extract_cond_operands.
1796 (rewrite_use_compare): Use extract_cond_operands and
1797 force_gimple_operand_bsi. Do not call update_stmt.
1798 (determine_use_iv_cost_condition): Use extract_cond_operands.
1799 Return cheaper of using original bound and changing the exit bound.
1801 2007-01-11 Zdenek Dvorak <dvorakz@suse.cz>
1803 PR tree-optimization/29516
1804 * tree-ssa-address.c (tree_mem_ref_addr, add_to_parts,
1805 most_expensive_mult_to_index, addr_to_parts,
1806 create_mem_ref, maybe_fold_tmr): Make the type of
1807 fields of TARGET_MEM_REF sizetype.
1808 (move_fixed_address_to_symbol, move_pointer_to_base):
1809 New functions.
1810 * tree.def (TARGET_MEM_REF): Add comment on types of
1811 the operands.
1813 2007-01-11 Joseph Myers <joseph@codesourcery.com>
1815 * c-common.c (vector_types_convertible_p): Treat opaque types as
1816 always convertible if they have the same size, but not otherwise.
1818 2007-01-11 Steven Bosscher <steven@gcc.gnu.org>
1820 * ifcvt.c (struct noce_if_info): Add comments to the fields.
1821 Remove the b_unconditional field.
1822 (noce_try_sign_mask): Do not look at b_unconditional.
1823 (noce_process_if_block): Do not use merge_if_blocks. Update
1824 the CFG here. Do not set b_unconditional.
1825 (cond_move_process_if_block): Likewise.
1826 (find_cond_trap): Likewise.
1827 (check_cond_move_block): Require simple jump insns at the end
1828 of the basic block.
1830 2007-01-11 Jan Hubicka <jh@suse.cz>
1832 PR tree-optimization/1046
1833 * tree-tailcall.c (suitable_for_tail_call_opt_p): Use TREE_ADDRESSABLE
1834 when alias info is not ready.
1835 (pass_tail_recursion): Do not require aliasing.
1836 * tree-ssa-copyrename.c (pass_rename_ssa_cop): Likewise.
1837 * tree-ssa-ccp.c (pass_ccp, pass_fold_builtins): Likewise.
1838 * tree-ssa-copy.c (pass_copy_prop): Likewise.
1839 * tree-ssa-forwprop.c (pass_forwprop): Likewise.
1840 * tree-ssa-dce.c (pass_dce, pass_dce_loop, pass_cd_dce): Likewise.
1841 * passes.c (init_optimization_passes): Execute rename_ssa_copies,
1842 ccp, forwprop, copy_prop, merge_phi, copy_prop, dce and tail recursion
1843 before inlining.
1844 * tree-ssa-operands.c (add_virtual_operand, get_indirect_ref_operand):
1845 When aliasing is not build, mark statement as volatile.
1847 2007-01-11 Tom Tromey <tromey@redhat.com>
1849 PR preprocessor/15185, PR preprocessor/20989:
1850 * doc/cppopts.texi <-MT>: Update description of algorithm for
1851 computing default target.
1852 <-M, -MD>: Reword "basename" text.
1854 2007-01-11 Roger Sayle <roger@eyesopen.com>
1856 * builtins.c (expand_builtin_pow, expand_builtin_powi,
1857 fold_builtin_cabs, fold_builtin_sqrt, fold_builtin_trunc,
1858 fold_builtin_floor, fold_builtin_ceil, fold_builtin_round,
1859 fold_builtin_int_int_roundingfn, fold_builtin_bitop,
1860 fold_builtin_bswap, real_constp, fold_builtin_pow,
1861 fold_builtin_powi, fold_builtin_signbit, fold_builtin_copysign,
1862 do_mpfr_arg1, do_mpfr_arg2, do_mpfr_arg3, do_mpfr_sincos): Replace
1863 uses of the macro TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
1864 * convert.c (convert_to_pointer): Likewise.
1865 * expr.c (highest_pow2_factor, expand_expr_real_1): Likewise.
1866 * fold-const.c (force_fit_type, fold_negate_expr, int_const_binop,
1867 const_binop, fold_convert_const_int_from_int,
1868 fold_convert_const_int_from_real,
1869 fold_convert_const_real_from_real, sign_bit_p,
1870 optimize_minmax_comparison, extract_muldiv_1, fold_div_compare,
1871 fold_sign_changed_comparison, fold_unary, fold_comparison,
1872 fold_binary, multiple_of_p, tree_Expr_non_zero_p,
1873 fold_negate_const, fold_abs_const, fold_not_const): Likewise.
1874 * print-tree.c (print_node_brief, print_node): Likewise.
1875 * stor-layout.c (place_field, layout_type): Likewise.
1876 * tree-chrec.c (keep_cast): Likewise.
1877 * tree.c (build_vector, build_real, build_real_from_int_cst,
1878 build_complex): Likewise.
1880 2007-01-11 Roger Sayle <roger@eyesopen.com>
1882 * tree.h (TREE_CONSTANT_OVERFLOW): Obsolete. For the time being,
1883 treat TREE_CONSTANT_OVERFLOW as a synonym of TREE_OVERFLOW.
1885 2007-01-11 Paolo Bonzini <bonzini@gnu.org>
1887 * configure.ac (strict1_warn): Rename to strict_warn.
1888 (WERROR, --enable-werror, symlink hacks, stage1_cflags,
1889 cc_set_by_configure, quoted_cc_set_by_configure,
1890 stage_prefix_set_by_configure, quoted_stage_prefix_set_by_configure,
1891 all_boot_languages, all_stagestuff): Remove.
1892 (target_list): Remove bootstrap targets.
1893 * Makefile.in (quickstrap): Unconditionally make a synonym of all.
1894 (BOOT_LANGUAGES, STAGE1_CFLAGS, STAGE1_CHECKING,
1895 REMAKEFLAGS, FLAGS_TO_PASS, PREPEND_DOTDOT_TO_RELATIVE_PATHS,
1896 SUBDIR_FLAGS_TO_PASS, WERROR_FLAGS, STRICT2_WARN, LANG_STAGESTUFF,
1897 VOL_FILES, POSTSTAGE1_FLAGS_TO_PASS, STAGE2_FLAGS_TO_PASS,
1898 STAGEPROFILE_FLAGS_TO_PASS, STAGEFEEDBACK_FLAGS_TO_PASS, stage1_build,
1899 stage1_copy, stage2_build, stage2_copy, stageprofile_build,
1900 stageprofile_copy, stage3_build, stage3_copy, stagefeedback_build,
1901 stagefeedback_copy, stage4_build, clean_s1, clean_sw, bootstrap,
1902 bootstrap-lean, bootstrap2, bootstrap2-lean, bootstrap3,
1903 bootstrap3-lean, bootstrap4, bootstrap4-lean, unstage1, unstage2,
1904 unstage3, unstage4, unstageprofile, unstagefeedback, restage, restage2,
1905 restage3, restage4, restageprofile, restagefeedback, bubbleestrap,
1906 cleanstrap, unstrap, restrap, *compare, *compare3, *compare4,
1907 *compare-lean, *compare3-lean, *compare4-lean, stage1-start, stage1,
1908 stage2-start, stage2, stage3-start, stage3, stage4-start, stage4,
1909 stageprofile-start, stageprofile, stagefeedback-start, stagefeedback,
1910 risky-stage1, risky-stage2, risky-stage3, risky-stage4): Remove.
1911 (ORDINARY_FLAGS_TO_PASS): Rename to FLAGS_TO_PASS.
1912 (STAGECOPYSTUFF, STAGEMOVESTUFF): Consolidate into MOSTLYCLEANFILES.
1913 (mostlyclean): Adjust.
1914 (clean, distclean): Don't mention bootstrap stuff.
1915 * configure: Regenerate.
1916 * ada/config-lang.in, cp/config-lang.in, forttran/config-lang.in,
1917 java/config-lang.in, objc/config-lang.in, objcp/config-lang.in,
1918 treelang/config-lang.in (stagestuff): Remove.
1919 * doc/sourcebuild.texi (stage1, stage2, stage3, stage4,
1920 stageprofile, stagefeedback, stagestuff): Remove mention.
1922 2007-01-11 Nick Clifton <nickc@redhat.com>
1924 * config/mcore/predicates.md (mcore_general_movesrc_operand):
1925 Accept CONSTs.
1926 (mcore_general_movdst_operand): Do not accept CONST_INTs.
1927 (mcore_arith_K_S_operand): Run the test for the S constraint not
1928 the test for the M constraint.
1929 (mcore_addsub_operand): Do not accept integer values that are
1930 larger than 32 bits.
1931 * config/mcore/mcore.md: Remove unused constraints from split.
1932 (andsi3): Use HOST_WIDE_INT instead of int to hold an INTVAL.
1933 (addsi3): Likewise.
1934 (allocate_stack): Likewise.
1935 * config/mcore/mcore.c (mcore_print_operand): Restrict output of P
1936 operands to 32 bits.
1937 (mcore_const_costs): Use HOST_WIDE_INT instead of int to hold an
1938 INTVAL.
1939 (mcore_and_cost, mcore_modify_comparison, const_ok_for_mcore,
1940 mcore_const_ok_for_inline, mcore_const_trick_uses_not,
1941 try_constant_tricks, mcore_num_ones, mcore_num_zeros,
1942 mcore_output_bclri, mcore_output_andn, output_inline_const,
1943 mcore_output_move, mcore_output_movedouble): Likewise.
1944 (mcore_output_cmov): Use CONST_OK_FOR_M and CONST_OK_FOR_N.
1945 (output_inline_const): Likewise.
1946 (output_inline_const): Fix format strings used in sprintf
1947 statements.
1948 * config/mcore/mcore-protos.h: Update prototypes for changed
1949 functions in mcore.c.
1950 * config/mcore/mcore.h (CONST_OK_FOR_I): Cast values to
1951 HOST_WIDE_INT and not int.
1952 (CONST_OK_FOR_J, CONST_OK_FOR_K, CONST_OK_FOR_L, CONST_OK_FOR_M,
1953 CONST_OK_FOR_N): Likewise.
1954 (LEGITIMATE_CONSTANT_P): Also check CONSTANT_P.
1955 (GO_IF_LEGITIMATE_INDEX): Use HOST_WIDE_INT instead of int to hold
1956 an INTVAL.
1958 2007-01-10 Jan Hubicka <jh@suse.cz>
1960 * tree-vrp.c (remove_range_assertions): Release defs.
1961 * tree-ssa-loop-ivopts.c (rmeove_statement): Likewise.
1962 * tree-ssa-dom.c (remove_stmt_or_phi): Likewise.
1964 2007-01-10 Paul Brook <paul@codesourcery.com>
1966 * config/arm/arm.c (arm_rtx_costs_1): Handle mutiply-subtract.
1967 * config/arm/arm.md (mulsi3subsi): New insn.
1969 2007-01-10 Zdenek Dvorak <dvorakz@suse.cz>
1971 * tree-ssa-loop-manip.c (tree_unroll_loop): Make it a wrapper over ...
1972 (tree_transform_and_unroll_loop): New.
1973 * tree-flow.h (transform_callback, tree_transform_and_unroll_loop):
1974 Declare.
1976 2007-01-10 Robert Kennedy <jimbob@google.com>
1978 * fold-const.c (fold_comparison): Fold comparisons like (x *
1979 1000 < 0) to (x < 0).
1981 2007-01-10 Ian Lance Taylor <iant@google.com>
1983 * tree-pretty-print.c (dump_generic_node): Print parentheses when
1984 operands have the same priority.
1986 2007-01-10 Tom Tromey <tromey@redhat.com>
1988 * fold-const.c (fold_truthop): Don't check can_use_bit_fields_p.
1989 (fold_binary): Likewise.
1990 * langhooks.c (lhd_can_use_bit_fields_p): Removed.
1991 * langhooks-def.h (lhd_can_use_bit_fields_p): Removed.
1992 (LANG_HOOKS_CAN_USE_BIT_FIELDS_P): Removed.
1993 (LANG_HOOKS_INITIALIZER): Remove LANG_HOOKS_CAN_USE_BIT_FIELDS_P.
1994 * langhooks.h (struct lang_hooks): Removed field
1995 'can_use_bit_fields_p'.
1997 2007-01-10 Ralf Corsépius <ralf.corsepius@rtems.org>
1999 * config/bfin/t-bfin, config/bfin/t-bfin-elf: Remove GCC_CFLAGS.
2001 2007-01-10 Razya Ladelsky <razya@il.ibm.com>
2003 * function.c (get_last_funcdef_no): New function.
2004 * function.h (get_last_funcdef_no): Declare.
2005 * tree-inline.c (initialize_cfun): Add initialization.
2006 (tree_function_versioning): Cleanup.
2008 2007-01-10 Jan Hubicka <jh@suse.cz>
2010 * tree-inline.c (setup_one_parameter): Do not propagate into abnormal
2011 PHIs.
2013 2007-01-10 Sa Liu <saliu@de.ibm.com>
2014 Ben Elliston <bje@au.ibm.com>
2016 * spu.h (STACK_SAVE_AREA): Use VOIDmode for SAVE_FUNCTION, SImode
2017 for SAVE_NONLOCAL and Pmode for any other save level.
2018 * spu-protos.h (spu_restore_stack_block): Declare.
2019 * spu.md (save_stack_block): Remove.
2020 (restore_stack_block): Call spu_restore_stack_block.
2021 * spu.c (spu_restore_stack_block): New function.
2022 (spu_expand_epilogue): Remove old comment.
2024 2007-01-09 Zdenek Dvorak <dvorakz@suse.cz>
2026 PR tree-optimization/30322
2027 * tree-ssa-loop-ivopts.c (fold_affine_expr, iv_value): Removed.
2028 (cand_value_at): Return the value as aff_tree.
2029 (may_eliminate_iv): Convert the bound from aff_tree to tree.
2030 * tree-affine.c (aff_combination_add_cst, aff_combination_add_product,
2031 aff_combination_mult): New functions.
2032 (aff_combination_add): Use aff_combination_add_cst.
2033 (aff_combination_convert): Allow conversions to a wider type.
2034 (tree_to_aff_combination): Handle BIT_NOT_EXPR.
2035 * tree-affine.h (aff_combination_mult): Declare.
2037 2007-01-09 Carlos O'Donell <carlos@codesourcery.com>
2039 * doc/tm.texi: Update documentation to reflect reality of exec
2040 and start file search behaviours. Update copyright year.
2041 * doc/invoke.texi: Explain how GCC_EXEC_PREFIX is used to find
2042 header file directories.
2044 2007-01-09 Uros Bizjak <ubizjak@gmail.com>
2046 * config/i386/i386.md (*sinxf2): Rename to *sinxf2_i387.
2047 (*cosxf2): Rename to cosxf2_i387.
2048 (*sindf2, *sinsf2): Extend operand 1 to XFmode. Macroize patterns
2049 using X87MODEF12 mode macro. Rename patterns to
2050 *sin_extend<mode>xf2_i387. Use SSE_FLOAT_MODE_P to disable patterns
2051 for SSE math.
2052 (*cosdf2, *cossf2): Ditto.
2053 (sincosdf3, sincossf3): Ditto. Rewrite corresponding splitters
2054 to match extended input operands.
2055 (sincos<mode>3): New expander.
2056 (*sinextendsfdf2, *cosextendsfdf2, *sincosextendsfdf3): Remove
2057 insn patterns and corresponding splitters.
2059 2007-01-09 Kaz Kojima <kkojima@gcc.gnu.org>
2061 * config/sh/t-linux (TARGET_LIBGCC2_CFLAGS): Delete.
2062 (SHLIB_MAPFILES, SHLIB_LINK, SHLIB_INSTALL): Likewise.
2064 2007-01-09 Nicolas Pitre <nico@cam.org>
2066 PR target/30173
2067 * arm/ieee754-df.S (Lad_s): Also test the low word of X for zero.
2069 2007-01-08 Geoffrey Keating <geoffk@apple.com>
2071 * target.h (struct gcc_target): New field library_rtti_comdat.
2072 * target-def.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): New.
2073 (TARGET_CXX): Add TARGET_CXX_LIBRARY_RTTI_COMDAT.
2074 * doc/tm.texi (C++ ABI): Document TARGET_CXX_LIBRARY_RTTI_COMDAT.
2075 * config/darwin.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
2077 2007-01-08 Geoffrey Keating <geoffk@apple.com>
2079 * doc/invoke.texi (Optimize Options): Correct description of -O0.
2081 2007-01-08 Richard Guenther <rguenther@suse.de>
2083 * tree.h (force_fit_type_double): Export.
2084 (force_fit_type): Remove.
2085 * fold-const.c (force_fit_type_double): New function.
2086 (force_fit_type): Remove.
2087 (int_const_binop): Use it.
2088 (fold_convert_const_int_from_int): Likewise.
2089 (fold_convert_const_int_from_real): Likewise.
2090 (fold_div_compare): Likewise.
2091 (fold_sign_changed_comparison): Likewise.
2092 (fold_unary): Likewise.
2093 (fold_negate_const): Likewise.
2094 (fold_abs_const): Likewise.
2095 (fold_not_const): Likewise.
2096 * c-common.c (shorten_compare): Use force_fit_type_double.
2097 * convert.c (convert_to_pointer): Likewise.
2099 2007-01-08 Richard Guenther <rguenther@suse.de>
2101 * tree.h (build_int_cst_wide_type): Export.
2102 * tree.c (build_int_cst_wide_type): New function.
2103 (build_int_cst_wide): Fix comment.
2104 * builtins.c (fold_builtin_object_size): Use build_int_cst
2105 to build -1 or 0 of the correct type. Use fit_double_type
2106 to check for overflow.
2107 * fold-const.c (optimize_bit_field_compare): Use build_int_cst_type
2108 to build the mask.
2109 (decode_field_reference): Likewise.
2110 (all_ones_mask_p): Likewise.
2111 (native_interpret_int): Use build_int_cst_wide_type.
2112 (fold_binary): Use build_int_cst_type to build an all-ones
2113 value.
2114 * stor-layout.c (set_sizetype): Use build_int_cst_wide_type.
2116 2007-01-08 Daniel Jacobowitz <dan@codesourcery.com>
2118 * config/pa/t-pa64 (libgcc_stub.a): Use $(T).
2120 2007-01-09 Ben Elliston <bje@au.ibm.com>
2122 * genautomata.c (STATS_OPTION): New option.
2123 (stats_flag): New flag.
2124 (gen_automata_option): Handle it.
2125 (initiate_automaton_gen): Ditto.
2126 (write_automata): Output statistics only if stats_flag is
2127 set. Likewise, output time statistics only if time_flag is set.
2128 * doc/md.texi (Processor pipeline description): Document new flag.
2130 2007-01-08 Richard Guenther <rguenther@suse.de>
2132 * builtins.c (fold_builtin_int_roundingfn): Use fit_double_type.
2133 * tree.c (build_int_cst_type): Likewise.
2134 (size_in_bytes): Don't call force_fit_type on the result.
2135 (int_fits_type_p): Use fit_double_type.
2136 * fold-const.c (fit_double_type): New function.
2137 (force_fit_type): Use it.
2138 * tree.h (fit_double_type): Export.
2140 2007-01-08 Jan Hubicka <jh@suse.cz>
2142 * tree-vectorizer.c (gate_increase_alignment): Fix return type.
2143 * ipa.c (function_and_variable_visibility): Fix return type.
2145 2007-01-08 Richard Guenther <rguenther@suse.de>
2147 * tree-ssa-ccp.c (maybe_fold_offset_to_array_ref): Use type
2148 of offset to build the index.
2149 * tree-pretty-print.c (dump_generic_node): Don't build negated
2150 const just for printing.
2151 * c-pretty-print.c (pp_c_integer_constant): Likewise.
2152 * builtins.c (fold_builtin_int_roundingfn): Check if result
2153 fits the type by using force_fit_type and comparing the result.
2154 * predict.c (predict_loops): Use compare_tree_int for comparison.
2155 * tree.c (build_int_cst): Fall back to integer_type_node for
2156 NULL_TREE type.
2157 (build_int_cst_wide): Assert type is non-null.
2159 2007-01-08 Roberto Costa <roberto.costa@st.com>
2161 * tree-vrp.c (extract_range_from_cond_expr): New.
2162 (extract_range_from_expr): Handle COND_EXPR nodes used as expressions.
2163 * tree-ssa-ccp.c (get_maxval_strlen): Handle COND_EXPR nodes used
2164 as expressions.
2165 (fold_stmt): Bug fix, avoid infinite recursion when folding COND_EXPRs.
2166 * tree-ssa-forwprop.c (simplify_cond, forward_propagate_into_cond,
2167 tree_ssa_forward_propagate_single_use_vars): Handle COND_EXPR nodes
2168 used as expressions.
2169 * tree-object-size.c (cond_expr_object_size): New.
2170 (collect_object_sizes_for): Handle COND_EXPR nodes used as expressions.
2172 2007-01-08 Jan Hubicka <jh@suse.cz>
2174 * tree-ssa-forwprop.c (forward_propagate_into_cond,
2175 tree_ssa_forward_propagate_single_use_va): Release defs of propagated
2176 statement.
2178 2007-01-08 Richard Guenther <rguenther@suse.de>
2180 PR tree-optimization/23603
2181 * tree-vrp.c (set_value_range_to_truthvalue): New function.
2182 (extract_range_from_binary): Fall back to truthvalue instead of
2183 varying for TRUTH_*_EXPR.
2184 (extract_range_from_comparison): Fall back to truthvalue instead of
2185 varying.
2186 (vrp_visit_phi_node): Don't adjust new range bounds to +INF/-INF
2187 if all visited PHI values were constant.
2189 2007-01-08 Jan Hubicka <jh@suse.cz>
2191 * cgraphunit.c (cgraph_process_new_functions): Reset reachable flag.
2192 (cgraph_analyze_function): break out from ...
2193 (cgraph_finalize_compilation_unit): ... here.
2194 (cgraph_expand_function): Remove forgoten commented out line.
2195 (cgraph_optimize): Analyze functions.
2197 2007-01-08 Jan Hubicka <jh@suse.cz>
2199 * tree-pas.h (TODO_remove_function): New flag.
2200 (TODO_update*): Renumber.
2201 (pass_ipa_increase_alignment,
2202 pass_ipa_function_and_variable_visibility): New passes.
2203 * cgraphunit.c (cgraph_increase_alignment): Move to tree-vectorizer.c
2204 (cgraph_function_and_variable_visibility): Move to ipa.c
2205 (cgraph_optimize): Don't call cgraph_function_and_variable_visibility,
2206 cgraph_increase_alignment.
2207 * ipa-inline.c (cgraph_decide_inlining): Don't push timevar.
2208 (cgraph_decide_inlining_incrementally): Push TV_INTEGRATION before
2209 calling tree-inline.
2210 (cgraph_early_inlining): Do not call cgraph_remove_unreachable_nodes.
2211 (pass_ipa_inline, pass_early_ipa_inlining): Set TODO_remove_functions
2212 * tree-vectorizer.c (increase_alignment): Move here from cgraphunit.c
2213 (gate_increase_alignment): New function.
2214 (pass_ipa_increase_alignment): New pass.
2215 * ipa.c: Inline tree-pass.h and timevar.h
2216 (function_and_variable_visibility): Move here from cgraphunit.c
2217 * tree-optimize.c (pass_early_local_passes): Add TODO_remove_functions.
2218 * passes.c (init_optimization_passes): Add the two new passes.
2219 (execute_todo): Handle cgraph_remove_functions.
2221 2007-01-08 Nick Clifton <nickc@redhat.com>
2223 * config/frv/predicates.md (reg_or_0_operand): Accept
2224 CONST_DOUBLEs.
2226 2007-01-08 Ralf Corsépius <ralf.corsepius@rtems.org>
2228 * config/bfin/rtems.h, config/bfin/t-rtems: New.
2229 * config.gcc: Add bfin*-rtems*.
2231 2007-01-08 Mark Shinwell <shinwell@codesourcery.com>
2233 * c.opt: Add -flax-vector-conversions.
2234 * c-typeck.c (convert_for_assignment): Pass flag to
2235 vector_types_convertible_p to allow emission of note.
2236 (digest_init): Likewise.
2237 * c-opts.c: Handle -flax-vector-conversions.
2238 * c-common.c (flag_lax_vector_conversions): New.
2239 (vector_types_convertible_p): Unless -flax-vector conversions
2240 has been passed, disallow conversions between vectors with
2241 differing numbers of subparts and/or element types. If such
2242 a conversion is disallowed, possibly emit a note on the first
2243 occasion only to inform the user of -flax-vector-conversions.
2244 The new last argument specifies this.
2245 * c-common.h (flag_lax_vector_conversions): New.
2246 (vector_types_convertible_p): Add extra argument.
2247 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Use
2248 char_type_node for V*QI type vectors.
2249 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins):
2250 Update to satisfy new typechecking rules.
2251 * config/rs6000/altivec.h (vec_cmple): Use vec_cmpge, for both
2252 C and C++ variants.
2253 * doc/invoke.texi (C Dialect Options): Document
2254 -flax-vector-conversions.
2256 2007-01-08 Mark Shinwell <shinwell@codesourcery.com>
2258 PR tree-optimization/29877
2259 * tree-ssa-ter.c (is_replaceable_p): Deem assignments with
2260 a register variable on the RHS to not be replaceable.
2262 2007-01-08 Chen Liqin <liqin@sunnorth.com.cn>
2263 * config/score/t-score-elf (MULTILIB_OPTIONS): Change.
2264 * config/score/predicates.md (const_uimm5, sr0_operand, const_simm12,
2265 const_simm15, const_pow2, const_npow2): Added.
2266 * config/score/misc.md (insv, extv, extzv, movmemsi,
2267 move_lbu_a/b, mov_lhu_a/b etc): Added and fix some bug.
2268 * config/score/score.c (score_address_cost, score_select_cc_mode):
2269 Added.
2270 Change CONST_OK_FOR_LETTER_P/EXTRA_CONSTRAINT define.
2271 Update score_rtx_costs for MACRO TARGET_RTX_COSTS.
2272 Update score_print_operand.
2273 * config/score/score.h (DATA_ALIGNMENT, SELECT_CC_MODE): Added.
2274 Adjust register allocate order and update some macro define.
2275 * config/score/score-mdaux.c (mdx_unaligned_load, mdx_unsigned_store,
2276 mdx_block_move_straight, mdx_block_move_loop_head,
2277 mdx_block_move_loop_body, mdx_block_move_loop_foot, mdx_block_move_loop,
2278 mdx_block_move): Added.
2279 (mdx_movsicc, mdp_select_add_imm, mdp_select, mds_zero_extract_andi,
2280 mdp_limm): Updated and fix some bug and typo.
2281 * config/score/score.md (movqi/hi/si, add/sub/zero/ext): Updated.
2282 (movsf, movdf, doloop_end): Added.
2284 2007-01-08 Kazu Hirata <kazu@codesourcery.com>
2286 * config/arm/arm.c, config/arm/arm.h, config/arm/arm.md,
2287 config/arm/thumb2.md: Fix comment typos.
2288 * doc/extend.texi: Fix a typo.
2290 2007-01-07 Eric Christopher <echristo@apple.com>
2292 * configure.ac: Check for __stack_chk_fail for darwin.
2293 * configure: Regenerate.
2295 2007-01-07 Richard Guenther <rguenther@suse.de>
2297 * tree-vrp.c (extract_range_from_assert): CSE calls to
2298 compare_values where possible.
2299 (extract_range_from_unary_expr): Likewise.
2301 2007-01-07 Anatoly Sokolov <aesok@post.ru>
2303 * config/avr/avr-protos.h (call_insn_operand): Delete prototype.
2304 * config/avr/avr.c (call_insn_operand): Delete function.
2305 * config/avr/avr.md (*pushqi, *pushhi, *pushsi, *pushsf): Use REG_SP
2306 instead of register number. Use predicates.
2307 * config/avr/predicates.md (const0_operand, reg_or_0_operand,
2308 call_insn_operand): Add.
2310 2007-01-06 Jan Hubicka <jh@suse.cz>
2312 * tree-pass.h (pass_build_cgraph_edges): Declare.
2313 * cgraphunit.c (record_refernece): Move to cgraphbuild.c
2314 (visited_nodes): Remove.
2315 (cgraph_create_edges): Move to cgraphbuild.c; rename to
2316 build_cgrpah_edges; make visited_nodes local.
2317 (cgraph_process_new_functions): DO not call initialize_inline_failed.
2318 (record_references_in_initializer): Move to cgraphbuild.c
2319 (initialize_inline_failed, rebuild_cgraph_edges,
2320 pass_rebuild_cgraph_edges): Move to cgraphbuild.c.
2321 (verify_cgraph_node): Make visited_nodes local.
2322 (cgraph_analyze_function): Do not call cgraph_create_edges and
2323 initialize_inline_failed.
2324 (cgraph_expand_function): Do not call cgraph_lower_function;
2325 assert that function is already lowered.
2326 * Makefile.in (cgraphbuild.o): New.
2327 * passes.c (init_optimization_passes): Add pass_build_cgraph_edges
2328 at the end of lowering passes.
2330 2007-01-06 Steven Bosscher <steven@gcc.gnu.org>
2332 * ifcvt.c (cond_move_convert_if_block): New function, code
2333 factored out from...
2334 (cond_move_process_if_block): ...here. Call the new function
2335 on the THEN and ELSE blocks.
2336 (merge_if_block): Do not copy global_live_at_end, merge_blocks
2337 already takes care of this.
2339 2007-01-05 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
2341 PR c/19978
2342 * tree.h (TREE_OVERFLOW_P): New.
2343 * c-typeck.c (parser_build_unary_op): Warn only if result
2344 overflowed and operands did not.
2345 (parser_build_binary_op): Likewise.
2346 (convert_for_assignment): Remove redundant overflow_warning.
2347 * c-common.c (overflow_warning): Don't check or set TREE_OVERFLOW.
2349 2007-01-05 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
2351 * c-typeck.c (store_init_value): Split over two lines to follow
2352 the GNU coding style.
2354 2007-01-05 Benjamin Kosnik <bkoz@redhat.com>
2356 * c-cppbuiltin.c (c_cpp_builtins): __GXX_EXPERIMENTAL_CPP0X__ to
2357 __GXX_EXPERIMENTAL_CXX0X__.
2358 * doc/cpp.texi: Same.
2360 2007-01-05 Richard Guenther <rguenther@suse.de>
2362 PR middle-end/27826
2363 * tree.c (get_narrower): Do not construct COMPONENT_REFs
2364 with mismatched types. Instead explicitly build a
2365 conversion NOP_EXPR.
2367 2007-01-05 Ian Lance Taylor <iant@google.com>
2369 * c-common.c (decl_with_nonnull_addr_p): New function.
2370 (c_common_truthvalue_conversion): Call it.
2371 * c-typeck.c (build_binary_op): Likewise.
2372 * c-common.h (decl_with_nonnull_addr_p): Declare.
2374 2007-01-05 Jakub Jelinek <jakub@redhat.com>
2376 PR c/30360
2377 * libgcc2.c (__divdc3): Compare c and d against 0.0 instead of
2378 denom against 0.0.
2380 2007-01-05 Joel Brobecker <brobecker@adacore.com>
2382 * doc/install.texi (Final install): Document the fact that
2383 the GNAT runtime should not be stripped.
2385 2007-01-04 Jan Hubicka <jh@suse.cz>
2387 * tree-inline.c (fold_marked_statements): Update operand caches
2388 and EH after folding
2390 2007-01-04 Ian Lance Taylor <iant@google.com>
2392 * c-common.c (check_function_nonnull): Whitespace fix.
2394 2007-01-04 Jan Hubicka <jh@suse.cz>
2396 * tree-optimize.c (execute_fixup_cfg): Correct previously mistakely
2397 comitted older version of patch.
2398 (pass_fixup_cfg): Add TODOs to verify flow and statements, dump
2399 function, celanup cfg and collect garbage.
2401 2007-01-04 Mike Stump <mrs@apple.com>
2403 * Makefile.in (mostlyclean): Don't remove libgcc anymore.
2404 (clean): Likewise.
2406 2007-01-04 Eric Christopher <echristo@apple.com>
2408 * libgcc2.c (__bswapsi2): Use SItype.
2409 (__bswapdi2): Use DItype.
2410 * libgcc2.h: Update for above.
2412 2007-01-04 Paul Brook <paul@codesourcery.com>
2414 * config/arm/arm.md (arm_mulsi3, thumb_mulsi3, mulsi3_compare0,
2415 mulsi_compare0_scratch, mulsi3addsi, mulsi3addsi_compare0,
2416 mulsi3addsi_compare0_scratch, mulsidi3adddi, mulsidi3,
2417 umulsidi3, umulsidi3adddi, smulsi3_highpart,
2418 umulsi3_highpart): Make conditional on !arm_arch6.
2419 (arm_mulsi3_v6, thumb_mulsi3_v6, mulsi3_compare0_v6,
2420 mulsi_compare0_scratch_v6, mulsi3addsi_v6, mulsi3addsi_compare0_v6,
2421 mulsi3addsi_compare0_scratch_v6, mulsidi3adddi_v6, mulsidi3_v6,
2422 umulsidi3_v6, umulsidi3adddi_v6, smulsi3_highpart_v6,
2423 umulsi3_highpart_v6): New insns.
2425 2007-01-04 Roger Sayle <roger@eyesopen.com>
2427 * fold-const.c (fold_convert): When casting an expression to void,
2428 fold_ignored_result may discover a GIMPLE_MODIFY_STMT which doesn't
2429 have a type. Instead of attempting to build a NOP_EXPR, return
2430 these "special" trees directly.
2432 2007-01-04 Joseph Myers <joseph@codesourcery.com>
2434 * config/rs6000/rs6000.c (rs6000_rtx_costs): Make adjustment for
2435 MULT inside MINUS as either argument. Use rs6000_cost->dmul -
2436 rs6000_cost->fp not 0 as adjustment for outer NEG.
2438 2007-01-04 Jan Hubicka <jh@suse.cz>
2440 * cgraph.c (cgraph_release_function_body): New function.
2441 (cgraph_remove_node): Use it.
2442 * cgraph.h (cgraph_release_function_body): Declare.
2443 * cgraphunit.c (cgraph_expand_function): Use it.
2444 * ipa.c (cgraph_remove_unreahchable_nodes): Use it.
2445 * tree-ssa.c (delete_tree_ssa): Allow to be called before aliasing
2446 is initialized and while compilation of other function is running.
2447 * tree-optimize.c (execute_free_cfg_annotations): Move code to clear
2448 statement CFG annotations from here to ...
2449 * tree-cfg.c (delete_tree_cfg_annotations): ... here.
2451 2007-01-04 Zdenek Dvorak <dvorakz@suse.cz>
2453 * cfgloop.h (enum li_flags): Make the constants powers of two.
2455 2007-01-04 Jan Hubicka <jh@suse.cz>
2457 * tree-inline.c (copy_bb): Insert new statements to statements_to_fold
2458 set.
2459 (fold_marked_statements): New function.
2460 (optimize_inline_calls, tree_function_versioning): Fold new statements.
2461 * tree-inline.h (copy_body_data): Add statements_to_fold.
2463 2007-01-03 Daniel Jacobowitz <dan@codesourcery.com>
2465 * config.gcc: Mention libgcc/config.host.
2466 * Makefile.in: Update comments mentioning libgcc.
2467 (LIBGCC, INSTALL_LIBGCC, GCC_PARTS, mklibgcc): Delete.
2468 (all.cross, start.encap, rest.encap, rest.cross): Update
2469 dependencies for libgcc move.
2470 (libgcc.mk, LIBGCC_DEPS, libgcov.a, libgcc.a, stmp-multilib)
2471 (clean-target, clean-target-libgcc): Delete.
2472 (srcdirify, GCC_EXTRA_PARTS): New macros.
2473 (libgcc-support, libgcc.mvars): New rules.
2474 (distclean): Remove mention of mklibgcc.
2475 (install): Don't reference INSTALL_LIBGCC.
2476 (install-common): Don't reference EXTRA_PARTS.
2477 (install-libgcc, install-multilib): Delete rules.
2478 * mklibgcc.in: Delete file.
2479 * doc/configfiles.texi: Don't mention mklibgcc.
2481 * config/i386/t-darwin (SHLIB_VERPFX): Delete (moved to libgcc).
2482 * config/i386/t-darwin64 (SHLIB_VERPFX): Likewise.
2483 * config/rs6000/t-darwin (SHLIB_VERPFX): Likewise.
2484 * config/rs6000/t-ppccomm (TARGET_LIBGCC2_CFLAGS, SHLIB_MAPFILES)
2485 (mklibgcc, ldblspecs): Likewise.
2487 * config/i386/t-nwld (libgcc.def, libc.def, libpcre.def)
2488 (posixpre.def): Use $(T).
2489 (SHLIB_EXT, SHLIB_NAME, SHLIB_SLIBDIR_QUAL, SHLIB_DEF, SHLIB_MAP)
2490 (SHLIB_SRC, SHLIB_INSTALL): Delete.
2491 (SHLIB_LINK): Make dummy.
2492 * config/t-slibgcc-darwin: Delete contents except for dummy SHLIB_LINK.
2494 * config/frv/t-linux (EXTRA_MULTILIB_PARTS): Clear.
2496 * config/alpha/t-crtfm: Use $(T) in rules for EXTRA_PARTS.
2497 * config/alpha/t-vms, config/alpha/t-vms64, config/fr30/t-fr30,
2498 config/i386/t-rtems-i386, config/ia64/t-ia64, config/rs6000/t-beos,
2499 config/rs6000/t-newas, config/sparc/t-elf: Likewise.
2501 * configure.ac (all_outputs): Remove mklibgcc.
2502 * configure: Regenerated.
2504 2007-01-03 Josh Conner <jconner@apple.com>
2506 PR middle-end/29683
2507 * calls.c (compute_argument_addresses): Set stack and stack_slot
2508 for partial args, too.
2509 (store_one_arg): Use locate.size.constant for the size when
2510 generating a save_area.
2512 2007-01-03 Robert Kennedy <jimbob@google.com>
2514 * tree-cfg.c (tree_merge_blocks): Release SSA_NAME phi results
2515 whose definitions are deleted due to basic block merging.
2517 2007-01-03 Paul Brook <paul@codesourcery.com>
2519 PR target/16634
2520 * config/arm/arm.c (output_return_instruction): Pop PC in interrupt
2521 functions.
2522 (use_return_insn): Return 0 for Thumb interrupt functions.
2523 (print_multi_reg): Add rfe argument for IRQ returns.
2524 (arm_output_epilogue): Pop interrupt return address directly into PC.
2525 (arm_expand_prologue): Only adjust IRQ return address in Arm mode.
2527 2007-01-03 Paul Brook <paul@codesourcery.com>
2529 Merge from sourcerygxx-4_1.
2530 * config/arm/thumb2.md: New file.
2531 * config/arm/elf.h (JUMP_TABLES_IN_TEXT_SECTION): Return True for
2532 Thumb-2.
2533 * config/arm/coff.h (JUMP_TABLES_IN_TEXT_SECTION): Ditto.
2534 * config/arm/aout.h (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
2535 (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump tables.
2536 * config/arm/aof.h (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump
2537 tables.
2538 (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
2539 * config/arm/ieee754-df.S: Use macros for Thumb-2/Unified asm
2540 comptibility.
2541 * config/arm/ieee754-sf.S: Ditto.
2542 * config/arm/arm.c (thumb_base_register_rtx_p): Rename...
2543 (thumb1_base_register_rtx_p): ... to this.
2544 (thumb_index_register_rtx_p): Rename...
2545 (thumb1_index_register_rtx_p): ... to this.
2546 (thumb_output_function_prologue): Rename...
2547 (thumb1_output_function_prologue): ... to this.
2548 (thumb_legitimate_address_p): Rename...
2549 (thumb1_legitimate_address_p): ... to this.
2550 (thumb_rtx_costs): Rename...
2551 (thumb1_rtx_costs): ... to this.
2552 (thumb_compute_save_reg_mask): Rename...
2553 (thumb1_compute_save_reg_mask): ... to this.
2554 (thumb_final_prescan_insn): Rename...
2555 (thumb1_final_prescan_insn): ... to this.
2556 (thumb_expand_epilogue): Rename...
2557 (thumb1_expand_epilogue): ... to this.
2558 (arm_unwind_emit_stm): Rename...
2559 (arm_unwind_emit_sequence): ... to this.
2560 (thumb2_legitimate_index_p, thumb2_legitimate_address_p,
2561 thumb1_compute_save_reg_mask, arm_dwarf_handle_frame_unspec,
2562 thumb2_index_mul_operand, output_move_vfp, arm_shift_nmem,
2563 arm_save_coproc_regs, thumb_set_frame_pointer, arm_print_condition,
2564 thumb2_final_prescan_insn, thumb2_asm_output_opcode, arm_output_shift,
2565 thumb2_output_casesi): New functions.
2566 (TARGET_DWARF_HANDLE_FRAME_UNSPEC): Define.
2567 (FL_THUMB2, FL_NOTM, FL_DIV, FL_FOR_ARCH6T2, FL_FOR_ARCH7,
2568 FL_FOR_ARCH7A, FL_FOR_ARCH7R, FL_FOR_ARCH7M, ARM_LSL_NAME,
2569 THUMB2_WORK_REGS): Define.
2570 (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv, arm_condexec_count,
2571 arm_condexec_mask, arm_condexec_masklen)): New variables.
2572 (all_architectures): Add armv6t2, armv7, armv7a, armv7r and armv7m.
2573 (arm_override_options): Check new CPU capabilities.
2574 Set new architecture flag variables.
2575 (arm_isr_value): Handle v7m interrupt functions.
2576 (user_return_insn): Return 0 for v7m interrupt functions. Handle
2577 Thumb-2.
2578 (const_ok_for_arm): Handle Thumb-2 constants.
2579 (arm_gen_constant): Ditto. Use movw when available.
2580 (arm_function_ok_for_sibcall): Return false for v7m interrupt
2581 functions.
2582 (legitimize_pic_address, arm_call_tls_get_addr): Handle Thumb-2.
2583 (thumb_find_work_register, arm_load_pic_register,
2584 legitimize_tls_address, arm_address_cost, load_multiple_sequence,
2585 emit_ldm_seq, emit_stm_seq, arm_select_cc_mode, get_jump_table_size,
2586 print_multi_reg, output_mov_long_double_fpa_from_arm,
2587 output_mov_long_double_arm_from_fpa, output_mov_double_fpa_from_arm,
2588 output_mov_double_fpa_from_arm, output_move_double,
2589 arm_compute_save_reg_mask, arm_compute_save_reg0_reg12_mask,
2590 output_return_instruction, arm_output_function_prologue,
2591 arm_output_epilogue, arm_get_frame_offsets, arm_regno_class,
2592 arm_output_mi_thunk, thumb_set_return_address): Ditto.
2593 (arm_expand_prologue): Handle Thumb-2. Use arm_save_coproc_regs.
2594 (arm_coproc_mem_operand): Allow POST_INC/PRE_DEC.
2595 (arithmetic_instr, shift_op): Use arm_shift_nmem.
2596 (arm_print_operand): Use arm_print_condition. Handle '(', ')', '.',
2597 '!' and 'L'.
2598 (arm_final_prescan_insn): Use extract_constrain_insn_cached.
2599 (thumb_expand_prologue): Use thumb_set_frame_pointer.
2600 (arm_file_start): Output directive for unified syntax.
2601 (arm_unwind_emit_set): Handle stack alignment instruction.
2602 * config/arm/lib1funcs.asm: Remove default for __ARM_ARCH__.
2603 Add v6t2, v7, v7a, v7r and v7m.
2604 (RETLDM): Add Thumb-2 code.
2605 (do_it, shift1, do_push, do_pop, COND, THUMB_SYNTAX): New macros.
2606 * config/arm/arm.h (TARGET_CPU_CPP_BUILTINS): Define __thumb2__.
2607 (TARGET_THUMB1, TARGET_32BIT, TARGET_THUMB2, TARGET_DSP_MULTIPLY,
2608 TARGET_INT_SIMD, TARGET_UNIFIED_ASM, ARM_FT_STACKALIGN, IS_STACKALIGN,
2609 THUMB2_TRAMPOLINE_TEMPLATE, TRAMPOLINE_ADJUST_ADDRESS,
2610 ASM_OUTPUT_OPCODE, THUMB2_GO_IF_LEGITIMATE_ADDRESS,
2611 THUMB2_LEGITIMIZE_ADDRESS, CASE_VECTOR_PC_RELATIVE,
2612 CASE_VECTOR_SHORTEN_MODE, ADDR_VEC_ALIGN, ASM_OUTPUT_CASE_END,
2613 ADJUST_INSN_LENGTH): Define.
2614 (TARGET_REALLY_IWMMXT, TARGET_IWMMXT_ABI, CONDITIONAL_REGISTER_USAGE,
2615 STATIC_CHAIN_REGNUM, HARD_REGNO_NREGS, INDEX_REG_CLASS,
2616 BASE_REG_CLASS, MODE_BASE_REG_CLASS, SMALL_REGISTER_CLASSES,
2617 PREFERRED_RELOAD_CLASS, SECONDARY_OUTPUT_RELOAD_CLASS,
2618 SECONDARY_INPUT_RELOAD_CLASS, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P,
2619 TRAMPOLINE_SIZE, INITIALIZE_TRAMPOLINE, HAVE_PRE_INCREMENT,
2620 HAVE_POST_DECREMENT, HAVE_PRE_DECREMENT, HAVE_PRE_MODIFY_DISP,
2621 HAVE_POST_MODIFY_DISP, HAVE_PRE_MODIFY_REG, HAVE_POST_MODIFY_REG,
2622 REGNO_MODE_OK_FOR_BASE_P, LEGITIMATE_CONSTANT_P,
2623 REG_MODE_OK_FOR_BASE_P, REG_OK_FOR_INDEX_P, GO_IF_LEGITIMATE_ADDRESS,
2624 LEGITIMIZE_ADDRESS, THUMB2_LEGITIMIZE_ADDRESS,
2625 GO_IF_MODE_DEPENDENT_ADDRESS, MEMORY_MOVE_COST, BRANCH_COST,
2626 ASM_APP_OFF, ASM_OUTPUT_CASE_LABEL, ARM_DECLARE_FUNCTION_NAME,
2627 FINAL_PRESCAN_INSN, PRINT_OPERAND_PUNCT_VALID_P,
2628 PRINT_OPERAND_ADDRESS): Adjust for Thumb-2.
2629 (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv): New declarations.
2630 * config/arm/arm-cores.def: Add arm1156t2-s, cortex-a8, cortex-r4 and
2631 cortex-m3.
2632 * config/arm/arm-tune.md: Regenerate.
2633 * config/arm/arm-protos.h: Update prototypes.
2634 * config/arm/vfp.md: Enable patterns for Thumb-2.
2635 (arm_movsi_vfp): Add movw alternative. Use output_move_vfp.
2636 (arm_movdi_vfp, movsf_vfp, movdf_vfp): Use output_move_vfp.
2637 (thumb2_movsi_vfp, thumb2_movdi_vfp, thumb2_movsf_vfp,
2638 thumb2_movdf_vfp, thumb2_movsfcc_vfp, thumb2_movdfcc_vfp): New.
2639 * config/arm/libunwind.S: Add Thumb-2 code.
2640 * config/arm/constraints.md: Update include Thumb-2.
2641 * config/arm/ieee754-sf.S: Add Thumb-2/Unified asm support.
2642 * config/arm/ieee754-df.S: Ditto.
2643 * config/arm/bpabi.S: Ditto.
2644 * config/arm/t-arm (MD_INCLUDES): Add thumb2.md.
2645 * config/arm/predicates.md (low_register_operand,
2646 low_reg_or_int_operand, thumb_16bit_operator): New.
2647 (thumb_cmp_operand, thumb_cmpneg_operand): Rename...
2648 (thumb1_cmp_operand, thumb1_cmpneg_operand): ... to this.
2649 * config/arm/t-arm-elf: Add armv7 multilib.
2650 * config/arm/arm.md: Update patterns for Thumb-2 and Unified asm.
2651 Include thumb2.md.
2652 (UNSPEC_STACK_ALIGN, ce_count): New.
2653 (arm_incscc, arm_decscc, arm_umaxsi3, arm_uminsi3,
2654 arm_zero_extendsidi2, arm_zero_extendqidi2): New
2655 insns/expanders.
2656 * config/arm/fpa.md: Update patterns for Thumb-2 and Unified asm.
2657 (thumb2_movsf_fpa, thumb2_movdf_fpa, thumb2_movxf_fpa,
2658 thumb2_movsfcc_fpa, thumb2_movdfcc_fpa): New insns.
2659 * config/arm/cirrus.md: Update patterns for Thumb-2 and Unified asm.
2660 (cirrus_thumb2_movdi, cirrus_thumb2_movsi_insn,
2661 thumb2_cirrus_movsf_hard_insn, thumb2_cirrus_movdf_hard_insn): New
2662 insns.
2663 * doc/extend.texi: Document ARMv7-M interrupt functions.
2664 * doc/invoke.texi: Document Thumb-2 new cores+architectures.
2666 2007-01-03 Jakub Jelinek <jakub@redhat.com>
2668 * unwind-dw2.c (SIGNAL_FRAME_BIT, EXTENDED_CONTEXT_BIT): Define.
2669 (struct _Unwind_Context): Rename args_size to flags, remove
2670 signal_frame field, add a new args_size field and version field.
2671 (_Unwind_IsSignalFrame, _Unwind_SetSignalFrame,
2672 _Unwind_IsExtendedContext): New inline functions.
2673 (_Unwind_GetGR, _Unwind_SetGR, _Unwind_GetGRPtr, _Unwind_SetGRPtr):
2674 Assume by_value array is only present if _Unwind_IsExtendedContext.
2675 (_Unwind_GetIPInfo, execute_cfa_program, uw_frame_state_for): Use
2676 _Unwind_IsSignalFrame.
2677 (__frame_state_for): Initialize context.flags to EXTENDED_CONTEXT_BIT.
2678 (uw_update_context_1): Use _Unwind_SetSignalFrame.
2679 (uw_init_context_1): Initialize context->flags to
2680 EXTENDED_CONTEXT_BIT.
2681 * config/rs6000/linux-unwind.h (frob_update_context): Use
2682 _Unwind_SetSignalFrame.
2684 2007-01-03 Andrew Pinski <andrew_pinski@playstation.sony.com>
2686 PR middle-end/30353
2687 * gimplify.c (gimplify_modify_expr_complex_part): Move below
2688 tree_to_gimple_tuple. Call tree_to_gimple_tuple when we need
2689 the value.
2691 2007-01-03 Kazu Hirata <kazu@codesourcery.com>
2693 * config/i386/i386.h (NON_STACK_REG_P, REGNO_OK_FOR_SIREG_P,
2694 REGNO_OK_FOR_DIREG_P, REWRITE_ADDRESS, ASM_OPERAND_LETTER,
2695 RET, AT_SP): Remove.
2696 * config/i386/i386.md (*sse_prologue_save_insn): Use return
2697 instead of RET.
2699 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
2700 * cfgbuild.c (find_basic_blocks): Likewise.
2701 * cfgrtl.c (rtl_create_basic_block): Likewise.
2702 * function.c (temp_slots_at_level): Likewise.
2703 * reg-stack.c (stack_regs_mentioned): Likewise.
2704 * regclass.c (allocate_reg_info): Likewise.
2705 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
2706 set_bb_for_stmt, move_block_to_fn): Likewise.
2707 * tree-complex.c (tree_lower_complex): Likewise.
2708 * vec.h (VEC_safe_grow_cleared): New.
2710 * cgraphunit.c, tree-ssa-alias.c: Fix comment typos.
2712 2007-01-03 Zdenek Dvorak <dvorakz@suse.cz>
2714 * loop-unswitch.c (unswitch_loop): Pass probabilities to loopify.
2715 * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Pass probabilities
2716 to loop_version.
2717 * cfgloopmanip.c (scale_loop_frequencies): Export.
2718 (loopify): Scale the frequencies by prescribed coefficients.
2719 (set_zero_probability): New function.
2720 (duplicate_loop_to_header_edge): Improve updating of frequencies.
2721 (lv_adjust_loop_entry_edge, loop_version): Set probabilities
2722 and frequencies according to arguments.
2723 * tree-ssa-loop-manip.c (tree_unroll_loop): Set probabilities
2724 correctly.
2725 * cfg.c (scale_bbs_frequencies_int): Allow scaling the frequencies up.
2726 * modulo-sched.c (sms_schedule): Set probabilities for entering
2727 versioned loop correctly.
2728 * tree-vect-transform.c (vect_transform_loop): Ditto.
2729 * cfgloop.h (loopify, loop_version): Declaration changed.
2730 (scale_loop_frequencies): Declared.
2732 2007-01-02 Jan Hubicka <jh@suse.cz>
2734 * cgraph.c: Include tree-flow.h
2735 (cgraph_add_new-function): Handle IPA_SSA mode; execute
2736 early_local_passes.
2737 * cgraph.h (enum cgraph_state): Add CGRAPH_STATE_IPA_SSA.
2738 * tree-pass.h (pass_all_early_optimizations): Declare.
2739 * cgraphunit.c (cgraph_process_new_functions): Add IPA_SSA; execute
2740 early_local_passes.
2741 (cgraph_analyze_function): Do early_local_passes.
2742 * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
2743 Do not add referenced vars.
2744 * tree-optimize.c (gate_all_optimizations): Do not execute when not in
2745 SSA form.
2746 (gate_all_early_local_passes): New gate.
2747 (pass_early_local_passes): Use new gate.
2748 (execute_early_local_optimizations): New functions.
2749 (gate_all_early_optimizations): New gate.
2750 (pass_all_early_optimizations): New pass.
2751 (execute_free_datastructures): Free SSA only when initialized.
2752 (gate_init_datastructures): Init only when optimizing.
2753 (tree_lowering_passes): Do early local passes when called late.
2754 * tree-profile.c (do_tree_profiling): Don't profile functions added
2755 late.
2756 (do_early_tree_profiling, pass_early_tree_profile): Kill.
2757 * tree-cfg.c (update_modified_stmts): Do not update when operands are
2758 not active.
2759 * passes.c (init_optimizations_passes): Reorder so we go into SSA
2760 during early_local_passes.
2761 * Makefile.in (cgraph.o): Add dependency on tree-flow.h.
2764 2007-01-02 Carlos O'Donell <carlos@codesourcery.com>
2766 * Makefile.in: Update copyright year.
2768 2007-01-02 Carlos O'Donell <carlos@codesourcery.com>
2770 * Makefile.in: Export GCC_EXEC_PREFIX before calling $(RUNTEST)
2771 in $(lang_checks) and check-consistency targets.
2773 2007-01-02 Jan Hubicka <jh@suse.cz>
2775 * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
2776 Do not add referenced vars.
2777 * tree-cfg.c (update_modified_stmts): Do not update when SSA operands
2778 are not active.
2779 * passes.c (init_optimization_passes): Put mudflap_2 after
2780 free_datastructures.
2782 2007-01-02 Jan Hubicka <jh@suse.cz>
2784 * tree-optimize (execute_fixup_cfg): Set after_inlining flag.
2785 Set NOTHROW flag on call statements proved to be nothrow.
2786 Update statement of local calls so new pure/const functions are
2787 updated. Update_ssa when in ssa form. Mark PHI nodes of nonlocal
2788 goto receivers.
2789 (tree_rest_of_compilation): Register hooks and initialize bitmap
2790 early. Do not set after_inlining flag.
2792 2007-01-02 Steve Ellcey <sje@cup.hp.com>
2794 * sbitmap.c (HOST_BITS_PER_LONG_LONG): Change to
2795 HOST_BITS_PER_LONGLONG
2797 2007-01-02 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
2799 PR c/19977
2800 * c-typeck.c (store_init_value): Don't emit pedantic overflow
2801 warning for non-static initializers.
2803 2007-01-02 Steven Bosscher <steven@gcc.gnu.org>
2805 * config/alpha/alpha.md, arm/arm.c, darwin.c, frv/frv.md,
2806 m32r/m32r.c, m32r/m32r.c, mn10300/mn10300.md, pa/pa.c,
2807 rs6000/rs6000.c, s390/s390.md, sh/sh.md, sparc/sparc.c:
2808 Always use set_unique_reg_note to add REG_EQUAL notes.
2810 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
2812 Revert:
2813 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
2815 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
2816 * cfgbuild.c (find_basic_blocks): Likewise.
2817 * cfgrtl.c (rtl_create_basic_block): Likewise.
2818 * function.c (temp_slots_at_level): Likewise.
2819 * reg-stack.c (stack_regs_mentioned): Likewise.
2820 * regclass.c (allocate_reg_info): Likewise.
2821 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
2822 set_bb_for_stmt, move_block_to_fn): Likewise.
2823 * tree-complex.c (tree_lower_complex): Likewise.
2824 * vec.h (VEC_safe_grow_cleared): New.
2826 2007-01-02 Ian Lance Taylor <iant@google.com>
2828 * c-common.c (c_common_truthvalue_conversion): When warning about
2829 using an assignment as a truth value, set TREE_NO_WARNING.
2831 2007-01-02 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
2833 PR middle-end/7651
2834 * c.opt (Wold-style-declaration): New.
2835 * doc/invoke.texi (C-only Warning Options): New.
2836 (Wold-style-declaration): Document it.
2837 (Wextra): Enabled by -Wextra.
2838 * c-opts.c (c_common_post_options): Enabled by -Wextra.
2839 * c-decl.c (declspecs_add_scspec): Replace -Wextra with
2840 -Wold-style-declaration.
2842 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
2844 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
2845 * cfgbuild.c (find_basic_blocks): Likewise.
2846 * cfgrtl.c (rtl_create_basic_block): Likewise.
2847 * function.c (temp_slots_at_level): Likewise.
2848 * reg-stack.c (stack_regs_mentioned): Likewise.
2849 * regclass.c (allocate_reg_info): Likewise.
2850 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
2851 set_bb_for_stmt, move_block_to_fn): Likewise.
2852 * tree-complex.c (tree_lower_complex): Likewise.
2853 * vec.h (VEC_safe_grow_cleared): New.
2855 2007-01-02 Douglas Gregor <doug.gregor@gmail.com>
2857 * c-common.c(c_common_nodes_and_builtins): Since variants of
2858 void_type_node get built before it is given a name, we need to
2859 give those variants the name, too.
2860 (complete_array_type): We need to work with the canonical main
2861 type of the array, from which we will build the qualified version.
2862 * params.def (PARAM_VERIFY_CANONICAL_TYPES): New.
2863 * print-tree.c (print_node): Display canonical type information
2864 for each type.
2865 * stor-layout.c (layout_type): When we don't know the
2866 alignment of a type for which we're building an array, we end up
2867 guessing wrong, so make the type require structural equality.
2868 * tree.c (make_node_stat): When we build a new type, it is its
2869 own canonical type.
2870 (build_type_attribute_qual_variant): When building an attribute
2871 variant, its canonical type is the non-attribute variant. However,
2872 if the attributes are target-dependent and they differ, we need to
2873 use structural equality checks for this type.
2874 (build_qualified_type): A qualified type is not equivalent to its
2875 unqualified variant; set the canonical type appropriately.
2876 (build_distinct_type_copy): When building a distinct type from
2877 another type, the new type is its own canonical type.
2878 (build_variant_type_copy): When building a new type variant, we
2879 assume that it is equivalent to the original type.
2880 (build_pointer_type_for_mode): When building a pointer type, also
2881 build a canonical type pointer.
2882 (build_reference_type_for_mode): When building a reference type,
2883 also build a canonical type reference.
2884 (build_index_type): When we can't hash an index type (e.g.,
2885 because its maximum value is negative), the index type requires
2886 structural equality tests.
2887 (build_array_type): Build the canonical form of an array type.
2888 (build_function_type): Function types require structural equality,
2889 because they contain default arguments, attributes, etc.
2890 (build_method_type_directly): Ditto for method types.
2891 (build_offset_type): Build the canonical offset type.
2892 (build_complex_type): Build the canonical vector type.
2893 (make_vector_type): Build the canonical vector type.
2894 * tree.h (TYPE_CANONICAL): New.
2895 (TYPE_STRUCTURAL_EQUALITY_P): New.
2896 (SET_TYPE_STRUCTURAL_EQUALITY): New.
2897 (struct tree_type): Added "canonical" field.
2898 * params.h (VERIFY_CANONICAL_TYPES): New.
2899 * doc/c-tree.texi (TYPE_CANONICAL): Document.
2900 (TYPE_STRUCTURAL_EQUALITY_P): Document.
2901 (SET_TYPE_STRUCTURAL_EQUALITY): Document.
2902 * doc/invoke.texi (verify-canonical-types): Document --param
2903 parameter for verifying canonical types.
2905 2007-01-02 Joseph Myers <joseph@codesourcery.com>
2907 * config.gcc (powerpc-*-eabispe*, powerpc-*-eabisimaltivec*,
2908 powerpc-*-eabisim*, powerpc-*-eabialtivec*, powerpc-*-eabi*,
2909 powerpc-*-rtems*, powerpc-wrs-vxworks, powerpc-wrs-vxworksae,
2910 powerpcle-*-eabisim*, powerpcle-*-eabi*): Add rs6000/e500.h to
2911 tm_file.
2912 * config/rs6000/e500.h: New.
2913 * config/rs6000/eabi.h (TARGET_SPE_ABI, TARGET_SPE, TARGET_E500,
2914 TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE, TARGET_E500_DOUBLE):
2915 Remove.
2916 * config/rs6000/linuxspe.h (TARGET_SPE_ABI, TARGET_SPE,
2917 TARGET_E500, TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE,
2918 TARGET_E500_DOUBLE): Remove.
2919 * config/rs6000/vxworks.h (TARGET_SPE_ABI, TARGET_SPE,
2920 TARGET_E500, TARGET_ISEL, TARGET_FPRS): Remove.
2921 * config/rs6000/rs6000.h (CHECK_E500_OPTIONS): Define.
2922 * config/rs6000/rs6000.c (rs6000_override_options): Use
2923 CHECK_E500_OPTIONS.
2925 2007-01-02 Joseph Myers <joseph@codesourcery.com>
2927 * config/rs6000/rs6000.c (print_operand): Check (TARGET_SPE ||
2928 TARGET_E500_DOUBLE), not TARGET_E500, for %y.
2929 (rs6000_generate_compare, rs6000_emit_sCOND, output_cbranch,
2930 rs6000_emit_cmove): Don't check TARGET_E500.
2931 * config/rs6000/rs6000.md (bunordered, bordered, sunordered,
2932 sordered): Don't check TARGET_E500.
2934 2007-01-01 Eric Christopher <echristo@apple.com>
2936 * config/mips/mips.c (mips_regno_mode_ok_for_base_p): Use
2937 HARD_REGISTER_NUM_P.
2939 2007-01-01 Roger Sayle <roger@eyesopen.com>
2941 * fold-const.c (fold_binary) <EQ_EXPR>: Fold "(X^C1) eq/ne C2" into
2942 "X eq/ne (C1^C2)". Fold "(X^Z) eq/ne (Y^Z)" as "X eq/ne Y" when Z
2943 has no side-effects. Fold "(X^C1) eq/ne (Y^C2)" as "(X^(C1^C2))
2944 eq/ne Y".
2946 2007-01-01 Mike Stump <mrs@apple.com>
2948 * configure.ac: Remove support for building with Apple's gcc-3.1.
2950 2007-01-02 Joseph Myers <joseph@codesourcery.com>
2952 PR middle-end/30311
2953 * caller-save.c (add_stored_regs): Only handle SUBREGs if inner
2954 REG is a hard register. Do not modify REG before calling
2955 subreg_nregs.
2956 * rtlanal.c (subreg_get_info): Don't assert size of XMODE is a
2957 multiple of the size of YMODE for certain lowpart cases.
2959 2007-01-01 Andrew Pinski <pinskia@gmail.com>
2961 PR middle-end/30253
2962 * gimplify (voidify_wrapper_expr): Update for
2963 GIMPLE_MODIFY_STMT.
2965 2007-01-01 Andreas Schwab <schwab@suse.de>
2967 PR target/29166
2968 * config/ia64/ia64.c (ia64_compute_frame_size): Account space for
2969 save of BR0 in extra_spill_size instead of spill_size.
2970 (ia64_expand_prologue): Save BR0 outside of the gr/br/fr spill
2971 area.
2972 (ia64_expand_epilogue): Restore BR0 from its new location.
2974 2007-01-01 Andrew Pinski <pinskia@gmail.com>
2976 * gimplify.c (gimplify_init_constructor <case VECTOR_TYPE>):
2977 Use a temporary variable if the left hand side is not a gimple
2978 register.
2980 2007-01-01 Andrew Pinski <pinskia@gmail.com>
2982 * gimplify.c (gimplify_return_expr): Make the temporary variable
2983 for the return expression, a gimple register variable.
2985 2007-01-01 Jan Hubicka <jh@suse.cz>
2987 * emit-rtl.c (emit_copy_of_insn_after): Do not call copy_insn_1 for
2988 INSN_LIST.
2990 2007-01-01 Mike Stump <mrs@apple.com>
2992 * configure.ac (HAVE_GAS_LITERAL16): Add autoconf check for
2993 .literal16.
2994 * config/darwin.c (machopic_select_rtx_section): Use
2995 HAVE_GAS_LITERAL16.
2996 (darwin_mergeable_constant_section): Likewise.
2997 * configure: Regenerate.
2998 * config.in: Regenerate.
3000 2007-01-01 Jan Hubicka <jh@suse.cz>
3001 Andrew Pinski <pinskia@gmail.com>
3003 * cgraphunit.c (cgraph_optimize): Call cgraph_add_new_functions
3004 before starting IPA passes.