1 2007-01-29 Roger Sayle <roger@eyesopen.com>
2 Richard Guenther <rguenther@suse.de>
4 * fold-const.c (round_up): Make HIGH an unsigned HOST_WIDE_INT to
5 avoid undefined behaviour on overflow. Use force_fit_type_double
6 to construct the constant with the specified TREE_OVERFLOW.
8 2007-01-29 Janis Johnson <janis187@us.ibm.com>
10 * config/dfp-bit.c: Add parameterized support for fp exceptions.
11 * config/dfp-bit.h: Ditto.
13 2007-01-29 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
15 * c-decl.c (pop_scope): Replace warnings with call to
16 warn_for_unused_label.
17 * c-common.h (warn_for_unused_label): Declare.
18 * c-common.c (warn_for_unused_label): Define.
20 2007-01-29 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
22 * tree-optimize.c (update_inlined_to_pointers): Delete unused
25 2007-01-29 Janis Johnson <janis187@us.ibm.com>
27 * Makefile.in (USER_H): Remove decfloat.h.
28 * ginclude/decfloat.h: Delete, moving contents to ...
29 * ginclude/float.h: Add support for decimal floating point,
30 guarded by __STDC_WANT_DEC_FP__.
32 2007-01-29 Mike Stump <mrs@apple.com>
34 * doc/gccint.texi (Top): Rename Loop Representation to Loop
35 Analysis and Representation to resolve case insensitive conflict.
36 * doc/loop.texi (Loop Analysis and Representation): Likewise.
38 2007-01-28 Daniel Berlin <dberlin@dberlin.org>
40 * tree.h (struct tree_memory_tag): Add aliases member.
41 (MTAG_ALIASES): New macro.
42 * tree-ssa-alias.c (alias_bitmap_obstack): New variable.
43 (add_may_alias): Remove pointer-set. Update for may_aliases being
45 (mark_aliases_call_clobbered): Update for may_aliases being a
47 (compute_tag_properties): Ditto.
48 (create_partition_for): Ditto.
49 (compute_memory_partitions): Ditto.
50 (dump_may_aliases_for): Ditto.
51 (is_aliased_with): Ditto.
52 (add_may_alias_for_new_tag): Ditto.
53 (rewrite_alias_set_for): Rewrite for may_aliases being a bitmap.
54 (compute_is_aliased): New function.
55 (compute_may_aliases): Call compute_is_aliased).
56 (init_alias_info): Initialize alias_bitmap_obstack.
57 (union_alias_set_into): New function.
58 (compute_flow_sensitive_aliasing): Use union_aliases_into.
59 (have_common_aliases_p): Rewrite to take two bitmaps and use
61 (compute_flow_insensitive_aliasing): Stop using pointer-sets.
63 (finalize_ref_all_pointers): Update for add_may_alias changes.
64 (new_type_alias): Ditto.
65 * tree-flow-inline.h (may_aliases): Return a bitmap.
66 * tree-dfa.c (dump_variable): Check for MTAG_P'ness.
67 * tree-ssa.c (verify_flow_insensitive_alias_info): Update for
68 may_aliases being a bitmap.
69 * tree-flow.h (struct var_ann_d): Remove may_aliases member.
70 may_aliases now returns a bitmap.
71 * tree-ssa-structalias.c (merge_smts_into): Update for may_aliases
73 * tree-ssa-operands.c (add_virtual_operand): Update for
74 may_aliases being a bitmap.
76 2007-01-29 Daniel Berlin <dberlin@dberlin.org>
78 PR tree-optimization/30630
79 * tree-ssa-structalias.c (do_complex_constraint): Mark correct
82 2007-01-29 Simon Martin <simartin@users.sourceforge.net>
85 * gimplify.c (gimplify_target_expr): Make sure that the TARGET_EXPR is
86 expanded only once even if an error occurs.
88 2007-01-29 Ben Elliston <bje@au.ibm.com>
90 * gcov-io.h (__gcov_indirect_call_profiler): Declare.
91 (__gcov_average_profiler): Likewise.
92 (__gcov_ior_profiler): Likewise.
93 (__gcov_merge_ior): Likewise.
95 2007-01-28 Jan Hubicka <jh@suse.cz>
97 * builtins.c (expand_builtin_memset): Fix typo in my last patch.
98 * value-prof.c (stringop_block_profile): Likewise.
100 2007-01-28 Jan Hubicka <jh@suse.cz>
102 * expr.c (emit_block_move_via_movmem, emit_block_move_via_libcall): Add
103 variant handling histograms; add wrapper.
104 (clear_storage_via_libcall): Export.
105 (emit_block_move_hints): Break out from ...; add histograms.
106 (emit_block_move): ... this one.
107 (clear_storage_hints): Break out from ...; add histograms.
108 (clear_storage): ... this one.
109 (set_storage_via_memset): Handle histogram.
110 * expr.h (emit_block_move_via_libcall, emit_block_move_hints): Declare.
111 (clear_storage_hints, clear_storage_via_libcall): Declare.
112 (set_storage_via_setmem): Update prototype.
113 * doc/md.texi (movmem, setmem): Document new arguments.
115 * value-prof.c (dump_histogram_value, tree_find_values_to_profile): Add
117 (stringop_block_profile): New global function.
118 (tree_stringops_values_to_profile): Profile block size and alignment.
119 * value-prof.h (enum hist_type): add HIST_TYPE_AVERAGE and
121 (struct profile_hooks): Add gen_average_profiler and gen_ior_profiler.
122 (stringop_block_profile): Declare.
123 * builtins.c: Include value-prof.h.
124 (expand_builtin_memcpy, expand_builtin_memset): Pass block profile.
125 * gcov-ui.h (GCOV_COUNTER_NAMES): Add new counter.
126 (GCOV_COUNTER_AVERAGE, GCOV_COUNTER_IOR): New constants.
127 (GCOV_COUNTERS, GCOV_LAST_VALUE_COUNTER): Update.
128 * profile.c (instrument_values): Add new counters.
129 * cfgexpand.c (expand_gimple_basic_block): Propagate histograms to
131 * tree-profile.c (tree_average_profiler_fn, tree_ior_profiler_fn): New.
132 (tree_init_edge_profiler): Build new profilers.
133 (tree_gen_average_profiler, tree_gen_ior_profiler): New.
134 (pass_tree_profile): Add dump.
135 (tree_profile_hooks): Update.
136 * Makefile.in (LIBGCOV): Add new constants.
137 * libgcov.c (__gcov_merge_ior, __gcov_average_profiler,
138 __gcov_ior_profiler): New.
139 * i386.md (movmem/setmem expanders): Add new optional arguments.
141 2007-01-28 David Edelsohn <edelsohn@gnu.org>
143 * doc/md.texi (Standard Pattern Names): Document blockage pattern.
145 2007-01-28 Zdenek Dvorak <dvorakz@suse.cz>
147 * tree-ssa-loop-unswitch.c: Include tree-inline.h.
148 (tree_unswitch_single_loop): Pass eni_size_weights to
150 * tree-ssa-loop-manip.c: Include tree-inline.h.
151 (can_unroll_loop_p): Pass eni_size_weights to
153 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p):
154 Pass eni_size_weights to estimate_num_insns.
155 * tree.h (init_inline_once): Export.
156 * toplev.c (backend_init): Call init_inline_once.
157 * cgraphunit.c (cgraph_process_new_functions,
158 cgraph_analyze_function): Pass eni_inlining_weights to
160 * ipa-inline.c (compute_inline_parameters): Ditto.
161 * tree-ssa-loop-ivcanon.c (tree_num_loop_insns): Pass weights
162 to estimate_num_insns.
163 (try_unroll_loop_completely): Pass eni_size_weights to
165 * tree-eh.c (decide_copy_try_finally): Pass eni_size_weights
166 ot estimate_num_insns.
167 * tree-ssa-loop-prefetch.c: Include tree-inline.h.
168 (loop_prefetch_arrays): Pass eni_time_weights to tree_num_loop_insns.
169 * tree-inline.c (eni_inlining_weights, eni_size_weights,
170 eni_time_weights): New variables.
171 (init_inline_once): Initialize them.
172 (struct eni_data): Mew.
173 (estimate_num_insns_1, estimate_num_insns): Use weights.
174 * tree-inline.h (struct eni_weights_d): New.
175 (eni_inlining_weights, eni_size_weights, eni_time_weights): Declare.
176 (estimate_num_insns): Declaration changed.
177 * cfgloop.h (tree_num_loop_insns): Declaration changed.
178 * Makefile.in (tree-ssa-loop-unswitch.o, tree-ssa-loop-prefetch.o,
179 tree-ssa-loop-manip.o): Add TREE_INLINE_H dependency.
181 2007-01-28 Zdenek Dvorak <dvorakz@suse.cz>
183 * tree-data-ref.c (conflict_fn): Assert that the number of affine
184 relations in the conflict function is valid.
186 2007-01-27 Ian Lance Taylor <iant@google.com>
188 * common.opt: Add fstrict-overflow.
189 * opts.c (decode_options): Set flag_strict_overflow if -O2.
190 * flags.h (TYPE_OVERFLOW_WRAPS): Define.
191 (TYPE_OVERFLOW_UNDEFINED): Define.
192 (TYPE_OVERFLOW_TRAPS): Define. This replaces TYPE_TRAP_SIGNED.
194 * tree.h (TYPE_TRAP_SIGNED): Don't define.
195 * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_UNDEFINED.
196 (fold_negate_expr): Likewise.
197 (make_range): Likewise.
198 (extract_muldiv_1): Likewise.
199 (maybe_canonicalize_comparison): Likewise.
200 (fold_comparison): Likewise.
201 (fold_binary): Likewise.
202 (tree_expr_nonnegative_p): Likewise.
203 (tree_expr_nonzero_p): Likewise.
204 * tree-vrp.c (compare_values): Likewise.
205 (extract_range_from_binary_expr): Likewise.
206 (extract_range_from_unary_expr): Likewise.
207 * tree-ssa-loop-niter.c (infer_loop_bounds_from_signedness):
209 (nowrap_type_p): Likewise.
210 * tree-scalar-evolution.c (simple_iv): Likewise.
211 * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_WRAPS.
212 (build_range_check): Likewise.
213 (extract_muldiv_1): Likewise.
214 (fold_comparison): Likewise.
215 * tree-vrp.c (vrp_int_const_binop): Likewise.
216 (extract_range_from_unary_expr): Likewise.
217 * convert.c (convert_to_integer): Likewise.
218 * fold-const.c (fold_negate_expr): Use TYPE_OVERFLOW_TRAPS.
219 (fold_comparison): Likewise.
220 (fold_binary): Likewise.
221 * optabs.c (optab_for_tree_code): Likewise.
222 * tree-vectorizer.c (vect_is_simple_reduction): Likewise.
223 * simplify-rtx.c (simplify_const_relational_operation): Check
224 flag_strict_overflow and flag_trapv.
225 (simplify_const_relational_operation): Likewise.
226 * doc/invoke.texi (Option Summary): Mention -fstrict-overflow.
227 (Optimize Options): Add -fstrict-overflow to -O2 list. Document
230 2007-01-27 Roger Sayle <roger@eyesopen.com>
232 * tree.c (tree_fold_gcd): Delete.
233 * tree.h (tree_fold_gcd): Remove prototype.
234 * tree-data-ref.c (tree_fold_divides_p): Don't use tree_fold_gcd to
235 test whether one constant integer is a multiple of another. Instead
236 call int_const_binop with TRUNC_MOD_EXPR and test for a zero result.
237 * fold-const.c (multiple_of_p): We've determined both TOP and
238 BOTTOM are integer constants so we can call int_const_binop directly
239 instead of the more generic const_binop.
241 2007-01-27 Roger Sayle <roger@eyesopen.com>
243 * fold-const.c (size_binop): In the fast-paths for X+0, 0+X, X-0 and
244 1*X check that the constant hasn't overflowed, to preserve the
246 (round_up): Provide an efficient implementation when rouding-up an
247 INTEGER_CST to a power-of-two.
249 2007-01-28 Ralf Wildenhues <Ralf.Wildenhues@gmx.de>
251 * doc/sourcebuild.texi: Add comma for clarity.
252 * doc/extend.texi: Fix some typos.
253 * doc/passes.texi: Likewise.
254 * doc/cppinternals.texi: Likewise.
255 * doc/c-tree.texi: Likewise.
256 * doc/tree-ssa.texi: Likewise.
257 * doc/install.texi: Likewise.
259 2007-01-27 Jan Hubicka <jh@suse.cz>
261 * tree-sra.c (sra_walk_function): Don't rely on aliases being build.
262 (pass_sra): Do not require alias information.
263 * passes.c (init_optimization_passes): Add SRA
265 2007-01-27 Steven Bosscher <steven@gcc.gnu.org>
267 * tracer.c (rest_of_handle_tracer): We already cleaned
268 up the CFG in tracer() so don't do it here again.
269 * cfgcleanup.c (rest_of_handle_jump2): Don't repeat
270 cleanup_cfg here, either. And don't call renumber_insns.
272 * cfgrtl.c (rtl_verify_flow_info_1): Don't verify that BB_END
273 and BB_HEAD are in the insn stream here. Instead make sure
274 that BB_INSN is valid on all insns. Also, do check here that
275 there are no pending branch predictions...
276 (rtl_verify_flow_info): ...instead of doing it here. Checks
277 for BB_END and BB_HEAD moved from rtl_verify_flow_info_1 to
280 2007-01-26 Roger Sayle <roger@eyesopen.com>
282 * config/i386/i386.c (ix86_swap_binary_operands_p): New helper
283 function to simplify/factorize operand order canonicalization.
284 (ix86_fixup_binary_operands): Reorganize using the above function.
285 (ix86_binary_operator_ok): Likewise.
287 2007-01-27 Jakub Jelinek <jakub@redhat.com>
289 * genattrtab.c (struct attr_value_list, insn_code_values): Move to
290 file scope from optimize_attrs.
291 (simplify_test_exp): If insn_code_values is not NULL, use it to speed
293 (optimize_attrs): Clear insn_code_values after freeing it.
295 2007-01-26 Zdenek Dvorak <dvorakz@suse.cz>
297 * tree-ssa-address.c (create_mem_ref): Remove ", bsi" from
298 a parts.base assignment.
300 2007-01-26 Zdenek Dvorak <dvorakz@suse.cz>
302 * tree-data-ref.c (dump_subscript): Use dump_conflict_function.
303 (compute_subscript_distance, initialize_data_dependence_relation,
304 finalize_ddr_dependent, analyze_ziv_subscript,
305 analyze_siv_subscript_cst_affine,
306 compute_overlap_steps_for_affine_univar,
307 compute_overlap_steps_for_affine_1_2, analyze_subscript_affine_affine,
308 analyze_siv_subscript, analyze_miv_subscript,
309 analyze_overlapping_iterations, subscript_dependence_tester_1,
310 compute_self_dependence, free_dependence_relation): Work
311 with affine_fn instead of chrecs.
312 (dump_affine_function, dump_conflict_function, affine_function_equal_p,
313 common_affine_function, affine_function_base,
314 affine_function_constant_p, affine_fn_op, affine_fn_plus,
315 affine_fn_minus, affine_fn_free, conflict_fn_not_known,
316 conflict_fn_no_dependence, free_conflict_function, free_subscripts,
317 conflict_fn, affine_fn_cst, affine_fn_univar): New functions.
318 (all_chrecs_equal_p): Removed.
319 * tree-data-ref.h (affine_fn, conflict_function): New types.
320 (struct subscript): Change type of conflicting_iterations_in_a
321 and conflicting_iterations_in_b.
323 2007-01-26 Steve Ellcey <sje@cup.hp.com>
326 * config/pa/pa.h (TARGET_HPUX_11): New.
327 * config/pa/pa-hpux11.h (TARGET_HPUX_11): New.
328 * config/pa/pa.c (pa_init_builtins): Use TARGET_HPUX_11.
330 2007-01-26 Daniel Berlin <dberlin@dberlin.org>
331 Richard Guenther <rguenther@suse.de>
333 * tree-ssa-structalias.c (solve_graph): Handle case
334 we merged the variable to another.
336 2007-01-25 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
338 * builtins.c (fold_builtin_1): Treat ccos and ccosh as 'even'
341 * fold-const.c (negate_mathfn_p): Treat casin, casinh, catan,
342 catanh, cproj, csin, csinh, ctan and ctanh as 'odd' functions.
344 2007-01-25 DJ Delorie <dj@redhat.com>
346 * config/m32c/m32c.c (m32c_cannot_change_mode_class): We don't
347 allow changes to modes which don't fit in those registers.
349 * reload1.c (choose_reload_regs): Check for invalid subregs before
350 computing their locations, not after.
352 2007-01-25 Geoffrey Keating <geoffk@apple.com>
355 * config/rs6000/rs6000.c (first_altivec_reg_to_save): On Darwin,
356 save Altivec registers in an eh_return function.
357 (compute_vrsave_mask): Likewise.
358 (rs6000_stack_info): Correct AIX/Darwin stack alignment computation
359 for saving Altivec registers.
360 (rs6000_emit_prologue): Don't allocate stack twice in
361 eh_return function. Correct expected value of altivec_save_offset
362 when using save_world. Describe save of R0 to stack when using
363 save_world. Describe stack pointer adjustment when using
364 save_world. Remove duplicated eh_return parameter register saving.
365 Update sp_offset variable after save_world.
366 * config/rs6000/t-darwin (LIB2FUNCS_STATIC_EXTRA): Remove
368 (LIB2FUNCS_EXTRA): Add darwin-world.asm.
369 * config/rs6000/darwin.h (SUBTARGET_OVERRIDE_OPTIONS): -m64
372 2007-01-25 Steve Ellcey <sje@cup.hp.com>
374 * config.gcc (ia64*-*-hpux*): Make posix threads the default.
376 2007-01-25 Steve Ellcey <sje@cup.hp.com>
379 * config/pa/pa.c (pa_init_builtins): Set asm names for finite routines.
380 * config/ia64/ia64.c (ia64_init_builtins): Ditto.
382 2007-01-25 Richard Guenther <rguenther@suse.de>
384 * doc/invoke.texi (-Wcoverage-mismatch): Document.
385 * common.opt (-Wcoverage-mismatch): New warning option.
386 * coverage.c (get_coverage_counts): Ignore coverage mismatch
387 if -Wcoverage-mismatch is given.
389 2007-01-25 Razya Ladelsky <razya@il.ibm.com>
391 * ipa-cp.c (ipcp_insert_stage, ipcp_driver): Support for SSA.
392 (ipcp_driver): Change to static definition.
393 Add dumping of the ifunctions.
394 (constant_val_insert): Remove unused parameter. Support for SSA.
395 (ipcp_propagate_const): Support for SSA.
396 (ipcp_profile_bb_print): Print only analyzed nodes.
397 (ipcp_replace_map_create): Remove support for Fortran constant
399 * ipa-prop.c (ipa_method_modify_stmt,
400 ipa_callsite_compute_param): Support for SSA.
401 * ipa-prop.h (ipcp_driver): Remove declaration.
402 (IS_VALID_TREE_MAP_INDEX): Add define.
404 2007-01-24 Geoffrey Keating <geoffk@apple.com>
406 * unwind-dw2.c (execute_stack_op): Handle DW_OP_swap.
408 2007-01-24 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
411 * builtins.c (fold_builtin_cabs): Use MPFR to evaluate a
412 constant argument to cabs and do it without checking for
413 -funsafe-math-optimizations.
415 2007-01-24 Douglas Gregor <dgregor@osl.iu.edu>
417 * c-common.h (RID_FIRST_CXX0X): New.
418 (RID_LAST_CXX0X): New.
419 * c-opts.c (c_common_handle_option): -Wc++0x-compat is triggered
421 * c.opt (Wc++0x-compat): New.
422 * doc/invoke.texi (-Wc++0x-compat): Document.
424 2007-01-24 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
426 * builtins.c (fold_builtin_carg): New.
427 (fold_builtin_1): Use it.
429 2007-01-24 Jan Hubicka <jh@suse.cz>
431 * ipa-inline.c (cgraph_decide_inlining): Initialize initial_insns.
433 * ipa-inline.c (initial_insns, max_insns): Delete.
434 (compute_max_insns): New function.
435 (cgraph_decide_inlining_of_small_function): Use it; take minimal amount
436 of insns as base for code growth.
437 (cgraph_decide_inlining): Make initial_insns local; do not compute
439 * params.def (PARAM_INLINE_UNIT_GROWTH): Set to 60.
440 * doc/invoke.texi (inline-unit-growth): Update docs.
442 2007-01-24 Jakub Jelinek <jakub@redhat.com>
444 * config/i386/i386.h (x86_cmpxchg16b): Remove const.
445 (TARGET_CMPXCHG16B): Define to x86_cmpxchg16b.
446 * config/i386/i386.c (x86_cmpxchg16b): Remove const.
447 (override_options): Add PTA_CX16 flag. Set x86_cmpxchg16b
448 for CPUs that have PTA_CX16 set.
451 * gimplify.c (omp_check_private): New function.
452 (gimplify_scan_omp_clauses): Use it for
453 firstprivate/lastprivate/reduction.
456 * gimplify.c (omp_add_variable): Don't call omp_notice_variable
457 on TYPE_SIZE_UNIT for GOVD_LOCAL VLAs.
460 * omp-low.c (lower_omp_for_lastprivate): Add dlist argument.
461 If lower_lastprivate_clauses emits some statements, append them
462 to dlist rather than body_p and to body_p append an initializer.
463 (lower_omp_for): Adjust caller.
465 2007-01-24 Steve Ellcey <sje@cup.hp.com>
467 * target.h (globalize_decl_name): New.
468 * target-def.h (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
469 * output.h (default_globalize_decl_name): New.
470 * varasm.c (asm_output_bss): Use globalize_decl_name instead of
472 (globalize_decl): Ditto.
473 (default_globalize_decl_name): New.
474 * config/ia64/ia64.c (ia64_globalize_decl_name): New.
475 (ia64_handle_version_id_attribute): New.
476 (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
477 (ia64_asm_output_external): Use globalize_decl_name instead
479 * doc/extend.texi (version_id): New pragma.
480 * doc/tm.texi (ARGET_ASM_GLOBALIZE_DECL_NAME): New target hook.
482 2007-01-24 Andreas Krebbel <krebbel1@de.ibm.com>
484 * unwind-dw2-fde.c (get_cie_encoding): Replaced _Unwind_Word with
485 _uleb128_t and _Unwind_SWord with _sleb128_t.
486 * unwind-dw2.c (extract_cie_info, execute_stack_op, execute_cfa_program,
487 uw_frame_state_for, uw_update_context_1): Likewise.
488 * unwind-c.c (parse_lsda_header, PERSONALITY_FUNCTION): Likewise.
489 * unwind-pe.h (read_uleb128, read_sleb128,
490 read_encoded_value_with_base): Likewise.
491 * unwind-generic.h: Define _sleb128_t and _uleb128_t types.
493 2007-01-24 Richard Guenther <rguenther@suse.de>
495 * builtins.c (expand_builtin_cexpi): Get the fndecl
496 for cexp in the correct way.
498 2007-01-24 Jan Hubicka <jh@suse.cz>
500 * tree-ssa-dce.c (eliminate_unnecesary_stmts): Remove dead LHS of calls.
502 2007-01-24 Andreas Krebbel <krebbel1@de.ibm.com>
504 * c-cppbuiltin.c (builtin_define_type_sizeof): New function.
505 (c_cpp_builtins): New builtin macros: __SIZEOF_INT__, __SIZEOF_LONG__,
506 __SIZEOF_LONG_LONG__, __SIZEOF_SHORT__, __SIZEOF_POINTER__,
507 __SIZEOF_FLOAT__, __SIZEOF_DOUBLE__, __SIZEOF_LONG_DOUBLE__,
508 __SIZEOF_SIZE_T__, __SIZEOF_WCHAR_T__, __SIZEOF_WINT_T__ and
509 __SIZEOF_PTRDIFF_T__.
510 * doc/cpp.texi: Documentation for the new builtin macros added.
512 2007-01-24 Uros Bizjak <ubizjak@gmail.com>
514 * config/i386/i386.md (tanxf2, tan<mode>2, atan<mode>2, log<mode>2,
515 log10<mode>2, log2<mode>2, expxf2, exp10xf2, exp2xf2): Use op2
516 instead of operands[2] to avoid access past the end of array.
518 2007-01-24 Richard Sandiford <richard@codesourcery.com>
520 * reload1.c (emit_reload_insns): Pass the reload register
521 for a non-spill output reload through forget_old_reloads_1.
523 2007-01-23 Joseph Myers <joseph@codesourcery.com>
525 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Define
526 _SOFT_DOUBLE if doubles use software floating-point.
527 * config/rs6000/libgcc-ppc-glibc.ver: Export additional long
528 double functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
529 * config/rs6000/darwin-ldouble.c: Also compile functions for
530 hard-float without FPRs. Use fmsub function for all __NO_FPRS__
531 cases. Compile extra functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
532 * config/rs6000/linuxspe.h (SUBSUBTARGET_OVERRIDE_OPTIONS): Remove
533 commented-out long double override.
534 (CPP_LONGDOUBLE_DEFAULT_SPEC): Likewise.
535 * config/rs6000/eabispe.h: Likewise.
536 * config/rs6000/rs6000.c (rs6000_override_options): Don't override
537 long double for non-SPE.
538 (rs6000_handle_option): Likewise.
539 (invalid_e500_subreg): Disallow more subregs involding DImode,
540 DFmode, TImode or TFmode.
541 (rs6000_legitimate_offset_address_p): Check TFmode offsets for
543 (legitimate_lo_sum_address_p): Also check for TFmode for E500
545 (rs6000_legitimize_address): Also handle TFmode for E500 double.
546 (rs6000_legitimize_reload_address): Also handle TFmode for E500
548 (rs6000_legitimate_address): Also check for TFmode for E500
550 (rs6000_emit_move): Use DFmode subregs of TFmode for E500 double.
551 (spe_build_register_parallel): Handle TFmode and TCmode.
552 (rs6000_spe_function_arg): Handle TFmode and TCmode for E500
554 (function_arg): Handle TFmode and TCmode for E500 double.
555 (rs6000_init_libfuncs): Initialize extra libfuncs for soft double
557 (print_operand): Handle TFmode and TImode for %y.
558 (rs6000_generate_compare): Handle TFmode comparisons for E500
560 (spe_func_has_64bit_regs_p): Check for TFmode for E500 double.
561 (rs6000_function_value): Handle TFmode and TCmode for E500 double.
562 (rs6000_libcall_value): Handle TFmode and TCmode for E500 double.
563 * config/rs6000/rs6000.h (CANNOT_CHANGE_MODE_CLASS): Check for
564 TFmode for E500 double.
565 * config/rs6000/rs6000.md (FP): Allow TF for E500 double.
566 (floatsidf2): Enable for E500 double.
567 (movtf_softfloat): Use rs6000_nonimmediate_operand.
568 (extenddftf2): Change to extenddftf2_fprs.
569 (extenddftf2): Call gen_spe_extenddftf2 or gen_extenddftf2_fprs
570 depending on TARGET_E500_DOUBLE.
571 (extendsftf2): Enable for E500 double.
572 (trunctfdf2): Enable for E500 double.
573 (trunctfsf2): Change to trunctfsf2_fprs.
574 (trunctfsf2): Call gen_spe_trunctfsf2 or gen_trunctfsf2_fprs
575 depending on TARGET_E500_DOUBLE.
576 (floatsitf2): Enable for E500 double.
577 (fix_trunctfsi2): Change to fix_trunctfsi2_fprs.
578 (fix_trunctfsi2): Call gen_spe_fix_trunctfsi2 or
579 gen_fix_trunctfsi2_fprs depending on TARGET_E500_DOUBLE.
580 (negtf2): Change to negtf2_internal.
581 (negtf2): New expander.
582 (abstf2): Enable for E500 double. Call gen_spe_abstf2_tst,
583 gen_spe_abstf2_cmp or gen_abstf2_internal depending on
584 TARGET_E500_DOUBLE and flag_unsafe_math_optimizations.
585 (movdi_internal32): Use rs6000_nonimmediate_operand.
586 (unnamed splitter): Likewise.
587 * config/rs6000/spe.md (CMPTFEQ_GPR, TSTTFEQ_GPR, CMPTFGT_GPR,
588 TSTTFGT_GPR, CMPTFLT_GPR, TSTTFLT_GPR): New unspecs.
589 (SPE64TF, DITI): New mode macros.
590 (frob_df_di): Change to frob_<SPE64:mode>_<DITI:mode>; allow more
593 (frob_<mode>_di_2): New.
594 (frob_tf_di_8_2): New.
595 (frob_di_df): Change to frob_di_<mode>; allow more modes.
597 (frob_di_df_2): Change to frob_<DITI:mode>_<SPE64:mode>_2; allow
599 (frob_ti_<mode>_8_2): New.
601 (mov_si<mode>_e500_subreg0, mov_si<mode>_e500_subreg0_2,
602 mov_si<mode>_e500_subreg4, mov_si<mode>_e500_subreg4_2): Allow
604 (mov_sitf_e500_subreg8, mov_sitf_e500_subreg8_2,
605 mov_sitf_e500_subreg12, mov_sitf_e500_subreg12_2): New.
606 (spe_trunctfdf2_internal1, spe_trunctfsf2, spe_extenddftf2,
607 spe_fix_trunctfsi2, spe_fix_trunctfsi2_internal,
608 spe_negtf2_internal, spe_abstf2_cmp, spe_abstf2_tst): New.
609 (cmptfeq_gpr, tsttfeq_gpr, cmptfgt_gpr, tsttfgt_gpr, cmptflt_gpr,
612 2007-01-23 Ian Lance Taylor <iant@google.com>
614 * Makefile.in (OBJS-common): Reformat, alphabetize, but put
616 (OBJS-archive): Reformat, alphabetize.
617 (OBJS): Change out_object_file to OBJS-md.
619 2007-01-23 Uros Bizjak <ubizjak@gmail.com>
621 * config/i386/i386.md: Use REG_P, MEM_P, CONST_INT_P, LABEL_P,
622 JUMP_P and CALL_P predicates where applicable.
623 * config/i386/i386.c: Ditto.
624 * config/i386/i386.md: Ditto.
625 * config/i386/mmx.md: Ditto.
626 * config/i386/predicates.md: Ditto.
628 2007-01-22 Andreas Schwab <schwab@suse.de>
630 * config/m68k/m68k.h: Fix comment.
632 2007-01-22 Jan Hubicka <jh@suse.cz>
634 * passes.c (init_optimization_passes): Do not rerun
635 pass_early_warn_uninitialized.
637 2007-01-22 Richard Guenther <rguenther@suse.de>
639 PR tree-optimization/30038
640 * tree-ssa-math-opts.c (maybe_record_sincos): New static helper
642 (execute_cse_sincos_1): Likewise.
643 (execute_cse_sincos): Likewise.
644 (gate_cse_sincos): Likewise.
645 (pass_cse_sincos): New pass CSEing sin() and cos() calls using
646 the cexpi() canonicalization of sincos().
647 * tree-pass.h (pass_cse_sincos): Declare.
648 * passes.c (init_optimization_passes): New pass pas_cse_sincos.
650 2007-01-21 Eric Botcazou <ebotcazou@libertysurf.fr>
652 PR rtl-optimization/29329
653 * combine.c (replaced_rhs_insn): Rename to i2mod.
654 (replaced_rhs_value): Rename to i2mod_new_rhs.
655 (i2mod_old_rhs): New global variable.
656 (combine_instructions): Adjust for above change. Save a copy of
657 the old RHS into i2mod_old_rhs when the contents of a REG_EQUAL
658 note are substituted in the second instruction.
659 (distribute_notes) <REG_DEAD>: Adjust for above change. Do not
660 ditch the note if it pertains to the second eliminated register
661 and this register is mentioned in i2mod_old_rhs.
664 2006-09-12 Eric Botcazou <ebotcazou@libertysurf.fr>
666 * combine.c (distribute_notes) <REG_DEAD>: Do not consider SETs past
667 the insn to which the note was originally attached.
669 2007-01-21 Jan Hubicka <jh@suse.cz>
671 * ipa-inline.c (inlining_mode): Comment, move up.
672 (cgraph_decide_inlining_incrementally): Do not perform inlining
673 itself; fix handling of flattening of self recursive functions.
674 (cgraph_find_cycles): Remove.
675 (cgraph_flatten_node): Remove.
676 (cgraph_decide_inlining): Use incremental inliner to handle flattening.
677 (try_inline): New function.
678 (cgraph_early_inlining): Update call of
679 cgraph_decide_inlining_incrementally. Apply inlining here.
680 (apply_inline): Update call of cgraph_decide_inlining_incrementally.
682 2007-01-21 Dirk Mueller <dmueller@suse.de>
685 * tree-vrp.c (check_array_bounds): do not warn
686 about ADDR_EXPR's of ARRAY_REF's which are immediately
687 used in binary expressions.
689 2007-01-21 Ira Rosen <irar@il.ibm.com>
691 * tree-vectorizer.h (struct _stmt_vec_info): Add new field
692 read_write_dep and macros for its access.
693 * tree-vectorizer.c (new_stmt_vec_info): Initialize the new field.
694 * tree-vect-analyze.c (vect_analyze_data_ref_dependence): Remove
695 argument, call vect_check_interleaving for every independent pair of
696 data-refs. Mark loads that access the same memory location as a store
698 (vect_check_dependences): Remove.
699 (vect_analyze_data_ref_dependences): Remove vect_check_dependences
700 call, fix the call to vect_analyze_data_ref_dependence.
701 (vect_analyze_data_ref_access): For statements that access the same
702 data-ref, check that they are not stores; for loads, check that there
703 is no store that access the same location.
705 2007-01-20 Roger Sayle <roger@eyesopen.com>
706 Joseph Myers <joseph@codesourcery.com>
708 * doc/invoke.texi (-fdump-rtl-combine): Fix under/overfull hbox.
709 (-fdump-rtl-stack): Likewise.
710 (-fno-signed-zeros): Use @minus{} for a minus sign. Correct typo.
711 (-mcheck-zero-division, -mcpu): Fix under/overfull hbox.
712 (-mpt-fixed): Use @minus{} for minus sign.
713 (Using Precompiled Headers): Fix under/overfull hbox.
715 2007-01-20 Jan Hubicka <jh@suse.cz>
717 * tree-flow.h (struct stmt_ann_d): Move references_memory to proper
718 place within annotation.
720 2007-01-20 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
722 * pa.c (output_move_double): Change array size of xoperands to 4.
724 2007-01-20 Jan Hubicka <jh@suse.cz>
726 * tree-tailcall.c (adjust_return_value): Do not use RESULT_DECL
729 2007-01-19 Ian Lance Taylor <iant@google.com>
731 * expmed.c (expand_divmod) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]: Cast
732 constant to unsigned HOST_WIDE_INT before negating.
734 2007-01-19 Ian Lance Taylor <iant@google.com>
736 * tree-ssa-operands.h (struct vuse_vec_d): Change num_vuse field
738 (VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Use unsigned comparison.
739 (VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
740 (SET_VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Likewise.
741 (SET_VUSE_ELEMENT_VAR) [ENABLE_CHECKING]: Likewise.
742 (SET_VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
743 (realloc_vdef, realloc_vuse): Change second parameter to
745 (ssa_operand_iterator_d): Change vuse_index and mayuse_index
747 * tree-ssa-operands.c (realloc_vop): Change num_elem parameter to
748 unsigned. Change x and lim locals to unsigned.
749 (realloc_vdef, realloc_vuse): Change num_elem parameter to
751 (finalize_ssa_vuse_ops): Change old_i local to unsigned.
752 (copy_virtual_operands): Change i and n locals to unsigned.
754 2007-01-19 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
757 * builtins.c (fold_builtin_1): Handle builtin fdim.
759 2007-01-20 Jan Hubicka <jh@suse.cz>
761 * tree-ssa.c (init_tree_ssa): Do not call init_alias_heapvars.
762 * tree-ssa-structalias.c (compute_points_to_sets): Do call
764 (init_alias_heapvars): Initialize only when not already initialized.
765 (delete_alias_heapvars): Set heapvar_for_stmt to NULL.
767 2007-01-19 Roger Sayle <roger@eyesopen.com>
769 * common.opt (fsigned-zeros): New command line option.
770 * flags.h (HONOR_SIGNED_ZEROS): Control via flag_signed_zeros instead
771 of flag_unsafe_math_optimizations.
772 * opts.c (set_fast_math_flags): The -ffast-math command line option
773 implies -fno-signed-zeros.
774 (fast_math_flags_set_p): Likewise.
776 * doc/invoke.texi: Document new -fno-signed-zeros option, and update
777 the documentation of -ffast-math appropriately. Wrap long lines.
779 2007-01-19 Steve Ellcey <sje@cup.hp.com>
781 * system.h (ASM_MAKE_LABEL_LINKONCE): Poison.
782 * varasm.c (globalize_decl): Remove ASM_MAKE_LABEL_LINKONCE ifdef.
784 2007-01-19 Tomas Bily <tbily@suse.cz>
786 * cgraphunit.c (cgraph_finalize_function): Updating of pid
788 (tree_init_ic_make_global_vars): New function
789 (tree_init_edge_profiler): call of tree_init_ic_make_global_vars
790 (tree_gen_ic_profiler): New function
791 (tree_gen_ic_func_profiler): New function
792 (tree_profiling): Added calling of tree_gen_ic_func_profiler
793 (tree_profile_hooks): Added hook for indirec/virtual calls
794 * value-prof.c (tree_find_values_to_profile): New case for
796 (tree_values_to_profile): Call for determining indirect/virtual
798 (tree_indirect_call_to_profile): New function
799 (tree_ic_transform): New function
800 (tree_ic): New function
801 (find_func_by_pid): New function
802 (init_pid_map): New function
803 (tree_value_profile_transformations): Added check for
804 indirect/virtual call transformation
805 * value-prof.h (enum hist_type): New counter type for
806 indirect/virtual calls
807 (profile_hooks): Added new hook for profiling indirect/virtual
809 * profile.c (instrument_values): New case for indirect/virtual
811 * gcov-io.h (GCOV_LAST_VALUE_COUNTER): Changed to 6
812 (GCOV_COUNTER_V_INDIR): New counter type
813 (GCOV_COUNTER_NAMES): New name of counter "indirect" added
814 (GCOV_MERGE_FUNCTIONS): New merge function for indirect/virtual
816 * cgraph.c: Definition of cgraph_max_pid
817 (cgraph_create_node): Default init of pid attribute
818 * cgraph.h: Declaration of cgraph_max_pid
819 (struct cgraph_node): Added pid attribute
820 * libgcov.c (__gcov_indirect_call_profiler): New function
821 (__gcov_one_value_profiler_body): New function
822 (__gcov_one_value_profiler): Body was moved to
823 __gcov_one_value_profiler_body and calls it
825 2007-01-19 Basile Starynkevitch <basile@starynkevitch.net>
827 * doc/gty.texi (Options): Document the mark_hook option to GTY.
828 * gengtype.c (write_types_data, write_func_for_structure,
829 write_types, ggc_wtd, pch_wtd): Add skip_hooks to
830 write_types_data, ggc_wtd, pch_wtd for processing mark_hook.
831 (walk_type, write_func_for_structure): Generate the mark_hook if
834 2007-01-19 Jan Hubicka <jh@suse.cz>
836 * ipa-inline.c (cgraph_decide_inlining_incrementally): Instead of
837 'early' argument take inlining mode argument specifying whether to
838 inline for size/speeed or all functions; add support for flattening;
840 (cgraph_early_inlining): Update call of decide_inlining_incrementally.
842 2007-01-19 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
845 * toplev.c (warn_deprecated_use): Use %qD instead of %qs to print
846 the name of the declared identifier.
848 2007-01-19 Dirk Mueller <dmueller@suse.de>
850 * config/i386.h (CONDITIONAL_REGISTER_USAGE): Store
851 result of PIC_OFFSET_TABLE_REGNUM in temporary variable to avoid
852 duplicate evaluation.
854 2007-01-19 Uros Bizjak <ubizjak@gmail.com>
856 * config/i386/i386.md (acos<mode>2): Rename from acossf2 and acosdf2.
857 Macroize expander using X87MODEF12 mode macro. Extend operand 1
858 to XFMode, use acosxf2 and truncate result to requested mode.
859 Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
860 (asin<mode>2): Similarly, with asin expanders.
861 (*fscalexf4_i387): Rename from *fscalexf4.
862 (expNcorexf3): New expander.
863 (expxf2, exp10xf2, exp2xf2): Use expNcorexf3 expander.
864 (exp<mode>2): Rename from expsf2 and expdf2. Macroize expander using
865 X87MODEF12 mode macro. Extend operand 1 to XFMode, use expxf2 and
866 truncate result to requested mode. Use SSE_FLOAT_MODE_P to disable
867 patterns for SSE math.
868 (exp10<mode>2): Similarly, with exp10 expanders.
869 (exp2<mode>2): Similarly, with exp2 expanders.
870 (expm1<mode>2): Similarly, with expm1 expanders.
871 (ldexp<mode>3): Similarly, with ldexp expanders.
872 (log<mode>2, log10<mode>2, log2<mode>2, log1p<mode>2, logb<mode>2):
873 Use gen_truncxf<mode>2_i387_noop to truncate result.
875 2007-01-19 Richard Sandiford <richard@codesourcery.com>
877 * config/m68k/m68k.h (PREFERRED_STACK_BOUNDARY): Define to 32
878 for ColdFire targets.
880 2007-01-19 Nathan Sidwell <nathan@codesourcery.com>
881 Richard Sandiford <richard@codesourcery.com>
883 * config/m68k/m68k.h (M68K_STATIC_CHAIN_REG_NAME): New macro.
884 (INITIALIZE_TRAMPOLINE): Use STATIC_CHAIN_REGNUM.
885 (__transfer_from_trampoline): Use M68K_STATIC_CHAIN_REG_NAME.
886 * config/m68k/m68kelf.h (STATIC_CHAIN_REGNUM): Override.
887 (M68K_STATIC_CHAIN_REG_NAME): Likewise.
888 * config/m68k/netbsd-elf.h (M68K_STATIC_CHAIN_REG_NAME): Likewise.
890 2007-01-19 Richard Sandiford <richard@codesourcery.com>
892 * config/m68k/m68k.md (adddi_dilshr32): Rename to...
893 (*adddi_dilshr32): ...this. Fix formatting. Remove commented-out
894 non-canonical pattern. Restrict to !TARGET_COLDFIRE.
895 (*adddi_dilshr32_cf): New pattern.
896 (adddi3, subdi3): Remove first alternatives.
898 2007-01-19 Richard Sandiford <richard@codesourcery.com>
900 * config/m68k/m68k.c (notice_update_cc): If an SFmode move is
901 implemented using move.l, do not use its cc result for floating-point
904 2007-01-19 Richard Sandiford <richard@codesourcery.com>
906 * config/m68k/m68k.h (EXTRA_CONSTRAINT): Stop the 'T' constraint
907 from accepting 's' constraints if flag_pic.
909 2007-01-19 Richard Sandiford <richard@codesourcery.com>
911 * config/m68k/m68k.md (bordered, bunordered, buneq, bunge, bungt)
912 (bunle, bunlt, bltgt, bordered_rev, bunordered_rev, buneq_rev)
913 (bunge_rev, bungt_rev, bunle_rev, bunlt_rev, bltgt_rev): Change
914 condition from TARGET_68881 to TARGET_HARD_FLOAT.
916 2007-01-19 Sandra Loosemore <sandra@codesourcery.com>
918 * longlong.h (count_leading_zeros, COUNT_LEADING_ZEROS_0): Add
919 ColdFire alternatives.
920 * config/m68k/m68k.h (CLZ_DEFINED_VALUE_AT_ZERO): New macro.
921 * config/m68k/m68k.md (clzsi2): Define for ColdFire
922 architectures that support ff1 instruction.
924 2007-01-19 Richard Sandiford <richard@codesourcery.com>
925 Julian Brown <julian@codesourcery.com>
927 * config/m68k/m68k.h (CONST_OK_FOR_LETTER_P): Add an 'R' case.
928 * config/m68k/m68k.md (*movsi_cfv4): Fold into...
929 (*movsi_cf): ...here. Remove unnecessary 'R' from 'Rg'.
932 2007-01-19 Richard Sandiford <richard@codesourcery.com>
934 * config/m68k/m68k-protos.h (valid_mov3q_const): Take a HOST_WIDE_INT
936 (output_move_const_into_data_reg, output_move_simode_const): Delete.
937 * config/m68k/m68k.c (const_method, const_int_cost): Take a
938 HOST_WIDE_INT instead of an rtx.
939 (m68k_rtx_costs): Update call accordingly.
940 (output_move_const_into_data_reg): Likewise. Fix formatting.
941 (valid_mov3q_const): Take a HOST_WIDE_INT instead of an rtx.
943 (output_move_simode_const): Update calls after above changes.
944 Rework to use automatic variables and predicates like MEM_P.
945 * config/m68k/m68k.md (pushexthisi_const): Update call to
948 2007-01-19 Dirk Mueller <dmueller@suse.de>
950 * tree-ssa-alias.c (perform_var_substitution): Fix typo
953 2007-01-19 Richard Guenther <rguenther@suse.de>
955 * builtins.c (expand_builtin_cexpi): Fall back to expanding
956 via cexp in case sincos is not available.
958 2007-01-19 Richard Guenther <rguenther@suse.de>
960 * doc/tm.texi (TARGET_HAS_SINCOS): Document new target macro.
961 * defaults.h (TARGET_HAS_SINCOS): Default to off.
962 * config/linux.h (TARGET_HAS_SINCOS): Set to on if we have glibc.
963 * config/alpha/linux.h (TARGET_HAS_SINCOS): Likewise.
964 * config/sparc/linux.h (TARGET_HAS_SINCOS): Likewise.
965 * config/sparc/linux64.h (TARGET_HAS_SINCOS): Likewise.
966 * config/rs6000/linux.h (TARGET_HAS_SINCOS): Likewise.
967 * config/rs6000/linux64.h (TARGET_HAS_SINCOS): Likewise.
969 2007-01-19 Uros Bizjak <ubizjak@gmail.com>
971 * config/i386/i386.md (*fpatanxf3_i387, fpatan_extend<mode>xf3_i387):
973 (atan2sf3_1, atan2df3_1, atan2xf3_1): Remove insn patterns.
974 (atan2xf3): Directly generate RTL pattern.
975 (atan2<mode>3): Rename from atan2sf3 and atan2df3 and macroize insn
976 patterns using X87MODEF12 mode macro. Use fpatan_extend<mode>xf3_i387
977 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
978 disable patterns for SSE math.
979 (atan<mode>2): Rename from atansf2 and atandf2 and macroize insn
980 patterns using X87MODEF12 mode macro. Use fpatan_extend<mode>xf3_i387
981 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
982 disable patterns for SSE math.
984 2007-01-19 Alexandre Oliva <aoliva@redhat.com>
986 * libgcc-std.ver: Fix typo in %inherit for GCC_4.3.0.
988 2007-01-18 Roger Sayle <roger@eyesopen.com>
990 * fold-const.c (fold_unary) <VIEW_CONVERT_EXPR>: Optimize away a
991 VIEW_CONVERT_EXPR to the same type as it's operand.
993 2007-01-18 David Edelsohn <edelsohn@gnu.org>
995 * config/rs6000/darwin-ldouble.c: Only build _SOFT_FLOAT if
996 configured for long double 128.
998 2007-01-18 Mike Stump <mrs@apple.com>
1000 * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Fix build
1003 2007-01-18 Michael Meissner <michael.meissner@amd.com>
1005 * i386.c (ix86_compute_frame_layout): Make fprintf's in #if 0 code
1008 2007-01-18 Jan Hubicka <jh@suse.cz>
1010 * tree-ssa-operands.c (vop_free_bucket_size): Never return value
1011 greater than NUM_VOP_FREE_BUCKETS.
1013 2007-01-18 Daniel Berlin <dberlin@dberlin.org>
1015 * tree-ssa-structalias.c: Update comments.
1016 (ptabitmap_obstack): Removed.
1018 (oldpta_obstack): Ditto.
1019 (stats): Add a few members.
1020 (struct variable_info): Remove node, complex, address_taken, and
1021 indirect_target members. Add oldsolution member.
1022 (new_var_info): Do not initialize removed members.
1023 (constraint_expr_type): Remove INCLUDES.
1024 (constraint_graph): Add size, implicit_preds, rep,
1025 indirect_cycles, eq_rep, label, direct_nodes, and complex members.
1026 (FIRST_REF_NODE): New macro.
1027 (LAST_REF_NODE): Ditto.
1028 (FIRST_ADDR_NODE): Ditto.
1029 (find): New function.
1031 (dump_constraint): Do not handle INCLUDES.
1032 (insert_into_complex): Do not insert duplicate constraints.
1033 (condense_varmap_nodes): Renamed and rewritten into ...
1034 (merge_node_constraints): This. Also fix bug in handling of
1035 offseted copy constraints.
1036 (clear_edges_for_node): No longer need to deal with preds at all,
1037 or removing associated preds/succs.
1038 (merge_graph_nodes): Deal with indirect_cycles.
1039 Don't deal with predecessors.
1040 (add_implicit_graph_edge): New function.
1041 (add_pred_graph_edge): Ditto.
1042 (add_graph_edge): Don't deal with predecessors.
1043 (build_constraint_graph): Removed.
1044 (build_pred_graph): New function.
1045 (build_succ_graph): Ditto.
1046 (struct scc_info): Removed in_component. Added roots, dfs, and
1047 node_mapping. Remove visited_index, unification_queue.
1048 (scc_visit): Deal with union-find we do now.
1049 Deal with cycles with REF nodes.
1050 (collapse_nodes): Renamed and rewritten to ...
1051 (unify_nodes): This.
1052 (process_unification_queue): Removed.
1053 (topo_visit): Cleanup
1054 (do_da_constraint): Use find.
1055 (do_sd_constraint): Ditto.
1056 (do_ds_constraint): Ditto.
1057 (do_complex_constraint): Ditto.
1058 (init_scc_info): Update for removed and added members.
1059 (find_and_collapse_graph_cycles): Renamed and rewritten into ...
1060 (find_indirect_cycles): This.
1061 (equivalence_class): New variable.
1062 (label_visit): New function.
1063 (perform_variable_substitution): Rewritten.
1064 (free_var_substitution_info): New function.
1065 (find_equivalent_node): Ditto.
1066 (move_complex_constraints): Ditto.
1067 (eliminate_indirect_cycles): Ditto.
1068 (solve_graph): Only propagate changed bits.
1069 Use indirect cycle elimination.
1071 (tree_id_t): Rename to tree_vi_t, delete id member, add vi member.
1072 (tree_id_eq): Renamed to ...
1073 (tree_vi_eq): This. Update for member change
1074 (insert_id_for_tree): Renamed and rewritten to ...
1075 (insert_vi_for_tree): This.
1076 (lookup_id_for_tree): Renamed and rewritten to ...
1077 (lookup_vi_for_tree): This.
1078 (get_id_for_tree): Renamed and rewritten to ...
1079 (get_vi_for_tree): Ditto.
1080 (get_constraint_exp_from_ssa_var): Update to use get_vi_for_tree.
1081 (process_constraint): Don't handle INCLUDES.
1082 Remove special ADDRESSOF case.
1083 (find_func_aliases): Rewrite to use vi functions instead of id
1085 (create_function_info_for): Ditto.
1086 (create_variable_info_for): Ditto.
1087 (intra_create_variable_infos): Ditto.
1088 (merge_smts_into): Ditto.
1089 (find_what_p_points_to): Ditto.
1090 (init_base_vars): Ditto.
1091 (init_alias_vars): Ditto.
1092 (remove_preds_and_fake_succs): New function.
1093 (dump_sa_points_to_info): Dump new stats.
1094 (dump_solution_for_var): Use find.
1095 (set_used_smts): Fix formatting.
1096 (compute_points_to_sets): Updated for new functions.
1097 (ipa_pta_execute): Ditto.
1099 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
1100 Richard Sandiford <richard@codesourcery.com>
1102 * doc/tm.texi (TARGET_FUNCTION_VALUE): Expand documentation of
1104 * calls.c (expand_call): If the return value is a PARALLEL,
1105 extract its first member.
1106 * config/m68k/linux.h (FUNCTION_EXTRA_EPILOGUE): Remove.
1107 * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
1108 use FUNCTION_EXTRA_EPILOGUE.
1109 (m68k_function_value): Return a PARALLEL if the return value
1110 is of a pointer type.
1111 * config/m68k/netbsd-elf.h (current_function_returns_pointer)
1112 (FUNCTION_EXTRA_EPILOGUE): Remove.
1113 * config/m68k/m68k.md (D0_REG): New constant.
1115 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
1117 * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
1118 output a NOP for empty epilogues.
1120 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1122 * config/m68k/m68k.c (m68k_use_return_insn): Update comments
1123 before function. Extend register save check to include all
1124 registers, not just integer ones.
1126 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
1128 * config/m68k/m68k-protos.h (use_return_insn): Rename to...
1129 (m68k_use_return_insn): ...this.
1130 * config/m68k/m68k.h (USE_RETURN_INSN): Delete.
1131 * config/m68k/m68k.c (use_return_insn): Rename to...
1132 (m68k_use_return_insn): ...this.
1133 * config/m68k/m68k.md (return): Use m68k_use_return_insn instead
1136 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
1138 * config/m68k/fpgnulib.c (__truncdfsf2): Implement round to
1139 nearest even, fix denormal rounding overflow.
1141 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1143 * config/m68k/m68k.md (movsf_cf_hard): Use fsmove instead of
1144 f%$smove and f%$move.
1145 (movdf_cf_hard): Use fdmove for cases 0 and 3 and fmove for case 1.
1146 (extendsfdf2_cf): Use fdmove instead of f%&move.
1147 (truncdfsf2_cf): Use fsmove instead of f%$smove.
1148 (add<mode>3_cf, sub<mode>3_cf): Use <FP:prec> instead of <FP:round>.
1150 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1152 * config/m68k/m68k.md (movdf_cf_hard): Use output_move_double for
1155 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1157 * real.h (real_format): Add a canonical_nan_lsbs_set field.
1158 (coldfire_single_format): Declare.
1159 (coldfire_double_format): Likewise.
1160 * real.c (encode_ieee_single): Use canonical_nan_lsbs_set instead
1161 of qnan_msb_set to determine the lower bits of a canonical
1163 (encode_ieee_double): Likewise.
1164 (encode_ieee_quad): Likewise.
1165 (ieee_single_format): Initialize canonical_nan_lsbs_set.
1166 (mips_single_format): Likewise.
1167 (ieee_double_format): Likewise.
1168 (mips_double_format): Likewise.
1169 (ieee_extended_motorola_format): Likewise.
1170 (ieee_extended_intel_96_format): Likewise.
1171 (ieee_extended_intel_128_format): Likewise.
1172 (ieee_extended_intel_96_round_53_format): Likewise.
1173 (ibm_extended_format): Likewise.
1174 (mips_extended_format): Likewise.
1175 (ieee_quad_format): Likewise.
1176 (mips_quad_format): Likewise.
1177 (vax_f_format): Likewise.
1178 (vax_d_format): Likewise.
1179 (vax_g_format): Likewise.
1180 (i370_single_format): Likewise.
1181 (i370_double_format): Likewise.
1182 (decimal_single_format): Likewise.
1183 (decimal_double_format): Likewise.
1184 (decimal_quad_format): Likewise.
1185 (c4x_single_format): Likewise.
1186 (c4x_extended_format): Likewise.
1187 (real_internal_format): Likewise.
1188 (coldfire_single_format): New real_format.
1189 (coldfire_double_format): Likewise.
1190 * config/pdp11/pdp11.c (pdp11_f_format): Initialize
1191 canonical_nan_lsbs_set.
1192 (pdp11_d_format): Likewise.
1193 * config/m68k/m68k.c (override_options): Override REAL_FORMAT_MODE
1194 if TARGET_COLDFIRE_CPU.
1196 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1198 * config/m68k/m68k-protos.h (m68k_output_pic_call): Delete.
1199 (output_call, m68k_legitimize_call_address): Declare.
1200 * config/m68k/m68k.h (EXTRA_CONSTRAINT): Remove unnecessary
1201 parenthesees. Add support for a 'W' constraint.
1202 (LEGITIMATE_PIC_OPERAND_P): Remove SYMBOL_REF_FLAG handling.
1203 (PRINT_OPERAND_PUNCT_VALID_P): Remove comment about 'o'.
1204 (m68k_symbolic_call, m68k_symbolic_jump): Declare.
1205 * config/m68k/m68k.c (m68k_symbolic_call, m68k_symbolic_jump): New
1207 (override_options): Initialize them. Do not set flag_no_function_cse
1208 for TARGET_ID_SHARED_LIBRARY.
1209 (m68k_output_pic_call): Delete.
1210 (m68k_legitimize_call_address): New function.
1211 (print_operand): Remove the %o prefix. Handle the %p prefix.
1212 (output_call): New function.
1213 (m68k_output_mi_thunk): Use m68k_symbolic_jump. Always load the
1214 target address from the GOT if symbolic jumps are not allowed.
1215 * config/m68k/m68k.md (call, general_operand): Do not set
1216 SYMBOL_REF_FLAG. Use m68k_legitimize_call_address instead.
1217 Merge separate flag_pic and !flag_pic define_insns into...
1218 (*call, *call_value): ...these new patterns. Match the address
1219 rather than the containing MEM and require it to be a call_operand.
1220 Use output_call to generate the asm template.
1221 * config/m68k/predicates.md (const_call_operand): New predicate.
1222 (call_operand): Likewise.
1224 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
1226 * config/m68k/m68k.h (REGISTER_MOVE_COST): Simplify definition.
1227 (STACK_GROWS_DOWNWARD): Define to 1.
1228 (FUNCTION_VALUE, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P): Equivocate
1229 comments, emphasizing that these values are only defaults.
1230 * config/m68k/linux.h (LINK_SPEC): Fix formatting in #undef.
1231 * config/m68k/m68k.c (const_method): Remove trailing whitespace.
1233 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1235 * config/m68k/m68k.md (cmpsi): Remove outdated flag_pic handling.
1237 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
1238 Richard Sandiford <richard@codesourcery.com>
1240 * config/m68k/m68k.h (DATA_REGNO_P, ADDRESS_REGNO_P, INT_REGNO_P)
1241 (FP_REGNO_P): New macros.
1242 (REGNO_OK_FOR_INDEX_P, REGNO_OK_FOR_BASE_P, REGNO_OK_FOR_DATA_P)
1243 (REGNO_OK_FOR_FP_P, REG_OK_FOR_INDEX_P, REG_OK_FOR_BASE_P): Use them.
1244 (EH_RETURN_STACKADJ_RTX): Use A0_REG.
1245 * config/m68k/m68k.c (m68k_regno_mode_ok): Use the new REGNO macros.
1247 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
1249 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
1250 (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-rtems*): Add t-floatlib
1252 * config/m68k/t-floatlib: New file, extracting common code from...
1253 * config/m68k/t-m68kbare, config/m68k/t-m68kelf,
1254 * config/m68k/t-uclinux: Here.
1255 * config/m68k/fpgnulib.c: Do not compile extendeed precision
1256 routines on ColdFire targets.
1258 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
1260 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
1261 (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*)
1262 (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-linux*)
1263 (m68k-*-rtems*): Use tm_file rather than m68k/m68k.h and
1264 explicitly set MOTOROLA to 1.
1265 * config/m68k/m68k.h (MOTOROLA): Simplify definition accordingly.
1267 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1268 Nathan Sidwell <nathan@codesourcery.com>
1270 * config/m68k/m68k.h (PCC_STATIC_STRUCT_RETURN): Delete.
1271 (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Add MOTOROLA cases.
1272 * config/m68k/coff.h (REGISTER_PREFIX_MD): Delete.
1273 * config/m68k/m68020-elf.h (LIB_SPEC): Delete.
1274 * config/m68k/m68k-none.h (CC1_SPEC, CPP_SUBTARGET_SPEC): Delete.
1275 * config/m68k/m68kelf.h (IMMEDIATE_PREFIX, REGISTER_PREFIX_MD)
1276 (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Delete.
1277 (NO_DOLLAR_IN_LABEL, PCC_STATIC_STRUCT_RETURN): Don't undefine.
1278 (BSS_ASM_OP, ASM_OUTPUT_SKIP, ASM_OUTPUT_ASCII): Delete.
1279 * config/m68k/m68kemb.h (PCC_STATIC_STRUCT_RETURN): Don't undefine.
1280 * config/m68k/linux.h (SIZE_TYPE, PTRDIFF_TYPE, WCHAR_TYPE)
1281 (WCHAR_TYPE_SIZE, TARGET_OBJFMT_CPP_BUILTINS): Delete.
1282 (TARGET_OS_CPP_BUILTINS): Don't define mc68000 and mc68020 here.
1283 (DBX_REGISTER_NUMBER): Delete.
1284 * config/m68k/netbsd-elf.h (IMMEDIATE_PREFIX): Delete.
1285 (PCC_STATIC_STRUCT_RETURN): Don't undefine.
1286 * config/m68k/openbsd.h (PCC_STATIC_STRUCT_RETURN): Define.
1288 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1290 * config.gcc (m68k-*-uclinux*): Add flat.h to $tm_file.
1291 * config/flat.h: New file.
1292 * crtstuff.c (USE_PT_GNU_EH_FRAME): Don't define if
1294 * config/m68k/m68k.h (ASM_PREFERRED_EH_DATA_FORMAT): Do not use
1295 indirect references for -msep-data or -mid-shared-library.
1296 Do not use PC-relative code addresses either.
1297 * config/m68k/m68k.c (override_options): Restrict -fPIC error
1299 * config/m68k/uclinux.h (STARTFILE_SPEC): Define. Use Scrt1.o
1300 for shared libraries and crt1.o for executables. Use crti.o and
1302 (ENDFILE_SPEC): Use crtend.o and crtn.o.
1303 (LIB_SPEC): Suppress -Rlibc.gdb if -static-libc is given.
1304 Do not add -elf2flt or -shared-lib-id options here.
1305 (LINK_SPEC): Define. Pass -elf2flt if no -elf2flt option is given.
1306 Pass -shared-lib-id if -mid-shared-library, taking the library
1307 identifier from -mshared-library-id if given, otherwise
1309 (EH_FRAME_IN_DATA_SECTION): Do not undefine.
1310 (INIT_SECTION_ASM_OP, FINI_SECTION_ASM_OP): Likewise.
1311 (TARGET_OS_CPP_BUILTINS): Define __GXX_MERGED_TYPEINFO_NAMES=0
1312 and __GXX_TYPEINFO_EQUALITY_INLINE=0 if -mid-shared-library.
1313 (DRIVER_SELF_SPECS): Map unadorned PIC options to -msep-data.
1314 * config/m68k/t-uclinux (EXTRA_MULTILIB_PARTS): Add crtbegin.o
1316 * config/m68k/lb1sf68.asm (PICCALL): Use an lea and pc-relative
1317 jump sequence for ISA A and ISA A+.
1318 (PICJUMP): Likewise.
1320 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1322 * config.gcc (m68k-*-uclinux*): Base the port on the common
1323 and m68k GNU/Linux files rather than on the generic ELF ones.
1324 * config/m68k/uclinux.h (TARGET_VERSION): Override.
1325 (TARGET_OS_CPP_BUILTINS): Use LINUX_TARGET_OS_CPP_BUILTINS.
1327 2007-01-18 Julian Brown <julian@codesourcery.com>
1328 Richard Sandiford <richard@codesourcery.com>
1330 * config/m68k/m68k.h (LONG_DOUBLE_TYPE_SIZE): Make 64-bit on ColdFire.
1331 (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
1332 * config/m68k/netbsd-elf.h (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Undefine
1334 * config/m68k/uclinux-oldabi.h (LONG_DOUBLE_TYPE_SIZE): Redefine to
1336 (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
1338 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1340 * doc/install.texi: Document m68k-uclinuxoldabi.
1341 * config.gcc (m68k-*-uclinuxoldabi): New configuration.
1342 * config/m68k/uclinux-oldabi.h: New file, copied from
1343 config/m68k/uclinux.h.
1345 2007-01-18 Kaz Kojima <kkojima@rr.iij4u.or.jp>
1347 * config/m32r/m32r-protos.h (m32r_expand_epilogue): Declare it.
1348 * config/m32r/m32r.c (m32r_setup_incoming_varargs): Use gen_frame_mem.
1349 (m32r_compute_frame_size): Use unsigned for regno.
1350 (m32r_reload_lr): Use gen_frame_mem.
1352 (m32r_output_function_epilogue): Don't output the function epilogue
1354 (m32r_expand_epilogue): New.
1355 (direct_return): Return false if the function has the interrupt
1357 (m32r_hard_regno_rename_ok): Remove code for the textual epilogue.
1358 * config/m32r/m32r.md (epilogue): New expander.
1359 (return_lr, return_rte): New insns.
1360 (return): Make it expander.
1361 (return_normal): New expander.
1363 2007-01-18 Josh Conner <jconner@apple.com>
1366 * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Add
1367 support for UNLE, UNLT, UNGE, and UNGT.
1369 2007-01-18 Jan Hubicka <jh@suse.cz>
1371 * tree-vrp.c (finalize_jump_threads): Do not call cleanup_cfg by hand.
1372 * tree-tailcall (add_virtual_phis): Likewise.
1373 (optimize_tail_call): Return TODOs.
1374 (execute_tail_calls): Return TODOs.
1375 * tree-ssa-ccp (execute_fold_all_builtins): Do cleanup_cfg via TODO.
1376 * tree-cfgcleanup.c (cleanup_tree_cfg_loop): Return if something
1378 * tree-ssa-forwprop.c (tree_ssa_forward_propagate_single_use_value):
1379 Cleanup cfg using TODO.
1380 * tree-flow.h (cleanup_tree_cfg_loop): Update prototype.
1381 * passes.c (execute_function_todo): When cleanup did something, remove
1383 * tree-cfg.c (pass_build_cfg): Add cleanup_cfg TODO.
1384 (make_edges): Don't cleanup_cfg.
1386 2007-01-18 Uros Bizjak <ubizjak@gmail.com>
1388 * reg-stack.c (subst_stack_regs_pat) [UNSPEC_SINCOS_COS,
1389 UNSPEC_XTRACT_FRACT]: Use generic code for instructions that
1390 operate on the top of stack.
1391 [UNSPEC_SINCOS_SIN, UNSPEC_XTRACT_EXP, UNSPEC_TAN]: Rewrite
1392 register handling of instructions that output to the second
1394 [UNSPEC_TAN_ONE, UNSPEC_TAN_TAN]: Remove.
1395 (move_for_stack_reg): Special-case check for dead destination
1396 stack slot for constant load of 1.0 inside UNSPEC_TAN.
1398 * config/i386/i386.md (UNSPEC_TAN): New constant.
1399 (UNSPEC_TAN_ONE, UNSPEC_TAN_TAN): Remove.
1400 (fptanxf4_i387, fptan_extend<mode>xf4_i387): New patterns
1401 to correctly model move of constant 1.0 to top stack slot.
1402 (*tandf3_1, *tansf3_1, *tanxf3_1): Remove insn patterns.
1403 (unnamed peephole2 pattern): Remove corresponding peephole2
1404 pattern that optimizes tan insn and loading of constant 1.0.
1405 (tanxf2): Use fptanxf4_i387.
1406 (tan<mode>2): Rename from tansf2 and tandf2 and macroize insn
1407 patterns using X87MODEF12 mode macro. Use fptan_extend<mode>xf4_i387
1408 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
1409 disable patterns for SSE math.
1410 (sincos<mode>3): Use truncxf<mode>2_i387_noop for truncation.
1411 (fyl2x_extend<mode>xf3_i387): Use X87MODEF12 for operand 1.
1413 2007-01-18 Dirk Mueller <dmueller@suse.de>
1414 Richard Guenther <rguenther@suse.de>
1417 * doc/invoke.texi (Warray-bounds): Document -Warray-bounds.
1418 * common.opt (Warray-bounds): Add new warning option.
1419 * c-opts.c (c_common_handle_option): Define -Warray-bounds
1421 * Makefile.in: make tree-vrp.o depend on toplev.h
1422 * tree-vrp.c (vrp_finalize): Call check_array_refs if -Warray-bounds
1424 (check_array_refs, check_array_bounds, check_array_ref): New.
1426 2007-01-18 Jan Hubicka <jh@suse.cz>
1428 * tree-ssa-ccp.c (ccp_finalize): Return if something changed.
1429 (execute_ssa_ccp): Return flags conditionally.
1430 * tree-ssa-propagate.c (substitue_and_fold): Return if something was
1432 * tree-ssa-propagate.h (substitute_and_fold): Update prototype.
1434 2007-01-18 Steven Bosscher <steven@gcc.gnu.org>
1436 * cfgcleanup.c (cleanup_cfg): Detect cfglayout mode and set
1437 the CLEANUP_CFGLAYOUT flag when in cfglayout mode.
1439 * Makefile.c (GTFILES): Add cfglayout.h.
1440 * gengtype.c (open_base_files): Likewise.
1441 * cfglayout.c (cfg_layout_function_footer,
1442 cfg_layout_function_header) Reindent to make gengtype happy.
1443 * cfglayout.h (cfg_layout_function_footer,
1444 cfg_layout_function_header): Add GTY(()) marker.
1446 * ifcvt.c (noce_try_sign_mask): Make sure INSN_B is non-null.
1448 2007-01-18 Ben Elliston <bje@au.ibm.com>
1450 * genautomata.c (write_automata): Include xstrerror output in the
1451 error message if writing the DFA description file fails.
1453 2007-01-17 H.J. Lu <hongjiu.lu@intel.com>
1455 * config/mips/mips-protos.h (mips_output_external): Make it
1457 * config/mips/iris.h (TARGET_ASM_EXTERNAL_LIBCALL): Removed.
1458 * config/mips/mips.c (irix_output_external_libcall): Likewise.
1459 (extern_list): Likewise.
1460 (extern_head): Likewise.
1461 (TARGET_ASM_FILE_END): Likewise.
1462 (mips_file_end): Likewise.
1463 (mips_output_external): Rewritten.
1465 2007-01-18 Ben Elliston <bje@au.ibm.com>
1467 * genpreds.c (write_insn_preds_c): Only write out the function
1468 body for regclass_for_constraint if we have register constraints.
1470 2007-01-17 Tom Tromey <tromey@redhat.com>
1472 * doc/sourcebuild.texi (libgcj Tests): Use sourceware.org.
1473 * doc/install.texi (Testing): Use sourceware.org.
1474 (Binaries): Likewise.
1475 (Specific): Likewise.
1476 * doc/contrib.texi (Contributors): Use sourceware.org.
1478 2007-01-17 Anatoly Sokolov <aesok@post.ru>
1480 * config/avr/avr.h (AVR_HAVE_LPMX): New macro.
1481 (AVR_ENHANCED): Rename to ...
1482 (AVR_HAVE_MUL): ... new.
1483 (avr_enhanced_p): Rename to ...
1484 (avr_have_mul_p): ... new.
1485 (TARGET_CPU_CPP_BUILTINS): Use 'avr_have_mul_p' instead of
1486 'avr_enhanced_p' for "__AVR_ENHANCED__". Define "__AVR_HAVE_MUL__".
1487 * config/avr/avr.c (avr_enhanced_p): Rename to ...
1488 (avr_have_mul_p): ... new.
1489 (base_arch_s): Rename 'enhanced' to 'have_mul'.
1490 (avr_override_options): Use 'avr_have_mul_p' and 'have_mul' instead of
1491 'avr_enhanced_p' and 'enhanced'.
1492 (ashlhi3_out, ashrhi3_out, lshrhi3_out, avr_rtx_costs): Use
1493 AVR_HAVE_MUL instead of AVR_ENHANCED.
1494 * avr.md (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
1495 (mulqi3, *mulqi3_enh, *mulqi3_call, mulqihi3, umulqihi3, mulhi3,
1496 *mulhi3_enh, *mulhi3_call, mulsi3, *mulsi3_call): Use AVR_HAVE_MUL
1497 instead of AVR_ENHANCED.
1498 (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
1499 * libgcc.S: Use __AVR_HAVE_MUL__ instead of __AVR_ENHANCED__.
1500 (__tablejump__): Use __AVR_HAVE_LPMX__ instead of __AVR_ENHANCED__.
1502 2007-01-17 Ian Lance Taylor <iant@google.com>
1504 * vec.h (VEC_reserve_exact): Define.
1505 (vec_gc_p_reserve_exact): Declare.
1506 (vec_gc_o_reserve_exact): Declare.
1507 (vec_heap_p_reserve_exact): Declare.
1508 (vec_heap_o_reserve_exact): Declare.
1509 (VEC_OP (T,A,reserve_exact)): New static inline function, three
1511 (VEC_OP (T,A,reserve)) [all versions]: Remove handling of
1513 (VEC_OP (T,A,alloc)) [all versions]: Call ...reserve_exact.
1514 (VEC_OP (T,A,copy)) [all versions]: Likewise.
1515 (VEC_OP (T,a,safe_grow)) [all versions]: Likewise.
1516 * vec.c (calculate_allocation): Add exact parameter. Change all
1518 (vec_gc_o_reserve_1): New static function, from vec_gc_o_reserve.
1519 (vec_gc_p_reserve, vec_gc_o_reserve): Call vec_gc_o_reserve_1.
1520 (vec_gc_p_reserve_exact, vec_gc_o_reserve_exact): New functions.
1521 (vec_heap_o_reserve_1): New static function, from vec_heap_o_reserve.
1522 (vec_heap_p_reserve, vec_heap_o_reserve): Call vec_heap_o_reserve_1.
1523 (vec_heap_p_reserve_exact): New function.
1524 (vec_heap_o_reserve_exact): New function.
1526 2007-01-17 Jan Hubicka <jh@suse.cz>
1528 * ipa-type-escape.c (look_for_casts): Revamp using handled_component_p.
1530 2007-01-17 Eric Christopher <echristo@apple.com>
1532 * config.gcc: Support core2 processor.
1534 2007-01-16 Jan Hubicka <jh@suse.cz>
1536 * tree-ssanames.c (release_dead_ssa_names): Instead of ggc_freeing
1537 the names, just unlink the chain so we don't crash on dangling pointers
1540 2007-01-16 Jan Hubicka <jh@suse.cz>
1542 * cgraph.h (cgraph_decide_inlining_incrementally): Kill.
1543 * tree-pass.h: Reorder to make IPA passes appear toegher.
1544 (pass_early_inline, pass_inline_parameters, pass_apply_inline): Declare.
1545 * cgraphunit.c (cgraph_finalize_function): Do not compute inling
1546 parameters, do not call early inliner.
1547 * ipa-inline.c: Update comments. Include tree-flow.h
1548 (cgraph_decide_inlining): Do not compute inlining parameters.
1549 (cgraph_decide_inlining_incrementally): Return TODOs; assume to
1550 be called with function context set up.
1551 (pass_ipa_inline): Remove unreachable functions before pass.
1552 (cgraph_early_inlining): Simplify assuming to be called from the
1554 (pass_early_inline): New pass.
1555 (cgraph_gate_ipa_early_inlining): New gate.
1556 (pass_ipa_early_inline): Turn into simple wrapper.
1557 (compute_inline_parameters): New function.
1558 (gate_inline_passes): New gate.
1559 (pass_inline_parameters): New pass.
1560 (apply_inline): Move here from tree-optimize.c
1561 (pass_apply_inline): New pass.
1562 * ipa.c (cgraph_remove_unreachable_nodes): Verify cgraph after
1564 * tree-inline.c (optimize_inline_calls): Return TODOs rather than
1566 (tree_function_versioning): Do not allocate dummy struct function.
1567 * tree-inline.h (optimize_inline_calls): Update prototype.
1568 * tree-optimize.c (execute_fixup_cfg): Export.
1569 (pass_fixup_cfg): Remove
1570 (tree_rest_of_compilation): Do not apply inlines.
1571 * tree-flow.h (execute_fixup_cfg): Declare.
1572 * Makefile.in (gt-passes.c): New.
1573 * passes.c: Include gt-passes.h
1574 (init_optimization_passes): New passes.
1575 (nnodes, order): New static vars.
1576 (do_per_function_toporder): New function.
1577 (execute_one_pass): Dump current pass here.
1578 (execute_ipa_pass_list): Don't dump current pass here.
1580 2007-01-16 Janis Johnson <janis187@us.ibm.com>
1582 * config/dfp-bit.c (dfp_compare_op): Return separate value for NaN.
1583 (DFP_NE, DFP_LE, DFP_GE): Return false for NaN.
1585 2007-01-16 David Edelsohn <edelsohn@gnu.org>
1587 * config/rs6000/darwin-ldouble.c: Build file for SOFT_FLOAT.
1588 (strong_alias): Define.
1589 (__gcc_qmul): Provide non-FMA for soft-float.
1595 (__gcc_qunord): New.
1605 * config/rs6000/rs6000.c (rs6000_init_libfuncs): Initialize
1606 soft-float functions.
1607 * config/rs6000/libgcc-ppc-glibc.ver: Version soft-float symbols.
1608 * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Do not warn
1609 about long double soft float.
1611 2007-01-16 Dorit Nuzman <dorit@il.ibm.com>
1612 Tehila Meyzels <tehila@il.ibm.com>
1614 * tree-vectorizer.h (is_pattern_stmt_p): New.
1615 * tree-vect-analyze.c (vect_determine_vectorization_factor): Fix
1616 formatting (tabs instead of spaces). Cleanup and clarify setting
1617 of STMT_VINFO_VECTYPE. Call is_pattern_stmt_p.
1618 * tree-vect-transform.c (vect_get_vec_def_for_operand): Fix typo.
1619 (vectorizable_type_demotion): Check that types are integral.
1620 (vectorizable_type_promotion): Likewise.
1621 (vectorizable_store): Fix typo. Eliminate new-line at end of
1624 2007-01-16 Jan Hubicka <jh@suse.cz>
1626 * tree-ssanames.c (release_dead_ssa_names): Remove invalidated
1629 2007-01-15 Eric Christopher <echristo@apple.com>
1631 * ifcvt.c: Include vec.h, vecprim.h.
1632 (check_cond_move_block): New argument regs.
1633 Reorganize. Add registers used to regs.
1634 (cond_move_process_if_block): Use regs set above as
1637 2007-01-15 Eric Christopher <echristo@apple.com>
1639 * config/darwin.h: Update copyright.
1640 (TARGET_OPTION_TRANSLATE_TABLE): Add umbrella.
1641 (LINK_COMMAND_SPEC): Add -u.
1642 (LINK_SPEC): Fix umbrella for above.
1644 2007-01-15 Joseph S. Myers <joseph@codesourcery.com>
1646 * config/soft-fp/op-common.h, config/soft-fp/op-4.h: Update from
1649 2007-01-15 Tom Tromey <tromey@redhat.com>
1651 * doc/sourcebuild.texi (libgcj Tests): Don't mention jacks.
1652 * doc/install.texi (Testing): Don't mention jacks.
1653 (Configuration): Document --enable-java-maintainer-mode. Move
1654 --with-java-home to libgcj-specific section. Document
1656 (Prerequisites): Mention --enable-java-maintainer-mode, ecj1.
1658 2007-01-15 Jan Hubicka <jh@suse.cz>
1660 * tree-ssa-dce.c (DCE_TODOs): New.
1661 (propagate_necessity): Return if something changed.
1662 (eliminate_unnecessary_stmts): Likewise.
1663 (perform_tree_ssa_dce): Return TODO flags when needed.
1664 (pass_dce, pass_dce_loop, pass_cd_dce): Remove TODO flags.
1666 2007-01-15 Uros Bizjak <ubizjak@gmail.com>
1668 * config/i386/i386.md (fyl2xxf3_i387): Rename from fyl2x_xf3.
1669 (fyl2x_extend<mode>xf3_i387): New insn pattern.
1670 (log<mode>2): Rename from logsf2 and logdf2 and macroize insn
1671 patterns using X87MODEF12 mode macro. Extend operand 1
1672 to XFmode. Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
1673 (log10<mode>2): Ditto.
1674 (log2<mode>2): Ditto.
1675 (log1p<mode>2): Ditto.
1676 (logb<mode>2): Ditto.
1677 (fyl2xp1xf3_i387): Rename from fyl2xp1_xf3.
1678 (fyl2xp1_extend<mode>xf3_i387): New insn pattern.
1679 (*fxtractxf3_i387): Rename from *fxtractxf3.
1680 (fxtract_extend<mode>xf3_i387): New insn pattern.
1681 (ilogbsi2): Use match_dup 3, not match_operand:XF 3.
1682 * config/i386/i386.c (ix86_emit_i387_log1p): Use gen_fyl2xp1xf3_i387()
1683 and gen_fyl2xxf3_i387().
1685 2007-01-14 Zdenek Dvorak <dvorakz@suse.cz>
1687 * loop-unswitch.c (unswitch_loop): Do not call fix_loop_placement.
1688 * cfgloopmanip.c (fix_loop_placement): Made static. Use
1689 get_loop_exit_edges. Changed return type to bool.
1690 * cfgloop.h (fix_loop_placement): Declaration removed.
1692 2007-01-14 Dorit Nuzman <dorit@il.ibm.com>
1694 * param.h (MIN_VECT_LOOP_BOUND): New.
1695 * params.def (MIN_VECT_LOOP_BOUND): New.
1696 * tree-vectorizer.c (slpeel_tree_peel_loop_to_edge): Takes another
1697 argument - minimum threshold for number of iterations.
1698 * tree-vectorizer.h (slpeel_tree_peel_loop_to_edge): Add another
1699 argument to declaration.
1700 * tree-vect-analyze.c (vect_analyze_operations): Check value of
1701 MIN_VECT_LOOP_BOUND.
1702 * tree-vect-transform.c (vect_do_peeling_for_loop_bound): Call
1703 slpeel_tree_peel_loop_to_edge with additional argument.
1704 (vect_do_peeling_for_alignment): Likewise.
1705 * doc/invoke.texi (min-vect-loop-bound): Document new param option.
1707 2007-01-14 Uros Bizjak <ubizjak@gmail.com>
1710 * config/i386/i386.c (print_operand) ['z']: Output 'b' for
1713 2007-01-14 Jan Hubicka <jh@suse.cz>
1715 * tree-dfa.c (remove_referenced_var): New function.
1716 * tree-ssa-live.c (remove_unused_locals): Walk referenced vars and
1717 prune referenced vars list too.
1718 * tree-flow.h (remove_referenced_var): Declare.
1720 2007-01-14 Jan Hubicka <jh@suse.cz>
1722 * tree-eh.c (add_stmt_to_eh_region_fn): Do not add call_exprs
1724 (remove_stmt_from_eh_region_fn): Do not remove call_exprs.
1725 (verify_eh_throw_stmt_node, verify_eh_throw_table_statements): Kill.
1726 * except.h (verify_eh_throw_table_statements): Kill prototype.
1727 * cfgexpand.c (expand_gimple_basic_block): Propagate Eh regions
1728 into call exrepssions.
1729 * tree-optimize.c (execute_free_cfg_annotatiosn): Do not call
1731 * tree-cfg.c: Include pointer-set.h.
1732 (verify_node_sharing): Work on pointer set.
1733 (verify_eh_throw_stmt_node): New.
1734 (verify_stmts): Use pointers sets, verify throw_stmt.
1736 2007-01-13 Zdenek Dvorak <dvorakz@suse.cz>
1738 * ipa-reference.c (analyze_function): Consider also addresses taken
1741 2007-01-12 Roger Sayle <roger@eyesopen.com>
1743 * c-typeck.c (null_pointer_constant_p): Replace use of
1744 TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
1745 (build_c_cast): Likewise.
1747 2007-01-12 Roger Sayle <roger@eyesopen.com>
1749 * tree.h (force_fit_type_double): Remove unused final argument.
1750 * c-common.c (constant_expression_warning): Replace use of
1751 TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
1752 (convert_and_check): Likewise.
1753 (shorten_compare): Update call to force_fit_type_double.
1754 (c_common_truthvalue_conversion) <INTEGER_CST>: Use integer_zerop.
1755 * convert.c (convert_to_pointer): Update call to
1756 force_fit_type_double.
1757 * fold-const.c (force_fit_type_double): Remove overflowed_const
1759 (int_const_binop, fold_convert_const_int_from_int,
1760 fold_convert_const_int_from_real, fold_div_compare,
1761 fold_sign_changed_comparison, fold_unary, fold_negate_const,
1762 fold_abs_const, fold_not_const): Remove the final argument from
1763 calls to force_fit_type_double.
1765 2007-01-12 Andrew Pinski <andrew_pinski@playstation.sony.com>
1767 * configure.ac: Set insn to "nop" for spu-*-* also.
1768 * configure: Regenerate.
1770 2007-01-12 Olga Golovanevsky <olga@il.ibm.com>
1772 * builtins.def : Add BUILT_IN_FREE.
1774 2007-01-12 Jan Hubicka <jh@suse.cz>
1776 PR tree-optimization/30443
1777 * tree-inline.c (tree_function_versioning): Do not optimize when
1778 cloning for inlining.
1780 2007-01-12 Zdenek Dvorak <dvorakz@suse.cz>
1782 * doc/loop.texi: Document recording of loop exits.
1783 * cfgloopmanip.c (loopify, duplicate_loop): Use alloc_loop.
1784 (update_single_exits_after_duplication,
1785 update_single_exit_for_duplicated_loop,
1786 update_single_exit_for_duplicated_loops): Removed.
1787 (duplicate_loop_to_header_edge): Do not call
1788 update_single_exits_after_duplication and
1789 update_single_exit_for_duplicated_loops.
1790 (loop_version): Do not update single_exit information.
1791 (fix_loop_structure): Use record_loop_exits instead of
1792 mark_single_exit_loops.
1793 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update
1794 the lists of loop exits.
1795 * cfghooks.c (redirect_edge_and_branch, redirect_edge_and_branch_force,
1796 split_edge, merge_blocks): Update the lists of loop exits.
1797 * modulo-sched.c (sms_schedule): Pass LOOPS_HAVE_RECORDED_EXITS to
1798 loop_optimizer_init.
1799 * loop-init.c (loop_optimizer_init): Call record_loop_exits instead
1800 of mark_single_exit_loops.
1801 (loop_optimizer_finalize): Call release_recorded_exits.
1802 * tree-ssa-loop.c (tree_loop_optimizer_init): Pass
1803 LOOPS_HAVE_RECORDED_EXITS to loop_optimizer_init.
1804 * tree-vectorizer.c (slpeel_tree_duplicate_loop_to_edge_cfg): Do not
1805 update single exit information.
1806 * lambda-code.c (perfect_nestify): Ditto.
1807 * cfgloop.c (flow_loop_free): Destroy the list of exits of the loop.
1808 (mark_single_exit_loops): Removed.
1809 (alloc_loop, loop_exit_hash, loop_exit_eq, loop_exit_free,
1810 get_exit_descriptions, rescan_loop_exit, record_loop_exits,
1811 dump_recorded_exit, dump_recorded_exits, release_recorded_exits): New
1813 (get_loop_exit_edges, single_exit): Use recorded exit lists.
1814 (add_bb_to_loop, remove_bb_from_loops): Update the lists of loop exits.
1815 (verify_loop_structure): Verify consistency of the exit lists.
1816 (flow_loops_find): Use alloc_loop. Initialize exits hash.
1817 (set_single_exit): Removed.
1818 * cfgloop.h (struct loop_exit): New function.
1819 (struct loop): single_exit_ field replaced by exits field.
1820 (LOOPS_HAVE_MARKED_SINGLE_EXITS): Replaced by LOOPS_HAVE_RECORDED_EXITS.
1821 (struct loops): Added exits hash.
1822 (mark_single_exit_loops, set_single_exit): Declaration removed.
1823 (release_recorded_exits, record_loop_exits, rescan_loop_exit): Declare.
1825 2007-01-12 Richard Sandiford <richard@codesourcery.com>
1827 * doc/invoke.texi: Avoid use of @headitem.
1829 2007-01-12 Richard Sandiford <richard@codesourcery.com>
1831 * cse.c (cse_insn): Move HAVE_CC0 code after declarations.
1833 2007-01-12 Richard Sandiford <richard@codesourcery.com>
1835 * doc/install.texi: Fix m68k-*-* anchor and add m68k-*-* to the
1838 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
1839 Richard Sandiford <richard@codesourcery.com>
1841 * doc/invoke.texi: Document -mno-bitfield, -mno-rtd and -mno-short.
1842 * config/m68k/m68k.opt: Resort options.
1843 (mbitfield, mrtd, mshort): Remove RejectNegative properties.
1845 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
1846 Richard Sandiford <richard@codesourcery.com>
1848 * doc/invoke.texi: Document the macros that are defined by
1849 m68k's -mtune and -mhard-float options.
1850 * config/m68k/m68k-protos.h (m68k_cpp_cpu_ident) Declare.
1851 (m68k_cpp_cpu_family): Likewise.
1852 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Add a full set
1853 of __ucfv*__ macros. Define __mcffpu__ if generating code for
1854 ColdFire FPUs. Define __mcf_cpu_* and __mcf_family_* macros.
1855 * config/m68k/m68k.c (m68k_cpp_cpu_ident): New function.
1856 (m68k_cpp_cpu_family): Likewise.
1858 2007-01-12 Richard Sandiford <richard@codesourcery.com>
1860 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Treat all mc68*
1861 macros besides mc68000 as tuning macros. Use a switch statement
1862 to set them and mcpu32.
1864 2007-01-12 Julian Brown <julian@codesourcery.com>
1866 * config/m68k/m68k.h: Use TARGET_68040 instead of TARGET_68040_ONLY.
1867 (TARGET_68040_ONLY): Rename to...
1868 (TARGET_68040): ...this.
1869 * config/m68k/m68k.c: Use TARGET_68040 instead of TARGET_68040_ONLY.
1870 * config/m68k/m68k.md: Likewise.
1872 2007-01-12 Julian Brown <julian@codesourcery.com>
1873 Nathan Sidwell <nathan@codesourcery.com>
1874 Richard Sandiford <richard@codesourcery.com>
1876 * config.gcc (m680[012]0-*-*, m68k*-*-*): Set m68k_cpu_ident to
1877 the -mcpu= argument associated with the --with-cpu setting.
1878 Define M68K_DEFAULT_TUNE to the default -mtune= option,
1879 if different from the one implied by the -mcpu setting.
1880 Accept --with-cpu=FOO if FOO is listed in m68k-devices.def,
1881 using mcpu=FOO as the default CPU option. Set target_cpu_default2.
1882 * doc/invoke.texi: Mention ColdFire in the introduction to the
1883 m68k options. Document the new -march, -mcpu, -mtune, -mdiv,
1884 -mno-div and -mhard-float options. Make -m68881 a synonym for
1885 -mhard-float. Document the previously-undocumented -m5206e,
1886 -m528x, -m5307 and -m5407 options. Tweak the existing option
1887 documentation for consistency.
1888 * doc/install.texi: Mention new --with-cpu arguments.
1889 * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Only use the
1890 default CPU if neither -mcpu nor -march are specified.
1891 (ASM_CPU_SPEC): Pass down -mcpu and -march options.
1892 (TARGET_CPU_CPP_BUILTINS): Set __mcfisa*__ macros from
1893 TARGET_ISA*. Set the legacy __mcf*__ cpu macros in the same way,
1894 using m68k_tune to decide between families that implement the
1895 same ISA. Use m68k_tune to set __mcfv4e__.
1896 (FL_BITFIELD, FL_68881, FL_COLDFIRE, FL_CF_HWDIV, FL_CF_MAC)
1897 (FL_CF_EMAC, FL_CF_EMAC_B, FL_CF_USP, FL_CF_FPU, FL_ISA_68000)
1898 (FL_ISA_68010, FL_ISA_68020, FL_ISA_68040, FL_ISA_A, FL_ISA_B)
1899 (FL_ISA_C, FL_ISA_MMU): New macros.
1900 (MASK_COLDFIRE): Delete.
1901 (TARGET_68010, TARGET_68020, TARGET_68040_ONLY, TARGET_COLDFIRE)
1902 (TARGET_ISAB): Redefine in terms of m68k_cpu_flags.
1903 (TARGET_68881, TARGET_COLDFIRE_FPU): Redefine in terms of m68k_fpu.
1904 (TARGET_HARD_FLOAT): Do not define here.
1905 (TARGET_ISAAPLUS, TARGET_ISAC): New macros.
1906 (TUNE_68000): New macro.
1907 (TUNE_68000_10): Redefine in terms of TUNE_68000 and TUNE_68010.
1908 (TUNE_68010, TUNE_68030, TUNE_68040, TUNE_68060, TUNE_CPU32)
1909 (TUNE_CFV2): Redefine in terms of m68k_tune.
1910 (uarch_type, target_device, fpu_type): New enums.
1911 (m68k_cpu, m68k_tune, m68k_fpu, m68k_cpu_flags): Declare.
1912 * config/m68k/m68k.c (TARGET_DEFAULT): Remove MASK_68881.
1913 (FL_FOR_isa_00, FL_FOR_isa_10, FL_FOR_isa_20, FL_FOR_isa_40)
1914 (FL_FOR_isa_cpu32, FL_FOR_isa_a, FL_FOR_isa_aplus, FL_FOR_isa_b)
1915 (FL_FOR_isa_c): New macros.
1916 (m68k_isa): New enum.
1917 (m68k_target_selection): New structure.
1918 (all_devices, all_isas, all_microarchs): New tables.
1919 (m68k_cpu_entry, m68k_arch_entry, m68k_tune_entry, m68k_cpu)
1920 (m68k_tune, m68k_fpu, m68k_cpu_flags): New variables.
1921 (MASK_ALL_CPU_BITS): Delete.
1922 (m68k_find_selection): New function.
1923 (m68k_handle_option): Handle -mcpu=, -march= and -mtune=.
1924 Map the legacy target options to a combination of the new ones.
1925 (override_options): Set m68k_cpu, m68k_tune, m68k_fpu and
1926 m68k_cpu_flags. Handle M68K_DEFAULT_TUNE. Use m68k_cpu_flags
1927 to derive default MASK_BITFIELD, MASK_CF_HWDIV and MASK_HARD_FLOAT
1929 * config/m68k/m68k.opt (m5200, m5206e, m528x, m5307, m5407, mcfv4e)
1930 (m68010, m68020, m68020-40, m68020-60, m68030, m68040): Remove Mask
1932 (m68881, msoft-float): Change mask from 68881 to HARD_FLOAT.
1933 (march=, mcpu=, mdiv, mhard-float, mtune=): New options.
1934 * config/m68k/m68k-devices.def: New file.
1936 2007-01-12 Richard Sandiford <richard@codesourcery.com>
1937 Nathan Sidwell <nathan@codesourcery.com>
1939 * config/m68k/m68k.h (ASM_CPU_SPEC, ASM_SPEC, EXTRA_SPECS)
1940 (SUBTARGET_EXTRA_SPECS): New macros.
1941 * config/m68k/linux.h (ASM_SPEC): Remove CPU flags;
1942 use %(asm_cpu_spec) instead.
1943 * config/m68k/m68k-none.h (ASM_SPEC): Likewise.
1944 * config/m68k/openbsd.h (ASM_SPEC): Likewise.
1945 * config/m68k/netbsd-elf.h (ASM_SPEC): Likewise.
1946 (EXTRA_SPECS): Rename to...
1947 (SUBTARGET_EXTRA_SPECS): ...this.
1949 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
1950 Richard Sandiford <richard@codesourcery.com>
1951 Julian Brown <julian@codesourcery.com>
1953 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
1954 (m68k-*-uclinux*, m68k-*-linux*, m68k-*-rtems*): Set default_m68k_cpu
1955 to the configuration's default CPU.
1956 (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*): Likewise.
1957 Remove default masks.
1958 (m680[012]0-*-*): Set the default with_cpu to the first part of
1960 (m68k*-*-*): Set the default with_cpu to m$default_m68k_cpu.
1961 (m68k*-*-linux): Extend the --with-cpu handling to...
1962 (m680[012]0-*-*, m68k*-*-*): ...these configurations. Allow m68000
1963 and m68010. Don't set target_cpu_default2.
1964 * doc/install.texi: Document --with-cpu for m68k.
1965 * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Define.
1966 * config/m68k/m68k-none.h (TARGET_CPU_DEFAULT, M68K_CPU_m68k)
1967 (M68K_CPU_m68000, M68K_CPU_m68010, M68K_CPU_m68020, M68K_CPU_m68030)
1968 (M68K_CPU_m68040, M68K_CPU_m68302, M68K_CPU_m68332, TARGET_DEFAULT)
1969 (ASM_CPU_DEFAULT_SPEC, CC1_CPU_DEFAULT_SPEC): Delete.
1970 (ASM_SPEC): Remove use of %(asm_cpu_default).
1971 (EXTRA_SPECS, SUBTARGET_EXTRA_SPECS, MULTILIB_DEFAULTS): Delete.
1972 * config/m68k/linux.h (TARGET_DEFAULT): Delete.
1973 (CPP_SPEC): Merge definitions. Do not handle __HAVE_68881__ here.
1974 * config/m68k/netbsd-elf.h (TARGET_OS_CPP_BUILTINS): Define
1975 __HAVE_FPU__ if TARGET_HARD_FLOAT.
1976 (TARGET_DEFAULT): Delete.
1977 (EXTRA_SPECS): Delete cpp_cpu_default_spec, cpp_cpu_spec,
1978 cpp_fpu_spec, asm_default_spec and netbsd_cpp_spec.
1979 (CPP_CPU_SPEC): Delete.
1980 (TARGET_VERSION): Merge definitions, using TARGET_68010 to pick
1981 the appropriate string.
1982 (CPP_CPU_DEFAULT_SPEC, ASM_DEFAULT_SPEC, CPP_FPU_SPEC): Delete.
1983 (CPP_SPEC): Define to NETBSD_CPP_SPEC.
1984 (ASM_SPEC): Don't use %(asm_default_spec).
1985 * config/m68k/m68k.c (TARGET_DEFAULT_TARGET_FLAGS): Remove
1986 TARGET_DEFAULT and add MASK_68881.
1987 * config/m68k/m68k.md: Remove mention of TARGET_DEFAULT from comments.
1989 2007-01-12 Richard Sandiford <richard@codesourcery.com>
1991 * config.gcc (m68010-*-netbsdelf*): Add MASK_68010.
1992 (m68k*-*-netbsdelf*, m68k*-*-openbsd*, m68k*-linux*): Add
1993 MASK_68010 alongside MASK_68020.
1994 * doc/invoke.texi: Document -m68010.
1995 * config/m68k/m68k.opt (m68010): New.
1996 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Define mc68010
1998 (TUNE_68010): New macro.
1999 * config/m68k/m68k-none.h (M68K_CPU_m68k, M68K_CPU_m68010)
2000 (M68K_CPU_m68020, M68K_CPU_m68030, M68K_CPU_m68040)
2001 (M68K_CPU_m68332): Add MASK_68010.
2002 * config/m68k/linux.h (TARGET_DEFAULT): Add MASK_68010 to
2003 fallback definition.
2004 * config/m68k/netbsd-elf.h (CPP_CPU_SPEC): Remove now-redundant
2006 * config/m68k/m68k.c (MASK_ALL_CPU_BITS): Add MASK_68010.
2007 (m68k_handle_option): Handle OPT_m68010. Add MASK_68010
2008 to all entries that use MASK_68020.
2009 (output_move_simode_const, output_move_himode, output_move_qimode)
2010 (output_move_stricthi, output_move_strictqi): Use TARGET_68010
2011 instead of TARGET_68020 to select clr behavior. Remove comment
2012 about there being no TARGET_68010.
2013 * config/m68k/m68k.md: Likewise throughout.
2015 2007-01-12 Julian Brown <julian@codesourcery.com>
2017 * config/m68k/m68k.h (TARGET_ISAB): New macro.
2018 * config/m68k/m68k.c: Use TARGET_ISAB rather than TARGET_CFV4.
2019 * config/m68k/m68k.md: Likewise.
2021 2007-01-12 Julian Brown <julian@codesourcery.com>
2023 * config/m68k/m68k.h (LEGITIMATE_INDEX_P, LEGITIMIZE_ADDRESS): Use
2024 TARGET_COLDFIRE_FPU instead of TARGET_CFV4E.
2026 2007-01-12 Julian Brown <julian@codesourcery.com>
2028 * config/m68k/m68k.h (TUNE_68040_60): New macro.
2029 * config/m68k/m68k.c (standard_68881_constant_p): Use it.
2030 * config/m68k/m68k.md: Likewise.
2032 2007-01-12 Julian Brown <julian@codesourcery.com>
2033 Richard Sandiford <richard@codesourcery.com>
2035 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Use TUNE_68030
2036 instead of TARGET_68030, TUNE_68040 instead of TARGET_68040,
2037 TUNE_68060 instead of TARGET_68060 and TUNE_CPU32 instead of
2039 (TARGET_CPU32): Rename to...
2040 (TUNE_CPU32): ...this.
2041 (TUNE_68000_10, TUNE_68030, TUNE_68040, TUNE_68060)
2042 (TUNE_CFV2): New macros.
2043 * config/m68k/netbsd-elf.h (LONG_DOUBLE_TYPE_SIZE): Simplify;
2044 remove conditions that are implied by TARGET_68020.
2045 * config/m68k/m68k.c (m68k_output_function_prologue): Use TUNE_68040
2046 instead of TARGET_68040 and TUNE_CPU32 instead of TARGET_CPU32.
2047 (m68k_output_function_epilogue): Likewise.
2048 (m68k_rtx_costs): Likewise. Use TUNE_68060 instead of TARGET_68060
2049 and TUNE_CFV2 instead of TARGET_5200. Use TUNE_68000_10 instead of
2050 "!TARGET_68020 && !TARGET_COLDFIRE" to choose between 68000 and
2051 non-68000 timings. Refactor multiplication and division costs.
2052 (output_addsi3): Use TUNE_68040 instead of TARGET_68040 and
2053 TUNE_CPU32 instead of TARGET_CPU32.
2054 (standard_68881_constant_p): Use TUNE_68040 instead of TARGET_68040
2055 and TUNE_68060 instead of TARGET_68060.
2056 * config/m68k/m68k.md: Use TUNE_68040 instead of TARGET_68040,
2057 TUNE_68060 instead of TARGET_68060, and TUNE_CPU32 instead of
2059 (movsi_const0): Use TUNE_68000_10 rather than "!TARGET_68020
2060 && !TARGET_COLDFIRE" to choose between moveq and clr.
2061 Likewise in the unnamed movsf pattern.
2062 (ashlsi_17_24, lshrsi_17_24): Guard with TUNE_68000_10 rather than
2063 "!TARGET_68020 && !TARGET_COLDFIRE". Likewise the unnamed
2066 2007-01-12 Richard Sandiford <richard@codesourcery.com>
2068 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Increase amount
2069 of tabbing before backslashes.
2071 2007-01-11 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
2073 * pa-linux.h (ASM_OUTPUT_INTERNAL_LABEL): Undefine.
2074 * pa.h (ASM_OUTPUT_LABEL): Output colon when using GAS.
2075 (ASM_OUTPUT_INTERNAL_LABEL): Define.
2077 2007-01-11 Zdenek Dvorak <dvorakz@suse.cz>
2079 * tree-ssa-loop-ivopts.c (extract_cond_operands): Split from
2080 find_interesting_uses_cond.
2081 (find_interesting_uses_cond): Use extract_cond_operands.
2082 (rewrite_use_compare): Use extract_cond_operands and
2083 force_gimple_operand_bsi. Do not call update_stmt.
2084 (determine_use_iv_cost_condition): Use extract_cond_operands.
2085 Return cheaper of using original bound and changing the exit bound.
2087 2007-01-11 Zdenek Dvorak <dvorakz@suse.cz>
2089 PR tree-optimization/29516
2090 * tree-ssa-address.c (tree_mem_ref_addr, add_to_parts,
2091 most_expensive_mult_to_index, addr_to_parts,
2092 create_mem_ref, maybe_fold_tmr): Make the type of
2093 fields of TARGET_MEM_REF sizetype.
2094 (move_fixed_address_to_symbol, move_pointer_to_base):
2096 * tree.def (TARGET_MEM_REF): Add comment on types of
2099 2007-01-11 Joseph Myers <joseph@codesourcery.com>
2101 * c-common.c (vector_types_convertible_p): Treat opaque types as
2102 always convertible if they have the same size, but not otherwise.
2104 2007-01-11 Steven Bosscher <steven@gcc.gnu.org>
2106 * ifcvt.c (struct noce_if_info): Add comments to the fields.
2107 Remove the b_unconditional field.
2108 (noce_try_sign_mask): Do not look at b_unconditional.
2109 (noce_process_if_block): Do not use merge_if_blocks. Update
2110 the CFG here. Do not set b_unconditional.
2111 (cond_move_process_if_block): Likewise.
2112 (find_cond_trap): Likewise.
2113 (check_cond_move_block): Require simple jump insns at the end
2116 2007-01-11 Jan Hubicka <jh@suse.cz>
2118 PR tree-optimization/1046
2119 * tree-tailcall.c (suitable_for_tail_call_opt_p): Use TREE_ADDRESSABLE
2120 when alias info is not ready.
2121 (pass_tail_recursion): Do not require aliasing.
2122 * tree-ssa-copyrename.c (pass_rename_ssa_cop): Likewise.
2123 * tree-ssa-ccp.c (pass_ccp, pass_fold_builtins): Likewise.
2124 * tree-ssa-copy.c (pass_copy_prop): Likewise.
2125 * tree-ssa-forwprop.c (pass_forwprop): Likewise.
2126 * tree-ssa-dce.c (pass_dce, pass_dce_loop, pass_cd_dce): Likewise.
2127 * passes.c (init_optimization_passes): Execute rename_ssa_copies,
2128 ccp, forwprop, copy_prop, merge_phi, copy_prop, dce and tail recursion
2130 * tree-ssa-operands.c (add_virtual_operand, get_indirect_ref_operand):
2131 When aliasing is not build, mark statement as volatile.
2133 2007-01-11 Tom Tromey <tromey@redhat.com>
2135 PR preprocessor/15185, PR preprocessor/20989:
2136 * doc/cppopts.texi <-MT>: Update description of algorithm for
2137 computing default target.
2138 <-M, -MD>: Reword "basename" text.
2140 2007-01-11 Roger Sayle <roger@eyesopen.com>
2142 * builtins.c (expand_builtin_pow, expand_builtin_powi,
2143 fold_builtin_cabs, fold_builtin_sqrt, fold_builtin_trunc,
2144 fold_builtin_floor, fold_builtin_ceil, fold_builtin_round,
2145 fold_builtin_int_int_roundingfn, fold_builtin_bitop,
2146 fold_builtin_bswap, real_constp, fold_builtin_pow,
2147 fold_builtin_powi, fold_builtin_signbit, fold_builtin_copysign,
2148 do_mpfr_arg1, do_mpfr_arg2, do_mpfr_arg3, do_mpfr_sincos): Replace
2149 uses of the macro TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
2150 * convert.c (convert_to_pointer): Likewise.
2151 * expr.c (highest_pow2_factor, expand_expr_real_1): Likewise.
2152 * fold-const.c (force_fit_type, fold_negate_expr, int_const_binop,
2153 const_binop, fold_convert_const_int_from_int,
2154 fold_convert_const_int_from_real,
2155 fold_convert_const_real_from_real, sign_bit_p,
2156 optimize_minmax_comparison, extract_muldiv_1, fold_div_compare,
2157 fold_sign_changed_comparison, fold_unary, fold_comparison,
2158 fold_binary, multiple_of_p, tree_Expr_non_zero_p,
2159 fold_negate_const, fold_abs_const, fold_not_const): Likewise.
2160 * print-tree.c (print_node_brief, print_node): Likewise.
2161 * stor-layout.c (place_field, layout_type): Likewise.
2162 * tree-chrec.c (keep_cast): Likewise.
2163 * tree.c (build_vector, build_real, build_real_from_int_cst,
2164 build_complex): Likewise.
2166 2007-01-11 Roger Sayle <roger@eyesopen.com>
2168 * tree.h (TREE_CONSTANT_OVERFLOW): Obsolete. For the time being,
2169 treat TREE_CONSTANT_OVERFLOW as a synonym of TREE_OVERFLOW.
2171 2007-01-11 Paolo Bonzini <bonzini@gnu.org>
2173 * configure.ac (strict1_warn): Rename to strict_warn.
2174 (WERROR, --enable-werror, symlink hacks, stage1_cflags,
2175 cc_set_by_configure, quoted_cc_set_by_configure,
2176 stage_prefix_set_by_configure, quoted_stage_prefix_set_by_configure,
2177 all_boot_languages, all_stagestuff): Remove.
2178 (target_list): Remove bootstrap targets.
2179 * Makefile.in (quickstrap): Unconditionally make a synonym of all.
2180 (BOOT_LANGUAGES, STAGE1_CFLAGS, STAGE1_CHECKING,
2181 REMAKEFLAGS, FLAGS_TO_PASS, PREPEND_DOTDOT_TO_RELATIVE_PATHS,
2182 SUBDIR_FLAGS_TO_PASS, WERROR_FLAGS, STRICT2_WARN, LANG_STAGESTUFF,
2183 VOL_FILES, POSTSTAGE1_FLAGS_TO_PASS, STAGE2_FLAGS_TO_PASS,
2184 STAGEPROFILE_FLAGS_TO_PASS, STAGEFEEDBACK_FLAGS_TO_PASS, stage1_build,
2185 stage1_copy, stage2_build, stage2_copy, stageprofile_build,
2186 stageprofile_copy, stage3_build, stage3_copy, stagefeedback_build,
2187 stagefeedback_copy, stage4_build, clean_s1, clean_sw, bootstrap,
2188 bootstrap-lean, bootstrap2, bootstrap2-lean, bootstrap3,
2189 bootstrap3-lean, bootstrap4, bootstrap4-lean, unstage1, unstage2,
2190 unstage3, unstage4, unstageprofile, unstagefeedback, restage, restage2,
2191 restage3, restage4, restageprofile, restagefeedback, bubbleestrap,
2192 cleanstrap, unstrap, restrap, *compare, *compare3, *compare4,
2193 *compare-lean, *compare3-lean, *compare4-lean, stage1-start, stage1,
2194 stage2-start, stage2, stage3-start, stage3, stage4-start, stage4,
2195 stageprofile-start, stageprofile, stagefeedback-start, stagefeedback,
2196 risky-stage1, risky-stage2, risky-stage3, risky-stage4): Remove.
2197 (ORDINARY_FLAGS_TO_PASS): Rename to FLAGS_TO_PASS.
2198 (STAGECOPYSTUFF, STAGEMOVESTUFF): Consolidate into MOSTLYCLEANFILES.
2199 (mostlyclean): Adjust.
2200 (clean, distclean): Don't mention bootstrap stuff.
2201 * configure: Regenerate.
2202 * ada/config-lang.in, cp/config-lang.in, forttran/config-lang.in,
2203 java/config-lang.in, objc/config-lang.in, objcp/config-lang.in,
2204 treelang/config-lang.in (stagestuff): Remove.
2205 * doc/sourcebuild.texi (stage1, stage2, stage3, stage4,
2206 stageprofile, stagefeedback, stagestuff): Remove mention.
2208 2007-01-11 Nick Clifton <nickc@redhat.com>
2210 * config/mcore/predicates.md (mcore_general_movesrc_operand):
2212 (mcore_general_movdst_operand): Do not accept CONST_INTs.
2213 (mcore_arith_K_S_operand): Run the test for the S constraint not
2214 the test for the M constraint.
2215 (mcore_addsub_operand): Do not accept integer values that are
2216 larger than 32 bits.
2217 * config/mcore/mcore.md: Remove unused constraints from split.
2218 (andsi3): Use HOST_WIDE_INT instead of int to hold an INTVAL.
2220 (allocate_stack): Likewise.
2221 * config/mcore/mcore.c (mcore_print_operand): Restrict output of P
2222 operands to 32 bits.
2223 (mcore_const_costs): Use HOST_WIDE_INT instead of int to hold an
2225 (mcore_and_cost, mcore_modify_comparison, const_ok_for_mcore,
2226 mcore_const_ok_for_inline, mcore_const_trick_uses_not,
2227 try_constant_tricks, mcore_num_ones, mcore_num_zeros,
2228 mcore_output_bclri, mcore_output_andn, output_inline_const,
2229 mcore_output_move, mcore_output_movedouble): Likewise.
2230 (mcore_output_cmov): Use CONST_OK_FOR_M and CONST_OK_FOR_N.
2231 (output_inline_const): Likewise.
2232 (output_inline_const): Fix format strings used in sprintf
2234 * config/mcore/mcore-protos.h: Update prototypes for changed
2235 functions in mcore.c.
2236 * config/mcore/mcore.h (CONST_OK_FOR_I): Cast values to
2237 HOST_WIDE_INT and not int.
2238 (CONST_OK_FOR_J, CONST_OK_FOR_K, CONST_OK_FOR_L, CONST_OK_FOR_M,
2239 CONST_OK_FOR_N): Likewise.
2240 (LEGITIMATE_CONSTANT_P): Also check CONSTANT_P.
2241 (GO_IF_LEGITIMATE_INDEX): Use HOST_WIDE_INT instead of int to hold
2244 2007-01-10 Jan Hubicka <jh@suse.cz>
2246 * tree-vrp.c (remove_range_assertions): Release defs.
2247 * tree-ssa-loop-ivopts.c (rmeove_statement): Likewise.
2248 * tree-ssa-dom.c (remove_stmt_or_phi): Likewise.
2250 2007-01-10 Paul Brook <paul@codesourcery.com>
2252 * config/arm/arm.c (arm_rtx_costs_1): Handle mutiply-subtract.
2253 * config/arm/arm.md (mulsi3subsi): New insn.
2255 2007-01-10 Zdenek Dvorak <dvorakz@suse.cz>
2257 * tree-ssa-loop-manip.c (tree_unroll_loop): Make it a wrapper over ...
2258 (tree_transform_and_unroll_loop): New.
2259 * tree-flow.h (transform_callback, tree_transform_and_unroll_loop):
2262 2007-01-10 Robert Kennedy <jimbob@google.com>
2264 * fold-const.c (fold_comparison): Fold comparisons like (x *
2265 1000 < 0) to (x < 0).
2267 2007-01-10 Ian Lance Taylor <iant@google.com>
2269 * tree-pretty-print.c (dump_generic_node): Print parentheses when
2270 operands have the same priority.
2272 2007-01-10 Tom Tromey <tromey@redhat.com>
2274 * fold-const.c (fold_truthop): Don't check can_use_bit_fields_p.
2275 (fold_binary): Likewise.
2276 * langhooks.c (lhd_can_use_bit_fields_p): Removed.
2277 * langhooks-def.h (lhd_can_use_bit_fields_p): Removed.
2278 (LANG_HOOKS_CAN_USE_BIT_FIELDS_P): Removed.
2279 (LANG_HOOKS_INITIALIZER): Remove LANG_HOOKS_CAN_USE_BIT_FIELDS_P.
2280 * langhooks.h (struct lang_hooks): Removed field
2281 'can_use_bit_fields_p'.
2283 2007-01-10 Ralf Corsépius <ralf.corsepius@rtems.org>
2285 * config/bfin/t-bfin, config/bfin/t-bfin-elf: Remove GCC_CFLAGS.
2287 2007-01-10 Razya Ladelsky <razya@il.ibm.com>
2289 * function.c (get_last_funcdef_no): New function.
2290 * function.h (get_last_funcdef_no): Declare.
2291 * tree-inline.c (initialize_cfun): Add initialization.
2292 (tree_function_versioning): Cleanup.
2294 2007-01-10 Jan Hubicka <jh@suse.cz>
2296 * tree-inline.c (setup_one_parameter): Do not propagate into abnormal
2299 2007-01-10 Sa Liu <saliu@de.ibm.com>
2300 Ben Elliston <bje@au.ibm.com>
2302 * spu.h (STACK_SAVE_AREA): Use VOIDmode for SAVE_FUNCTION, SImode
2303 for SAVE_NONLOCAL and Pmode for any other save level.
2304 * spu-protos.h (spu_restore_stack_block): Declare.
2305 * spu.md (save_stack_block): Remove.
2306 (restore_stack_block): Call spu_restore_stack_block.
2307 * spu.c (spu_restore_stack_block): New function.
2308 (spu_expand_epilogue): Remove old comment.
2310 2007-01-09 Zdenek Dvorak <dvorakz@suse.cz>
2312 PR tree-optimization/30322
2313 * tree-ssa-loop-ivopts.c (fold_affine_expr, iv_value): Removed.
2314 (cand_value_at): Return the value as aff_tree.
2315 (may_eliminate_iv): Convert the bound from aff_tree to tree.
2316 * tree-affine.c (aff_combination_add_cst, aff_combination_add_product,
2317 aff_combination_mult): New functions.
2318 (aff_combination_add): Use aff_combination_add_cst.
2319 (aff_combination_convert): Allow conversions to a wider type.
2320 (tree_to_aff_combination): Handle BIT_NOT_EXPR.
2321 * tree-affine.h (aff_combination_mult): Declare.
2323 2007-01-09 Carlos O'Donell <carlos@codesourcery.com>
2325 * doc/tm.texi: Update documentation to reflect reality of exec
2326 and start file search behaviours. Update copyright year.
2327 * doc/invoke.texi: Explain how GCC_EXEC_PREFIX is used to find
2328 header file directories.
2330 2007-01-09 Uros Bizjak <ubizjak@gmail.com>
2332 * config/i386/i386.md (*sinxf2): Rename to *sinxf2_i387.
2333 (*cosxf2): Rename to cosxf2_i387.
2334 (*sindf2, *sinsf2): Extend operand 1 to XFmode. Macroize patterns
2335 using X87MODEF12 mode macro. Rename patterns to
2336 *sin_extend<mode>xf2_i387. Use SSE_FLOAT_MODE_P to disable patterns
2338 (*cosdf2, *cossf2): Ditto.
2339 (sincosdf3, sincossf3): Ditto. Rewrite corresponding splitters
2340 to match extended input operands.
2341 (sincos<mode>3): New expander.
2342 (*sinextendsfdf2, *cosextendsfdf2, *sincosextendsfdf3): Remove
2343 insn patterns and corresponding splitters.
2345 2007-01-09 Kaz Kojima <kkojima@gcc.gnu.org>
2347 * config/sh/t-linux (TARGET_LIBGCC2_CFLAGS): Delete.
2348 (SHLIB_MAPFILES, SHLIB_LINK, SHLIB_INSTALL): Likewise.
2350 2007-01-09 Nicolas Pitre <nico@cam.org>
2353 * arm/ieee754-df.S (Lad_s): Also test the low word of X for zero.
2355 2007-01-08 Geoffrey Keating <geoffk@apple.com>
2357 * target.h (struct gcc_target): New field library_rtti_comdat.
2358 * target-def.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): New.
2359 (TARGET_CXX): Add TARGET_CXX_LIBRARY_RTTI_COMDAT.
2360 * doc/tm.texi (C++ ABI): Document TARGET_CXX_LIBRARY_RTTI_COMDAT.
2361 * config/darwin.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
2363 2007-01-08 Geoffrey Keating <geoffk@apple.com>
2365 * doc/invoke.texi (Optimize Options): Correct description of -O0.
2367 2007-01-08 Richard Guenther <rguenther@suse.de>
2369 * tree.h (force_fit_type_double): Export.
2370 (force_fit_type): Remove.
2371 * fold-const.c (force_fit_type_double): New function.
2372 (force_fit_type): Remove.
2373 (int_const_binop): Use it.
2374 (fold_convert_const_int_from_int): Likewise.
2375 (fold_convert_const_int_from_real): Likewise.
2376 (fold_div_compare): Likewise.
2377 (fold_sign_changed_comparison): Likewise.
2378 (fold_unary): Likewise.
2379 (fold_negate_const): Likewise.
2380 (fold_abs_const): Likewise.
2381 (fold_not_const): Likewise.
2382 * c-common.c (shorten_compare): Use force_fit_type_double.
2383 * convert.c (convert_to_pointer): Likewise.
2385 2007-01-08 Richard Guenther <rguenther@suse.de>
2387 * tree.h (build_int_cst_wide_type): Export.
2388 * tree.c (build_int_cst_wide_type): New function.
2389 (build_int_cst_wide): Fix comment.
2390 * builtins.c (fold_builtin_object_size): Use build_int_cst
2391 to build -1 or 0 of the correct type. Use fit_double_type
2392 to check for overflow.
2393 * fold-const.c (optimize_bit_field_compare): Use build_int_cst_type
2395 (decode_field_reference): Likewise.
2396 (all_ones_mask_p): Likewise.
2397 (native_interpret_int): Use build_int_cst_wide_type.
2398 (fold_binary): Use build_int_cst_type to build an all-ones
2400 * stor-layout.c (set_sizetype): Use build_int_cst_wide_type.
2402 2007-01-08 Daniel Jacobowitz <dan@codesourcery.com>
2404 * config/pa/t-pa64 (libgcc_stub.a): Use $(T).
2406 2007-01-09 Ben Elliston <bje@au.ibm.com>
2408 * genautomata.c (STATS_OPTION): New option.
2409 (stats_flag): New flag.
2410 (gen_automata_option): Handle it.
2411 (initiate_automaton_gen): Ditto.
2412 (write_automata): Output statistics only if stats_flag is
2413 set. Likewise, output time statistics only if time_flag is set.
2414 * doc/md.texi (Processor pipeline description): Document new flag.
2416 2007-01-08 Richard Guenther <rguenther@suse.de>
2418 * builtins.c (fold_builtin_int_roundingfn): Use fit_double_type.
2419 * tree.c (build_int_cst_type): Likewise.
2420 (size_in_bytes): Don't call force_fit_type on the result.
2421 (int_fits_type_p): Use fit_double_type.
2422 * fold-const.c (fit_double_type): New function.
2423 (force_fit_type): Use it.
2424 * tree.h (fit_double_type): Export.
2426 2007-01-08 Jan Hubicka <jh@suse.cz>
2428 * tree-vectorizer.c (gate_increase_alignment): Fix return type.
2429 * ipa.c (function_and_variable_visibility): Fix return type.
2431 2007-01-08 Richard Guenther <rguenther@suse.de>
2433 * tree-ssa-ccp.c (maybe_fold_offset_to_array_ref): Use type
2434 of offset to build the index.
2435 * tree-pretty-print.c (dump_generic_node): Don't build negated
2436 const just for printing.
2437 * c-pretty-print.c (pp_c_integer_constant): Likewise.
2438 * builtins.c (fold_builtin_int_roundingfn): Check if result
2439 fits the type by using force_fit_type and comparing the result.
2440 * predict.c (predict_loops): Use compare_tree_int for comparison.
2441 * tree.c (build_int_cst): Fall back to integer_type_node for
2443 (build_int_cst_wide): Assert type is non-null.
2445 2007-01-08 Roberto Costa <roberto.costa@st.com>
2447 * tree-vrp.c (extract_range_from_cond_expr): New.
2448 (extract_range_from_expr): Handle COND_EXPR nodes used as expressions.
2449 * tree-ssa-ccp.c (get_maxval_strlen): Handle COND_EXPR nodes used
2451 (fold_stmt): Bug fix, avoid infinite recursion when folding COND_EXPRs.
2452 * tree-ssa-forwprop.c (simplify_cond, forward_propagate_into_cond,
2453 tree_ssa_forward_propagate_single_use_vars): Handle COND_EXPR nodes
2454 used as expressions.
2455 * tree-object-size.c (cond_expr_object_size): New.
2456 (collect_object_sizes_for): Handle COND_EXPR nodes used as expressions.
2458 2007-01-08 Jan Hubicka <jh@suse.cz>
2460 * tree-ssa-forwprop.c (forward_propagate_into_cond,
2461 tree_ssa_forward_propagate_single_use_va): Release defs of propagated
2464 2007-01-08 Richard Guenther <rguenther@suse.de>
2466 PR tree-optimization/23603
2467 * tree-vrp.c (set_value_range_to_truthvalue): New function.
2468 (extract_range_from_binary): Fall back to truthvalue instead of
2469 varying for TRUTH_*_EXPR.
2470 (extract_range_from_comparison): Fall back to truthvalue instead of
2472 (vrp_visit_phi_node): Don't adjust new range bounds to +INF/-INF
2473 if all visited PHI values were constant.
2475 2007-01-08 Jan Hubicka <jh@suse.cz>
2477 * cgraphunit.c (cgraph_process_new_functions): Reset reachable flag.
2478 (cgraph_analyze_function): break out from ...
2479 (cgraph_finalize_compilation_unit): ... here.
2480 (cgraph_expand_function): Remove forgoten commented out line.
2481 (cgraph_optimize): Analyze functions.
2483 2007-01-08 Jan Hubicka <jh@suse.cz>
2485 * tree-pas.h (TODO_remove_function): New flag.
2486 (TODO_update*): Renumber.
2487 (pass_ipa_increase_alignment,
2488 pass_ipa_function_and_variable_visibility): New passes.
2489 * cgraphunit.c (cgraph_increase_alignment): Move to tree-vectorizer.c
2490 (cgraph_function_and_variable_visibility): Move to ipa.c
2491 (cgraph_optimize): Don't call cgraph_function_and_variable_visibility,
2492 cgraph_increase_alignment.
2493 * ipa-inline.c (cgraph_decide_inlining): Don't push timevar.
2494 (cgraph_decide_inlining_incrementally): Push TV_INTEGRATION before
2495 calling tree-inline.
2496 (cgraph_early_inlining): Do not call cgraph_remove_unreachable_nodes.
2497 (pass_ipa_inline, pass_early_ipa_inlining): Set TODO_remove_functions
2498 * tree-vectorizer.c (increase_alignment): Move here from cgraphunit.c
2499 (gate_increase_alignment): New function.
2500 (pass_ipa_increase_alignment): New pass.
2501 * ipa.c: Inline tree-pass.h and timevar.h
2502 (function_and_variable_visibility): Move here from cgraphunit.c
2503 * tree-optimize.c (pass_early_local_passes): Add TODO_remove_functions.
2504 * passes.c (init_optimization_passes): Add the two new passes.
2505 (execute_todo): Handle cgraph_remove_functions.
2507 2007-01-08 Nick Clifton <nickc@redhat.com>
2509 * config/frv/predicates.md (reg_or_0_operand): Accept
2512 2007-01-08 Ralf Corsépius <ralf.corsepius@rtems.org>
2514 * config/bfin/rtems.h, config/bfin/t-rtems: New.
2515 * config.gcc: Add bfin*-rtems*.
2517 2007-01-08 Mark Shinwell <shinwell@codesourcery.com>
2519 * c.opt: Add -flax-vector-conversions.
2520 * c-typeck.c (convert_for_assignment): Pass flag to
2521 vector_types_convertible_p to allow emission of note.
2522 (digest_init): Likewise.
2523 * c-opts.c: Handle -flax-vector-conversions.
2524 * c-common.c (flag_lax_vector_conversions): New.
2525 (vector_types_convertible_p): Unless -flax-vector conversions
2526 has been passed, disallow conversions between vectors with
2527 differing numbers of subparts and/or element types. If such
2528 a conversion is disallowed, possibly emit a note on the first
2529 occasion only to inform the user of -flax-vector-conversions.
2530 The new last argument specifies this.
2531 * c-common.h (flag_lax_vector_conversions): New.
2532 (vector_types_convertible_p): Add extra argument.
2533 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Use
2534 char_type_node for V*QI type vectors.
2535 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins):
2536 Update to satisfy new typechecking rules.
2537 * config/rs6000/altivec.h (vec_cmple): Use vec_cmpge, for both
2539 * doc/invoke.texi (C Dialect Options): Document
2540 -flax-vector-conversions.
2542 2007-01-08 Mark Shinwell <shinwell@codesourcery.com>
2544 PR tree-optimization/29877
2545 * tree-ssa-ter.c (is_replaceable_p): Deem assignments with
2546 a register variable on the RHS to not be replaceable.
2548 2007-01-08 Chen Liqin <liqin@sunnorth.com.cn>
2549 * config/score/t-score-elf (MULTILIB_OPTIONS): Change.
2550 * config/score/predicates.md (const_uimm5, sr0_operand, const_simm12,
2551 const_simm15, const_pow2, const_npow2): Added.
2552 * config/score/misc.md (insv, extv, extzv, movmemsi,
2553 move_lbu_a/b, mov_lhu_a/b etc): Added and fix some bug.
2554 * config/score/score.c (score_address_cost, score_select_cc_mode):
2556 Change CONST_OK_FOR_LETTER_P/EXTRA_CONSTRAINT define.
2557 Update score_rtx_costs for MACRO TARGET_RTX_COSTS.
2558 Update score_print_operand.
2559 * config/score/score.h (DATA_ALIGNMENT, SELECT_CC_MODE): Added.
2560 Adjust register allocate order and update some macro define.
2561 * config/score/score-mdaux.c (mdx_unaligned_load, mdx_unsigned_store,
2562 mdx_block_move_straight, mdx_block_move_loop_head,
2563 mdx_block_move_loop_body, mdx_block_move_loop_foot, mdx_block_move_loop,
2564 mdx_block_move): Added.
2565 (mdx_movsicc, mdp_select_add_imm, mdp_select, mds_zero_extract_andi,
2566 mdp_limm): Updated and fix some bug and typo.
2567 * config/score/score.md (movqi/hi/si, add/sub/zero/ext): Updated.
2568 (movsf, movdf, doloop_end): Added.
2570 2007-01-08 Kazu Hirata <kazu@codesourcery.com>
2572 * config/arm/arm.c, config/arm/arm.h, config/arm/arm.md,
2573 config/arm/thumb2.md: Fix comment typos.
2574 * doc/extend.texi: Fix a typo.
2576 2007-01-07 Eric Christopher <echristo@apple.com>
2578 * configure.ac: Check for __stack_chk_fail for darwin.
2579 * configure: Regenerate.
2581 2007-01-07 Richard Guenther <rguenther@suse.de>
2583 * tree-vrp.c (extract_range_from_assert): CSE calls to
2584 compare_values where possible.
2585 (extract_range_from_unary_expr): Likewise.
2587 2007-01-07 Anatoly Sokolov <aesok@post.ru>
2589 * config/avr/avr-protos.h (call_insn_operand): Delete prototype.
2590 * config/avr/avr.c (call_insn_operand): Delete function.
2591 * config/avr/avr.md (*pushqi, *pushhi, *pushsi, *pushsf): Use REG_SP
2592 instead of register number. Use predicates.
2593 * config/avr/predicates.md (const0_operand, reg_or_0_operand,
2594 call_insn_operand): Add.
2596 2007-01-06 Jan Hubicka <jh@suse.cz>
2598 * tree-pass.h (pass_build_cgraph_edges): Declare.
2599 * cgraphunit.c (record_refernece): Move to cgraphbuild.c
2600 (visited_nodes): Remove.
2601 (cgraph_create_edges): Move to cgraphbuild.c; rename to
2602 build_cgrpah_edges; make visited_nodes local.
2603 (cgraph_process_new_functions): DO not call initialize_inline_failed.
2604 (record_references_in_initializer): Move to cgraphbuild.c
2605 (initialize_inline_failed, rebuild_cgraph_edges,
2606 pass_rebuild_cgraph_edges): Move to cgraphbuild.c.
2607 (verify_cgraph_node): Make visited_nodes local.
2608 (cgraph_analyze_function): Do not call cgraph_create_edges and
2609 initialize_inline_failed.
2610 (cgraph_expand_function): Do not call cgraph_lower_function;
2611 assert that function is already lowered.
2612 * Makefile.in (cgraphbuild.o): New.
2613 * passes.c (init_optimization_passes): Add pass_build_cgraph_edges
2614 at the end of lowering passes.
2616 2007-01-06 Steven Bosscher <steven@gcc.gnu.org>
2618 * ifcvt.c (cond_move_convert_if_block): New function, code
2619 factored out from...
2620 (cond_move_process_if_block): ...here. Call the new function
2621 on the THEN and ELSE blocks.
2622 (merge_if_block): Do not copy global_live_at_end, merge_blocks
2623 already takes care of this.
2625 2007-01-05 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
2628 * tree.h (TREE_OVERFLOW_P): New.
2629 * c-typeck.c (parser_build_unary_op): Warn only if result
2630 overflowed and operands did not.
2631 (parser_build_binary_op): Likewise.
2632 (convert_for_assignment): Remove redundant overflow_warning.
2633 * c-common.c (overflow_warning): Don't check or set TREE_OVERFLOW.
2635 2007-01-05 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
2637 * c-typeck.c (store_init_value): Split over two lines to follow
2638 the GNU coding style.
2640 2007-01-05 Benjamin Kosnik <bkoz@redhat.com>
2642 * c-cppbuiltin.c (c_cpp_builtins): __GXX_EXPERIMENTAL_CPP0X__ to
2643 __GXX_EXPERIMENTAL_CXX0X__.
2644 * doc/cpp.texi: Same.
2646 2007-01-05 Richard Guenther <rguenther@suse.de>
2649 * tree.c (get_narrower): Do not construct COMPONENT_REFs
2650 with mismatched types. Instead explicitly build a
2651 conversion NOP_EXPR.
2653 2007-01-05 Ian Lance Taylor <iant@google.com>
2655 * c-common.c (decl_with_nonnull_addr_p): New function.
2656 (c_common_truthvalue_conversion): Call it.
2657 * c-typeck.c (build_binary_op): Likewise.
2658 * c-common.h (decl_with_nonnull_addr_p): Declare.
2660 2007-01-05 Jakub Jelinek <jakub@redhat.com>
2663 * libgcc2.c (__divdc3): Compare c and d against 0.0 instead of
2666 2007-01-05 Joel Brobecker <brobecker@adacore.com>
2668 * doc/install.texi (Final install): Document the fact that
2669 the GNAT runtime should not be stripped.
2671 2007-01-04 Jan Hubicka <jh@suse.cz>
2673 * tree-inline.c (fold_marked_statements): Update operand caches
2674 and EH after folding
2676 2007-01-04 Ian Lance Taylor <iant@google.com>
2678 * c-common.c (check_function_nonnull): Whitespace fix.
2680 2007-01-04 Jan Hubicka <jh@suse.cz>
2682 * tree-optimize.c (execute_fixup_cfg): Correct previously mistakely
2683 comitted older version of patch.
2684 (pass_fixup_cfg): Add TODOs to verify flow and statements, dump
2685 function, celanup cfg and collect garbage.
2687 2007-01-04 Mike Stump <mrs@apple.com>
2689 * Makefile.in (mostlyclean): Don't remove libgcc anymore.
2692 2007-01-04 Eric Christopher <echristo@apple.com>
2694 * libgcc2.c (__bswapsi2): Use SItype.
2695 (__bswapdi2): Use DItype.
2696 * libgcc2.h: Update for above.
2698 2007-01-04 Paul Brook <paul@codesourcery.com>
2700 * config/arm/arm.md (arm_mulsi3, thumb_mulsi3, mulsi3_compare0,
2701 mulsi_compare0_scratch, mulsi3addsi, mulsi3addsi_compare0,
2702 mulsi3addsi_compare0_scratch, mulsidi3adddi, mulsidi3,
2703 umulsidi3, umulsidi3adddi, smulsi3_highpart,
2704 umulsi3_highpart): Make conditional on !arm_arch6.
2705 (arm_mulsi3_v6, thumb_mulsi3_v6, mulsi3_compare0_v6,
2706 mulsi_compare0_scratch_v6, mulsi3addsi_v6, mulsi3addsi_compare0_v6,
2707 mulsi3addsi_compare0_scratch_v6, mulsidi3adddi_v6, mulsidi3_v6,
2708 umulsidi3_v6, umulsidi3adddi_v6, smulsi3_highpart_v6,
2709 umulsi3_highpart_v6): New insns.
2711 2007-01-04 Roger Sayle <roger@eyesopen.com>
2713 * fold-const.c (fold_convert): When casting an expression to void,
2714 fold_ignored_result may discover a GIMPLE_MODIFY_STMT which doesn't
2715 have a type. Instead of attempting to build a NOP_EXPR, return
2716 these "special" trees directly.
2718 2007-01-04 Joseph Myers <joseph@codesourcery.com>
2720 * config/rs6000/rs6000.c (rs6000_rtx_costs): Make adjustment for
2721 MULT inside MINUS as either argument. Use rs6000_cost->dmul -
2722 rs6000_cost->fp not 0 as adjustment for outer NEG.
2724 2007-01-04 Jan Hubicka <jh@suse.cz>
2726 * cgraph.c (cgraph_release_function_body): New function.
2727 (cgraph_remove_node): Use it.
2728 * cgraph.h (cgraph_release_function_body): Declare.
2729 * cgraphunit.c (cgraph_expand_function): Use it.
2730 * ipa.c (cgraph_remove_unreahchable_nodes): Use it.
2731 * tree-ssa.c (delete_tree_ssa): Allow to be called before aliasing
2732 is initialized and while compilation of other function is running.
2733 * tree-optimize.c (execute_free_cfg_annotations): Move code to clear
2734 statement CFG annotations from here to ...
2735 * tree-cfg.c (delete_tree_cfg_annotations): ... here.
2737 2007-01-04 Zdenek Dvorak <dvorakz@suse.cz>
2739 * cfgloop.h (enum li_flags): Make the constants powers of two.
2741 2007-01-04 Jan Hubicka <jh@suse.cz>
2743 * tree-inline.c (copy_bb): Insert new statements to statements_to_fold
2745 (fold_marked_statements): New function.
2746 (optimize_inline_calls, tree_function_versioning): Fold new statements.
2747 * tree-inline.h (copy_body_data): Add statements_to_fold.
2749 2007-01-03 Daniel Jacobowitz <dan@codesourcery.com>
2751 * config.gcc: Mention libgcc/config.host.
2752 * Makefile.in: Update comments mentioning libgcc.
2753 (LIBGCC, INSTALL_LIBGCC, GCC_PARTS, mklibgcc): Delete.
2754 (all.cross, start.encap, rest.encap, rest.cross): Update
2755 dependencies for libgcc move.
2756 (libgcc.mk, LIBGCC_DEPS, libgcov.a, libgcc.a, stmp-multilib)
2757 (clean-target, clean-target-libgcc): Delete.
2758 (srcdirify, GCC_EXTRA_PARTS): New macros.
2759 (libgcc-support, libgcc.mvars): New rules.
2760 (distclean): Remove mention of mklibgcc.
2761 (install): Don't reference INSTALL_LIBGCC.
2762 (install-common): Don't reference EXTRA_PARTS.
2763 (install-libgcc, install-multilib): Delete rules.
2764 * mklibgcc.in: Delete file.
2765 * doc/configfiles.texi: Don't mention mklibgcc.
2767 * config/i386/t-darwin (SHLIB_VERPFX): Delete (moved to libgcc).
2768 * config/i386/t-darwin64 (SHLIB_VERPFX): Likewise.
2769 * config/rs6000/t-darwin (SHLIB_VERPFX): Likewise.
2770 * config/rs6000/t-ppccomm (TARGET_LIBGCC2_CFLAGS, SHLIB_MAPFILES)
2771 (mklibgcc, ldblspecs): Likewise.
2773 * config/i386/t-nwld (libgcc.def, libc.def, libpcre.def)
2774 (posixpre.def): Use $(T).
2775 (SHLIB_EXT, SHLIB_NAME, SHLIB_SLIBDIR_QUAL, SHLIB_DEF, SHLIB_MAP)
2776 (SHLIB_SRC, SHLIB_INSTALL): Delete.
2777 (SHLIB_LINK): Make dummy.
2778 * config/t-slibgcc-darwin: Delete contents except for dummy SHLIB_LINK.
2780 * config/frv/t-linux (EXTRA_MULTILIB_PARTS): Clear.
2782 * config/alpha/t-crtfm: Use $(T) in rules for EXTRA_PARTS.
2783 * config/alpha/t-vms, config/alpha/t-vms64, config/fr30/t-fr30,
2784 config/i386/t-rtems-i386, config/ia64/t-ia64, config/rs6000/t-beos,
2785 config/rs6000/t-newas, config/sparc/t-elf: Likewise.
2787 * configure.ac (all_outputs): Remove mklibgcc.
2788 * configure: Regenerated.
2790 2007-01-03 Josh Conner <jconner@apple.com>
2793 * calls.c (compute_argument_addresses): Set stack and stack_slot
2794 for partial args, too.
2795 (store_one_arg): Use locate.size.constant for the size when
2796 generating a save_area.
2798 2007-01-03 Robert Kennedy <jimbob@google.com>
2800 * tree-cfg.c (tree_merge_blocks): Release SSA_NAME phi results
2801 whose definitions are deleted due to basic block merging.
2803 2007-01-03 Paul Brook <paul@codesourcery.com>
2806 * config/arm/arm.c (output_return_instruction): Pop PC in interrupt
2808 (use_return_insn): Return 0 for Thumb interrupt functions.
2809 (print_multi_reg): Add rfe argument for IRQ returns.
2810 (arm_output_epilogue): Pop interrupt return address directly into PC.
2811 (arm_expand_prologue): Only adjust IRQ return address in Arm mode.
2813 2007-01-03 Paul Brook <paul@codesourcery.com>
2815 Merge from sourcerygxx-4_1.
2816 * config/arm/thumb2.md: New file.
2817 * config/arm/elf.h (JUMP_TABLES_IN_TEXT_SECTION): Return True for
2819 * config/arm/coff.h (JUMP_TABLES_IN_TEXT_SECTION): Ditto.
2820 * config/arm/aout.h (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
2821 (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump tables.
2822 * config/arm/aof.h (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump
2824 (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
2825 * config/arm/ieee754-df.S: Use macros for Thumb-2/Unified asm
2827 * config/arm/ieee754-sf.S: Ditto.
2828 * config/arm/arm.c (thumb_base_register_rtx_p): Rename...
2829 (thumb1_base_register_rtx_p): ... to this.
2830 (thumb_index_register_rtx_p): Rename...
2831 (thumb1_index_register_rtx_p): ... to this.
2832 (thumb_output_function_prologue): Rename...
2833 (thumb1_output_function_prologue): ... to this.
2834 (thumb_legitimate_address_p): Rename...
2835 (thumb1_legitimate_address_p): ... to this.
2836 (thumb_rtx_costs): Rename...
2837 (thumb1_rtx_costs): ... to this.
2838 (thumb_compute_save_reg_mask): Rename...
2839 (thumb1_compute_save_reg_mask): ... to this.
2840 (thumb_final_prescan_insn): Rename...
2841 (thumb1_final_prescan_insn): ... to this.
2842 (thumb_expand_epilogue): Rename...
2843 (thumb1_expand_epilogue): ... to this.
2844 (arm_unwind_emit_stm): Rename...
2845 (arm_unwind_emit_sequence): ... to this.
2846 (thumb2_legitimate_index_p, thumb2_legitimate_address_p,
2847 thumb1_compute_save_reg_mask, arm_dwarf_handle_frame_unspec,
2848 thumb2_index_mul_operand, output_move_vfp, arm_shift_nmem,
2849 arm_save_coproc_regs, thumb_set_frame_pointer, arm_print_condition,
2850 thumb2_final_prescan_insn, thumb2_asm_output_opcode, arm_output_shift,
2851 thumb2_output_casesi): New functions.
2852 (TARGET_DWARF_HANDLE_FRAME_UNSPEC): Define.
2853 (FL_THUMB2, FL_NOTM, FL_DIV, FL_FOR_ARCH6T2, FL_FOR_ARCH7,
2854 FL_FOR_ARCH7A, FL_FOR_ARCH7R, FL_FOR_ARCH7M, ARM_LSL_NAME,
2855 THUMB2_WORK_REGS): Define.
2856 (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv, arm_condexec_count,
2857 arm_condexec_mask, arm_condexec_masklen)): New variables.
2858 (all_architectures): Add armv6t2, armv7, armv7a, armv7r and armv7m.
2859 (arm_override_options): Check new CPU capabilities.
2860 Set new architecture flag variables.
2861 (arm_isr_value): Handle v7m interrupt functions.
2862 (user_return_insn): Return 0 for v7m interrupt functions. Handle
2864 (const_ok_for_arm): Handle Thumb-2 constants.
2865 (arm_gen_constant): Ditto. Use movw when available.
2866 (arm_function_ok_for_sibcall): Return false for v7m interrupt
2868 (legitimize_pic_address, arm_call_tls_get_addr): Handle Thumb-2.
2869 (thumb_find_work_register, arm_load_pic_register,
2870 legitimize_tls_address, arm_address_cost, load_multiple_sequence,
2871 emit_ldm_seq, emit_stm_seq, arm_select_cc_mode, get_jump_table_size,
2872 print_multi_reg, output_mov_long_double_fpa_from_arm,
2873 output_mov_long_double_arm_from_fpa, output_mov_double_fpa_from_arm,
2874 output_mov_double_fpa_from_arm, output_move_double,
2875 arm_compute_save_reg_mask, arm_compute_save_reg0_reg12_mask,
2876 output_return_instruction, arm_output_function_prologue,
2877 arm_output_epilogue, arm_get_frame_offsets, arm_regno_class,
2878 arm_output_mi_thunk, thumb_set_return_address): Ditto.
2879 (arm_expand_prologue): Handle Thumb-2. Use arm_save_coproc_regs.
2880 (arm_coproc_mem_operand): Allow POST_INC/PRE_DEC.
2881 (arithmetic_instr, shift_op): Use arm_shift_nmem.
2882 (arm_print_operand): Use arm_print_condition. Handle '(', ')', '.',
2884 (arm_final_prescan_insn): Use extract_constrain_insn_cached.
2885 (thumb_expand_prologue): Use thumb_set_frame_pointer.
2886 (arm_file_start): Output directive for unified syntax.
2887 (arm_unwind_emit_set): Handle stack alignment instruction.
2888 * config/arm/lib1funcs.asm: Remove default for __ARM_ARCH__.
2889 Add v6t2, v7, v7a, v7r and v7m.
2890 (RETLDM): Add Thumb-2 code.
2891 (do_it, shift1, do_push, do_pop, COND, THUMB_SYNTAX): New macros.
2892 * config/arm/arm.h (TARGET_CPU_CPP_BUILTINS): Define __thumb2__.
2893 (TARGET_THUMB1, TARGET_32BIT, TARGET_THUMB2, TARGET_DSP_MULTIPLY,
2894 TARGET_INT_SIMD, TARGET_UNIFIED_ASM, ARM_FT_STACKALIGN, IS_STACKALIGN,
2895 THUMB2_TRAMPOLINE_TEMPLATE, TRAMPOLINE_ADJUST_ADDRESS,
2896 ASM_OUTPUT_OPCODE, THUMB2_GO_IF_LEGITIMATE_ADDRESS,
2897 THUMB2_LEGITIMIZE_ADDRESS, CASE_VECTOR_PC_RELATIVE,
2898 CASE_VECTOR_SHORTEN_MODE, ADDR_VEC_ALIGN, ASM_OUTPUT_CASE_END,
2899 ADJUST_INSN_LENGTH): Define.
2900 (TARGET_REALLY_IWMMXT, TARGET_IWMMXT_ABI, CONDITIONAL_REGISTER_USAGE,
2901 STATIC_CHAIN_REGNUM, HARD_REGNO_NREGS, INDEX_REG_CLASS,
2902 BASE_REG_CLASS, MODE_BASE_REG_CLASS, SMALL_REGISTER_CLASSES,
2903 PREFERRED_RELOAD_CLASS, SECONDARY_OUTPUT_RELOAD_CLASS,
2904 SECONDARY_INPUT_RELOAD_CLASS, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P,
2905 TRAMPOLINE_SIZE, INITIALIZE_TRAMPOLINE, HAVE_PRE_INCREMENT,
2906 HAVE_POST_DECREMENT, HAVE_PRE_DECREMENT, HAVE_PRE_MODIFY_DISP,
2907 HAVE_POST_MODIFY_DISP, HAVE_PRE_MODIFY_REG, HAVE_POST_MODIFY_REG,
2908 REGNO_MODE_OK_FOR_BASE_P, LEGITIMATE_CONSTANT_P,
2909 REG_MODE_OK_FOR_BASE_P, REG_OK_FOR_INDEX_P, GO_IF_LEGITIMATE_ADDRESS,
2910 LEGITIMIZE_ADDRESS, THUMB2_LEGITIMIZE_ADDRESS,
2911 GO_IF_MODE_DEPENDENT_ADDRESS, MEMORY_MOVE_COST, BRANCH_COST,
2912 ASM_APP_OFF, ASM_OUTPUT_CASE_LABEL, ARM_DECLARE_FUNCTION_NAME,
2913 FINAL_PRESCAN_INSN, PRINT_OPERAND_PUNCT_VALID_P,
2914 PRINT_OPERAND_ADDRESS): Adjust for Thumb-2.
2915 (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv): New declarations.
2916 * config/arm/arm-cores.def: Add arm1156t2-s, cortex-a8, cortex-r4 and
2918 * config/arm/arm-tune.md: Regenerate.
2919 * config/arm/arm-protos.h: Update prototypes.
2920 * config/arm/vfp.md: Enable patterns for Thumb-2.
2921 (arm_movsi_vfp): Add movw alternative. Use output_move_vfp.
2922 (arm_movdi_vfp, movsf_vfp, movdf_vfp): Use output_move_vfp.
2923 (thumb2_movsi_vfp, thumb2_movdi_vfp, thumb2_movsf_vfp,
2924 thumb2_movdf_vfp, thumb2_movsfcc_vfp, thumb2_movdfcc_vfp): New.
2925 * config/arm/libunwind.S: Add Thumb-2 code.
2926 * config/arm/constraints.md: Update include Thumb-2.
2927 * config/arm/ieee754-sf.S: Add Thumb-2/Unified asm support.
2928 * config/arm/ieee754-df.S: Ditto.
2929 * config/arm/bpabi.S: Ditto.
2930 * config/arm/t-arm (MD_INCLUDES): Add thumb2.md.
2931 * config/arm/predicates.md (low_register_operand,
2932 low_reg_or_int_operand, thumb_16bit_operator): New.
2933 (thumb_cmp_operand, thumb_cmpneg_operand): Rename...
2934 (thumb1_cmp_operand, thumb1_cmpneg_operand): ... to this.
2935 * config/arm/t-arm-elf: Add armv7 multilib.
2936 * config/arm/arm.md: Update patterns for Thumb-2 and Unified asm.
2938 (UNSPEC_STACK_ALIGN, ce_count): New.
2939 (arm_incscc, arm_decscc, arm_umaxsi3, arm_uminsi3,
2940 arm_zero_extendsidi2, arm_zero_extendqidi2): New
2942 * config/arm/fpa.md: Update patterns for Thumb-2 and Unified asm.
2943 (thumb2_movsf_fpa, thumb2_movdf_fpa, thumb2_movxf_fpa,
2944 thumb2_movsfcc_fpa, thumb2_movdfcc_fpa): New insns.
2945 * config/arm/cirrus.md: Update patterns for Thumb-2 and Unified asm.
2946 (cirrus_thumb2_movdi, cirrus_thumb2_movsi_insn,
2947 thumb2_cirrus_movsf_hard_insn, thumb2_cirrus_movdf_hard_insn): New
2949 * doc/extend.texi: Document ARMv7-M interrupt functions.
2950 * doc/invoke.texi: Document Thumb-2 new cores+architectures.
2952 2007-01-03 Jakub Jelinek <jakub@redhat.com>
2954 * unwind-dw2.c (SIGNAL_FRAME_BIT, EXTENDED_CONTEXT_BIT): Define.
2955 (struct _Unwind_Context): Rename args_size to flags, remove
2956 signal_frame field, add a new args_size field and version field.
2957 (_Unwind_IsSignalFrame, _Unwind_SetSignalFrame,
2958 _Unwind_IsExtendedContext): New inline functions.
2959 (_Unwind_GetGR, _Unwind_SetGR, _Unwind_GetGRPtr, _Unwind_SetGRPtr):
2960 Assume by_value array is only present if _Unwind_IsExtendedContext.
2961 (_Unwind_GetIPInfo, execute_cfa_program, uw_frame_state_for): Use
2962 _Unwind_IsSignalFrame.
2963 (__frame_state_for): Initialize context.flags to EXTENDED_CONTEXT_BIT.
2964 (uw_update_context_1): Use _Unwind_SetSignalFrame.
2965 (uw_init_context_1): Initialize context->flags to
2966 EXTENDED_CONTEXT_BIT.
2967 * config/rs6000/linux-unwind.h (frob_update_context): Use
2968 _Unwind_SetSignalFrame.
2970 2007-01-03 Andrew Pinski <andrew_pinski@playstation.sony.com>
2973 * gimplify.c (gimplify_modify_expr_complex_part): Move below
2974 tree_to_gimple_tuple. Call tree_to_gimple_tuple when we need
2977 2007-01-03 Kazu Hirata <kazu@codesourcery.com>
2979 * config/i386/i386.h (NON_STACK_REG_P, REGNO_OK_FOR_SIREG_P,
2980 REGNO_OK_FOR_DIREG_P, REWRITE_ADDRESS, ASM_OPERAND_LETTER,
2981 RET, AT_SP): Remove.
2982 * config/i386/i386.md (*sse_prologue_save_insn): Use return
2985 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
2986 * cfgbuild.c (find_basic_blocks): Likewise.
2987 * cfgrtl.c (rtl_create_basic_block): Likewise.
2988 * function.c (temp_slots_at_level): Likewise.
2989 * reg-stack.c (stack_regs_mentioned): Likewise.
2990 * regclass.c (allocate_reg_info): Likewise.
2991 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
2992 set_bb_for_stmt, move_block_to_fn): Likewise.
2993 * tree-complex.c (tree_lower_complex): Likewise.
2994 * vec.h (VEC_safe_grow_cleared): New.
2996 * cgraphunit.c, tree-ssa-alias.c: Fix comment typos.
2998 2007-01-03 Zdenek Dvorak <dvorakz@suse.cz>
3000 * loop-unswitch.c (unswitch_loop): Pass probabilities to loopify.
3001 * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Pass probabilities
3003 * cfgloopmanip.c (scale_loop_frequencies): Export.
3004 (loopify): Scale the frequencies by prescribed coefficients.
3005 (set_zero_probability): New function.
3006 (duplicate_loop_to_header_edge): Improve updating of frequencies.
3007 (lv_adjust_loop_entry_edge, loop_version): Set probabilities
3008 and frequencies according to arguments.
3009 * tree-ssa-loop-manip.c (tree_unroll_loop): Set probabilities
3011 * cfg.c (scale_bbs_frequencies_int): Allow scaling the frequencies up.
3012 * modulo-sched.c (sms_schedule): Set probabilities for entering
3013 versioned loop correctly.
3014 * tree-vect-transform.c (vect_transform_loop): Ditto.
3015 * cfgloop.h (loopify, loop_version): Declaration changed.
3016 (scale_loop_frequencies): Declared.
3018 2007-01-02 Jan Hubicka <jh@suse.cz>
3020 * cgraph.c: Include tree-flow.h
3021 (cgraph_add_new-function): Handle IPA_SSA mode; execute
3023 * cgraph.h (enum cgraph_state): Add CGRAPH_STATE_IPA_SSA.
3024 * tree-pass.h (pass_all_early_optimizations): Declare.
3025 * cgraphunit.c (cgraph_process_new_functions): Add IPA_SSA; execute
3027 (cgraph_analyze_function): Do early_local_passes.
3028 * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
3029 Do not add referenced vars.
3030 * tree-optimize.c (gate_all_optimizations): Do not execute when not in
3032 (gate_all_early_local_passes): New gate.
3033 (pass_early_local_passes): Use new gate.
3034 (execute_early_local_optimizations): New functions.
3035 (gate_all_early_optimizations): New gate.
3036 (pass_all_early_optimizations): New pass.
3037 (execute_free_datastructures): Free SSA only when initialized.
3038 (gate_init_datastructures): Init only when optimizing.
3039 (tree_lowering_passes): Do early local passes when called late.
3040 * tree-profile.c (do_tree_profiling): Don't profile functions added
3042 (do_early_tree_profiling, pass_early_tree_profile): Kill.
3043 * tree-cfg.c (update_modified_stmts): Do not update when operands are
3045 * passes.c (init_optimizations_passes): Reorder so we go into SSA
3046 during early_local_passes.
3047 * Makefile.in (cgraph.o): Add dependency on tree-flow.h.
3050 2007-01-02 Carlos O'Donell <carlos@codesourcery.com>
3052 * Makefile.in: Update copyright year.
3054 2007-01-02 Carlos O'Donell <carlos@codesourcery.com>
3056 * Makefile.in: Export GCC_EXEC_PREFIX before calling $(RUNTEST)
3057 in $(lang_checks) and check-consistency targets.
3059 2007-01-02 Jan Hubicka <jh@suse.cz>
3061 * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
3062 Do not add referenced vars.
3063 * tree-cfg.c (update_modified_stmts): Do not update when SSA operands
3065 * passes.c (init_optimization_passes): Put mudflap_2 after
3066 free_datastructures.
3068 2007-01-02 Jan Hubicka <jh@suse.cz>
3070 * tree-optimize (execute_fixup_cfg): Set after_inlining flag.
3071 Set NOTHROW flag on call statements proved to be nothrow.
3072 Update statement of local calls so new pure/const functions are
3073 updated. Update_ssa when in ssa form. Mark PHI nodes of nonlocal
3075 (tree_rest_of_compilation): Register hooks and initialize bitmap
3076 early. Do not set after_inlining flag.
3078 2007-01-02 Steve Ellcey <sje@cup.hp.com>
3080 * sbitmap.c (HOST_BITS_PER_LONG_LONG): Change to
3081 HOST_BITS_PER_LONGLONG
3083 2007-01-02 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
3086 * c-typeck.c (store_init_value): Don't emit pedantic overflow
3087 warning for non-static initializers.
3089 2007-01-02 Steven Bosscher <steven@gcc.gnu.org>
3091 * config/alpha/alpha.md, arm/arm.c, darwin.c, frv/frv.md,
3092 m32r/m32r.c, m32r/m32r.c, mn10300/mn10300.md, pa/pa.c,
3093 rs6000/rs6000.c, s390/s390.md, sh/sh.md, sparc/sparc.c:
3094 Always use set_unique_reg_note to add REG_EQUAL notes.
3096 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
3099 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
3101 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
3102 * cfgbuild.c (find_basic_blocks): Likewise.
3103 * cfgrtl.c (rtl_create_basic_block): Likewise.
3104 * function.c (temp_slots_at_level): Likewise.
3105 * reg-stack.c (stack_regs_mentioned): Likewise.
3106 * regclass.c (allocate_reg_info): Likewise.
3107 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
3108 set_bb_for_stmt, move_block_to_fn): Likewise.
3109 * tree-complex.c (tree_lower_complex): Likewise.
3110 * vec.h (VEC_safe_grow_cleared): New.
3112 2007-01-02 Ian Lance Taylor <iant@google.com>
3114 * c-common.c (c_common_truthvalue_conversion): When warning about
3115 using an assignment as a truth value, set TREE_NO_WARNING.
3117 2007-01-02 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
3120 * c.opt (Wold-style-declaration): New.
3121 * doc/invoke.texi (C-only Warning Options): New.
3122 (Wold-style-declaration): Document it.
3123 (Wextra): Enabled by -Wextra.
3124 * c-opts.c (c_common_post_options): Enabled by -Wextra.
3125 * c-decl.c (declspecs_add_scspec): Replace -Wextra with
3126 -Wold-style-declaration.
3128 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
3130 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
3131 * cfgbuild.c (find_basic_blocks): Likewise.
3132 * cfgrtl.c (rtl_create_basic_block): Likewise.
3133 * function.c (temp_slots_at_level): Likewise.
3134 * reg-stack.c (stack_regs_mentioned): Likewise.
3135 * regclass.c (allocate_reg_info): Likewise.
3136 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
3137 set_bb_for_stmt, move_block_to_fn): Likewise.
3138 * tree-complex.c (tree_lower_complex): Likewise.
3139 * vec.h (VEC_safe_grow_cleared): New.
3141 2007-01-02 Douglas Gregor <doug.gregor@gmail.com>
3143 * c-common.c (c_common_nodes_and_builtins): Since variants of
3144 void_type_node get built before it is given a name, we need to
3145 give those variants the name, too.
3146 (complete_array_type): We need to work with the canonical main
3147 type of the array, from which we will build the qualified version.
3148 * params.def (PARAM_VERIFY_CANONICAL_TYPES): New.
3149 * print-tree.c (print_node): Display canonical type information
3151 * stor-layout.c (layout_type): When we don't know the
3152 alignment of a type for which we're building an array, we end up
3153 guessing wrong, so make the type require structural equality.
3154 * tree.c (make_node_stat): When we build a new type, it is its
3156 (build_type_attribute_qual_variant): When building an attribute
3157 variant, its canonical type is the non-attribute variant. However,
3158 if the attributes are target-dependent and they differ, we need to
3159 use structural equality checks for this type.
3160 (build_qualified_type): A qualified type is not equivalent to its
3161 unqualified variant; set the canonical type appropriately.
3162 (build_distinct_type_copy): When building a distinct type from
3163 another type, the new type is its own canonical type.
3164 (build_variant_type_copy): When building a new type variant, we
3165 assume that it is equivalent to the original type.
3166 (build_pointer_type_for_mode): When building a pointer type, also
3167 build a canonical type pointer.
3168 (build_reference_type_for_mode): When building a reference type,
3169 also build a canonical type reference.
3170 (build_index_type): When we can't hash an index type (e.g.,
3171 because its maximum value is negative), the index type requires
3172 structural equality tests.
3173 (build_array_type): Build the canonical form of an array type.
3174 (build_function_type): Function types require structural equality,
3175 because they contain default arguments, attributes, etc.
3176 (build_method_type_directly): Ditto for method types.
3177 (build_offset_type): Build the canonical offset type.
3178 (build_complex_type): Build the canonical vector type.
3179 (make_vector_type): Build the canonical vector type.
3180 * tree.h (TYPE_CANONICAL): New.
3181 (TYPE_STRUCTURAL_EQUALITY_P): New.
3182 (SET_TYPE_STRUCTURAL_EQUALITY): New.
3183 (struct tree_type): Added "canonical" field.
3184 * params.h (VERIFY_CANONICAL_TYPES): New.
3185 * doc/c-tree.texi (TYPE_CANONICAL): Document.
3186 (TYPE_STRUCTURAL_EQUALITY_P): Document.
3187 (SET_TYPE_STRUCTURAL_EQUALITY): Document.
3188 * doc/invoke.texi (verify-canonical-types): Document --param
3189 parameter for verifying canonical types.
3191 2007-01-02 Joseph Myers <joseph@codesourcery.com>
3193 * config.gcc (powerpc-*-eabispe*, powerpc-*-eabisimaltivec*,
3194 powerpc-*-eabisim*, powerpc-*-eabialtivec*, powerpc-*-eabi*,
3195 powerpc-*-rtems*, powerpc-wrs-vxworks, powerpc-wrs-vxworksae,
3196 powerpcle-*-eabisim*, powerpcle-*-eabi*): Add rs6000/e500.h to
3198 * config/rs6000/e500.h: New.
3199 * config/rs6000/eabi.h (TARGET_SPE_ABI, TARGET_SPE, TARGET_E500,
3200 TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE, TARGET_E500_DOUBLE):
3202 * config/rs6000/linuxspe.h (TARGET_SPE_ABI, TARGET_SPE,
3203 TARGET_E500, TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE,
3204 TARGET_E500_DOUBLE): Remove.
3205 * config/rs6000/vxworks.h (TARGET_SPE_ABI, TARGET_SPE,
3206 TARGET_E500, TARGET_ISEL, TARGET_FPRS): Remove.
3207 * config/rs6000/rs6000.h (CHECK_E500_OPTIONS): Define.
3208 * config/rs6000/rs6000.c (rs6000_override_options): Use
3211 2007-01-02 Joseph Myers <joseph@codesourcery.com>
3213 * config/rs6000/rs6000.c (print_operand): Check (TARGET_SPE ||
3214 TARGET_E500_DOUBLE), not TARGET_E500, for %y.
3215 (rs6000_generate_compare, rs6000_emit_sCOND, output_cbranch,
3216 rs6000_emit_cmove): Don't check TARGET_E500.
3217 * config/rs6000/rs6000.md (bunordered, bordered, sunordered,
3218 sordered): Don't check TARGET_E500.
3220 2007-01-01 Eric Christopher <echristo@apple.com>
3222 * config/mips/mips.c (mips_regno_mode_ok_for_base_p): Use
3223 HARD_REGISTER_NUM_P.
3225 2007-01-01 Roger Sayle <roger@eyesopen.com>
3227 * fold-const.c (fold_binary) <EQ_EXPR>: Fold "(X^C1) eq/ne C2" into
3228 "X eq/ne (C1^C2)". Fold "(X^Z) eq/ne (Y^Z)" as "X eq/ne Y" when Z
3229 has no side-effects. Fold "(X^C1) eq/ne (Y^C2)" as "(X^(C1^C2))
3232 2007-01-01 Mike Stump <mrs@apple.com>
3234 * configure.ac: Remove support for building with Apple's gcc-3.1.
3236 2007-01-02 Joseph Myers <joseph@codesourcery.com>
3239 * caller-save.c (add_stored_regs): Only handle SUBREGs if inner
3240 REG is a hard register. Do not modify REG before calling
3242 * rtlanal.c (subreg_get_info): Don't assert size of XMODE is a
3243 multiple of the size of YMODE for certain lowpart cases.
3245 2007-01-01 Andrew Pinski <pinskia@gmail.com>
3248 * gimplify (voidify_wrapper_expr): Update for
3251 2007-01-01 Andreas Schwab <schwab@suse.de>
3254 * config/ia64/ia64.c (ia64_compute_frame_size): Account space for
3255 save of BR0 in extra_spill_size instead of spill_size.
3256 (ia64_expand_prologue): Save BR0 outside of the gr/br/fr spill
3258 (ia64_expand_epilogue): Restore BR0 from its new location.
3260 2007-01-01 Andrew Pinski <pinskia@gmail.com>
3262 * gimplify.c (gimplify_init_constructor <case VECTOR_TYPE>):
3263 Use a temporary variable if the left hand side is not a gimple
3266 2007-01-01 Andrew Pinski <pinskia@gmail.com>
3268 * gimplify.c (gimplify_return_expr): Make the temporary variable
3269 for the return expression, a gimple register variable.
3271 2007-01-01 Jan Hubicka <jh@suse.cz>
3273 * emit-rtl.c (emit_copy_of_insn_after): Do not call copy_insn_1 for
3276 2007-01-01 Mike Stump <mrs@apple.com>
3278 * configure.ac (HAVE_GAS_LITERAL16): Add autoconf check for
3280 * config/darwin.c (machopic_select_rtx_section): Use
3282 (darwin_mergeable_constant_section): Likewise.
3283 * configure: Regenerate.
3284 * config.in: Regenerate.
3286 2007-01-01 Jan Hubicka <jh@suse.cz>
3287 Andrew Pinski <pinskia@gmail.com>
3289 * cgraphunit.c (cgraph_optimize): Call cgraph_add_new_functions
3290 before starting IPA passes.