1 2007-01-28 Ralf Wildenhues <Ralf.Wildenhues@gmx.de>
3 * doc/sourcebuild.texi: Add comma for clarity.
4 * doc/extend.texi: Fix some typos.
5 * doc/passes.texi: Likewise.
6 * doc/cppinternals.texi: Likewise.
7 * doc/c-tree.texi: Likewise.
8 * doc/tree-ssa.texi: Likewise.
9 * doc/install.texi: Likewise.
11 2007-01-27 Jan Hubicka <jh@suse.cz>
13 * tree-sra.c (sra_walk_function): Don't rely on aliases being build.
14 (pass_sra): Do not require alias information.
15 * passes.c (init_optimization_passes): Add SRA
17 2007-01-27 Steven Bosscher <steven@gcc.gnu.org>
19 * tracer.c (rest_of_handle_tracer): We already cleaned
20 up the CFG in tracer() so don't do it here again.
21 * cfgcleanup.c (rest_of_handle_jump2): Don't repeat
22 cleanup_cfg here, either. And don't call renumber_insns.
24 * cfgrtl.c (rtl_verify_flow_info_1): Don't verify that BB_END
25 and BB_HEAD are in the insn stream here. Instead make sure
26 that BB_INSN is valid on all insns. Also, do check here that
27 there are no pending branch predictions...
28 (rtl_verify_flow_info): ...instead of doing it here. Checks
29 for BB_END and BB_HEAD moved from rtl_verify_flow_info_1 to
32 2007-01-26 Roger Sayle <roger@eyesopen.com>
34 * config/i386/i386.c (ix86_swap_binary_operands_p): New helper
35 function to simplify/factorize operand order canonicalization.
36 (ix86_fixup_binary_operands): Reorganize using the above function.
37 (ix86_binary_operator_ok): Likewise.
39 2007-01-27 Jakub Jelinek <jakub@redhat.com>
41 * genattrtab.c (struct attr_value_list, insn_code_values): Move to
42 file scope from optimize_attrs.
43 (simplify_test_exp): If insn_code_values is not NULL, use it to speed
45 (optimize_attrs): Clear insn_code_values after freeing it.
47 2007-01-26 Zdenek Dvorak <dvorakz@suse.cz>
49 * tree-ssa-address.c (create_mem_ref): Remove ", bsi" from
50 a parts.base assignment.
52 2007-01-26 Zdenek Dvorak <dvorakz@suse.cz>
54 * tree-data-ref.c (dump_subscript): Use dump_conflict_function.
55 (compute_subscript_distance, initialize_data_dependence_relation,
56 finalize_ddr_dependent, analyze_ziv_subscript,
57 analyze_siv_subscript_cst_affine,
58 compute_overlap_steps_for_affine_univar,
59 compute_overlap_steps_for_affine_1_2, analyze_subscript_affine_affine,
60 analyze_siv_subscript, analyze_miv_subscript,
61 analyze_overlapping_iterations, subscript_dependence_tester_1,
62 compute_self_dependence, free_dependence_relation): Work
63 with affine_fn instead of chrecs.
64 (dump_affine_function, dump_conflict_function, affine_function_equal_p,
65 common_affine_function, affine_function_base,
66 affine_function_constant_p, affine_fn_op, affine_fn_plus,
67 affine_fn_minus, affine_fn_free, conflict_fn_not_known,
68 conflict_fn_no_dependence, free_conflict_function, free_subscripts,
69 conflict_fn, affine_fn_cst, affine_fn_univar): New functions.
70 (all_chrecs_equal_p): Removed.
71 * tree-data-ref.h (affine_fn, conflict_function): New types.
72 (struct subscript): Change type of conflicting_iterations_in_a
73 and conflicting_iterations_in_b.
75 2007-01-26 Steve Ellcey <sje@cup.hp.com>
78 * config/pa/pa.h (TARGET_HPUX_11): New.
79 * config/pa/pa-hpux11.h (TARGET_HPUX_11): New.
80 * config/pa/pa.c (pa_init_builtins): Use TARGET_HPUX_11.
82 2007-01-26 Daniel Berlin <dberlin@dberlin.org>
83 Richard Guenther <rguenther@suse.de>
85 * tree-ssa-structalias.c (solve_graph): Handle case
86 we merged the variable to another.
88 2007-01-25 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
90 * builtins.c (fold_builtin_1): Treat ccos and ccosh as 'even'
93 * fold-const.c (negate_mathfn_p): Treat casin, casinh, catan,
94 catanh, cproj, csin, csinh, ctan and ctanh as 'odd' functions.
96 2007-01-25 DJ Delorie <dj@redhat.com>
98 * config/m32c/m32c.c (m32c_cannot_change_mode_class): We don't
99 allow changes to modes which don't fit in those registers.
101 * reload1.c (choose_reload_regs): Check for invalid subregs before
102 computing their locations, not after.
104 2007-01-25 Geoffrey Keating <geoffk@apple.com>
107 * config/rs6000/rs6000.c (first_altivec_reg_to_save): On Darwin,
108 save Altivec registers in an eh_return function.
109 (compute_vrsave_mask): Likewise.
110 (rs6000_stack_info): Correct AIX/Darwin stack alignment computation
111 for saving Altivec registers.
112 (rs6000_emit_prologue): Don't allocate stack twice in
113 eh_return function. Correct expected value of altivec_save_offset
114 when using save_world. Describe save of R0 to stack when using
115 save_world. Describe stack pointer adjustment when using
116 save_world. Remove duplicated eh_return parameter register saving.
117 Update sp_offset variable after save_world.
118 * config/rs6000/t-darwin (LIB2FUNCS_STATIC_EXTRA): Remove
120 (LIB2FUNCS_EXTRA): Add darwin-world.asm.
121 * config/rs6000/darwin.h (SUBTARGET_OVERRIDE_OPTIONS): -m64
124 2007-01-25 Steve Ellcey <sje@cup.hp.com>
126 * config.gcc (ia64*-*-hpux*): Make posix threads the default.
128 2007-01-25 Steve Ellcey <sje@cup.hp.com>
131 * config/pa/pa.c (pa_init_builtins): Set asm names for finite routines.
132 * config/ia64/ia64.c (ia64_init_builtins): Ditto.
134 2007-01-25 Richard Guenther <rguenther@suse.de>
136 * doc/invoke.texi (-Wcoverage-mismatch): Document.
137 * common.opt (-Wcoverage-mismatch): New warning option.
138 * coverage.c (get_coverage_counts): Ignore coverage mismatch
139 if -Wcoverage-mismatch is given.
141 2007-01-25 Razya Ladelsky <razya@il.ibm.com>
143 * ipa-cp.c (ipcp_insert_stage, ipcp_driver): Support for SSA.
144 (ipcp_driver): Change to static definition.
145 Add dumping of the ifunctions.
146 (constant_val_insert): Remove unused parameter. Support for SSA.
147 (ipcp_propagate_const): Support for SSA.
148 (ipcp_profile_bb_print): Print only analyzed nodes.
149 (ipcp_replace_map_create): Remove support for Fortran constant
151 * ipa-prop.c (ipa_method_modify_stmt,
152 ipa_callsite_compute_param): Support for SSA.
153 * ipa-prop.h (ipcp_driver): Remove declaration.
154 (IS_VALID_TREE_MAP_INDEX): Add define.
156 2007-01-24 Geoffrey Keating <geoffk@apple.com>
158 * unwind-dw2.c (execute_stack_op): Handle DW_OP_swap.
160 2007-01-24 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
163 * builtins.c (fold_builtin_cabs): Use MPFR to evaluate a
164 constant argument to cabs and do it without checking for
165 -funsafe-math-optimizations.
167 2007-01-24 Douglas Gregor <dgregor@osl.iu.edu>
169 * c-common.h (RID_FIRST_CXX0X): New.
170 (RID_LAST_CXX0X): New.
171 * c-opts.c (c_common_handle_option): -Wc++0x-compat is triggered
173 * c.opt (Wc++0x-compat): New.
174 * doc/invoke.texi (-Wc++0x-compat): Document.
176 2007-01-24 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
178 * builtins.c (fold_builtin_carg): New.
179 (fold_builtin_1): Use it.
181 2007-01-24 Jan Hubicka <jh@suse.cz>
183 * ipa-inline.c (cgraph_decide_inlining): Initialize initial_insns.
185 * ipa-inline.c (initial_insns, max_insns): Delete.
186 (compute_max_insns): New function.
187 (cgraph_decide_inlining_of_small_function): Use it; take minimal amount
188 of insns as base for code growth.
189 (cgraph_decide_inlining): Make initial_insns local; do not compute
191 * params.def (PARAM_INLINE_UNIT_GROWTH): Set to 60.
192 * doc/invoke.texi (inline-unit-growth): Update docs.
194 2007-01-24 Jakub Jelinek <jakub@redhat.com>
196 * config/i386/i386.h (x86_cmpxchg16b): Remove const.
197 (TARGET_CMPXCHG16B): Define to x86_cmpxchg16b.
198 * config/i386/i386.c (x86_cmpxchg16b): Remove const.
199 (override_options): Add PTA_CX16 flag. Set x86_cmpxchg16b
200 for CPUs that have PTA_CX16 set.
203 * gimplify.c (omp_check_private): New function.
204 (gimplify_scan_omp_clauses): Use it for
205 firstprivate/lastprivate/reduction.
208 * gimplify.c (omp_add_variable): Don't call omp_notice_variable
209 on TYPE_SIZE_UNIT for GOVD_LOCAL VLAs.
212 * omp-low.c (lower_omp_for_lastprivate): Add dlist argument.
213 If lower_lastprivate_clauses emits some statements, append them
214 to dlist rather than body_p and to body_p append an initializer.
215 (lower_omp_for): Adjust caller.
217 2007-01-24 Steve Ellcey <sje@cup.hp.com>
219 * target.h (globalize_decl_name): New.
220 * target-def.h (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
221 * output.h (default_globalize_decl_name): New.
222 * varasm.c (asm_output_bss): Use globalize_decl_name instead of
224 (globalize_decl): Ditto.
225 (default_globalize_decl_name): New.
226 * config/ia64/ia64.c (ia64_globalize_decl_name): New.
227 (ia64_handle_version_id_attribute): New.
228 (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
229 (ia64_asm_output_external): Use globalize_decl_name instead
231 * doc/extend.texi (version_id): New pragma.
232 * doc/tm.texi (ARGET_ASM_GLOBALIZE_DECL_NAME): New target hook.
234 2007-01-24 Andreas Krebbel <krebbel1@de.ibm.com>
236 * unwind-dw2-fde.c (get_cie_encoding): Replaced _Unwind_Word with
237 _uleb128_t and _Unwind_SWord with _sleb128_t.
238 * unwind-dw2.c (extract_cie_info, execute_stack_op, execute_cfa_program,
239 uw_frame_state_for, uw_update_context_1): Likewise.
240 * unwind-c.c (parse_lsda_header, PERSONALITY_FUNCTION): Likewise.
241 * unwind-pe.h (read_uleb128, read_sleb128,
242 read_encoded_value_with_base): Likewise.
243 * unwind-generic.h: Define _sleb128_t and _uleb128_t types.
245 2007-01-24 Richard Guenther <rguenther@suse.de>
247 * builtins.c (expand_builtin_cexpi): Get the fndecl
248 for cexp in the correct way.
250 2007-01-24 Jan Hubicka <jh@suse.cz>
252 * tree-ssa-dce.c (eliminate_unnecesary_stmts): Remove dead LHS of calls.
254 2007-01-24 Andreas Krebbel <krebbel1@de.ibm.com>
256 * c-cppbuiltin.c (builtin_define_type_sizeof): New function.
257 (c_cpp_builtins): New builtin macros: __SIZEOF_INT__, __SIZEOF_LONG__,
258 __SIZEOF_LONG_LONG__, __SIZEOF_SHORT__, __SIZEOF_POINTER__,
259 __SIZEOF_FLOAT__, __SIZEOF_DOUBLE__, __SIZEOF_LONG_DOUBLE__,
260 __SIZEOF_SIZE_T__, __SIZEOF_WCHAR_T__, __SIZEOF_WINT_T__ and
261 __SIZEOF_PTRDIFF_T__.
262 * doc/cpp.texi: Documentation for the new builtin macros added.
264 2007-01-24 Uros Bizjak <ubizjak@gmail.com>
266 * config/i386/i386.md (tanxf2, tan<mode>2, atan<mode>2, log<mode>2,
267 log10<mode>2, log2<mode>2, expxf2, exp10xf2, exp2xf2): Use op2
268 instead of operands[2] to avoid access past the end of array.
270 2007-01-24 Richard Sandiford <richard@codesourcery.com>
272 * reload1.c (emit_reload_insns): Pass the reload register
273 for a non-spill output reload through forget_old_reloads_1.
275 2007-01-23 Joseph Myers <joseph@codesourcery.com>
277 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Define
278 _SOFT_DOUBLE if doubles use software floating-point.
279 * config/rs6000/libgcc-ppc-glibc.ver: Export additional long
280 double functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
281 * config/rs6000/darwin-ldouble.c: Also compile functions for
282 hard-float without FPRs. Use fmsub function for all __NO_FPRS__
283 cases. Compile extra functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
284 * config/rs6000/linuxspe.h (SUBSUBTARGET_OVERRIDE_OPTIONS): Remove
285 commented-out long double override.
286 (CPP_LONGDOUBLE_DEFAULT_SPEC): Likewise.
287 * config/rs6000/eabispe.h: Likewise.
288 * config/rs6000/rs6000.c (rs6000_override_options): Don't override
289 long double for non-SPE.
290 (rs6000_handle_option): Likewise.
291 (invalid_e500_subreg): Disallow more subregs involding DImode,
292 DFmode, TImode or TFmode.
293 (rs6000_legitimate_offset_address_p): Check TFmode offsets for
295 (legitimate_lo_sum_address_p): Also check for TFmode for E500
297 (rs6000_legitimize_address): Also handle TFmode for E500 double.
298 (rs6000_legitimize_reload_address): Also handle TFmode for E500
300 (rs6000_legitimate_address): Also check for TFmode for E500
302 (rs6000_emit_move): Use DFmode subregs of TFmode for E500 double.
303 (spe_build_register_parallel): Handle TFmode and TCmode.
304 (rs6000_spe_function_arg): Handle TFmode and TCmode for E500
306 (function_arg): Handle TFmode and TCmode for E500 double.
307 (rs6000_init_libfuncs): Initialize extra libfuncs for soft double
309 (print_operand): Handle TFmode and TImode for %y.
310 (rs6000_generate_compare): Handle TFmode comparisons for E500
312 (spe_func_has_64bit_regs_p): Check for TFmode for E500 double.
313 (rs6000_function_value): Handle TFmode and TCmode for E500 double.
314 (rs6000_libcall_value): Handle TFmode and TCmode for E500 double.
315 * config/rs6000/rs6000.h (CANNOT_CHANGE_MODE_CLASS): Check for
316 TFmode for E500 double.
317 * config/rs6000/rs6000.md (FP): Allow TF for E500 double.
318 (floatsidf2): Enable for E500 double.
319 (movtf_softfloat): Use rs6000_nonimmediate_operand.
320 (extenddftf2): Change to extenddftf2_fprs.
321 (extenddftf2): Call gen_spe_extenddftf2 or gen_extenddftf2_fprs
322 depending on TARGET_E500_DOUBLE.
323 (extendsftf2): Enable for E500 double.
324 (trunctfdf2): Enable for E500 double.
325 (trunctfsf2): Change to trunctfsf2_fprs.
326 (trunctfsf2): Call gen_spe_trunctfsf2 or gen_trunctfsf2_fprs
327 depending on TARGET_E500_DOUBLE.
328 (floatsitf2): Enable for E500 double.
329 (fix_trunctfsi2): Change to fix_trunctfsi2_fprs.
330 (fix_trunctfsi2): Call gen_spe_fix_trunctfsi2 or
331 gen_fix_trunctfsi2_fprs depending on TARGET_E500_DOUBLE.
332 (negtf2): Change to negtf2_internal.
333 (negtf2): New expander.
334 (abstf2): Enable for E500 double. Call gen_spe_abstf2_tst,
335 gen_spe_abstf2_cmp or gen_abstf2_internal depending on
336 TARGET_E500_DOUBLE and flag_unsafe_math_optimizations.
337 (movdi_internal32): Use rs6000_nonimmediate_operand.
338 (unnamed splitter): Likewise.
339 * config/rs6000/spe.md (CMPTFEQ_GPR, TSTTFEQ_GPR, CMPTFGT_GPR,
340 TSTTFGT_GPR, CMPTFLT_GPR, TSTTFLT_GPR): New unspecs.
341 (SPE64TF, DITI): New mode macros.
342 (frob_df_di): Change to frob_<SPE64:mode>_<DITI:mode>; allow more
345 (frob_<mode>_di_2): New.
346 (frob_tf_di_8_2): New.
347 (frob_di_df): Change to frob_di_<mode>; allow more modes.
349 (frob_di_df_2): Change to frob_<DITI:mode>_<SPE64:mode>_2; allow
351 (frob_ti_<mode>_8_2): New.
353 (mov_si<mode>_e500_subreg0, mov_si<mode>_e500_subreg0_2,
354 mov_si<mode>_e500_subreg4, mov_si<mode>_e500_subreg4_2): Allow
356 (mov_sitf_e500_subreg8, mov_sitf_e500_subreg8_2,
357 mov_sitf_e500_subreg12, mov_sitf_e500_subreg12_2): New.
358 (spe_trunctfdf2_internal1, spe_trunctfsf2, spe_extenddftf2,
359 spe_fix_trunctfsi2, spe_fix_trunctfsi2_internal,
360 spe_negtf2_internal, spe_abstf2_cmp, spe_abstf2_tst): New.
361 (cmptfeq_gpr, tsttfeq_gpr, cmptfgt_gpr, tsttfgt_gpr, cmptflt_gpr,
364 2007-01-23 Ian Lance Taylor <iant@google.com>
366 * Makefile.in (OBJS-common): Reformat, alphabetize, but put
368 (OBJS-archive): Reformat, alphabetize.
369 (OBJS): Change out_object_file to OBJS-md.
371 2007-01-23 Uros Bizjak <ubizjak@gmail.com>
373 * config/i386/i386.md: Use REG_P, MEM_P, CONST_INT_P, LABEL_P,
374 JUMP_P and CALL_P predicates where applicable.
375 * config/i386/i386.c: Ditto.
376 * config/i386/i386.md: Ditto.
377 * config/i386/mmx.md: Ditto.
378 * config/i386/predicates.md: Ditto.
380 2007-01-22 Andreas Schwab <schwab@suse.de>
382 * config/m68k/m68k.h: Fix comment.
384 2007-01-22 Jan Hubicka <jh@suse.cz>
386 * passes.c (init_optimization_passes): Do not rerun
387 pass_early_warn_uninitialized.
389 2007-01-22 Richard Guenther <rguenther@suse.de>
391 PR tree-optimization/30038
392 * tree-ssa-math-opts.c (maybe_record_sincos): New static helper
394 (execute_cse_sincos_1): Likewise.
395 (execute_cse_sincos): Likewise.
396 (gate_cse_sincos): Likewise.
397 (pass_cse_sincos): New pass CSEing sin() and cos() calls using
398 the cexpi() canonicalization of sincos().
399 * tree-pass.h (pass_cse_sincos): Declare.
400 * passes.c (init_optimization_passes): New pass pas_cse_sincos.
402 2007-01-21 Eric Botcazou <ebotcazou@libertysurf.fr>
404 PR rtl-optimization/29329
405 * combine.c (replaced_rhs_insn): Rename to i2mod.
406 (replaced_rhs_value): Rename to i2mod_new_rhs.
407 (i2mod_old_rhs): New global variable.
408 (combine_instructions): Adjust for above change. Save a copy of
409 the old RHS into i2mod_old_rhs when the contents of a REG_EQUAL
410 note are substituted in the second instruction.
411 (distribute_notes) <REG_DEAD>: Adjust for above change. Do not
412 ditch the note if it pertains to the second eliminated register
413 and this register is mentioned in i2mod_old_rhs.
416 2006-09-12 Eric Botcazou <ebotcazou@libertysurf.fr>
418 * combine.c (distribute_notes) <REG_DEAD>: Do not consider SETs past
419 the insn to which the note was originally attached.
421 2007-01-21 Jan Hubicka <jh@suse.cz>
423 * ipa-inline.c (inlining_mode): Comment, move up.
424 (cgraph_decide_inlining_incrementally): Do not perform inlining
425 itself; fix handling of flattening of self recursive functions.
426 (cgraph_find_cycles): Remove.
427 (cgraph_flatten_node): Remove.
428 (cgraph_decide_inlining): Use incremental inliner to handle flattening.
429 (try_inline): New function.
430 (cgraph_early_inlining): Update call of
431 cgraph_decide_inlining_incrementally. Apply inlining here.
432 (apply_inline): Update call of cgraph_decide_inlining_incrementally.
434 2007-01-21 Dirk Mueller <dmueller@suse.de>
437 * tree-vrp.c (check_array_bounds): do not warn
438 about ADDR_EXPR's of ARRAY_REF's which are immediately
439 used in binary expressions.
441 2007-01-21 Ira Rosen <irar@il.ibm.com>
443 * tree-vectorizer.h (struct _stmt_vec_info): Add new field
444 read_write_dep and macros for its access.
445 * tree-vectorizer.c (new_stmt_vec_info): Initialize the new field.
446 * tree-vect-analyze.c (vect_analyze_data_ref_dependence): Remove
447 argument, call vect_check_interleaving for every independent pair of
448 data-refs. Mark loads that access the same memory location as a store
450 (vect_check_dependences): Remove.
451 (vect_analyze_data_ref_dependences): Remove vect_check_dependences
452 call, fix the call to vect_analyze_data_ref_dependence.
453 (vect_analyze_data_ref_access): For statements that access the same
454 data-ref, check that they are not stores; for loads, check that there
455 is no store that access the same location.
457 2007-01-20 Roger Sayle <roger@eyesopen.com>
458 Joseph Myers <joseph@codesourcery.com>
460 * doc/invoke.texi (-fdump-rtl-combine): Fix under/overfull hbox.
461 (-fdump-rtl-stack): Likewise.
462 (-fno-signed-zeros): Use @minus{} for a minus sign. Correct typo.
463 (-mcheck-zero-division, -mcpu): Fix under/overfull hbox.
464 (-mpt-fixed): Use @minus{} for minus sign.
465 (Using Precompiled Headers): Fix under/overfull hbox.
467 2007-01-20 Jan Hubicka <jh@suse.cz>
469 * tree-flow.h (struct stmt_ann_d): Move references_memory to proper
470 place within annotation.
472 2007-01-20 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
474 * pa.c (output_move_double): Change array size of xoperands to 4.
476 2007-01-20 Jan Hubicka <jh@suse.cz>
478 * tree-tailcall.c (adjust_return_value): Do not use RESULT_DECL
481 2007-01-19 Ian Lance Taylor <iant@google.com>
483 * expmed.c (expand_divmod) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]: Cast
484 constant to unsigned HOST_WIDE_INT before negating.
486 2007-01-19 Ian Lance Taylor <iant@google.com>
488 * tree-ssa-operands.h (struct vuse_vec_d): Change num_vuse field
490 (VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Use unsigned comparison.
491 (VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
492 (SET_VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Likewise.
493 (SET_VUSE_ELEMENT_VAR) [ENABLE_CHECKING]: Likewise.
494 (SET_VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
495 (realloc_vdef, realloc_vuse): Change second parameter to
497 (ssa_operand_iterator_d): Change vuse_index and mayuse_index
499 * tree-ssa-operands.c (realloc_vop): Change num_elem parameter to
500 unsigned. Change x and lim locals to unsigned.
501 (realloc_vdef, realloc_vuse): Change num_elem parameter to
503 (finalize_ssa_vuse_ops): Change old_i local to unsigned.
504 (copy_virtual_operands): Change i and n locals to unsigned.
506 2007-01-19 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
509 * builtins.c (fold_builtin_1): Handle builtin fdim.
511 2007-01-20 Jan Hubicka <jh@suse.cz>
513 * tree-ssa.c (init_tree_ssa): Do not call init_alias_heapvars.
514 * tree-ssa-structalias.c (compute_points_to_sets): Do call
516 (init_alias_heapvars): Initialize only when not already initialized.
517 (delete_alias_heapvars): Set heapvar_for_stmt to NULL.
519 2007-01-19 Roger Sayle <roger@eyesopen.com>
521 * common.opt (fsigned-zeros): New command line option.
522 * flags.h (HONOR_SIGNED_ZEROS): Control via flag_signed_zeros instead
523 of flag_unsafe_math_optimizations.
524 * opts.c (set_fast_math_flags): The -ffast-math command line option
525 implies -fno-signed-zeros.
526 (fast_math_flags_set_p): Likewise.
528 * doc/invoke.texi: Document new -fno-signed-zeros option, and update
529 the documentation of -ffast-math appropriately. Wrap long lines.
531 2007-01-19 Steve Ellcey <sje@cup.hp.com>
533 * system.h (ASM_MAKE_LABEL_LINKONCE): Poison.
534 * varasm.c (globalize_decl): Remove ASM_MAKE_LABEL_LINKONCE ifdef.
536 2007-01-19 Tomas Bily <tbily@suse.cz>
538 * cgraphunit.c (cgraph_finalize_function): Updating of pid
540 (tree_init_ic_make_global_vars): New function
541 (tree_init_edge_profiler): call of tree_init_ic_make_global_vars
542 (tree_gen_ic_profiler): New function
543 (tree_gen_ic_func_profiler): New function
544 (tree_profiling): Added calling of tree_gen_ic_func_profiler
545 (tree_profile_hooks): Added hook for indirec/virtual calls
546 * value-prof.c (tree_find_values_to_profile): New case for
548 (tree_values_to_profile): Call for determining indirect/virtual
550 (tree_indirect_call_to_profile): New function
551 (tree_ic_transform): New function
552 (tree_ic): New function
553 (find_func_by_pid): New function
554 (init_pid_map): New function
555 (tree_value_profile_transformations): Added check for
556 indirect/virtual call transformation
557 * value-prof.h (enum hist_type): New counter type for
558 indirect/virtual calls
559 (profile_hooks): Added new hook for profiling indirect/virtual
561 * profile.c (instrument_values): New case for indirect/virtual
563 * gcov-io.h (GCOV_LAST_VALUE_COUNTER): Changed to 6
564 (GCOV_COUNTER_V_INDIR): New counter type
565 (GCOV_COUNTER_NAMES): New name of counter "indirect" added
566 (GCOV_MERGE_FUNCTIONS): New merge function for indirect/virtual
568 * cgraph.c: Definition of cgraph_max_pid
569 (cgraph_create_node): Default init of pid attribute
570 * cgraph.h: Declaration of cgraph_max_pid
571 (struct cgraph_node): Added pid attribute
572 * libgcov.c (__gcov_indirect_call_profiler): New function
573 (__gcov_one_value_profiler_body): New function
574 (__gcov_one_value_profiler): Body was moved to
575 __gcov_one_value_profiler_body and calls it
577 2007-01-19 Basile Starynkevitch <basile@starynkevitch.net>
579 * doc/gty.texi (Options): Document the mark_hook option to GTY.
580 * gengtype.c (write_types_data, write_func_for_structure,
581 write_types, ggc_wtd, pch_wtd): Add skip_hooks to
582 write_types_data, ggc_wtd, pch_wtd for processing mark_hook.
583 (walk_type, write_func_for_structure): Generate the mark_hook if
586 2007-01-19 Jan Hubicka <jh@suse.cz>
588 * ipa-inline.c (cgraph_decide_inlining_incrementally): Instead of
589 'early' argument take inlining mode argument specifying whether to
590 inline for size/speeed or all functions; add support for flattening;
592 (cgraph_early_inlining): Update call of decide_inlining_incrementally.
594 2007-01-19 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
597 * toplev.c (warn_deprecated_use): Use %qD instead of %qs to print
598 the name of the declared identifier.
600 2007-01-19 Dirk Mueller <dmueller@suse.de>
602 * config/i386.h (CONDITIONAL_REGISTER_USAGE): Store
603 result of PIC_OFFSET_TABLE_REGNUM in temporary variable to avoid
604 duplicate evaluation.
606 2007-01-19 Uros Bizjak <ubizjak@gmail.com>
608 * config/i386/i386.md (acos<mode>2): Rename from acossf2 and acosdf2.
609 Macroize expander using X87MODEF12 mode macro. Extend operand 1
610 to XFMode, use acosxf2 and truncate result to requested mode.
611 Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
612 (asin<mode>2): Similarly, with asin expanders.
613 (*fscalexf4_i387): Rename from *fscalexf4.
614 (expNcorexf3): New expander.
615 (expxf2, exp10xf2, exp2xf2): Use expNcorexf3 expander.
616 (exp<mode>2): Rename from expsf2 and expdf2. Macroize expander using
617 X87MODEF12 mode macro. Extend operand 1 to XFMode, use expxf2 and
618 truncate result to requested mode. Use SSE_FLOAT_MODE_P to disable
619 patterns for SSE math.
620 (exp10<mode>2): Similarly, with exp10 expanders.
621 (exp2<mode>2): Similarly, with exp2 expanders.
622 (expm1<mode>2): Similarly, with expm1 expanders.
623 (ldexp<mode>3): Similarly, with ldexp expanders.
624 (log<mode>2, log10<mode>2, log2<mode>2, log1p<mode>2, logb<mode>2):
625 Use gen_truncxf<mode>2_i387_noop to truncate result.
627 2007-01-19 Richard Sandiford <richard@codesourcery.com>
629 * config/m68k/m68k.h (PREFERRED_STACK_BOUNDARY): Define to 32
630 for ColdFire targets.
632 2007-01-19 Nathan Sidwell <nathan@codesourcery.com>
633 Richard Sandiford <richard@codesourcery.com>
635 * config/m68k/m68k.h (M68K_STATIC_CHAIN_REG_NAME): New macro.
636 (INITIALIZE_TRAMPOLINE): Use STATIC_CHAIN_REGNUM.
637 (__transfer_from_trampoline): Use M68K_STATIC_CHAIN_REG_NAME.
638 * config/m68k/m68kelf.h (STATIC_CHAIN_REGNUM): Override.
639 (M68K_STATIC_CHAIN_REG_NAME): Likewise.
640 * config/m68k/netbsd-elf.h (M68K_STATIC_CHAIN_REG_NAME): Likewise.
642 2007-01-19 Richard Sandiford <richard@codesourcery.com>
644 * config/m68k/m68k.md (adddi_dilshr32): Rename to...
645 (*adddi_dilshr32): ...this. Fix formatting. Remove commented-out
646 non-canonical pattern. Restrict to !TARGET_COLDFIRE.
647 (*adddi_dilshr32_cf): New pattern.
648 (adddi3, subdi3): Remove first alternatives.
650 2007-01-19 Richard Sandiford <richard@codesourcery.com>
652 * config/m68k/m68k.c (notice_update_cc): If an SFmode move is
653 implemented using move.l, do not use its cc result for floating-point
656 2007-01-19 Richard Sandiford <richard@codesourcery.com>
658 * config/m68k/m68k.h (EXTRA_CONSTRAINT): Stop the 'T' constraint
659 from accepting 's' constraints if flag_pic.
661 2007-01-19 Richard Sandiford <richard@codesourcery.com>
663 * config/m68k/m68k.md (bordered, bunordered, buneq, bunge, bungt)
664 (bunle, bunlt, bltgt, bordered_rev, bunordered_rev, buneq_rev)
665 (bunge_rev, bungt_rev, bunle_rev, bunlt_rev, bltgt_rev): Change
666 condition from TARGET_68881 to TARGET_HARD_FLOAT.
668 2007-01-19 Sandra Loosemore <sandra@codesourcery.com>
670 * longlong.h (count_leading_zeros, COUNT_LEADING_ZEROS_0): Add
671 ColdFire alternatives.
672 * config/m68k/m68k.h (CLZ_DEFINED_VALUE_AT_ZERO): New macro.
673 * config/m68k/m68k.md (clzsi2): Define for ColdFire
674 architectures that support ff1 instruction.
676 2007-01-19 Richard Sandiford <richard@codesourcery.com>
677 Julian Brown <julian@codesourcery.com>
679 * config/m68k/m68k.h (CONST_OK_FOR_LETTER_P): Add an 'R' case.
680 * config/m68k/m68k.md (*movsi_cfv4): Fold into...
681 (*movsi_cf): ...here. Remove unnecessary 'R' from 'Rg'.
684 2007-01-19 Richard Sandiford <richard@codesourcery.com>
686 * config/m68k/m68k-protos.h (valid_mov3q_const): Take a HOST_WIDE_INT
688 (output_move_const_into_data_reg, output_move_simode_const): Delete.
689 * config/m68k/m68k.c (const_method, const_int_cost): Take a
690 HOST_WIDE_INT instead of an rtx.
691 (m68k_rtx_costs): Update call accordingly.
692 (output_move_const_into_data_reg): Likewise. Fix formatting.
693 (valid_mov3q_const): Take a HOST_WIDE_INT instead of an rtx.
695 (output_move_simode_const): Update calls after above changes.
696 Rework to use automatic variables and predicates like MEM_P.
697 * config/m68k/m68k.md (pushexthisi_const): Update call to
700 2007-01-19 Dirk Mueller <dmueller@suse.de>
702 * tree-ssa-alias.c (perform_var_substitution): Fix typo
705 2007-01-19 Richard Guenther <rguenther@suse.de>
707 * builtins.c (expand_builtin_cexpi): Fall back to expanding
708 via cexp in case sincos is not available.
710 2007-01-19 Richard Guenther <rguenther@suse.de>
712 * doc/tm.texi (TARGET_HAS_SINCOS): Document new target macro.
713 * defaults.h (TARGET_HAS_SINCOS): Default to off.
714 * config/linux.h (TARGET_HAS_SINCOS): Set to on if we have glibc.
715 * config/alpha/linux.h (TARGET_HAS_SINCOS): Likewise.
716 * config/sparc/linux.h (TARGET_HAS_SINCOS): Likewise.
717 * config/sparc/linux64.h (TARGET_HAS_SINCOS): Likewise.
718 * config/rs6000/linux.h (TARGET_HAS_SINCOS): Likewise.
719 * config/rs6000/linux64.h (TARGET_HAS_SINCOS): Likewise.
721 2007-01-19 Uros Bizjak <ubizjak@gmail.com>
723 * config/i386/i386.md (*fpatanxf3_i387, fpatan_extend<mode>xf3_i387):
725 (atan2sf3_1, atan2df3_1, atan2xf3_1): Remove insn patterns.
726 (atan2xf3): Directly generate RTL pattern.
727 (atan2<mode>3): Rename from atan2sf3 and atan2df3 and macroize insn
728 patterns using X87MODEF12 mode macro. Use fpatan_extend<mode>xf3_i387
729 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
730 disable patterns for SSE math.
731 (atan<mode>2): Rename from atansf2 and atandf2 and macroize insn
732 patterns using X87MODEF12 mode macro. Use fpatan_extend<mode>xf3_i387
733 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
734 disable patterns for SSE math.
736 2007-01-19 Alexandre Oliva <aoliva@redhat.com>
738 * libgcc-std.ver: Fix typo in %inherit for GCC_4.3.0.
740 2007-01-18 Roger Sayle <roger@eyesopen.com>
742 * fold-const.c (fold_unary) <VIEW_CONVERT_EXPR>: Optimize away a
743 VIEW_CONVERT_EXPR to the same type as it's operand.
745 2007-01-18 David Edelsohn <edelsohn@gnu.org>
747 * config/rs6000/darwin-ldouble.c: Only build _SOFT_FLOAT if
748 configured for long double 128.
750 2007-01-18 Mike Stump <mrs@apple.com>
752 * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Fix build
755 2007-01-18 Michael Meissner <michael.meissner@amd.com>
757 * i386.c (ix86_compute_frame_layout): Make fprintf's in #if 0 code
760 2007-01-18 Jan Hubicka <jh@suse.cz>
762 * tree-ssa-operands.c (vop_free_bucket_size): Never return value
763 greater than NUM_VOP_FREE_BUCKETS.
765 2007-01-18 Daniel Berlin <dberlin@dberlin.org>
767 * tree-ssa-structalias.c: Update comments.
768 (ptabitmap_obstack): Removed.
770 (oldpta_obstack): Ditto.
771 (stats): Add a few members.
772 (struct variable_info): Remove node, complex, address_taken, and
773 indirect_target members. Add oldsolution member.
774 (new_var_info): Do not initialize removed members.
775 (constraint_expr_type): Remove INCLUDES.
776 (constraint_graph): Add size, implicit_preds, rep,
777 indirect_cycles, eq_rep, label, direct_nodes, and complex members.
778 (FIRST_REF_NODE): New macro.
779 (LAST_REF_NODE): Ditto.
780 (FIRST_ADDR_NODE): Ditto.
781 (find): New function.
783 (dump_constraint): Do not handle INCLUDES.
784 (insert_into_complex): Do not insert duplicate constraints.
785 (condense_varmap_nodes): Renamed and rewritten into ...
786 (merge_node_constraints): This. Also fix bug in handling of
787 offseted copy constraints.
788 (clear_edges_for_node): No longer need to deal with preds at all,
789 or removing associated preds/succs.
790 (merge_graph_nodes): Deal with indirect_cycles.
791 Don't deal with predecessors.
792 (add_implicit_graph_edge): New function.
793 (add_pred_graph_edge): Ditto.
794 (add_graph_edge): Don't deal with predecessors.
795 (build_constraint_graph): Removed.
796 (build_pred_graph): New function.
797 (build_succ_graph): Ditto.
798 (struct scc_info): Removed in_component. Added roots, dfs, and
799 node_mapping. Remove visited_index, unification_queue.
800 (scc_visit): Deal with union-find we do now.
801 Deal with cycles with REF nodes.
802 (collapse_nodes): Renamed and rewritten to ...
804 (process_unification_queue): Removed.
805 (topo_visit): Cleanup
806 (do_da_constraint): Use find.
807 (do_sd_constraint): Ditto.
808 (do_ds_constraint): Ditto.
809 (do_complex_constraint): Ditto.
810 (init_scc_info): Update for removed and added members.
811 (find_and_collapse_graph_cycles): Renamed and rewritten into ...
812 (find_indirect_cycles): This.
813 (equivalence_class): New variable.
814 (label_visit): New function.
815 (perform_variable_substitution): Rewritten.
816 (free_var_substitution_info): New function.
817 (find_equivalent_node): Ditto.
818 (move_complex_constraints): Ditto.
819 (eliminate_indirect_cycles): Ditto.
820 (solve_graph): Only propagate changed bits.
821 Use indirect cycle elimination.
823 (tree_id_t): Rename to tree_vi_t, delete id member, add vi member.
824 (tree_id_eq): Renamed to ...
825 (tree_vi_eq): This. Update for member change
826 (insert_id_for_tree): Renamed and rewritten to ...
827 (insert_vi_for_tree): This.
828 (lookup_id_for_tree): Renamed and rewritten to ...
829 (lookup_vi_for_tree): This.
830 (get_id_for_tree): Renamed and rewritten to ...
831 (get_vi_for_tree): Ditto.
832 (get_constraint_exp_from_ssa_var): Update to use get_vi_for_tree.
833 (process_constraint): Don't handle INCLUDES.
834 Remove special ADDRESSOF case.
835 (find_func_aliases): Rewrite to use vi functions instead of id
837 (create_function_info_for): Ditto.
838 (create_variable_info_for): Ditto.
839 (intra_create_variable_infos): Ditto.
840 (merge_smts_into): Ditto.
841 (find_what_p_points_to): Ditto.
842 (init_base_vars): Ditto.
843 (init_alias_vars): Ditto.
844 (remove_preds_and_fake_succs): New function.
845 (dump_sa_points_to_info): Dump new stats.
846 (dump_solution_for_var): Use find.
847 (set_used_smts): Fix formatting.
848 (compute_points_to_sets): Updated for new functions.
849 (ipa_pta_execute): Ditto.
851 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
852 Richard Sandiford <richard@codesourcery.com>
854 * doc/tm.texi (TARGET_FUNCTION_VALUE): Expand documentation of
856 * calls.c (expand_call): If the return value is a PARALLEL,
857 extract its first member.
858 * config/m68k/linux.h (FUNCTION_EXTRA_EPILOGUE): Remove.
859 * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
860 use FUNCTION_EXTRA_EPILOGUE.
861 (m68k_function_value): Return a PARALLEL if the return value
862 is of a pointer type.
863 * config/m68k/netbsd-elf.h (current_function_returns_pointer)
864 (FUNCTION_EXTRA_EPILOGUE): Remove.
865 * config/m68k/m68k.md (D0_REG): New constant.
867 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
869 * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
870 output a NOP for empty epilogues.
872 2007-01-18 Richard Sandiford <richard@codesourcery.com>
874 * config/m68k/m68k.c (m68k_use_return_insn): Update comments
875 before function. Extend register save check to include all
876 registers, not just integer ones.
878 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
880 * config/m68k/m68k-protos.h (use_return_insn): Rename to...
881 (m68k_use_return_insn): ...this.
882 * config/m68k/m68k.h (USE_RETURN_INSN): Delete.
883 * config/m68k/m68k.c (use_return_insn): Rename to...
884 (m68k_use_return_insn): ...this.
885 * config/m68k/m68k.md (return): Use m68k_use_return_insn instead
888 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
890 * config/m68k/fpgnulib.c (__truncdfsf2): Implement round to
891 nearest even, fix denormal rounding overflow.
893 2007-01-18 Richard Sandiford <richard@codesourcery.com>
895 * config/m68k/m68k.md (movsf_cf_hard): Use fsmove instead of
896 f%$smove and f%$move.
897 (movdf_cf_hard): Use fdmove for cases 0 and 3 and fmove for case 1.
898 (extendsfdf2_cf): Use fdmove instead of f%&move.
899 (truncdfsf2_cf): Use fsmove instead of f%$smove.
900 (add<mode>3_cf, sub<mode>3_cf): Use <FP:prec> instead of <FP:round>.
902 2007-01-18 Richard Sandiford <richard@codesourcery.com>
904 * config/m68k/m68k.md (movdf_cf_hard): Use output_move_double for
907 2007-01-18 Richard Sandiford <richard@codesourcery.com>
909 * real.h (real_format): Add a canonical_nan_lsbs_set field.
910 (coldfire_single_format): Declare.
911 (coldfire_double_format): Likewise.
912 * real.c (encode_ieee_single): Use canonical_nan_lsbs_set instead
913 of qnan_msb_set to determine the lower bits of a canonical
915 (encode_ieee_double): Likewise.
916 (encode_ieee_quad): Likewise.
917 (ieee_single_format): Initialize canonical_nan_lsbs_set.
918 (mips_single_format): Likewise.
919 (ieee_double_format): Likewise.
920 (mips_double_format): Likewise.
921 (ieee_extended_motorola_format): Likewise.
922 (ieee_extended_intel_96_format): Likewise.
923 (ieee_extended_intel_128_format): Likewise.
924 (ieee_extended_intel_96_round_53_format): Likewise.
925 (ibm_extended_format): Likewise.
926 (mips_extended_format): Likewise.
927 (ieee_quad_format): Likewise.
928 (mips_quad_format): Likewise.
929 (vax_f_format): Likewise.
930 (vax_d_format): Likewise.
931 (vax_g_format): Likewise.
932 (i370_single_format): Likewise.
933 (i370_double_format): Likewise.
934 (decimal_single_format): Likewise.
935 (decimal_double_format): Likewise.
936 (decimal_quad_format): Likewise.
937 (c4x_single_format): Likewise.
938 (c4x_extended_format): Likewise.
939 (real_internal_format): Likewise.
940 (coldfire_single_format): New real_format.
941 (coldfire_double_format): Likewise.
942 * config/pdp11/pdp11.c (pdp11_f_format): Initialize
943 canonical_nan_lsbs_set.
944 (pdp11_d_format): Likewise.
945 * config/m68k/m68k.c (override_options): Override REAL_FORMAT_MODE
946 if TARGET_COLDFIRE_CPU.
948 2007-01-18 Richard Sandiford <richard@codesourcery.com>
950 * config/m68k/m68k-protos.h (m68k_output_pic_call): Delete.
951 (output_call, m68k_legitimize_call_address): Declare.
952 * config/m68k/m68k.h (EXTRA_CONSTRAINT): Remove unnecessary
953 parenthesees. Add support for a 'W' constraint.
954 (LEGITIMATE_PIC_OPERAND_P): Remove SYMBOL_REF_FLAG handling.
955 (PRINT_OPERAND_PUNCT_VALID_P): Remove comment about 'o'.
956 (m68k_symbolic_call, m68k_symbolic_jump): Declare.
957 * config/m68k/m68k.c (m68k_symbolic_call, m68k_symbolic_jump): New
959 (override_options): Initialize them. Do not set flag_no_function_cse
960 for TARGET_ID_SHARED_LIBRARY.
961 (m68k_output_pic_call): Delete.
962 (m68k_legitimize_call_address): New function.
963 (print_operand): Remove the %o prefix. Handle the %p prefix.
964 (output_call): New function.
965 (m68k_output_mi_thunk): Use m68k_symbolic_jump. Always load the
966 target address from the GOT if symbolic jumps are not allowed.
967 * config/m68k/m68k.md (call, general_operand): Do not set
968 SYMBOL_REF_FLAG. Use m68k_legitimize_call_address instead.
969 Merge separate flag_pic and !flag_pic define_insns into...
970 (*call, *call_value): ...these new patterns. Match the address
971 rather than the containing MEM and require it to be a call_operand.
972 Use output_call to generate the asm template.
973 * config/m68k/predicates.md (const_call_operand): New predicate.
974 (call_operand): Likewise.
976 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
978 * config/m68k/m68k.h (REGISTER_MOVE_COST): Simplify definition.
979 (STACK_GROWS_DOWNWARD): Define to 1.
980 (FUNCTION_VALUE, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P): Equivocate
981 comments, emphasizing that these values are only defaults.
982 * config/m68k/linux.h (LINK_SPEC): Fix formatting in #undef.
983 * config/m68k/m68k.c (const_method): Remove trailing whitespace.
985 2007-01-18 Richard Sandiford <richard@codesourcery.com>
987 * config/m68k/m68k.md (cmpsi): Remove outdated flag_pic handling.
989 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
990 Richard Sandiford <richard@codesourcery.com>
992 * config/m68k/m68k.h (DATA_REGNO_P, ADDRESS_REGNO_P, INT_REGNO_P)
993 (FP_REGNO_P): New macros.
994 (REGNO_OK_FOR_INDEX_P, REGNO_OK_FOR_BASE_P, REGNO_OK_FOR_DATA_P)
995 (REGNO_OK_FOR_FP_P, REG_OK_FOR_INDEX_P, REG_OK_FOR_BASE_P): Use them.
996 (EH_RETURN_STACKADJ_RTX): Use A0_REG.
997 * config/m68k/m68k.c (m68k_regno_mode_ok): Use the new REGNO macros.
999 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
1001 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
1002 (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-rtems*): Add t-floatlib
1004 * config/m68k/t-floatlib: New file, extracting common code from...
1005 * config/m68k/t-m68kbare, config/m68k/t-m68kelf,
1006 * config/m68k/t-uclinux: Here.
1007 * config/m68k/fpgnulib.c: Do not compile extendeed precision
1008 routines on ColdFire targets.
1010 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
1012 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
1013 (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*)
1014 (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-linux*)
1015 (m68k-*-rtems*): Use tm_file rather than m68k/m68k.h and
1016 explicitly set MOTOROLA to 1.
1017 * config/m68k/m68k.h (MOTOROLA): Simplify definition accordingly.
1019 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1020 Nathan Sidwell <nathan@codesourcery.com>
1022 * config/m68k/m68k.h (PCC_STATIC_STRUCT_RETURN): Delete.
1023 (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Add MOTOROLA cases.
1024 * config/m68k/coff.h (REGISTER_PREFIX_MD): Delete.
1025 * config/m68k/m68020-elf.h (LIB_SPEC): Delete.
1026 * config/m68k/m68k-none.h (CC1_SPEC, CPP_SUBTARGET_SPEC): Delete.
1027 * config/m68k/m68kelf.h (IMMEDIATE_PREFIX, REGISTER_PREFIX_MD)
1028 (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Delete.
1029 (NO_DOLLAR_IN_LABEL, PCC_STATIC_STRUCT_RETURN): Don't undefine.
1030 (BSS_ASM_OP, ASM_OUTPUT_SKIP, ASM_OUTPUT_ASCII): Delete.
1031 * config/m68k/m68kemb.h (PCC_STATIC_STRUCT_RETURN): Don't undefine.
1032 * config/m68k/linux.h (SIZE_TYPE, PTRDIFF_TYPE, WCHAR_TYPE)
1033 (WCHAR_TYPE_SIZE, TARGET_OBJFMT_CPP_BUILTINS): Delete.
1034 (TARGET_OS_CPP_BUILTINS): Don't define mc68000 and mc68020 here.
1035 (DBX_REGISTER_NUMBER): Delete.
1036 * config/m68k/netbsd-elf.h (IMMEDIATE_PREFIX): Delete.
1037 (PCC_STATIC_STRUCT_RETURN): Don't undefine.
1038 * config/m68k/openbsd.h (PCC_STATIC_STRUCT_RETURN): Define.
1040 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1042 * config.gcc (m68k-*-uclinux*): Add flat.h to $tm_file.
1043 * config/flat.h: New file.
1044 * crtstuff.c (USE_PT_GNU_EH_FRAME): Don't define if
1046 * config/m68k/m68k.h (ASM_PREFERRED_EH_DATA_FORMAT): Do not use
1047 indirect references for -msep-data or -mid-shared-library.
1048 Do not use PC-relative code addresses either.
1049 * config/m68k/m68k.c (override_options): Restrict -fPIC error
1051 * config/m68k/uclinux.h (STARTFILE_SPEC): Define. Use Scrt1.o
1052 for shared libraries and crt1.o for executables. Use crti.o and
1054 (ENDFILE_SPEC): Use crtend.o and crtn.o.
1055 (LIB_SPEC): Suppress -Rlibc.gdb if -static-libc is given.
1056 Do not add -elf2flt or -shared-lib-id options here.
1057 (LINK_SPEC): Define. Pass -elf2flt if no -elf2flt option is given.
1058 Pass -shared-lib-id if -mid-shared-library, taking the library
1059 identifier from -mshared-library-id if given, otherwise
1061 (EH_FRAME_IN_DATA_SECTION): Do not undefine.
1062 (INIT_SECTION_ASM_OP, FINI_SECTION_ASM_OP): Likewise.
1063 (TARGET_OS_CPP_BUILTINS): Define __GXX_MERGED_TYPEINFO_NAMES=0
1064 and __GXX_TYPEINFO_EQUALITY_INLINE=0 if -mid-shared-library.
1065 (DRIVER_SELF_SPECS): Map unadorned PIC options to -msep-data.
1066 * config/m68k/t-uclinux (EXTRA_MULTILIB_PARTS): Add crtbegin.o
1068 * config/m68k/lb1sf68.asm (PICCALL): Use an lea and pc-relative
1069 jump sequence for ISA A and ISA A+.
1070 (PICJUMP): Likewise.
1072 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1074 * config.gcc (m68k-*-uclinux*): Base the port on the common
1075 and m68k GNU/Linux files rather than on the generic ELF ones.
1076 * config/m68k/uclinux.h (TARGET_VERSION): Override.
1077 (TARGET_OS_CPP_BUILTINS): Use LINUX_TARGET_OS_CPP_BUILTINS.
1079 2007-01-18 Julian Brown <julian@codesourcery.com>
1080 Richard Sandiford <richard@codesourcery.com>
1082 * config/m68k/m68k.h (LONG_DOUBLE_TYPE_SIZE): Make 64-bit on ColdFire.
1083 (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
1084 * config/m68k/netbsd-elf.h (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Undefine
1086 * config/m68k/uclinux-oldabi.h (LONG_DOUBLE_TYPE_SIZE): Redefine to
1088 (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
1090 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1092 * doc/install.texi: Document m68k-uclinuxoldabi.
1093 * config.gcc (m68k-*-uclinuxoldabi): New configuration.
1094 * config/m68k/uclinux-oldabi.h: New file, copied from
1095 config/m68k/uclinux.h.
1097 2007-01-18 Kaz Kojima <kkojima@rr.iij4u.or.jp>
1099 * config/m32r/m32r-protos.h (m32r_expand_epilogue): Declare it.
1100 * config/m32r/m32r.c (m32r_setup_incoming_varargs): Use gen_frame_mem.
1101 (m32r_compute_frame_size): Use unsigned for regno.
1102 (m32r_reload_lr): Use gen_frame_mem.
1104 (m32r_output_function_epilogue): Don't output the function epilogue
1106 (m32r_expand_epilogue): New.
1107 (direct_return): Return false if the function has the interrupt
1109 (m32r_hard_regno_rename_ok): Remove code for the textual epilogue.
1110 * config/m32r/m32r.md (epilogue): New expander.
1111 (return_lr, return_rte): New insns.
1112 (return): Make it expander.
1113 (return_normal): New expander.
1115 2007-01-18 Josh Conner <jconner@apple.com>
1118 * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Add
1119 support for UNLE, UNLT, UNGE, and UNGT.
1121 2007-01-18 Jan Hubicka <jh@suse.cz>
1123 * tree-vrp.c (finalize_jump_threads): Do not call cleanup_cfg by hand.
1124 * tree-tailcall (add_virtual_phis): Likewise.
1125 (optimize_tail_call): Return TODOs.
1126 (execute_tail_calls): Return TODOs.
1127 * tree-ssa-ccp (execute_fold_all_builtins): Do cleanup_cfg via TODO.
1128 * tree-cfgcleanup.c (cleanup_tree_cfg_loop): Return if something
1130 * tree-ssa-forwprop.c (tree_ssa_forward_propagate_single_use_value):
1131 Cleanup cfg using TODO.
1132 * tree-flow.h (cleanup_tree_cfg_loop): Update prototype.
1133 * passes.c (execute_function_todo): When cleanup did something, remove
1135 * tree-cfg.c (pass_build_cfg): Add cleanup_cfg TODO.
1136 (make_edges): Don't cleanup_cfg.
1138 2007-01-18 Uros Bizjak <ubizjak@gmail.com>
1140 * reg-stack.c (subst_stack_regs_pat) [UNSPEC_SINCOS_COS,
1141 UNSPEC_XTRACT_FRACT]: Use generic code for instructions that
1142 operate on the top of stack.
1143 [UNSPEC_SINCOS_SIN, UNSPEC_XTRACT_EXP, UNSPEC_TAN]: Rewrite
1144 register handling of instructions that output to the second
1146 [UNSPEC_TAN_ONE, UNSPEC_TAN_TAN]: Remove.
1147 (move_for_stack_reg): Special-case check for dead destination
1148 stack slot for constant load of 1.0 inside UNSPEC_TAN.
1150 * config/i386/i386.md (UNSPEC_TAN): New constant.
1151 (UNSPEC_TAN_ONE, UNSPEC_TAN_TAN): Remove.
1152 (fptanxf4_i387, fptan_extend<mode>xf4_i387): New patterns
1153 to correctly model move of constant 1.0 to top stack slot.
1154 (*tandf3_1, *tansf3_1, *tanxf3_1): Remove insn patterns.
1155 (unnamed peephole2 pattern): Remove corresponding peephole2
1156 pattern that optimizes tan insn and loading of constant 1.0.
1157 (tanxf2): Use fptanxf4_i387.
1158 (tan<mode>2): Rename from tansf2 and tandf2 and macroize insn
1159 patterns using X87MODEF12 mode macro. Use fptan_extend<mode>xf4_i387
1160 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
1161 disable patterns for SSE math.
1162 (sincos<mode>3): Use truncxf<mode>2_i387_noop for truncation.
1163 (fyl2x_extend<mode>xf3_i387): Use X87MODEF12 for operand 1.
1165 2007-01-18 Dirk Mueller <dmueller@suse.de>
1166 Richard Guenther <rguenther@suse.de>
1169 * doc/invoke.texi (Warray-bounds): Document -Warray-bounds.
1170 * common.opt (Warray-bounds): Add new warning option.
1171 * c-opts.c (c_common_handle_option): Define -Warray-bounds
1173 * Makefile.in: make tree-vrp.o depend on toplev.h
1174 * tree-vrp.c (vrp_finalize): Call check_array_refs if -Warray-bounds
1176 (check_array_refs, check_array_bounds, check_array_ref): New.
1178 2007-01-18 Jan Hubicka <jh@suse.cz>
1180 * tree-ssa-ccp.c (ccp_finalize): Return if something changed.
1181 (execute_ssa_ccp): Return flags conditionally.
1182 * tree-ssa-propagate.c (substitue_and_fold): Return if something was
1184 * tree-ssa-propagate.h (substitute_and_fold): Update prototype.
1186 2007-01-18 Steven Bosscher <steven@gcc.gnu.org>
1188 * cfgcleanup.c (cleanup_cfg): Detect cfglayout mode and set
1189 the CLEANUP_CFGLAYOUT flag when in cfglayout mode.
1191 * Makefile.c (GTFILES): Add cfglayout.h.
1192 * gengtype.c (open_base_files): Likewise.
1193 * cfglayout.c (cfg_layout_function_footer,
1194 cfg_layout_function_header) Reindent to make gengtype happy.
1195 * cfglayout.h (cfg_layout_function_footer,
1196 cfg_layout_function_header): Add GTY(()) marker.
1198 * ifcvt.c (noce_try_sign_mask): Make sure INSN_B is non-null.
1200 2007-01-18 Ben Elliston <bje@au.ibm.com>
1202 * genautomata.c (write_automata): Include xstrerror output in the
1203 error message if writing the DFA description file fails.
1205 2007-01-17 H.J. Lu <hongjiu.lu@intel.com>
1207 * config/mips/mips-protos.h (mips_output_external): Make it
1209 * config/mips/iris.h (TARGET_ASM_EXTERNAL_LIBCALL): Removed.
1210 * config/mips/mips.c (irix_output_external_libcall): Likewise.
1211 (extern_list): Likewise.
1212 (extern_head): Likewise.
1213 (TARGET_ASM_FILE_END): Likewise.
1214 (mips_file_end): Likewise.
1215 (mips_output_external): Rewritten.
1217 2007-01-18 Ben Elliston <bje@au.ibm.com>
1219 * genpreds.c (write_insn_preds_c): Only write out the function
1220 body for regclass_for_constraint if we have register constraints.
1222 2007-01-17 Tom Tromey <tromey@redhat.com>
1224 * doc/sourcebuild.texi (libgcj Tests): Use sourceware.org.
1225 * doc/install.texi (Testing): Use sourceware.org.
1226 (Binaries): Likewise.
1227 (Specific): Likewise.
1228 * doc/contrib.texi (Contributors): Use sourceware.org.
1230 2007-01-17 Anatoly Sokolov <aesok@post.ru>
1232 * config/avr/avr.h (AVR_HAVE_LPMX): New macro.
1233 (AVR_ENHANCED): Rename to ...
1234 (AVR_HAVE_MUL): ... new.
1235 (avr_enhanced_p): Rename to ...
1236 (avr_have_mul_p): ... new.
1237 (TARGET_CPU_CPP_BUILTINS): Use 'avr_have_mul_p' instead of
1238 'avr_enhanced_p' for "__AVR_ENHANCED__". Define "__AVR_HAVE_MUL__".
1239 * config/avr/avr.c (avr_enhanced_p): Rename to ...
1240 (avr_have_mul_p): ... new.
1241 (base_arch_s): Rename 'enhanced' to 'have_mul'.
1242 (avr_override_options): Use 'avr_have_mul_p' and 'have_mul' instead of
1243 'avr_enhanced_p' and 'enhanced'.
1244 (ashlhi3_out, ashrhi3_out, lshrhi3_out, avr_rtx_costs): Use
1245 AVR_HAVE_MUL instead of AVR_ENHANCED.
1246 * avr.md (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
1247 (mulqi3, *mulqi3_enh, *mulqi3_call, mulqihi3, umulqihi3, mulhi3,
1248 *mulhi3_enh, *mulhi3_call, mulsi3, *mulsi3_call): Use AVR_HAVE_MUL
1249 instead of AVR_ENHANCED.
1250 (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
1251 * libgcc.S: Use __AVR_HAVE_MUL__ instead of __AVR_ENHANCED__.
1252 (__tablejump__): Use __AVR_HAVE_LPMX__ instead of __AVR_ENHANCED__.
1254 2007-01-17 Ian Lance Taylor <iant@google.com>
1256 * vec.h (VEC_reserve_exact): Define.
1257 (vec_gc_p_reserve_exact): Declare.
1258 (vec_gc_o_reserve_exact): Declare.
1259 (vec_heap_p_reserve_exact): Declare.
1260 (vec_heap_o_reserve_exact): Declare.
1261 (VEC_OP (T,A,reserve_exact)): New static inline function, three
1263 (VEC_OP (T,A,reserve)) [all versions]: Remove handling of
1265 (VEC_OP (T,A,alloc)) [all versions]: Call ...reserve_exact.
1266 (VEC_OP (T,A,copy)) [all versions]: Likewise.
1267 (VEC_OP (T,a,safe_grow)) [all versions]: Likewise.
1268 * vec.c (calculate_allocation): Add exact parameter. Change all
1270 (vec_gc_o_reserve_1): New static function, from vec_gc_o_reserve.
1271 (vec_gc_p_reserve, vec_gc_o_reserve): Call vec_gc_o_reserve_1.
1272 (vec_gc_p_reserve_exact, vec_gc_o_reserve_exact): New functions.
1273 (vec_heap_o_reserve_1): New static function, from vec_heap_o_reserve.
1274 (vec_heap_p_reserve, vec_heap_o_reserve): Call vec_heap_o_reserve_1.
1275 (vec_heap_p_reserve_exact): New function.
1276 (vec_heap_o_reserve_exact): New function.
1278 2007-01-17 Jan Hubicka <jh@suse.cz>
1280 * ipa-type-escape.c (look_for_casts): Revamp using handled_component_p.
1282 2007-01-17 Eric Christopher <echristo@apple.com>
1284 * config.gcc: Support core2 processor.
1286 2007-01-16 Jan Hubicka <jh@suse.cz>
1288 * tree-ssanames.c (release_dead_ssa_names): Instead of ggc_freeing
1289 the names, just unlink the chain so we don't crash on dangling pointers
1292 2007-01-16 Jan Hubicka <jh@suse.cz>
1294 * cgraph.h (cgraph_decide_inlining_incrementally): Kill.
1295 * tree-pass.h: Reorder to make IPA passes appear toegher.
1296 (pass_early_inline, pass_inline_parameters, pass_apply_inline): Declare.
1297 * cgraphunit.c (cgraph_finalize_function): Do not compute inling
1298 parameters, do not call early inliner.
1299 * ipa-inline.c: Update comments. Include tree-flow.h
1300 (cgraph_decide_inlining): Do not compute inlining parameters.
1301 (cgraph_decide_inlining_incrementally): Return TODOs; assume to
1302 be called with function context set up.
1303 (pass_ipa_inline): Remove unreachable functions before pass.
1304 (cgraph_early_inlining): Simplify assuming to be called from the
1306 (pass_early_inline): New pass.
1307 (cgraph_gate_ipa_early_inlining): New gate.
1308 (pass_ipa_early_inline): Turn into simple wrapper.
1309 (compute_inline_parameters): New function.
1310 (gate_inline_passes): New gate.
1311 (pass_inline_parameters): New pass.
1312 (apply_inline): Move here from tree-optimize.c
1313 (pass_apply_inline): New pass.
1314 * ipa.c (cgraph_remove_unreachable_nodes): Verify cgraph after
1316 * tree-inline.c (optimize_inline_calls): Return TODOs rather than
1318 (tree_function_versioning): Do not allocate dummy struct function.
1319 * tree-inline.h (optimize_inline_calls): Update prototype.
1320 * tree-optimize.c (execute_fixup_cfg): Export.
1321 (pass_fixup_cfg): Remove
1322 (tree_rest_of_compilation): Do not apply inlines.
1323 * tree-flow.h (execute_fixup_cfg): Declare.
1324 * Makefile.in (gt-passes.c): New.
1325 * passes.c: Include gt-passes.h
1326 (init_optimization_passes): New passes.
1327 (nnodes, order): New static vars.
1328 (do_per_function_toporder): New function.
1329 (execute_one_pass): Dump current pass here.
1330 (execute_ipa_pass_list): Don't dump current pass here.
1332 2007-01-16 Janis Johnson <janis187@us.ibm.com>
1334 * config/dfp-bit.c (dfp_compare_op): Return separate value for NaN.
1335 (DFP_NE, DFP_LE, DFP_GE): Return false for NaN.
1337 2007-01-16 David Edelsohn <edelsohn@gnu.org>
1339 * config/rs6000/darwin-ldouble.c: Build file for SOFT_FLOAT.
1340 (strong_alias): Define.
1341 (__gcc_qmul): Provide non-FMA for soft-float.
1347 (__gcc_qunord): New.
1357 * config/rs6000/rs6000.c (rs6000_init_libfuncs): Initialize
1358 soft-float functions.
1359 * config/rs6000/libgcc-ppc-glibc.ver: Version soft-float symbols.
1360 * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Do not warn
1361 about long double soft float.
1363 2007-01-16 Dorit Nuzman <dorit@il.ibm.com>
1364 Tehila Meyzels <tehila@il.ibm.com>
1366 * tree-vectorizer.h (is_pattern_stmt_p): New.
1367 * tree-vect-analyze.c (vect_determine_vectorization_factor): Fix
1368 formatting (tabs instead of spaces). Cleanup and clarify setting
1369 of STMT_VINFO_VECTYPE. Call is_pattern_stmt_p.
1370 * tree-vect-transform.c (vect_get_vec_def_for_operand): Fix typo.
1371 (vectorizable_type_demotion): Check that types are integral.
1372 (vectorizable_type_promotion): Likewise.
1373 (vectorizable_store): Fix typo. Eliminate new-line at end of
1376 2007-01-16 Jan Hubicka <jh@suse.cz>
1378 * tree-ssanames.c (release_dead_ssa_names): Remove invalidated
1381 2007-01-15 Eric Christopher <echristo@apple.com>
1383 * ifcvt.c: Include vec.h, vecprim.h.
1384 (check_cond_move_block): New argument regs.
1385 Reorganize. Add registers used to regs.
1386 (cond_move_process_if_block): Use regs set above as
1389 2007-01-15 Eric Christopher <echristo@apple.com>
1391 * config/darwin.h: Update copyright.
1392 (TARGET_OPTION_TRANSLATE_TABLE): Add umbrella.
1393 (LINK_COMMAND_SPEC): Add -u.
1394 (LINK_SPEC): Fix umbrella for above.
1396 2007-01-15 Joseph S. Myers <joseph@codesourcery.com>
1398 * config/soft-fp/op-common.h, config/soft-fp/op-4.h: Update from
1401 2007-01-15 Tom Tromey <tromey@redhat.com>
1403 * doc/sourcebuild.texi (libgcj Tests): Don't mention jacks.
1404 * doc/install.texi (Testing): Don't mention jacks.
1405 (Configuration): Document --enable-java-maintainer-mode. Move
1406 --with-java-home to libgcj-specific section. Document
1408 (Prerequisites): Mention --enable-java-maintainer-mode, ecj1.
1410 2007-01-15 Jan Hubicka <jh@suse.cz>
1412 * tree-ssa-dce.c (DCE_TODOs): New.
1413 (propagate_necessity): Return if something changed.
1414 (eliminate_unnecessary_stmts): Likewise.
1415 (perform_tree_ssa_dce): Return TODO flags when needed.
1416 (pass_dce, pass_dce_loop, pass_cd_dce): Remove TODO flags.
1418 2007-01-15 Uros Bizjak <ubizjak@gmail.com>
1420 * config/i386/i386.md (fyl2xxf3_i387): Rename from fyl2x_xf3.
1421 (fyl2x_extend<mode>xf3_i387): New insn pattern.
1422 (log<mode>2): Rename from logsf2 and logdf2 and macroize insn
1423 patterns using X87MODEF12 mode macro. Extend operand 1
1424 to XFmode. Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
1425 (log10<mode>2): Ditto.
1426 (log2<mode>2): Ditto.
1427 (log1p<mode>2): Ditto.
1428 (logb<mode>2): Ditto.
1429 (fyl2xp1xf3_i387): Rename from fyl2xp1_xf3.
1430 (fyl2xp1_extend<mode>xf3_i387): New insn pattern.
1431 (*fxtractxf3_i387): Rename from *fxtractxf3.
1432 (fxtract_extend<mode>xf3_i387): New insn pattern.
1433 (ilogbsi2): Use match_dup 3, not match_operand:XF 3.
1434 * config/i386/i386.c (ix86_emit_i387_log1p): Use gen_fyl2xp1xf3_i387()
1435 and gen_fyl2xxf3_i387().
1437 2007-01-14 Zdenek Dvorak <dvorakz@suse.cz>
1439 * loop-unswitch.c (unswitch_loop): Do not call fix_loop_placement.
1440 * cfgloopmanip.c (fix_loop_placement): Made static. Use
1441 get_loop_exit_edges. Changed return type to bool.
1442 * cfgloop.h (fix_loop_placement): Declaration removed.
1444 2007-01-14 Dorit Nuzman <dorit@il.ibm.com>
1446 * param.h (MIN_VECT_LOOP_BOUND): New.
1447 * params.def (MIN_VECT_LOOP_BOUND): New.
1448 * tree-vectorizer.c (slpeel_tree_peel_loop_to_edge): Takes another
1449 argument - minimum threshold for number of iterations.
1450 * tree-vectorizer.h (slpeel_tree_peel_loop_to_edge): Add another
1451 argument to declaration.
1452 * tree-vect-analyze.c (vect_analyze_operations): Check value of
1453 MIN_VECT_LOOP_BOUND.
1454 * tree-vect-transform.c (vect_do_peeling_for_loop_bound): Call
1455 slpeel_tree_peel_loop_to_edge with additional argument.
1456 (vect_do_peeling_for_alignment): Likewise.
1457 * doc/invoke.texi (min-vect-loop-bound): Document new param option.
1459 2007-01-14 Uros Bizjak <ubizjak@gmail.com>
1462 * config/i386/i386.c (print_operand) ['z']: Output 'b' for
1465 2007-01-14 Jan Hubicka <jh@suse.cz>
1467 * tree-dfa.c (remove_referenced_var): New function.
1468 * tree-ssa-live.c (remove_unused_locals): Walk referenced vars and
1469 prune referenced vars list too.
1470 * tree-flow.h (remove_referenced_var): Declare.
1472 2007-01-14 Jan Hubicka <jh@suse.cz>
1474 * tree-eh.c (add_stmt_to_eh_region_fn): Do not add call_exprs
1476 (remove_stmt_from_eh_region_fn): Do not remove call_exprs.
1477 (verify_eh_throw_stmt_node, verify_eh_throw_table_statements): Kill.
1478 * except.h (verify_eh_throw_table_statements): Kill prototype.
1479 * cfgexpand.c (expand_gimple_basic_block): Propagate Eh regions
1480 into call exrepssions.
1481 * tree-optimize.c (execute_free_cfg_annotatiosn): Do not call
1483 * tree-cfg.c: Include pointer-set.h.
1484 (verify_node_sharing): Work on pointer set.
1485 (verify_eh_throw_stmt_node): New.
1486 (verify_stmts): Use pointers sets, verify throw_stmt.
1488 2007-01-13 Zdenek Dvorak <dvorakz@suse.cz>
1490 * ipa-reference.c (analyze_function): Consider also addresses taken
1493 2007-01-12 Roger Sayle <roger@eyesopen.com>
1495 * c-typeck.c (null_pointer_constant_p): Replace use of
1496 TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
1497 (build_c_cast): Likewise.
1499 2007-01-12 Roger Sayle <roger@eyesopen.com>
1501 * tree.h (force_fit_type_double): Remove unused final argument.
1502 * c-common.c (constant_expression_warning): Replace use of
1503 TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
1504 (convert_and_check): Likewise.
1505 (shorten_compare): Update call to force_fit_type_double.
1506 (c_common_truthvalue_conversion) <INTEGER_CST>: Use integer_zerop.
1507 * convert.c (convert_to_pointer): Update call to
1508 force_fit_type_double.
1509 * fold-const.c (force_fit_type_double): Remove overflowed_const
1511 (int_const_binop, fold_convert_const_int_from_int,
1512 fold_convert_const_int_from_real, fold_div_compare,
1513 fold_sign_changed_comparison, fold_unary, fold_negate_const,
1514 fold_abs_const, fold_not_const): Remove the final argument from
1515 calls to force_fit_type_double.
1517 2007-01-12 Andrew Pinski <andrew_pinski@playstation.sony.com>
1519 * configure.ac: Set insn to "nop" for spu-*-* also.
1520 * configure: Regenerate.
1522 2007-01-12 Olga Golovanevsky <olga@il.ibm.com>
1524 * builtins.def : Add BUILT_IN_FREE.
1526 2007-01-12 Jan Hubicka <jh@suse.cz>
1528 PR tree-optimization/30443
1529 * tree-inline.c (tree_function_versioning): Do not optimize when
1530 cloning for inlining.
1532 2007-01-12 Zdenek Dvorak <dvorakz@suse.cz>
1534 * doc/loop.texi: Document recording of loop exits.
1535 * cfgloopmanip.c (loopify, duplicate_loop): Use alloc_loop.
1536 (update_single_exits_after_duplication,
1537 update_single_exit_for_duplicated_loop,
1538 update_single_exit_for_duplicated_loops): Removed.
1539 (duplicate_loop_to_header_edge): Do not call
1540 update_single_exits_after_duplication and
1541 update_single_exit_for_duplicated_loops.
1542 (loop_version): Do not update single_exit information.
1543 (fix_loop_structure): Use record_loop_exits instead of
1544 mark_single_exit_loops.
1545 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update
1546 the lists of loop exits.
1547 * cfghooks.c (redirect_edge_and_branch, redirect_edge_and_branch_force,
1548 split_edge, merge_blocks): Update the lists of loop exits.
1549 * modulo-sched.c (sms_schedule): Pass LOOPS_HAVE_RECORDED_EXITS to
1550 loop_optimizer_init.
1551 * loop-init.c (loop_optimizer_init): Call record_loop_exits instead
1552 of mark_single_exit_loops.
1553 (loop_optimizer_finalize): Call release_recorded_exits.
1554 * tree-ssa-loop.c (tree_loop_optimizer_init): Pass
1555 LOOPS_HAVE_RECORDED_EXITS to loop_optimizer_init.
1556 * tree-vectorizer.c (slpeel_tree_duplicate_loop_to_edge_cfg): Do not
1557 update single exit information.
1558 * lambda-code.c (perfect_nestify): Ditto.
1559 * cfgloop.c (flow_loop_free): Destroy the list of exits of the loop.
1560 (mark_single_exit_loops): Removed.
1561 (alloc_loop, loop_exit_hash, loop_exit_eq, loop_exit_free,
1562 get_exit_descriptions, rescan_loop_exit, record_loop_exits,
1563 dump_recorded_exit, dump_recorded_exits, release_recorded_exits): New
1565 (get_loop_exit_edges, single_exit): Use recorded exit lists.
1566 (add_bb_to_loop, remove_bb_from_loops): Update the lists of loop exits.
1567 (verify_loop_structure): Verify consistency of the exit lists.
1568 (flow_loops_find): Use alloc_loop. Initialize exits hash.
1569 (set_single_exit): Removed.
1570 * cfgloop.h (struct loop_exit): New function.
1571 (struct loop): single_exit_ field replaced by exits field.
1572 (LOOPS_HAVE_MARKED_SINGLE_EXITS): Replaced by LOOPS_HAVE_RECORDED_EXITS.
1573 (struct loops): Added exits hash.
1574 (mark_single_exit_loops, set_single_exit): Declaration removed.
1575 (release_recorded_exits, record_loop_exits, rescan_loop_exit): Declare.
1577 2007-01-12 Richard Sandiford <richard@codesourcery.com>
1579 * doc/invoke.texi: Avoid use of @headitem.
1581 2007-01-12 Richard Sandiford <richard@codesourcery.com>
1583 * cse.c (cse_insn): Move HAVE_CC0 code after declarations.
1585 2007-01-12 Richard Sandiford <richard@codesourcery.com>
1587 * doc/install.texi: Fix m68k-*-* anchor and add m68k-*-* to the
1590 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
1591 Richard Sandiford <richard@codesourcery.com>
1593 * doc/invoke.texi: Document -mno-bitfield, -mno-rtd and -mno-short.
1594 * config/m68k/m68k.opt: Resort options.
1595 (mbitfield, mrtd, mshort): Remove RejectNegative properties.
1597 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
1598 Richard Sandiford <richard@codesourcery.com>
1600 * doc/invoke.texi: Document the macros that are defined by
1601 m68k's -mtune and -mhard-float options.
1602 * config/m68k/m68k-protos.h (m68k_cpp_cpu_ident) Declare.
1603 (m68k_cpp_cpu_family): Likewise.
1604 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Add a full set
1605 of __ucfv*__ macros. Define __mcffpu__ if generating code for
1606 ColdFire FPUs. Define __mcf_cpu_* and __mcf_family_* macros.
1607 * config/m68k/m68k.c (m68k_cpp_cpu_ident): New function.
1608 (m68k_cpp_cpu_family): Likewise.
1610 2007-01-12 Richard Sandiford <richard@codesourcery.com>
1612 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Treat all mc68*
1613 macros besides mc68000 as tuning macros. Use a switch statement
1614 to set them and mcpu32.
1616 2007-01-12 Julian Brown <julian@codesourcery.com>
1618 * config/m68k/m68k.h: Use TARGET_68040 instead of TARGET_68040_ONLY.
1619 (TARGET_68040_ONLY): Rename to...
1620 (TARGET_68040): ...this.
1621 * config/m68k/m68k.c: Use TARGET_68040 instead of TARGET_68040_ONLY.
1622 * config/m68k/m68k.md: Likewise.
1624 2007-01-12 Julian Brown <julian@codesourcery.com>
1625 Nathan Sidwell <nathan@codesourcery.com>
1626 Richard Sandiford <richard@codesourcery.com>
1628 * config.gcc (m680[012]0-*-*, m68k*-*-*): Set m68k_cpu_ident to
1629 the -mcpu= argument associated with the --with-cpu setting.
1630 Define M68K_DEFAULT_TUNE to the default -mtune= option,
1631 if different from the one implied by the -mcpu setting.
1632 Accept --with-cpu=FOO if FOO is listed in m68k-devices.def,
1633 using mcpu=FOO as the default CPU option. Set target_cpu_default2.
1634 * doc/invoke.texi: Mention ColdFire in the introduction to the
1635 m68k options. Document the new -march, -mcpu, -mtune, -mdiv,
1636 -mno-div and -mhard-float options. Make -m68881 a synonym for
1637 -mhard-float. Document the previously-undocumented -m5206e,
1638 -m528x, -m5307 and -m5407 options. Tweak the existing option
1639 documentation for consistency.
1640 * doc/install.texi: Mention new --with-cpu arguments.
1641 * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Only use the
1642 default CPU if neither -mcpu nor -march are specified.
1643 (ASM_CPU_SPEC): Pass down -mcpu and -march options.
1644 (TARGET_CPU_CPP_BUILTINS): Set __mcfisa*__ macros from
1645 TARGET_ISA*. Set the legacy __mcf*__ cpu macros in the same way,
1646 using m68k_tune to decide between families that implement the
1647 same ISA. Use m68k_tune to set __mcfv4e__.
1648 (FL_BITFIELD, FL_68881, FL_COLDFIRE, FL_CF_HWDIV, FL_CF_MAC)
1649 (FL_CF_EMAC, FL_CF_EMAC_B, FL_CF_USP, FL_CF_FPU, FL_ISA_68000)
1650 (FL_ISA_68010, FL_ISA_68020, FL_ISA_68040, FL_ISA_A, FL_ISA_B)
1651 (FL_ISA_C, FL_ISA_MMU): New macros.
1652 (MASK_COLDFIRE): Delete.
1653 (TARGET_68010, TARGET_68020, TARGET_68040_ONLY, TARGET_COLDFIRE)
1654 (TARGET_ISAB): Redefine in terms of m68k_cpu_flags.
1655 (TARGET_68881, TARGET_COLDFIRE_FPU): Redefine in terms of m68k_fpu.
1656 (TARGET_HARD_FLOAT): Do not define here.
1657 (TARGET_ISAAPLUS, TARGET_ISAC): New macros.
1658 (TUNE_68000): New macro.
1659 (TUNE_68000_10): Redefine in terms of TUNE_68000 and TUNE_68010.
1660 (TUNE_68010, TUNE_68030, TUNE_68040, TUNE_68060, TUNE_CPU32)
1661 (TUNE_CFV2): Redefine in terms of m68k_tune.
1662 (uarch_type, target_device, fpu_type): New enums.
1663 (m68k_cpu, m68k_tune, m68k_fpu, m68k_cpu_flags): Declare.
1664 * config/m68k/m68k.c (TARGET_DEFAULT): Remove MASK_68881.
1665 (FL_FOR_isa_00, FL_FOR_isa_10, FL_FOR_isa_20, FL_FOR_isa_40)
1666 (FL_FOR_isa_cpu32, FL_FOR_isa_a, FL_FOR_isa_aplus, FL_FOR_isa_b)
1667 (FL_FOR_isa_c): New macros.
1668 (m68k_isa): New enum.
1669 (m68k_target_selection): New structure.
1670 (all_devices, all_isas, all_microarchs): New tables.
1671 (m68k_cpu_entry, m68k_arch_entry, m68k_tune_entry, m68k_cpu)
1672 (m68k_tune, m68k_fpu, m68k_cpu_flags): New variables.
1673 (MASK_ALL_CPU_BITS): Delete.
1674 (m68k_find_selection): New function.
1675 (m68k_handle_option): Handle -mcpu=, -march= and -mtune=.
1676 Map the legacy target options to a combination of the new ones.
1677 (override_options): Set m68k_cpu, m68k_tune, m68k_fpu and
1678 m68k_cpu_flags. Handle M68K_DEFAULT_TUNE. Use m68k_cpu_flags
1679 to derive default MASK_BITFIELD, MASK_CF_HWDIV and MASK_HARD_FLOAT
1681 * config/m68k/m68k.opt (m5200, m5206e, m528x, m5307, m5407, mcfv4e)
1682 (m68010, m68020, m68020-40, m68020-60, m68030, m68040): Remove Mask
1684 (m68881, msoft-float): Change mask from 68881 to HARD_FLOAT.
1685 (march=, mcpu=, mdiv, mhard-float, mtune=): New options.
1686 * config/m68k/m68k-devices.def: New file.
1688 2007-01-12 Richard Sandiford <richard@codesourcery.com>
1689 Nathan Sidwell <nathan@codesourcery.com>
1691 * config/m68k/m68k.h (ASM_CPU_SPEC, ASM_SPEC, EXTRA_SPECS)
1692 (SUBTARGET_EXTRA_SPECS): New macros.
1693 * config/m68k/linux.h (ASM_SPEC): Remove CPU flags;
1694 use %(asm_cpu_spec) instead.
1695 * config/m68k/m68k-none.h (ASM_SPEC): Likewise.
1696 * config/m68k/openbsd.h (ASM_SPEC): Likewise.
1697 * config/m68k/netbsd-elf.h (ASM_SPEC): Likewise.
1698 (EXTRA_SPECS): Rename to...
1699 (SUBTARGET_EXTRA_SPECS): ...this.
1701 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
1702 Richard Sandiford <richard@codesourcery.com>
1703 Julian Brown <julian@codesourcery.com>
1705 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
1706 (m68k-*-uclinux*, m68k-*-linux*, m68k-*-rtems*): Set default_m68k_cpu
1707 to the configuration's default CPU.
1708 (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*): Likewise.
1709 Remove default masks.
1710 (m680[012]0-*-*): Set the default with_cpu to the first part of
1712 (m68k*-*-*): Set the default with_cpu to m$default_m68k_cpu.
1713 (m68k*-*-linux): Extend the --with-cpu handling to...
1714 (m680[012]0-*-*, m68k*-*-*): ...these configurations. Allow m68000
1715 and m68010. Don't set target_cpu_default2.
1716 * doc/install.texi: Document --with-cpu for m68k.
1717 * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Define.
1718 * config/m68k/m68k-none.h (TARGET_CPU_DEFAULT, M68K_CPU_m68k)
1719 (M68K_CPU_m68000, M68K_CPU_m68010, M68K_CPU_m68020, M68K_CPU_m68030)
1720 (M68K_CPU_m68040, M68K_CPU_m68302, M68K_CPU_m68332, TARGET_DEFAULT)
1721 (ASM_CPU_DEFAULT_SPEC, CC1_CPU_DEFAULT_SPEC): Delete.
1722 (ASM_SPEC): Remove use of %(asm_cpu_default).
1723 (EXTRA_SPECS, SUBTARGET_EXTRA_SPECS, MULTILIB_DEFAULTS): Delete.
1724 * config/m68k/linux.h (TARGET_DEFAULT): Delete.
1725 (CPP_SPEC): Merge definitions. Do not handle __HAVE_68881__ here.
1726 * config/m68k/netbsd-elf.h (TARGET_OS_CPP_BUILTINS): Define
1727 __HAVE_FPU__ if TARGET_HARD_FLOAT.
1728 (TARGET_DEFAULT): Delete.
1729 (EXTRA_SPECS): Delete cpp_cpu_default_spec, cpp_cpu_spec,
1730 cpp_fpu_spec, asm_default_spec and netbsd_cpp_spec.
1731 (CPP_CPU_SPEC): Delete.
1732 (TARGET_VERSION): Merge definitions, using TARGET_68010 to pick
1733 the appropriate string.
1734 (CPP_CPU_DEFAULT_SPEC, ASM_DEFAULT_SPEC, CPP_FPU_SPEC): Delete.
1735 (CPP_SPEC): Define to NETBSD_CPP_SPEC.
1736 (ASM_SPEC): Don't use %(asm_default_spec).
1737 * config/m68k/m68k.c (TARGET_DEFAULT_TARGET_FLAGS): Remove
1738 TARGET_DEFAULT and add MASK_68881.
1739 * config/m68k/m68k.md: Remove mention of TARGET_DEFAULT from comments.
1741 2007-01-12 Richard Sandiford <richard@codesourcery.com>
1743 * config.gcc (m68010-*-netbsdelf*): Add MASK_68010.
1744 (m68k*-*-netbsdelf*, m68k*-*-openbsd*, m68k*-linux*): Add
1745 MASK_68010 alongside MASK_68020.
1746 * doc/invoke.texi: Document -m68010.
1747 * config/m68k/m68k.opt (m68010): New.
1748 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Define mc68010
1750 (TUNE_68010): New macro.
1751 * config/m68k/m68k-none.h (M68K_CPU_m68k, M68K_CPU_m68010)
1752 (M68K_CPU_m68020, M68K_CPU_m68030, M68K_CPU_m68040)
1753 (M68K_CPU_m68332): Add MASK_68010.
1754 * config/m68k/linux.h (TARGET_DEFAULT): Add MASK_68010 to
1755 fallback definition.
1756 * config/m68k/netbsd-elf.h (CPP_CPU_SPEC): Remove now-redundant
1758 * config/m68k/m68k.c (MASK_ALL_CPU_BITS): Add MASK_68010.
1759 (m68k_handle_option): Handle OPT_m68010. Add MASK_68010
1760 to all entries that use MASK_68020.
1761 (output_move_simode_const, output_move_himode, output_move_qimode)
1762 (output_move_stricthi, output_move_strictqi): Use TARGET_68010
1763 instead of TARGET_68020 to select clr behavior. Remove comment
1764 about there being no TARGET_68010.
1765 * config/m68k/m68k.md: Likewise throughout.
1767 2007-01-12 Julian Brown <julian@codesourcery.com>
1769 * config/m68k/m68k.h (TARGET_ISAB): New macro.
1770 * config/m68k/m68k.c: Use TARGET_ISAB rather than TARGET_CFV4.
1771 * config/m68k/m68k.md: Likewise.
1773 2007-01-12 Julian Brown <julian@codesourcery.com>
1775 * config/m68k/m68k.h (LEGITIMATE_INDEX_P, LEGITIMIZE_ADDRESS): Use
1776 TARGET_COLDFIRE_FPU instead of TARGET_CFV4E.
1778 2007-01-12 Julian Brown <julian@codesourcery.com>
1780 * config/m68k/m68k.h (TUNE_68040_60): New macro.
1781 * config/m68k/m68k.c (standard_68881_constant_p): Use it.
1782 * config/m68k/m68k.md: Likewise.
1784 2007-01-12 Julian Brown <julian@codesourcery.com>
1785 Richard Sandiford <richard@codesourcery.com>
1787 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Use TUNE_68030
1788 instead of TARGET_68030, TUNE_68040 instead of TARGET_68040,
1789 TUNE_68060 instead of TARGET_68060 and TUNE_CPU32 instead of
1791 (TARGET_CPU32): Rename to...
1792 (TUNE_CPU32): ...this.
1793 (TUNE_68000_10, TUNE_68030, TUNE_68040, TUNE_68060)
1794 (TUNE_CFV2): New macros.
1795 * config/m68k/netbsd-elf.h (LONG_DOUBLE_TYPE_SIZE): Simplify;
1796 remove conditions that are implied by TARGET_68020.
1797 * config/m68k/m68k.c (m68k_output_function_prologue): Use TUNE_68040
1798 instead of TARGET_68040 and TUNE_CPU32 instead of TARGET_CPU32.
1799 (m68k_output_function_epilogue): Likewise.
1800 (m68k_rtx_costs): Likewise. Use TUNE_68060 instead of TARGET_68060
1801 and TUNE_CFV2 instead of TARGET_5200. Use TUNE_68000_10 instead of
1802 "!TARGET_68020 && !TARGET_COLDFIRE" to choose between 68000 and
1803 non-68000 timings. Refactor multiplication and division costs.
1804 (output_addsi3): Use TUNE_68040 instead of TARGET_68040 and
1805 TUNE_CPU32 instead of TARGET_CPU32.
1806 (standard_68881_constant_p): Use TUNE_68040 instead of TARGET_68040
1807 and TUNE_68060 instead of TARGET_68060.
1808 * config/m68k/m68k.md: Use TUNE_68040 instead of TARGET_68040,
1809 TUNE_68060 instead of TARGET_68060, and TUNE_CPU32 instead of
1811 (movsi_const0): Use TUNE_68000_10 rather than "!TARGET_68020
1812 && !TARGET_COLDFIRE" to choose between moveq and clr.
1813 Likewise in the unnamed movsf pattern.
1814 (ashlsi_17_24, lshrsi_17_24): Guard with TUNE_68000_10 rather than
1815 "!TARGET_68020 && !TARGET_COLDFIRE". Likewise the unnamed
1818 2007-01-12 Richard Sandiford <richard@codesourcery.com>
1820 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Increase amount
1821 of tabbing before backslashes.
1823 2007-01-11 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
1825 * pa-linux.h (ASM_OUTPUT_INTERNAL_LABEL): Undefine.
1826 * pa.h (ASM_OUTPUT_LABEL): Output colon when using GAS.
1827 (ASM_OUTPUT_INTERNAL_LABEL): Define.
1829 2007-01-11 Zdenek Dvorak <dvorakz@suse.cz>
1831 * tree-ssa-loop-ivopts.c (extract_cond_operands): Split from
1832 find_interesting_uses_cond.
1833 (find_interesting_uses_cond): Use extract_cond_operands.
1834 (rewrite_use_compare): Use extract_cond_operands and
1835 force_gimple_operand_bsi. Do not call update_stmt.
1836 (determine_use_iv_cost_condition): Use extract_cond_operands.
1837 Return cheaper of using original bound and changing the exit bound.
1839 2007-01-11 Zdenek Dvorak <dvorakz@suse.cz>
1841 PR tree-optimization/29516
1842 * tree-ssa-address.c (tree_mem_ref_addr, add_to_parts,
1843 most_expensive_mult_to_index, addr_to_parts,
1844 create_mem_ref, maybe_fold_tmr): Make the type of
1845 fields of TARGET_MEM_REF sizetype.
1846 (move_fixed_address_to_symbol, move_pointer_to_base):
1848 * tree.def (TARGET_MEM_REF): Add comment on types of
1851 2007-01-11 Joseph Myers <joseph@codesourcery.com>
1853 * c-common.c (vector_types_convertible_p): Treat opaque types as
1854 always convertible if they have the same size, but not otherwise.
1856 2007-01-11 Steven Bosscher <steven@gcc.gnu.org>
1858 * ifcvt.c (struct noce_if_info): Add comments to the fields.
1859 Remove the b_unconditional field.
1860 (noce_try_sign_mask): Do not look at b_unconditional.
1861 (noce_process_if_block): Do not use merge_if_blocks. Update
1862 the CFG here. Do not set b_unconditional.
1863 (cond_move_process_if_block): Likewise.
1864 (find_cond_trap): Likewise.
1865 (check_cond_move_block): Require simple jump insns at the end
1868 2007-01-11 Jan Hubicka <jh@suse.cz>
1870 PR tree-optimization/1046
1871 * tree-tailcall.c (suitable_for_tail_call_opt_p): Use TREE_ADDRESSABLE
1872 when alias info is not ready.
1873 (pass_tail_recursion): Do not require aliasing.
1874 * tree-ssa-copyrename.c (pass_rename_ssa_cop): Likewise.
1875 * tree-ssa-ccp.c (pass_ccp, pass_fold_builtins): Likewise.
1876 * tree-ssa-copy.c (pass_copy_prop): Likewise.
1877 * tree-ssa-forwprop.c (pass_forwprop): Likewise.
1878 * tree-ssa-dce.c (pass_dce, pass_dce_loop, pass_cd_dce): Likewise.
1879 * passes.c (init_optimization_passes): Execute rename_ssa_copies,
1880 ccp, forwprop, copy_prop, merge_phi, copy_prop, dce and tail recursion
1882 * tree-ssa-operands.c (add_virtual_operand, get_indirect_ref_operand):
1883 When aliasing is not build, mark statement as volatile.
1885 2007-01-11 Tom Tromey <tromey@redhat.com>
1887 PR preprocessor/15185, PR preprocessor/20989:
1888 * doc/cppopts.texi <-MT>: Update description of algorithm for
1889 computing default target.
1890 <-M, -MD>: Reword "basename" text.
1892 2007-01-11 Roger Sayle <roger@eyesopen.com>
1894 * builtins.c (expand_builtin_pow, expand_builtin_powi,
1895 fold_builtin_cabs, fold_builtin_sqrt, fold_builtin_trunc,
1896 fold_builtin_floor, fold_builtin_ceil, fold_builtin_round,
1897 fold_builtin_int_int_roundingfn, fold_builtin_bitop,
1898 fold_builtin_bswap, real_constp, fold_builtin_pow,
1899 fold_builtin_powi, fold_builtin_signbit, fold_builtin_copysign,
1900 do_mpfr_arg1, do_mpfr_arg2, do_mpfr_arg3, do_mpfr_sincos): Replace
1901 uses of the macro TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
1902 * convert.c (convert_to_pointer): Likewise.
1903 * expr.c (highest_pow2_factor, expand_expr_real_1): Likewise.
1904 * fold-const.c (force_fit_type, fold_negate_expr, int_const_binop,
1905 const_binop, fold_convert_const_int_from_int,
1906 fold_convert_const_int_from_real,
1907 fold_convert_const_real_from_real, sign_bit_p,
1908 optimize_minmax_comparison, extract_muldiv_1, fold_div_compare,
1909 fold_sign_changed_comparison, fold_unary, fold_comparison,
1910 fold_binary, multiple_of_p, tree_Expr_non_zero_p,
1911 fold_negate_const, fold_abs_const, fold_not_const): Likewise.
1912 * print-tree.c (print_node_brief, print_node): Likewise.
1913 * stor-layout.c (place_field, layout_type): Likewise.
1914 * tree-chrec.c (keep_cast): Likewise.
1915 * tree.c (build_vector, build_real, build_real_from_int_cst,
1916 build_complex): Likewise.
1918 2007-01-11 Roger Sayle <roger@eyesopen.com>
1920 * tree.h (TREE_CONSTANT_OVERFLOW): Obsolete. For the time being,
1921 treat TREE_CONSTANT_OVERFLOW as a synonym of TREE_OVERFLOW.
1923 2007-01-11 Paolo Bonzini <bonzini@gnu.org>
1925 * configure.ac (strict1_warn): Rename to strict_warn.
1926 (WERROR, --enable-werror, symlink hacks, stage1_cflags,
1927 cc_set_by_configure, quoted_cc_set_by_configure,
1928 stage_prefix_set_by_configure, quoted_stage_prefix_set_by_configure,
1929 all_boot_languages, all_stagestuff): Remove.
1930 (target_list): Remove bootstrap targets.
1931 * Makefile.in (quickstrap): Unconditionally make a synonym of all.
1932 (BOOT_LANGUAGES, STAGE1_CFLAGS, STAGE1_CHECKING,
1933 REMAKEFLAGS, FLAGS_TO_PASS, PREPEND_DOTDOT_TO_RELATIVE_PATHS,
1934 SUBDIR_FLAGS_TO_PASS, WERROR_FLAGS, STRICT2_WARN, LANG_STAGESTUFF,
1935 VOL_FILES, POSTSTAGE1_FLAGS_TO_PASS, STAGE2_FLAGS_TO_PASS,
1936 STAGEPROFILE_FLAGS_TO_PASS, STAGEFEEDBACK_FLAGS_TO_PASS, stage1_build,
1937 stage1_copy, stage2_build, stage2_copy, stageprofile_build,
1938 stageprofile_copy, stage3_build, stage3_copy, stagefeedback_build,
1939 stagefeedback_copy, stage4_build, clean_s1, clean_sw, bootstrap,
1940 bootstrap-lean, bootstrap2, bootstrap2-lean, bootstrap3,
1941 bootstrap3-lean, bootstrap4, bootstrap4-lean, unstage1, unstage2,
1942 unstage3, unstage4, unstageprofile, unstagefeedback, restage, restage2,
1943 restage3, restage4, restageprofile, restagefeedback, bubbleestrap,
1944 cleanstrap, unstrap, restrap, *compare, *compare3, *compare4,
1945 *compare-lean, *compare3-lean, *compare4-lean, stage1-start, stage1,
1946 stage2-start, stage2, stage3-start, stage3, stage4-start, stage4,
1947 stageprofile-start, stageprofile, stagefeedback-start, stagefeedback,
1948 risky-stage1, risky-stage2, risky-stage3, risky-stage4): Remove.
1949 (ORDINARY_FLAGS_TO_PASS): Rename to FLAGS_TO_PASS.
1950 (STAGECOPYSTUFF, STAGEMOVESTUFF): Consolidate into MOSTLYCLEANFILES.
1951 (mostlyclean): Adjust.
1952 (clean, distclean): Don't mention bootstrap stuff.
1953 * configure: Regenerate.
1954 * ada/config-lang.in, cp/config-lang.in, forttran/config-lang.in,
1955 java/config-lang.in, objc/config-lang.in, objcp/config-lang.in,
1956 treelang/config-lang.in (stagestuff): Remove.
1957 * doc/sourcebuild.texi (stage1, stage2, stage3, stage4,
1958 stageprofile, stagefeedback, stagestuff): Remove mention.
1960 2007-01-11 Nick Clifton <nickc@redhat.com>
1962 * config/mcore/predicates.md (mcore_general_movesrc_operand):
1964 (mcore_general_movdst_operand): Do not accept CONST_INTs.
1965 (mcore_arith_K_S_operand): Run the test for the S constraint not
1966 the test for the M constraint.
1967 (mcore_addsub_operand): Do not accept integer values that are
1968 larger than 32 bits.
1969 * config/mcore/mcore.md: Remove unused constraints from split.
1970 (andsi3): Use HOST_WIDE_INT instead of int to hold an INTVAL.
1972 (allocate_stack): Likewise.
1973 * config/mcore/mcore.c (mcore_print_operand): Restrict output of P
1974 operands to 32 bits.
1975 (mcore_const_costs): Use HOST_WIDE_INT instead of int to hold an
1977 (mcore_and_cost, mcore_modify_comparison, const_ok_for_mcore,
1978 mcore_const_ok_for_inline, mcore_const_trick_uses_not,
1979 try_constant_tricks, mcore_num_ones, mcore_num_zeros,
1980 mcore_output_bclri, mcore_output_andn, output_inline_const,
1981 mcore_output_move, mcore_output_movedouble): Likewise.
1982 (mcore_output_cmov): Use CONST_OK_FOR_M and CONST_OK_FOR_N.
1983 (output_inline_const): Likewise.
1984 (output_inline_const): Fix format strings used in sprintf
1986 * config/mcore/mcore-protos.h: Update prototypes for changed
1987 functions in mcore.c.
1988 * config/mcore/mcore.h (CONST_OK_FOR_I): Cast values to
1989 HOST_WIDE_INT and not int.
1990 (CONST_OK_FOR_J, CONST_OK_FOR_K, CONST_OK_FOR_L, CONST_OK_FOR_M,
1991 CONST_OK_FOR_N): Likewise.
1992 (LEGITIMATE_CONSTANT_P): Also check CONSTANT_P.
1993 (GO_IF_LEGITIMATE_INDEX): Use HOST_WIDE_INT instead of int to hold
1996 2007-01-10 Jan Hubicka <jh@suse.cz>
1998 * tree-vrp.c (remove_range_assertions): Release defs.
1999 * tree-ssa-loop-ivopts.c (rmeove_statement): Likewise.
2000 * tree-ssa-dom.c (remove_stmt_or_phi): Likewise.
2002 2007-01-10 Paul Brook <paul@codesourcery.com>
2004 * config/arm/arm.c (arm_rtx_costs_1): Handle mutiply-subtract.
2005 * config/arm/arm.md (mulsi3subsi): New insn.
2007 2007-01-10 Zdenek Dvorak <dvorakz@suse.cz>
2009 * tree-ssa-loop-manip.c (tree_unroll_loop): Make it a wrapper over ...
2010 (tree_transform_and_unroll_loop): New.
2011 * tree-flow.h (transform_callback, tree_transform_and_unroll_loop):
2014 2007-01-10 Robert Kennedy <jimbob@google.com>
2016 * fold-const.c (fold_comparison): Fold comparisons like (x *
2017 1000 < 0) to (x < 0).
2019 2007-01-10 Ian Lance Taylor <iant@google.com>
2021 * tree-pretty-print.c (dump_generic_node): Print parentheses when
2022 operands have the same priority.
2024 2007-01-10 Tom Tromey <tromey@redhat.com>
2026 * fold-const.c (fold_truthop): Don't check can_use_bit_fields_p.
2027 (fold_binary): Likewise.
2028 * langhooks.c (lhd_can_use_bit_fields_p): Removed.
2029 * langhooks-def.h (lhd_can_use_bit_fields_p): Removed.
2030 (LANG_HOOKS_CAN_USE_BIT_FIELDS_P): Removed.
2031 (LANG_HOOKS_INITIALIZER): Remove LANG_HOOKS_CAN_USE_BIT_FIELDS_P.
2032 * langhooks.h (struct lang_hooks): Removed field
2033 'can_use_bit_fields_p'.
2035 2007-01-10 Ralf Corsépius <ralf.corsepius@rtems.org>
2037 * config/bfin/t-bfin, config/bfin/t-bfin-elf: Remove GCC_CFLAGS.
2039 2007-01-10 Razya Ladelsky <razya@il.ibm.com>
2041 * function.c (get_last_funcdef_no): New function.
2042 * function.h (get_last_funcdef_no): Declare.
2043 * tree-inline.c (initialize_cfun): Add initialization.
2044 (tree_function_versioning): Cleanup.
2046 2007-01-10 Jan Hubicka <jh@suse.cz>
2048 * tree-inline.c (setup_one_parameter): Do not propagate into abnormal
2051 2007-01-10 Sa Liu <saliu@de.ibm.com>
2052 Ben Elliston <bje@au.ibm.com>
2054 * spu.h (STACK_SAVE_AREA): Use VOIDmode for SAVE_FUNCTION, SImode
2055 for SAVE_NONLOCAL and Pmode for any other save level.
2056 * spu-protos.h (spu_restore_stack_block): Declare.
2057 * spu.md (save_stack_block): Remove.
2058 (restore_stack_block): Call spu_restore_stack_block.
2059 * spu.c (spu_restore_stack_block): New function.
2060 (spu_expand_epilogue): Remove old comment.
2062 2007-01-09 Zdenek Dvorak <dvorakz@suse.cz>
2064 PR tree-optimization/30322
2065 * tree-ssa-loop-ivopts.c (fold_affine_expr, iv_value): Removed.
2066 (cand_value_at): Return the value as aff_tree.
2067 (may_eliminate_iv): Convert the bound from aff_tree to tree.
2068 * tree-affine.c (aff_combination_add_cst, aff_combination_add_product,
2069 aff_combination_mult): New functions.
2070 (aff_combination_add): Use aff_combination_add_cst.
2071 (aff_combination_convert): Allow conversions to a wider type.
2072 (tree_to_aff_combination): Handle BIT_NOT_EXPR.
2073 * tree-affine.h (aff_combination_mult): Declare.
2075 2007-01-09 Carlos O'Donell <carlos@codesourcery.com>
2077 * doc/tm.texi: Update documentation to reflect reality of exec
2078 and start file search behaviours. Update copyright year.
2079 * doc/invoke.texi: Explain how GCC_EXEC_PREFIX is used to find
2080 header file directories.
2082 2007-01-09 Uros Bizjak <ubizjak@gmail.com>
2084 * config/i386/i386.md (*sinxf2): Rename to *sinxf2_i387.
2085 (*cosxf2): Rename to cosxf2_i387.
2086 (*sindf2, *sinsf2): Extend operand 1 to XFmode. Macroize patterns
2087 using X87MODEF12 mode macro. Rename patterns to
2088 *sin_extend<mode>xf2_i387. Use SSE_FLOAT_MODE_P to disable patterns
2090 (*cosdf2, *cossf2): Ditto.
2091 (sincosdf3, sincossf3): Ditto. Rewrite corresponding splitters
2092 to match extended input operands.
2093 (sincos<mode>3): New expander.
2094 (*sinextendsfdf2, *cosextendsfdf2, *sincosextendsfdf3): Remove
2095 insn patterns and corresponding splitters.
2097 2007-01-09 Kaz Kojima <kkojima@gcc.gnu.org>
2099 * config/sh/t-linux (TARGET_LIBGCC2_CFLAGS): Delete.
2100 (SHLIB_MAPFILES, SHLIB_LINK, SHLIB_INSTALL): Likewise.
2102 2007-01-09 Nicolas Pitre <nico@cam.org>
2105 * arm/ieee754-df.S (Lad_s): Also test the low word of X for zero.
2107 2007-01-08 Geoffrey Keating <geoffk@apple.com>
2109 * target.h (struct gcc_target): New field library_rtti_comdat.
2110 * target-def.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): New.
2111 (TARGET_CXX): Add TARGET_CXX_LIBRARY_RTTI_COMDAT.
2112 * doc/tm.texi (C++ ABI): Document TARGET_CXX_LIBRARY_RTTI_COMDAT.
2113 * config/darwin.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
2115 2007-01-08 Geoffrey Keating <geoffk@apple.com>
2117 * doc/invoke.texi (Optimize Options): Correct description of -O0.
2119 2007-01-08 Richard Guenther <rguenther@suse.de>
2121 * tree.h (force_fit_type_double): Export.
2122 (force_fit_type): Remove.
2123 * fold-const.c (force_fit_type_double): New function.
2124 (force_fit_type): Remove.
2125 (int_const_binop): Use it.
2126 (fold_convert_const_int_from_int): Likewise.
2127 (fold_convert_const_int_from_real): Likewise.
2128 (fold_div_compare): Likewise.
2129 (fold_sign_changed_comparison): Likewise.
2130 (fold_unary): Likewise.
2131 (fold_negate_const): Likewise.
2132 (fold_abs_const): Likewise.
2133 (fold_not_const): Likewise.
2134 * c-common.c (shorten_compare): Use force_fit_type_double.
2135 * convert.c (convert_to_pointer): Likewise.
2137 2007-01-08 Richard Guenther <rguenther@suse.de>
2139 * tree.h (build_int_cst_wide_type): Export.
2140 * tree.c (build_int_cst_wide_type): New function.
2141 (build_int_cst_wide): Fix comment.
2142 * builtins.c (fold_builtin_object_size): Use build_int_cst
2143 to build -1 or 0 of the correct type. Use fit_double_type
2144 to check for overflow.
2145 * fold-const.c (optimize_bit_field_compare): Use build_int_cst_type
2147 (decode_field_reference): Likewise.
2148 (all_ones_mask_p): Likewise.
2149 (native_interpret_int): Use build_int_cst_wide_type.
2150 (fold_binary): Use build_int_cst_type to build an all-ones
2152 * stor-layout.c (set_sizetype): Use build_int_cst_wide_type.
2154 2007-01-08 Daniel Jacobowitz <dan@codesourcery.com>
2156 * config/pa/t-pa64 (libgcc_stub.a): Use $(T).
2158 2007-01-09 Ben Elliston <bje@au.ibm.com>
2160 * genautomata.c (STATS_OPTION): New option.
2161 (stats_flag): New flag.
2162 (gen_automata_option): Handle it.
2163 (initiate_automaton_gen): Ditto.
2164 (write_automata): Output statistics only if stats_flag is
2165 set. Likewise, output time statistics only if time_flag is set.
2166 * doc/md.texi (Processor pipeline description): Document new flag.
2168 2007-01-08 Richard Guenther <rguenther@suse.de>
2170 * builtins.c (fold_builtin_int_roundingfn): Use fit_double_type.
2171 * tree.c (build_int_cst_type): Likewise.
2172 (size_in_bytes): Don't call force_fit_type on the result.
2173 (int_fits_type_p): Use fit_double_type.
2174 * fold-const.c (fit_double_type): New function.
2175 (force_fit_type): Use it.
2176 * tree.h (fit_double_type): Export.
2178 2007-01-08 Jan Hubicka <jh@suse.cz>
2180 * tree-vectorizer.c (gate_increase_alignment): Fix return type.
2181 * ipa.c (function_and_variable_visibility): Fix return type.
2183 2007-01-08 Richard Guenther <rguenther@suse.de>
2185 * tree-ssa-ccp.c (maybe_fold_offset_to_array_ref): Use type
2186 of offset to build the index.
2187 * tree-pretty-print.c (dump_generic_node): Don't build negated
2188 const just for printing.
2189 * c-pretty-print.c (pp_c_integer_constant): Likewise.
2190 * builtins.c (fold_builtin_int_roundingfn): Check if result
2191 fits the type by using force_fit_type and comparing the result.
2192 * predict.c (predict_loops): Use compare_tree_int for comparison.
2193 * tree.c (build_int_cst): Fall back to integer_type_node for
2195 (build_int_cst_wide): Assert type is non-null.
2197 2007-01-08 Roberto Costa <roberto.costa@st.com>
2199 * tree-vrp.c (extract_range_from_cond_expr): New.
2200 (extract_range_from_expr): Handle COND_EXPR nodes used as expressions.
2201 * tree-ssa-ccp.c (get_maxval_strlen): Handle COND_EXPR nodes used
2203 (fold_stmt): Bug fix, avoid infinite recursion when folding COND_EXPRs.
2204 * tree-ssa-forwprop.c (simplify_cond, forward_propagate_into_cond,
2205 tree_ssa_forward_propagate_single_use_vars): Handle COND_EXPR nodes
2206 used as expressions.
2207 * tree-object-size.c (cond_expr_object_size): New.
2208 (collect_object_sizes_for): Handle COND_EXPR nodes used as expressions.
2210 2007-01-08 Jan Hubicka <jh@suse.cz>
2212 * tree-ssa-forwprop.c (forward_propagate_into_cond,
2213 tree_ssa_forward_propagate_single_use_va): Release defs of propagated
2216 2007-01-08 Richard Guenther <rguenther@suse.de>
2218 PR tree-optimization/23603
2219 * tree-vrp.c (set_value_range_to_truthvalue): New function.
2220 (extract_range_from_binary): Fall back to truthvalue instead of
2221 varying for TRUTH_*_EXPR.
2222 (extract_range_from_comparison): Fall back to truthvalue instead of
2224 (vrp_visit_phi_node): Don't adjust new range bounds to +INF/-INF
2225 if all visited PHI values were constant.
2227 2007-01-08 Jan Hubicka <jh@suse.cz>
2229 * cgraphunit.c (cgraph_process_new_functions): Reset reachable flag.
2230 (cgraph_analyze_function): break out from ...
2231 (cgraph_finalize_compilation_unit): ... here.
2232 (cgraph_expand_function): Remove forgoten commented out line.
2233 (cgraph_optimize): Analyze functions.
2235 2007-01-08 Jan Hubicka <jh@suse.cz>
2237 * tree-pas.h (TODO_remove_function): New flag.
2238 (TODO_update*): Renumber.
2239 (pass_ipa_increase_alignment,
2240 pass_ipa_function_and_variable_visibility): New passes.
2241 * cgraphunit.c (cgraph_increase_alignment): Move to tree-vectorizer.c
2242 (cgraph_function_and_variable_visibility): Move to ipa.c
2243 (cgraph_optimize): Don't call cgraph_function_and_variable_visibility,
2244 cgraph_increase_alignment.
2245 * ipa-inline.c (cgraph_decide_inlining): Don't push timevar.
2246 (cgraph_decide_inlining_incrementally): Push TV_INTEGRATION before
2247 calling tree-inline.
2248 (cgraph_early_inlining): Do not call cgraph_remove_unreachable_nodes.
2249 (pass_ipa_inline, pass_early_ipa_inlining): Set TODO_remove_functions
2250 * tree-vectorizer.c (increase_alignment): Move here from cgraphunit.c
2251 (gate_increase_alignment): New function.
2252 (pass_ipa_increase_alignment): New pass.
2253 * ipa.c: Inline tree-pass.h and timevar.h
2254 (function_and_variable_visibility): Move here from cgraphunit.c
2255 * tree-optimize.c (pass_early_local_passes): Add TODO_remove_functions.
2256 * passes.c (init_optimization_passes): Add the two new passes.
2257 (execute_todo): Handle cgraph_remove_functions.
2259 2007-01-08 Nick Clifton <nickc@redhat.com>
2261 * config/frv/predicates.md (reg_or_0_operand): Accept
2264 2007-01-08 Ralf Corsépius <ralf.corsepius@rtems.org>
2266 * config/bfin/rtems.h, config/bfin/t-rtems: New.
2267 * config.gcc: Add bfin*-rtems*.
2269 2007-01-08 Mark Shinwell <shinwell@codesourcery.com>
2271 * c.opt: Add -flax-vector-conversions.
2272 * c-typeck.c (convert_for_assignment): Pass flag to
2273 vector_types_convertible_p to allow emission of note.
2274 (digest_init): Likewise.
2275 * c-opts.c: Handle -flax-vector-conversions.
2276 * c-common.c (flag_lax_vector_conversions): New.
2277 (vector_types_convertible_p): Unless -flax-vector conversions
2278 has been passed, disallow conversions between vectors with
2279 differing numbers of subparts and/or element types. If such
2280 a conversion is disallowed, possibly emit a note on the first
2281 occasion only to inform the user of -flax-vector-conversions.
2282 The new last argument specifies this.
2283 * c-common.h (flag_lax_vector_conversions): New.
2284 (vector_types_convertible_p): Add extra argument.
2285 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Use
2286 char_type_node for V*QI type vectors.
2287 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins):
2288 Update to satisfy new typechecking rules.
2289 * config/rs6000/altivec.h (vec_cmple): Use vec_cmpge, for both
2291 * doc/invoke.texi (C Dialect Options): Document
2292 -flax-vector-conversions.
2294 2007-01-08 Mark Shinwell <shinwell@codesourcery.com>
2296 PR tree-optimization/29877
2297 * tree-ssa-ter.c (is_replaceable_p): Deem assignments with
2298 a register variable on the RHS to not be replaceable.
2300 2007-01-08 Chen Liqin <liqin@sunnorth.com.cn>
2301 * config/score/t-score-elf (MULTILIB_OPTIONS): Change.
2302 * config/score/predicates.md (const_uimm5, sr0_operand, const_simm12,
2303 const_simm15, const_pow2, const_npow2): Added.
2304 * config/score/misc.md (insv, extv, extzv, movmemsi,
2305 move_lbu_a/b, mov_lhu_a/b etc): Added and fix some bug.
2306 * config/score/score.c (score_address_cost, score_select_cc_mode):
2308 Change CONST_OK_FOR_LETTER_P/EXTRA_CONSTRAINT define.
2309 Update score_rtx_costs for MACRO TARGET_RTX_COSTS.
2310 Update score_print_operand.
2311 * config/score/score.h (DATA_ALIGNMENT, SELECT_CC_MODE): Added.
2312 Adjust register allocate order and update some macro define.
2313 * config/score/score-mdaux.c (mdx_unaligned_load, mdx_unsigned_store,
2314 mdx_block_move_straight, mdx_block_move_loop_head,
2315 mdx_block_move_loop_body, mdx_block_move_loop_foot, mdx_block_move_loop,
2316 mdx_block_move): Added.
2317 (mdx_movsicc, mdp_select_add_imm, mdp_select, mds_zero_extract_andi,
2318 mdp_limm): Updated and fix some bug and typo.
2319 * config/score/score.md (movqi/hi/si, add/sub/zero/ext): Updated.
2320 (movsf, movdf, doloop_end): Added.
2322 2007-01-08 Kazu Hirata <kazu@codesourcery.com>
2324 * config/arm/arm.c, config/arm/arm.h, config/arm/arm.md,
2325 config/arm/thumb2.md: Fix comment typos.
2326 * doc/extend.texi: Fix a typo.
2328 2007-01-07 Eric Christopher <echristo@apple.com>
2330 * configure.ac: Check for __stack_chk_fail for darwin.
2331 * configure: Regenerate.
2333 2007-01-07 Richard Guenther <rguenther@suse.de>
2335 * tree-vrp.c (extract_range_from_assert): CSE calls to
2336 compare_values where possible.
2337 (extract_range_from_unary_expr): Likewise.
2339 2007-01-07 Anatoly Sokolov <aesok@post.ru>
2341 * config/avr/avr-protos.h (call_insn_operand): Delete prototype.
2342 * config/avr/avr.c (call_insn_operand): Delete function.
2343 * config/avr/avr.md (*pushqi, *pushhi, *pushsi, *pushsf): Use REG_SP
2344 instead of register number. Use predicates.
2345 * config/avr/predicates.md (const0_operand, reg_or_0_operand,
2346 call_insn_operand): Add.
2348 2007-01-06 Jan Hubicka <jh@suse.cz>
2350 * tree-pass.h (pass_build_cgraph_edges): Declare.
2351 * cgraphunit.c (record_refernece): Move to cgraphbuild.c
2352 (visited_nodes): Remove.
2353 (cgraph_create_edges): Move to cgraphbuild.c; rename to
2354 build_cgrpah_edges; make visited_nodes local.
2355 (cgraph_process_new_functions): DO not call initialize_inline_failed.
2356 (record_references_in_initializer): Move to cgraphbuild.c
2357 (initialize_inline_failed, rebuild_cgraph_edges,
2358 pass_rebuild_cgraph_edges): Move to cgraphbuild.c.
2359 (verify_cgraph_node): Make visited_nodes local.
2360 (cgraph_analyze_function): Do not call cgraph_create_edges and
2361 initialize_inline_failed.
2362 (cgraph_expand_function): Do not call cgraph_lower_function;
2363 assert that function is already lowered.
2364 * Makefile.in (cgraphbuild.o): New.
2365 * passes.c (init_optimization_passes): Add pass_build_cgraph_edges
2366 at the end of lowering passes.
2368 2007-01-06 Steven Bosscher <steven@gcc.gnu.org>
2370 * ifcvt.c (cond_move_convert_if_block): New function, code
2371 factored out from...
2372 (cond_move_process_if_block): ...here. Call the new function
2373 on the THEN and ELSE blocks.
2374 (merge_if_block): Do not copy global_live_at_end, merge_blocks
2375 already takes care of this.
2377 2007-01-05 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
2380 * tree.h (TREE_OVERFLOW_P): New.
2381 * c-typeck.c (parser_build_unary_op): Warn only if result
2382 overflowed and operands did not.
2383 (parser_build_binary_op): Likewise.
2384 (convert_for_assignment): Remove redundant overflow_warning.
2385 * c-common.c (overflow_warning): Don't check or set TREE_OVERFLOW.
2387 2007-01-05 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
2389 * c-typeck.c (store_init_value): Split over two lines to follow
2390 the GNU coding style.
2392 2007-01-05 Benjamin Kosnik <bkoz@redhat.com>
2394 * c-cppbuiltin.c (c_cpp_builtins): __GXX_EXPERIMENTAL_CPP0X__ to
2395 __GXX_EXPERIMENTAL_CXX0X__.
2396 * doc/cpp.texi: Same.
2398 2007-01-05 Richard Guenther <rguenther@suse.de>
2401 * tree.c (get_narrower): Do not construct COMPONENT_REFs
2402 with mismatched types. Instead explicitly build a
2403 conversion NOP_EXPR.
2405 2007-01-05 Ian Lance Taylor <iant@google.com>
2407 * c-common.c (decl_with_nonnull_addr_p): New function.
2408 (c_common_truthvalue_conversion): Call it.
2409 * c-typeck.c (build_binary_op): Likewise.
2410 * c-common.h (decl_with_nonnull_addr_p): Declare.
2412 2007-01-05 Jakub Jelinek <jakub@redhat.com>
2415 * libgcc2.c (__divdc3): Compare c and d against 0.0 instead of
2418 2007-01-05 Joel Brobecker <brobecker@adacore.com>
2420 * doc/install.texi (Final install): Document the fact that
2421 the GNAT runtime should not be stripped.
2423 2007-01-04 Jan Hubicka <jh@suse.cz>
2425 * tree-inline.c (fold_marked_statements): Update operand caches
2426 and EH after folding
2428 2007-01-04 Ian Lance Taylor <iant@google.com>
2430 * c-common.c (check_function_nonnull): Whitespace fix.
2432 2007-01-04 Jan Hubicka <jh@suse.cz>
2434 * tree-optimize.c (execute_fixup_cfg): Correct previously mistakely
2435 comitted older version of patch.
2436 (pass_fixup_cfg): Add TODOs to verify flow and statements, dump
2437 function, celanup cfg and collect garbage.
2439 2007-01-04 Mike Stump <mrs@apple.com>
2441 * Makefile.in (mostlyclean): Don't remove libgcc anymore.
2444 2007-01-04 Eric Christopher <echristo@apple.com>
2446 * libgcc2.c (__bswapsi2): Use SItype.
2447 (__bswapdi2): Use DItype.
2448 * libgcc2.h: Update for above.
2450 2007-01-04 Paul Brook <paul@codesourcery.com>
2452 * config/arm/arm.md (arm_mulsi3, thumb_mulsi3, mulsi3_compare0,
2453 mulsi_compare0_scratch, mulsi3addsi, mulsi3addsi_compare0,
2454 mulsi3addsi_compare0_scratch, mulsidi3adddi, mulsidi3,
2455 umulsidi3, umulsidi3adddi, smulsi3_highpart,
2456 umulsi3_highpart): Make conditional on !arm_arch6.
2457 (arm_mulsi3_v6, thumb_mulsi3_v6, mulsi3_compare0_v6,
2458 mulsi_compare0_scratch_v6, mulsi3addsi_v6, mulsi3addsi_compare0_v6,
2459 mulsi3addsi_compare0_scratch_v6, mulsidi3adddi_v6, mulsidi3_v6,
2460 umulsidi3_v6, umulsidi3adddi_v6, smulsi3_highpart_v6,
2461 umulsi3_highpart_v6): New insns.
2463 2007-01-04 Roger Sayle <roger@eyesopen.com>
2465 * fold-const.c (fold_convert): When casting an expression to void,
2466 fold_ignored_result may discover a GIMPLE_MODIFY_STMT which doesn't
2467 have a type. Instead of attempting to build a NOP_EXPR, return
2468 these "special" trees directly.
2470 2007-01-04 Joseph Myers <joseph@codesourcery.com>
2472 * config/rs6000/rs6000.c (rs6000_rtx_costs): Make adjustment for
2473 MULT inside MINUS as either argument. Use rs6000_cost->dmul -
2474 rs6000_cost->fp not 0 as adjustment for outer NEG.
2476 2007-01-04 Jan Hubicka <jh@suse.cz>
2478 * cgraph.c (cgraph_release_function_body): New function.
2479 (cgraph_remove_node): Use it.
2480 * cgraph.h (cgraph_release_function_body): Declare.
2481 * cgraphunit.c (cgraph_expand_function): Use it.
2482 * ipa.c (cgraph_remove_unreahchable_nodes): Use it.
2483 * tree-ssa.c (delete_tree_ssa): Allow to be called before aliasing
2484 is initialized and while compilation of other function is running.
2485 * tree-optimize.c (execute_free_cfg_annotations): Move code to clear
2486 statement CFG annotations from here to ...
2487 * tree-cfg.c (delete_tree_cfg_annotations): ... here.
2489 2007-01-04 Zdenek Dvorak <dvorakz@suse.cz>
2491 * cfgloop.h (enum li_flags): Make the constants powers of two.
2493 2007-01-04 Jan Hubicka <jh@suse.cz>
2495 * tree-inline.c (copy_bb): Insert new statements to statements_to_fold
2497 (fold_marked_statements): New function.
2498 (optimize_inline_calls, tree_function_versioning): Fold new statements.
2499 * tree-inline.h (copy_body_data): Add statements_to_fold.
2501 2007-01-03 Daniel Jacobowitz <dan@codesourcery.com>
2503 * config.gcc: Mention libgcc/config.host.
2504 * Makefile.in: Update comments mentioning libgcc.
2505 (LIBGCC, INSTALL_LIBGCC, GCC_PARTS, mklibgcc): Delete.
2506 (all.cross, start.encap, rest.encap, rest.cross): Update
2507 dependencies for libgcc move.
2508 (libgcc.mk, LIBGCC_DEPS, libgcov.a, libgcc.a, stmp-multilib)
2509 (clean-target, clean-target-libgcc): Delete.
2510 (srcdirify, GCC_EXTRA_PARTS): New macros.
2511 (libgcc-support, libgcc.mvars): New rules.
2512 (distclean): Remove mention of mklibgcc.
2513 (install): Don't reference INSTALL_LIBGCC.
2514 (install-common): Don't reference EXTRA_PARTS.
2515 (install-libgcc, install-multilib): Delete rules.
2516 * mklibgcc.in: Delete file.
2517 * doc/configfiles.texi: Don't mention mklibgcc.
2519 * config/i386/t-darwin (SHLIB_VERPFX): Delete (moved to libgcc).
2520 * config/i386/t-darwin64 (SHLIB_VERPFX): Likewise.
2521 * config/rs6000/t-darwin (SHLIB_VERPFX): Likewise.
2522 * config/rs6000/t-ppccomm (TARGET_LIBGCC2_CFLAGS, SHLIB_MAPFILES)
2523 (mklibgcc, ldblspecs): Likewise.
2525 * config/i386/t-nwld (libgcc.def, libc.def, libpcre.def)
2526 (posixpre.def): Use $(T).
2527 (SHLIB_EXT, SHLIB_NAME, SHLIB_SLIBDIR_QUAL, SHLIB_DEF, SHLIB_MAP)
2528 (SHLIB_SRC, SHLIB_INSTALL): Delete.
2529 (SHLIB_LINK): Make dummy.
2530 * config/t-slibgcc-darwin: Delete contents except for dummy SHLIB_LINK.
2532 * config/frv/t-linux (EXTRA_MULTILIB_PARTS): Clear.
2534 * config/alpha/t-crtfm: Use $(T) in rules for EXTRA_PARTS.
2535 * config/alpha/t-vms, config/alpha/t-vms64, config/fr30/t-fr30,
2536 config/i386/t-rtems-i386, config/ia64/t-ia64, config/rs6000/t-beos,
2537 config/rs6000/t-newas, config/sparc/t-elf: Likewise.
2539 * configure.ac (all_outputs): Remove mklibgcc.
2540 * configure: Regenerated.
2542 2007-01-03 Josh Conner <jconner@apple.com>
2545 * calls.c (compute_argument_addresses): Set stack and stack_slot
2546 for partial args, too.
2547 (store_one_arg): Use locate.size.constant for the size when
2548 generating a save_area.
2550 2007-01-03 Robert Kennedy <jimbob@google.com>
2552 * tree-cfg.c (tree_merge_blocks): Release SSA_NAME phi results
2553 whose definitions are deleted due to basic block merging.
2555 2007-01-03 Paul Brook <paul@codesourcery.com>
2558 * config/arm/arm.c (output_return_instruction): Pop PC in interrupt
2560 (use_return_insn): Return 0 for Thumb interrupt functions.
2561 (print_multi_reg): Add rfe argument for IRQ returns.
2562 (arm_output_epilogue): Pop interrupt return address directly into PC.
2563 (arm_expand_prologue): Only adjust IRQ return address in Arm mode.
2565 2007-01-03 Paul Brook <paul@codesourcery.com>
2567 Merge from sourcerygxx-4_1.
2568 * config/arm/thumb2.md: New file.
2569 * config/arm/elf.h (JUMP_TABLES_IN_TEXT_SECTION): Return True for
2571 * config/arm/coff.h (JUMP_TABLES_IN_TEXT_SECTION): Ditto.
2572 * config/arm/aout.h (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
2573 (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump tables.
2574 * config/arm/aof.h (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump
2576 (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
2577 * config/arm/ieee754-df.S: Use macros for Thumb-2/Unified asm
2579 * config/arm/ieee754-sf.S: Ditto.
2580 * config/arm/arm.c (thumb_base_register_rtx_p): Rename...
2581 (thumb1_base_register_rtx_p): ... to this.
2582 (thumb_index_register_rtx_p): Rename...
2583 (thumb1_index_register_rtx_p): ... to this.
2584 (thumb_output_function_prologue): Rename...
2585 (thumb1_output_function_prologue): ... to this.
2586 (thumb_legitimate_address_p): Rename...
2587 (thumb1_legitimate_address_p): ... to this.
2588 (thumb_rtx_costs): Rename...
2589 (thumb1_rtx_costs): ... to this.
2590 (thumb_compute_save_reg_mask): Rename...
2591 (thumb1_compute_save_reg_mask): ... to this.
2592 (thumb_final_prescan_insn): Rename...
2593 (thumb1_final_prescan_insn): ... to this.
2594 (thumb_expand_epilogue): Rename...
2595 (thumb1_expand_epilogue): ... to this.
2596 (arm_unwind_emit_stm): Rename...
2597 (arm_unwind_emit_sequence): ... to this.
2598 (thumb2_legitimate_index_p, thumb2_legitimate_address_p,
2599 thumb1_compute_save_reg_mask, arm_dwarf_handle_frame_unspec,
2600 thumb2_index_mul_operand, output_move_vfp, arm_shift_nmem,
2601 arm_save_coproc_regs, thumb_set_frame_pointer, arm_print_condition,
2602 thumb2_final_prescan_insn, thumb2_asm_output_opcode, arm_output_shift,
2603 thumb2_output_casesi): New functions.
2604 (TARGET_DWARF_HANDLE_FRAME_UNSPEC): Define.
2605 (FL_THUMB2, FL_NOTM, FL_DIV, FL_FOR_ARCH6T2, FL_FOR_ARCH7,
2606 FL_FOR_ARCH7A, FL_FOR_ARCH7R, FL_FOR_ARCH7M, ARM_LSL_NAME,
2607 THUMB2_WORK_REGS): Define.
2608 (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv, arm_condexec_count,
2609 arm_condexec_mask, arm_condexec_masklen)): New variables.
2610 (all_architectures): Add armv6t2, armv7, armv7a, armv7r and armv7m.
2611 (arm_override_options): Check new CPU capabilities.
2612 Set new architecture flag variables.
2613 (arm_isr_value): Handle v7m interrupt functions.
2614 (user_return_insn): Return 0 for v7m interrupt functions. Handle
2616 (const_ok_for_arm): Handle Thumb-2 constants.
2617 (arm_gen_constant): Ditto. Use movw when available.
2618 (arm_function_ok_for_sibcall): Return false for v7m interrupt
2620 (legitimize_pic_address, arm_call_tls_get_addr): Handle Thumb-2.
2621 (thumb_find_work_register, arm_load_pic_register,
2622 legitimize_tls_address, arm_address_cost, load_multiple_sequence,
2623 emit_ldm_seq, emit_stm_seq, arm_select_cc_mode, get_jump_table_size,
2624 print_multi_reg, output_mov_long_double_fpa_from_arm,
2625 output_mov_long_double_arm_from_fpa, output_mov_double_fpa_from_arm,
2626 output_mov_double_fpa_from_arm, output_move_double,
2627 arm_compute_save_reg_mask, arm_compute_save_reg0_reg12_mask,
2628 output_return_instruction, arm_output_function_prologue,
2629 arm_output_epilogue, arm_get_frame_offsets, arm_regno_class,
2630 arm_output_mi_thunk, thumb_set_return_address): Ditto.
2631 (arm_expand_prologue): Handle Thumb-2. Use arm_save_coproc_regs.
2632 (arm_coproc_mem_operand): Allow POST_INC/PRE_DEC.
2633 (arithmetic_instr, shift_op): Use arm_shift_nmem.
2634 (arm_print_operand): Use arm_print_condition. Handle '(', ')', '.',
2636 (arm_final_prescan_insn): Use extract_constrain_insn_cached.
2637 (thumb_expand_prologue): Use thumb_set_frame_pointer.
2638 (arm_file_start): Output directive for unified syntax.
2639 (arm_unwind_emit_set): Handle stack alignment instruction.
2640 * config/arm/lib1funcs.asm: Remove default for __ARM_ARCH__.
2641 Add v6t2, v7, v7a, v7r and v7m.
2642 (RETLDM): Add Thumb-2 code.
2643 (do_it, shift1, do_push, do_pop, COND, THUMB_SYNTAX): New macros.
2644 * config/arm/arm.h (TARGET_CPU_CPP_BUILTINS): Define __thumb2__.
2645 (TARGET_THUMB1, TARGET_32BIT, TARGET_THUMB2, TARGET_DSP_MULTIPLY,
2646 TARGET_INT_SIMD, TARGET_UNIFIED_ASM, ARM_FT_STACKALIGN, IS_STACKALIGN,
2647 THUMB2_TRAMPOLINE_TEMPLATE, TRAMPOLINE_ADJUST_ADDRESS,
2648 ASM_OUTPUT_OPCODE, THUMB2_GO_IF_LEGITIMATE_ADDRESS,
2649 THUMB2_LEGITIMIZE_ADDRESS, CASE_VECTOR_PC_RELATIVE,
2650 CASE_VECTOR_SHORTEN_MODE, ADDR_VEC_ALIGN, ASM_OUTPUT_CASE_END,
2651 ADJUST_INSN_LENGTH): Define.
2652 (TARGET_REALLY_IWMMXT, TARGET_IWMMXT_ABI, CONDITIONAL_REGISTER_USAGE,
2653 STATIC_CHAIN_REGNUM, HARD_REGNO_NREGS, INDEX_REG_CLASS,
2654 BASE_REG_CLASS, MODE_BASE_REG_CLASS, SMALL_REGISTER_CLASSES,
2655 PREFERRED_RELOAD_CLASS, SECONDARY_OUTPUT_RELOAD_CLASS,
2656 SECONDARY_INPUT_RELOAD_CLASS, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P,
2657 TRAMPOLINE_SIZE, INITIALIZE_TRAMPOLINE, HAVE_PRE_INCREMENT,
2658 HAVE_POST_DECREMENT, HAVE_PRE_DECREMENT, HAVE_PRE_MODIFY_DISP,
2659 HAVE_POST_MODIFY_DISP, HAVE_PRE_MODIFY_REG, HAVE_POST_MODIFY_REG,
2660 REGNO_MODE_OK_FOR_BASE_P, LEGITIMATE_CONSTANT_P,
2661 REG_MODE_OK_FOR_BASE_P, REG_OK_FOR_INDEX_P, GO_IF_LEGITIMATE_ADDRESS,
2662 LEGITIMIZE_ADDRESS, THUMB2_LEGITIMIZE_ADDRESS,
2663 GO_IF_MODE_DEPENDENT_ADDRESS, MEMORY_MOVE_COST, BRANCH_COST,
2664 ASM_APP_OFF, ASM_OUTPUT_CASE_LABEL, ARM_DECLARE_FUNCTION_NAME,
2665 FINAL_PRESCAN_INSN, PRINT_OPERAND_PUNCT_VALID_P,
2666 PRINT_OPERAND_ADDRESS): Adjust for Thumb-2.
2667 (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv): New declarations.
2668 * config/arm/arm-cores.def: Add arm1156t2-s, cortex-a8, cortex-r4 and
2670 * config/arm/arm-tune.md: Regenerate.
2671 * config/arm/arm-protos.h: Update prototypes.
2672 * config/arm/vfp.md: Enable patterns for Thumb-2.
2673 (arm_movsi_vfp): Add movw alternative. Use output_move_vfp.
2674 (arm_movdi_vfp, movsf_vfp, movdf_vfp): Use output_move_vfp.
2675 (thumb2_movsi_vfp, thumb2_movdi_vfp, thumb2_movsf_vfp,
2676 thumb2_movdf_vfp, thumb2_movsfcc_vfp, thumb2_movdfcc_vfp): New.
2677 * config/arm/libunwind.S: Add Thumb-2 code.
2678 * config/arm/constraints.md: Update include Thumb-2.
2679 * config/arm/ieee754-sf.S: Add Thumb-2/Unified asm support.
2680 * config/arm/ieee754-df.S: Ditto.
2681 * config/arm/bpabi.S: Ditto.
2682 * config/arm/t-arm (MD_INCLUDES): Add thumb2.md.
2683 * config/arm/predicates.md (low_register_operand,
2684 low_reg_or_int_operand, thumb_16bit_operator): New.
2685 (thumb_cmp_operand, thumb_cmpneg_operand): Rename...
2686 (thumb1_cmp_operand, thumb1_cmpneg_operand): ... to this.
2687 * config/arm/t-arm-elf: Add armv7 multilib.
2688 * config/arm/arm.md: Update patterns for Thumb-2 and Unified asm.
2690 (UNSPEC_STACK_ALIGN, ce_count): New.
2691 (arm_incscc, arm_decscc, arm_umaxsi3, arm_uminsi3,
2692 arm_zero_extendsidi2, arm_zero_extendqidi2): New
2694 * config/arm/fpa.md: Update patterns for Thumb-2 and Unified asm.
2695 (thumb2_movsf_fpa, thumb2_movdf_fpa, thumb2_movxf_fpa,
2696 thumb2_movsfcc_fpa, thumb2_movdfcc_fpa): New insns.
2697 * config/arm/cirrus.md: Update patterns for Thumb-2 and Unified asm.
2698 (cirrus_thumb2_movdi, cirrus_thumb2_movsi_insn,
2699 thumb2_cirrus_movsf_hard_insn, thumb2_cirrus_movdf_hard_insn): New
2701 * doc/extend.texi: Document ARMv7-M interrupt functions.
2702 * doc/invoke.texi: Document Thumb-2 new cores+architectures.
2704 2007-01-03 Jakub Jelinek <jakub@redhat.com>
2706 * unwind-dw2.c (SIGNAL_FRAME_BIT, EXTENDED_CONTEXT_BIT): Define.
2707 (struct _Unwind_Context): Rename args_size to flags, remove
2708 signal_frame field, add a new args_size field and version field.
2709 (_Unwind_IsSignalFrame, _Unwind_SetSignalFrame,
2710 _Unwind_IsExtendedContext): New inline functions.
2711 (_Unwind_GetGR, _Unwind_SetGR, _Unwind_GetGRPtr, _Unwind_SetGRPtr):
2712 Assume by_value array is only present if _Unwind_IsExtendedContext.
2713 (_Unwind_GetIPInfo, execute_cfa_program, uw_frame_state_for): Use
2714 _Unwind_IsSignalFrame.
2715 (__frame_state_for): Initialize context.flags to EXTENDED_CONTEXT_BIT.
2716 (uw_update_context_1): Use _Unwind_SetSignalFrame.
2717 (uw_init_context_1): Initialize context->flags to
2718 EXTENDED_CONTEXT_BIT.
2719 * config/rs6000/linux-unwind.h (frob_update_context): Use
2720 _Unwind_SetSignalFrame.
2722 2007-01-03 Andrew Pinski <andrew_pinski@playstation.sony.com>
2725 * gimplify.c (gimplify_modify_expr_complex_part): Move below
2726 tree_to_gimple_tuple. Call tree_to_gimple_tuple when we need
2729 2007-01-03 Kazu Hirata <kazu@codesourcery.com>
2731 * config/i386/i386.h (NON_STACK_REG_P, REGNO_OK_FOR_SIREG_P,
2732 REGNO_OK_FOR_DIREG_P, REWRITE_ADDRESS, ASM_OPERAND_LETTER,
2733 RET, AT_SP): Remove.
2734 * config/i386/i386.md (*sse_prologue_save_insn): Use return
2737 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
2738 * cfgbuild.c (find_basic_blocks): Likewise.
2739 * cfgrtl.c (rtl_create_basic_block): Likewise.
2740 * function.c (temp_slots_at_level): Likewise.
2741 * reg-stack.c (stack_regs_mentioned): Likewise.
2742 * regclass.c (allocate_reg_info): Likewise.
2743 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
2744 set_bb_for_stmt, move_block_to_fn): Likewise.
2745 * tree-complex.c (tree_lower_complex): Likewise.
2746 * vec.h (VEC_safe_grow_cleared): New.
2748 * cgraphunit.c, tree-ssa-alias.c: Fix comment typos.
2750 2007-01-03 Zdenek Dvorak <dvorakz@suse.cz>
2752 * loop-unswitch.c (unswitch_loop): Pass probabilities to loopify.
2753 * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Pass probabilities
2755 * cfgloopmanip.c (scale_loop_frequencies): Export.
2756 (loopify): Scale the frequencies by prescribed coefficients.
2757 (set_zero_probability): New function.
2758 (duplicate_loop_to_header_edge): Improve updating of frequencies.
2759 (lv_adjust_loop_entry_edge, loop_version): Set probabilities
2760 and frequencies according to arguments.
2761 * tree-ssa-loop-manip.c (tree_unroll_loop): Set probabilities
2763 * cfg.c (scale_bbs_frequencies_int): Allow scaling the frequencies up.
2764 * modulo-sched.c (sms_schedule): Set probabilities for entering
2765 versioned loop correctly.
2766 * tree-vect-transform.c (vect_transform_loop): Ditto.
2767 * cfgloop.h (loopify, loop_version): Declaration changed.
2768 (scale_loop_frequencies): Declared.
2770 2007-01-02 Jan Hubicka <jh@suse.cz>
2772 * cgraph.c: Include tree-flow.h
2773 (cgraph_add_new-function): Handle IPA_SSA mode; execute
2775 * cgraph.h (enum cgraph_state): Add CGRAPH_STATE_IPA_SSA.
2776 * tree-pass.h (pass_all_early_optimizations): Declare.
2777 * cgraphunit.c (cgraph_process_new_functions): Add IPA_SSA; execute
2779 (cgraph_analyze_function): Do early_local_passes.
2780 * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
2781 Do not add referenced vars.
2782 * tree-optimize.c (gate_all_optimizations): Do not execute when not in
2784 (gate_all_early_local_passes): New gate.
2785 (pass_early_local_passes): Use new gate.
2786 (execute_early_local_optimizations): New functions.
2787 (gate_all_early_optimizations): New gate.
2788 (pass_all_early_optimizations): New pass.
2789 (execute_free_datastructures): Free SSA only when initialized.
2790 (gate_init_datastructures): Init only when optimizing.
2791 (tree_lowering_passes): Do early local passes when called late.
2792 * tree-profile.c (do_tree_profiling): Don't profile functions added
2794 (do_early_tree_profiling, pass_early_tree_profile): Kill.
2795 * tree-cfg.c (update_modified_stmts): Do not update when operands are
2797 * passes.c (init_optimizations_passes): Reorder so we go into SSA
2798 during early_local_passes.
2799 * Makefile.in (cgraph.o): Add dependency on tree-flow.h.
2802 2007-01-02 Carlos O'Donell <carlos@codesourcery.com>
2804 * Makefile.in: Update copyright year.
2806 2007-01-02 Carlos O'Donell <carlos@codesourcery.com>
2808 * Makefile.in: Export GCC_EXEC_PREFIX before calling $(RUNTEST)
2809 in $(lang_checks) and check-consistency targets.
2811 2007-01-02 Jan Hubicka <jh@suse.cz>
2813 * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
2814 Do not add referenced vars.
2815 * tree-cfg.c (update_modified_stmts): Do not update when SSA operands
2817 * passes.c (init_optimization_passes): Put mudflap_2 after
2818 free_datastructures.
2820 2007-01-02 Jan Hubicka <jh@suse.cz>
2822 * tree-optimize (execute_fixup_cfg): Set after_inlining flag.
2823 Set NOTHROW flag on call statements proved to be nothrow.
2824 Update statement of local calls so new pure/const functions are
2825 updated. Update_ssa when in ssa form. Mark PHI nodes of nonlocal
2827 (tree_rest_of_compilation): Register hooks and initialize bitmap
2828 early. Do not set after_inlining flag.
2830 2007-01-02 Steve Ellcey <sje@cup.hp.com>
2832 * sbitmap.c (HOST_BITS_PER_LONG_LONG): Change to
2833 HOST_BITS_PER_LONGLONG
2835 2007-01-02 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
2838 * c-typeck.c (store_init_value): Don't emit pedantic overflow
2839 warning for non-static initializers.
2841 2007-01-02 Steven Bosscher <steven@gcc.gnu.org>
2843 * config/alpha/alpha.md, arm/arm.c, darwin.c, frv/frv.md,
2844 m32r/m32r.c, m32r/m32r.c, mn10300/mn10300.md, pa/pa.c,
2845 rs6000/rs6000.c, s390/s390.md, sh/sh.md, sparc/sparc.c:
2846 Always use set_unique_reg_note to add REG_EQUAL notes.
2848 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
2851 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
2853 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
2854 * cfgbuild.c (find_basic_blocks): Likewise.
2855 * cfgrtl.c (rtl_create_basic_block): Likewise.
2856 * function.c (temp_slots_at_level): Likewise.
2857 * reg-stack.c (stack_regs_mentioned): Likewise.
2858 * regclass.c (allocate_reg_info): Likewise.
2859 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
2860 set_bb_for_stmt, move_block_to_fn): Likewise.
2861 * tree-complex.c (tree_lower_complex): Likewise.
2862 * vec.h (VEC_safe_grow_cleared): New.
2864 2007-01-02 Ian Lance Taylor <iant@google.com>
2866 * c-common.c (c_common_truthvalue_conversion): When warning about
2867 using an assignment as a truth value, set TREE_NO_WARNING.
2869 2007-01-02 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
2872 * c.opt (Wold-style-declaration): New.
2873 * doc/invoke.texi (C-only Warning Options): New.
2874 (Wold-style-declaration): Document it.
2875 (Wextra): Enabled by -Wextra.
2876 * c-opts.c (c_common_post_options): Enabled by -Wextra.
2877 * c-decl.c (declspecs_add_scspec): Replace -Wextra with
2878 -Wold-style-declaration.
2880 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
2882 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
2883 * cfgbuild.c (find_basic_blocks): Likewise.
2884 * cfgrtl.c (rtl_create_basic_block): Likewise.
2885 * function.c (temp_slots_at_level): Likewise.
2886 * reg-stack.c (stack_regs_mentioned): Likewise.
2887 * regclass.c (allocate_reg_info): Likewise.
2888 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
2889 set_bb_for_stmt, move_block_to_fn): Likewise.
2890 * tree-complex.c (tree_lower_complex): Likewise.
2891 * vec.h (VEC_safe_grow_cleared): New.
2893 2007-01-02 Douglas Gregor <doug.gregor@gmail.com>
2895 * c-common.c(c_common_nodes_and_builtins): Since variants of
2896 void_type_node get built before it is given a name, we need to
2897 give those variants the name, too.
2898 (complete_array_type): We need to work with the canonical main
2899 type of the array, from which we will build the qualified version.
2900 * params.def (PARAM_VERIFY_CANONICAL_TYPES): New.
2901 * print-tree.c (print_node): Display canonical type information
2903 * stor-layout.c (layout_type): When we don't know the
2904 alignment of a type for which we're building an array, we end up
2905 guessing wrong, so make the type require structural equality.
2906 * tree.c (make_node_stat): When we build a new type, it is its
2908 (build_type_attribute_qual_variant): When building an attribute
2909 variant, its canonical type is the non-attribute variant. However,
2910 if the attributes are target-dependent and they differ, we need to
2911 use structural equality checks for this type.
2912 (build_qualified_type): A qualified type is not equivalent to its
2913 unqualified variant; set the canonical type appropriately.
2914 (build_distinct_type_copy): When building a distinct type from
2915 another type, the new type is its own canonical type.
2916 (build_variant_type_copy): When building a new type variant, we
2917 assume that it is equivalent to the original type.
2918 (build_pointer_type_for_mode): When building a pointer type, also
2919 build a canonical type pointer.
2920 (build_reference_type_for_mode): When building a reference type,
2921 also build a canonical type reference.
2922 (build_index_type): When we can't hash an index type (e.g.,
2923 because its maximum value is negative), the index type requires
2924 structural equality tests.
2925 (build_array_type): Build the canonical form of an array type.
2926 (build_function_type): Function types require structural equality,
2927 because they contain default arguments, attributes, etc.
2928 (build_method_type_directly): Ditto for method types.
2929 (build_offset_type): Build the canonical offset type.
2930 (build_complex_type): Build the canonical vector type.
2931 (make_vector_type): Build the canonical vector type.
2932 * tree.h (TYPE_CANONICAL): New.
2933 (TYPE_STRUCTURAL_EQUALITY_P): New.
2934 (SET_TYPE_STRUCTURAL_EQUALITY): New.
2935 (struct tree_type): Added "canonical" field.
2936 * params.h (VERIFY_CANONICAL_TYPES): New.
2937 * doc/c-tree.texi (TYPE_CANONICAL): Document.
2938 (TYPE_STRUCTURAL_EQUALITY_P): Document.
2939 (SET_TYPE_STRUCTURAL_EQUALITY): Document.
2940 * doc/invoke.texi (verify-canonical-types): Document --param
2941 parameter for verifying canonical types.
2943 2007-01-02 Joseph Myers <joseph@codesourcery.com>
2945 * config.gcc (powerpc-*-eabispe*, powerpc-*-eabisimaltivec*,
2946 powerpc-*-eabisim*, powerpc-*-eabialtivec*, powerpc-*-eabi*,
2947 powerpc-*-rtems*, powerpc-wrs-vxworks, powerpc-wrs-vxworksae,
2948 powerpcle-*-eabisim*, powerpcle-*-eabi*): Add rs6000/e500.h to
2950 * config/rs6000/e500.h: New.
2951 * config/rs6000/eabi.h (TARGET_SPE_ABI, TARGET_SPE, TARGET_E500,
2952 TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE, TARGET_E500_DOUBLE):
2954 * config/rs6000/linuxspe.h (TARGET_SPE_ABI, TARGET_SPE,
2955 TARGET_E500, TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE,
2956 TARGET_E500_DOUBLE): Remove.
2957 * config/rs6000/vxworks.h (TARGET_SPE_ABI, TARGET_SPE,
2958 TARGET_E500, TARGET_ISEL, TARGET_FPRS): Remove.
2959 * config/rs6000/rs6000.h (CHECK_E500_OPTIONS): Define.
2960 * config/rs6000/rs6000.c (rs6000_override_options): Use
2963 2007-01-02 Joseph Myers <joseph@codesourcery.com>
2965 * config/rs6000/rs6000.c (print_operand): Check (TARGET_SPE ||
2966 TARGET_E500_DOUBLE), not TARGET_E500, for %y.
2967 (rs6000_generate_compare, rs6000_emit_sCOND, output_cbranch,
2968 rs6000_emit_cmove): Don't check TARGET_E500.
2969 * config/rs6000/rs6000.md (bunordered, bordered, sunordered,
2970 sordered): Don't check TARGET_E500.
2972 2007-01-01 Eric Christopher <echristo@apple.com>
2974 * config/mips/mips.c (mips_regno_mode_ok_for_base_p): Use
2975 HARD_REGISTER_NUM_P.
2977 2007-01-01 Roger Sayle <roger@eyesopen.com>
2979 * fold-const.c (fold_binary) <EQ_EXPR>: Fold "(X^C1) eq/ne C2" into
2980 "X eq/ne (C1^C2)". Fold "(X^Z) eq/ne (Y^Z)" as "X eq/ne Y" when Z
2981 has no side-effects. Fold "(X^C1) eq/ne (Y^C2)" as "(X^(C1^C2))
2984 2007-01-01 Mike Stump <mrs@apple.com>
2986 * configure.ac: Remove support for building with Apple's gcc-3.1.
2988 2007-01-02 Joseph Myers <joseph@codesourcery.com>
2991 * caller-save.c (add_stored_regs): Only handle SUBREGs if inner
2992 REG is a hard register. Do not modify REG before calling
2994 * rtlanal.c (subreg_get_info): Don't assert size of XMODE is a
2995 multiple of the size of YMODE for certain lowpart cases.
2997 2007-01-01 Andrew Pinski <pinskia@gmail.com>
3000 * gimplify (voidify_wrapper_expr): Update for
3003 2007-01-01 Andreas Schwab <schwab@suse.de>
3006 * config/ia64/ia64.c (ia64_compute_frame_size): Account space for
3007 save of BR0 in extra_spill_size instead of spill_size.
3008 (ia64_expand_prologue): Save BR0 outside of the gr/br/fr spill
3010 (ia64_expand_epilogue): Restore BR0 from its new location.
3012 2007-01-01 Andrew Pinski <pinskia@gmail.com>
3014 * gimplify.c (gimplify_init_constructor <case VECTOR_TYPE>):
3015 Use a temporary variable if the left hand side is not a gimple
3018 2007-01-01 Andrew Pinski <pinskia@gmail.com>
3020 * gimplify.c (gimplify_return_expr): Make the temporary variable
3021 for the return expression, a gimple register variable.
3023 2007-01-01 Jan Hubicka <jh@suse.cz>
3025 * emit-rtl.c (emit_copy_of_insn_after): Do not call copy_insn_1 for
3028 2007-01-01 Mike Stump <mrs@apple.com>
3030 * configure.ac (HAVE_GAS_LITERAL16): Add autoconf check for
3032 * config/darwin.c (machopic_select_rtx_section): Use
3034 (darwin_mergeable_constant_section): Likewise.
3035 * configure: Regenerate.
3036 * config.in: Regenerate.
3038 2007-01-01 Jan Hubicka <jh@suse.cz>
3039 Andrew Pinski <pinskia@gmail.com>
3041 * cgraphunit.c (cgraph_optimize): Call cgraph_add_new_functions
3042 before starting IPA passes.