1 2007-02-15 Alexandre Oliva <aoliva@redhat.com>
3 * dwarf2out.c (dwarf2out_finish): Accept namespaces as context of
6 2007-02-14 Joseph Myers <joseph@codesourcery.com>
8 * emit-rtl.c (set_mem_attributes_minus_bitpos): Treat complex
9 types as aggregates not scalars.
10 * function.c (assign_stack_temp_for_type): Likewise.
12 2007-02-14 Roger Sayle <roger@eyesopen.com>
13 Zdenek Dvorak <dvorakz@suse.cz>
15 * tree-dump.c (dump_switch_p_1): Require exact match of the option
18 2007-02-14 Zdenek Dvorak <dvorakz@suse.cz>
20 * passes.c (next_pass_1): Clear the next field of the copied
23 2007-02-14 Richard Henderson <rth@redhat.com>
25 * tree-sra.c (early_sra): New.
26 (decl_can_be_decomposed_p): Deny va_list if early_sra.
27 (tree_sra_early, pass_sra_early): New.
28 * tree-pass.h (pass_sra_early): Declare.
29 * passes.c (init_optimization_passes): Use it.
31 2007-02-14 Richard Guenther <rguenther@suse.de>
33 * flags.h (issue_strict_overflow_warning): Convert to a macro.
35 2007-02-14 Dorit Nuzman <dorit@il.ibm.com>
37 PR tree-optimization/30771
38 * tree-vect-analyze.c (vect_determine_vectorization_factor): Traverse
40 (vect_analyze_operations): Induction phis can now be marked as
42 (vect_mark_stmts_to_be_vectorized): No special treatment for phis.
43 Update documentation accordingly.
45 2007-02-14 Nick Clifton <nickc@redhat.com>
47 * builtin-types.def (DEF_FUNCTION_TYPE_x): Do not imply that at
48 most 3 arguments are supported.
49 (DEF_FUNCTION_TYPE_VAR_5): Fix typo in its description.
51 2007-02-13 Seongbae Park <seongbae.park@gmail.com>
53 * bitmap.c (bitmap_and, bitmap_and_compl, bitmap_xor):
54 Ensure dst->current is valid.
56 2007-02-13 Paul Brook <paul@codesourcery.com>
58 * config.gcc: Add arm*-*-uclinux-*eabi.
59 * config/arm/uclinux-elf.h (TARGET_OS_CPP_BUILTINS): Define.
60 (SUBTARGET_EXTRA_LINK_SPEC): Define.
61 (STARTFILE_SPEC, ENDFILE_SPEC): Remove broken -shared handling.
62 (LINK_GCC_C_SEQUENCE_SPEC): Undef.
65 * config/arm/arm.c (arm_override_options): Use r9 as EABI PIC
67 * config/arm/uclinux-eabi.h: New file.
68 * config/arm/linux-eabi.h (WCHAR_TYPE): Remove.
69 * config/arm/linux-gas.h (WCHAR_TYPE): Use unsigned long on AAPCS
72 2007-02-13 Ian Lance Taylor <iant@google.com>
74 * common.opt: Add Wstrict-overflow and Wstrict-overflow=.
75 * flags.h (warn_strict_overflow): Declare.
76 (enum warn_strict_overflow_code): Define.
77 (issue_strict_overflow_warning): New static inline function.
78 * opts.c (warn_strict_overflow): New variable.
79 (common_handle_option): Handle OPT_Wstrict_overflow and
80 OPT_Wstrict_overflow_.
81 * c-opts.c (c_common_handle_option): Set warn_strict_overflow for
83 * fold-const.c: Include intl.h.
84 (fold_deferring_overflow_warnings): New static variable.
85 (fold_deferred_overflow_warning): New static variable.
86 (fold_deferred_overflow_code): New static variable.
87 (fold_defer_overflow_warnings): New function.
88 (fold_undefer_overflow_warnings): New function.
89 (fold_undefer_and_ignore_overflow_warnings): New function.
90 (fold_deferring_overflow_warnings_p): New function.
91 (fold_overflow_warning): New static function.
92 (make_range): Add strict_overflow_p parameter. Change all
94 (extract_muldiv, extract_muldiv_1): Likewise.
95 (fold_unary) [ABS_EXPR]: Check ABS_EXPR before calling
96 tree_expr_nonnegative_p.
97 (fold_negate_expr): Call fold_overflow_warning.
98 (fold_range_test): Likewise.
99 (fold_comparison): Likewise.
100 (fold_binary): Likewise. Call tree_expr_nonnegative_warnv_p
101 instead of tree_expr_nonnegative_p.
102 (tree_expr_nonnegative_warnv_p): Rename from
103 tree_expr_nonnegative_p, add strict_overflow_p parameter.
104 (tree_expr_nonnegative_p): New function.
105 (tree_expr_nonzero_warnv_p): Rename from tree_expr_nonzero_p, add
106 strict_overflow_p parameter.
107 (tree_expr_nonzero_p): New function.
108 * passes.c (verify_interpass_invariants): New static function.
109 (execute_one_pass): Call it.
110 * tree-ssa-loop-niter.c (expand_simple_operations): Ignore fold
112 (number_of_iterations_exit, loop_niter_by_eval): Likewise.
113 (estimate_numbers_of_iterations): Likewise.
114 (scev_probably_wraps_p): Likewise.
115 * tree-ssa-ccp.c: Include "toplev.h".
116 (evaluate_stmt): Defer fold overflow warnings until we know we are
118 (struct fold_stmt_r_data): Add stmt field.
119 (fold_stmt_r): Defer fold overflow warnings until we know we
121 (fold_stmt): Initialize stmt field of fold_stmt_r_data.
122 (fold_stmt_inplace): Likewise.
123 * tree-cfgcleanup.c: Include "toplev.h" rather than "errors.h".
124 (cleanup_control_expr_graph): Defer fold overflow warnings until
125 we know we are going to optimize.
126 * tree-cfg.c (fold_cond_expr_cond): Likewise.
127 * tree-ssa-threadedge.c (simplify_control_stmt_condition):
129 * tree-vrp.c (vrp_expr_computes_nonnegative): Call
130 tree_expr_nonnegative_warnv_p instead of tree_expr_nonnegative_p.
131 * tree-ssa-loop-manip.c (create_iv): Likewise.
132 * c-typeck.c (build_conditional_expr): Likewise.
133 (build_binary_op): Likewise.
134 * tree-vrp.c (vrp_expr_computes_nonzero): Call
135 tree_expr_nonzero_warnv_p instead of tree_expr_nonzero_p.
136 (extract_range_from_unary_expr): Likewise.
137 * simplify-rtx.c (simplify_const_relational_operation): Warn when
138 assuming that signed overflow does not occur.
139 * c-common.c (pointer_int_sum): Ignore fold overflow warnings.
140 * tree.h (tree_expr_nonnegative_warnv_p): Declare.
141 (fold_defer_overflow_warnings): Declare.
142 (fold_undefer_overflow_warnings): Declare.
143 (fold_undefer_and_ignore_overflow_warnings): Declare.
144 (fold_deferring_overflow_warnings_p): Declare.
145 (tree_expr_nonzero_warnv_p): Declare.
146 * doc/invoke.texi (Option Summary): Add -Wstrict-overflow to list
148 (Warning Options): Document -Wstrict-overflow.
149 * Makefile.in (tree-ssa-threadedge.o): Depend on toplev.h.
150 (tree-ssa-ccp.o): Likewise.
151 (tree-cfgcleanup.o): Change errors.h dependency to toplev.h.
152 (fold-const.o): Depend on intl.h.
154 2007-02-13 Ian Lance Taylor <iant@google.com>
157 * lower-subreg.c (resolve_simple_move): Decompose subregs in
160 2007-02-13 Stuart Hastings <stuart@apple.com>
162 * gcc/config/i386/i386.md (fixuns_truncdfhi2): Require SSE2.
164 2007-02-13 Richard Henderson <rth@redhat.com>
166 * config/alpha/alpha.c (alpha_stdarg_optimize_hook): Strip
167 handled_component_p before looking for the indirect_ref.
169 2007-02-13 Richard Henderson <rth@redhat.com>
171 * config/i386/i386.md (bswapsi_1): Rename from bswapsi2,
172 remove flags clobber.
173 (bswapsi2): New expander, emit code for !TARGET_BSWAP.
174 (bswaphi_lowpart): New.
175 (bswapdi2): Rename from bswapdi2_rex, remove flags clobber,
176 remove TARGET_BSWAP test. Delete expander of the same name.
178 * optabs.c (widen_bswap, expand_doubleword_bswap): New.
179 (expand_unop): Use them.
181 2007-02-13 Uros Bizjak <ubizjak@gmail.com>
183 * config/i386/i386.md (cmpdi_ccno_1_rex64, *cmpsi_ccno_1,
184 *cmphi_ccno_1, *cmpqi_ccno_1, *movsi_xor, *movstricthi_xor,
185 *movstrictqi_xor, *movdi_xor_rex64, *ashldi3_1_rex64,
186 *ashldi3_cmp_rex64, *ashldi3_cconly_rex64, ashlsi3, *ashlsi3_1_zext,
187 *ashlsi3_cmp, *ashlsi3_cconly, *ashlsi3_cmp_zext, *ashlhi3_1_lea,
188 *ashlhi3_1, *ashlhi3_cmp, *ashlhi3_cconly, *ashlqi3_1_lea,
189 *ashlqi3_1, *ashlqi3_cmp, *ashlqi3_cconly): Remove equivalent
190 assembler dialect choice from asm templates.
192 2007-02-12 Richard Henderson <rth@redhat.com>
194 * config/i386/i386.md (fixuns_trunc<SSEMODEF>si_1): New insn.
195 (fixuns_trunc<SSEMODEF>si2): Use it.
196 * config/i386/sse.md (vec_setv4sf_0): Export.
197 * config/i386/i386.c (ix86_build_const_vector): Export.
198 (ix86_split_convert_uns_si_sse): Rename from
199 ix86_expand_convert_uns_si_sse and rewrite as a splitter.
200 * config/i386/i386-protos.h: Update.
202 2007-02-13 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
205 * c-typeck.c (c_finish_return): Improve warning message.
207 2007-02-12 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
209 * alias.c (find_symbolic_term): Delete unused function.
211 2007-02-12 Uros Bizjak <ubizjak@gmail.com>
213 * config/i386/i386.md (paritydi2, paritysi2): New expanders.
214 (paritydi2_cmp, paritydi2_cmp): New insn and split patterns.
215 (*parityhi2_cmp, *parityqi2_cmp): New insn patterns.
217 2007-02-12 Eric Botcazou <ebotcazou@adacore.com>
219 * tree.h (DECL_IGNORED_P): Document further effect for FUNCTION_DECL.
220 * cgraphunit.c (cgraph_expand_function): If DECL_IGNORED_P is set on
221 the function, temporarily point the debug interface to the null one.
223 2007-02-12 Eric Botcazou <ebotcazou@adacore.com>
225 * dwarf2out.c (round_up_to_align): New static function.
226 (field_byte_offset): Use it to round the offset.
228 2007-02-12 Richard Henderson <rth@redhat.com>
230 * config/alpha/alpha.md (bswapsi2, bswapdi2): New.
231 (inswl_const): Export.
233 2007-02-12 Richard Henderson <rth@redhat.com>
235 * calls.c (emit_library_call_value_1): If PROMOTE_MODE modifed the
236 result mode of the libcall, convert back to outmode.
238 2007-02-12 Roger Sayle <roger@eyesopen.com>
240 * config/i386/i386.md (*bswapdi2_rex): Renamed from bswapdi2.
241 (bswapdi2): New define_expand to implement 32-bit implementation.
243 2007-02-12 Nick Clifton <nickc@redhat.com>
245 * doc/invoke.texi (Overall Options): Document --help=.
246 * gcc.c (target_help_flag): Rename to print_subprocess_flag.
247 (cc1_options): Pass --help= on to cc1.
248 (display_help): Add description of --help=.
249 (process_command): Add code to handle --help=. Allow translated
250 --help and --target-help switches to be passed on to compiler
252 (main): Remove unused if statement.
253 * opts.c (columns): Remove.
254 (LEFT_COLUMN): Define.
255 (wrap_help): Add columns argument.
256 (print_filtered_help): Change parameters to be an include bitmask,
257 an exclude bitmask, an any bitmask and the column width. Move the
258 code to display the params list here. Add code to display the
259 status of options rather than their descriptions if the quiet flag
261 (print_specific_help): Change parameters to be an include bitmask,
262 an exclude bitmask and an any bitmask. Move code to look up the
263 column width here. Decide upon the title for an options listing.
264 (common_handle_options): Add code to handle --help=. Adapt code
265 for --help and --target-help to use the revised form of the
266 print_specific_help function.
267 (print_help): Delete.
268 (print_param_help): Delete.
269 (print_switch): Delete.
270 * opts.h (cl_lang_count): Add prototype.
271 (CL_PARAMS, CL_WARNING, CL_OPTIMIZATION, CL_MIN_OPTION_CLASS,
272 CL_MAX_OPTION_CLASS): New defines.
273 * optc-gen.awk: Add construction of cl_lang_count.
274 * c.opt: Add Warning attribute to warning options and Optimization
275 attribute to optimization options.
276 * common.opt: Likewise.
278 Add -fhelp and -ftarget-help as aliases for the transformed --help
279 and --target-help options.
280 * opt-functions.awk: Add code to handle Warning and Optimization
283 2007-02-12 Richard Henderson <rth@redhat.com>
285 * config/alpha/constraints.md: New file.
286 * config/alpha/alpha.c: Include tm-constrs.h.
287 (alpha_const_ok_for_letter_p, alpha_const_double_ok_for_letter_p,
288 alpha_extra_constraint): Remove.
289 (alpha_emit_conditional_branch): Use satisfies_constraint_*.
290 * config/alpha/alpha-protos.h: Update.
291 * config/alpha/alpha.h (REG_CLASS_FROM_LETTER): Remove.
292 (CONST_OK_FOR_LETTER_P, CONST_DOUBLE_OK_FOR_LETTER_P): Remove.
293 (EXTRA_CONSTRAINT): Remove.
294 * config/alpha/alpha.md: Include constraints.md.
295 (adddi splitter): Use satisfies_constraint_*.
296 * config/alpha/predicates.md (add_operand): Likewise.
297 (sext_add_operand, addition_operation): Likewise.
299 2007-02-12 Dorit Nuzman <dorit@il.ibm.com>
301 PR tree-optimization/29145
302 * tree-data-ref.c (base_addr_differ_p): Make us more conservative
303 in our handling of restrict qualified pointers.
305 2007-02-12 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
308 * doc/invoke.texi (Wunused-value): Update description.
309 (Wextra): Delete item.
310 * opts.c (set_Wextra): Don't use the value of Wextra to set the
311 value of Wunused-value.
312 * c-typeck.c (c_process_expr_stmt): Don't check extra_warnings.
313 (c_finish_stmt_expr): Don't check extra_warnings.
314 (emit_side_effect_warnings): The caller is responsible to check
317 2007-02-11 Roger Sayle <roger@eyesopen.com>
318 Matt Thomas <matt@3am-software.com>
320 * simplify-rtx.c (simplify_relational_operation_1): Correct typo.
322 2007-02-11 Roger Sayle <roger@eyesopen.com>
324 * simplify-rtx.c (simplify_relational_operation_1): Optimize
325 comparisons of POPCOUNT against zero.
326 (simplify_const_relational_operation): Likewise.
328 2007-02-11 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
330 * doc/invoke.texi (Wextra): Delete outdated paragraph.
332 2007-02-11 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
334 * dwarf2out.c (root_type): Delete unused function.
336 2007-02-11 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
338 * genattrtab.c (contained_in_p): Delete unused function.
339 (write_expr_attr_cache): Likewise.
341 2007-02-11 Jan Hubicka <jh@suse.cz>
343 * ipa-inline.c (cgraph_edge_badness): Add "else" missing in the
346 2007-02-11 Steven Bosscher <steven@gcc.gnu.org>
348 * fwprop.c (try_fwprop_subst): Use set_unique_reg_note
349 to add the REG_EQ* note.
350 * see.c (see_merge_one_use_extension): Likewise.
351 * local-alloc.c (update_equiv_regs): Likewise. Also don't
352 turn REG_EQUAL notes into REG_EQUIV notes if the target
353 register may have more than one set.
354 * function.c (assign_parm_setup_reg): Use set_unique_reg_note.
355 * gcse.c (try_replace_reg): Likewise.
356 * alias.c (init_alias_analysis): Use find_reg_equal_equiv_note.
357 * calls.c (fixup_tail_calls): Likewise. Abort if there is
358 more than one REG_EQUIV note.
359 * reload1.c (gen_reload): Use set_unique_reg_note.
361 2007-02-11 Uros Bizjak <ubizjak@gmail.com>
363 * config/i386/i386.c (TARGET_VECTORIZE_BUILTIN_CONVERSION): Define.
364 (ix86_builtin_conversion): New function.
366 2007-02-06 Mark Mitchell <mark@codesourcery.com>
369 * tree.h (DECL_REPLACEABLE_P): New macro.
370 * except.c (set_nothrow_function_flags): Likewise.
372 2007-02-11 Tehila Meyzels <tehila@il.ibm.com>
373 Ira Rosen <irar@il.ibm.com>
374 Dorit Nuzman <dorit@il.ibm.com>
376 * doc/tm.texi (TARGET_VECTORIZE_BUILTIN_CONVERSION): New target hook.
377 * targhooks.c (default_builtin_vectorized_conversion): New.
378 * targhooks.h (default_builtin_vectorized_function): New declaration.
379 * target.h (struct vectorize): Add builtin_conversion field.
380 * tree-vectorizer.h (type_conversion_vec_info_type): New enum
381 stmt_vec_info_type value.
382 (vectorizable_conversion): New declaration.
383 * tree-vect-analyze.c (vect_analyze_operations): Add
384 vectorizable_conversion call.
385 * target-def.h (TARGET_VECTORIZE_BUILTIN_CONVERSION): New.
386 * tree-vect-transform.c (vectorizable_conversion): New function.
387 (vect_transform_stmt): Add case for type_conversion_vec_info_type.
388 * tree-vect-generic.c (expand_vector_operations_1): Consider correct
390 * config/rs6000/rs6000.c (rs6000_builtin_conversion): New.
391 (TARGET_VECTORIZE_BUILTIN_CONVERSION): Defined.
392 (rs6000_expand_builtin): Add handling a case of ALTIVEC_BUILTIN_VCFUX
393 or ALTIVEC_BUILTIN_VCFSX.
395 2007-02-10 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
398 * pa.md (movdf): For 64-bit target, fail if operand 1 is a non-zero
399 CONST_DOUBLE and operand 0 is a hard register.
400 (movdi): For 64-bit target, remove code to force CONST_DOUBLE to
401 memory. Fail if operand 1 is a non-zero CONST_INT and operand 0
402 is a hard floating-point register.
404 2007-02-10 Richard Henderson <rth@redhat.com>
405 Jakub Jelinek <jakub@redhat.com>
406 Alexandre Oliva <aoliva@redhat.com>
408 * Makefile.in (libgcc-support, libgcc.mvars): Add emutls.c.
409 * builtin-types.def (BT_WORD): Make unsigned.
410 (BT_FN_VOID_PTR_WORD_WORD_PTR): New.
411 * builtins.def (BUILT_IN_EMUTLS_GET_ADDRESS): New.
412 (BUILT_IN_EMUTLS_REGISTER_COMMON): New.
413 * c-decl.c (grokdeclarator): Don't error if !have_tls.
414 * c-parser.c (c_parser_omp_threadprivate): Likewise.
415 * dwarf2out.c (loc_descriptor_from_tree_1): Don't do anything for
417 * expr.c (emutls_var_address): New.
418 (expand_expr_real_1): Expand emulated tls.
419 (expand_expr_addr_expr_1): Likewise.
420 * libgcc-std.ver: Add __emutls_get_address, __emutls_register_common.
421 * output.h (emutls_finish): Declare.
422 * toplev.c (compile_file): Call it.
423 * tree-ssa-address.c (gen_addr_rtx): Check for const-ness of the
424 address before wrapping in CONST.
425 * varasm.c (emutls_htab, emutls_object_type): New.
426 (EMUTLS_VAR_PREFIX, EMUTLS_TMPL_PREFIX): New.
427 (get_emutls_object_name, get_emutls_object_type): New.
428 (get_emutls_init_templ_addr, emutls_decl): New.
429 (emutls_common_1, emutls_finish): New.
430 (assemble_variable): When emulating tls, swap decls; generate
431 constructor for the emutls objects.
432 (do_assemble_alias): When emulating tls, swap decl and target name.
433 (default_encode_section_info): Don't add SYMBOL_FLAG_TLS_SHIFT
435 * varpool.c (decide_is_variable_needed): Look at force_output.
436 Recurse for emulated tls.
437 (cgraph_varpool_remove_unreferenced_decls): Remove checks redundant
438 with decide_is_variable_needed.
439 * emutls.c: New file.
440 * config/sparc/sol2.h (ASM_DECLARE_OBJECT_NAME): Only emit
441 tls_object for real tls.
443 2007-02-10 Kaz Kojima <kkojima@gcc.gnu.org>
445 PR rtl-optimization/29599
446 * reload1.c (eliminate_regs_in_insn): Take the destination
447 mode into account when computing the offset.
449 2007-02-09 Stuart Hastings <stuart@apple.com>
450 Richard Henderson <rth@redhat.com>
452 * gcc/config/i386/i386.h (TARGET_KEEPS_VECTOR_ALIGNED_STACK): New.
453 * gcc/config/i386/darwin.h: (TARGET_KEEPS_VECTOR_ALIGNED_STACK): New.
454 * gcc/config/i386/i386.md (fixuns_trunc<mode>si2, fixuns_truncsfhi2,
455 fixuns_truncdfhi2): New.
456 (fix_truncsfdi_sse): Call ix86_expand_convert_sign_didf_sse.
457 (floatunsdidf2): Call ix86_expand_convert_uns_didf_sse.
458 (floatunssisf2): Add call to ix86_expand_convert_uns_sisf_sse.
459 (floatunssidf2): Allow nonimmediate source.
460 * gcc/config/i386/sse.md (movdi_to_sse): New.
461 (vec_concatv2di): Drop '*'.
462 * gcc/config/i386/i386-protos.h (ix86_expand_convert_uns_si_sse,
463 ix86_expand_convert_uns_didf_sse, ix86_expand_convert_uns_sidf_sse,
464 ix86_expand_convert_uns_sisf_sse, ix86_expand_convert_sign_didf_sse):
466 * gcc/config/i386/i386.c (ix86_expand_convert_uns_si_sse,
467 ix86_expand_convert_uns_didf_sse, ix86_expand_convert_uns_sidf_sse,
468 ix86_expand_convert_uns_sisf_sse, ix86_expand_convert_sign_didf_sse,
469 ix86_build_const_vector, ix86_expand_vector_init_one_nonzero): New.
470 (ix86_build_signbit_mask): Fix decl of v, refactor to call
471 ix86_build_const_vector.
472 (x86_emit_floatuns): Rewrite.
474 2007-02-10 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
476 * genautomata.c (longest_path_length): Delete unused function.
477 (struct state): Delete unused longest_path_length.
478 (UNDEFINED_LONGEST_PATH_LENGTH): Delete unused macro.
479 (get_free_state): Delete unused.
481 2007-02-09 Jan Hubicka <jh@suse.cz>
483 * params.def (PARAM_INLINE_UNIT_GROWTH): Set to 30.
484 * doc/invoke.texi (inline-unit-growth): Update default value.
486 * Makefile.in (passes.o, ipa-inline.o): Add dependencies.
487 * cgraphbuild.c (build_cgraph_edges): Compute frequencies.
488 (rebuild_cgraph_edges): Likewise.
489 * cgraph.c (cgraph_set_call_stmt): Add new argument frequency.
490 (dump_cgraph_node): Dump frequencies.
491 (cgraph_clone_edge): Add frequency scales.
492 (cgraph_clone_node): Add freuqnecy.
493 * cgraph.h (cgraph_edge): Add freuqnecy argument.
494 (CGRAPH_FREQ_BASE, CGRAPH_FREQ_MAX): New constants.
495 (cgraph_create_edge, cgraph_clone_edge, cgraph_clone_node): Update.
496 * tree-pass.h (TODO_rebuild_frequencies): New constant.
497 * cgraphunit.c (verify_cgraph_node): Verify frequencies.
498 (cgraph_copy_node_for_versioning): Update call of cgraph_clone_edge.
499 (save_inline_function_body): Likewise.
500 * ipa-inline.c: inluce rtl.h
501 (cgraph_clone_inlined_nods): Update call of cgraph_clone_node.
502 (cgraph_edge_badness): Use frequencies.
503 (cgraph_decide_recursive_inlining): Update clonning.
504 (cgraph_decide_inlining_of_small_function): Dump frequency.
505 * predict.c (estimate_bb_frequencies): Export.
506 * predict.h (estimate_bb_frequencies): Declare.
507 * tree-inline.c (copy_bb): Watch overflows.
508 (expand_call_inline): Update call of cgraph_create_edge.
509 (optimize_inline_calls): Use TODO flags to update frequnecies.
510 * passes.h: Include predict.h
511 (init_optimization_passes): Move profile ahead.
512 (execute_function_todo): Handle TODO_rebuild_frequencies.
514 2007-02-09 Roger Sayle <roger@eyesopen.com>
516 * config/alpha/alpha.c (emit_insxl): Force the first operand of
517 the insbl or inswl pattern into a register.
519 2007-02-09 Roger Sayle <roger@eyesopen.com>
521 * config/ia64/ia64.md (bswapdi2): New define_insn.
523 2007-02-09 Richard Henderson <rth@redhat.com>
525 * config/i386/constraints.md (Ym): New constraint.
526 * config/i386/i386.md (movsi_1): Change Y2 to Yi constraints.
527 (movdi_1_rex64): Split sse and xmm general register moves from
528 memory move alternatives. Use conditional register constraints.
529 (movsf_1, movdf_integer): Likewise.
530 (zero_extendsidi2_32, zero_extendsidi2_rex64): Likewise.
531 (movdf_integer_rex64): New.
532 (pushsf_rex64): Fix output constraints.
533 * config/i386/sse.md (sse2_loadld): Split rm alternative, use Yi.
534 (sse2_stored): Likewise.
535 (sse2_storeq_rex64): New.
536 * config/i386/i386.c (x86_inter_unit_moves): Enable for not
538 (ix86_secondary_memory_needed): Don't bypass TARGET_INTER_UNIT_MOVES
539 for optimize_size. Remove SF/DFmode hack.
541 2007-02-09 Dwarakanath Rajagopal <dwarak.rajagopal@amd.com>
543 * config/i386/driver-i386.c: Turn on -mtune=native for AMDFAM10.
546 2007-02-09 Nathan Sidwell <nathan@codesourcery.com>
547 Richard Sandiford <richard@codesourcery.com>
549 * config.gcc (m68010-*-netbsdelf*, m68k*-*-netbsdelf*)
550 (m68k*-*-openbsd*, m68k-*-linux*): Set default_cf_cpu.
551 (m68k-*-aout*, m68k-*-coff*, m68k-*-uclinux*, m68k-*-rtems*): Add
552 m68k/t-mlib to tmake_file.
553 (m68020-*-elf*, m68k-*-elf*): Likewise. Add t-m68kbare as well.
554 (m68k*-*-*): Use --with-arch to pick a default for --with-cpu.
555 (m680[012]0-*-*, m68k*-*-*): Add support for --with-arch.
556 Allow it to be cf or m68k. Set m68k_arch_family. If that
557 variable is not empty, add t-$m68k_arch_family to tmake_file.
558 Add t-mlibs to tmake_file.
559 * doc/install.texi: Document --with-arch=m68k and --with-arch=cf.
560 * config/m68k/t-cf: New file.
561 * config/m68k/t-m68k: Likewise.
562 * config/m68k/t-mlibs: Likewise.
563 * config/m68k/t-m68kbare (MULTILIB_OPTIONS, MULTILIB_DIRNAMES)
564 (MULTILIB_MATCHES, MULTILIB_EXCEPTIONS): Delete.
565 (M68K_MLIB_DIRNAMES, M68K_MLIB_OPTIONS): Define.
566 * config/m68k/t-m68kelf (MULTILIB_OPTIONS, MULTILIB_DIRNAMES)
567 (MULTILIB_MATCHES, MULTILIB_EXCEPTIONS, LIBGCC, INSTALL_LIBGCC):
569 * config/m68k/t-openbsd (MULTILIB_OPTIONS, LIBGCC): Delete.
570 (INSTALL_LIBGCC): Delete.
571 (M68K_MLIB_DIRNAMES, M68K_MLIB_OPTIONS): Define.
572 * config/m68k/t-rtems (MULTILIB_OPTIONS, MULTILIB_DIRNAMES)
573 (MULTILIB_MATCHES, MULTILIB_EXCEPTIONS): Delete.
574 (M68K_MLIB_CPU): Define.
575 * config/m68k/t-uclinux (MULTILIB_OPTIONS, MULTILIB_DIRNAMES)
576 (MULTILIB_MATCHES, MULTILIB_EXCEPTIONS): Delete.
577 (M68K_MLIB_CPU, M68K_MLIB_OPTIONS, M68K_MLIB_DIRNAMES): Define.
579 2007-02-09 Zdenek Dvorak <dvorakz@suse.cz>
580 Richard Guenther <rguenther@suse.de>
583 * fold-const.c (fold_comparison): Handle obfuscated comparisons
584 against INT_MIN/INT_MAX.
585 * tree-ssa-loop-ivcanon.c (remove_empty_loop): Print to dump
586 file if a loop is removed.
588 2007-02-09 Joseph Myers <joseph@codesourcery.com>
590 * calls.c (store_one_arg): Pass correct alignment to
591 emit_push_insn for non-BLKmode values.
592 * expr.c (emit_push_insn): If STRICT_ALIGNMENT, copy to an
593 unaligned stack slot via a suitably aligned slot.
595 2007-02-08 DJ Delorie <dj@redhat.com>
597 * config/m32c/m32c.c (m32c_unpend_compare): Add default to silence
599 (legal_subregs): Use unsigned char, make const.
600 (m32c_illegal_subreg_p): Use ARRAY_SIZE. Delete unused variables.
602 2007-02-08 Paul Brook <paul@codesourcery.com>
604 * config/arm/lib1funcs.asm (RETLDM): Pop directly into PC when no
605 special interworking needed.
607 2007-02-08 Harsha Jagasia <harsha.jagasia@amd.com>
609 * config/i386/xmmintrin.h: Make inclusion of emmintrin.h
610 conditional to __SSE2__.
611 (Entries below should have been added to first ChangeLog
612 entry for amdfam10 dated 2007-02-05)
613 * config/i386/emmintrin.h: Generate #error if __SSE2__ is not
615 * config/i386/pmmintrin.h: Generate #error if __SSE3__ is not
617 * config/i386/tmmintrin.h: Generate #error if __SSSE3__ is not
620 2007-02-08 DJ Delorie <dj@redhat.com>
622 * config/m32c/m32c-protos.h (m32c_illegal_subreg_p): New.
623 * config/m32c/m32c.c (legal_subregs): New.
624 (m32c_illegal_subreg_p): New.
625 * config/m32c/predicates.md (m32c_any_operand): Use it to reject
626 unsupported subregs of hard regs.
628 2007-02-08 Jan Hubicka <jh@suse.cz>
630 * tree-cfg.c (bsi_replace): Shortcut when replacing the statement with
631 the same one; always update histograms.
633 2007-02-08 Diego Novillo <dnovillo@redhat.com>
635 * passes.c (init_optimization_passes): Tidy comment.
637 2007-02-08 Roger Sayle <roger@eyesopen.com>
639 * simplify-rtx.c (simplify_unary_operation_1) <POPCOUNT>: We can
640 strip zero_extend, bswap and rotates from POCOUNT's argument.
641 <PARITY>: Likewise, we can strip not, bswap, sign_extend,
642 zero_extend and rotates from PARITY's argument.
643 <BSWAP>: A byte-swap followed by a byte-swap is an identity.
644 (simplify_const_unary_operation) <BSWAP>: Evaluate the byte-swap
645 of an integer constant at compile-time.
647 2007-02-08 Diego Novillo <dnovillo@redhat.com>
650 * tree-flow.h (struct var_ann_d): Remove field 'is_used'.
652 * tree-ssa-alias.c (compute_is_aliased): Remove. Update all
655 * tree-ssa-live.c (remove_unused_locals): Do not remove
656 TREE_ADDRESSABLE variables.
657 * tree-ssa-structalias.c (compute_points_to_sets): Tidy.
658 * tree-ssa-operands.c (add_virtual_operand): Remove argument
659 FOR_CLOBBER. Update all users.
660 If VAR has an associated alias set, add a virtual operand for
661 it if no alias is found to conflict with the memory reference.
663 2007-02-07 Jan Hubicka <jh@suse.cz>
664 Robert Kidd <rkidd@crhc.uiuc.edu>
666 * value-prof.c (visit_hist, free_hist): Return 1 instead of 0.
668 2007-02-07 Ian Lance Taylor <iant@google.com>
670 * lower-subreg.c (simple_move): Reject PARTIAL_INT modes.
672 2007-02-07 Roger Sayle <roger@eyesopen.com>
674 * config/rs6000/rs6000.md (ctz<mode>2, ffs<mode>2, popcount<mode>2,
675 parity<mode>2, smulsi3_highpart, abstf2_internal, allocate_stack,
676 tablejumpdi, movsi_to_cr_one): Remove constraints from
677 define_expand's match_operands.
679 2007-02-07 Roger Sayle <roger@eyesopen.com>
681 * global.c (compute_regsets): Move declatation of "i" inside of
682 #ifdef ELIMINABLE_REGS to avoid unused variable bootstrap failure.
684 2007-02-07 Jakub Jelinek <jakub@redhat.com>
687 * gimplify.c (gimplify_scan_omp_clauses): Remove special casing
688 of INDIRECT_REF <RESULT_DECL>.
690 * config/i386/i386.c (override_options): Set PTA_SSSE3 for core2.
692 2007-02-06 J"orn Rennecke <joern.rennecke@arc.com>
693 Kaz Kojima <kkojima@gcc.gnu.org>
696 * config/sh/sh.c (expand_cbranchdi4): Use scratch register
698 (sh_initialize_trampoline): Add parentheses to avoid a warning.
700 2007-02-06 Zdenek Dvorak <dvorakz@suse.cz>
702 * doc/loop.texi: Document possibility not to perform disambiguation
703 of loops with multiple latches.
704 * cfgloopmanip.c (alp_enum_p): Removed.
705 (add_loop): Handle subloops. Use get_loop_body_with_size.
706 (create_preheader): Do not allow ENTRY_BLOCK_PTR to be preheader.
707 * cfghooks.c (redirect_edge_and_branch_force): Set dominator for
708 the new forwarder block.
709 (make_forwarder_block): Only call new_bb_cbk if it is not NULL.
710 Handle the case latch is NULL.
711 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Avoid cfg modifications
712 when marking loop exits.
713 * ifcvt.c (if_convert): Ditto. Mark loop exits even if cfg cannot
715 * loop-init.c (loop_optimizer_init): Do not modify cfg. Call
716 disambiguate_loops_with_multiple_latches.
717 * tree-cfgcleanup.c (cleanup_tree_cfg_loop): Calculate dominators
718 before fix_loop_structure.
719 * cfgloop.c: Include pointer-set.h and output.h.
720 (canonicalize_loop_headers, HEADER_BLOCK, LATCH_EDGE,
721 update_latch_info, mfb_keep_just, mfb_keep_nonlatch): Removed.
722 (get_loop_latch_edges, find_subloop_latch_edge_by_profile,
723 find_subloop_latch_edge_by_ivs, find_subloop_latch_edge,
724 mfb_redirect_edges_in_set, form_subloop, merge_latch_edges,
725 disambiguate_multiple_latches, get_loop_body_with_size,
726 disambiguate_loops_with_multiple_latches): New functions.
727 (flow_loop_dump): Dump multiple latch edges.
728 (flow_loop_nodes_find): Handle loops with multiple latches.
729 (flow_loops_find): Ditto. Do not call canonicalize_loop_headers.
730 (glb_enum_p): Modified.
731 (get_loop_body): Use get_loop_body_with_size.
732 * cfgloop.h (LOOPS_HAVE_RECORDED_EXITS): New flag.
733 (AVOID_CFG_MODIFICATIONS): New constant.
734 (disambiguate_loops_with_multiple_latches, add_loop,
735 get_loop_body_with_size): Declare.
736 * Makefile.in (cfgloop.o): Add pointer-set.h and output.h.
738 2007-02-06 Seongbae Park <seongbae.park@gmail.com>
741 * global.c (compute_regsets): New function.
742 (global_alloc): Refactored ELIMINABLE_REGSET
743 and NO_GLOBAL_ALLOC_REGS computation out.
744 (rest_of_handle_global_alloc): Call compute_regsets()
745 for non-optimizing case.
747 2007-02-06 Richard Henderson <rth@redhat.com>
749 * config/i386/constraints.md (Y2): Rename from Y.
750 (Yi): New constraint.
751 * config/i386/i386.md (movsi_1, movdi_2, pushdf_nointeger,
752 pushdf_integer, movdf_nointeger, movdf_integer, zero_extendsidi2_32,
753 zero_extendsidi2_rex64, truncxfdf2_mixed): Change Y constraints to Y2.
754 (extendsfdf2_mixed, extendsfdf2_sse, truncdfsf_fast_mixed,
755 truncdfsf_fast_sse, truncdfsf_mixed, fix_truncdfdi_sse,
756 fix_truncdfsi_sse, floatsidf2_mixed, floatsidf2_sse,
757 floatdidf2_mixed, floatdidf2_sse, absnegdf2_mixed,
758 absnegdf2_sse, sse_setccdf, fop_df_comm_mixed, fop_df_comm_sse,
759 fop_df_1_mixed, fop_df_1_sse): Change Y constraints to x.
760 * config/i386/mmx.md (mov<MMXMODEI>_internal_rex64,
761 mov<MMXMODEI>_internal, movv2sf_internal_rex64, movv2sf_internal,
762 vec_extractv2si_1): Change Y constraints to Y2.
763 * config/i386/sse.md (vec_setv4sf_0, vec_concatv2df, vec_dupv4si,
764 vec_dupv2di, sse2_concatv2si, vec_concatv4si_1, vec_concatv2di):
765 Change Y constraints to Y2.
766 (sse2_loadld): Change Y constraints to x.
768 2007-02-06 Roger Sayle <roger@eyesopen.com>
770 * config/rs6000/rs6000.md (popcount<mode>2): Rewrite.
771 (parity<mode>2): New define_expand using rs6000_emit_parity.
772 * config/rs6000/rs6000.c (rs6000_emit_popcount,
773 rs6000_emit_parity): New functions.
774 * config/rs6000/rs6000-protos.h (rs6000_emit_popcount,
775 rs6000_emit_parity): Prototype here.
777 2007-02-06 Ian Lance Taylor <iant@google.com>
779 * lower-subreg.c (simple_move_operand): Reject CONST.
780 (resolve_clobber): Call validate_change rather than directly
781 assigning to XEXP (pat, 0).
783 2006-02-06 Paolo Bonzini <bonzini@gnu.org>
785 * Makefile.in (tree-ssa-loop-ivopts.o): Add pointer-set.h dependency.
786 (tree-ssa-reassoc.o): Add pointer-set.h dependency.
787 (tree-cfg.o): Remove hashtab.h dependency.
789 * tree-ssa-loop-ivopts.c: Include pointer-set.h.
790 (struct ivopts_data): Change niters to pointer_map_t.
791 (struct nfe_cache_elt, nfe_hash, nfe_eq): Delete.
792 (niter_for_exit): Create pointer_map on demand. Change for
794 (tree_ssa_iv_optimize_init): Initialize data->niters to NULL.
795 (free_loop_data): Destroy data->niters if created and reset field.
796 (tree_ssa_iv_optimize_finalize): Don't delete data->niters here.
797 (tree_ssa_iv_optimize_loop): Check for presence of stale data.
799 * tree-ssa-reassoc.c: Include pointer-set.h.
800 (bb_rank): Change to long *.
801 (operand_rank): Change to pointer_map_t.
802 (find_operand_rank): Return long, -1 if not found. Declare as inline.
803 (insert_operand_rank): Accept long.
804 (operand_entry_hash, operand_entry_eq): Remove.
805 (get_rank): Return long. Adjust for changes above.
806 (init_reassoc): Change rank type to long. Adjust creation of bb_rank
808 (fini_reassoc): Delete operand_rank with pointer_map_destroy.
810 * tree-ssa-structalias.c (vi_for_tree): Change to pointer_map.
811 (struct tree_vi, tree_vi_t, tree_vi_hash, tree_vi_eq): Delete.
812 (insert_vi_for_tree): Rewrite for pointer_map API. Assert argument
814 (lookup_vi_for_tree): Rewrite for pointer_map API. Return varinfo_t
815 directly since it cannot be NULL.
816 (get_vi_for_tree): Rewrite for pointer_map API.
817 (find_what_p_points_to): Adjust for change to lookup_vi_for_tree.
818 (init_alias_vars): Create vi_for_tree as pointer_map.
819 (delete_points_to_sets): Delete vi_for_tree using pointer_map_destroy.
821 * tree-cfg.c: Don't include hashtab.h.
822 (edge_to_cases): Declare as pointer_map.
823 (struct edge_to_cases_elt, edge_to_cases_hash, edge_to_cases_eq):
825 (edge_to_cases_cleanup): Rewrite as pointer_map_traverse callback.
826 (start_recording_case_labels): Create edge_to_cases as pointer_map.
827 (end_recoding_case_labels): Cleanup edge_to_cases manually before
829 (record_switch_edge): Delete.
830 (get_cases_for_edge): Adjust for pointer_map API, inline
831 record_switch_edge (rewritten for new API), remove goto.
833 2006-02-06 Paolo Bonzini <bonzini@gnu.org>
835 * Makefile.in (tree-nested.o): Add pointer-set.h dependency.
836 * tree-nested.c: Include pointer-set.h.
837 (var_map_elt, var_map_eq, var_map_hash): Delete.
838 (struct nesting_info): Remove GTY marker. Change the two htab_t's
840 (nesting_info_bitmap_obstack): New.
841 (lookup_field_for_decl): Adjust for pointer_map API.
842 (lookup_tramp_for_decl): Adjust for pointer_map API.
843 (get_nonlocal_debug_decl): Adjust for pointer_map API.
844 (get_local_debug_decl): Adjust for pointer_map API.
845 (convert_nl_goto_reference): Adjust for pointer_map API.
846 (convert_nl_goto_receiver): Adjust for pointer_map API.
847 (create_nesting_tree): Create outside GGC space. Create bitmap on
848 the new obstack. Create field_map and var_map as pointer_maps.
849 (free_nesting_tree): Adjust for changes to create_nesting_tree.
851 (lower_nested_functions): Move root here, no need to NULL it.
852 Initialize and release the obstack.
854 2007-02-06 Paolo Bonzini <bonzini@gnu.org>
856 * tree.c (tree_int_map_hash, tree_int_map_eq, tree_int_map_marked_p):
857 Remove prototypes and make them non-static.
858 (struct tree_int_map): Remove.
859 * tree.h (struct tree_int_map): Move here, turning TO into an
861 (tree_int_map_hash, tree_int_map_eq, tree_int_map_marked_p): Declare.
863 * tree.h (TREE_COMPLEXITY): Remove.
864 (struct tree_exp): Remove complexity field.
865 * tree.c (build1_stat): Don't set it.
867 2007-02-06 Dorit Nuzman <dorit@il.ibm.com>
868 Victor Kaplansky <victork@il.ibm.com>
870 * tree-vectorizer.c (vect_is_simple_use): Support induction.
871 (vect_is_simple_reduction): Support reduction with induction as
873 (vect_is_simple_iv_evolution): Fix formatting.
874 * tree-vect-analyze.c (vect_mark_stmts_to_be_vectorized): Fix
875 formatting. Don't mark induction phis for vectorization.
876 (vect_analyze_scalar_cycles): Analyze all inductions, then reductions.
877 * tree-vect-transform.c (get_initial_def_for_induction): New function.
878 (vect_get_vec_def_for_operand): Support induction.
879 (vect_get_vec_def_for_stmt_copy): Fix formatting and add check for
881 (vectorizable_reduction): Support reduction with induction as one of
883 (vectorizable_type_demotion): Use def-type of stmt argument rather
886 * tree-ssa-loop.c (gate_scev_const_prop): Return the value of
887 flag_tree_scev_cprop.
888 * common.opt (tree-scev-cprop): New flag.
890 * tree-vect-transform.c (vect_create_destination_var): Use 'kind' in
891 call to vect_get_new_vect_var.
893 2007-02-06 Ira Rosen <irar@il.ibm.com>
895 * tree-vect-patterns.c (vect_recog_widen_mult_pattern): Check that
897 (vect_pattern_recog_1): Likewise.
899 2007-02-05 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
901 * fold-const.c (negate_expr_p): Handle CONJ_EXPR.
902 (fold_negate_expr): Likewise.
904 2007-02-05 Alexandre Oliva <aoliva@redhat.com>
907 * dwarf2out.c (modified_type_die): Follow DECL_ORIGINAL_TYPE
908 even if cv-qualification is the same.
910 2007-02-05 Geoffrey Keating <geoffk@apple.com>
912 * config/rs6000/darwin-tramp.asm (__trampoline_setup): Call
913 __enable_execute_stack on completion.
915 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
917 * config/i386/athlon.md (athlon_fldxf_k8, athlon_fld_k8,
918 athlon_fstxf_k8, athlon_fst_k8, athlon_fist, athlon_fmov,
919 athlon_fadd_load, athlon_fadd_load_k8, athlon_fadd, athlon_fmul,
920 athlon_fmul_load, athlon_fmul_load_k8, athlon_fsgn,
921 athlon_fdiv_load, athlon_fdiv_load_k8, athlon_fdiv_k8,
922 athlon_fpspc_load, athlon_fpspc, athlon_fcmov_load,
923 athlon_fcmov_load_k8, athlon_fcmov_k8, athlon_fcomi_load_k8,
924 athlon_fcomi, athlon_fcom_load_k8, athlon_fcom): Added amdfam10.
926 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
928 * config/i386/i386.md (x86_sahf_1, cmpfp_i_mixed, cmpfp_i_sse,
929 cmpfp_i_i387, cmpfp_iu_mixed, cmpfp_iu_sse, cmpfp_iu_387,
930 swapsi, swaphi_1, swapqi_1, swapdi_rex64, fix_truncsfdi_sse,
931 fix_truncdfdi_sse, fix_truncsfsi_sse, fix_truncdfsi_sse,
932 x86_fldcw_1, floatsisf2_mixed, floatsisf2_sse, floatdisf2_mixed,
933 floatdisf2_sse, floatsidf2_mixed, floatsidf2_sse,
934 floatdidf2_mixed, floatdidf2_sse, muldi3_1_rex64, mulsi3_1,
935 mulsi3_1_zext, mulhi3_1, mulqi3_1, umulqihi3_1, mulqihi3_insn,
936 umulditi3_insn, umulsidi3_insn, mulditi3_insn, mulsidi3_insn,
937 umuldi3_highpart_rex64, umulsi3_highpart_insn,
938 umulsi3_highpart_zext, smuldi3_highpart_rex64,
939 smulsi3_highpart_insn, smulsi3_highpart_zext, x86_64_shld,
940 x86_shld_1, x86_64_shrd, sqrtsf2_mixed, sqrtsf2_sse,
941 sqrtsf2_i387, sqrtdf2_mixed, sqrtdf2_sse, sqrtdf2_i387,
942 sqrtextendsfdf2_i387, sqrtxf2, sqrtextendsfxf2_i387,
943 sqrtextenddfxf2_i387): Added amdfam10_decode.
945 * config/i386/athlon.md (athlon_idirect_amdfam10,
946 athlon_ivector_amdfam10, athlon_idirect_load_amdfam10,
947 athlon_ivector_load_amdfam10, athlon_idirect_both_amdfam10,
948 athlon_ivector_both_amdfam10, athlon_idirect_store_amdfam10,
949 athlon_ivector_store_amdfam10): New define_insn_reservation.
950 (athlon_idirect_loadmov, athlon_idirect_movstore): Added
953 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
955 * config/i386/athlon.md (athlon_call_amdfam10,
956 athlon_pop_amdfam10, athlon_lea_amdfam10): New
957 define_insn_reservation.
958 (athlon_branch, athlon_push, athlon_leave_k8, athlon_imul_k8,
959 athlon_imul_k8_DI, athlon_imul_mem_k8, athlon_imul_mem_k8_DI,
960 athlon_idiv, athlon_idiv_mem, athlon_str): Added amdfam10.
962 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
964 * config/i386/athlon.md (athlon_sseld_amdfam10,
965 athlon_mmxld_amdfam10, athlon_ssest_amdfam10,
966 athlon_mmxssest_short_amdfam10): New define_insn_reservation.
968 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
970 * config/i386/athlon.md (athlon_sseins_amdfam10): New
971 define_insn_reservation.
972 * config/i386/i386.md (sseins): Added sseins to define_attr type
973 and define_attr unit.
974 * config/i386/sse.md: Set type attribute to sseins for insertq
977 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
979 * config/i386/athlon.md (sselog_load_amdfam10, sselog_amdfam10,
980 ssecmpvector_load_amdfam10, ssecmpvector_amdfam10,
981 ssecomi_load_amdfam10, ssecomi_amdfam10,
982 sseaddvector_load_amdfam10, sseaddvector_amdfam10): New
983 define_insn_reservation.
984 (ssecmp_load_k8, ssecmp, sseadd_load_k8, seadd): Added amdfam10.
986 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
988 * config/i386/athlon.md (cvtss2sd_load_amdfam10,
989 cvtss2sd_amdfam10, cvtps2pd_load_amdfam10, cvtps2pd_amdfam10,
990 cvtsi2sd_load_amdfam10, cvtsi2ss_load_amdfam10,
991 cvtsi2sd_amdfam10, cvtsi2ss_amdfam10, cvtsd2ss_load_amdfam10,
992 cvtsd2ss_amdfam10, cvtpd2ps_load_amdfam10, cvtpd2ps_amdfam10,
993 cvtsX2si_load_amdfam10, cvtsX2si_amdfam10): New
994 define_insn_reservation.
996 * config/i386/sse.md (cvtsi2ss, cvtsi2ssq, cvtss2si,
997 cvtss2siq, cvttss2si, cvttss2siq, cvtsi2sd, cvtsi2sdq,
998 cvtsd2si, cvtsd2siq, cvttsd2si, cvttsd2siq,
999 cvtpd2dq, cvttpd2dq, cvtsd2ss, cvtss2sd,
1000 cvtpd2ps, cvtps2pd): Added amdfam10_decode attribute.
1002 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
1004 * config/i386/athlon.md (athlon_ssedivvector_amdfam10,
1005 athlon_ssedivvector_load_amdfam10, athlon_ssemulvector_amdfam10,
1006 athlon_ssemulvector_load_amdfam10): New define_insn_reservation.
1007 (athlon_ssediv, athlon_ssediv_load_k8, athlon_ssemul,
1008 athlon_ssemul_load_k8): Added amdfam10.
1010 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
1012 * config/i386/i386.h (TARGET_SSE_UNALIGNED_MOVE_OPTIMAL): New macro.
1013 (x86_sse_unaligned_move_optimal): New variable.
1015 * config/i386/i386.c (x86_sse_unaligned_move_optimal): Enable for
1017 (ix86_expand_vector_move_misalign): Add code to generate movupd/movups
1018 for unaligned vector SSE double/single precision loads for AMDFAM10.
1020 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
1022 * config/i386/i386.h (TARGET_AMDFAM10): New macro.
1023 (TARGET_CPU_CPP_BUILTINS): Add code for amdfam10.
1024 Define TARGET_CPU_DEFAULT_amdfam10.
1025 (TARGET_CPU_DEFAULT_NAMES): Add amdfam10.
1026 (processor_type): Add PROCESSOR_AMDFAM10.
1028 * config/i386/i386.md: Add amdfam10 as a new cpu attribute to match
1029 processor_type in config/i386/i386.h.
1030 Enable imul peepholes for TARGET_AMDFAM10.
1032 * config.gcc: Add support for --with-cpu option for amdfam10.
1034 * config/i386/i386.c (amdfam10_cost): New variable.
1035 (m_AMDFAM10): New macro.
1036 (m_ATHLON_K8_AMDFAM10): New macro.
1037 (x86_use_leave, x86_push_memory, x86_movx, x86_unroll_strlen,
1038 x86_cmove, x86_3dnow_a, x86_deep_branch, x86_use_simode_fiop,
1039 x86_promote_QImode, x86_integer_DFmode_moves,
1040 x86_partial_reg_dependency, x86_memory_mismatch_stall,
1041 x86_accumulate_outgoing_args, x86_arch_always_fancy_math_387,
1042 x86_sse_partial_reg_dependency, x86_sse_typeless_stores,
1043 x86_use_ffreep, x86_use_incdec, x86_four_jump_limit,
1044 x86_schedule, x86_use_bt, x86_cmpxchg16b, x86_pad_returns):
1045 Enable/disable for amdfam10.
1046 (override_options): Add amdfam10_cost to processor_target_table.
1047 Set up PROCESSOR_AMDFAM10 for amdfam10 entry in
1048 processor_alias_table.
1049 (ix86_issue_rate): Add PROCESSOR_AMDFAM10.
1050 (ix86_adjust_cost): Add code for amdfam10.
1052 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
1054 * config/i386/i386.opt: Add new Advanced Bit Manipulation (-mabm)
1055 instruction set feature flag. Add new (-mpopcnt) flag for popcnt
1056 instruction. Add new SSE4A (-msse4a) instruction set feature flag.
1057 * config/i386/i386.h: Add builtin definition for SSE4A.
1058 * config/i386/i386.md: Add support for ABM instructions
1060 * config/i386/sse.md: Add support for SSE4A instructions
1061 (movntss, movntsd, extrq, insertq).
1062 * config/i386/i386.c: Add support for ABM and SSE4A builtins.
1063 Add -march=amdfam10 flag.
1064 * config/i386/ammintrin.h: Add support for SSE4A intrinsics.
1065 * doc/invoke.texi: Add documentation on flags for sse4a, abm, popcnt
1067 * doc/extend.texi: Add documentation for SSE4A builtins.
1069 2007-02-05 Bob Wilson <bob.wilson@acm.org>
1071 * config/xtensa/xtensa.c (constantpool_mem_p): Skip over SUBREGs.
1073 2007-02-05 Richard Guenther <rguenther@suse.de>
1075 * tree-vectorizer.h (vectorizable_function): Add argument type
1076 argument, change return type.
1077 * tree-vect-patterns.c (vect_recog_pow_pattern): Adjust caller.
1078 * tree-vect-transform.c (vectorizable_function): Handle extra
1079 argument, return vectorized function decl.
1080 (build_vectorized_function_call): Remove.
1081 (vectorizable_call): Handle calls with result and argument types
1082 differing. Handle loop vectorization factor correctly.
1083 * targhooks.c (default_builtin_vectorized_function): Adjust for
1085 * targhooks.h (default_builtin_vectorized_function): Likewise.
1086 * target.h (builtin_vectorized_function): Add argument type
1088 * config/i386/i386.c (ix86_builtin_vectorized_function): Handle
1089 extra argument, allow vectorizing of lrintf.
1090 * doc/tm.texi (TARGET_VECTORIZE_BUILTIN_VECTORIZED_FUNCTION): Adjust
1091 documentation of target hook.
1093 2007-02-05 Hans-Peter Nilsson <hp@axis.com>
1096 * config/cris/cris.md ("*andsi_movu", "*andsi_clear", "*andhi_movu")
1097 ("*andhi_clear", andu (casesi+45)): For size-changed operand where
1098 memory is allowed, require !side_effects_p, not just !MEM_VOLATILE_P.
1100 2007-02-05 Roger Sayle <roger@eyesopen.com>
1102 * fold-const.c (fold_unary) <REAL_PART>: Test for availability of
1103 BUILT_IN_COS before simplifying REAL_PART(CEXPI)) to COS.
1104 <IMAG_PART>: Likewise, check for availability of BUILT_IN_SIN.
1105 * builtins.c (fold_builtin_sincos): Check for TARGET_C99_FUNCTIONS
1106 before canonicalizing sincos to cexpi.
1107 (fold_builtin_cexp): Likewise, for canonicalizing cexp to cexpi.
1109 2007-02-05 Roger Sayle <roger@eyesopen.com>
1111 * config/alpha/alpha.c (alpha_add_builtins): New Helper function.
1112 Set TREE_READONLY and TREE_NOTHROW directly, not via attributes.
1113 (alpha_init_builtins): Use alpha_add_builtins to process tables.
1115 2007-02-05 Roger Sayle <roger@eyesopen.com>
1117 * mips-tfile.c (initialize_init_file): Correct endianness test.
1119 2007-02-05 Kazu Hirata <kazu@codesourcery.com>
1121 * config/m68k/m68k.md (pushdi-1, pushdi, movsi+1): Don't use
1124 2007-02-05 Richard Sandiford <richard@codesourcery.com>
1126 * dwarf2out.c (dwarf2out_frame_debug_expr): Record the register
1127 saves in a PARALLEL before the register assignments.
1129 2007-02-05 Richard Sandiford <richard@codesourcery.com>
1131 * doc/tm.texi (DWARF_ALT_FRAME_RETURN_COLUMN): Do not require
1132 DWARF_FRAME_RETURN_COLUMN to be a general register.
1133 * dwarf2out.c (init_return_column_size): New function, split from...
1134 (expand_builtin_init_dwarf_reg_sizes): ...here. Allow both
1135 DWARF_FRAME_RETURN_COLUMN and DWARF_ALT_FRAME_RETURN_COLUMN
1136 to be nongeneral registers.
1137 * config/m68k/m68k.h (DWARF_FRAME_REGNUM): Only map FP and
1139 (DWARF_FRAME_REGISTERS, DWARF_FRAME_RETURN_COLUMN): Define.
1140 (DWARF_ALT_FRAME_RETURN_COLUMN): Define.
1142 2007-02-04 Zdenek Dvorak <dvorakz@suse.cz>
1144 * cfgcleanup.c (try_optimize_cfg): Avoid removing ENTRY_BLOCK_PTR.
1146 2007-02-04 Zdenek Dvorak <dvorakz@suse.cz>
1148 * cfgloopmanip.c (loop_delete_branch_edge): Removed.
1149 (remove_path): Use can_remove_branch_p and remove_branch instead
1150 of loop_delete_branch_edge.
1151 * tree-ssa-loop-manip.c (scale_dominated_blocks_in_loop): New function.
1152 (tree_transform_and_unroll_loop): Remove dead branches immediately.
1153 Update profile using scale_dominated_blocks_in_loop.
1154 * cfghooks.c (can_remove_branch_p, remove_branch): New functions.
1155 * cfghooks.h (struct cfg_hooks): Add can_remove_branch_p.
1156 (can_remove_branch_p, remove_branch): Declare.
1157 * tree-cfg.c (tree_can_remove_branch_p): New function.
1158 (tree_cfg_hooks): Add tree_can_remove_branch_p.
1159 * cfgrtl.c (rtl_can_remove_branch_p): New function.
1160 (rtl_cfg_hooks, cfg_layout_rtl_cfg_hook): Add rtl_can_remove_branch_p.
1162 2007-02-05 Jan Hubicka <jh@suse.cz>
1165 * ipa-inline.c (cgraph_clone_inlined_nodes): When there are unanalyzed
1166 nodes in cgraph, don't remove offline copy of the function.
1168 2007-02-04 Jan Hubicka <jh@suse.cz>
1170 * tree-sra.c (sra_walk_expr): Add linebreaks. BITFIELD_REFs into
1171 vectors might cause maybe_lookup_element_for_expr to be called
1172 on non-sra-candidate.
1174 2007-02-04 Kazu Hirata <kazu@codesourcery.com>
1176 * config/bfin/bfin-modes.def, config/bfin/bfin.c,
1177 config/bfin/bfin.md, config/bfin/predicates.md: Follow
1178 spelling conventions.
1180 2007-02-04 Richard Guenther <rguenther@suse.de>
1183 * fold-const.c (try_move_mult_to_index): Make sure to not
1184 overflow one dimension of a multi-dimensional array access.
1186 2007-02-04 Jan Hubicka <jh@suse.cz>
1188 * passes.c (init_optimization_passes): Reindent.
1190 2007-02-04 Jan Hubicka <jh@suse.cz>
1191 Eric Botcazou <ebotcazou@adacore.com>
1193 * tree-optimize.c (has_abnormal_outgoing_edge_p): Move to...
1194 (execute_fixup_cfg): Break out the abnormal goto code.
1195 * tree-inline.c (has_abnormal_outgoing_edge_p): ...here.
1196 (make_nonlocal_label_edges): Move here from execute_fixup_cfg.
1197 (optimize_inline_calls): Call make_nonlocal_label_edges.
1199 2007-02-04 Jan Hubicka <jh@suse.cz>
1201 * tree-ssa-copyrename.c (copy_rename_partition_coalesce): Return
1202 true when something was changed.
1203 (rename_ssa_copies): When something was changed, do
1204 TODO_remove_unused_locals.
1205 * tree-ssa-forwprop.c (tree_ssa_forward_propagate_single_use_value):
1206 add TODO_remove_unused_locals when instruction was removed.
1208 2007-02-04 Jan Hubicka <jh@suse.cz>
1210 * ipa-inline.c (try_inline): Improve debug output; work on already
1212 (cgraph_decide_inlining_incrementally): Indent; improve debug output;
1213 call try_inline for already inlined edges too when flattening;
1214 inline also functions that make callee growth but overall unit size
1217 2007-02-04 Kazu Hirata <kazu@codesourcery.com>
1219 * config/m32c/bitops.md, config/m32c/jump.md,
1220 config/m32c/m32c.c, config/m32c/m32c.h, config/m32r/m32r.c,
1221 config/m32r/m32r.h, config/m32r/m32r.md,
1222 config/m32r/predicates.md, config/m68hc11/larith.asm,
1223 config/m68hc11/m68hc11.c, config/m68hc11/m68hc11.h,
1224 config/m68k/m68k.h, config/mcore/mcore.md, config/mips/4k.md,
1225 config/mips/mips-protos.h, config/mips/mips.c,
1226 config/mips/mips.h, config/mips/mips.md, config/mips/mips16.S,
1227 config/mn10300/mn10300.h, config/mn10300/predicates.md,
1228 config/mt/mt.c, config/mt/mt.h, config/mt/mt.md: Follow
1229 spelling conventions.
1231 * config/v850/v850.c, config/v850/v850.h, config/v850/v850.md:
1232 Follow spelling conventions.
1234 2007-02-03 Douglas Gregor <doug.gregor@gmail.com>
1236 * c-opts.c (c_common_post_options): If C++0x mode is enabled, don't
1237 warn about C++0x compatibility.
1239 2007-02-04 Kazu Hirata <kazu@codesourcery.com>
1241 * config/h8300/h8300.c, config/h8300/h8300.h,
1242 config/h8300/h8300.md: Follow spelling conventions.
1244 2007-02-03 Uros Bizjak <ubizjak@gmail.com>
1247 * combine.c (try_combine): Do not substitute source operand
1248 with constants wider than 2 * HOST_BITS_PER_WIDE_INT.
1250 2007-02-03 Jan Hubicka <jh@suse.cz>
1252 PR gcov-profile/30650
1253 * value-prof.c (stringop_block_profile): Fix handling of size counter;
1254 do not divide by zero for never executed counters.
1255 (tree_find_values_to_profile): Fix counters.
1256 * gcov-ui.h (GCOV_COUNTER_AVERAGE, GCOV_COUNTER_IOR): Fix comments.
1258 2007-02-03 Ian Lance Taylor <iant@google.com>
1260 * lower-subreg.c (simple_move_operand): New static function,
1261 broken out of simple_move. Reject LABEL_REF, SYMBOL_REF, and HIGH
1263 (simple_move): Call simple_move_operand.
1264 (find_decomposable_subregs): Add special handling of MEMs.
1265 (can_decompose_p): Rename from cannot_decompose_p. Reverse
1266 meaning of return value. If we see a hard register, test whether
1267 it can store a word_mode value. Change all callers.
1269 2007-02-03 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
1271 * pa.md (addvdi3, addvsi3, subvdi3, subvsi3, negvdi2, negvsi2): New
1272 ftrapv insns and expanders.
1273 (subdi3): Change define_expand operand 1 to arith11_operand, and
1274 operand 2 to reg_or_0_operand. Change constraints of 64-bit insn
1275 pattern to handle reg_or_0 operands. Revise 32-bit insn pattern to
1276 handle 11-bit constants and reg_or_0 operands in operands 1 and 2,
1280 * varasm.c (notice_global_symbol): Treat global objects as weak when
1283 2007-02-03 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1285 * emit-rtl.c (dconstpi): Delete.
1287 (init_emit_once): Delete dconstpi and init dconstsqrt2.
1288 * real.h (dconstpi): Delete.
1290 * builtins.c (fold_builtin_cabs): Use dconstsqrt2.
1291 (fold_builtin_hypot): Likewise.
1293 2007-02-03 Tom Tromey <tromey@redhat.com>
1296 * gcc.c (cpp_unique_options): Any of -ggdb3, -gstabs3,
1297 -gcoff3, -gxcoff3, -gvms3 implies -dD.
1299 2007-02-03 Kazu Hirata <kazu@codesourcery.com>
1301 * c-decl.c, config/avr/avr.c, config/avr/avr.h,
1302 config/m68k/m68k.c, config/m68k/netbsd-elf.h,
1303 config/mn10300/mn10300.c, config/pdp11/pdp11.h,
1304 config/rs6000/cell.md, config/rs6000/darwin.h,
1305 config/sh/sh.md, config/sh/sh4-300.md, config/spu/spu.c,
1306 config/spu/spu.md, cselib.c, expr.c, haifa-sched.c, hwint.h,
1307 jump.c, reload.c, sched-deps.c, sched-int.h, tree-inline.c,
1308 tree-profile.c, tree-ssa-live.h, tree-vrp.c: Fix comment
1309 typos. Follow spelling conventions.
1310 * doc/invoke.texi: Follow spelling conventions.
1312 2007-02-03 Roger Sayle <roger@eyesopen.com>
1314 * simplify-rtx.c (simplify_relational_operation_1): Implement some
1315 canonicalization transformations that attempt to simplify integer
1316 constant comparisons to become comparisons against zero.
1318 2007-02-03 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1320 * builtins.c (fold_builtin_cabs): Fold cabs(x+xi) into
1322 * fold-const.c (fold_binary): Fix comment typos. Fold complex
1323 (x,0)-(0,y) into (x,-y). Likewise (0,y)-(x,0) into (-x,y).
1325 2007-02-02 Mike Stump <mrs@apple.com>
1327 * config/darwin9.h (DARWIN_LINKER_GENERATES_ISLANDS): Add.
1328 * config/rs6000/rs6000.c (DARWIN_GENERATE_ISLANDS): Add.
1329 (output_call): Use DARWIN_GENERATE_ISLANDS to decide when to
1330 generate a branch island.
1332 2007-02-02 Bob Wilson <bob.wilson@acm.org>
1334 * config/xtensa/xtensa.c (smalloffset_mem_p): Use BASE_REG_P.
1335 (xtensa_legitimate_address_p): New.
1336 (xtensa_legitimize_address): New.
1337 (xtensa_output_addr_const_extra): New.
1338 * config/xtensa/xtensa.h (REG_OK_STRICT_FLAG): Define.
1340 (REG_OK_FOR_BASE_P): Use BASE_REG_P.
1341 (GO_IF_LEGITIMATE_ADDRESS): Move code to xtensa_legitimate_address_p.
1342 (LEGITIMIZE_ADDRESS): Move code to xtensa_legitimize_address.
1343 (OUTPUT_ADDR_CONST_EXTRA): Move code to xtensa_output_addr_const_extra.
1344 * config/xtensa/xtensa-protos.h (xtensa_legitimate_address_p): New.
1345 (xtensa_legitimize_address): New.
1346 (xtensa_output_addr_const_extra): New.
1348 2007-02-02 Steve Ellcey <sje@cup.hp.com>
1350 * config/ia64/ia64.c (ia64_print_operand): Fix compare strings.
1352 2007-02-02 Ian Lance Taylor <iant@google.com>
1354 * expmed.c (expand_divmod): Add comment.
1356 2007-02-02 Kazu Hirata <kazu@codesourcery.com>
1358 * emit-rtl.c (renumber_insns): Remove.
1359 * flags.h: Remove the extern for flag_renumber_insns.
1360 * rtl.h: Remove the prototype for renumber_insns.
1361 * toplev.c (flag_renumber_insns): Remove.
1363 2007-02-02 Hui-May Chang <hm.chang@apple.com>
1365 Revert for x86 darwin:
1366 2005-06-19 Uros Bizjak <uros@kss-loka.si>
1368 * config/i386/i386.c (ix86_function_arg_regno_p): Put back the
1369 code before the following patch under TARGET_MACHO.
1370 (ix86_function_value_regno_p): Likewise.
1372 2007-02-02 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1374 * fold-const.c (negate_expr_p, fold_negate_expr): Handle
1377 2007-02-02 Maxim Kuvyrkov <mkuvyrkov@ispras.ru>
1380 * config/ia64/ia64.c (ia64_speculate_insn): Restrict to memory loads to
1381 general or fp registers. Add comments.
1382 * config/ia64/ia64.md (reg_pred_prefix): Add comment.
1384 2007-02-02 Paolo Bonzini <bonzini@gnu.org>
1386 * pointer-set.c (insert_aux): Only return insertion slot.
1387 (pointer_set_insert): Adjust.
1388 (pointer_set_traverse, struct pointer_map_t, pointer_map_create,
1389 pointer_map_destroy, pointer_map_insert, pointer_map_contains,
1390 pointer_map_traverse): New.
1391 * pointer-set.h (pointer_set_traverse, struct pointer_map_t,
1392 pointer_map_create, pointer_map_destroy, pointer_map_insert,
1393 pointer_map_contains, pointer_map_traverse): Declare.
1395 2007-02-02 Jakub Jelinek <jakub@redhat.com>
1398 * builtins.c (fold_builtin_sprintf): Do not attempt to optimize
1399 sprintf (str, "%s"). Do not optimize sprintf (str, "nopercent", p++).
1401 2007-02-02 Maxim Kuvyrkov <mkuvyrkov@ispras.ru>
1403 * sched-int.h (ds_to_dk, dk_to_ds): Declare functions.
1405 (struct _dep): New type.
1406 (dep_t): New typedef.
1407 (DEP_PRO, DEP_CON, DEP_KIND): New access macros.
1408 (DEP_STATUS): New access macro. The macro with the same name was
1409 renamed to DEP_LINK_STATUS.
1410 (dep_init): Declare function
1412 (struct _dep_link): New type.
1413 (dep_link_t): New typedef.
1414 (DEP_LINK_NODE, DEP_LINK_NEXT, DEP_LINK_PREV_NEXTP): New access macros.
1415 (DEP_LINK_DEP, DEP_LINK_PRO, DEP_LINK_CON, DEP_LINK_KIND): New macros.
1416 (DEP_LINK_STATUS): New macro.
1417 (debug_dep_links): New debug function.
1419 (struct _deps_list): New type.
1420 (deps_list_t): New typedef.
1421 (DEPS_LIST_FIRST): New access macro.
1422 (FOR_EACH_DEP_LINK): New cycle macro.
1423 (create_deps_list, free_deps_list, delete_deps_list): Declare
1425 (deps_list_empty_p, debug_deps_list, add_back_dep_to_deps_list): Ditto.
1426 (find_link_by_pro_in_deps_list, find_link_by_con_in_deps_list): Ditto.
1427 (copy_deps_list_change_con): Ditto.
1429 (move_dep_link): Declare function.
1431 (struct _dep_node): New type.
1432 (dep_node_t): New typedef.
1433 (DEP_NODE_BACK, DEP_NODE_DEP, DEP_NODE_FORW): New access macros.
1435 (struct haifa_insn_data.back_deps): New field to hold backward
1436 dependencies of the insn.
1437 (struct haifa_insn_data.depend): Rename to forw_deps. Change its type
1439 (struct haifa_insn_data.resolved_deps): Rename to resolved_back_deps.
1440 Change its type to deps_list_t.
1441 (INSN_BACK_DEPS): New access macro to use instead of LOG_LINKS.
1442 (INSN_DEPEND): Rename to INSN_FORW_DEPS.
1443 (RESOLVED_DEPS): Rename to INSN_RESOLVED_BACK_DEPS.
1445 (INSN_COST): Move to haifa-sched.c. Use insn_cost () instead.
1447 (DEP_STATUS): Rename to DEP_LINK_STATUS. Fix typo in the comment.
1449 (add_forw_dep, delete_back_forw_dep, insn_cost): Update declaration and
1451 (dep_cost): Declare.
1453 * sched-deps.c (CHECK): New macro to (en/dis)able sanity checks.
1454 (ds_to_dk, dk_to_ds): New functions.
1456 (init_dep_1): New static function.
1457 (init_dep): New function.
1458 (copy_dep): New static function.
1460 (dep_link_consistent_p, attach_dep_link, add_to_deps_list): New static
1462 (detach_dep_link): New static function.
1463 (move_dep_link): New function.
1465 (dep_links_consistent_p, dump_dep_links): New static functions.
1466 (debug_dep_links): New debugging function.
1468 (deps_obstack, dl_obstack, dn_obstack): New static variables.
1470 (alloc_deps_list, init_deps_list): New static functions.
1471 (create_deps_list): New function.
1472 (clear_deps_list): New static function.
1473 (free_deps_list, delete_deps_list, deps_list_empty_p): New functions.
1474 (deps_list_consistent_p, dump_deps_list): New static functions.
1475 (debug_deps_list): New function.
1476 (add_back_dep_to_deps_list, find_link_by_pro_in_deps_list): New
1478 (find_link_by_con_in_deps_list, copy_deps_list_change_con): Ditto.
1480 (maybe_add_or_update_back_dep_1, add_or_update_back_dep_1): Update to
1481 use new scheduler dependencies lists.
1482 (add_back_dep, delete_all_dependences, fixup_sched_groups): Ditto.
1483 (sched_analyze): Ditto. Initialize dependencies lists.
1484 (add_forw_dep, compute_forward_dependences): Update to use new
1485 scheduler dependencies lists.
1487 (init_dependency_caches): Init deps_obstack.
1488 (free_dependency_caches): Free deps_obstack.
1490 (adjust_add_sorted_back_dep, adjust_back_add_forw_dep): Update to use
1491 new scheduler dependencies lists.
1492 (delete_forw_dep, add_or_update_back_forw_dep): Ditto.
1493 (add_back_forw_dep, delete_back_forw_dep): Ditto.
1495 * sched-rgn.c (set_spec_fed, find_conditional_protection, is_pfree):
1496 Update to use new scheduler dependencies lists.
1497 (is_conditionally_protected, is_prisky, add_branch_dependences): Ditto.
1498 (debug_dependencies): Ditto.
1499 (schedule_region): Update comments.
1501 * sched-ebb.c (earliest_block_with_similiar_load): Update to use new
1502 scheduler dependencies lists.
1503 (schedule_ebb): Update comments.
1505 * rtl.def (DEPS_LIST): Remove.
1507 * lists.c (unused_deps_list): Remove.
1508 (free_list): Update assertions.
1510 (alloc_DEPS_LIST, free_DEPS_LIST_list, free_DEPS_LIST_node): Remove.
1511 (remove_free_DEPS_LIST_elem, copy_DEPS_LIST_list): Ditto.
1513 * rtl.h (free_DEPS_LIST_list, alloc_DEPS_LIST): Remove declarations.
1514 (remove_free_DEPS_LIST_elem, copy_DEPS_LIST_list): Ditto.
1516 * haifa-sched.c (comments): Update.
1517 (insn_cost1): Remove. Inline the code into insn_cost ().
1518 (insn_cost): Update to use new scheduler dependencies lists. Move
1519 processing of the dependency cost to dep_cost ().
1520 (dep_cost): New function. Use it instead of insn_cost () when
1521 evaluating cost of the dependency. Use compatible interface to
1522 interact with the target.
1523 (priority): Update to use new scheduler dependencies lists.
1524 (rank_for_schedule): Ditto. Optimize heuristic that prefers the insn
1525 with greater number of insns that depend on the insn.
1526 (schedule_insn): Update to use new scheduler dependencies lists. Add
1527 code to free backward dependencies lists. Inline and optimize code
1528 from resolve_dep () - see PR28071.
1529 (ok_for_early_queue_removal): Update to use new scheduler dependencies
1530 lists. Update call to targetm.sched.is_costly_dependence hook.
1532 (fix_inter_tick, try_ready, fix_tick_ready): Update to use new
1533 scheduler dependencies lists.
1535 (resolve_dep): Remove. Move the logic to schedule_insn ().
1536 (init_h_i_d): Initialize dependencies lists.
1538 (process_insn_depend_be_in_spec): Rename to
1539 process_insn_forw_deps_be_in_spec. Update to use new scheduler
1541 (add_to_speculative_block, create_check_block_twin, fix_recovery_deps):
1542 Update to use new scheduler dependencies lists.
1543 (clear_priorities, calc_priorities, add_jump_dependencies): Ditto.
1545 * ddg.c (create_ddg_dependence, create_ddg_dep_no_link): Update to use
1546 new scheduler dependencies lists.
1547 (build_intra_loop_deps): Ditto.
1549 * target.h (struct _dep): Declare to use in
1550 gcc_target.sched.is_costly_dependence.
1551 (struct gcc_target.sched.adjust_cost): Fix typo.
1552 (struct gcc_target.sched.is_costly_dependence): Change signature to use
1553 single dep_t parameter instead of an equivalent triad.
1554 (struct gcc_target.sched.adjust_cost_2): Remove.
1556 * target-def.h (TARGET_SCHED_ADJUST_COST_2): Remove.
1558 * reg-notes.def (DEP_TRUE, DEP_OUTPUT, DEP_ANTI): Update comments.
1560 * doc/tm.texi (TARGET_SCHED_IS_COSTLY_DEPENDENCE): Update
1562 (TARGET_SCHED_ADJUST_COST_2): Remove documentation.
1564 * doc/rtl.texi (LOG_LINKS): Remove part about instruction scheduler.
1565 (REG_DEP_TRUE): Document.
1567 * config/ia64/ia64.c (ia64_adjust_cost_2): Rename to ia64_adjust_cost.
1568 Change signature to correspond to the targetm.sched.adjust_cost hook.
1569 Update use in TARGET_SCHED_ADJUST_COST_2.
1570 (TARGET_SCHED_ADJUST_COST_2): Rename to TARGET_SCHED_ADJUST_COST.
1571 (ia64_dependencies_evaluation_hook, ia64_dfa_new_cycle): Update to use
1572 new scheduler dependencies lists.
1573 (ia64_gen_check): Ditto.
1575 * config/mips/mips.c (vr4130_swap_insns_p): Update to use new scheduler
1578 * config/rs6000/rs6000.c (rs6000_is_costly_dependence): Change
1579 signature to correspond to the targetm.sched.is_costly_dependence hook.
1580 (is_costly_group): Update to use new scheduler dependencies lists.
1582 * config/spu/spu.c (spu_sched_adjust_cost): Use insn_cost () function
1583 instead of INSN_COST () macro.
1585 2007-02-01 Ian Lance Taylor <iant@google.com>
1587 * lower-subreg.c (resolve_clobber): Handle a subreg of a concatn.
1589 2007-02-01 Guy Martin <gmsoft@gentoo.org>
1591 * pa.md (tp_load): Correct mfctl instruction syntax.
1593 2007-02-01 Geoffrey Keating <geoffk@apple.com>
1595 * config/rs6000/rs6000.c (rs6000_stack_info): Correct
1596 altivec_padding_size calculation on AIX. Improve comment, add
1597 assert to verify that it's right.
1599 * config/rs6000/darwin.h (MD_UNWIND_SUPPORT): Don't define for
1602 2007-2-01 Seongbae Park <seongbae.park@gmail.com>
1605 * global.c (global_alloc): Add mising initialization of
1608 2007-02-01 Roger Sayle <roger@eyesopen.com>
1610 * alias.c (init_alias_analysis): Correct whitespace.
1611 * bb-reorder.c (fix_edges_for_rarely_executed_code,
1612 partition_hot_cold_basic_blocks): Likewise.
1613 * builtins.c (expand_builtin_printf, expand_builtin_fprintf,
1614 expand_builtin_sprintf, fold_builtin_carg, fold_builtin_sprintf,
1615 maybe_emit_sprintf_chk_warning, fold_builtin_sprintf_chk,
1616 fold_builtin_snprintf_chk, fold_builtin_printf,
1617 fold_builtin_fprintf, do_mpfr_ckconv, do_mpfr_arg1, do_mpfr_arg2,
1618 do_mpfr_arg3, do_mpfr_sincos): Likewise.
1619 * cfgcleanup.c (cleanup_cfg): Likewise.
1620 * cfgexpand.c (tree_expand_cfg): Likewise.
1621 * fold-const.c (fold_binary) <RDIV_EXPR>: Likewise.
1622 * function.c (get_next_funcdef_no): Likewise.
1623 * gengtype.c (main): Likewise.
1624 * genmodes.c (main): Likewise.
1625 * gcse.c (bypass_conditional_jumps, print_ldst_list): Likewise.
1626 * haifa-sched.c (schedule_block, extend_h_i_d): Likewise.
1627 * ifcvt.c (noce_emit_move_insn): Likewise.
1628 * modulo-sched.c (generate_prolog_epilog, sms_schedule_by_order):
1630 * stor-layout.c (get_best_mode): Likewise.
1631 * tree-ssa-loop-niter.c (get_val_for): Likewise.
1632 * tree-ssa-structalias.c (get_varinfo, get_varinfo_fc,
1633 scc_visit, do_ds_constraint, do_complex_constraint, label_visit,
1634 perform_var_substitution, solve_graph): Likewise.
1635 * tree-vrp.c (vrp_finalize): Likewise.
1637 2007-02-01 Ian Lance Taylor <iant@google.com>
1639 * lower-subreg.c (simplify_gen_subreg_concatn): If we ask for the
1640 high part of a paradoxical subreg, return a constant zero.
1642 2007-02-01 Zdenek Dvorak <dvorakz@suse.cz>
1644 * toplev.c (lang_dependent_init): Call init_set_costs.
1645 * loop-init.c (loop_optimizer_init): Do not call init_set_costs.
1647 2007-02-01 Richard Guenther <rguenther@suse.de>
1650 * fold-const.c (fold_negate_expr): Allow negating a
1651 constant if overflow does not change.
1653 2007-02-01 Ralf Wildenhues <Ralf.Wildenhues@gmx.de>
1655 * doc/c-tree.texi (Expression trees): Improve markup.
1656 * doc/tm.texi (Register Classes, Addressing Modes)
1657 (Floating Point): Fix spacing after abbreviations. Fix some
1660 2007-02-01 Ben Elliston <bje@au.ibm.com>
1662 * doc/invoke.texi: Replace "bugfix" with "bug fix" throughout.
1663 * doc/contrib.texi: Likewise.
1664 * doc/install.texi: Likewise.
1666 2007-01-31 Richard Henderson <rth@redhat.com>
1667 Ian Lance Taylor <iant@google.com>
1669 * lower-subreg.c: New file.
1670 * rtl.def (CONCATN): Define.
1671 * passes.c (init_optimization_passes): Add pass_lower_subreg and
1673 * emit-rtl.c (update_reg_offset): New static function, broken out
1674 of gen_rtx_REG_offset.
1675 (gen_rtx_REG_offset): Call update_reg_offset.
1676 (gen_reg_rtx_offset): New function.
1677 * regclass.c: Revert patch of 2006-03-05, restoring
1679 (clear_reg_info_regno): New function.
1680 * dwarf2out.c (concatn_loc_descriptor): New static function.
1681 (loc_descriptor): Handle CONCATN.
1682 * common.opt (fsplit_wide_types): New option.
1683 * opts.c (decode_options): Set flag_split_wide_types when
1685 * timevar.def (TV_LOWER_SUBREG): Define.
1686 * rtl.h (gen_reg_rtx_offset): Declare.
1687 (reg_scan_update): Declare.
1688 * regs.h (clear_reg_info_regno): Declare.
1689 * tree-pass.h (pass_lower_subreg): Declare.
1690 (pass_lower_subreg2): Declare.
1691 * doc/invoke.texi (Option Summary): List -fno-split-wide-types.
1692 (Optimize Options): Add -fsplit-wide-types to -O1 list. Document
1694 * doc/rtl.texi (Regs and Memory): Document concat and concatn.
1695 * Makefile.in (OBJS-common): Add lower-subreg.o.
1696 (lower-subreg.o): New target.
1698 2007-01-31 Kazu Hirata <kazu@codesourcery.com>
1700 * config/sh/sh.h (HAVE_SECONDARY_RELOADS): Remove.
1702 2007-01-31 Anatoly Sokolov <aesok@post.ru>
1705 * config/avr/avr.c (DWARF2_ADDR_SIZE): Define.
1707 2007-01-31 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1710 * builtins.c (fold_builtin_sqrt): Use MPFR for constant args.
1712 2007-01-31 Zdenek Dvorak <dvorakz@suse.cz>
1714 * cfgloop.h: Include vec-prim.h.
1715 (enum li_flags): Remove LI_ONLY_OLD.
1716 (loop_iterator): Changed.
1717 (fel_next, fel_init): Iterate over loop tree.
1718 (FOR_EACH_LOOP_BREAK): New macro.
1719 * loop-unswitch.c (unswitch_loops): Do not pass LI_ONLY_OLD to
1721 * tree-ssa-loop-unswitch.c (tree_ssa_unswitch_loops): Ditto.
1722 * modulo-sched.c (sms_schedule): Ditto.
1723 * tree-vectorizer.c (vectorize_loops): Ditto.
1724 * doc/loop.texi: Update information on loop numbering and behavior of
1725 FOR_EACH_LOOP wrto new loops.
1726 * tree-scalar-evolution.c (compute_overall_effect_of_inner_loop,
1727 add_to_evolution_1): Test nestedness of loops instead of comparing
1729 * tree-chrec.c (chrec_fold_plus_poly_poly,
1730 chrec_fold_multiply_poly_poly, chrec_evaluate,
1731 hide_evolution_in_other_loops_than_loop, chrec_component_in_loop_num,
1732 reset_evolution_in_loop): Ditto.
1733 * Makefile.in (CFGLOOP_H): Add vecprim.h dependency.
1735 2007-01-31 Dirk Mueller <dmueller@suse.de>
1737 * c-common.c (warn_about_parentheses): Separate warning about
1738 un-parenthized sequence of comparison operators from the one
1739 which is supposed to warn about x <= y <= z.
1741 2007-01-31 Uros Bizjak <ubizjak@gmail.com>
1743 * optabs.h (enum optab_index): Add new OTI_isinf.
1744 (isinf_optab): Define corresponding macro.
1745 * optabs.c (init_optabs): Initialize isinf_optab.
1746 * genopinit.c (optabs): Implement isinf_optab using isinf?f2
1748 * builtins.c (mathfn_built_in): Handle BUILT_IN_ISINF{,F,L}.
1749 (expand_builtin_interclass_mathfn): Expand BUILT_IN_ISINF{,F,L}
1751 (expand_builtin): Expand BUILT_IN_ISINF{,F,L} using
1752 expand_builtin_interclass_mathfn.
1753 * reg_stack.c (subst_stack_regs_pat): Handle UNSPEC_FXAM.
1754 * config/i386/i386.md (UNSPEC_FXAM): New constant.
1755 (fxam<mode>2_i387): New insn pattern.
1756 (isinf<mode>2) New expander to implement isinf, isinff and isinfl
1757 built-in functions as x87 inline asm.
1759 2007-01-31 Kazu Hirata <kazu@codesourcery.com>
1761 * gcc/config/arm/unwind-arm.h (_sleb128_t, _uleb128_t): New.
1763 2007-01-30 Eric Christopher <echristo@apple.com>
1765 * config.gcc: Add geode.
1767 2007-01-31 Kazu Hirata <kazu@codesourcery.com>
1769 * cgraphunit.c, config/arm/arm.c, config/m68k/m68k.c,
1770 ipa-inline.c, tree-profile.c, tree-ssa-live.c,
1771 tree-ssa-math-opts.c, tree-ssanames.c, tree-vect-analyze.c,
1772 value-prof.c: Fix comment typos.
1774 2007-01-30 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
1777 * doc/invoke.texi (Wpointer-arith): Document warning.
1779 2007-01-30 Janis Johnson <janis187@us.ibm.com>
1781 * doc/extend.texi (Decimal Floating Types): Remove decfloat.h from
1782 the list of discrepancies from the draft TR.
1784 2007-01-30 Dirk Mueller <dmueller@suse.de>
1787 * doc/invoke.texi (-Wreturn-type): Update description to
1790 2007-01-30 Richard Sandiford <richard@codesourcery.com>
1792 * cfgrtl.c (try_redirect_by_replacing_jump): Check only_sets_cc0_p.
1794 2007-01-30 Uros Bizjak <ubizjak@gmail.com>
1796 * builtins.c (expand_builtin_int_interclass_roundingfn): New function
1797 to handle optabs that operate on floating point input argument and
1798 output to integer output.
1799 (expand_builtin_mathfn) [BUILT_IN_ILOGB]: Move from here ...
1800 (expand_builtin_interclass_mathfn) [BUILT_IN_ILOGB]: ... to here.
1801 (expand_builtin): Expand BUILT_IN_ILOGB{,F,L} using
1802 expand_builtin_interclass_mathfn ().
1803 * config/i386/i386.md (fxtractxf3_i387): Rename from *fxtractxf3_i387.
1805 (ilogbxf2, ilogb<mode>2): New expanders to implement ilogb, ilogbf and
1806 ilogbl built-in functions as x87 intrinsics.
1808 2007-01-30 Richard Guenther <rguenther@suse.de>
1811 * dwarf2out.c (reference_to_unused): Query varpool if the
1812 variable was output.
1814 2007-01-30 Richard Guenther <rguenther@suse.de>
1817 * passes.c (execute_one_pass): Reset in_gimple_form to not
1818 confuse non-unit-at-a-time mode.
1820 2007-01-29 Roger Sayle <roger@eyesopen.com>
1821 Richard Guenther <rguenther@suse.de>
1823 * fold-const.c (round_up): Make HIGH an unsigned HOST_WIDE_INT to
1824 avoid undefined behaviour on overflow. Use force_fit_type_double
1825 to construct the constant with the specified TREE_OVERFLOW.
1827 2007-01-29 Janis Johnson <janis187@us.ibm.com>
1829 * config/dfp-bit.c: Add parameterized support for fp exceptions.
1830 * config/dfp-bit.h: Ditto.
1832 2007-01-29 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
1834 * c-decl.c (pop_scope): Replace warnings with call to
1835 warn_for_unused_label.
1836 * c-common.h (warn_for_unused_label): Declare.
1837 * c-common.c (warn_for_unused_label): Define.
1839 2007-01-29 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
1841 * tree-optimize.c (update_inlined_to_pointers): Delete unused
1844 2007-01-29 Janis Johnson <janis187@us.ibm.com>
1846 * Makefile.in (USER_H): Remove decfloat.h.
1847 * ginclude/decfloat.h: Delete, moving contents to ...
1848 * ginclude/float.h: Add support for decimal floating point,
1849 guarded by __STDC_WANT_DEC_FP__.
1851 2007-01-29 Mike Stump <mrs@apple.com>
1853 * doc/gccint.texi (Top): Rename Loop Representation to Loop
1854 Analysis and Representation to resolve case insensitive conflict.
1855 * doc/loop.texi (Loop Analysis and Representation): Likewise.
1857 2007-01-28 Daniel Berlin <dberlin@dberlin.org>
1859 * tree.h (struct tree_memory_tag): Add aliases member.
1860 (MTAG_ALIASES): New macro.
1861 * tree-ssa-alias.c (alias_bitmap_obstack): New variable.
1862 (add_may_alias): Remove pointer-set. Update for may_aliases being
1864 (mark_aliases_call_clobbered): Update for may_aliases being a
1866 (compute_tag_properties): Ditto.
1867 (create_partition_for): Ditto.
1868 (compute_memory_partitions): Ditto.
1869 (dump_may_aliases_for): Ditto.
1870 (is_aliased_with): Ditto.
1871 (add_may_alias_for_new_tag): Ditto.
1872 (rewrite_alias_set_for): Rewrite for may_aliases being a bitmap.
1873 (compute_is_aliased): New function.
1874 (compute_may_aliases): Call compute_is_aliased).
1875 (init_alias_info): Initialize alias_bitmap_obstack.
1876 (union_alias_set_into): New function.
1877 (compute_flow_sensitive_aliasing): Use union_aliases_into.
1878 (have_common_aliases_p): Rewrite to take two bitmaps and use
1880 (compute_flow_insensitive_aliasing): Stop using pointer-sets.
1882 (finalize_ref_all_pointers): Update for add_may_alias changes.
1883 (new_type_alias): Ditto.
1884 * tree-flow-inline.h (may_aliases): Return a bitmap.
1885 * tree-dfa.c (dump_variable): Check for MTAG_P'ness.
1886 * tree-ssa.c (verify_flow_insensitive_alias_info): Update for
1887 may_aliases being a bitmap.
1888 * tree-flow.h (struct var_ann_d): Remove may_aliases member.
1889 may_aliases now returns a bitmap.
1890 * tree-ssa-structalias.c (merge_smts_into): Update for may_aliases
1892 * tree-ssa-operands.c (add_virtual_operand): Update for
1893 may_aliases being a bitmap.
1895 2007-01-29 Daniel Berlin <dberlin@dberlin.org>
1897 PR tree-optimization/30630
1898 * tree-ssa-structalias.c (do_complex_constraint): Mark correct
1899 variable as changed.
1901 2007-01-29 Simon Martin <simartin@users.sourceforge.net>
1904 * gimplify.c (gimplify_target_expr): Make sure that the TARGET_EXPR is
1905 expanded only once even if an error occurs.
1907 2007-01-29 Ben Elliston <bje@au.ibm.com>
1909 * gcov-io.h (__gcov_indirect_call_profiler): Declare.
1910 (__gcov_average_profiler): Likewise.
1911 (__gcov_ior_profiler): Likewise.
1912 (__gcov_merge_ior): Likewise.
1914 2007-01-28 Jan Hubicka <jh@suse.cz>
1916 * builtins.c (expand_builtin_memset): Fix typo in my last patch.
1917 * value-prof.c (stringop_block_profile): Likewise.
1919 2007-01-28 Jan Hubicka <jh@suse.cz>
1921 * expr.c (emit_block_move_via_movmem, emit_block_move_via_libcall): Add
1922 variant handling histograms; add wrapper.
1923 (clear_storage_via_libcall): Export.
1924 (emit_block_move_hints): Break out from ...; add histograms.
1925 (emit_block_move): ... this one.
1926 (clear_storage_hints): Break out from ...; add histograms.
1927 (clear_storage): ... this one.
1928 (set_storage_via_memset): Handle histogram.
1929 * expr.h (emit_block_move_via_libcall, emit_block_move_hints): Declare.
1930 (clear_storage_hints, clear_storage_via_libcall): Declare.
1931 (set_storage_via_setmem): Update prototype.
1932 * doc/md.texi (movmem, setmem): Document new arguments.
1934 * value-prof.c (dump_histogram_value, tree_find_values_to_profile): Add
1936 (stringop_block_profile): New global function.
1937 (tree_stringops_values_to_profile): Profile block size and alignment.
1938 * value-prof.h (enum hist_type): add HIST_TYPE_AVERAGE and
1940 (struct profile_hooks): Add gen_average_profiler and gen_ior_profiler.
1941 (stringop_block_profile): Declare.
1942 * builtins.c: Include value-prof.h.
1943 (expand_builtin_memcpy, expand_builtin_memset): Pass block profile.
1944 * gcov-ui.h (GCOV_COUNTER_NAMES): Add new counter.
1945 (GCOV_COUNTER_AVERAGE, GCOV_COUNTER_IOR): New constants.
1946 (GCOV_COUNTERS, GCOV_LAST_VALUE_COUNTER): Update.
1947 * profile.c (instrument_values): Add new counters.
1948 * cfgexpand.c (expand_gimple_basic_block): Propagate histograms to
1950 * tree-profile.c (tree_average_profiler_fn, tree_ior_profiler_fn): New.
1951 (tree_init_edge_profiler): Build new profilers.
1952 (tree_gen_average_profiler, tree_gen_ior_profiler): New.
1953 (pass_tree_profile): Add dump.
1954 (tree_profile_hooks): Update.
1955 * Makefile.in (LIBGCOV): Add new constants.
1956 * libgcov.c (__gcov_merge_ior, __gcov_average_profiler,
1957 __gcov_ior_profiler): New.
1958 * i386.md (movmem/setmem expanders): Add new optional arguments.
1960 2007-01-28 David Edelsohn <edelsohn@gnu.org>
1962 * doc/md.texi (Standard Pattern Names): Document blockage pattern.
1964 2007-01-28 Zdenek Dvorak <dvorakz@suse.cz>
1966 * tree-ssa-loop-unswitch.c: Include tree-inline.h.
1967 (tree_unswitch_single_loop): Pass eni_size_weights to
1968 tree_num_loop_insns.
1969 * tree-ssa-loop-manip.c: Include tree-inline.h.
1970 (can_unroll_loop_p): Pass eni_size_weights to
1971 tree_num_loop_insns.
1972 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p):
1973 Pass eni_size_weights to estimate_num_insns.
1974 * tree.h (init_inline_once): Export.
1975 * toplev.c (backend_init): Call init_inline_once.
1976 * cgraphunit.c (cgraph_process_new_functions,
1977 cgraph_analyze_function): Pass eni_inlining_weights to
1979 * ipa-inline.c (compute_inline_parameters): Ditto.
1980 * tree-ssa-loop-ivcanon.c (tree_num_loop_insns): Pass weights
1981 to estimate_num_insns.
1982 (try_unroll_loop_completely): Pass eni_size_weights to
1983 tree_num_loop_insns.
1984 * tree-eh.c (decide_copy_try_finally): Pass eni_size_weights
1985 ot estimate_num_insns.
1986 * tree-ssa-loop-prefetch.c: Include tree-inline.h.
1987 (loop_prefetch_arrays): Pass eni_time_weights to tree_num_loop_insns.
1988 * tree-inline.c (eni_inlining_weights, eni_size_weights,
1989 eni_time_weights): New variables.
1990 (init_inline_once): Initialize them.
1991 (struct eni_data): Mew.
1992 (estimate_num_insns_1, estimate_num_insns): Use weights.
1993 * tree-inline.h (struct eni_weights_d): New.
1994 (eni_inlining_weights, eni_size_weights, eni_time_weights): Declare.
1995 (estimate_num_insns): Declaration changed.
1996 * cfgloop.h (tree_num_loop_insns): Declaration changed.
1997 * Makefile.in (tree-ssa-loop-unswitch.o, tree-ssa-loop-prefetch.o,
1998 tree-ssa-loop-manip.o): Add TREE_INLINE_H dependency.
2000 2007-01-28 Zdenek Dvorak <dvorakz@suse.cz>
2002 * tree-data-ref.c (conflict_fn): Assert that the number of affine
2003 relations in the conflict function is valid.
2005 2007-01-27 Ian Lance Taylor <iant@google.com>
2007 * common.opt: Add fstrict-overflow.
2008 * opts.c (decode_options): Set flag_strict_overflow if -O2.
2009 * flags.h (TYPE_OVERFLOW_WRAPS): Define.
2010 (TYPE_OVERFLOW_UNDEFINED): Define.
2011 (TYPE_OVERFLOW_TRAPS): Define. This replaces TYPE_TRAP_SIGNED.
2013 * tree.h (TYPE_TRAP_SIGNED): Don't define.
2014 * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_UNDEFINED.
2015 (fold_negate_expr): Likewise.
2016 (make_range): Likewise.
2017 (extract_muldiv_1): Likewise.
2018 (maybe_canonicalize_comparison): Likewise.
2019 (fold_comparison): Likewise.
2020 (fold_binary): Likewise.
2021 (tree_expr_nonnegative_p): Likewise.
2022 (tree_expr_nonzero_p): Likewise.
2023 * tree-vrp.c (compare_values): Likewise.
2024 (extract_range_from_binary_expr): Likewise.
2025 (extract_range_from_unary_expr): Likewise.
2026 * tree-ssa-loop-niter.c (infer_loop_bounds_from_signedness):
2028 (nowrap_type_p): Likewise.
2029 * tree-scalar-evolution.c (simple_iv): Likewise.
2030 * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_WRAPS.
2031 (build_range_check): Likewise.
2032 (extract_muldiv_1): Likewise.
2033 (fold_comparison): Likewise.
2034 * tree-vrp.c (vrp_int_const_binop): Likewise.
2035 (extract_range_from_unary_expr): Likewise.
2036 * convert.c (convert_to_integer): Likewise.
2037 * fold-const.c (fold_negate_expr): Use TYPE_OVERFLOW_TRAPS.
2038 (fold_comparison): Likewise.
2039 (fold_binary): Likewise.
2040 * optabs.c (optab_for_tree_code): Likewise.
2041 * tree-vectorizer.c (vect_is_simple_reduction): Likewise.
2042 * simplify-rtx.c (simplify_const_relational_operation): Check
2043 flag_strict_overflow and flag_trapv.
2044 (simplify_const_relational_operation): Likewise.
2045 * doc/invoke.texi (Option Summary): Mention -fstrict-overflow.
2046 (Optimize Options): Add -fstrict-overflow to -O2 list. Document
2049 2007-01-27 Roger Sayle <roger@eyesopen.com>
2051 * tree.c (tree_fold_gcd): Delete.
2052 * tree.h (tree_fold_gcd): Remove prototype.
2053 * tree-data-ref.c (tree_fold_divides_p): Don't use tree_fold_gcd to
2054 test whether one constant integer is a multiple of another. Instead
2055 call int_const_binop with TRUNC_MOD_EXPR and test for a zero result.
2056 * fold-const.c (multiple_of_p): We've determined both TOP and
2057 BOTTOM are integer constants so we can call int_const_binop directly
2058 instead of the more generic const_binop.
2060 2007-01-27 Roger Sayle <roger@eyesopen.com>
2062 * fold-const.c (size_binop): In the fast-paths for X+0, 0+X, X-0 and
2063 1*X check that the constant hasn't overflowed, to preserve the
2065 (round_up): Provide an efficient implementation when rouding-up an
2066 INTEGER_CST to a power-of-two.
2068 2007-01-28 Ralf Wildenhues <Ralf.Wildenhues@gmx.de>
2070 * doc/sourcebuild.texi: Add comma for clarity.
2071 * doc/extend.texi: Fix some typos.
2072 * doc/passes.texi: Likewise.
2073 * doc/cppinternals.texi: Likewise.
2074 * doc/c-tree.texi: Likewise.
2075 * doc/tree-ssa.texi: Likewise.
2076 * doc/install.texi: Likewise.
2078 2007-01-27 Jan Hubicka <jh@suse.cz>
2080 * tree-sra.c (sra_walk_function): Don't rely on aliases being build.
2081 (pass_sra): Do not require alias information.
2082 * passes.c (init_optimization_passes): Add SRA
2084 2007-01-27 Steven Bosscher <steven@gcc.gnu.org>
2086 * tracer.c (rest_of_handle_tracer): We already cleaned
2087 up the CFG in tracer() so don't do it here again.
2088 * cfgcleanup.c (rest_of_handle_jump2): Don't repeat
2089 cleanup_cfg here, either. And don't call renumber_insns.
2091 * cfgrtl.c (rtl_verify_flow_info_1): Don't verify that BB_END
2092 and BB_HEAD are in the insn stream here. Instead make sure
2093 that BB_INSN is valid on all insns. Also, do check here that
2094 there are no pending branch predictions...
2095 (rtl_verify_flow_info): ...instead of doing it here. Checks
2096 for BB_END and BB_HEAD moved from rtl_verify_flow_info_1 to
2099 2007-01-26 Roger Sayle <roger@eyesopen.com>
2101 * config/i386/i386.c (ix86_swap_binary_operands_p): New helper
2102 function to simplify/factorize operand order canonicalization.
2103 (ix86_fixup_binary_operands): Reorganize using the above function.
2104 (ix86_binary_operator_ok): Likewise.
2106 2007-01-27 Jakub Jelinek <jakub@redhat.com>
2108 * genattrtab.c (struct attr_value_list, insn_code_values): Move to
2109 file scope from optimize_attrs.
2110 (simplify_test_exp): If insn_code_values is not NULL, use it to speed
2112 (optimize_attrs): Clear insn_code_values after freeing it.
2114 2007-01-26 Zdenek Dvorak <dvorakz@suse.cz>
2116 * tree-ssa-address.c (create_mem_ref): Remove ", bsi" from
2117 a parts.base assignment.
2119 2007-01-26 Zdenek Dvorak <dvorakz@suse.cz>
2121 * tree-data-ref.c (dump_subscript): Use dump_conflict_function.
2122 (compute_subscript_distance, initialize_data_dependence_relation,
2123 finalize_ddr_dependent, analyze_ziv_subscript,
2124 analyze_siv_subscript_cst_affine,
2125 compute_overlap_steps_for_affine_univar,
2126 compute_overlap_steps_for_affine_1_2, analyze_subscript_affine_affine,
2127 analyze_siv_subscript, analyze_miv_subscript,
2128 analyze_overlapping_iterations, subscript_dependence_tester_1,
2129 compute_self_dependence, free_dependence_relation): Work
2130 with affine_fn instead of chrecs.
2131 (dump_affine_function, dump_conflict_function, affine_function_equal_p,
2132 common_affine_function, affine_function_base,
2133 affine_function_constant_p, affine_fn_op, affine_fn_plus,
2134 affine_fn_minus, affine_fn_free, conflict_fn_not_known,
2135 conflict_fn_no_dependence, free_conflict_function, free_subscripts,
2136 conflict_fn, affine_fn_cst, affine_fn_univar): New functions.
2137 (all_chrecs_equal_p): Removed.
2138 * tree-data-ref.h (affine_fn, conflict_function): New types.
2139 (struct subscript): Change type of conflicting_iterations_in_a
2140 and conflicting_iterations_in_b.
2142 2007-01-26 Steve Ellcey <sje@cup.hp.com>
2145 * config/pa/pa.h (TARGET_HPUX_11): New.
2146 * config/pa/pa-hpux11.h (TARGET_HPUX_11): New.
2147 * config/pa/pa.c (pa_init_builtins): Use TARGET_HPUX_11.
2149 2007-01-26 Daniel Berlin <dberlin@dberlin.org>
2150 Richard Guenther <rguenther@suse.de>
2152 * tree-ssa-structalias.c (solve_graph): Handle case
2153 we merged the variable to another.
2155 2007-01-25 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
2157 * builtins.c (fold_builtin_1): Treat ccos and ccosh as 'even'
2160 * fold-const.c (negate_mathfn_p): Treat casin, casinh, catan,
2161 catanh, cproj, csin, csinh, ctan and ctanh as 'odd' functions.
2163 2007-01-25 DJ Delorie <dj@redhat.com>
2165 * config/m32c/m32c.c (m32c_cannot_change_mode_class): We don't
2166 allow changes to modes which don't fit in those registers.
2168 * reload1.c (choose_reload_regs): Check for invalid subregs before
2169 computing their locations, not after.
2171 2007-01-25 Geoffrey Keating <geoffk@apple.com>
2174 * config/rs6000/rs6000.c (first_altivec_reg_to_save): On Darwin,
2175 save Altivec registers in an eh_return function.
2176 (compute_vrsave_mask): Likewise.
2177 (rs6000_stack_info): Correct AIX/Darwin stack alignment computation
2178 for saving Altivec registers.
2179 (rs6000_emit_prologue): Don't allocate stack twice in
2180 eh_return function. Correct expected value of altivec_save_offset
2181 when using save_world. Describe save of R0 to stack when using
2182 save_world. Describe stack pointer adjustment when using
2183 save_world. Remove duplicated eh_return parameter register saving.
2184 Update sp_offset variable after save_world.
2185 * config/rs6000/t-darwin (LIB2FUNCS_STATIC_EXTRA): Remove
2187 (LIB2FUNCS_EXTRA): Add darwin-world.asm.
2188 * config/rs6000/darwin.h (SUBTARGET_OVERRIDE_OPTIONS): -m64
2191 2007-01-25 Steve Ellcey <sje@cup.hp.com>
2193 * config.gcc (ia64*-*-hpux*): Make posix threads the default.
2195 2007-01-25 Steve Ellcey <sje@cup.hp.com>
2198 * config/pa/pa.c (pa_init_builtins): Set asm names for finite routines.
2199 * config/ia64/ia64.c (ia64_init_builtins): Ditto.
2201 2007-01-25 Richard Guenther <rguenther@suse.de>
2203 * doc/invoke.texi (-Wcoverage-mismatch): Document.
2204 * common.opt (-Wcoverage-mismatch): New warning option.
2205 * coverage.c (get_coverage_counts): Ignore coverage mismatch
2206 if -Wcoverage-mismatch is given.
2208 2007-01-25 Razya Ladelsky <razya@il.ibm.com>
2210 * ipa-cp.c (ipcp_insert_stage, ipcp_driver): Support for SSA.
2211 (ipcp_driver): Change to static definition.
2212 Add dumping of the ifunctions.
2213 (constant_val_insert): Remove unused parameter. Support for SSA.
2214 (ipcp_propagate_const): Support for SSA.
2215 (ipcp_profile_bb_print): Print only analyzed nodes.
2216 (ipcp_replace_map_create): Remove support for Fortran constant
2218 * ipa-prop.c (ipa_method_modify_stmt,
2219 ipa_callsite_compute_param): Support for SSA.
2220 * ipa-prop.h (ipcp_driver): Remove declaration.
2221 (IS_VALID_TREE_MAP_INDEX): Add define.
2223 2007-01-24 Geoffrey Keating <geoffk@apple.com>
2225 * unwind-dw2.c (execute_stack_op): Handle DW_OP_swap.
2227 2007-01-24 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
2230 * builtins.c (fold_builtin_cabs): Use MPFR to evaluate a
2231 constant argument to cabs and do it without checking for
2232 -funsafe-math-optimizations.
2234 2007-01-24 Douglas Gregor <dgregor@osl.iu.edu>
2236 * c-common.h (RID_FIRST_CXX0X): New.
2237 (RID_LAST_CXX0X): New.
2238 * c-opts.c (c_common_handle_option): -Wc++0x-compat is triggered
2240 * c.opt (Wc++0x-compat): New.
2241 * doc/invoke.texi (-Wc++0x-compat): Document.
2243 2007-01-24 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
2245 * builtins.c (fold_builtin_carg): New.
2246 (fold_builtin_1): Use it.
2248 2007-01-24 Jan Hubicka <jh@suse.cz>
2250 * ipa-inline.c (cgraph_decide_inlining): Initialize initial_insns.
2252 * ipa-inline.c (initial_insns, max_insns): Delete.
2253 (compute_max_insns): New function.
2254 (cgraph_decide_inlining_of_small_function): Use it; take minimal amount
2255 of insns as base for code growth.
2256 (cgraph_decide_inlining): Make initial_insns local; do not compute
2258 * params.def (PARAM_INLINE_UNIT_GROWTH): Set to 60.
2259 * doc/invoke.texi (inline-unit-growth): Update docs.
2261 2007-01-24 Jakub Jelinek <jakub@redhat.com>
2263 * config/i386/i386.h (x86_cmpxchg16b): Remove const.
2264 (TARGET_CMPXCHG16B): Define to x86_cmpxchg16b.
2265 * config/i386/i386.c (x86_cmpxchg16b): Remove const.
2266 (override_options): Add PTA_CX16 flag. Set x86_cmpxchg16b
2267 for CPUs that have PTA_CX16 set.
2270 * gimplify.c (omp_check_private): New function.
2271 (gimplify_scan_omp_clauses): Use it for
2272 firstprivate/lastprivate/reduction.
2275 * gimplify.c (omp_add_variable): Don't call omp_notice_variable
2276 on TYPE_SIZE_UNIT for GOVD_LOCAL VLAs.
2279 * omp-low.c (lower_omp_for_lastprivate): Add dlist argument.
2280 If lower_lastprivate_clauses emits some statements, append them
2281 to dlist rather than body_p and to body_p append an initializer.
2282 (lower_omp_for): Adjust caller.
2284 2007-01-24 Steve Ellcey <sje@cup.hp.com>
2286 * target.h (globalize_decl_name): New.
2287 * target-def.h (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
2288 * output.h (default_globalize_decl_name): New.
2289 * varasm.c (asm_output_bss): Use globalize_decl_name instead of
2291 (globalize_decl): Ditto.
2292 (default_globalize_decl_name): New.
2293 * config/ia64/ia64.c (ia64_globalize_decl_name): New.
2294 (ia64_handle_version_id_attribute): New.
2295 (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
2296 (ia64_asm_output_external): Use globalize_decl_name instead
2298 * doc/extend.texi (version_id): New pragma.
2299 * doc/tm.texi (ARGET_ASM_GLOBALIZE_DECL_NAME): New target hook.
2301 2007-01-24 Andreas Krebbel <krebbel1@de.ibm.com>
2303 * unwind-dw2-fde.c (get_cie_encoding): Replaced _Unwind_Word with
2304 _uleb128_t and _Unwind_SWord with _sleb128_t.
2305 * unwind-dw2.c (extract_cie_info, execute_stack_op, execute_cfa_program,
2306 uw_frame_state_for, uw_update_context_1): Likewise.
2307 * unwind-c.c (parse_lsda_header, PERSONALITY_FUNCTION): Likewise.
2308 * unwind-pe.h (read_uleb128, read_sleb128,
2309 read_encoded_value_with_base): Likewise.
2310 * unwind-generic.h: Define _sleb128_t and _uleb128_t types.
2312 2007-01-24 Richard Guenther <rguenther@suse.de>
2314 * builtins.c (expand_builtin_cexpi): Get the fndecl
2315 for cexp in the correct way.
2317 2007-01-24 Jan Hubicka <jh@suse.cz>
2319 * tree-ssa-dce.c (eliminate_unnecesary_stmts): Remove dead LHS of calls.
2321 2007-01-24 Andreas Krebbel <krebbel1@de.ibm.com>
2323 * c-cppbuiltin.c (builtin_define_type_sizeof): New function.
2324 (c_cpp_builtins): New builtin macros: __SIZEOF_INT__, __SIZEOF_LONG__,
2325 __SIZEOF_LONG_LONG__, __SIZEOF_SHORT__, __SIZEOF_POINTER__,
2326 __SIZEOF_FLOAT__, __SIZEOF_DOUBLE__, __SIZEOF_LONG_DOUBLE__,
2327 __SIZEOF_SIZE_T__, __SIZEOF_WCHAR_T__, __SIZEOF_WINT_T__ and
2328 __SIZEOF_PTRDIFF_T__.
2329 * doc/cpp.texi: Documentation for the new builtin macros added.
2331 2007-01-24 Uros Bizjak <ubizjak@gmail.com>
2333 * config/i386/i386.md (tanxf2, tan<mode>2, atan<mode>2, log<mode>2,
2334 log10<mode>2, log2<mode>2, expxf2, exp10xf2, exp2xf2): Use op2
2335 instead of operands[2] to avoid access past the end of array.
2337 2007-01-24 Richard Sandiford <richard@codesourcery.com>
2339 * reload1.c (emit_reload_insns): Pass the reload register
2340 for a non-spill output reload through forget_old_reloads_1.
2342 2007-01-23 Joseph Myers <joseph@codesourcery.com>
2344 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Define
2345 _SOFT_DOUBLE if doubles use software floating-point.
2346 * config/rs6000/libgcc-ppc-glibc.ver: Export additional long
2347 double functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
2348 * config/rs6000/darwin-ldouble.c: Also compile functions for
2349 hard-float without FPRs. Use fmsub function for all __NO_FPRS__
2350 cases. Compile extra functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
2351 * config/rs6000/linuxspe.h (SUBSUBTARGET_OVERRIDE_OPTIONS): Remove
2352 commented-out long double override.
2353 (CPP_LONGDOUBLE_DEFAULT_SPEC): Likewise.
2354 * config/rs6000/eabispe.h: Likewise.
2355 * config/rs6000/rs6000.c (rs6000_override_options): Don't override
2356 long double for non-SPE.
2357 (rs6000_handle_option): Likewise.
2358 (invalid_e500_subreg): Disallow more subregs involding DImode,
2359 DFmode, TImode or TFmode.
2360 (rs6000_legitimate_offset_address_p): Check TFmode offsets for
2362 (legitimate_lo_sum_address_p): Also check for TFmode for E500
2364 (rs6000_legitimize_address): Also handle TFmode for E500 double.
2365 (rs6000_legitimize_reload_address): Also handle TFmode for E500
2367 (rs6000_legitimate_address): Also check for TFmode for E500
2369 (rs6000_emit_move): Use DFmode subregs of TFmode for E500 double.
2370 (spe_build_register_parallel): Handle TFmode and TCmode.
2371 (rs6000_spe_function_arg): Handle TFmode and TCmode for E500
2373 (function_arg): Handle TFmode and TCmode for E500 double.
2374 (rs6000_init_libfuncs): Initialize extra libfuncs for soft double
2376 (print_operand): Handle TFmode and TImode for %y.
2377 (rs6000_generate_compare): Handle TFmode comparisons for E500
2379 (spe_func_has_64bit_regs_p): Check for TFmode for E500 double.
2380 (rs6000_function_value): Handle TFmode and TCmode for E500 double.
2381 (rs6000_libcall_value): Handle TFmode and TCmode for E500 double.
2382 * config/rs6000/rs6000.h (CANNOT_CHANGE_MODE_CLASS): Check for
2383 TFmode for E500 double.
2384 * config/rs6000/rs6000.md (FP): Allow TF for E500 double.
2385 (floatsidf2): Enable for E500 double.
2386 (movtf_softfloat): Use rs6000_nonimmediate_operand.
2387 (extenddftf2): Change to extenddftf2_fprs.
2388 (extenddftf2): Call gen_spe_extenddftf2 or gen_extenddftf2_fprs
2389 depending on TARGET_E500_DOUBLE.
2390 (extendsftf2): Enable for E500 double.
2391 (trunctfdf2): Enable for E500 double.
2392 (trunctfsf2): Change to trunctfsf2_fprs.
2393 (trunctfsf2): Call gen_spe_trunctfsf2 or gen_trunctfsf2_fprs
2394 depending on TARGET_E500_DOUBLE.
2395 (floatsitf2): Enable for E500 double.
2396 (fix_trunctfsi2): Change to fix_trunctfsi2_fprs.
2397 (fix_trunctfsi2): Call gen_spe_fix_trunctfsi2 or
2398 gen_fix_trunctfsi2_fprs depending on TARGET_E500_DOUBLE.
2399 (negtf2): Change to negtf2_internal.
2400 (negtf2): New expander.
2401 (abstf2): Enable for E500 double. Call gen_spe_abstf2_tst,
2402 gen_spe_abstf2_cmp or gen_abstf2_internal depending on
2403 TARGET_E500_DOUBLE and flag_unsafe_math_optimizations.
2404 (movdi_internal32): Use rs6000_nonimmediate_operand.
2405 (unnamed splitter): Likewise.
2406 * config/rs6000/spe.md (CMPTFEQ_GPR, TSTTFEQ_GPR, CMPTFGT_GPR,
2407 TSTTFGT_GPR, CMPTFLT_GPR, TSTTFLT_GPR): New unspecs.
2408 (SPE64TF, DITI): New mode macros.
2409 (frob_df_di): Change to frob_<SPE64:mode>_<DITI:mode>; allow more
2412 (frob_<mode>_di_2): New.
2413 (frob_tf_di_8_2): New.
2414 (frob_di_df): Change to frob_di_<mode>; allow more modes.
2416 (frob_di_df_2): Change to frob_<DITI:mode>_<SPE64:mode>_2; allow
2418 (frob_ti_<mode>_8_2): New.
2419 (frob_ti_tf_2): New.
2420 (mov_si<mode>_e500_subreg0, mov_si<mode>_e500_subreg0_2,
2421 mov_si<mode>_e500_subreg4, mov_si<mode>_e500_subreg4_2): Allow
2423 (mov_sitf_e500_subreg8, mov_sitf_e500_subreg8_2,
2424 mov_sitf_e500_subreg12, mov_sitf_e500_subreg12_2): New.
2425 (spe_trunctfdf2_internal1, spe_trunctfsf2, spe_extenddftf2,
2426 spe_fix_trunctfsi2, spe_fix_trunctfsi2_internal,
2427 spe_negtf2_internal, spe_abstf2_cmp, spe_abstf2_tst): New.
2428 (cmptfeq_gpr, tsttfeq_gpr, cmptfgt_gpr, tsttfgt_gpr, cmptflt_gpr,
2431 2007-01-23 Ian Lance Taylor <iant@google.com>
2433 * Makefile.in (OBJS-common): Reformat, alphabetize, but put
2435 (OBJS-archive): Reformat, alphabetize.
2436 (OBJS): Change out_object_file to OBJS-md.
2438 2007-01-23 Uros Bizjak <ubizjak@gmail.com>
2440 * config/i386/i386.md: Use REG_P, MEM_P, CONST_INT_P, LABEL_P,
2441 JUMP_P and CALL_P predicates where applicable.
2442 * config/i386/i386.c: Ditto.
2443 * config/i386/i386.md: Ditto.
2444 * config/i386/mmx.md: Ditto.
2445 * config/i386/predicates.md: Ditto.
2447 2007-01-22 Andreas Schwab <schwab@suse.de>
2449 * config/m68k/m68k.h: Fix comment.
2451 2007-01-22 Jan Hubicka <jh@suse.cz>
2453 * passes.c (init_optimization_passes): Do not rerun
2454 pass_early_warn_uninitialized.
2456 2007-01-22 Richard Guenther <rguenther@suse.de>
2458 PR tree-optimization/30038
2459 * tree-ssa-math-opts.c (maybe_record_sincos): New static helper
2461 (execute_cse_sincos_1): Likewise.
2462 (execute_cse_sincos): Likewise.
2463 (gate_cse_sincos): Likewise.
2464 (pass_cse_sincos): New pass CSEing sin() and cos() calls using
2465 the cexpi() canonicalization of sincos().
2466 * tree-pass.h (pass_cse_sincos): Declare.
2467 * passes.c (init_optimization_passes): New pass pas_cse_sincos.
2469 2007-01-21 Eric Botcazou <ebotcazou@libertysurf.fr>
2471 PR rtl-optimization/29329
2472 * combine.c (replaced_rhs_insn): Rename to i2mod.
2473 (replaced_rhs_value): Rename to i2mod_new_rhs.
2474 (i2mod_old_rhs): New global variable.
2475 (combine_instructions): Adjust for above change. Save a copy of
2476 the old RHS into i2mod_old_rhs when the contents of a REG_EQUAL
2477 note are substituted in the second instruction.
2478 (distribute_notes) <REG_DEAD>: Adjust for above change. Do not
2479 ditch the note if it pertains to the second eliminated register
2480 and this register is mentioned in i2mod_old_rhs.
2483 2006-09-12 Eric Botcazou <ebotcazou@libertysurf.fr>
2485 * combine.c (distribute_notes) <REG_DEAD>: Do not consider SETs past
2486 the insn to which the note was originally attached.
2488 2007-01-21 Jan Hubicka <jh@suse.cz>
2490 * ipa-inline.c (inlining_mode): Comment, move up.
2491 (cgraph_decide_inlining_incrementally): Do not perform inlining
2492 itself; fix handling of flattening of self recursive functions.
2493 (cgraph_find_cycles): Remove.
2494 (cgraph_flatten_node): Remove.
2495 (cgraph_decide_inlining): Use incremental inliner to handle flattening.
2496 (try_inline): New function.
2497 (cgraph_early_inlining): Update call of
2498 cgraph_decide_inlining_incrementally. Apply inlining here.
2499 (apply_inline): Update call of cgraph_decide_inlining_incrementally.
2501 2007-01-21 Dirk Mueller <dmueller@suse.de>
2504 * tree-vrp.c (check_array_bounds): do not warn
2505 about ADDR_EXPR's of ARRAY_REF's which are immediately
2506 used in binary expressions.
2508 2007-01-21 Ira Rosen <irar@il.ibm.com>
2510 * tree-vectorizer.h (struct _stmt_vec_info): Add new field
2511 read_write_dep and macros for its access.
2512 * tree-vectorizer.c (new_stmt_vec_info): Initialize the new field.
2513 * tree-vect-analyze.c (vect_analyze_data_ref_dependence): Remove
2514 argument, call vect_check_interleaving for every independent pair of
2515 data-refs. Mark loads that access the same memory location as a store
2517 (vect_check_dependences): Remove.
2518 (vect_analyze_data_ref_dependences): Remove vect_check_dependences
2519 call, fix the call to vect_analyze_data_ref_dependence.
2520 (vect_analyze_data_ref_access): For statements that access the same
2521 data-ref, check that they are not stores; for loads, check that there
2522 is no store that access the same location.
2524 2007-01-20 Roger Sayle <roger@eyesopen.com>
2525 Joseph Myers <joseph@codesourcery.com>
2527 * doc/invoke.texi (-fdump-rtl-combine): Fix under/overfull hbox.
2528 (-fdump-rtl-stack): Likewise.
2529 (-fno-signed-zeros): Use @minus{} for a minus sign. Correct typo.
2530 (-mcheck-zero-division, -mcpu): Fix under/overfull hbox.
2531 (-mpt-fixed): Use @minus{} for minus sign.
2532 (Using Precompiled Headers): Fix under/overfull hbox.
2534 2007-01-20 Jan Hubicka <jh@suse.cz>
2536 * tree-flow.h (struct stmt_ann_d): Move references_memory to proper
2537 place within annotation.
2539 2007-01-20 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
2541 * pa.c (output_move_double): Change array size of xoperands to 4.
2543 2007-01-20 Jan Hubicka <jh@suse.cz>
2545 * tree-tailcall.c (adjust_return_value): Do not use RESULT_DECL
2548 2007-01-19 Ian Lance Taylor <iant@google.com>
2550 * expmed.c (expand_divmod) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]: Cast
2551 constant to unsigned HOST_WIDE_INT before negating.
2553 2007-01-19 Ian Lance Taylor <iant@google.com>
2555 * tree-ssa-operands.h (struct vuse_vec_d): Change num_vuse field
2557 (VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Use unsigned comparison.
2558 (VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
2559 (SET_VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Likewise.
2560 (SET_VUSE_ELEMENT_VAR) [ENABLE_CHECKING]: Likewise.
2561 (SET_VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
2562 (realloc_vdef, realloc_vuse): Change second parameter to
2564 (ssa_operand_iterator_d): Change vuse_index and mayuse_index
2566 * tree-ssa-operands.c (realloc_vop): Change num_elem parameter to
2567 unsigned. Change x and lim locals to unsigned.
2568 (realloc_vdef, realloc_vuse): Change num_elem parameter to
2570 (finalize_ssa_vuse_ops): Change old_i local to unsigned.
2571 (copy_virtual_operands): Change i and n locals to unsigned.
2573 2007-01-19 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
2576 * builtins.c (fold_builtin_1): Handle builtin fdim.
2578 2007-01-20 Jan Hubicka <jh@suse.cz>
2580 * tree-ssa.c (init_tree_ssa): Do not call init_alias_heapvars.
2581 * tree-ssa-structalias.c (compute_points_to_sets): Do call
2582 init_alias_heapvars.
2583 (init_alias_heapvars): Initialize only when not already initialized.
2584 (delete_alias_heapvars): Set heapvar_for_stmt to NULL.
2586 2007-01-19 Roger Sayle <roger@eyesopen.com>
2588 * common.opt (fsigned-zeros): New command line option.
2589 * flags.h (HONOR_SIGNED_ZEROS): Control via flag_signed_zeros instead
2590 of flag_unsafe_math_optimizations.
2591 * opts.c (set_fast_math_flags): The -ffast-math command line option
2592 implies -fno-signed-zeros.
2593 (fast_math_flags_set_p): Likewise.
2595 * doc/invoke.texi: Document new -fno-signed-zeros option, and update
2596 the documentation of -ffast-math appropriately. Wrap long lines.
2598 2007-01-19 Steve Ellcey <sje@cup.hp.com>
2600 * system.h (ASM_MAKE_LABEL_LINKONCE): Poison.
2601 * varasm.c (globalize_decl): Remove ASM_MAKE_LABEL_LINKONCE ifdef.
2603 2007-01-19 Tomas Bily <tbily@suse.cz>
2605 * cgraphunit.c (cgraph_finalize_function): Updating of pid
2607 (tree_init_ic_make_global_vars): New function
2608 (tree_init_edge_profiler): call of tree_init_ic_make_global_vars
2609 (tree_gen_ic_profiler): New function
2610 (tree_gen_ic_func_profiler): New function
2611 (tree_profiling): Added calling of tree_gen_ic_func_profiler
2612 (tree_profile_hooks): Added hook for indirec/virtual calls
2613 * value-prof.c (tree_find_values_to_profile): New case for
2615 (tree_values_to_profile): Call for determining indirect/virtual
2617 (tree_indirect_call_to_profile): New function
2618 (tree_ic_transform): New function
2619 (tree_ic): New function
2620 (find_func_by_pid): New function
2621 (init_pid_map): New function
2622 (tree_value_profile_transformations): Added check for
2623 indirect/virtual call transformation
2624 * value-prof.h (enum hist_type): New counter type for
2625 indirect/virtual calls
2626 (profile_hooks): Added new hook for profiling indirect/virtual
2628 * profile.c (instrument_values): New case for indirect/virtual
2630 * gcov-io.h (GCOV_LAST_VALUE_COUNTER): Changed to 6
2631 (GCOV_COUNTER_V_INDIR): New counter type
2632 (GCOV_COUNTER_NAMES): New name of counter "indirect" added
2633 (GCOV_MERGE_FUNCTIONS): New merge function for indirect/virtual
2635 * cgraph.c: Definition of cgraph_max_pid
2636 (cgraph_create_node): Default init of pid attribute
2637 * cgraph.h: Declaration of cgraph_max_pid
2638 (struct cgraph_node): Added pid attribute
2639 * libgcov.c (__gcov_indirect_call_profiler): New function
2640 (__gcov_one_value_profiler_body): New function
2641 (__gcov_one_value_profiler): Body was moved to
2642 __gcov_one_value_profiler_body and calls it
2644 2007-01-19 Basile Starynkevitch <basile@starynkevitch.net>
2646 * doc/gty.texi (Options): Document the mark_hook option to GTY.
2647 * gengtype.c (write_types_data, write_func_for_structure,
2648 write_types, ggc_wtd, pch_wtd): Add skip_hooks to
2649 write_types_data, ggc_wtd, pch_wtd for processing mark_hook.
2650 (walk_type, write_func_for_structure): Generate the mark_hook if
2653 2007-01-19 Jan Hubicka <jh@suse.cz>
2655 * ipa-inline.c (cgraph_decide_inlining_incrementally): Instead of
2656 'early' argument take inlining mode argument specifying whether to
2657 inline for size/speeed or all functions; add support for flattening;
2659 (cgraph_early_inlining): Update call of decide_inlining_incrementally.
2661 2007-01-19 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
2664 * toplev.c (warn_deprecated_use): Use %qD instead of %qs to print
2665 the name of the declared identifier.
2667 2007-01-19 Dirk Mueller <dmueller@suse.de>
2669 * config/i386.h (CONDITIONAL_REGISTER_USAGE): Store
2670 result of PIC_OFFSET_TABLE_REGNUM in temporary variable to avoid
2671 duplicate evaluation.
2673 2007-01-19 Uros Bizjak <ubizjak@gmail.com>
2675 * config/i386/i386.md (acos<mode>2): Rename from acossf2 and acosdf2.
2676 Macroize expander using X87MODEF12 mode macro. Extend operand 1
2677 to XFMode, use acosxf2 and truncate result to requested mode.
2678 Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
2679 (asin<mode>2): Similarly, with asin expanders.
2680 (*fscalexf4_i387): Rename from *fscalexf4.
2681 (expNcorexf3): New expander.
2682 (expxf2, exp10xf2, exp2xf2): Use expNcorexf3 expander.
2683 (exp<mode>2): Rename from expsf2 and expdf2. Macroize expander using
2684 X87MODEF12 mode macro. Extend operand 1 to XFMode, use expxf2 and
2685 truncate result to requested mode. Use SSE_FLOAT_MODE_P to disable
2686 patterns for SSE math.
2687 (exp10<mode>2): Similarly, with exp10 expanders.
2688 (exp2<mode>2): Similarly, with exp2 expanders.
2689 (expm1<mode>2): Similarly, with expm1 expanders.
2690 (ldexp<mode>3): Similarly, with ldexp expanders.
2691 (log<mode>2, log10<mode>2, log2<mode>2, log1p<mode>2, logb<mode>2):
2692 Use gen_truncxf<mode>2_i387_noop to truncate result.
2694 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2696 * config/m68k/m68k.h (PREFERRED_STACK_BOUNDARY): Define to 32
2697 for ColdFire targets.
2699 2007-01-19 Nathan Sidwell <nathan@codesourcery.com>
2700 Richard Sandiford <richard@codesourcery.com>
2702 * config/m68k/m68k.h (M68K_STATIC_CHAIN_REG_NAME): New macro.
2703 (INITIALIZE_TRAMPOLINE): Use STATIC_CHAIN_REGNUM.
2704 (__transfer_from_trampoline): Use M68K_STATIC_CHAIN_REG_NAME.
2705 * config/m68k/m68kelf.h (STATIC_CHAIN_REGNUM): Override.
2706 (M68K_STATIC_CHAIN_REG_NAME): Likewise.
2707 * config/m68k/netbsd-elf.h (M68K_STATIC_CHAIN_REG_NAME): Likewise.
2709 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2711 * config/m68k/m68k.md (adddi_dilshr32): Rename to...
2712 (*adddi_dilshr32): ...this. Fix formatting. Remove commented-out
2713 non-canonical pattern. Restrict to !TARGET_COLDFIRE.
2714 (*adddi_dilshr32_cf): New pattern.
2715 (adddi3, subdi3): Remove first alternatives.
2717 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2719 * config/m68k/m68k.c (notice_update_cc): If an SFmode move is
2720 implemented using move.l, do not use its cc result for floating-point
2723 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2725 * config/m68k/m68k.h (EXTRA_CONSTRAINT): Stop the 'T' constraint
2726 from accepting 's' constraints if flag_pic.
2728 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2730 * config/m68k/m68k.md (bordered, bunordered, buneq, bunge, bungt)
2731 (bunle, bunlt, bltgt, bordered_rev, bunordered_rev, buneq_rev)
2732 (bunge_rev, bungt_rev, bunle_rev, bunlt_rev, bltgt_rev): Change
2733 condition from TARGET_68881 to TARGET_HARD_FLOAT.
2735 2007-01-19 Sandra Loosemore <sandra@codesourcery.com>
2737 * longlong.h (count_leading_zeros, COUNT_LEADING_ZEROS_0): Add
2738 ColdFire alternatives.
2739 * config/m68k/m68k.h (CLZ_DEFINED_VALUE_AT_ZERO): New macro.
2740 * config/m68k/m68k.md (clzsi2): Define for ColdFire
2741 architectures that support ff1 instruction.
2743 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2744 Julian Brown <julian@codesourcery.com>
2746 * config/m68k/m68k.h (CONST_OK_FOR_LETTER_P): Add an 'R' case.
2747 * config/m68k/m68k.md (*movsi_cfv4): Fold into...
2748 (*movsi_cf): ...here. Remove unnecessary 'R' from 'Rg'.
2751 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2753 * config/m68k/m68k-protos.h (valid_mov3q_const): Take a HOST_WIDE_INT
2755 (output_move_const_into_data_reg, output_move_simode_const): Delete.
2756 * config/m68k/m68k.c (const_method, const_int_cost): Take a
2757 HOST_WIDE_INT instead of an rtx.
2758 (m68k_rtx_costs): Update call accordingly.
2759 (output_move_const_into_data_reg): Likewise. Fix formatting.
2760 (valid_mov3q_const): Take a HOST_WIDE_INT instead of an rtx.
2762 (output_move_simode_const): Update calls after above changes.
2763 Rework to use automatic variables and predicates like MEM_P.
2764 * config/m68k/m68k.md (pushexthisi_const): Update call to
2767 2007-01-19 Dirk Mueller <dmueller@suse.de>
2769 * tree-ssa-alias.c (perform_var_substitution): Fix typo
2772 2007-01-19 Richard Guenther <rguenther@suse.de>
2774 * builtins.c (expand_builtin_cexpi): Fall back to expanding
2775 via cexp in case sincos is not available.
2777 2007-01-19 Richard Guenther <rguenther@suse.de>
2779 * doc/tm.texi (TARGET_HAS_SINCOS): Document new target macro.
2780 * defaults.h (TARGET_HAS_SINCOS): Default to off.
2781 * config/linux.h (TARGET_HAS_SINCOS): Set to on if we have glibc.
2782 * config/alpha/linux.h (TARGET_HAS_SINCOS): Likewise.
2783 * config/sparc/linux.h (TARGET_HAS_SINCOS): Likewise.
2784 * config/sparc/linux64.h (TARGET_HAS_SINCOS): Likewise.
2785 * config/rs6000/linux.h (TARGET_HAS_SINCOS): Likewise.
2786 * config/rs6000/linux64.h (TARGET_HAS_SINCOS): Likewise.
2788 2007-01-19 Uros Bizjak <ubizjak@gmail.com>
2790 * config/i386/i386.md (*fpatanxf3_i387, fpatan_extend<mode>xf3_i387):
2792 (atan2sf3_1, atan2df3_1, atan2xf3_1): Remove insn patterns.
2793 (atan2xf3): Directly generate RTL pattern.
2794 (atan2<mode>3): Rename from atan2sf3 and atan2df3 and macroize insn
2795 patterns using X87MODEF12 mode macro. Use fpatan_extend<mode>xf3_i387
2796 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
2797 disable patterns for SSE math.
2798 (atan<mode>2): Rename from atansf2 and atandf2 and macroize insn
2799 patterns using X87MODEF12 mode macro. Use fpatan_extend<mode>xf3_i387
2800 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
2801 disable patterns for SSE math.
2803 2007-01-19 Alexandre Oliva <aoliva@redhat.com>
2805 * libgcc-std.ver: Fix typo in %inherit for GCC_4.3.0.
2807 2007-01-18 Roger Sayle <roger@eyesopen.com>
2809 * fold-const.c (fold_unary) <VIEW_CONVERT_EXPR>: Optimize away a
2810 VIEW_CONVERT_EXPR to the same type as it's operand.
2812 2007-01-18 David Edelsohn <edelsohn@gnu.org>
2814 * config/rs6000/darwin-ldouble.c: Only build _SOFT_FLOAT if
2815 configured for long double 128.
2817 2007-01-18 Mike Stump <mrs@apple.com>
2819 * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Fix build
2822 2007-01-18 Michael Meissner <michael.meissner@amd.com>
2824 * i386.c (ix86_compute_frame_layout): Make fprintf's in #if 0 code
2827 2007-01-18 Jan Hubicka <jh@suse.cz>
2829 * tree-ssa-operands.c (vop_free_bucket_size): Never return value
2830 greater than NUM_VOP_FREE_BUCKETS.
2832 2007-01-18 Daniel Berlin <dberlin@dberlin.org>
2834 * tree-ssa-structalias.c: Update comments.
2835 (ptabitmap_obstack): Removed.
2837 (oldpta_obstack): Ditto.
2838 (stats): Add a few members.
2839 (struct variable_info): Remove node, complex, address_taken, and
2840 indirect_target members. Add oldsolution member.
2841 (new_var_info): Do not initialize removed members.
2842 (constraint_expr_type): Remove INCLUDES.
2843 (constraint_graph): Add size, implicit_preds, rep,
2844 indirect_cycles, eq_rep, label, direct_nodes, and complex members.
2845 (FIRST_REF_NODE): New macro.
2846 (LAST_REF_NODE): Ditto.
2847 (FIRST_ADDR_NODE): Ditto.
2848 (find): New function.
2850 (dump_constraint): Do not handle INCLUDES.
2851 (insert_into_complex): Do not insert duplicate constraints.
2852 (condense_varmap_nodes): Renamed and rewritten into ...
2853 (merge_node_constraints): This. Also fix bug in handling of
2854 offseted copy constraints.
2855 (clear_edges_for_node): No longer need to deal with preds at all,
2856 or removing associated preds/succs.
2857 (merge_graph_nodes): Deal with indirect_cycles.
2858 Don't deal with predecessors.
2859 (add_implicit_graph_edge): New function.
2860 (add_pred_graph_edge): Ditto.
2861 (add_graph_edge): Don't deal with predecessors.
2862 (build_constraint_graph): Removed.
2863 (build_pred_graph): New function.
2864 (build_succ_graph): Ditto.
2865 (struct scc_info): Removed in_component. Added roots, dfs, and
2866 node_mapping. Remove visited_index, unification_queue.
2867 (scc_visit): Deal with union-find we do now.
2868 Deal with cycles with REF nodes.
2869 (collapse_nodes): Renamed and rewritten to ...
2870 (unify_nodes): This.
2871 (process_unification_queue): Removed.
2872 (topo_visit): Cleanup
2873 (do_da_constraint): Use find.
2874 (do_sd_constraint): Ditto.
2875 (do_ds_constraint): Ditto.
2876 (do_complex_constraint): Ditto.
2877 (init_scc_info): Update for removed and added members.
2878 (find_and_collapse_graph_cycles): Renamed and rewritten into ...
2879 (find_indirect_cycles): This.
2880 (equivalence_class): New variable.
2881 (label_visit): New function.
2882 (perform_variable_substitution): Rewritten.
2883 (free_var_substitution_info): New function.
2884 (find_equivalent_node): Ditto.
2885 (move_complex_constraints): Ditto.
2886 (eliminate_indirect_cycles): Ditto.
2887 (solve_graph): Only propagate changed bits.
2888 Use indirect cycle elimination.
2890 (tree_id_t): Rename to tree_vi_t, delete id member, add vi member.
2891 (tree_id_eq): Renamed to ...
2892 (tree_vi_eq): This. Update for member change
2893 (insert_id_for_tree): Renamed and rewritten to ...
2894 (insert_vi_for_tree): This.
2895 (lookup_id_for_tree): Renamed and rewritten to ...
2896 (lookup_vi_for_tree): This.
2897 (get_id_for_tree): Renamed and rewritten to ...
2898 (get_vi_for_tree): Ditto.
2899 (get_constraint_exp_from_ssa_var): Update to use get_vi_for_tree.
2900 (process_constraint): Don't handle INCLUDES.
2901 Remove special ADDRESSOF case.
2902 (find_func_aliases): Rewrite to use vi functions instead of id
2904 (create_function_info_for): Ditto.
2905 (create_variable_info_for): Ditto.
2906 (intra_create_variable_infos): Ditto.
2907 (merge_smts_into): Ditto.
2908 (find_what_p_points_to): Ditto.
2909 (init_base_vars): Ditto.
2910 (init_alias_vars): Ditto.
2911 (remove_preds_and_fake_succs): New function.
2912 (dump_sa_points_to_info): Dump new stats.
2913 (dump_solution_for_var): Use find.
2914 (set_used_smts): Fix formatting.
2915 (compute_points_to_sets): Updated for new functions.
2916 (ipa_pta_execute): Ditto.
2918 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
2919 Richard Sandiford <richard@codesourcery.com>
2921 * doc/tm.texi (TARGET_FUNCTION_VALUE): Expand documentation of
2923 * calls.c (expand_call): If the return value is a PARALLEL,
2924 extract its first member.
2925 * config/m68k/linux.h (FUNCTION_EXTRA_EPILOGUE): Remove.
2926 * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
2927 use FUNCTION_EXTRA_EPILOGUE.
2928 (m68k_function_value): Return a PARALLEL if the return value
2929 is of a pointer type.
2930 * config/m68k/netbsd-elf.h (current_function_returns_pointer)
2931 (FUNCTION_EXTRA_EPILOGUE): Remove.
2932 * config/m68k/m68k.md (D0_REG): New constant.
2934 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
2936 * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
2937 output a NOP for empty epilogues.
2939 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2941 * config/m68k/m68k.c (m68k_use_return_insn): Update comments
2942 before function. Extend register save check to include all
2943 registers, not just integer ones.
2945 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
2947 * config/m68k/m68k-protos.h (use_return_insn): Rename to...
2948 (m68k_use_return_insn): ...this.
2949 * config/m68k/m68k.h (USE_RETURN_INSN): Delete.
2950 * config/m68k/m68k.c (use_return_insn): Rename to...
2951 (m68k_use_return_insn): ...this.
2952 * config/m68k/m68k.md (return): Use m68k_use_return_insn instead
2955 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
2957 * config/m68k/fpgnulib.c (__truncdfsf2): Implement round to
2958 nearest even, fix denormal rounding overflow.
2960 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2962 * config/m68k/m68k.md (movsf_cf_hard): Use fsmove instead of
2963 f%$smove and f%$move.
2964 (movdf_cf_hard): Use fdmove for cases 0 and 3 and fmove for case 1.
2965 (extendsfdf2_cf): Use fdmove instead of f%&move.
2966 (truncdfsf2_cf): Use fsmove instead of f%$smove.
2967 (add<mode>3_cf, sub<mode>3_cf): Use <FP:prec> instead of <FP:round>.
2969 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2971 * config/m68k/m68k.md (movdf_cf_hard): Use output_move_double for
2974 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2976 * real.h (real_format): Add a canonical_nan_lsbs_set field.
2977 (coldfire_single_format): Declare.
2978 (coldfire_double_format): Likewise.
2979 * real.c (encode_ieee_single): Use canonical_nan_lsbs_set instead
2980 of qnan_msb_set to determine the lower bits of a canonical
2982 (encode_ieee_double): Likewise.
2983 (encode_ieee_quad): Likewise.
2984 (ieee_single_format): Initialize canonical_nan_lsbs_set.
2985 (mips_single_format): Likewise.
2986 (ieee_double_format): Likewise.
2987 (mips_double_format): Likewise.
2988 (ieee_extended_motorola_format): Likewise.
2989 (ieee_extended_intel_96_format): Likewise.
2990 (ieee_extended_intel_128_format): Likewise.
2991 (ieee_extended_intel_96_round_53_format): Likewise.
2992 (ibm_extended_format): Likewise.
2993 (mips_extended_format): Likewise.
2994 (ieee_quad_format): Likewise.
2995 (mips_quad_format): Likewise.
2996 (vax_f_format): Likewise.
2997 (vax_d_format): Likewise.
2998 (vax_g_format): Likewise.
2999 (i370_single_format): Likewise.
3000 (i370_double_format): Likewise.
3001 (decimal_single_format): Likewise.
3002 (decimal_double_format): Likewise.
3003 (decimal_quad_format): Likewise.
3004 (c4x_single_format): Likewise.
3005 (c4x_extended_format): Likewise.
3006 (real_internal_format): Likewise.
3007 (coldfire_single_format): New real_format.
3008 (coldfire_double_format): Likewise.
3009 * config/pdp11/pdp11.c (pdp11_f_format): Initialize
3010 canonical_nan_lsbs_set.
3011 (pdp11_d_format): Likewise.
3012 * config/m68k/m68k.c (override_options): Override REAL_FORMAT_MODE
3013 if TARGET_COLDFIRE_CPU.
3015 2007-01-18 Richard Sandiford <richard@codesourcery.com>
3017 * config/m68k/m68k-protos.h (m68k_output_pic_call): Delete.
3018 (output_call, m68k_legitimize_call_address): Declare.
3019 * config/m68k/m68k.h (EXTRA_CONSTRAINT): Remove unnecessary
3020 parenthesees. Add support for a 'W' constraint.
3021 (LEGITIMATE_PIC_OPERAND_P): Remove SYMBOL_REF_FLAG handling.
3022 (PRINT_OPERAND_PUNCT_VALID_P): Remove comment about 'o'.
3023 (m68k_symbolic_call, m68k_symbolic_jump): Declare.
3024 * config/m68k/m68k.c (m68k_symbolic_call, m68k_symbolic_jump): New
3026 (override_options): Initialize them. Do not set flag_no_function_cse
3027 for TARGET_ID_SHARED_LIBRARY.
3028 (m68k_output_pic_call): Delete.
3029 (m68k_legitimize_call_address): New function.
3030 (print_operand): Remove the %o prefix. Handle the %p prefix.
3031 (output_call): New function.
3032 (m68k_output_mi_thunk): Use m68k_symbolic_jump. Always load the
3033 target address from the GOT if symbolic jumps are not allowed.
3034 * config/m68k/m68k.md (call, general_operand): Do not set
3035 SYMBOL_REF_FLAG. Use m68k_legitimize_call_address instead.
3036 Merge separate flag_pic and !flag_pic define_insns into...
3037 (*call, *call_value): ...these new patterns. Match the address
3038 rather than the containing MEM and require it to be a call_operand.
3039 Use output_call to generate the asm template.
3040 * config/m68k/predicates.md (const_call_operand): New predicate.
3041 (call_operand): Likewise.
3043 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
3045 * config/m68k/m68k.h (REGISTER_MOVE_COST): Simplify definition.
3046 (STACK_GROWS_DOWNWARD): Define to 1.
3047 (FUNCTION_VALUE, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P): Equivocate
3048 comments, emphasizing that these values are only defaults.
3049 * config/m68k/linux.h (LINK_SPEC): Fix formatting in #undef.
3050 * config/m68k/m68k.c (const_method): Remove trailing whitespace.
3052 2007-01-18 Richard Sandiford <richard@codesourcery.com>
3054 * config/m68k/m68k.md (cmpsi): Remove outdated flag_pic handling.
3056 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
3057 Richard Sandiford <richard@codesourcery.com>
3059 * config/m68k/m68k.h (DATA_REGNO_P, ADDRESS_REGNO_P, INT_REGNO_P)
3060 (FP_REGNO_P): New macros.
3061 (REGNO_OK_FOR_INDEX_P, REGNO_OK_FOR_BASE_P, REGNO_OK_FOR_DATA_P)
3062 (REGNO_OK_FOR_FP_P, REG_OK_FOR_INDEX_P, REG_OK_FOR_BASE_P): Use them.
3063 (EH_RETURN_STACKADJ_RTX): Use A0_REG.
3064 * config/m68k/m68k.c (m68k_regno_mode_ok): Use the new REGNO macros.
3066 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
3068 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
3069 (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-rtems*): Add t-floatlib
3071 * config/m68k/t-floatlib: New file, extracting common code from...
3072 * config/m68k/t-m68kbare, config/m68k/t-m68kelf,
3073 * config/m68k/t-uclinux: Here.
3074 * config/m68k/fpgnulib.c: Do not compile extendeed precision
3075 routines on ColdFire targets.
3077 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
3079 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
3080 (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*)
3081 (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-linux*)
3082 (m68k-*-rtems*): Use tm_file rather than m68k/m68k.h and
3083 explicitly set MOTOROLA to 1.
3084 * config/m68k/m68k.h (MOTOROLA): Simplify definition accordingly.
3086 2007-01-18 Richard Sandiford <richard@codesourcery.com>
3087 Nathan Sidwell <nathan@codesourcery.com>
3089 * config/m68k/m68k.h (PCC_STATIC_STRUCT_RETURN): Delete.
3090 (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Add MOTOROLA cases.
3091 * config/m68k/coff.h (REGISTER_PREFIX_MD): Delete.
3092 * config/m68k/m68020-elf.h (LIB_SPEC): Delete.
3093 * config/m68k/m68k-none.h (CC1_SPEC, CPP_SUBTARGET_SPEC): Delete.
3094 * config/m68k/m68kelf.h (IMMEDIATE_PREFIX, REGISTER_PREFIX_MD)
3095 (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Delete.
3096 (NO_DOLLAR_IN_LABEL, PCC_STATIC_STRUCT_RETURN): Don't undefine.
3097 (BSS_ASM_OP, ASM_OUTPUT_SKIP, ASM_OUTPUT_ASCII): Delete.
3098 * config/m68k/m68kemb.h (PCC_STATIC_STRUCT_RETURN): Don't undefine.
3099 * config/m68k/linux.h (SIZE_TYPE, PTRDIFF_TYPE, WCHAR_TYPE)
3100 (WCHAR_TYPE_SIZE, TARGET_OBJFMT_CPP_BUILTINS): Delete.
3101 (TARGET_OS_CPP_BUILTINS): Don't define mc68000 and mc68020 here.
3102 (DBX_REGISTER_NUMBER): Delete.
3103 * config/m68k/netbsd-elf.h (IMMEDIATE_PREFIX): Delete.
3104 (PCC_STATIC_STRUCT_RETURN): Don't undefine.
3105 * config/m68k/openbsd.h (PCC_STATIC_STRUCT_RETURN): Define.
3107 2007-01-18 Richard Sandiford <richard@codesourcery.com>
3109 * config.gcc (m68k-*-uclinux*): Add flat.h to $tm_file.
3110 * config/flat.h: New file.
3111 * crtstuff.c (USE_PT_GNU_EH_FRAME): Don't define if
3113 * config/m68k/m68k.h (ASM_PREFERRED_EH_DATA_FORMAT): Do not use
3114 indirect references for -msep-data or -mid-shared-library.
3115 Do not use PC-relative code addresses either.
3116 * config/m68k/m68k.c (override_options): Restrict -fPIC error
3118 * config/m68k/uclinux.h (STARTFILE_SPEC): Define. Use Scrt1.o
3119 for shared libraries and crt1.o for executables. Use crti.o and
3121 (ENDFILE_SPEC): Use crtend.o and crtn.o.
3122 (LIB_SPEC): Suppress -Rlibc.gdb if -static-libc is given.
3123 Do not add -elf2flt or -shared-lib-id options here.
3124 (LINK_SPEC): Define. Pass -elf2flt if no -elf2flt option is given.
3125 Pass -shared-lib-id if -mid-shared-library, taking the library
3126 identifier from -mshared-library-id if given, otherwise
3128 (EH_FRAME_IN_DATA_SECTION): Do not undefine.
3129 (INIT_SECTION_ASM_OP, FINI_SECTION_ASM_OP): Likewise.
3130 (TARGET_OS_CPP_BUILTINS): Define __GXX_MERGED_TYPEINFO_NAMES=0
3131 and __GXX_TYPEINFO_EQUALITY_INLINE=0 if -mid-shared-library.
3132 (DRIVER_SELF_SPECS): Map unadorned PIC options to -msep-data.
3133 * config/m68k/t-uclinux (EXTRA_MULTILIB_PARTS): Add crtbegin.o
3135 * config/m68k/lb1sf68.asm (PICCALL): Use an lea and pc-relative
3136 jump sequence for ISA A and ISA A+.
3137 (PICJUMP): Likewise.
3139 2007-01-18 Richard Sandiford <richard@codesourcery.com>
3141 * config.gcc (m68k-*-uclinux*): Base the port on the common
3142 and m68k GNU/Linux files rather than on the generic ELF ones.
3143 * config/m68k/uclinux.h (TARGET_VERSION): Override.
3144 (TARGET_OS_CPP_BUILTINS): Use LINUX_TARGET_OS_CPP_BUILTINS.
3146 2007-01-18 Julian Brown <julian@codesourcery.com>
3147 Richard Sandiford <richard@codesourcery.com>
3149 * config/m68k/m68k.h (LONG_DOUBLE_TYPE_SIZE): Make 64-bit on ColdFire.
3150 (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
3151 * config/m68k/netbsd-elf.h (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Undefine
3153 * config/m68k/uclinux-oldabi.h (LONG_DOUBLE_TYPE_SIZE): Redefine to
3155 (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
3157 2007-01-18 Richard Sandiford <richard@codesourcery.com>
3159 * doc/install.texi: Document m68k-uclinuxoldabi.
3160 * config.gcc (m68k-*-uclinuxoldabi): New configuration.
3161 * config/m68k/uclinux-oldabi.h: New file, copied from
3162 config/m68k/uclinux.h.
3164 2007-01-18 Kaz Kojima <kkojima@rr.iij4u.or.jp>
3166 * config/m32r/m32r-protos.h (m32r_expand_epilogue): Declare it.
3167 * config/m32r/m32r.c (m32r_setup_incoming_varargs): Use gen_frame_mem.
3168 (m32r_compute_frame_size): Use unsigned for regno.
3169 (m32r_reload_lr): Use gen_frame_mem.
3171 (m32r_output_function_epilogue): Don't output the function epilogue
3173 (m32r_expand_epilogue): New.
3174 (direct_return): Return false if the function has the interrupt
3176 (m32r_hard_regno_rename_ok): Remove code for the textual epilogue.
3177 * config/m32r/m32r.md (epilogue): New expander.
3178 (return_lr, return_rte): New insns.
3179 (return): Make it expander.
3180 (return_normal): New expander.
3182 2007-01-18 Josh Conner <jconner@apple.com>
3185 * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Add
3186 support for UNLE, UNLT, UNGE, and UNGT.
3188 2007-01-18 Jan Hubicka <jh@suse.cz>
3190 * tree-vrp.c (finalize_jump_threads): Do not call cleanup_cfg by hand.
3191 * tree-tailcall (add_virtual_phis): Likewise.
3192 (optimize_tail_call): Return TODOs.
3193 (execute_tail_calls): Return TODOs.
3194 * tree-ssa-ccp (execute_fold_all_builtins): Do cleanup_cfg via TODO.
3195 * tree-cfgcleanup.c (cleanup_tree_cfg_loop): Return if something
3197 * tree-ssa-forwprop.c (tree_ssa_forward_propagate_single_use_value):
3198 Cleanup cfg using TODO.
3199 * tree-flow.h (cleanup_tree_cfg_loop): Update prototype.
3200 * passes.c (execute_function_todo): When cleanup did something, remove
3202 * tree-cfg.c (pass_build_cfg): Add cleanup_cfg TODO.
3203 (make_edges): Don't cleanup_cfg.
3205 2007-01-18 Uros Bizjak <ubizjak@gmail.com>
3207 * reg-stack.c (subst_stack_regs_pat) [UNSPEC_SINCOS_COS,
3208 UNSPEC_XTRACT_FRACT]: Use generic code for instructions that
3209 operate on the top of stack.
3210 [UNSPEC_SINCOS_SIN, UNSPEC_XTRACT_EXP, UNSPEC_TAN]: Rewrite
3211 register handling of instructions that output to the second
3213 [UNSPEC_TAN_ONE, UNSPEC_TAN_TAN]: Remove.
3214 (move_for_stack_reg): Special-case check for dead destination
3215 stack slot for constant load of 1.0 inside UNSPEC_TAN.
3217 * config/i386/i386.md (UNSPEC_TAN): New constant.
3218 (UNSPEC_TAN_ONE, UNSPEC_TAN_TAN): Remove.
3219 (fptanxf4_i387, fptan_extend<mode>xf4_i387): New patterns
3220 to correctly model move of constant 1.0 to top stack slot.
3221 (*tandf3_1, *tansf3_1, *tanxf3_1): Remove insn patterns.
3222 (unnamed peephole2 pattern): Remove corresponding peephole2
3223 pattern that optimizes tan insn and loading of constant 1.0.
3224 (tanxf2): Use fptanxf4_i387.
3225 (tan<mode>2): Rename from tansf2 and tandf2 and macroize insn
3226 patterns using X87MODEF12 mode macro. Use fptan_extend<mode>xf4_i387
3227 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
3228 disable patterns for SSE math.
3229 (sincos<mode>3): Use truncxf<mode>2_i387_noop for truncation.
3230 (fyl2x_extend<mode>xf3_i387): Use X87MODEF12 for operand 1.
3232 2007-01-18 Dirk Mueller <dmueller@suse.de>
3233 Richard Guenther <rguenther@suse.de>
3236 * doc/invoke.texi (Warray-bounds): Document -Warray-bounds.
3237 * common.opt (Warray-bounds): Add new warning option.
3238 * c-opts.c (c_common_handle_option): Define -Warray-bounds
3240 * Makefile.in: make tree-vrp.o depend on toplev.h
3241 * tree-vrp.c (vrp_finalize): Call check_array_refs if -Warray-bounds
3243 (check_array_refs, check_array_bounds, check_array_ref): New.
3245 2007-01-18 Jan Hubicka <jh@suse.cz>
3247 * tree-ssa-ccp.c (ccp_finalize): Return if something changed.
3248 (execute_ssa_ccp): Return flags conditionally.
3249 * tree-ssa-propagate.c (substitue_and_fold): Return if something was
3251 * tree-ssa-propagate.h (substitute_and_fold): Update prototype.
3253 2007-01-18 Steven Bosscher <steven@gcc.gnu.org>
3255 * cfgcleanup.c (cleanup_cfg): Detect cfglayout mode and set
3256 the CLEANUP_CFGLAYOUT flag when in cfglayout mode.
3258 * Makefile.c (GTFILES): Add cfglayout.h.
3259 * gengtype.c (open_base_files): Likewise.
3260 * cfglayout.c (cfg_layout_function_footer,
3261 cfg_layout_function_header) Reindent to make gengtype happy.
3262 * cfglayout.h (cfg_layout_function_footer,
3263 cfg_layout_function_header): Add GTY(()) marker.
3265 * ifcvt.c (noce_try_sign_mask): Make sure INSN_B is non-null.
3267 2007-01-18 Ben Elliston <bje@au.ibm.com>
3269 * genautomata.c (write_automata): Include xstrerror output in the
3270 error message if writing the DFA description file fails.
3272 2007-01-17 H.J. Lu <hongjiu.lu@intel.com>
3274 * config/mips/mips-protos.h (mips_output_external): Make it
3276 * config/mips/iris.h (TARGET_ASM_EXTERNAL_LIBCALL): Removed.
3277 * config/mips/mips.c (irix_output_external_libcall): Likewise.
3278 (extern_list): Likewise.
3279 (extern_head): Likewise.
3280 (TARGET_ASM_FILE_END): Likewise.
3281 (mips_file_end): Likewise.
3282 (mips_output_external): Rewritten.
3284 2007-01-18 Ben Elliston <bje@au.ibm.com>
3286 * genpreds.c (write_insn_preds_c): Only write out the function
3287 body for regclass_for_constraint if we have register constraints.
3289 2007-01-17 Tom Tromey <tromey@redhat.com>
3291 * doc/sourcebuild.texi (libgcj Tests): Use sourceware.org.
3292 * doc/install.texi (Testing): Use sourceware.org.
3293 (Binaries): Likewise.
3294 (Specific): Likewise.
3295 * doc/contrib.texi (Contributors): Use sourceware.org.
3297 2007-01-17 Anatoly Sokolov <aesok@post.ru>
3299 * config/avr/avr.h (AVR_HAVE_LPMX): New macro.
3300 (AVR_ENHANCED): Rename to ...
3301 (AVR_HAVE_MUL): ... new.
3302 (avr_enhanced_p): Rename to ...
3303 (avr_have_mul_p): ... new.
3304 (TARGET_CPU_CPP_BUILTINS): Use 'avr_have_mul_p' instead of
3305 'avr_enhanced_p' for "__AVR_ENHANCED__". Define "__AVR_HAVE_MUL__".
3306 * config/avr/avr.c (avr_enhanced_p): Rename to ...
3307 (avr_have_mul_p): ... new.
3308 (base_arch_s): Rename 'enhanced' to 'have_mul'.
3309 (avr_override_options): Use 'avr_have_mul_p' and 'have_mul' instead of
3310 'avr_enhanced_p' and 'enhanced'.
3311 (ashlhi3_out, ashrhi3_out, lshrhi3_out, avr_rtx_costs): Use
3312 AVR_HAVE_MUL instead of AVR_ENHANCED.
3313 * avr.md (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
3314 (mulqi3, *mulqi3_enh, *mulqi3_call, mulqihi3, umulqihi3, mulhi3,
3315 *mulhi3_enh, *mulhi3_call, mulsi3, *mulsi3_call): Use AVR_HAVE_MUL
3316 instead of AVR_ENHANCED.
3317 (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
3318 * libgcc.S: Use __AVR_HAVE_MUL__ instead of __AVR_ENHANCED__.
3319 (__tablejump__): Use __AVR_HAVE_LPMX__ instead of __AVR_ENHANCED__.
3321 2007-01-17 Ian Lance Taylor <iant@google.com>
3323 * vec.h (VEC_reserve_exact): Define.
3324 (vec_gc_p_reserve_exact): Declare.
3325 (vec_gc_o_reserve_exact): Declare.
3326 (vec_heap_p_reserve_exact): Declare.
3327 (vec_heap_o_reserve_exact): Declare.
3328 (VEC_OP (T,A,reserve_exact)): New static inline function, three
3330 (VEC_OP (T,A,reserve)) [all versions]: Remove handling of
3332 (VEC_OP (T,A,alloc)) [all versions]: Call ...reserve_exact.
3333 (VEC_OP (T,A,copy)) [all versions]: Likewise.
3334 (VEC_OP (T,a,safe_grow)) [all versions]: Likewise.
3335 * vec.c (calculate_allocation): Add exact parameter. Change all
3337 (vec_gc_o_reserve_1): New static function, from vec_gc_o_reserve.
3338 (vec_gc_p_reserve, vec_gc_o_reserve): Call vec_gc_o_reserve_1.
3339 (vec_gc_p_reserve_exact, vec_gc_o_reserve_exact): New functions.
3340 (vec_heap_o_reserve_1): New static function, from vec_heap_o_reserve.
3341 (vec_heap_p_reserve, vec_heap_o_reserve): Call vec_heap_o_reserve_1.
3342 (vec_heap_p_reserve_exact): New function.
3343 (vec_heap_o_reserve_exact): New function.
3345 2007-01-17 Jan Hubicka <jh@suse.cz>
3347 * ipa-type-escape.c (look_for_casts): Revamp using handled_component_p.
3349 2007-01-17 Eric Christopher <echristo@apple.com>
3351 * config.gcc: Support core2 processor.
3353 2007-01-16 Jan Hubicka <jh@suse.cz>
3355 * tree-ssanames.c (release_dead_ssa_names): Instead of ggc_freeing
3356 the names, just unlink the chain so we don't crash on dangling pointers
3359 2007-01-16 Jan Hubicka <jh@suse.cz>
3361 * cgraph.h (cgraph_decide_inlining_incrementally): Kill.
3362 * tree-pass.h: Reorder to make IPA passes appear toegher.
3363 (pass_early_inline, pass_inline_parameters, pass_apply_inline): Declare.
3364 * cgraphunit.c (cgraph_finalize_function): Do not compute inling
3365 parameters, do not call early inliner.
3366 * ipa-inline.c: Update comments. Include tree-flow.h
3367 (cgraph_decide_inlining): Do not compute inlining parameters.
3368 (cgraph_decide_inlining_incrementally): Return TODOs; assume to
3369 be called with function context set up.
3370 (pass_ipa_inline): Remove unreachable functions before pass.
3371 (cgraph_early_inlining): Simplify assuming to be called from the
3373 (pass_early_inline): New pass.
3374 (cgraph_gate_ipa_early_inlining): New gate.
3375 (pass_ipa_early_inline): Turn into simple wrapper.
3376 (compute_inline_parameters): New function.
3377 (gate_inline_passes): New gate.
3378 (pass_inline_parameters): New pass.
3379 (apply_inline): Move here from tree-optimize.c
3380 (pass_apply_inline): New pass.
3381 * ipa.c (cgraph_remove_unreachable_nodes): Verify cgraph after
3383 * tree-inline.c (optimize_inline_calls): Return TODOs rather than
3385 (tree_function_versioning): Do not allocate dummy struct function.
3386 * tree-inline.h (optimize_inline_calls): Update prototype.
3387 * tree-optimize.c (execute_fixup_cfg): Export.
3388 (pass_fixup_cfg): Remove
3389 (tree_rest_of_compilation): Do not apply inlines.
3390 * tree-flow.h (execute_fixup_cfg): Declare.
3391 * Makefile.in (gt-passes.c): New.
3392 * passes.c: Include gt-passes.h
3393 (init_optimization_passes): New passes.
3394 (nnodes, order): New static vars.
3395 (do_per_function_toporder): New function.
3396 (execute_one_pass): Dump current pass here.
3397 (execute_ipa_pass_list): Don't dump current pass here.
3399 2007-01-16 Janis Johnson <janis187@us.ibm.com>
3401 * config/dfp-bit.c (dfp_compare_op): Return separate value for NaN.
3402 (DFP_NE, DFP_LE, DFP_GE): Return false for NaN.
3404 2007-01-16 David Edelsohn <edelsohn@gnu.org>
3406 * config/rs6000/darwin-ldouble.c: Build file for SOFT_FLOAT.
3407 (strong_alias): Define.
3408 (__gcc_qmul): Provide non-FMA for soft-float.
3414 (__gcc_qunord): New.
3424 * config/rs6000/rs6000.c (rs6000_init_libfuncs): Initialize
3425 soft-float functions.
3426 * config/rs6000/libgcc-ppc-glibc.ver: Version soft-float symbols.
3427 * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Do not warn
3428 about long double soft float.
3430 2007-01-16 Dorit Nuzman <dorit@il.ibm.com>
3431 Tehila Meyzels <tehila@il.ibm.com>
3433 * tree-vectorizer.h (is_pattern_stmt_p): New.
3434 * tree-vect-analyze.c (vect_determine_vectorization_factor): Fix
3435 formatting (tabs instead of spaces). Cleanup and clarify setting
3436 of STMT_VINFO_VECTYPE. Call is_pattern_stmt_p.
3437 * tree-vect-transform.c (vect_get_vec_def_for_operand): Fix typo.
3438 (vectorizable_type_demotion): Check that types are integral.
3439 (vectorizable_type_promotion): Likewise.
3440 (vectorizable_store): Fix typo. Eliminate new-line at end of
3443 2007-01-16 Jan Hubicka <jh@suse.cz>
3445 * tree-ssanames.c (release_dead_ssa_names): Remove invalidated
3448 2007-01-15 Eric Christopher <echristo@apple.com>
3450 * ifcvt.c: Include vec.h, vecprim.h.
3451 (check_cond_move_block): New argument regs.
3452 Reorganize. Add registers used to regs.
3453 (cond_move_process_if_block): Use regs set above as
3456 2007-01-15 Eric Christopher <echristo@apple.com>
3458 * config/darwin.h: Update copyright.
3459 (TARGET_OPTION_TRANSLATE_TABLE): Add umbrella.
3460 (LINK_COMMAND_SPEC): Add -u.
3461 (LINK_SPEC): Fix umbrella for above.
3463 2007-01-15 Joseph S. Myers <joseph@codesourcery.com>
3465 * config/soft-fp/op-common.h, config/soft-fp/op-4.h: Update from
3468 2007-01-15 Tom Tromey <tromey@redhat.com>
3470 * doc/sourcebuild.texi (libgcj Tests): Don't mention jacks.
3471 * doc/install.texi (Testing): Don't mention jacks.
3472 (Configuration): Document --enable-java-maintainer-mode. Move
3473 --with-java-home to libgcj-specific section. Document
3475 (Prerequisites): Mention --enable-java-maintainer-mode, ecj1.
3477 2007-01-15 Jan Hubicka <jh@suse.cz>
3479 * tree-ssa-dce.c (DCE_TODOs): New.
3480 (propagate_necessity): Return if something changed.
3481 (eliminate_unnecessary_stmts): Likewise.
3482 (perform_tree_ssa_dce): Return TODO flags when needed.
3483 (pass_dce, pass_dce_loop, pass_cd_dce): Remove TODO flags.
3485 2007-01-15 Uros Bizjak <ubizjak@gmail.com>
3487 * config/i386/i386.md (fyl2xxf3_i387): Rename from fyl2x_xf3.
3488 (fyl2x_extend<mode>xf3_i387): New insn pattern.
3489 (log<mode>2): Rename from logsf2 and logdf2 and macroize insn
3490 patterns using X87MODEF12 mode macro. Extend operand 1
3491 to XFmode. Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
3492 (log10<mode>2): Ditto.
3493 (log2<mode>2): Ditto.
3494 (log1p<mode>2): Ditto.
3495 (logb<mode>2): Ditto.
3496 (fyl2xp1xf3_i387): Rename from fyl2xp1_xf3.
3497 (fyl2xp1_extend<mode>xf3_i387): New insn pattern.
3498 (*fxtractxf3_i387): Rename from *fxtractxf3.
3499 (fxtract_extend<mode>xf3_i387): New insn pattern.
3500 (ilogbsi2): Use match_dup 3, not match_operand:XF 3.
3501 * config/i386/i386.c (ix86_emit_i387_log1p): Use gen_fyl2xp1xf3_i387()
3502 and gen_fyl2xxf3_i387().
3504 2007-01-14 Zdenek Dvorak <dvorakz@suse.cz>
3506 * loop-unswitch.c (unswitch_loop): Do not call fix_loop_placement.
3507 * cfgloopmanip.c (fix_loop_placement): Made static. Use
3508 get_loop_exit_edges. Changed return type to bool.
3509 * cfgloop.h (fix_loop_placement): Declaration removed.
3511 2007-01-14 Dorit Nuzman <dorit@il.ibm.com>
3513 * param.h (MIN_VECT_LOOP_BOUND): New.
3514 * params.def (MIN_VECT_LOOP_BOUND): New.
3515 * tree-vectorizer.c (slpeel_tree_peel_loop_to_edge): Takes another
3516 argument - minimum threshold for number of iterations.
3517 * tree-vectorizer.h (slpeel_tree_peel_loop_to_edge): Add another
3518 argument to declaration.
3519 * tree-vect-analyze.c (vect_analyze_operations): Check value of
3520 MIN_VECT_LOOP_BOUND.
3521 * tree-vect-transform.c (vect_do_peeling_for_loop_bound): Call
3522 slpeel_tree_peel_loop_to_edge with additional argument.
3523 (vect_do_peeling_for_alignment): Likewise.
3524 * doc/invoke.texi (min-vect-loop-bound): Document new param option.
3526 2007-01-14 Uros Bizjak <ubizjak@gmail.com>
3529 * config/i386/i386.c (print_operand) ['z']: Output 'b' for
3532 2007-01-14 Jan Hubicka <jh@suse.cz>
3534 * tree-dfa.c (remove_referenced_var): New function.
3535 * tree-ssa-live.c (remove_unused_locals): Walk referenced vars and
3536 prune referenced vars list too.
3537 * tree-flow.h (remove_referenced_var): Declare.
3539 2007-01-14 Jan Hubicka <jh@suse.cz>
3541 * tree-eh.c (add_stmt_to_eh_region_fn): Do not add call_exprs
3543 (remove_stmt_from_eh_region_fn): Do not remove call_exprs.
3544 (verify_eh_throw_stmt_node, verify_eh_throw_table_statements): Kill.
3545 * except.h (verify_eh_throw_table_statements): Kill prototype.
3546 * cfgexpand.c (expand_gimple_basic_block): Propagate Eh regions
3547 into call exrepssions.
3548 * tree-optimize.c (execute_free_cfg_annotatiosn): Do not call
3550 * tree-cfg.c: Include pointer-set.h.
3551 (verify_node_sharing): Work on pointer set.
3552 (verify_eh_throw_stmt_node): New.
3553 (verify_stmts): Use pointers sets, verify throw_stmt.
3555 2007-01-13 Zdenek Dvorak <dvorakz@suse.cz>
3557 * ipa-reference.c (analyze_function): Consider also addresses taken
3560 2007-01-12 Roger Sayle <roger@eyesopen.com>
3562 * c-typeck.c (null_pointer_constant_p): Replace use of
3563 TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
3564 (build_c_cast): Likewise.
3566 2007-01-12 Roger Sayle <roger@eyesopen.com>
3568 * tree.h (force_fit_type_double): Remove unused final argument.
3569 * c-common.c (constant_expression_warning): Replace use of
3570 TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
3571 (convert_and_check): Likewise.
3572 (shorten_compare): Update call to force_fit_type_double.
3573 (c_common_truthvalue_conversion) <INTEGER_CST>: Use integer_zerop.
3574 * convert.c (convert_to_pointer): Update call to
3575 force_fit_type_double.
3576 * fold-const.c (force_fit_type_double): Remove overflowed_const
3578 (int_const_binop, fold_convert_const_int_from_int,
3579 fold_convert_const_int_from_real, fold_div_compare,
3580 fold_sign_changed_comparison, fold_unary, fold_negate_const,
3581 fold_abs_const, fold_not_const): Remove the final argument from
3582 calls to force_fit_type_double.
3584 2007-01-12 Andrew Pinski <andrew_pinski@playstation.sony.com>
3586 * configure.ac: Set insn to "nop" for spu-*-* also.
3587 * configure: Regenerate.
3589 2007-01-12 Olga Golovanevsky <olga@il.ibm.com>
3591 * builtins.def : Add BUILT_IN_FREE.
3593 2007-01-12 Jan Hubicka <jh@suse.cz>
3595 PR tree-optimization/30443
3596 * tree-inline.c (tree_function_versioning): Do not optimize when
3597 cloning for inlining.
3599 2007-01-12 Zdenek Dvorak <dvorakz@suse.cz>
3601 * doc/loop.texi: Document recording of loop exits.
3602 * cfgloopmanip.c (loopify, duplicate_loop): Use alloc_loop.
3603 (update_single_exits_after_duplication,
3604 update_single_exit_for_duplicated_loop,
3605 update_single_exit_for_duplicated_loops): Removed.
3606 (duplicate_loop_to_header_edge): Do not call
3607 update_single_exits_after_duplication and
3608 update_single_exit_for_duplicated_loops.
3609 (loop_version): Do not update single_exit information.
3610 (fix_loop_structure): Use record_loop_exits instead of
3611 mark_single_exit_loops.
3612 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update
3613 the lists of loop exits.
3614 * cfghooks.c (redirect_edge_and_branch, redirect_edge_and_branch_force,
3615 split_edge, merge_blocks): Update the lists of loop exits.
3616 * modulo-sched.c (sms_schedule): Pass LOOPS_HAVE_RECORDED_EXITS to
3617 loop_optimizer_init.
3618 * loop-init.c (loop_optimizer_init): Call record_loop_exits instead
3619 of mark_single_exit_loops.
3620 (loop_optimizer_finalize): Call release_recorded_exits.
3621 * tree-ssa-loop.c (tree_loop_optimizer_init): Pass
3622 LOOPS_HAVE_RECORDED_EXITS to loop_optimizer_init.
3623 * tree-vectorizer.c (slpeel_tree_duplicate_loop_to_edge_cfg): Do not
3624 update single exit information.
3625 * lambda-code.c (perfect_nestify): Ditto.
3626 * cfgloop.c (flow_loop_free): Destroy the list of exits of the loop.
3627 (mark_single_exit_loops): Removed.
3628 (alloc_loop, loop_exit_hash, loop_exit_eq, loop_exit_free,
3629 get_exit_descriptions, rescan_loop_exit, record_loop_exits,
3630 dump_recorded_exit, dump_recorded_exits, release_recorded_exits): New
3632 (get_loop_exit_edges, single_exit): Use recorded exit lists.
3633 (add_bb_to_loop, remove_bb_from_loops): Update the lists of loop exits.
3634 (verify_loop_structure): Verify consistency of the exit lists.
3635 (flow_loops_find): Use alloc_loop. Initialize exits hash.
3636 (set_single_exit): Removed.
3637 * cfgloop.h (struct loop_exit): New function.
3638 (struct loop): single_exit_ field replaced by exits field.
3639 (LOOPS_HAVE_MARKED_SINGLE_EXITS): Replaced by LOOPS_HAVE_RECORDED_EXITS.
3640 (struct loops): Added exits hash.
3641 (mark_single_exit_loops, set_single_exit): Declaration removed.
3642 (release_recorded_exits, record_loop_exits, rescan_loop_exit): Declare.
3644 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3646 * doc/invoke.texi: Avoid use of @headitem.
3648 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3650 * cse.c (cse_insn): Move HAVE_CC0 code after declarations.
3652 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3654 * doc/install.texi: Fix m68k-*-* anchor and add m68k-*-* to the
3657 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
3658 Richard Sandiford <richard@codesourcery.com>
3660 * doc/invoke.texi: Document -mno-bitfield, -mno-rtd and -mno-short.
3661 * config/m68k/m68k.opt: Resort options.
3662 (mbitfield, mrtd, mshort): Remove RejectNegative properties.
3664 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
3665 Richard Sandiford <richard@codesourcery.com>
3667 * doc/invoke.texi: Document the macros that are defined by
3668 m68k's -mtune and -mhard-float options.
3669 * config/m68k/m68k-protos.h (m68k_cpp_cpu_ident) Declare.
3670 (m68k_cpp_cpu_family): Likewise.
3671 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Add a full set
3672 of __ucfv*__ macros. Define __mcffpu__ if generating code for
3673 ColdFire FPUs. Define __mcf_cpu_* and __mcf_family_* macros.
3674 * config/m68k/m68k.c (m68k_cpp_cpu_ident): New function.
3675 (m68k_cpp_cpu_family): Likewise.
3677 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3679 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Treat all mc68*
3680 macros besides mc68000 as tuning macros. Use a switch statement
3681 to set them and mcpu32.
3683 2007-01-12 Julian Brown <julian@codesourcery.com>
3685 * config/m68k/m68k.h: Use TARGET_68040 instead of TARGET_68040_ONLY.
3686 (TARGET_68040_ONLY): Rename to...
3687 (TARGET_68040): ...this.
3688 * config/m68k/m68k.c: Use TARGET_68040 instead of TARGET_68040_ONLY.
3689 * config/m68k/m68k.md: Likewise.
3691 2007-01-12 Julian Brown <julian@codesourcery.com>
3692 Nathan Sidwell <nathan@codesourcery.com>
3693 Richard Sandiford <richard@codesourcery.com>
3695 * config.gcc (m680[012]0-*-*, m68k*-*-*): Set m68k_cpu_ident to
3696 the -mcpu= argument associated with the --with-cpu setting.
3697 Define M68K_DEFAULT_TUNE to the default -mtune= option,
3698 if different from the one implied by the -mcpu setting.
3699 Accept --with-cpu=FOO if FOO is listed in m68k-devices.def,
3700 using mcpu=FOO as the default CPU option. Set target_cpu_default2.
3701 * doc/invoke.texi: Mention ColdFire in the introduction to the
3702 m68k options. Document the new -march, -mcpu, -mtune, -mdiv,
3703 -mno-div and -mhard-float options. Make -m68881 a synonym for
3704 -mhard-float. Document the previously-undocumented -m5206e,
3705 -m528x, -m5307 and -m5407 options. Tweak the existing option
3706 documentation for consistency.
3707 * doc/install.texi: Mention new --with-cpu arguments.
3708 * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Only use the
3709 default CPU if neither -mcpu nor -march are specified.
3710 (ASM_CPU_SPEC): Pass down -mcpu and -march options.
3711 (TARGET_CPU_CPP_BUILTINS): Set __mcfisa*__ macros from
3712 TARGET_ISA*. Set the legacy __mcf*__ cpu macros in the same way,
3713 using m68k_tune to decide between families that implement the
3714 same ISA. Use m68k_tune to set __mcfv4e__.
3715 (FL_BITFIELD, FL_68881, FL_COLDFIRE, FL_CF_HWDIV, FL_CF_MAC)
3716 (FL_CF_EMAC, FL_CF_EMAC_B, FL_CF_USP, FL_CF_FPU, FL_ISA_68000)
3717 (FL_ISA_68010, FL_ISA_68020, FL_ISA_68040, FL_ISA_A, FL_ISA_B)
3718 (FL_ISA_C, FL_ISA_MMU): New macros.
3719 (MASK_COLDFIRE): Delete.
3720 (TARGET_68010, TARGET_68020, TARGET_68040_ONLY, TARGET_COLDFIRE)
3721 (TARGET_ISAB): Redefine in terms of m68k_cpu_flags.
3722 (TARGET_68881, TARGET_COLDFIRE_FPU): Redefine in terms of m68k_fpu.
3723 (TARGET_HARD_FLOAT): Do not define here.
3724 (TARGET_ISAAPLUS, TARGET_ISAC): New macros.
3725 (TUNE_68000): New macro.
3726 (TUNE_68000_10): Redefine in terms of TUNE_68000 and TUNE_68010.
3727 (TUNE_68010, TUNE_68030, TUNE_68040, TUNE_68060, TUNE_CPU32)
3728 (TUNE_CFV2): Redefine in terms of m68k_tune.
3729 (uarch_type, target_device, fpu_type): New enums.
3730 (m68k_cpu, m68k_tune, m68k_fpu, m68k_cpu_flags): Declare.
3731 * config/m68k/m68k.c (TARGET_DEFAULT): Remove MASK_68881.
3732 (FL_FOR_isa_00, FL_FOR_isa_10, FL_FOR_isa_20, FL_FOR_isa_40)
3733 (FL_FOR_isa_cpu32, FL_FOR_isa_a, FL_FOR_isa_aplus, FL_FOR_isa_b)
3734 (FL_FOR_isa_c): New macros.
3735 (m68k_isa): New enum.
3736 (m68k_target_selection): New structure.
3737 (all_devices, all_isas, all_microarchs): New tables.
3738 (m68k_cpu_entry, m68k_arch_entry, m68k_tune_entry, m68k_cpu)
3739 (m68k_tune, m68k_fpu, m68k_cpu_flags): New variables.
3740 (MASK_ALL_CPU_BITS): Delete.
3741 (m68k_find_selection): New function.
3742 (m68k_handle_option): Handle -mcpu=, -march= and -mtune=.
3743 Map the legacy target options to a combination of the new ones.
3744 (override_options): Set m68k_cpu, m68k_tune, m68k_fpu and
3745 m68k_cpu_flags. Handle M68K_DEFAULT_TUNE. Use m68k_cpu_flags
3746 to derive default MASK_BITFIELD, MASK_CF_HWDIV and MASK_HARD_FLOAT
3748 * config/m68k/m68k.opt (m5200, m5206e, m528x, m5307, m5407, mcfv4e)
3749 (m68010, m68020, m68020-40, m68020-60, m68030, m68040): Remove Mask
3751 (m68881, msoft-float): Change mask from 68881 to HARD_FLOAT.
3752 (march=, mcpu=, mdiv, mhard-float, mtune=): New options.
3753 * config/m68k/m68k-devices.def: New file.
3755 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3756 Nathan Sidwell <nathan@codesourcery.com>
3758 * config/m68k/m68k.h (ASM_CPU_SPEC, ASM_SPEC, EXTRA_SPECS)
3759 (SUBTARGET_EXTRA_SPECS): New macros.
3760 * config/m68k/linux.h (ASM_SPEC): Remove CPU flags;
3761 use %(asm_cpu_spec) instead.
3762 * config/m68k/m68k-none.h (ASM_SPEC): Likewise.
3763 * config/m68k/openbsd.h (ASM_SPEC): Likewise.
3764 * config/m68k/netbsd-elf.h (ASM_SPEC): Likewise.
3765 (EXTRA_SPECS): Rename to...
3766 (SUBTARGET_EXTRA_SPECS): ...this.
3768 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
3769 Richard Sandiford <richard@codesourcery.com>
3770 Julian Brown <julian@codesourcery.com>
3772 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
3773 (m68k-*-uclinux*, m68k-*-linux*, m68k-*-rtems*): Set default_m68k_cpu
3774 to the configuration's default CPU.
3775 (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*): Likewise.
3776 Remove default masks.
3777 (m680[012]0-*-*): Set the default with_cpu to the first part of
3779 (m68k*-*-*): Set the default with_cpu to m$default_m68k_cpu.
3780 (m68k*-*-linux): Extend the --with-cpu handling to...
3781 (m680[012]0-*-*, m68k*-*-*): ...these configurations. Allow m68000
3782 and m68010. Don't set target_cpu_default2.
3783 * doc/install.texi: Document --with-cpu for m68k.
3784 * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Define.
3785 * config/m68k/m68k-none.h (TARGET_CPU_DEFAULT, M68K_CPU_m68k)
3786 (M68K_CPU_m68000, M68K_CPU_m68010, M68K_CPU_m68020, M68K_CPU_m68030)
3787 (M68K_CPU_m68040, M68K_CPU_m68302, M68K_CPU_m68332, TARGET_DEFAULT)
3788 (ASM_CPU_DEFAULT_SPEC, CC1_CPU_DEFAULT_SPEC): Delete.
3789 (ASM_SPEC): Remove use of %(asm_cpu_default).
3790 (EXTRA_SPECS, SUBTARGET_EXTRA_SPECS, MULTILIB_DEFAULTS): Delete.
3791 * config/m68k/linux.h (TARGET_DEFAULT): Delete.
3792 (CPP_SPEC): Merge definitions. Do not handle __HAVE_68881__ here.
3793 * config/m68k/netbsd-elf.h (TARGET_OS_CPP_BUILTINS): Define
3794 __HAVE_FPU__ if TARGET_HARD_FLOAT.
3795 (TARGET_DEFAULT): Delete.
3796 (EXTRA_SPECS): Delete cpp_cpu_default_spec, cpp_cpu_spec,
3797 cpp_fpu_spec, asm_default_spec and netbsd_cpp_spec.
3798 (CPP_CPU_SPEC): Delete.
3799 (TARGET_VERSION): Merge definitions, using TARGET_68010 to pick
3800 the appropriate string.
3801 (CPP_CPU_DEFAULT_SPEC, ASM_DEFAULT_SPEC, CPP_FPU_SPEC): Delete.
3802 (CPP_SPEC): Define to NETBSD_CPP_SPEC.
3803 (ASM_SPEC): Don't use %(asm_default_spec).
3804 * config/m68k/m68k.c (TARGET_DEFAULT_TARGET_FLAGS): Remove
3805 TARGET_DEFAULT and add MASK_68881.
3806 * config/m68k/m68k.md: Remove mention of TARGET_DEFAULT from comments.
3808 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3810 * config.gcc (m68010-*-netbsdelf*): Add MASK_68010.
3811 (m68k*-*-netbsdelf*, m68k*-*-openbsd*, m68k*-linux*): Add
3812 MASK_68010 alongside MASK_68020.
3813 * doc/invoke.texi: Document -m68010.
3814 * config/m68k/m68k.opt (m68010): New.
3815 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Define mc68010
3817 (TUNE_68010): New macro.
3818 * config/m68k/m68k-none.h (M68K_CPU_m68k, M68K_CPU_m68010)
3819 (M68K_CPU_m68020, M68K_CPU_m68030, M68K_CPU_m68040)
3820 (M68K_CPU_m68332): Add MASK_68010.
3821 * config/m68k/linux.h (TARGET_DEFAULT): Add MASK_68010 to
3822 fallback definition.
3823 * config/m68k/netbsd-elf.h (CPP_CPU_SPEC): Remove now-redundant
3825 * config/m68k/m68k.c (MASK_ALL_CPU_BITS): Add MASK_68010.
3826 (m68k_handle_option): Handle OPT_m68010. Add MASK_68010
3827 to all entries that use MASK_68020.
3828 (output_move_simode_const, output_move_himode, output_move_qimode)
3829 (output_move_stricthi, output_move_strictqi): Use TARGET_68010
3830 instead of TARGET_68020 to select clr behavior. Remove comment
3831 about there being no TARGET_68010.
3832 * config/m68k/m68k.md: Likewise throughout.
3834 2007-01-12 Julian Brown <julian@codesourcery.com>
3836 * config/m68k/m68k.h (TARGET_ISAB): New macro.
3837 * config/m68k/m68k.c: Use TARGET_ISAB rather than TARGET_CFV4.
3838 * config/m68k/m68k.md: Likewise.
3840 2007-01-12 Julian Brown <julian@codesourcery.com>
3842 * config/m68k/m68k.h (LEGITIMATE_INDEX_P, LEGITIMIZE_ADDRESS): Use
3843 TARGET_COLDFIRE_FPU instead of TARGET_CFV4E.
3845 2007-01-12 Julian Brown <julian@codesourcery.com>
3847 * config/m68k/m68k.h (TUNE_68040_60): New macro.
3848 * config/m68k/m68k.c (standard_68881_constant_p): Use it.
3849 * config/m68k/m68k.md: Likewise.
3851 2007-01-12 Julian Brown <julian@codesourcery.com>
3852 Richard Sandiford <richard@codesourcery.com>
3854 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Use TUNE_68030
3855 instead of TARGET_68030, TUNE_68040 instead of TARGET_68040,
3856 TUNE_68060 instead of TARGET_68060 and TUNE_CPU32 instead of
3858 (TARGET_CPU32): Rename to...
3859 (TUNE_CPU32): ...this.
3860 (TUNE_68000_10, TUNE_68030, TUNE_68040, TUNE_68060)
3861 (TUNE_CFV2): New macros.
3862 * config/m68k/netbsd-elf.h (LONG_DOUBLE_TYPE_SIZE): Simplify;
3863 remove conditions that are implied by TARGET_68020.
3864 * config/m68k/m68k.c (m68k_output_function_prologue): Use TUNE_68040
3865 instead of TARGET_68040 and TUNE_CPU32 instead of TARGET_CPU32.
3866 (m68k_output_function_epilogue): Likewise.
3867 (m68k_rtx_costs): Likewise. Use TUNE_68060 instead of TARGET_68060
3868 and TUNE_CFV2 instead of TARGET_5200. Use TUNE_68000_10 instead of
3869 "!TARGET_68020 && !TARGET_COLDFIRE" to choose between 68000 and
3870 non-68000 timings. Refactor multiplication and division costs.
3871 (output_addsi3): Use TUNE_68040 instead of TARGET_68040 and
3872 TUNE_CPU32 instead of TARGET_CPU32.
3873 (standard_68881_constant_p): Use TUNE_68040 instead of TARGET_68040
3874 and TUNE_68060 instead of TARGET_68060.
3875 * config/m68k/m68k.md: Use TUNE_68040 instead of TARGET_68040,
3876 TUNE_68060 instead of TARGET_68060, and TUNE_CPU32 instead of
3878 (movsi_const0): Use TUNE_68000_10 rather than "!TARGET_68020
3879 && !TARGET_COLDFIRE" to choose between moveq and clr.
3880 Likewise in the unnamed movsf pattern.
3881 (ashlsi_17_24, lshrsi_17_24): Guard with TUNE_68000_10 rather than
3882 "!TARGET_68020 && !TARGET_COLDFIRE". Likewise the unnamed
3885 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3887 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Increase amount
3888 of tabbing before backslashes.
3890 2007-01-11 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
3892 * pa-linux.h (ASM_OUTPUT_INTERNAL_LABEL): Undefine.
3893 * pa.h (ASM_OUTPUT_LABEL): Output colon when using GAS.
3894 (ASM_OUTPUT_INTERNAL_LABEL): Define.
3896 2007-01-11 Zdenek Dvorak <dvorakz@suse.cz>
3898 * tree-ssa-loop-ivopts.c (extract_cond_operands): Split from
3899 find_interesting_uses_cond.
3900 (find_interesting_uses_cond): Use extract_cond_operands.
3901 (rewrite_use_compare): Use extract_cond_operands and
3902 force_gimple_operand_bsi. Do not call update_stmt.
3903 (determine_use_iv_cost_condition): Use extract_cond_operands.
3904 Return cheaper of using original bound and changing the exit bound.
3906 2007-01-11 Zdenek Dvorak <dvorakz@suse.cz>
3908 PR tree-optimization/29516
3909 * tree-ssa-address.c (tree_mem_ref_addr, add_to_parts,
3910 most_expensive_mult_to_index, addr_to_parts,
3911 create_mem_ref, maybe_fold_tmr): Make the type of
3912 fields of TARGET_MEM_REF sizetype.
3913 (move_fixed_address_to_symbol, move_pointer_to_base):
3915 * tree.def (TARGET_MEM_REF): Add comment on types of
3918 2007-01-11 Joseph Myers <joseph@codesourcery.com>
3920 * c-common.c (vector_types_convertible_p): Treat opaque types as
3921 always convertible if they have the same size, but not otherwise.
3923 2007-01-11 Steven Bosscher <steven@gcc.gnu.org>
3925 * ifcvt.c (struct noce_if_info): Add comments to the fields.
3926 Remove the b_unconditional field.
3927 (noce_try_sign_mask): Do not look at b_unconditional.
3928 (noce_process_if_block): Do not use merge_if_blocks. Update
3929 the CFG here. Do not set b_unconditional.
3930 (cond_move_process_if_block): Likewise.
3931 (find_cond_trap): Likewise.
3932 (check_cond_move_block): Require simple jump insns at the end
3935 2007-01-11 Jan Hubicka <jh@suse.cz>
3937 PR tree-optimization/1046
3938 * tree-tailcall.c (suitable_for_tail_call_opt_p): Use TREE_ADDRESSABLE
3939 when alias info is not ready.
3940 (pass_tail_recursion): Do not require aliasing.
3941 * tree-ssa-copyrename.c (pass_rename_ssa_cop): Likewise.
3942 * tree-ssa-ccp.c (pass_ccp, pass_fold_builtins): Likewise.
3943 * tree-ssa-copy.c (pass_copy_prop): Likewise.
3944 * tree-ssa-forwprop.c (pass_forwprop): Likewise.
3945 * tree-ssa-dce.c (pass_dce, pass_dce_loop, pass_cd_dce): Likewise.
3946 * passes.c (init_optimization_passes): Execute rename_ssa_copies,
3947 ccp, forwprop, copy_prop, merge_phi, copy_prop, dce and tail recursion
3949 * tree-ssa-operands.c (add_virtual_operand, get_indirect_ref_operand):
3950 When aliasing is not build, mark statement as volatile.
3952 2007-01-11 Tom Tromey <tromey@redhat.com>
3954 PR preprocessor/15185, PR preprocessor/20989:
3955 * doc/cppopts.texi <-MT>: Update description of algorithm for
3956 computing default target.
3957 <-M, -MD>: Reword "basename" text.
3959 2007-01-11 Roger Sayle <roger@eyesopen.com>
3961 * builtins.c (expand_builtin_pow, expand_builtin_powi,
3962 fold_builtin_cabs, fold_builtin_sqrt, fold_builtin_trunc,
3963 fold_builtin_floor, fold_builtin_ceil, fold_builtin_round,
3964 fold_builtin_int_int_roundingfn, fold_builtin_bitop,
3965 fold_builtin_bswap, real_constp, fold_builtin_pow,
3966 fold_builtin_powi, fold_builtin_signbit, fold_builtin_copysign,
3967 do_mpfr_arg1, do_mpfr_arg2, do_mpfr_arg3, do_mpfr_sincos): Replace
3968 uses of the macro TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
3969 * convert.c (convert_to_pointer): Likewise.
3970 * expr.c (highest_pow2_factor, expand_expr_real_1): Likewise.
3971 * fold-const.c (force_fit_type, fold_negate_expr, int_const_binop,
3972 const_binop, fold_convert_const_int_from_int,
3973 fold_convert_const_int_from_real,
3974 fold_convert_const_real_from_real, sign_bit_p,
3975 optimize_minmax_comparison, extract_muldiv_1, fold_div_compare,
3976 fold_sign_changed_comparison, fold_unary, fold_comparison,
3977 fold_binary, multiple_of_p, tree_Expr_non_zero_p,
3978 fold_negate_const, fold_abs_const, fold_not_const): Likewise.
3979 * print-tree.c (print_node_brief, print_node): Likewise.
3980 * stor-layout.c (place_field, layout_type): Likewise.
3981 * tree-chrec.c (keep_cast): Likewise.
3982 * tree.c (build_vector, build_real, build_real_from_int_cst,
3983 build_complex): Likewise.
3985 2007-01-11 Roger Sayle <roger@eyesopen.com>
3987 * tree.h (TREE_CONSTANT_OVERFLOW): Obsolete. For the time being,
3988 treat TREE_CONSTANT_OVERFLOW as a synonym of TREE_OVERFLOW.
3990 2007-01-11 Paolo Bonzini <bonzini@gnu.org>
3992 * configure.ac (strict1_warn): Rename to strict_warn.
3993 (WERROR, --enable-werror, symlink hacks, stage1_cflags,
3994 cc_set_by_configure, quoted_cc_set_by_configure,
3995 stage_prefix_set_by_configure, quoted_stage_prefix_set_by_configure,
3996 all_boot_languages, all_stagestuff): Remove.
3997 (target_list): Remove bootstrap targets.
3998 * Makefile.in (quickstrap): Unconditionally make a synonym of all.
3999 (BOOT_LANGUAGES, STAGE1_CFLAGS, STAGE1_CHECKING,
4000 REMAKEFLAGS, FLAGS_TO_PASS, PREPEND_DOTDOT_TO_RELATIVE_PATHS,
4001 SUBDIR_FLAGS_TO_PASS, WERROR_FLAGS, STRICT2_WARN, LANG_STAGESTUFF,
4002 VOL_FILES, POSTSTAGE1_FLAGS_TO_PASS, STAGE2_FLAGS_TO_PASS,
4003 STAGEPROFILE_FLAGS_TO_PASS, STAGEFEEDBACK_FLAGS_TO_PASS, stage1_build,
4004 stage1_copy, stage2_build, stage2_copy, stageprofile_build,
4005 stageprofile_copy, stage3_build, stage3_copy, stagefeedback_build,
4006 stagefeedback_copy, stage4_build, clean_s1, clean_sw, bootstrap,
4007 bootstrap-lean, bootstrap2, bootstrap2-lean, bootstrap3,
4008 bootstrap3-lean, bootstrap4, bootstrap4-lean, unstage1, unstage2,
4009 unstage3, unstage4, unstageprofile, unstagefeedback, restage, restage2,
4010 restage3, restage4, restageprofile, restagefeedback, bubbleestrap,
4011 cleanstrap, unstrap, restrap, *compare, *compare3, *compare4,
4012 *compare-lean, *compare3-lean, *compare4-lean, stage1-start, stage1,
4013 stage2-start, stage2, stage3-start, stage3, stage4-start, stage4,
4014 stageprofile-start, stageprofile, stagefeedback-start, stagefeedback,
4015 risky-stage1, risky-stage2, risky-stage3, risky-stage4): Remove.
4016 (ORDINARY_FLAGS_TO_PASS): Rename to FLAGS_TO_PASS.
4017 (STAGECOPYSTUFF, STAGEMOVESTUFF): Consolidate into MOSTLYCLEANFILES.
4018 (mostlyclean): Adjust.
4019 (clean, distclean): Don't mention bootstrap stuff.
4020 * configure: Regenerate.
4021 * ada/config-lang.in, cp/config-lang.in, forttran/config-lang.in,
4022 java/config-lang.in, objc/config-lang.in, objcp/config-lang.in,
4023 treelang/config-lang.in (stagestuff): Remove.
4024 * doc/sourcebuild.texi (stage1, stage2, stage3, stage4,
4025 stageprofile, stagefeedback, stagestuff): Remove mention.
4027 2007-01-11 Nick Clifton <nickc@redhat.com>
4029 * config/mcore/predicates.md (mcore_general_movesrc_operand):
4031 (mcore_general_movdst_operand): Do not accept CONST_INTs.
4032 (mcore_arith_K_S_operand): Run the test for the S constraint not
4033 the test for the M constraint.
4034 (mcore_addsub_operand): Do not accept integer values that are
4035 larger than 32 bits.
4036 * config/mcore/mcore.md: Remove unused constraints from split.
4037 (andsi3): Use HOST_WIDE_INT instead of int to hold an INTVAL.
4039 (allocate_stack): Likewise.
4040 * config/mcore/mcore.c (mcore_print_operand): Restrict output of P
4041 operands to 32 bits.
4042 (mcore_const_costs): Use HOST_WIDE_INT instead of int to hold an
4044 (mcore_and_cost, mcore_modify_comparison, const_ok_for_mcore,
4045 mcore_const_ok_for_inline, mcore_const_trick_uses_not,
4046 try_constant_tricks, mcore_num_ones, mcore_num_zeros,
4047 mcore_output_bclri, mcore_output_andn, output_inline_const,
4048 mcore_output_move, mcore_output_movedouble): Likewise.
4049 (mcore_output_cmov): Use CONST_OK_FOR_M and CONST_OK_FOR_N.
4050 (output_inline_const): Likewise.
4051 (output_inline_const): Fix format strings used in sprintf
4053 * config/mcore/mcore-protos.h: Update prototypes for changed
4054 functions in mcore.c.
4055 * config/mcore/mcore.h (CONST_OK_FOR_I): Cast values to
4056 HOST_WIDE_INT and not int.
4057 (CONST_OK_FOR_J, CONST_OK_FOR_K, CONST_OK_FOR_L, CONST_OK_FOR_M,
4058 CONST_OK_FOR_N): Likewise.
4059 (LEGITIMATE_CONSTANT_P): Also check CONSTANT_P.
4060 (GO_IF_LEGITIMATE_INDEX): Use HOST_WIDE_INT instead of int to hold
4063 2007-01-10 Jan Hubicka <jh@suse.cz>
4065 * tree-vrp.c (remove_range_assertions): Release defs.
4066 * tree-ssa-loop-ivopts.c (rmeove_statement): Likewise.
4067 * tree-ssa-dom.c (remove_stmt_or_phi): Likewise.
4069 2007-01-10 Paul Brook <paul@codesourcery.com>
4071 * config/arm/arm.c (arm_rtx_costs_1): Handle mutiply-subtract.
4072 * config/arm/arm.md (mulsi3subsi): New insn.
4074 2007-01-10 Zdenek Dvorak <dvorakz@suse.cz>
4076 * tree-ssa-loop-manip.c (tree_unroll_loop): Make it a wrapper over ...
4077 (tree_transform_and_unroll_loop): New.
4078 * tree-flow.h (transform_callback, tree_transform_and_unroll_loop):
4081 2007-01-10 Robert Kennedy <jimbob@google.com>
4083 * fold-const.c (fold_comparison): Fold comparisons like (x *
4084 1000 < 0) to (x < 0).
4086 2007-01-10 Ian Lance Taylor <iant@google.com>
4088 * tree-pretty-print.c (dump_generic_node): Print parentheses when
4089 operands have the same priority.
4091 2007-01-10 Tom Tromey <tromey@redhat.com>
4093 * fold-const.c (fold_truthop): Don't check can_use_bit_fields_p.
4094 (fold_binary): Likewise.
4095 * langhooks.c (lhd_can_use_bit_fields_p): Removed.
4096 * langhooks-def.h (lhd_can_use_bit_fields_p): Removed.
4097 (LANG_HOOKS_CAN_USE_BIT_FIELDS_P): Removed.
4098 (LANG_HOOKS_INITIALIZER): Remove LANG_HOOKS_CAN_USE_BIT_FIELDS_P.
4099 * langhooks.h (struct lang_hooks): Removed field
4100 'can_use_bit_fields_p'.
4102 2007-01-10 Ralf Corsépius <ralf.corsepius@rtems.org>
4104 * config/bfin/t-bfin, config/bfin/t-bfin-elf: Remove GCC_CFLAGS.
4106 2007-01-10 Razya Ladelsky <razya@il.ibm.com>
4108 * function.c (get_last_funcdef_no): New function.
4109 * function.h (get_last_funcdef_no): Declare.
4110 * tree-inline.c (initialize_cfun): Add initialization.
4111 (tree_function_versioning): Cleanup.
4113 2007-01-10 Jan Hubicka <jh@suse.cz>
4115 * tree-inline.c (setup_one_parameter): Do not propagate into abnormal
4118 2007-01-10 Sa Liu <saliu@de.ibm.com>
4119 Ben Elliston <bje@au.ibm.com>
4121 * spu.h (STACK_SAVE_AREA): Use VOIDmode for SAVE_FUNCTION, SImode
4122 for SAVE_NONLOCAL and Pmode for any other save level.
4123 * spu-protos.h (spu_restore_stack_block): Declare.
4124 * spu.md (save_stack_block): Remove.
4125 (restore_stack_block): Call spu_restore_stack_block.
4126 * spu.c (spu_restore_stack_block): New function.
4127 (spu_expand_epilogue): Remove old comment.
4129 2007-01-09 Zdenek Dvorak <dvorakz@suse.cz>
4131 PR tree-optimization/30322
4132 * tree-ssa-loop-ivopts.c (fold_affine_expr, iv_value): Removed.
4133 (cand_value_at): Return the value as aff_tree.
4134 (may_eliminate_iv): Convert the bound from aff_tree to tree.
4135 * tree-affine.c (aff_combination_add_cst, aff_combination_add_product,
4136 aff_combination_mult): New functions.
4137 (aff_combination_add): Use aff_combination_add_cst.
4138 (aff_combination_convert): Allow conversions to a wider type.
4139 (tree_to_aff_combination): Handle BIT_NOT_EXPR.
4140 * tree-affine.h (aff_combination_mult): Declare.
4142 2007-01-09 Carlos O'Donell <carlos@codesourcery.com>
4144 * doc/tm.texi: Update documentation to reflect reality of exec
4145 and start file search behaviours. Update copyright year.
4146 * doc/invoke.texi: Explain how GCC_EXEC_PREFIX is used to find
4147 header file directories.
4149 2007-01-09 Uros Bizjak <ubizjak@gmail.com>
4151 * config/i386/i386.md (*sinxf2): Rename to *sinxf2_i387.
4152 (*cosxf2): Rename to cosxf2_i387.
4153 (*sindf2, *sinsf2): Extend operand 1 to XFmode. Macroize patterns
4154 using X87MODEF12 mode macro. Rename patterns to
4155 *sin_extend<mode>xf2_i387. Use SSE_FLOAT_MODE_P to disable patterns
4157 (*cosdf2, *cossf2): Ditto.
4158 (sincosdf3, sincossf3): Ditto. Rewrite corresponding splitters
4159 to match extended input operands.
4160 (sincos<mode>3): New expander.
4161 (*sinextendsfdf2, *cosextendsfdf2, *sincosextendsfdf3): Remove
4162 insn patterns and corresponding splitters.
4164 2007-01-09 Kaz Kojima <kkojima@gcc.gnu.org>
4166 * config/sh/t-linux (TARGET_LIBGCC2_CFLAGS): Delete.
4167 (SHLIB_MAPFILES, SHLIB_LINK, SHLIB_INSTALL): Likewise.
4169 2007-01-09 Nicolas Pitre <nico@cam.org>
4172 * arm/ieee754-df.S (Lad_s): Also test the low word of X for zero.
4174 2007-01-08 Geoffrey Keating <geoffk@apple.com>
4176 * target.h (struct gcc_target): New field library_rtti_comdat.
4177 * target-def.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): New.
4178 (TARGET_CXX): Add TARGET_CXX_LIBRARY_RTTI_COMDAT.
4179 * doc/tm.texi (C++ ABI): Document TARGET_CXX_LIBRARY_RTTI_COMDAT.
4180 * config/darwin.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
4182 2007-01-08 Geoffrey Keating <geoffk@apple.com>
4184 * doc/invoke.texi (Optimize Options): Correct description of -O0.
4186 2007-01-08 Richard Guenther <rguenther@suse.de>
4188 * tree.h (force_fit_type_double): Export.
4189 (force_fit_type): Remove.
4190 * fold-const.c (force_fit_type_double): New function.
4191 (force_fit_type): Remove.
4192 (int_const_binop): Use it.
4193 (fold_convert_const_int_from_int): Likewise.
4194 (fold_convert_const_int_from_real): Likewise.
4195 (fold_div_compare): Likewise.
4196 (fold_sign_changed_comparison): Likewise.
4197 (fold_unary): Likewise.
4198 (fold_negate_const): Likewise.
4199 (fold_abs_const): Likewise.
4200 (fold_not_const): Likewise.
4201 * c-common.c (shorten_compare): Use force_fit_type_double.
4202 * convert.c (convert_to_pointer): Likewise.
4204 2007-01-08 Richard Guenther <rguenther@suse.de>
4206 * tree.h (build_int_cst_wide_type): Export.
4207 * tree.c (build_int_cst_wide_type): New function.
4208 (build_int_cst_wide): Fix comment.
4209 * builtins.c (fold_builtin_object_size): Use build_int_cst
4210 to build -1 or 0 of the correct type. Use fit_double_type
4211 to check for overflow.
4212 * fold-const.c (optimize_bit_field_compare): Use build_int_cst_type
4214 (decode_field_reference): Likewise.
4215 (all_ones_mask_p): Likewise.
4216 (native_interpret_int): Use build_int_cst_wide_type.
4217 (fold_binary): Use build_int_cst_type to build an all-ones
4219 * stor-layout.c (set_sizetype): Use build_int_cst_wide_type.
4221 2007-01-08 Daniel Jacobowitz <dan@codesourcery.com>
4223 * config/pa/t-pa64 (libgcc_stub.a): Use $(T).
4225 2007-01-09 Ben Elliston <bje@au.ibm.com>
4227 * genautomata.c (STATS_OPTION): New option.
4228 (stats_flag): New flag.
4229 (gen_automata_option): Handle it.
4230 (initiate_automaton_gen): Ditto.
4231 (write_automata): Output statistics only if stats_flag is
4232 set. Likewise, output time statistics only if time_flag is set.
4233 * doc/md.texi (Processor pipeline description): Document new flag.
4235 2007-01-08 Richard Guenther <rguenther@suse.de>
4237 * builtins.c (fold_builtin_int_roundingfn): Use fit_double_type.
4238 * tree.c (build_int_cst_type): Likewise.
4239 (size_in_bytes): Don't call force_fit_type on the result.
4240 (int_fits_type_p): Use fit_double_type.
4241 * fold-const.c (fit_double_type): New function.
4242 (force_fit_type): Use it.
4243 * tree.h (fit_double_type): Export.
4245 2007-01-08 Jan Hubicka <jh@suse.cz>
4247 * tree-vectorizer.c (gate_increase_alignment): Fix return type.
4248 * ipa.c (function_and_variable_visibility): Fix return type.
4250 2007-01-08 Richard Guenther <rguenther@suse.de>
4252 * tree-ssa-ccp.c (maybe_fold_offset_to_array_ref): Use type
4253 of offset to build the index.
4254 * tree-pretty-print.c (dump_generic_node): Don't build negated
4255 const just for printing.
4256 * c-pretty-print.c (pp_c_integer_constant): Likewise.
4257 * builtins.c (fold_builtin_int_roundingfn): Check if result
4258 fits the type by using force_fit_type and comparing the result.
4259 * predict.c (predict_loops): Use compare_tree_int for comparison.
4260 * tree.c (build_int_cst): Fall back to integer_type_node for
4262 (build_int_cst_wide): Assert type is non-null.
4264 2007-01-08 Roberto Costa <roberto.costa@st.com>
4266 * tree-vrp.c (extract_range_from_cond_expr): New.
4267 (extract_range_from_expr): Handle COND_EXPR nodes used as expressions.
4268 * tree-ssa-ccp.c (get_maxval_strlen): Handle COND_EXPR nodes used
4270 (fold_stmt): Bug fix, avoid infinite recursion when folding COND_EXPRs.
4271 * tree-ssa-forwprop.c (simplify_cond, forward_propagate_into_cond,
4272 tree_ssa_forward_propagate_single_use_vars): Handle COND_EXPR nodes
4273 used as expressions.
4274 * tree-object-size.c (cond_expr_object_size): New.
4275 (collect_object_sizes_for): Handle COND_EXPR nodes used as expressions.
4277 2007-01-08 Jan Hubicka <jh@suse.cz>
4279 * tree-ssa-forwprop.c (forward_propagate_into_cond,
4280 tree_ssa_forward_propagate_single_use_va): Release defs of propagated
4283 2007-01-08 Richard Guenther <rguenther@suse.de>
4285 PR tree-optimization/23603
4286 * tree-vrp.c (set_value_range_to_truthvalue): New function.
4287 (extract_range_from_binary): Fall back to truthvalue instead of
4288 varying for TRUTH_*_EXPR.
4289 (extract_range_from_comparison): Fall back to truthvalue instead of
4291 (vrp_visit_phi_node): Don't adjust new range bounds to +INF/-INF
4292 if all visited PHI values were constant.
4294 2007-01-08 Jan Hubicka <jh@suse.cz>
4296 * cgraphunit.c (cgraph_process_new_functions): Reset reachable flag.
4297 (cgraph_analyze_function): break out from ...
4298 (cgraph_finalize_compilation_unit): ... here.
4299 (cgraph_expand_function): Remove forgoten commented out line.
4300 (cgraph_optimize): Analyze functions.
4302 2007-01-08 Jan Hubicka <jh@suse.cz>
4304 * tree-pas.h (TODO_remove_function): New flag.
4305 (TODO_update*): Renumber.
4306 (pass_ipa_increase_alignment,
4307 pass_ipa_function_and_variable_visibility): New passes.
4308 * cgraphunit.c (cgraph_increase_alignment): Move to tree-vectorizer.c
4309 (cgraph_function_and_variable_visibility): Move to ipa.c
4310 (cgraph_optimize): Don't call cgraph_function_and_variable_visibility,
4311 cgraph_increase_alignment.
4312 * ipa-inline.c (cgraph_decide_inlining): Don't push timevar.
4313 (cgraph_decide_inlining_incrementally): Push TV_INTEGRATION before
4314 calling tree-inline.
4315 (cgraph_early_inlining): Do not call cgraph_remove_unreachable_nodes.
4316 (pass_ipa_inline, pass_early_ipa_inlining): Set TODO_remove_functions
4317 * tree-vectorizer.c (increase_alignment): Move here from cgraphunit.c
4318 (gate_increase_alignment): New function.
4319 (pass_ipa_increase_alignment): New pass.
4320 * ipa.c: Inline tree-pass.h and timevar.h
4321 (function_and_variable_visibility): Move here from cgraphunit.c
4322 * tree-optimize.c (pass_early_local_passes): Add TODO_remove_functions.
4323 * passes.c (init_optimization_passes): Add the two new passes.
4324 (execute_todo): Handle cgraph_remove_functions.
4326 2007-01-08 Nick Clifton <nickc@redhat.com>
4328 * config/frv/predicates.md (reg_or_0_operand): Accept
4331 2007-01-08 Ralf Corsépius <ralf.corsepius@rtems.org>
4333 * config/bfin/rtems.h, config/bfin/t-rtems: New.
4334 * config.gcc: Add bfin*-rtems*.
4336 2007-01-08 Mark Shinwell <shinwell@codesourcery.com>
4338 * c.opt: Add -flax-vector-conversions.
4339 * c-typeck.c (convert_for_assignment): Pass flag to
4340 vector_types_convertible_p to allow emission of note.
4341 (digest_init): Likewise.
4342 * c-opts.c: Handle -flax-vector-conversions.
4343 * c-common.c (flag_lax_vector_conversions): New.
4344 (vector_types_convertible_p): Unless -flax-vector conversions
4345 has been passed, disallow conversions between vectors with
4346 differing numbers of subparts and/or element types. If such
4347 a conversion is disallowed, possibly emit a note on the first
4348 occasion only to inform the user of -flax-vector-conversions.
4349 The new last argument specifies this.
4350 * c-common.h (flag_lax_vector_conversions): New.
4351 (vector_types_convertible_p): Add extra argument.
4352 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Use
4353 char_type_node for V*QI type vectors.
4354 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins):
4355 Update to satisfy new typechecking rules.
4356 * config/rs6000/altivec.h (vec_cmple): Use vec_cmpge, for both
4358 * doc/invoke.texi (C Dialect Options): Document
4359 -flax-vector-conversions.
4361 2007-01-08 Mark Shinwell <shinwell@codesourcery.com>
4363 PR tree-optimization/29877
4364 * tree-ssa-ter.c (is_replaceable_p): Deem assignments with
4365 a register variable on the RHS to not be replaceable.
4367 2007-01-08 Chen Liqin <liqin@sunnorth.com.cn>
4368 * config/score/t-score-elf (MULTILIB_OPTIONS): Change.
4369 * config/score/predicates.md (const_uimm5, sr0_operand, const_simm12,
4370 const_simm15, const_pow2, const_npow2): Added.
4371 * config/score/misc.md (insv, extv, extzv, movmemsi,
4372 move_lbu_a/b, mov_lhu_a/b etc): Added and fix some bug.
4373 * config/score/score.c (score_address_cost, score_select_cc_mode):
4375 Change CONST_OK_FOR_LETTER_P/EXTRA_CONSTRAINT define.
4376 Update score_rtx_costs for MACRO TARGET_RTX_COSTS.
4377 Update score_print_operand.
4378 * config/score/score.h (DATA_ALIGNMENT, SELECT_CC_MODE): Added.
4379 Adjust register allocate order and update some macro define.
4380 * config/score/score-mdaux.c (mdx_unaligned_load, mdx_unsigned_store,
4381 mdx_block_move_straight, mdx_block_move_loop_head,
4382 mdx_block_move_loop_body, mdx_block_move_loop_foot, mdx_block_move_loop,
4383 mdx_block_move): Added.
4384 (mdx_movsicc, mdp_select_add_imm, mdp_select, mds_zero_extract_andi,
4385 mdp_limm): Updated and fix some bug and typo.
4386 * config/score/score.md (movqi/hi/si, add/sub/zero/ext): Updated.
4387 (movsf, movdf, doloop_end): Added.
4389 2007-01-08 Kazu Hirata <kazu@codesourcery.com>
4391 * config/arm/arm.c, config/arm/arm.h, config/arm/arm.md,
4392 config/arm/thumb2.md: Fix comment typos.
4393 * doc/extend.texi: Fix a typo.
4395 2007-01-07 Eric Christopher <echristo@apple.com>
4397 * configure.ac: Check for __stack_chk_fail for darwin.
4398 * configure: Regenerate.
4400 2007-01-07 Richard Guenther <rguenther@suse.de>
4402 * tree-vrp.c (extract_range_from_assert): CSE calls to
4403 compare_values where possible.
4404 (extract_range_from_unary_expr): Likewise.
4406 2007-01-07 Anatoly Sokolov <aesok@post.ru>
4408 * config/avr/avr-protos.h (call_insn_operand): Delete prototype.
4409 * config/avr/avr.c (call_insn_operand): Delete function.
4410 * config/avr/avr.md (*pushqi, *pushhi, *pushsi, *pushsf): Use REG_SP
4411 instead of register number. Use predicates.
4412 * config/avr/predicates.md (const0_operand, reg_or_0_operand,
4413 call_insn_operand): Add.
4415 2007-01-06 Jan Hubicka <jh@suse.cz>
4417 * tree-pass.h (pass_build_cgraph_edges): Declare.
4418 * cgraphunit.c (record_refernece): Move to cgraphbuild.c
4419 (visited_nodes): Remove.
4420 (cgraph_create_edges): Move to cgraphbuild.c; rename to
4421 build_cgrpah_edges; make visited_nodes local.
4422 (cgraph_process_new_functions): DO not call initialize_inline_failed.
4423 (record_references_in_initializer): Move to cgraphbuild.c
4424 (initialize_inline_failed, rebuild_cgraph_edges,
4425 pass_rebuild_cgraph_edges): Move to cgraphbuild.c.
4426 (verify_cgraph_node): Make visited_nodes local.
4427 (cgraph_analyze_function): Do not call cgraph_create_edges and
4428 initialize_inline_failed.
4429 (cgraph_expand_function): Do not call cgraph_lower_function;
4430 assert that function is already lowered.
4431 * Makefile.in (cgraphbuild.o): New.
4432 * passes.c (init_optimization_passes): Add pass_build_cgraph_edges
4433 at the end of lowering passes.
4435 2007-01-06 Steven Bosscher <steven@gcc.gnu.org>
4437 * ifcvt.c (cond_move_convert_if_block): New function, code
4438 factored out from...
4439 (cond_move_process_if_block): ...here. Call the new function
4440 on the THEN and ELSE blocks.
4441 (merge_if_block): Do not copy global_live_at_end, merge_blocks
4442 already takes care of this.
4444 2007-01-05 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
4447 * tree.h (TREE_OVERFLOW_P): New.
4448 * c-typeck.c (parser_build_unary_op): Warn only if result
4449 overflowed and operands did not.
4450 (parser_build_binary_op): Likewise.
4451 (convert_for_assignment): Remove redundant overflow_warning.
4452 * c-common.c (overflow_warning): Don't check or set TREE_OVERFLOW.
4454 2007-01-05 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
4456 * c-typeck.c (store_init_value): Split over two lines to follow
4457 the GNU coding style.
4459 2007-01-05 Benjamin Kosnik <bkoz@redhat.com>
4461 * c-cppbuiltin.c (c_cpp_builtins): __GXX_EXPERIMENTAL_CPP0X__ to
4462 __GXX_EXPERIMENTAL_CXX0X__.
4463 * doc/cpp.texi: Same.
4465 2007-01-05 Richard Guenther <rguenther@suse.de>
4468 * tree.c (get_narrower): Do not construct COMPONENT_REFs
4469 with mismatched types. Instead explicitly build a
4470 conversion NOP_EXPR.
4472 2007-01-05 Ian Lance Taylor <iant@google.com>
4474 * c-common.c (decl_with_nonnull_addr_p): New function.
4475 (c_common_truthvalue_conversion): Call it.
4476 * c-typeck.c (build_binary_op): Likewise.
4477 * c-common.h (decl_with_nonnull_addr_p): Declare.
4479 2007-01-05 Jakub Jelinek <jakub@redhat.com>
4482 * libgcc2.c (__divdc3): Compare c and d against 0.0 instead of
4485 2007-01-05 Joel Brobecker <brobecker@adacore.com>
4487 * doc/install.texi (Final install): Document the fact that
4488 the GNAT runtime should not be stripped.
4490 2007-01-04 Jan Hubicka <jh@suse.cz>
4492 * tree-inline.c (fold_marked_statements): Update operand caches
4493 and EH after folding
4495 2007-01-04 Ian Lance Taylor <iant@google.com>
4497 * c-common.c (check_function_nonnull): Whitespace fix.
4499 2007-01-04 Jan Hubicka <jh@suse.cz>
4501 * tree-optimize.c (execute_fixup_cfg): Correct previously mistakely
4502 comitted older version of patch.
4503 (pass_fixup_cfg): Add TODOs to verify flow and statements, dump
4504 function, celanup cfg and collect garbage.
4506 2007-01-04 Mike Stump <mrs@apple.com>
4508 * Makefile.in (mostlyclean): Don't remove libgcc anymore.
4511 2007-01-04 Eric Christopher <echristo@apple.com>
4513 * libgcc2.c (__bswapsi2): Use SItype.
4514 (__bswapdi2): Use DItype.
4515 * libgcc2.h: Update for above.
4517 2007-01-04 Paul Brook <paul@codesourcery.com>
4519 * config/arm/arm.md (arm_mulsi3, thumb_mulsi3, mulsi3_compare0,
4520 mulsi_compare0_scratch, mulsi3addsi, mulsi3addsi_compare0,
4521 mulsi3addsi_compare0_scratch, mulsidi3adddi, mulsidi3,
4522 umulsidi3, umulsidi3adddi, smulsi3_highpart,
4523 umulsi3_highpart): Make conditional on !arm_arch6.
4524 (arm_mulsi3_v6, thumb_mulsi3_v6, mulsi3_compare0_v6,
4525 mulsi_compare0_scratch_v6, mulsi3addsi_v6, mulsi3addsi_compare0_v6,
4526 mulsi3addsi_compare0_scratch_v6, mulsidi3adddi_v6, mulsidi3_v6,
4527 umulsidi3_v6, umulsidi3adddi_v6, smulsi3_highpart_v6,
4528 umulsi3_highpart_v6): New insns.
4530 2007-01-04 Roger Sayle <roger@eyesopen.com>
4532 * fold-const.c (fold_convert): When casting an expression to void,
4533 fold_ignored_result may discover a GIMPLE_MODIFY_STMT which doesn't
4534 have a type. Instead of attempting to build a NOP_EXPR, return
4535 these "special" trees directly.
4537 2007-01-04 Joseph Myers <joseph@codesourcery.com>
4539 * config/rs6000/rs6000.c (rs6000_rtx_costs): Make adjustment for
4540 MULT inside MINUS as either argument. Use rs6000_cost->dmul -
4541 rs6000_cost->fp not 0 as adjustment for outer NEG.
4543 2007-01-04 Jan Hubicka <jh@suse.cz>
4545 * cgraph.c (cgraph_release_function_body): New function.
4546 (cgraph_remove_node): Use it.
4547 * cgraph.h (cgraph_release_function_body): Declare.
4548 * cgraphunit.c (cgraph_expand_function): Use it.
4549 * ipa.c (cgraph_remove_unreahchable_nodes): Use it.
4550 * tree-ssa.c (delete_tree_ssa): Allow to be called before aliasing
4551 is initialized and while compilation of other function is running.
4552 * tree-optimize.c (execute_free_cfg_annotations): Move code to clear
4553 statement CFG annotations from here to ...
4554 * tree-cfg.c (delete_tree_cfg_annotations): ... here.
4556 2007-01-04 Zdenek Dvorak <dvorakz@suse.cz>
4558 * cfgloop.h (enum li_flags): Make the constants powers of two.
4560 2007-01-04 Jan Hubicka <jh@suse.cz>
4562 * tree-inline.c (copy_bb): Insert new statements to statements_to_fold
4564 (fold_marked_statements): New function.
4565 (optimize_inline_calls, tree_function_versioning): Fold new statements.
4566 * tree-inline.h (copy_body_data): Add statements_to_fold.
4568 2007-01-03 Daniel Jacobowitz <dan@codesourcery.com>
4570 * config.gcc: Mention libgcc/config.host.
4571 * Makefile.in: Update comments mentioning libgcc.
4572 (LIBGCC, INSTALL_LIBGCC, GCC_PARTS, mklibgcc): Delete.
4573 (all.cross, start.encap, rest.encap, rest.cross): Update
4574 dependencies for libgcc move.
4575 (libgcc.mk, LIBGCC_DEPS, libgcov.a, libgcc.a, stmp-multilib)
4576 (clean-target, clean-target-libgcc): Delete.
4577 (srcdirify, GCC_EXTRA_PARTS): New macros.
4578 (libgcc-support, libgcc.mvars): New rules.
4579 (distclean): Remove mention of mklibgcc.
4580 (install): Don't reference INSTALL_LIBGCC.
4581 (install-common): Don't reference EXTRA_PARTS.
4582 (install-libgcc, install-multilib): Delete rules.
4583 * mklibgcc.in: Delete file.
4584 * doc/configfiles.texi: Don't mention mklibgcc.
4586 * config/i386/t-darwin (SHLIB_VERPFX): Delete (moved to libgcc).
4587 * config/i386/t-darwin64 (SHLIB_VERPFX): Likewise.
4588 * config/rs6000/t-darwin (SHLIB_VERPFX): Likewise.
4589 * config/rs6000/t-ppccomm (TARGET_LIBGCC2_CFLAGS, SHLIB_MAPFILES)
4590 (mklibgcc, ldblspecs): Likewise.
4592 * config/i386/t-nwld (libgcc.def, libc.def, libpcre.def)
4593 (posixpre.def): Use $(T).
4594 (SHLIB_EXT, SHLIB_NAME, SHLIB_SLIBDIR_QUAL, SHLIB_DEF, SHLIB_MAP)
4595 (SHLIB_SRC, SHLIB_INSTALL): Delete.
4596 (SHLIB_LINK): Make dummy.
4597 * config/t-slibgcc-darwin: Delete contents except for dummy SHLIB_LINK.
4599 * config/frv/t-linux (EXTRA_MULTILIB_PARTS): Clear.
4601 * config/alpha/t-crtfm: Use $(T) in rules for EXTRA_PARTS.
4602 * config/alpha/t-vms, config/alpha/t-vms64, config/fr30/t-fr30,
4603 config/i386/t-rtems-i386, config/ia64/t-ia64, config/rs6000/t-beos,
4604 config/rs6000/t-newas, config/sparc/t-elf: Likewise.
4606 * configure.ac (all_outputs): Remove mklibgcc.
4607 * configure: Regenerated.
4609 2007-01-03 Josh Conner <jconner@apple.com>
4612 * calls.c (compute_argument_addresses): Set stack and stack_slot
4613 for partial args, too.
4614 (store_one_arg): Use locate.size.constant for the size when
4615 generating a save_area.
4617 2007-01-03 Robert Kennedy <jimbob@google.com>
4619 * tree-cfg.c (tree_merge_blocks): Release SSA_NAME phi results
4620 whose definitions are deleted due to basic block merging.
4622 2007-01-03 Paul Brook <paul@codesourcery.com>
4625 * config/arm/arm.c (output_return_instruction): Pop PC in interrupt
4627 (use_return_insn): Return 0 for Thumb interrupt functions.
4628 (print_multi_reg): Add rfe argument for IRQ returns.
4629 (arm_output_epilogue): Pop interrupt return address directly into PC.
4630 (arm_expand_prologue): Only adjust IRQ return address in Arm mode.
4632 2007-01-03 Paul Brook <paul@codesourcery.com>
4634 Merge from sourcerygxx-4_1.
4635 * config/arm/thumb2.md: New file.
4636 * config/arm/elf.h (JUMP_TABLES_IN_TEXT_SECTION): Return True for
4638 * config/arm/coff.h (JUMP_TABLES_IN_TEXT_SECTION): Ditto.
4639 * config/arm/aout.h (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
4640 (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump tables.
4641 * config/arm/aof.h (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump
4643 (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
4644 * config/arm/ieee754-df.S: Use macros for Thumb-2/Unified asm
4646 * config/arm/ieee754-sf.S: Ditto.
4647 * config/arm/arm.c (thumb_base_register_rtx_p): Rename...
4648 (thumb1_base_register_rtx_p): ... to this.
4649 (thumb_index_register_rtx_p): Rename...
4650 (thumb1_index_register_rtx_p): ... to this.
4651 (thumb_output_function_prologue): Rename...
4652 (thumb1_output_function_prologue): ... to this.
4653 (thumb_legitimate_address_p): Rename...
4654 (thumb1_legitimate_address_p): ... to this.
4655 (thumb_rtx_costs): Rename...
4656 (thumb1_rtx_costs): ... to this.
4657 (thumb_compute_save_reg_mask): Rename...
4658 (thumb1_compute_save_reg_mask): ... to this.
4659 (thumb_final_prescan_insn): Rename...
4660 (thumb1_final_prescan_insn): ... to this.
4661 (thumb_expand_epilogue): Rename...
4662 (thumb1_expand_epilogue): ... to this.
4663 (arm_unwind_emit_stm): Rename...
4664 (arm_unwind_emit_sequence): ... to this.
4665 (thumb2_legitimate_index_p, thumb2_legitimate_address_p,
4666 thumb1_compute_save_reg_mask, arm_dwarf_handle_frame_unspec,
4667 thumb2_index_mul_operand, output_move_vfp, arm_shift_nmem,
4668 arm_save_coproc_regs, thumb_set_frame_pointer, arm_print_condition,
4669 thumb2_final_prescan_insn, thumb2_asm_output_opcode, arm_output_shift,
4670 thumb2_output_casesi): New functions.
4671 (TARGET_DWARF_HANDLE_FRAME_UNSPEC): Define.
4672 (FL_THUMB2, FL_NOTM, FL_DIV, FL_FOR_ARCH6T2, FL_FOR_ARCH7,
4673 FL_FOR_ARCH7A, FL_FOR_ARCH7R, FL_FOR_ARCH7M, ARM_LSL_NAME,
4674 THUMB2_WORK_REGS): Define.
4675 (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv, arm_condexec_count,
4676 arm_condexec_mask, arm_condexec_masklen)): New variables.
4677 (all_architectures): Add armv6t2, armv7, armv7a, armv7r and armv7m.
4678 (arm_override_options): Check new CPU capabilities.
4679 Set new architecture flag variables.
4680 (arm_isr_value): Handle v7m interrupt functions.
4681 (user_return_insn): Return 0 for v7m interrupt functions. Handle
4683 (const_ok_for_arm): Handle Thumb-2 constants.
4684 (arm_gen_constant): Ditto. Use movw when available.
4685 (arm_function_ok_for_sibcall): Return false for v7m interrupt
4687 (legitimize_pic_address, arm_call_tls_get_addr): Handle Thumb-2.
4688 (thumb_find_work_register, arm_load_pic_register,
4689 legitimize_tls_address, arm_address_cost, load_multiple_sequence,
4690 emit_ldm_seq, emit_stm_seq, arm_select_cc_mode, get_jump_table_size,
4691 print_multi_reg, output_mov_long_double_fpa_from_arm,
4692 output_mov_long_double_arm_from_fpa, output_mov_double_fpa_from_arm,
4693 output_mov_double_fpa_from_arm, output_move_double,
4694 arm_compute_save_reg_mask, arm_compute_save_reg0_reg12_mask,
4695 output_return_instruction, arm_output_function_prologue,
4696 arm_output_epilogue, arm_get_frame_offsets, arm_regno_class,
4697 arm_output_mi_thunk, thumb_set_return_address): Ditto.
4698 (arm_expand_prologue): Handle Thumb-2. Use arm_save_coproc_regs.
4699 (arm_coproc_mem_operand): Allow POST_INC/PRE_DEC.
4700 (arithmetic_instr, shift_op): Use arm_shift_nmem.
4701 (arm_print_operand): Use arm_print_condition. Handle '(', ')', '.',
4703 (arm_final_prescan_insn): Use extract_constrain_insn_cached.
4704 (thumb_expand_prologue): Use thumb_set_frame_pointer.
4705 (arm_file_start): Output directive for unified syntax.
4706 (arm_unwind_emit_set): Handle stack alignment instruction.
4707 * config/arm/lib1funcs.asm: Remove default for __ARM_ARCH__.
4708 Add v6t2, v7, v7a, v7r and v7m.
4709 (RETLDM): Add Thumb-2 code.
4710 (do_it, shift1, do_push, do_pop, COND, THUMB_SYNTAX): New macros.
4711 * config/arm/arm.h (TARGET_CPU_CPP_BUILTINS): Define __thumb2__.
4712 (TARGET_THUMB1, TARGET_32BIT, TARGET_THUMB2, TARGET_DSP_MULTIPLY,
4713 TARGET_INT_SIMD, TARGET_UNIFIED_ASM, ARM_FT_STACKALIGN, IS_STACKALIGN,
4714 THUMB2_TRAMPOLINE_TEMPLATE, TRAMPOLINE_ADJUST_ADDRESS,
4715 ASM_OUTPUT_OPCODE, THUMB2_GO_IF_LEGITIMATE_ADDRESS,
4716 THUMB2_LEGITIMIZE_ADDRESS, CASE_VECTOR_PC_RELATIVE,
4717 CASE_VECTOR_SHORTEN_MODE, ADDR_VEC_ALIGN, ASM_OUTPUT_CASE_END,
4718 ADJUST_INSN_LENGTH): Define.
4719 (TARGET_REALLY_IWMMXT, TARGET_IWMMXT_ABI, CONDITIONAL_REGISTER_USAGE,
4720 STATIC_CHAIN_REGNUM, HARD_REGNO_NREGS, INDEX_REG_CLASS,
4721 BASE_REG_CLASS, MODE_BASE_REG_CLASS, SMALL_REGISTER_CLASSES,
4722 PREFERRED_RELOAD_CLASS, SECONDARY_OUTPUT_RELOAD_CLASS,
4723 SECONDARY_INPUT_RELOAD_CLASS, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P,
4724 TRAMPOLINE_SIZE, INITIALIZE_TRAMPOLINE, HAVE_PRE_INCREMENT,
4725 HAVE_POST_DECREMENT, HAVE_PRE_DECREMENT, HAVE_PRE_MODIFY_DISP,
4726 HAVE_POST_MODIFY_DISP, HAVE_PRE_MODIFY_REG, HAVE_POST_MODIFY_REG,
4727 REGNO_MODE_OK_FOR_BASE_P, LEGITIMATE_CONSTANT_P,
4728 REG_MODE_OK_FOR_BASE_P, REG_OK_FOR_INDEX_P, GO_IF_LEGITIMATE_ADDRESS,
4729 LEGITIMIZE_ADDRESS, THUMB2_LEGITIMIZE_ADDRESS,
4730 GO_IF_MODE_DEPENDENT_ADDRESS, MEMORY_MOVE_COST, BRANCH_COST,
4731 ASM_APP_OFF, ASM_OUTPUT_CASE_LABEL, ARM_DECLARE_FUNCTION_NAME,
4732 FINAL_PRESCAN_INSN, PRINT_OPERAND_PUNCT_VALID_P,
4733 PRINT_OPERAND_ADDRESS): Adjust for Thumb-2.
4734 (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv): New declarations.
4735 * config/arm/arm-cores.def: Add arm1156t2-s, cortex-a8, cortex-r4 and
4737 * config/arm/arm-tune.md: Regenerate.
4738 * config/arm/arm-protos.h: Update prototypes.
4739 * config/arm/vfp.md: Enable patterns for Thumb-2.
4740 (arm_movsi_vfp): Add movw alternative. Use output_move_vfp.
4741 (arm_movdi_vfp, movsf_vfp, movdf_vfp): Use output_move_vfp.
4742 (thumb2_movsi_vfp, thumb2_movdi_vfp, thumb2_movsf_vfp,
4743 thumb2_movdf_vfp, thumb2_movsfcc_vfp, thumb2_movdfcc_vfp): New.
4744 * config/arm/libunwind.S: Add Thumb-2 code.
4745 * config/arm/constraints.md: Update include Thumb-2.
4746 * config/arm/ieee754-sf.S: Add Thumb-2/Unified asm support.
4747 * config/arm/ieee754-df.S: Ditto.
4748 * config/arm/bpabi.S: Ditto.
4749 * config/arm/t-arm (MD_INCLUDES): Add thumb2.md.
4750 * config/arm/predicates.md (low_register_operand,
4751 low_reg_or_int_operand, thumb_16bit_operator): New.
4752 (thumb_cmp_operand, thumb_cmpneg_operand): Rename...
4753 (thumb1_cmp_operand, thumb1_cmpneg_operand): ... to this.
4754 * config/arm/t-arm-elf: Add armv7 multilib.
4755 * config/arm/arm.md: Update patterns for Thumb-2 and Unified asm.
4757 (UNSPEC_STACK_ALIGN, ce_count): New.
4758 (arm_incscc, arm_decscc, arm_umaxsi3, arm_uminsi3,
4759 arm_zero_extendsidi2, arm_zero_extendqidi2): New
4761 * config/arm/fpa.md: Update patterns for Thumb-2 and Unified asm.
4762 (thumb2_movsf_fpa, thumb2_movdf_fpa, thumb2_movxf_fpa,
4763 thumb2_movsfcc_fpa, thumb2_movdfcc_fpa): New insns.
4764 * config/arm/cirrus.md: Update patterns for Thumb-2 and Unified asm.
4765 (cirrus_thumb2_movdi, cirrus_thumb2_movsi_insn,
4766 thumb2_cirrus_movsf_hard_insn, thumb2_cirrus_movdf_hard_insn): New
4768 * doc/extend.texi: Document ARMv7-M interrupt functions.
4769 * doc/invoke.texi: Document Thumb-2 new cores+architectures.
4771 2007-01-03 Jakub Jelinek <jakub@redhat.com>
4773 * unwind-dw2.c (SIGNAL_FRAME_BIT, EXTENDED_CONTEXT_BIT): Define.
4774 (struct _Unwind_Context): Rename args_size to flags, remove
4775 signal_frame field, add a new args_size field and version field.
4776 (_Unwind_IsSignalFrame, _Unwind_SetSignalFrame,
4777 _Unwind_IsExtendedContext): New inline functions.
4778 (_Unwind_GetGR, _Unwind_SetGR, _Unwind_GetGRPtr, _Unwind_SetGRPtr):
4779 Assume by_value array is only present if _Unwind_IsExtendedContext.
4780 (_Unwind_GetIPInfo, execute_cfa_program, uw_frame_state_for): Use
4781 _Unwind_IsSignalFrame.
4782 (__frame_state_for): Initialize context.flags to EXTENDED_CONTEXT_BIT.
4783 (uw_update_context_1): Use _Unwind_SetSignalFrame.
4784 (uw_init_context_1): Initialize context->flags to
4785 EXTENDED_CONTEXT_BIT.
4786 * config/rs6000/linux-unwind.h (frob_update_context): Use
4787 _Unwind_SetSignalFrame.
4789 2007-01-03 Andrew Pinski <andrew_pinski@playstation.sony.com>
4792 * gimplify.c (gimplify_modify_expr_complex_part): Move below
4793 tree_to_gimple_tuple. Call tree_to_gimple_tuple when we need
4796 2007-01-03 Kazu Hirata <kazu@codesourcery.com>
4798 * config/i386/i386.h (NON_STACK_REG_P, REGNO_OK_FOR_SIREG_P,
4799 REGNO_OK_FOR_DIREG_P, REWRITE_ADDRESS, ASM_OPERAND_LETTER,
4800 RET, AT_SP): Remove.
4801 * config/i386/i386.md (*sse_prologue_save_insn): Use return
4804 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
4805 * cfgbuild.c (find_basic_blocks): Likewise.
4806 * cfgrtl.c (rtl_create_basic_block): Likewise.
4807 * function.c (temp_slots_at_level): Likewise.
4808 * reg-stack.c (stack_regs_mentioned): Likewise.
4809 * regclass.c (allocate_reg_info): Likewise.
4810 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
4811 set_bb_for_stmt, move_block_to_fn): Likewise.
4812 * tree-complex.c (tree_lower_complex): Likewise.
4813 * vec.h (VEC_safe_grow_cleared): New.
4815 * cgraphunit.c, tree-ssa-alias.c: Fix comment typos.
4817 2007-01-03 Zdenek Dvorak <dvorakz@suse.cz>
4819 * loop-unswitch.c (unswitch_loop): Pass probabilities to loopify.
4820 * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Pass probabilities
4822 * cfgloopmanip.c (scale_loop_frequencies): Export.
4823 (loopify): Scale the frequencies by prescribed coefficients.
4824 (set_zero_probability): New function.
4825 (duplicate_loop_to_header_edge): Improve updating of frequencies.
4826 (lv_adjust_loop_entry_edge, loop_version): Set probabilities
4827 and frequencies according to arguments.
4828 * tree-ssa-loop-manip.c (tree_unroll_loop): Set probabilities
4830 * cfg.c (scale_bbs_frequencies_int): Allow scaling the frequencies up.
4831 * modulo-sched.c (sms_schedule): Set probabilities for entering
4832 versioned loop correctly.
4833 * tree-vect-transform.c (vect_transform_loop): Ditto.
4834 * cfgloop.h (loopify, loop_version): Declaration changed.
4835 (scale_loop_frequencies): Declared.
4837 2007-01-02 Jan Hubicka <jh@suse.cz>
4839 * cgraph.c: Include tree-flow.h
4840 (cgraph_add_new-function): Handle IPA_SSA mode; execute
4842 * cgraph.h (enum cgraph_state): Add CGRAPH_STATE_IPA_SSA.
4843 * tree-pass.h (pass_all_early_optimizations): Declare.
4844 * cgraphunit.c (cgraph_process_new_functions): Add IPA_SSA; execute
4846 (cgraph_analyze_function): Do early_local_passes.
4847 * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
4848 Do not add referenced vars.
4849 * tree-optimize.c (gate_all_optimizations): Do not execute when not in
4851 (gate_all_early_local_passes): New gate.
4852 (pass_early_local_passes): Use new gate.
4853 (execute_early_local_optimizations): New functions.
4854 (gate_all_early_optimizations): New gate.
4855 (pass_all_early_optimizations): New pass.
4856 (execute_free_datastructures): Free SSA only when initialized.
4857 (gate_init_datastructures): Init only when optimizing.
4858 (tree_lowering_passes): Do early local passes when called late.
4859 * tree-profile.c (do_tree_profiling): Don't profile functions added
4861 (do_early_tree_profiling, pass_early_tree_profile): Kill.
4862 * tree-cfg.c (update_modified_stmts): Do not update when operands are
4864 * passes.c (init_optimizations_passes): Reorder so we go into SSA
4865 during early_local_passes.
4866 * Makefile.in (cgraph.o): Add dependency on tree-flow.h.
4869 2007-01-02 Carlos O'Donell <carlos@codesourcery.com>
4871 * Makefile.in: Update copyright year.
4873 2007-01-02 Carlos O'Donell <carlos@codesourcery.com>
4875 * Makefile.in: Export GCC_EXEC_PREFIX before calling $(RUNTEST)
4876 in $(lang_checks) and check-consistency targets.
4878 2007-01-02 Jan Hubicka <jh@suse.cz>
4880 * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
4881 Do not add referenced vars.
4882 * tree-cfg.c (update_modified_stmts): Do not update when SSA operands
4884 * passes.c (init_optimization_passes): Put mudflap_2 after
4885 free_datastructures.
4887 2007-01-02 Jan Hubicka <jh@suse.cz>
4889 * tree-optimize (execute_fixup_cfg): Set after_inlining flag.
4890 Set NOTHROW flag on call statements proved to be nothrow.
4891 Update statement of local calls so new pure/const functions are
4892 updated. Update_ssa when in ssa form. Mark PHI nodes of nonlocal
4894 (tree_rest_of_compilation): Register hooks and initialize bitmap
4895 early. Do not set after_inlining flag.
4897 2007-01-02 Steve Ellcey <sje@cup.hp.com>
4899 * sbitmap.c (HOST_BITS_PER_LONG_LONG): Change to
4900 HOST_BITS_PER_LONGLONG
4902 2007-01-02 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
4905 * c-typeck.c (store_init_value): Don't emit pedantic overflow
4906 warning for non-static initializers.
4908 2007-01-02 Steven Bosscher <steven@gcc.gnu.org>
4910 * config/alpha/alpha.md, arm/arm.c, darwin.c, frv/frv.md,
4911 m32r/m32r.c, m32r/m32r.c, mn10300/mn10300.md, pa/pa.c,
4912 rs6000/rs6000.c, s390/s390.md, sh/sh.md, sparc/sparc.c:
4913 Always use set_unique_reg_note to add REG_EQUAL notes.
4915 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
4918 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
4920 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
4921 * cfgbuild.c (find_basic_blocks): Likewise.
4922 * cfgrtl.c (rtl_create_basic_block): Likewise.
4923 * function.c (temp_slots_at_level): Likewise.
4924 * reg-stack.c (stack_regs_mentioned): Likewise.
4925 * regclass.c (allocate_reg_info): Likewise.
4926 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
4927 set_bb_for_stmt, move_block_to_fn): Likewise.
4928 * tree-complex.c (tree_lower_complex): Likewise.
4929 * vec.h (VEC_safe_grow_cleared): New.
4931 2007-01-02 Ian Lance Taylor <iant@google.com>
4933 * c-common.c (c_common_truthvalue_conversion): When warning about
4934 using an assignment as a truth value, set TREE_NO_WARNING.
4936 2007-01-02 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
4939 * c.opt (Wold-style-declaration): New.
4940 * doc/invoke.texi (C-only Warning Options): New.
4941 (Wold-style-declaration): Document it.
4942 (Wextra): Enabled by -Wextra.
4943 * c-opts.c (c_common_post_options): Enabled by -Wextra.
4944 * c-decl.c (declspecs_add_scspec): Replace -Wextra with
4945 -Wold-style-declaration.
4947 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
4949 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
4950 * cfgbuild.c (find_basic_blocks): Likewise.
4951 * cfgrtl.c (rtl_create_basic_block): Likewise.
4952 * function.c (temp_slots_at_level): Likewise.
4953 * reg-stack.c (stack_regs_mentioned): Likewise.
4954 * regclass.c (allocate_reg_info): Likewise.
4955 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
4956 set_bb_for_stmt, move_block_to_fn): Likewise.
4957 * tree-complex.c (tree_lower_complex): Likewise.
4958 * vec.h (VEC_safe_grow_cleared): New.
4960 2007-01-02 Douglas Gregor <doug.gregor@gmail.com>
4962 * c-common.c (c_common_nodes_and_builtins): Since variants of
4963 void_type_node get built before it is given a name, we need to
4964 give those variants the name, too.
4965 (complete_array_type): We need to work with the canonical main
4966 type of the array, from which we will build the qualified version.
4967 * params.def (PARAM_VERIFY_CANONICAL_TYPES): New.
4968 * print-tree.c (print_node): Display canonical type information
4970 * stor-layout.c (layout_type): When we don't know the
4971 alignment of a type for which we're building an array, we end up
4972 guessing wrong, so make the type require structural equality.
4973 * tree.c (make_node_stat): When we build a new type, it is its
4975 (build_type_attribute_qual_variant): When building an attribute
4976 variant, its canonical type is the non-attribute variant. However,
4977 if the attributes are target-dependent and they differ, we need to
4978 use structural equality checks for this type.
4979 (build_qualified_type): A qualified type is not equivalent to its
4980 unqualified variant; set the canonical type appropriately.
4981 (build_distinct_type_copy): When building a distinct type from
4982 another type, the new type is its own canonical type.
4983 (build_variant_type_copy): When building a new type variant, we
4984 assume that it is equivalent to the original type.
4985 (build_pointer_type_for_mode): When building a pointer type, also
4986 build a canonical type pointer.
4987 (build_reference_type_for_mode): When building a reference type,
4988 also build a canonical type reference.
4989 (build_index_type): When we can't hash an index type (e.g.,
4990 because its maximum value is negative), the index type requires
4991 structural equality tests.
4992 (build_array_type): Build the canonical form of an array type.
4993 (build_function_type): Function types require structural equality,
4994 because they contain default arguments, attributes, etc.
4995 (build_method_type_directly): Ditto for method types.
4996 (build_offset_type): Build the canonical offset type.
4997 (build_complex_type): Build the canonical vector type.
4998 (make_vector_type): Build the canonical vector type.
4999 * tree.h (TYPE_CANONICAL): New.
5000 (TYPE_STRUCTURAL_EQUALITY_P): New.
5001 (SET_TYPE_STRUCTURAL_EQUALITY): New.
5002 (struct tree_type): Added "canonical" field.
5003 * params.h (VERIFY_CANONICAL_TYPES): New.
5004 * doc/c-tree.texi (TYPE_CANONICAL): Document.
5005 (TYPE_STRUCTURAL_EQUALITY_P): Document.
5006 (SET_TYPE_STRUCTURAL_EQUALITY): Document.
5007 * doc/invoke.texi (verify-canonical-types): Document --param
5008 parameter for verifying canonical types.
5010 2007-01-02 Joseph Myers <joseph@codesourcery.com>
5012 * config.gcc (powerpc-*-eabispe*, powerpc-*-eabisimaltivec*,
5013 powerpc-*-eabisim*, powerpc-*-eabialtivec*, powerpc-*-eabi*,
5014 powerpc-*-rtems*, powerpc-wrs-vxworks, powerpc-wrs-vxworksae,
5015 powerpcle-*-eabisim*, powerpcle-*-eabi*): Add rs6000/e500.h to
5017 * config/rs6000/e500.h: New.
5018 * config/rs6000/eabi.h (TARGET_SPE_ABI, TARGET_SPE, TARGET_E500,
5019 TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE, TARGET_E500_DOUBLE):
5021 * config/rs6000/linuxspe.h (TARGET_SPE_ABI, TARGET_SPE,
5022 TARGET_E500, TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE,
5023 TARGET_E500_DOUBLE): Remove.
5024 * config/rs6000/vxworks.h (TARGET_SPE_ABI, TARGET_SPE,
5025 TARGET_E500, TARGET_ISEL, TARGET_FPRS): Remove.
5026 * config/rs6000/rs6000.h (CHECK_E500_OPTIONS): Define.
5027 * config/rs6000/rs6000.c (rs6000_override_options): Use
5030 2007-01-02 Joseph Myers <joseph@codesourcery.com>
5032 * config/rs6000/rs6000.c (print_operand): Check (TARGET_SPE ||
5033 TARGET_E500_DOUBLE), not TARGET_E500, for %y.
5034 (rs6000_generate_compare, rs6000_emit_sCOND, output_cbranch,
5035 rs6000_emit_cmove): Don't check TARGET_E500.
5036 * config/rs6000/rs6000.md (bunordered, bordered, sunordered,
5037 sordered): Don't check TARGET_E500.
5039 2007-01-01 Eric Christopher <echristo@apple.com>
5041 * config/mips/mips.c (mips_regno_mode_ok_for_base_p): Use
5042 HARD_REGISTER_NUM_P.
5044 2007-01-01 Roger Sayle <roger@eyesopen.com>
5046 * fold-const.c (fold_binary) <EQ_EXPR>: Fold "(X^C1) eq/ne C2" into
5047 "X eq/ne (C1^C2)". Fold "(X^Z) eq/ne (Y^Z)" as "X eq/ne Y" when Z
5048 has no side-effects. Fold "(X^C1) eq/ne (Y^C2)" as "(X^(C1^C2))
5051 2007-01-01 Mike Stump <mrs@apple.com>
5053 * configure.ac: Remove support for building with Apple's gcc-3.1.
5055 2007-01-02 Joseph Myers <joseph@codesourcery.com>
5058 * caller-save.c (add_stored_regs): Only handle SUBREGs if inner
5059 REG is a hard register. Do not modify REG before calling
5061 * rtlanal.c (subreg_get_info): Don't assert size of XMODE is a
5062 multiple of the size of YMODE for certain lowpart cases.
5064 2007-01-01 Andrew Pinski <pinskia@gmail.com>
5067 * gimplify (voidify_wrapper_expr): Update for
5070 2007-01-01 Andreas Schwab <schwab@suse.de>
5073 * config/ia64/ia64.c (ia64_compute_frame_size): Account space for
5074 save of BR0 in extra_spill_size instead of spill_size.
5075 (ia64_expand_prologue): Save BR0 outside of the gr/br/fr spill
5077 (ia64_expand_epilogue): Restore BR0 from its new location.
5079 2007-01-01 Andrew Pinski <pinskia@gmail.com>
5081 * gimplify.c (gimplify_init_constructor <case VECTOR_TYPE>):
5082 Use a temporary variable if the left hand side is not a gimple
5085 2007-01-01 Andrew Pinski <pinskia@gmail.com>
5087 * gimplify.c (gimplify_return_expr): Make the temporary variable
5088 for the return expression, a gimple register variable.
5090 2007-01-01 Jan Hubicka <jh@suse.cz>
5092 * emit-rtl.c (emit_copy_of_insn_after): Do not call copy_insn_1 for
5095 2007-01-01 Mike Stump <mrs@apple.com>
5097 * configure.ac (HAVE_GAS_LITERAL16): Add autoconf check for
5099 * config/darwin.c (machopic_select_rtx_section): Use
5101 (darwin_mergeable_constant_section): Likewise.
5102 * configure: Regenerate.
5103 * config.in: Regenerate.
5105 2007-01-01 Jan Hubicka <jh@suse.cz>
5106 Andrew Pinski <pinskia@gmail.com>
5108 * cgraphunit.c (cgraph_optimize): Call cgraph_add_new_functions
5109 before starting IPA passes.