1 2007-02-13 Ian Lance Taylor <iant@google.com>
3 * common.opt: Add Wstrict-overflow and Wstrict-overflow=.
4 * flags.h (warn_strict_overflow): Declare.
5 (enum warn_strict_overflow_code): Define.
6 (issue_strict_overflow_warning): New static inline function.
7 * opts.c (warn_strict_overflow): New variable.
8 (common_handle_option): Handle OPT_Wstrict_overflow and
10 * c-opts.c (c_common_handle_option): Set warn_strict_overflow for
12 * fold-const.c: Include intl.h.
13 (fold_deferring_overflow_warnings): New static variable.
14 (fold_deferred_overflow_warning): New static variable.
15 (fold_deferred_overflow_code): New static variable.
16 (fold_defer_overflow_warnings): New function.
17 (fold_undefer_overflow_warnings): New function.
18 (fold_undefer_and_ignore_overflow_warnings): New function.
19 (fold_deferring_overflow_warnings_p): New function.
20 (fold_overflow_warning): New static function.
21 (make_range): Add strict_overflow_p parameter. Change all
23 (extract_muldiv, extract_muldiv_1): Likewise.
24 (fold_unary) [ABS_EXPR]: Check ABS_EXPR before calling
25 tree_expr_nonnegative_p.
26 (fold_negate_expr): Call fold_overflow_warning.
27 (fold_range_test): Likewise.
28 (fold_comparison): Likewise.
29 (fold_binary): Likewise. Call tree_expr_nonnegative_warnv_p
30 instead of tree_expr_nonnegative_p.
31 (tree_expr_nonnegative_warnv_p): Rename from
32 tree_expr_nonnegative_p, add strict_overflow_p parameter.
33 (tree_expr_nonnegative_p): New function.
34 (tree_expr_nonzero_warnv_p): Rename from tree_expr_nonzero_p, add
35 strict_overflow_p parameter.
36 (tree_expr_nonzero_p): New function.
37 * passes.c (verify_interpass_invariants): New static function.
38 (execute_one_pass): Call it.
39 * tree-ssa-loop-niter.c (expand_simple_operations): Ignore fold
41 (number_of_iterations_exit, loop_niter_by_eval): Likewise.
42 (estimate_numbers_of_iterations): Likewise.
43 (scev_probably_wraps_p): Likewise.
44 * tree-ssa-ccp.c: Include "toplev.h".
45 (evaluate_stmt): Defer fold overflow warnings until we know we are
47 (struct fold_stmt_r_data): Add stmt field.
48 (fold_stmt_r): Defer fold overflow warnings until we know we
50 (fold_stmt): Initialize stmt field of fold_stmt_r_data.
51 (fold_stmt_inplace): Likewise.
52 * tree-cfgcleanup.c: Include "toplev.h" rather than "errors.h".
53 (cleanup_control_expr_graph): Defer fold overflow warnings until
54 we know we are going to optimize.
55 * tree-cfg.c (fold_cond_expr_cond): Likewise.
56 * tree-ssa-threadedge.c (simplify_control_stmt_condition):
58 * tree-vrp.c (vrp_expr_computes_nonnegative): Call
59 tree_expr_nonnegative_warnv_p instead of tree_expr_nonnegative_p.
60 * tree-ssa-loop-manip.c (create_iv): Likewise.
61 * c-typeck.c (build_conditional_expr): Likewise.
62 (build_binary_op): Likewise.
63 * tree-vrp.c (vrp_expr_computes_nonzero): Call
64 tree_expr_nonzero_warnv_p instead of tree_expr_nonzero_p.
65 (extract_range_from_unary_expr): Likewise.
66 * simplify-rtx.c (simplify_const_relational_operation): Warn when
67 assuming that signed overflow does not occur.
68 * c-common.c (pointer_int_sum): Ignore fold overflow warnings.
69 * tree.h (tree_expr_nonnegative_warnv_p): Declare.
70 (fold_defer_overflow_warnings): Declare.
71 (fold_undefer_overflow_warnings): Declare.
72 (fold_undefer_and_ignore_overflow_warnings): Declare.
73 (fold_deferring_overflow_warnings_p): Declare.
74 (tree_expr_nonzero_warnv_p): Declare.
75 * doc/invoke.texi (Option Summary): Add -Wstrict-overflow to list
77 (Warning Options): Document -Wstrict-overflow.
78 * Makefile.in (tree-ssa-threadedge.o): Depend on toplev.h.
79 (tree-ssa-ccp.o): Likewise.
80 (tree-cfgcleanup.o): Change errors.h dependency to toplev.h.
81 (fold-const.o): Depend on intl.h.
83 2007-02-13 Ian Lance Taylor <iant@google.com>
86 * lower-subreg.c (resolve_simple_move): Decompose subregs in
89 2007-02-13 Stuart Hastings <stuart@apple.com>
91 * gcc/config/i386/i386.md (fixuns_truncdfhi2): Require SSE2.
93 2007-02-13 Richard Henderson <rth@redhat.com>
95 * config/alpha/alpha.c (alpha_stdarg_optimize_hook): Strip
96 handled_component_p before looking for the indirect_ref.
98 2007-02-13 Richard Henderson <rth@redhat.com>
100 * config/i386/i386.md (bswapsi_1): Rename from bswapsi2,
101 remove flags clobber.
102 (bswapsi2): New expander, emit code for !TARGET_BSWAP.
103 (bswaphi_lowpart): New.
104 (bswapdi2): Rename from bswapdi2_rex, remove flags clobber,
105 remove TARGET_BSWAP test. Delete expander of the same name.
107 * optabs.c (widen_bswap, expand_doubleword_bswap): New.
108 (expand_unop): Use them.
110 2007-02-13 Uros Bizjak <ubizjak@gmail.com>
112 * config/i386/i386.md (cmpdi_ccno_1_rex64, *cmpsi_ccno_1,
113 *cmphi_ccno_1, *cmpqi_ccno_1, *movsi_xor, *movstricthi_xor,
114 *movstrictqi_xor, *movdi_xor_rex64, *ashldi3_1_rex64,
115 *ashldi3_cmp_rex64, *ashldi3_cconly_rex64, ashlsi3, *ashlsi3_1_zext,
116 *ashlsi3_cmp, *ashlsi3_cconly, *ashlsi3_cmp_zext, *ashlhi3_1_lea,
117 *ashlhi3_1, *ashlhi3_cmp, *ashlhi3_cconly, *ashlqi3_1_lea,
118 *ashlqi3_1, *ashlqi3_cmp, *ashlqi3_cconly): Remove equivalent
119 assembler dialect choice from asm templates.
121 2007-02-12 Richard Henderson <rth@redhat.com>
123 * config/i386/i386.md (fixuns_trunc<SSEMODEF>si_1): New insn.
124 (fixuns_trunc<SSEMODEF>si2): Use it.
125 * config/i386/sse.md (vec_setv4sf_0): Export.
126 * config/i386/i386.c (ix86_build_const_vector): Export.
127 (ix86_split_convert_uns_si_sse): Rename from
128 ix86_expand_convert_uns_si_sse and rewrite as a splitter.
129 * config/i386/i386-protos.h: Update.
131 2007-02-13 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
134 * c-typeck.c (c_finish_return): Improve warning message.
136 2007-02-12 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
138 * alias.c (find_symbolic_term): Delete unused function.
140 2007-02-12 Uros Bizjak <ubizjak@gmail.com>
142 * config/i386/i386.md (paritydi2, paritysi2): New expanders.
143 (paritydi2_cmp, paritydi2_cmp): New insn and split patterns.
144 (*parityhi2_cmp, *parityqi2_cmp): New insn patterns.
146 2007-02-12 Eric Botcazou <ebotcazou@adacore.com>
148 * tree.h (DECL_IGNORED_P): Document further effect for FUNCTION_DECL.
149 * cgraphunit.c (cgraph_expand_function): If DECL_IGNORED_P is set on
150 the function, temporarily point the debug interface to the null one.
152 2007-02-12 Eric Botcazou <ebotcazou@adacore.com>
154 * dwarf2out.c (round_up_to_align): New static function.
155 (field_byte_offset): Use it to round the offset.
157 2007-02-12 Richard Henderson <rth@redhat.com>
159 * config/alpha/alpha.md (bswapsi2, bswapdi2): New.
160 (inswl_const): Export.
162 2007-02-12 Richard Henderson <rth@redhat.com>
164 * calls.c (emit_library_call_value_1): If PROMOTE_MODE modifed the
165 result mode of the libcall, convert back to outmode.
167 2007-02-12 Roger Sayle <roger@eyesopen.com>
169 * config/i386/i386.md (*bswapdi2_rex): Renamed from bswapdi2.
170 (bswapdi2): New define_expand to implement 32-bit implementation.
172 2007-02-12 Nick Clifton <nickc@redhat.com>
174 * doc/invoke.texi (Overall Options): Document --help=.
175 * gcc.c (target_help_flag): Rename to print_subprocess_flag.
176 (cc1_options): Pass --help= on to cc1.
177 (display_help): Add description of --help=.
178 (process_command): Add code to handle --help=. Allow translated
179 --help and --target-help switches to be passed on to compiler
181 (main): Remove unused if statement.
182 * opts.c (columns): Remove.
183 (LEFT_COLUMN): Define.
184 (wrap_help): Add columns argument.
185 (print_filtered_help): Change parameters to be an include bitmask,
186 an exclude bitmask, an any bitmask and the column width. Move the
187 code to display the params list here. Add code to display the
188 status of options rather than their descriptions if the quiet flag
190 (print_specific_help): Change parameters to be an include bitmask,
191 an exclude bitmask and an any bitmask. Move code to look up the
192 column width here. Decide upon the title for an options listing.
193 (common_handle_options): Add code to handle --help=. Adapt code
194 for --help and --target-help to use the revised form of the
195 print_specific_help function.
196 (print_help): Delete.
197 (print_param_help): Delete.
198 (print_switch): Delete.
199 * opts.h (cl_lang_count): Add prototype.
200 (CL_PARAMS, CL_WARNING, CL_OPTIMIZATION, CL_MIN_OPTION_CLASS,
201 CL_MAX_OPTION_CLASS): New defines.
202 * optc-gen.awk: Add construction of cl_lang_count.
203 * c.opt: Add Warning attribute to warning options and Optimization
204 attribute to optimization options.
205 * common.opt: Likewise.
207 Add -fhelp and -ftarget-help as aliases for the transformed --help
208 and --target-help options.
209 * opt-functions.awk: Add code to handle Warning and Optimization
212 2007-02-12 Richard Henderson <rth@redhat.com>
214 * config/alpha/constraints.md: New file.
215 * config/alpha/alpha.c: Include tm-constrs.h.
216 (alpha_const_ok_for_letter_p, alpha_const_double_ok_for_letter_p,
217 alpha_extra_constraint): Remove.
218 (alpha_emit_conditional_branch): Use satisfies_constraint_*.
219 * config/alpha/alpha-protos.h: Update.
220 * config/alpha/alpha.h (REG_CLASS_FROM_LETTER): Remove.
221 (CONST_OK_FOR_LETTER_P, CONST_DOUBLE_OK_FOR_LETTER_P): Remove.
222 (EXTRA_CONSTRAINT): Remove.
223 * config/alpha/alpha.md: Include constraints.md.
224 (adddi splitter): Use satisfies_constraint_*.
225 * config/alpha/predicates.md (add_operand): Likewise.
226 (sext_add_operand, addition_operation): Likewise.
228 2007-02-12 Dorit Nuzman <dorit@il.ibm.com>
230 PR tree-optimization/29145
231 * tree-data-ref.c (base_addr_differ_p): Make us more conservative
232 in our handling of restrict qualified pointers.
234 2007-02-12 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
237 * doc/invoke.texi (Wunused-value): Update description.
238 (Wextra): Delete item.
239 * opts.c (set_Wextra): Don't use the value of Wextra to set the
240 value of Wunused-value.
241 * c-typeck.c (c_process_expr_stmt): Don't check extra_warnings.
242 (c_finish_stmt_expr): Don't check extra_warnings.
243 (emit_side_effect_warnings): The caller is responsible to check
246 2007-02-11 Roger Sayle <roger@eyesopen.com>
247 Matt Thomas <matt@3am-software.com>
249 * simplify-rtx.c (simplify_relational_operation_1): Correct typo.
251 2007-02-11 Roger Sayle <roger@eyesopen.com>
253 * simplify-rtx.c (simplify_relational_operation_1): Optimize
254 comparisons of POPCOUNT against zero.
255 (simplify_const_relational_operation): Likewise.
257 2007-02-11 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
259 * doc/invoke.texi (Wextra): Delete outdated paragraph.
261 2007-02-11 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
263 * dwarf2out.c (root_type): Delete unused function.
265 2007-02-11 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
267 * genattrtab.c (contained_in_p): Delete unused function.
268 (write_expr_attr_cache): Likewise.
270 2007-02-11 Jan Hubicka <jh@suse.cz>
272 * ipa-inline.c (cgraph_edge_badness): Add "else" missing in the
275 2007-02-11 Steven Bosscher <steven@gcc.gnu.org>
277 * fwprop.c (try_fwprop_subst): Use set_unique_reg_note
278 to add the REG_EQ* note.
279 * see.c (see_merge_one_use_extension): Likewise.
280 * local-alloc.c (update_equiv_regs): Likewise. Also don't
281 turn REG_EQUAL notes into REG_EQUIV notes if the target
282 register may have more than one set.
283 * function.c (assign_parm_setup_reg): Use set_unique_reg_note.
284 * gcse.c (try_replace_reg): Likewise.
285 * alias.c (init_alias_analysis): Use find_reg_equal_equiv_note.
286 * calls.c (fixup_tail_calls): Likewise. Abort if there is
287 more than one REG_EQUIV note.
288 * reload1.c (gen_reload): Use set_unique_reg_note.
290 2007-02-11 Uros Bizjak <ubizjak@gmail.com>
292 * config/i386/i386.c (TARGET_VECTORIZE_BUILTIN_CONVERSION): Define.
293 (ix86_builtin_conversion): New function.
295 2007-02-06 Mark Mitchell <mark@codesourcery.com>
298 * tree.h (DECL_REPLACEABLE_P): New macro.
299 * except.c (set_nothrow_function_flags): Likewise.
301 2007-02-11 Tehila Meyzels <tehila@il.ibm.com>
302 Ira Rosen <irar@il.ibm.com>
303 Dorit Nuzman <dorit@il.ibm.com>
305 * doc/tm.texi (TARGET_VECTORIZE_BUILTIN_CONVERSION): New target hook.
306 * targhooks.c (default_builtin_vectorized_conversion): New.
307 * targhooks.h (default_builtin_vectorized_function): New declaration.
308 * target.h (struct vectorize): Add builtin_conversion field.
309 * tree-vectorizer.h (type_conversion_vec_info_type): New enum
310 stmt_vec_info_type value.
311 (vectorizable_conversion): New declaration.
312 * tree-vect-analyze.c (vect_analyze_operations): Add
313 vectorizable_conversion call.
314 * target-def.h (TARGET_VECTORIZE_BUILTIN_CONVERSION): New.
315 * tree-vect-transform.c (vectorizable_conversion): New function.
316 (vect_transform_stmt): Add case for type_conversion_vec_info_type.
317 * tree-vect-generic.c (expand_vector_operations_1): Consider correct
319 * config/rs6000/rs6000.c (rs6000_builtin_conversion): New.
320 (TARGET_VECTORIZE_BUILTIN_CONVERSION): Defined.
321 (rs6000_expand_builtin): Add handling a case of ALTIVEC_BUILTIN_VCFUX
322 or ALTIVEC_BUILTIN_VCFSX.
324 2007-02-10 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
327 * pa.md (movdf): For 64-bit target, fail if operand 1 is a non-zero
328 CONST_DOUBLE and operand 0 is a hard register.
329 (movdi): For 64-bit target, remove code to force CONST_DOUBLE to
330 memory. Fail if operand 1 is a non-zero CONST_INT and operand 0
331 is a hard floating-point register.
333 2007-02-10 Richard Henderson <rth@redhat.com>
334 Jakub Jelinek <jakub@redhat.com>
335 Alexandre Oliva <aoliva@redhat.com>
337 * Makefile.in (libgcc-support, libgcc.mvars): Add emutls.c.
338 * builtin-types.def (BT_WORD): Make unsigned.
339 (BT_FN_VOID_PTR_WORD_WORD_PTR): New.
340 * builtins.def (BUILT_IN_EMUTLS_GET_ADDRESS): New.
341 (BUILT_IN_EMUTLS_REGISTER_COMMON): New.
342 * c-decl.c (grokdeclarator): Don't error if !have_tls.
343 * c-parser.c (c_parser_omp_threadprivate): Likewise.
344 * dwarf2out.c (loc_descriptor_from_tree_1): Don't do anything for
346 * expr.c (emutls_var_address): New.
347 (expand_expr_real_1): Expand emulated tls.
348 (expand_expr_addr_expr_1): Likewise.
349 * libgcc-std.ver: Add __emutls_get_address, __emutls_register_common.
350 * output.h (emutls_finish): Declare.
351 * toplev.c (compile_file): Call it.
352 * tree-ssa-address.c (gen_addr_rtx): Check for const-ness of the
353 address before wrapping in CONST.
354 * varasm.c (emutls_htab, emutls_object_type): New.
355 (EMUTLS_VAR_PREFIX, EMUTLS_TMPL_PREFIX): New.
356 (get_emutls_object_name, get_emutls_object_type): New.
357 (get_emutls_init_templ_addr, emutls_decl): New.
358 (emutls_common_1, emutls_finish): New.
359 (assemble_variable): When emulating tls, swap decls; generate
360 constructor for the emutls objects.
361 (do_assemble_alias): When emulating tls, swap decl and target name.
362 (default_encode_section_info): Don't add SYMBOL_FLAG_TLS_SHIFT
364 * varpool.c (decide_is_variable_needed): Look at force_output.
365 Recurse for emulated tls.
366 (cgraph_varpool_remove_unreferenced_decls): Remove checks redundant
367 with decide_is_variable_needed.
368 * emutls.c: New file.
369 * config/sparc/sol2.h (ASM_DECLARE_OBJECT_NAME): Only emit
370 tls_object for real tls.
372 2007-02-10 Kaz Kojima <kkojima@gcc.gnu.org>
374 PR rtl-optimization/29599
375 * reload1.c (eliminate_regs_in_insn): Take the destination
376 mode into account when computing the offset.
378 2007-02-09 Stuart Hastings <stuart@apple.com>
379 Richard Henderson <rth@redhat.com>
381 * gcc/config/i386/i386.h (TARGET_KEEPS_VECTOR_ALIGNED_STACK): New.
382 * gcc/config/i386/darwin.h: (TARGET_KEEPS_VECTOR_ALIGNED_STACK): New.
383 * gcc/config/i386/i386.md (fixuns_trunc<mode>si2, fixuns_truncsfhi2,
384 fixuns_truncdfhi2): New.
385 (fix_truncsfdi_sse): Call ix86_expand_convert_sign_didf_sse.
386 (floatunsdidf2): Call ix86_expand_convert_uns_didf_sse.
387 (floatunssisf2): Add call to ix86_expand_convert_uns_sisf_sse.
388 (floatunssidf2): Allow nonimmediate source.
389 * gcc/config/i386/sse.md (movdi_to_sse): New.
390 (vec_concatv2di): Drop '*'.
391 * gcc/config/i386/i386-protos.h (ix86_expand_convert_uns_si_sse,
392 ix86_expand_convert_uns_didf_sse, ix86_expand_convert_uns_sidf_sse,
393 ix86_expand_convert_uns_sisf_sse, ix86_expand_convert_sign_didf_sse):
395 * gcc/config/i386/i386.c (ix86_expand_convert_uns_si_sse,
396 ix86_expand_convert_uns_didf_sse, ix86_expand_convert_uns_sidf_sse,
397 ix86_expand_convert_uns_sisf_sse, ix86_expand_convert_sign_didf_sse,
398 ix86_build_const_vector, ix86_expand_vector_init_one_nonzero): New.
399 (ix86_build_signbit_mask): Fix decl of v, refactor to call
400 ix86_build_const_vector.
401 (x86_emit_floatuns): Rewrite.
403 2007-02-10 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
405 * genautomata.c (longest_path_length): Delete unused function.
406 (struct state): Delete unused longest_path_length.
407 (UNDEFINED_LONGEST_PATH_LENGTH): Delete unused macro.
408 (get_free_state): Delete unused.
410 2007-02-09 Jan Hubicka <jh@suse.cz>
412 * params.def (PARAM_INLINE_UNIT_GROWTH): Set to 30.
413 * doc/invoke.texi (inline-unit-growth): Update default value.
415 * Makefile.in (passes.o, ipa-inline.o): Add dependencies.
416 * cgraphbuild.c (build_cgraph_edges): Compute frequencies.
417 (rebuild_cgraph_edges): Likewise.
418 * cgraph.c (cgraph_set_call_stmt): Add new argument frequency.
419 (dump_cgraph_node): Dump frequencies.
420 (cgraph_clone_edge): Add frequency scales.
421 (cgraph_clone_node): Add freuqnecy.
422 * cgraph.h (cgraph_edge): Add freuqnecy argument.
423 (CGRAPH_FREQ_BASE, CGRAPH_FREQ_MAX): New constants.
424 (cgraph_create_edge, cgraph_clone_edge, cgraph_clone_node): Update.
425 * tree-pass.h (TODO_rebuild_frequencies): New constant.
426 * cgraphunit.c (verify_cgraph_node): Verify frequencies.
427 (cgraph_copy_node_for_versioning): Update call of cgraph_clone_edge.
428 (save_inline_function_body): Likewise.
429 * ipa-inline.c: inluce rtl.h
430 (cgraph_clone_inlined_nods): Update call of cgraph_clone_node.
431 (cgraph_edge_badness): Use frequencies.
432 (cgraph_decide_recursive_inlining): Update clonning.
433 (cgraph_decide_inlining_of_small_function): Dump frequency.
434 * predict.c (estimate_bb_frequencies): Export.
435 * predict.h (estimate_bb_frequencies): Declare.
436 * tree-inline.c (copy_bb): Watch overflows.
437 (expand_call_inline): Update call of cgraph_create_edge.
438 (optimize_inline_calls): Use TODO flags to update frequnecies.
439 * passes.h: Include predict.h
440 (init_optimization_passes): Move profile ahead.
441 (execute_function_todo): Handle TODO_rebuild_frequencies.
443 2007-02-09 Roger Sayle <roger@eyesopen.com>
445 * config/alpha/alpha.c (emit_insxl): Force the first operand of
446 the insbl or inswl pattern into a register.
448 2007-02-09 Roger Sayle <roger@eyesopen.com>
450 * config/ia64/ia64.md (bswapdi2): New define_insn.
452 2007-02-09 Richard Henderson <rth@redhat.com>
454 * config/i386/constraints.md (Ym): New constraint.
455 * config/i386/i386.md (movsi_1): Change Y2 to Yi constraints.
456 (movdi_1_rex64): Split sse and xmm general register moves from
457 memory move alternatives. Use conditional register constraints.
458 (movsf_1, movdf_integer): Likewise.
459 (zero_extendsidi2_32, zero_extendsidi2_rex64): Likewise.
460 (movdf_integer_rex64): New.
461 (pushsf_rex64): Fix output constraints.
462 * config/i386/sse.md (sse2_loadld): Split rm alternative, use Yi.
463 (sse2_stored): Likewise.
464 (sse2_storeq_rex64): New.
465 * config/i386/i386.c (x86_inter_unit_moves): Enable for not
467 (ix86_secondary_memory_needed): Don't bypass TARGET_INTER_UNIT_MOVES
468 for optimize_size. Remove SF/DFmode hack.
470 2007-02-09 Dwarakanath Rajagopal <dwarak.rajagopal@amd.com>
472 * config/i386/driver-i386.c: Turn on -mtune=native for AMDFAM10.
475 2007-02-09 Nathan Sidwell <nathan@codesourcery.com>
476 Richard Sandiford <richard@codesourcery.com>
478 * config.gcc (m68010-*-netbsdelf*, m68k*-*-netbsdelf*)
479 (m68k*-*-openbsd*, m68k-*-linux*): Set default_cf_cpu.
480 (m68k-*-aout*, m68k-*-coff*, m68k-*-uclinux*, m68k-*-rtems*): Add
481 m68k/t-mlib to tmake_file.
482 (m68020-*-elf*, m68k-*-elf*): Likewise. Add t-m68kbare as well.
483 (m68k*-*-*): Use --with-arch to pick a default for --with-cpu.
484 (m680[012]0-*-*, m68k*-*-*): Add support for --with-arch.
485 Allow it to be cf or m68k. Set m68k_arch_family. If that
486 variable is not empty, add t-$m68k_arch_family to tmake_file.
487 Add t-mlibs to tmake_file.
488 * doc/install.texi: Document --with-arch=m68k and --with-arch=cf.
489 * config/m68k/t-cf: New file.
490 * config/m68k/t-m68k: Likewise.
491 * config/m68k/t-mlibs: Likewise.
492 * config/m68k/t-m68kbare (MULTILIB_OPTIONS, MULTILIB_DIRNAMES)
493 (MULTILIB_MATCHES, MULTILIB_EXCEPTIONS): Delete.
494 (M68K_MLIB_DIRNAMES, M68K_MLIB_OPTIONS): Define.
495 * config/m68k/t-m68kelf (MULTILIB_OPTIONS, MULTILIB_DIRNAMES)
496 (MULTILIB_MATCHES, MULTILIB_EXCEPTIONS, LIBGCC, INSTALL_LIBGCC):
498 * config/m68k/t-openbsd (MULTILIB_OPTIONS, LIBGCC): Delete.
499 (INSTALL_LIBGCC): Delete.
500 (M68K_MLIB_DIRNAMES, M68K_MLIB_OPTIONS): Define.
501 * config/m68k/t-rtems (MULTILIB_OPTIONS, MULTILIB_DIRNAMES)
502 (MULTILIB_MATCHES, MULTILIB_EXCEPTIONS): Delete.
503 (M68K_MLIB_CPU): Define.
504 * config/m68k/t-uclinux (MULTILIB_OPTIONS, MULTILIB_DIRNAMES)
505 (MULTILIB_MATCHES, MULTILIB_EXCEPTIONS): Delete.
506 (M68K_MLIB_CPU, M68K_MLIB_OPTIONS, M68K_MLIB_DIRNAMES): Define.
508 2007-02-09 Zdenek Dvorak <dvorakz@suse.cz>
509 Richard Guenther <rguenther@suse.de>
512 * fold-const.c (fold_comparison): Handle obfuscated comparisons
513 against INT_MIN/INT_MAX.
514 * tree-ssa-loop-ivcanon.c (remove_empty_loop): Print to dump
515 file if a loop is removed.
517 2007-02-09 Joseph Myers <joseph@codesourcery.com>
519 * calls.c (store_one_arg): Pass correct alignment to
520 emit_push_insn for non-BLKmode values.
521 * expr.c (emit_push_insn): If STRICT_ALIGNMENT, copy to an
522 unaligned stack slot via a suitably aligned slot.
524 2007-02-08 DJ Delorie <dj@redhat.com>
526 * config/m32c/m32c.c (m32c_unpend_compare): Add default to silence
528 (legal_subregs): Use unsigned char, make const.
529 (m32c_illegal_subreg_p): Use ARRAY_SIZE. Delete unused variables.
531 2007-02-08 Paul Brook <paul@codesourcery.com>
533 * config/arm/lib1funcs.asm (RETLDM): Pop directly into PC when no
534 special interworking needed.
536 2007-02-08 Harsha Jagasia <harsha.jagasia@amd.com>
538 * config/i386/xmmintrin.h: Make inclusion of emmintrin.h
539 conditional to __SSE2__.
540 (Entries below should have been added to first ChangeLog
541 entry for amdfam10 dated 2007-02-05)
542 * config/i386/emmintrin.h: Generate #error if __SSE2__ is not
544 * config/i386/pmmintrin.h: Generate #error if __SSE3__ is not
546 * config/i386/tmmintrin.h: Generate #error if __SSSE3__ is not
549 2007-02-08 DJ Delorie <dj@redhat.com>
551 * config/m32c/m32c-protos.h (m32c_illegal_subreg_p): New.
552 * config/m32c/m32c.c (legal_subregs): New.
553 (m32c_illegal_subreg_p): New.
554 * config/m32c/predicates.md (m32c_any_operand): Use it to reject
555 unsupported subregs of hard regs.
557 2007-02-08 Jan Hubicka <jh@suse.cz>
559 * tree-cfg.c (bsi_replace): Shortcut when replacing the statement with
560 the same one; always update histograms.
562 2007-02-08 Diego Novillo <dnovillo@redhat.com>
564 * passes.c (init_optimization_passes): Tidy comment.
566 2007-02-08 Roger Sayle <roger@eyesopen.com>
568 * simplify-rtx.c (simplify_unary_operation_1) <POPCOUNT>: We can
569 strip zero_extend, bswap and rotates from POCOUNT's argument.
570 <PARITY>: Likewise, we can strip not, bswap, sign_extend,
571 zero_extend and rotates from PARITY's argument.
572 <BSWAP>: A byte-swap followed by a byte-swap is an identity.
573 (simplify_const_unary_operation) <BSWAP>: Evaluate the byte-swap
574 of an integer constant at compile-time.
576 2007-02-08 Diego Novillo <dnovillo@redhat.com>
579 * tree-flow.h (struct var_ann_d): Remove field 'is_used'.
581 * tree-ssa-alias.c (compute_is_aliased): Remove. Update all
584 * tree-ssa-live.c (remove_unused_locals): Do not remove
585 TREE_ADDRESSABLE variables.
586 * tree-ssa-structalias.c (compute_points_to_sets): Tidy.
587 * tree-ssa-operands.c (add_virtual_operand): Remove argument
588 FOR_CLOBBER. Update all users.
589 If VAR has an associated alias set, add a virtual operand for
590 it if no alias is found to conflict with the memory reference.
592 2007-02-07 Jan Hubicka <jh@suse.cz>
593 Robert Kidd <rkidd@crhc.uiuc.edu>
595 * value-prof.c (visit_hist, free_hist): Return 1 instead of 0.
597 2007-02-07 Ian Lance Taylor <iant@google.com>
599 * lower-subreg.c (simple_move): Reject PARTIAL_INT modes.
601 2007-02-07 Roger Sayle <roger@eyesopen.com>
603 * config/rs6000/rs6000.md (ctz<mode>2, ffs<mode>2, popcount<mode>2,
604 parity<mode>2, smulsi3_highpart, abstf2_internal, allocate_stack,
605 tablejumpdi, movsi_to_cr_one): Remove constraints from
606 define_expand's match_operands.
608 2007-02-07 Roger Sayle <roger@eyesopen.com>
610 * global.c (compute_regsets): Move declatation of "i" inside of
611 #ifdef ELIMINABLE_REGS to avoid unused variable bootstrap failure.
613 2007-02-07 Jakub Jelinek <jakub@redhat.com>
616 * gimplify.c (gimplify_scan_omp_clauses): Remove special casing
617 of INDIRECT_REF <RESULT_DECL>.
619 * config/i386/i386.c (override_options): Set PTA_SSSE3 for core2.
621 2007-02-06 J"orn Rennecke <joern.rennecke@arc.com>
622 Kaz Kojima <kkojima@gcc.gnu.org>
625 * config/sh/sh.c (expand_cbranchdi4): Use scratch register
627 (sh_initialize_trampoline): Add parentheses to avoid a warning.
629 2007-02-06 Zdenek Dvorak <dvorakz@suse.cz>
631 * doc/loop.texi: Document possibility not to perform disambiguation
632 of loops with multiple latches.
633 * cfgloopmanip.c (alp_enum_p): Removed.
634 (add_loop): Handle subloops. Use get_loop_body_with_size.
635 (create_preheader): Do not allow ENTRY_BLOCK_PTR to be preheader.
636 * cfghooks.c (redirect_edge_and_branch_force): Set dominator for
637 the new forwarder block.
638 (make_forwarder_block): Only call new_bb_cbk if it is not NULL.
639 Handle the case latch is NULL.
640 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Avoid cfg modifications
641 when marking loop exits.
642 * ifcvt.c (if_convert): Ditto. Mark loop exits even if cfg cannot
644 * loop-init.c (loop_optimizer_init): Do not modify cfg. Call
645 disambiguate_loops_with_multiple_latches.
646 * tree-cfgcleanup.c (cleanup_tree_cfg_loop): Calculate dominators
647 before fix_loop_structure.
648 * cfgloop.c: Include pointer-set.h and output.h.
649 (canonicalize_loop_headers, HEADER_BLOCK, LATCH_EDGE,
650 update_latch_info, mfb_keep_just, mfb_keep_nonlatch): Removed.
651 (get_loop_latch_edges, find_subloop_latch_edge_by_profile,
652 find_subloop_latch_edge_by_ivs, find_subloop_latch_edge,
653 mfb_redirect_edges_in_set, form_subloop, merge_latch_edges,
654 disambiguate_multiple_latches, get_loop_body_with_size,
655 disambiguate_loops_with_multiple_latches): New functions.
656 (flow_loop_dump): Dump multiple latch edges.
657 (flow_loop_nodes_find): Handle loops with multiple latches.
658 (flow_loops_find): Ditto. Do not call canonicalize_loop_headers.
659 (glb_enum_p): Modified.
660 (get_loop_body): Use get_loop_body_with_size.
661 * cfgloop.h (LOOPS_HAVE_RECORDED_EXITS): New flag.
662 (AVOID_CFG_MODIFICATIONS): New constant.
663 (disambiguate_loops_with_multiple_latches, add_loop,
664 get_loop_body_with_size): Declare.
665 * Makefile.in (cfgloop.o): Add pointer-set.h and output.h.
667 2007-02-06 Seongbae Park <seongbae.park@gmail.com>
670 * global.c (compute_regsets): New function.
671 (global_alloc): Refactored ELIMINABLE_REGSET
672 and NO_GLOBAL_ALLOC_REGS computation out.
673 (rest_of_handle_global_alloc): Call compute_regsets()
674 for non-optimizing case.
676 2007-02-06 Richard Henderson <rth@redhat.com>
678 * config/i386/constraints.md (Y2): Rename from Y.
679 (Yi): New constraint.
680 * config/i386/i386.md (movsi_1, movdi_2, pushdf_nointeger,
681 pushdf_integer, movdf_nointeger, movdf_integer, zero_extendsidi2_32,
682 zero_extendsidi2_rex64, truncxfdf2_mixed): Change Y constraints to Y2.
683 (extendsfdf2_mixed, extendsfdf2_sse, truncdfsf_fast_mixed,
684 truncdfsf_fast_sse, truncdfsf_mixed, fix_truncdfdi_sse,
685 fix_truncdfsi_sse, floatsidf2_mixed, floatsidf2_sse,
686 floatdidf2_mixed, floatdidf2_sse, absnegdf2_mixed,
687 absnegdf2_sse, sse_setccdf, fop_df_comm_mixed, fop_df_comm_sse,
688 fop_df_1_mixed, fop_df_1_sse): Change Y constraints to x.
689 * config/i386/mmx.md (mov<MMXMODEI>_internal_rex64,
690 mov<MMXMODEI>_internal, movv2sf_internal_rex64, movv2sf_internal,
691 vec_extractv2si_1): Change Y constraints to Y2.
692 * config/i386/sse.md (vec_setv4sf_0, vec_concatv2df, vec_dupv4si,
693 vec_dupv2di, sse2_concatv2si, vec_concatv4si_1, vec_concatv2di):
694 Change Y constraints to Y2.
695 (sse2_loadld): Change Y constraints to x.
697 2007-02-06 Roger Sayle <roger@eyesopen.com>
699 * config/rs6000/rs6000.md (popcount<mode>2): Rewrite.
700 (parity<mode>2): New define_expand using rs6000_emit_parity.
701 * config/rs6000/rs6000.c (rs6000_emit_popcount,
702 rs6000_emit_parity): New functions.
703 * config/rs6000/rs6000-protos.h (rs6000_emit_popcount,
704 rs6000_emit_parity): Prototype here.
706 2007-02-06 Ian Lance Taylor <iant@google.com>
708 * lower-subreg.c (simple_move_operand): Reject CONST.
709 (resolve_clobber): Call validate_change rather than directly
710 assigning to XEXP (pat, 0).
712 2006-02-06 Paolo Bonzini <bonzini@gnu.org>
714 * Makefile.in (tree-ssa-loop-ivopts.o): Add pointer-set.h dependency.
715 (tree-ssa-reassoc.o): Add pointer-set.h dependency.
716 (tree-cfg.o): Remove hashtab.h dependency.
718 * tree-ssa-loop-ivopts.c: Include pointer-set.h.
719 (struct ivopts_data): Change niters to pointer_map_t.
720 (struct nfe_cache_elt, nfe_hash, nfe_eq): Delete.
721 (niter_for_exit): Create pointer_map on demand. Change for
723 (tree_ssa_iv_optimize_init): Initialize data->niters to NULL.
724 (free_loop_data): Destroy data->niters if created and reset field.
725 (tree_ssa_iv_optimize_finalize): Don't delete data->niters here.
726 (tree_ssa_iv_optimize_loop): Check for presence of stale data.
728 * tree-ssa-reassoc.c: Include pointer-set.h.
729 (bb_rank): Change to long *.
730 (operand_rank): Change to pointer_map_t.
731 (find_operand_rank): Return long, -1 if not found. Declare as inline.
732 (insert_operand_rank): Accept long.
733 (operand_entry_hash, operand_entry_eq): Remove.
734 (get_rank): Return long. Adjust for changes above.
735 (init_reassoc): Change rank type to long. Adjust creation of bb_rank
737 (fini_reassoc): Delete operand_rank with pointer_map_destroy.
739 * tree-ssa-structalias.c (vi_for_tree): Change to pointer_map.
740 (struct tree_vi, tree_vi_t, tree_vi_hash, tree_vi_eq): Delete.
741 (insert_vi_for_tree): Rewrite for pointer_map API. Assert argument
743 (lookup_vi_for_tree): Rewrite for pointer_map API. Return varinfo_t
744 directly since it cannot be NULL.
745 (get_vi_for_tree): Rewrite for pointer_map API.
746 (find_what_p_points_to): Adjust for change to lookup_vi_for_tree.
747 (init_alias_vars): Create vi_for_tree as pointer_map.
748 (delete_points_to_sets): Delete vi_for_tree using pointer_map_destroy.
750 * tree-cfg.c: Don't include hashtab.h.
751 (edge_to_cases): Declare as pointer_map.
752 (struct edge_to_cases_elt, edge_to_cases_hash, edge_to_cases_eq):
754 (edge_to_cases_cleanup): Rewrite as pointer_map_traverse callback.
755 (start_recording_case_labels): Create edge_to_cases as pointer_map.
756 (end_recoding_case_labels): Cleanup edge_to_cases manually before
758 (record_switch_edge): Delete.
759 (get_cases_for_edge): Adjust for pointer_map API, inline
760 record_switch_edge (rewritten for new API), remove goto.
762 2006-02-06 Paolo Bonzini <bonzini@gnu.org>
764 * Makefile.in (tree-nested.o): Add pointer-set.h dependency.
765 * tree-nested.c: Include pointer-set.h.
766 (var_map_elt, var_map_eq, var_map_hash): Delete.
767 (struct nesting_info): Remove GTY marker. Change the two htab_t's
769 (nesting_info_bitmap_obstack): New.
770 (lookup_field_for_decl): Adjust for pointer_map API.
771 (lookup_tramp_for_decl): Adjust for pointer_map API.
772 (get_nonlocal_debug_decl): Adjust for pointer_map API.
773 (get_local_debug_decl): Adjust for pointer_map API.
774 (convert_nl_goto_reference): Adjust for pointer_map API.
775 (convert_nl_goto_receiver): Adjust for pointer_map API.
776 (create_nesting_tree): Create outside GGC space. Create bitmap on
777 the new obstack. Create field_map and var_map as pointer_maps.
778 (free_nesting_tree): Adjust for changes to create_nesting_tree.
780 (lower_nested_functions): Move root here, no need to NULL it.
781 Initialize and release the obstack.
783 2007-02-06 Paolo Bonzini <bonzini@gnu.org>
785 * tree.c (tree_int_map_hash, tree_int_map_eq, tree_int_map_marked_p):
786 Remove prototypes and make them non-static.
787 (struct tree_int_map): Remove.
788 * tree.h (struct tree_int_map): Move here, turning TO into an
790 (tree_int_map_hash, tree_int_map_eq, tree_int_map_marked_p): Declare.
792 * tree.h (TREE_COMPLEXITY): Remove.
793 (struct tree_exp): Remove complexity field.
794 * tree.c (build1_stat): Don't set it.
796 2007-02-06 Dorit Nuzman <dorit@il.ibm.com>
797 Victor Kaplansky <victork@il.ibm.com>
799 * tree-vectorizer.c (vect_is_simple_use): Support induction.
800 (vect_is_simple_reduction): Support reduction with induction as
802 (vect_is_simple_iv_evolution): Fix formatting.
803 * tree-vect-analyze.c (vect_mark_stmts_to_be_vectorized): Fix
804 formatting. Don't mark induction phis for vectorization.
805 (vect_analyze_scalar_cycles): Analyze all inductions, then reductions.
806 * tree-vect-transform.c (get_initial_def_for_induction): New function.
807 (vect_get_vec_def_for_operand): Support induction.
808 (vect_get_vec_def_for_stmt_copy): Fix formatting and add check for
810 (vectorizable_reduction): Support reduction with induction as one of
812 (vectorizable_type_demotion): Use def-type of stmt argument rather
815 * tree-ssa-loop.c (gate_scev_const_prop): Return the value of
816 flag_tree_scev_cprop.
817 * common.opt (tree-scev-cprop): New flag.
819 * tree-vect-transform.c (vect_create_destination_var): Use 'kind' in
820 call to vect_get_new_vect_var.
822 2007-02-06 Ira Rosen <irar@il.ibm.com>
824 * tree-vect-patterns.c (vect_recog_widen_mult_pattern): Check that
826 (vect_pattern_recog_1): Likewise.
828 2007-02-05 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
830 * fold-const.c (negate_expr_p): Handle CONJ_EXPR.
831 (fold_negate_expr): Likewise.
833 2007-02-05 Alexandre Oliva <aoliva@redhat.com>
836 * dwarf2out.c (modified_type_die): Follow DECL_ORIGINAL_TYPE
837 even if cv-qualification is the same.
839 2007-02-05 Geoffrey Keating <geoffk@apple.com>
841 * config/rs6000/darwin-tramp.asm (__trampoline_setup): Call
842 __enable_execute_stack on completion.
844 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
846 * config/i386/athlon.md (athlon_fldxf_k8, athlon_fld_k8,
847 athlon_fstxf_k8, athlon_fst_k8, athlon_fist, athlon_fmov,
848 athlon_fadd_load, athlon_fadd_load_k8, athlon_fadd, athlon_fmul,
849 athlon_fmul_load, athlon_fmul_load_k8, athlon_fsgn,
850 athlon_fdiv_load, athlon_fdiv_load_k8, athlon_fdiv_k8,
851 athlon_fpspc_load, athlon_fpspc, athlon_fcmov_load,
852 athlon_fcmov_load_k8, athlon_fcmov_k8, athlon_fcomi_load_k8,
853 athlon_fcomi, athlon_fcom_load_k8, athlon_fcom): Added amdfam10.
855 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
857 * config/i386/i386.md (x86_sahf_1, cmpfp_i_mixed, cmpfp_i_sse,
858 cmpfp_i_i387, cmpfp_iu_mixed, cmpfp_iu_sse, cmpfp_iu_387,
859 swapsi, swaphi_1, swapqi_1, swapdi_rex64, fix_truncsfdi_sse,
860 fix_truncdfdi_sse, fix_truncsfsi_sse, fix_truncdfsi_sse,
861 x86_fldcw_1, floatsisf2_mixed, floatsisf2_sse, floatdisf2_mixed,
862 floatdisf2_sse, floatsidf2_mixed, floatsidf2_sse,
863 floatdidf2_mixed, floatdidf2_sse, muldi3_1_rex64, mulsi3_1,
864 mulsi3_1_zext, mulhi3_1, mulqi3_1, umulqihi3_1, mulqihi3_insn,
865 umulditi3_insn, umulsidi3_insn, mulditi3_insn, mulsidi3_insn,
866 umuldi3_highpart_rex64, umulsi3_highpart_insn,
867 umulsi3_highpart_zext, smuldi3_highpart_rex64,
868 smulsi3_highpart_insn, smulsi3_highpart_zext, x86_64_shld,
869 x86_shld_1, x86_64_shrd, sqrtsf2_mixed, sqrtsf2_sse,
870 sqrtsf2_i387, sqrtdf2_mixed, sqrtdf2_sse, sqrtdf2_i387,
871 sqrtextendsfdf2_i387, sqrtxf2, sqrtextendsfxf2_i387,
872 sqrtextenddfxf2_i387): Added amdfam10_decode.
874 * config/i386/athlon.md (athlon_idirect_amdfam10,
875 athlon_ivector_amdfam10, athlon_idirect_load_amdfam10,
876 athlon_ivector_load_amdfam10, athlon_idirect_both_amdfam10,
877 athlon_ivector_both_amdfam10, athlon_idirect_store_amdfam10,
878 athlon_ivector_store_amdfam10): New define_insn_reservation.
879 (athlon_idirect_loadmov, athlon_idirect_movstore): Added
882 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
884 * config/i386/athlon.md (athlon_call_amdfam10,
885 athlon_pop_amdfam10, athlon_lea_amdfam10): New
886 define_insn_reservation.
887 (athlon_branch, athlon_push, athlon_leave_k8, athlon_imul_k8,
888 athlon_imul_k8_DI, athlon_imul_mem_k8, athlon_imul_mem_k8_DI,
889 athlon_idiv, athlon_idiv_mem, athlon_str): Added amdfam10.
891 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
893 * config/i386/athlon.md (athlon_sseld_amdfam10,
894 athlon_mmxld_amdfam10, athlon_ssest_amdfam10,
895 athlon_mmxssest_short_amdfam10): New define_insn_reservation.
897 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
899 * config/i386/athlon.md (athlon_sseins_amdfam10): New
900 define_insn_reservation.
901 * config/i386/i386.md (sseins): Added sseins to define_attr type
902 and define_attr unit.
903 * config/i386/sse.md: Set type attribute to sseins for insertq
906 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
908 * config/i386/athlon.md (sselog_load_amdfam10, sselog_amdfam10,
909 ssecmpvector_load_amdfam10, ssecmpvector_amdfam10,
910 ssecomi_load_amdfam10, ssecomi_amdfam10,
911 sseaddvector_load_amdfam10, sseaddvector_amdfam10): New
912 define_insn_reservation.
913 (ssecmp_load_k8, ssecmp, sseadd_load_k8, seadd): Added amdfam10.
915 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
917 * config/i386/athlon.md (cvtss2sd_load_amdfam10,
918 cvtss2sd_amdfam10, cvtps2pd_load_amdfam10, cvtps2pd_amdfam10,
919 cvtsi2sd_load_amdfam10, cvtsi2ss_load_amdfam10,
920 cvtsi2sd_amdfam10, cvtsi2ss_amdfam10, cvtsd2ss_load_amdfam10,
921 cvtsd2ss_amdfam10, cvtpd2ps_load_amdfam10, cvtpd2ps_amdfam10,
922 cvtsX2si_load_amdfam10, cvtsX2si_amdfam10): New
923 define_insn_reservation.
925 * config/i386/sse.md (cvtsi2ss, cvtsi2ssq, cvtss2si,
926 cvtss2siq, cvttss2si, cvttss2siq, cvtsi2sd, cvtsi2sdq,
927 cvtsd2si, cvtsd2siq, cvttsd2si, cvttsd2siq,
928 cvtpd2dq, cvttpd2dq, cvtsd2ss, cvtss2sd,
929 cvtpd2ps, cvtps2pd): Added amdfam10_decode attribute.
931 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
933 * config/i386/athlon.md (athlon_ssedivvector_amdfam10,
934 athlon_ssedivvector_load_amdfam10, athlon_ssemulvector_amdfam10,
935 athlon_ssemulvector_load_amdfam10): New define_insn_reservation.
936 (athlon_ssediv, athlon_ssediv_load_k8, athlon_ssemul,
937 athlon_ssemul_load_k8): Added amdfam10.
939 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
941 * config/i386/i386.h (TARGET_SSE_UNALIGNED_MOVE_OPTIMAL): New macro.
942 (x86_sse_unaligned_move_optimal): New variable.
944 * config/i386/i386.c (x86_sse_unaligned_move_optimal): Enable for
946 (ix86_expand_vector_move_misalign): Add code to generate movupd/movups
947 for unaligned vector SSE double/single precision loads for AMDFAM10.
949 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
951 * config/i386/i386.h (TARGET_AMDFAM10): New macro.
952 (TARGET_CPU_CPP_BUILTINS): Add code for amdfam10.
953 Define TARGET_CPU_DEFAULT_amdfam10.
954 (TARGET_CPU_DEFAULT_NAMES): Add amdfam10.
955 (processor_type): Add PROCESSOR_AMDFAM10.
957 * config/i386/i386.md: Add amdfam10 as a new cpu attribute to match
958 processor_type in config/i386/i386.h.
959 Enable imul peepholes for TARGET_AMDFAM10.
961 * config.gcc: Add support for --with-cpu option for amdfam10.
963 * config/i386/i386.c (amdfam10_cost): New variable.
964 (m_AMDFAM10): New macro.
965 (m_ATHLON_K8_AMDFAM10): New macro.
966 (x86_use_leave, x86_push_memory, x86_movx, x86_unroll_strlen,
967 x86_cmove, x86_3dnow_a, x86_deep_branch, x86_use_simode_fiop,
968 x86_promote_QImode, x86_integer_DFmode_moves,
969 x86_partial_reg_dependency, x86_memory_mismatch_stall,
970 x86_accumulate_outgoing_args, x86_arch_always_fancy_math_387,
971 x86_sse_partial_reg_dependency, x86_sse_typeless_stores,
972 x86_use_ffreep, x86_use_incdec, x86_four_jump_limit,
973 x86_schedule, x86_use_bt, x86_cmpxchg16b, x86_pad_returns):
974 Enable/disable for amdfam10.
975 (override_options): Add amdfam10_cost to processor_target_table.
976 Set up PROCESSOR_AMDFAM10 for amdfam10 entry in
977 processor_alias_table.
978 (ix86_issue_rate): Add PROCESSOR_AMDFAM10.
979 (ix86_adjust_cost): Add code for amdfam10.
981 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
983 * config/i386/i386.opt: Add new Advanced Bit Manipulation (-mabm)
984 instruction set feature flag. Add new (-mpopcnt) flag for popcnt
985 instruction. Add new SSE4A (-msse4a) instruction set feature flag.
986 * config/i386/i386.h: Add builtin definition for SSE4A.
987 * config/i386/i386.md: Add support for ABM instructions
989 * config/i386/sse.md: Add support for SSE4A instructions
990 (movntss, movntsd, extrq, insertq).
991 * config/i386/i386.c: Add support for ABM and SSE4A builtins.
992 Add -march=amdfam10 flag.
993 * config/i386/ammintrin.h: Add support for SSE4A intrinsics.
994 * doc/invoke.texi: Add documentation on flags for sse4a, abm, popcnt
996 * doc/extend.texi: Add documentation for SSE4A builtins.
998 2007-02-05 Bob Wilson <bob.wilson@acm.org>
1000 * config/xtensa/xtensa.c (constantpool_mem_p): Skip over SUBREGs.
1002 2007-02-05 Richard Guenther <rguenther@suse.de>
1004 * tree-vectorizer.h (vectorizable_function): Add argument type
1005 argument, change return type.
1006 * tree-vect-patterns.c (vect_recog_pow_pattern): Adjust caller.
1007 * tree-vect-transform.c (vectorizable_function): Handle extra
1008 argument, return vectorized function decl.
1009 (build_vectorized_function_call): Remove.
1010 (vectorizable_call): Handle calls with result and argument types
1011 differing. Handle loop vectorization factor correctly.
1012 * targhooks.c (default_builtin_vectorized_function): Adjust for
1014 * targhooks.h (default_builtin_vectorized_function): Likewise.
1015 * target.h (builtin_vectorized_function): Add argument type
1017 * config/i386/i386.c (ix86_builtin_vectorized_function): Handle
1018 extra argument, allow vectorizing of lrintf.
1019 * doc/tm.texi (TARGET_VECTORIZE_BUILTIN_VECTORIZED_FUNCTION): Adjust
1020 documentation of target hook.
1022 2007-02-05 Hans-Peter Nilsson <hp@axis.com>
1025 * config/cris/cris.md ("*andsi_movu", "*andsi_clear", "*andhi_movu")
1026 ("*andhi_clear", andu (casesi+45)): For size-changed operand where
1027 memory is allowed, require !side_effects_p, not just !MEM_VOLATILE_P.
1029 2007-02-05 Roger Sayle <roger@eyesopen.com>
1031 * fold-const.c (fold_unary) <REAL_PART>: Test for availability of
1032 BUILT_IN_COS before simplifying REAL_PART(CEXPI)) to COS.
1033 <IMAG_PART>: Likewise, check for availability of BUILT_IN_SIN.
1034 * builtins.c (fold_builtin_sincos): Check for TARGET_C99_FUNCTIONS
1035 before canonicalizing sincos to cexpi.
1036 (fold_builtin_cexp): Likewise, for canonicalizing cexp to cexpi.
1038 2007-02-05 Roger Sayle <roger@eyesopen.com>
1040 * config/alpha/alpha.c (alpha_add_builtins): New Helper function.
1041 Set TREE_READONLY and TREE_NOTHROW directly, not via attributes.
1042 (alpha_init_builtins): Use alpha_add_builtins to process tables.
1044 2007-02-05 Roger Sayle <roger@eyesopen.com>
1046 * mips-tfile.c (initialize_init_file): Correct endianness test.
1048 2007-02-05 Kazu Hirata <kazu@codesourcery.com>
1050 * config/m68k/m68k.md (pushdi-1, pushdi, movsi+1): Don't use
1053 2007-02-05 Richard Sandiford <richard@codesourcery.com>
1055 * dwarf2out.c (dwarf2out_frame_debug_expr): Record the register
1056 saves in a PARALLEL before the register assignments.
1058 2007-02-05 Richard Sandiford <richard@codesourcery.com>
1060 * doc/tm.texi (DWARF_ALT_FRAME_RETURN_COLUMN): Do not require
1061 DWARF_FRAME_RETURN_COLUMN to be a general register.
1062 * dwarf2out.c (init_return_column_size): New function, split from...
1063 (expand_builtin_init_dwarf_reg_sizes): ...here. Allow both
1064 DWARF_FRAME_RETURN_COLUMN and DWARF_ALT_FRAME_RETURN_COLUMN
1065 to be nongeneral registers.
1066 * config/m68k/m68k.h (DWARF_FRAME_REGNUM): Only map FP and
1068 (DWARF_FRAME_REGISTERS, DWARF_FRAME_RETURN_COLUMN): Define.
1069 (DWARF_ALT_FRAME_RETURN_COLUMN): Define.
1071 2007-02-04 Zdenek Dvorak <dvorakz@suse.cz>
1073 * cfgcleanup.c (try_optimize_cfg): Avoid removing ENTRY_BLOCK_PTR.
1075 2007-02-04 Zdenek Dvorak <dvorakz@suse.cz>
1077 * cfgloopmanip.c (loop_delete_branch_edge): Removed.
1078 (remove_path): Use can_remove_branch_p and remove_branch instead
1079 of loop_delete_branch_edge.
1080 * tree-ssa-loop-manip.c (scale_dominated_blocks_in_loop): New function.
1081 (tree_transform_and_unroll_loop): Remove dead branches immediately.
1082 Update profile using scale_dominated_blocks_in_loop.
1083 * cfghooks.c (can_remove_branch_p, remove_branch): New functions.
1084 * cfghooks.h (struct cfg_hooks): Add can_remove_branch_p.
1085 (can_remove_branch_p, remove_branch): Declare.
1086 * tree-cfg.c (tree_can_remove_branch_p): New function.
1087 (tree_cfg_hooks): Add tree_can_remove_branch_p.
1088 * cfgrtl.c (rtl_can_remove_branch_p): New function.
1089 (rtl_cfg_hooks, cfg_layout_rtl_cfg_hook): Add rtl_can_remove_branch_p.
1091 2007-02-05 Jan Hubicka <jh@suse.cz>
1094 * ipa-inline.c (cgraph_clone_inlined_nodes): When there are unanalyzed
1095 nodes in cgraph, don't remove offline copy of the function.
1097 2007-02-04 Jan Hubicka <jh@suse.cz>
1099 * tree-sra.c (sra_walk_expr): Add linebreaks. BITFIELD_REFs into
1100 vectors might cause maybe_lookup_element_for_expr to be called
1101 on non-sra-candidate.
1103 2007-02-04 Kazu Hirata <kazu@codesourcery.com>
1105 * config/bfin/bfin-modes.def, config/bfin/bfin.c,
1106 config/bfin/bfin.md, config/bfin/predicates.md: Follow
1107 spelling conventions.
1109 2007-02-04 Richard Guenther <rguenther@suse.de>
1112 * fold-const.c (try_move_mult_to_index): Make sure to not
1113 overflow one dimension of a multi-dimensional array access.
1115 2007-02-04 Jan Hubicka <jh@suse.cz>
1117 * passes.c (init_optimization_passes): Reindent.
1119 2007-02-04 Jan Hubicka <jh@suse.cz>
1120 Eric Botcazou <ebotcazou@adacore.com>
1122 * tree-optimize.c (has_abnormal_outgoing_edge_p): Move to...
1123 (execute_fixup_cfg): Break out the abnormal goto code.
1124 * tree-inline.c (has_abnormal_outgoing_edge_p): ...here.
1125 (make_nonlocal_label_edges): Move here from execute_fixup_cfg.
1126 (optimize_inline_calls): Call make_nonlocal_label_edges.
1128 2007-02-04 Jan Hubicka <jh@suse.cz>
1130 * tree-ssa-copyrename.c (copy_rename_partition_coalesce): Return
1131 true when something was changed.
1132 (rename_ssa_copies): When something was changed, do
1133 TODO_remove_unused_locals.
1134 * tree-ssa-forwprop.c (tree_ssa_forward_propagate_single_use_value):
1135 add TODO_remove_unused_locals when instruction was removed.
1137 2007-02-04 Jan Hubicka <jh@suse.cz>
1139 * ipa-inline.c (try_inline): Improve debug output; work on already
1141 (cgraph_decide_inlining_incrementally): Indent; improve debug output;
1142 call try_inline for already inlined edges too when flattening;
1143 inline also functions that make callee growth but overall unit size
1146 2007-02-04 Kazu Hirata <kazu@codesourcery.com>
1148 * config/m32c/bitops.md, config/m32c/jump.md,
1149 config/m32c/m32c.c, config/m32c/m32c.h, config/m32r/m32r.c,
1150 config/m32r/m32r.h, config/m32r/m32r.md,
1151 config/m32r/predicates.md, config/m68hc11/larith.asm,
1152 config/m68hc11/m68hc11.c, config/m68hc11/m68hc11.h,
1153 config/m68k/m68k.h, config/mcore/mcore.md, config/mips/4k.md,
1154 config/mips/mips-protos.h, config/mips/mips.c,
1155 config/mips/mips.h, config/mips/mips.md, config/mips/mips16.S,
1156 config/mn10300/mn10300.h, config/mn10300/predicates.md,
1157 config/mt/mt.c, config/mt/mt.h, config/mt/mt.md: Follow
1158 spelling conventions.
1160 * config/v850/v850.c, config/v850/v850.h, config/v850/v850.md:
1161 Follow spelling conventions.
1163 2007-02-03 Douglas Gregor <doug.gregor@gmail.com>
1165 * c-opts.c (c_common_post_options): If C++0x mode is enabled, don't
1166 warn about C++0x compatibility.
1168 2007-02-04 Kazu Hirata <kazu@codesourcery.com>
1170 * config/h8300/h8300.c, config/h8300/h8300.h,
1171 config/h8300/h8300.md: Follow spelling conventions.
1173 2007-02-03 Uros Bizjak <ubizjak@gmail.com>
1176 * combine.c (try_combine): Do not substitute source operand
1177 with constants wider than 2 * HOST_BITS_PER_WIDE_INT.
1179 2007-02-03 Jan Hubicka <jh@suse.cz>
1181 PR gcov-profile/30650
1182 * value-prof.c (stringop_block_profile): Fix handling of size counter;
1183 do not divide by zero for never executed counters.
1184 (tree_find_values_to_profile): Fix counters.
1185 * gcov-ui.h (GCOV_COUNTER_AVERAGE, GCOV_COUNTER_IOR): Fix comments.
1187 2007-02-03 Ian Lance Taylor <iant@google.com>
1189 * lower-subreg.c (simple_move_operand): New static function,
1190 broken out of simple_move. Reject LABEL_REF, SYMBOL_REF, and HIGH
1192 (simple_move): Call simple_move_operand.
1193 (find_decomposable_subregs): Add special handling of MEMs.
1194 (can_decompose_p): Rename from cannot_decompose_p. Reverse
1195 meaning of return value. If we see a hard register, test whether
1196 it can store a word_mode value. Change all callers.
1198 2007-02-03 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
1200 * pa.md (addvdi3, addvsi3, subvdi3, subvsi3, negvdi2, negvsi2): New
1201 ftrapv insns and expanders.
1202 (subdi3): Change define_expand operand 1 to arith11_operand, and
1203 operand 2 to reg_or_0_operand. Change constraints of 64-bit insn
1204 pattern to handle reg_or_0 operands. Revise 32-bit insn pattern to
1205 handle 11-bit constants and reg_or_0 operands in operands 1 and 2,
1209 * varasm.c (notice_global_symbol): Treat global objects as weak when
1212 2007-02-03 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1214 * emit-rtl.c (dconstpi): Delete.
1216 (init_emit_once): Delete dconstpi and init dconstsqrt2.
1217 * real.h (dconstpi): Delete.
1219 * builtins.c (fold_builtin_cabs): Use dconstsqrt2.
1220 (fold_builtin_hypot): Likewise.
1222 2007-02-03 Tom Tromey <tromey@redhat.com>
1225 * gcc.c (cpp_unique_options): Any of -ggdb3, -gstabs3,
1226 -gcoff3, -gxcoff3, -gvms3 implies -dD.
1228 2007-02-03 Kazu Hirata <kazu@codesourcery.com>
1230 * c-decl.c, config/avr/avr.c, config/avr/avr.h,
1231 config/m68k/m68k.c, config/m68k/netbsd-elf.h,
1232 config/mn10300/mn10300.c, config/pdp11/pdp11.h,
1233 config/rs6000/cell.md, config/rs6000/darwin.h,
1234 config/sh/sh.md, config/sh/sh4-300.md, config/spu/spu.c,
1235 config/spu/spu.md, cselib.c, expr.c, haifa-sched.c, hwint.h,
1236 jump.c, reload.c, sched-deps.c, sched-int.h, tree-inline.c,
1237 tree-profile.c, tree-ssa-live.h, tree-vrp.c: Fix comment
1238 typos. Follow spelling conventions.
1239 * doc/invoke.texi: Follow spelling conventions.
1241 2007-02-03 Roger Sayle <roger@eyesopen.com>
1243 * simplify-rtx.c (simplify_relational_operation_1): Implement some
1244 canonicalization transformations that attempt to simplify integer
1245 constant comparisons to become comparisons against zero.
1247 2007-02-03 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1249 * builtins.c (fold_builtin_cabs): Fold cabs(x+xi) into
1251 * fold-const.c (fold_binary): Fix comment typos. Fold complex
1252 (x,0)-(0,y) into (x,-y). Likewise (0,y)-(x,0) into (-x,y).
1254 2007-02-02 Mike Stump <mrs@apple.com>
1256 * config/darwin9.h (DARWIN_LINKER_GENERATES_ISLANDS): Add.
1257 * config/rs6000/rs6000.c (DARWIN_GENERATE_ISLANDS): Add.
1258 (output_call): Use DARWIN_GENERATE_ISLANDS to decide when to
1259 generate a branch island.
1261 2007-02-02 Bob Wilson <bob.wilson@acm.org>
1263 * config/xtensa/xtensa.c (smalloffset_mem_p): Use BASE_REG_P.
1264 (xtensa_legitimate_address_p): New.
1265 (xtensa_legitimize_address): New.
1266 (xtensa_output_addr_const_extra): New.
1267 * config/xtensa/xtensa.h (REG_OK_STRICT_FLAG): Define.
1269 (REG_OK_FOR_BASE_P): Use BASE_REG_P.
1270 (GO_IF_LEGITIMATE_ADDRESS): Move code to xtensa_legitimate_address_p.
1271 (LEGITIMIZE_ADDRESS): Move code to xtensa_legitimize_address.
1272 (OUTPUT_ADDR_CONST_EXTRA): Move code to xtensa_output_addr_const_extra.
1273 * config/xtensa/xtensa-protos.h (xtensa_legitimate_address_p): New.
1274 (xtensa_legitimize_address): New.
1275 (xtensa_output_addr_const_extra): New.
1277 2007-02-02 Steve Ellcey <sje@cup.hp.com>
1279 * config/ia64/ia64.c (ia64_print_operand): Fix compare strings.
1281 2007-02-02 Ian Lance Taylor <iant@google.com>
1283 * expmed.c (expand_divmod): Add comment.
1285 2007-02-02 Kazu Hirata <kazu@codesourcery.com>
1287 * emit-rtl.c (renumber_insns): Remove.
1288 * flags.h: Remove the extern for flag_renumber_insns.
1289 * rtl.h: Remove the prototype for renumber_insns.
1290 * toplev.c (flag_renumber_insns): Remove.
1292 2007-02-02 Hui-May Chang <hm.chang@apple.com>
1294 Revert for x86 darwin:
1295 2005-06-19 Uros Bizjak <uros@kss-loka.si>
1297 * config/i386/i386.c (ix86_function_arg_regno_p): Put back the
1298 code before the following patch under TARGET_MACHO.
1299 (ix86_function_value_regno_p): Likewise.
1301 2007-02-02 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1303 * fold-const.c (negate_expr_p, fold_negate_expr): Handle
1306 2007-02-02 Maxim Kuvyrkov <mkuvyrkov@ispras.ru>
1309 * config/ia64/ia64.c (ia64_speculate_insn): Restrict to memory loads to
1310 general or fp registers. Add comments.
1311 * config/ia64/ia64.md (reg_pred_prefix): Add comment.
1313 2007-02-02 Paolo Bonzini <bonzini@gnu.org>
1315 * pointer-set.c (insert_aux): Only return insertion slot.
1316 (pointer_set_insert): Adjust.
1317 (pointer_set_traverse, struct pointer_map_t, pointer_map_create,
1318 pointer_map_destroy, pointer_map_insert, pointer_map_contains,
1319 pointer_map_traverse): New.
1320 * pointer-set.h (pointer_set_traverse, struct pointer_map_t,
1321 pointer_map_create, pointer_map_destroy, pointer_map_insert,
1322 pointer_map_contains, pointer_map_traverse): Declare.
1324 2007-02-02 Jakub Jelinek <jakub@redhat.com>
1327 * builtins.c (fold_builtin_sprintf): Do not attempt to optimize
1328 sprintf (str, "%s"). Do not optimize sprintf (str, "nopercent", p++).
1330 2007-02-02 Maxim Kuvyrkov <mkuvyrkov@ispras.ru>
1332 * sched-int.h (ds_to_dk, dk_to_ds): Declare functions.
1334 (struct _dep): New type.
1335 (dep_t): New typedef.
1336 (DEP_PRO, DEP_CON, DEP_KIND): New access macros.
1337 (DEP_STATUS): New access macro. The macro with the same name was
1338 renamed to DEP_LINK_STATUS.
1339 (dep_init): Declare function
1341 (struct _dep_link): New type.
1342 (dep_link_t): New typedef.
1343 (DEP_LINK_NODE, DEP_LINK_NEXT, DEP_LINK_PREV_NEXTP): New access macros.
1344 (DEP_LINK_DEP, DEP_LINK_PRO, DEP_LINK_CON, DEP_LINK_KIND): New macros.
1345 (DEP_LINK_STATUS): New macro.
1346 (debug_dep_links): New debug function.
1348 (struct _deps_list): New type.
1349 (deps_list_t): New typedef.
1350 (DEPS_LIST_FIRST): New access macro.
1351 (FOR_EACH_DEP_LINK): New cycle macro.
1352 (create_deps_list, free_deps_list, delete_deps_list): Declare
1354 (deps_list_empty_p, debug_deps_list, add_back_dep_to_deps_list): Ditto.
1355 (find_link_by_pro_in_deps_list, find_link_by_con_in_deps_list): Ditto.
1356 (copy_deps_list_change_con): Ditto.
1358 (move_dep_link): Declare function.
1360 (struct _dep_node): New type.
1361 (dep_node_t): New typedef.
1362 (DEP_NODE_BACK, DEP_NODE_DEP, DEP_NODE_FORW): New access macros.
1364 (struct haifa_insn_data.back_deps): New field to hold backward
1365 dependencies of the insn.
1366 (struct haifa_insn_data.depend): Rename to forw_deps. Change its type
1368 (struct haifa_insn_data.resolved_deps): Rename to resolved_back_deps.
1369 Change its type to deps_list_t.
1370 (INSN_BACK_DEPS): New access macro to use instead of LOG_LINKS.
1371 (INSN_DEPEND): Rename to INSN_FORW_DEPS.
1372 (RESOLVED_DEPS): Rename to INSN_RESOLVED_BACK_DEPS.
1374 (INSN_COST): Move to haifa-sched.c. Use insn_cost () instead.
1376 (DEP_STATUS): Rename to DEP_LINK_STATUS. Fix typo in the comment.
1378 (add_forw_dep, delete_back_forw_dep, insn_cost): Update declaration and
1380 (dep_cost): Declare.
1382 * sched-deps.c (CHECK): New macro to (en/dis)able sanity checks.
1383 (ds_to_dk, dk_to_ds): New functions.
1385 (init_dep_1): New static function.
1386 (init_dep): New function.
1387 (copy_dep): New static function.
1389 (dep_link_consistent_p, attach_dep_link, add_to_deps_list): New static
1391 (detach_dep_link): New static function.
1392 (move_dep_link): New function.
1394 (dep_links_consistent_p, dump_dep_links): New static functions.
1395 (debug_dep_links): New debugging function.
1397 (deps_obstack, dl_obstack, dn_obstack): New static variables.
1399 (alloc_deps_list, init_deps_list): New static functions.
1400 (create_deps_list): New function.
1401 (clear_deps_list): New static function.
1402 (free_deps_list, delete_deps_list, deps_list_empty_p): New functions.
1403 (deps_list_consistent_p, dump_deps_list): New static functions.
1404 (debug_deps_list): New function.
1405 (add_back_dep_to_deps_list, find_link_by_pro_in_deps_list): New
1407 (find_link_by_con_in_deps_list, copy_deps_list_change_con): Ditto.
1409 (maybe_add_or_update_back_dep_1, add_or_update_back_dep_1): Update to
1410 use new scheduler dependencies lists.
1411 (add_back_dep, delete_all_dependences, fixup_sched_groups): Ditto.
1412 (sched_analyze): Ditto. Initialize dependencies lists.
1413 (add_forw_dep, compute_forward_dependences): Update to use new
1414 scheduler dependencies lists.
1416 (init_dependency_caches): Init deps_obstack.
1417 (free_dependency_caches): Free deps_obstack.
1419 (adjust_add_sorted_back_dep, adjust_back_add_forw_dep): Update to use
1420 new scheduler dependencies lists.
1421 (delete_forw_dep, add_or_update_back_forw_dep): Ditto.
1422 (add_back_forw_dep, delete_back_forw_dep): Ditto.
1424 * sched-rgn.c (set_spec_fed, find_conditional_protection, is_pfree):
1425 Update to use new scheduler dependencies lists.
1426 (is_conditionally_protected, is_prisky, add_branch_dependences): Ditto.
1427 (debug_dependencies): Ditto.
1428 (schedule_region): Update comments.
1430 * sched-ebb.c (earliest_block_with_similiar_load): Update to use new
1431 scheduler dependencies lists.
1432 (schedule_ebb): Update comments.
1434 * rtl.def (DEPS_LIST): Remove.
1436 * lists.c (unused_deps_list): Remove.
1437 (free_list): Update assertions.
1439 (alloc_DEPS_LIST, free_DEPS_LIST_list, free_DEPS_LIST_node): Remove.
1440 (remove_free_DEPS_LIST_elem, copy_DEPS_LIST_list): Ditto.
1442 * rtl.h (free_DEPS_LIST_list, alloc_DEPS_LIST): Remove declarations.
1443 (remove_free_DEPS_LIST_elem, copy_DEPS_LIST_list): Ditto.
1445 * haifa-sched.c (comments): Update.
1446 (insn_cost1): Remove. Inline the code into insn_cost ().
1447 (insn_cost): Update to use new scheduler dependencies lists. Move
1448 processing of the dependency cost to dep_cost ().
1449 (dep_cost): New function. Use it instead of insn_cost () when
1450 evaluating cost of the dependency. Use compatible interface to
1451 interact with the target.
1452 (priority): Update to use new scheduler dependencies lists.
1453 (rank_for_schedule): Ditto. Optimize heuristic that prefers the insn
1454 with greater number of insns that depend on the insn.
1455 (schedule_insn): Update to use new scheduler dependencies lists. Add
1456 code to free backward dependencies lists. Inline and optimize code
1457 from resolve_dep () - see PR28071.
1458 (ok_for_early_queue_removal): Update to use new scheduler dependencies
1459 lists. Update call to targetm.sched.is_costly_dependence hook.
1461 (fix_inter_tick, try_ready, fix_tick_ready): Update to use new
1462 scheduler dependencies lists.
1464 (resolve_dep): Remove. Move the logic to schedule_insn ().
1465 (init_h_i_d): Initialize dependencies lists.
1467 (process_insn_depend_be_in_spec): Rename to
1468 process_insn_forw_deps_be_in_spec. Update to use new scheduler
1470 (add_to_speculative_block, create_check_block_twin, fix_recovery_deps):
1471 Update to use new scheduler dependencies lists.
1472 (clear_priorities, calc_priorities, add_jump_dependencies): Ditto.
1474 * ddg.c (create_ddg_dependence, create_ddg_dep_no_link): Update to use
1475 new scheduler dependencies lists.
1476 (build_intra_loop_deps): Ditto.
1478 * target.h (struct _dep): Declare to use in
1479 gcc_target.sched.is_costly_dependence.
1480 (struct gcc_target.sched.adjust_cost): Fix typo.
1481 (struct gcc_target.sched.is_costly_dependence): Change signature to use
1482 single dep_t parameter instead of an equivalent triad.
1483 (struct gcc_target.sched.adjust_cost_2): Remove.
1485 * target-def.h (TARGET_SCHED_ADJUST_COST_2): Remove.
1487 * reg-notes.def (DEP_TRUE, DEP_OUTPUT, DEP_ANTI): Update comments.
1489 * doc/tm.texi (TARGET_SCHED_IS_COSTLY_DEPENDENCE): Update
1491 (TARGET_SCHED_ADJUST_COST_2): Remove documentation.
1493 * doc/rtl.texi (LOG_LINKS): Remove part about instruction scheduler.
1494 (REG_DEP_TRUE): Document.
1496 * config/ia64/ia64.c (ia64_adjust_cost_2): Rename to ia64_adjust_cost.
1497 Change signature to correspond to the targetm.sched.adjust_cost hook.
1498 Update use in TARGET_SCHED_ADJUST_COST_2.
1499 (TARGET_SCHED_ADJUST_COST_2): Rename to TARGET_SCHED_ADJUST_COST.
1500 (ia64_dependencies_evaluation_hook, ia64_dfa_new_cycle): Update to use
1501 new scheduler dependencies lists.
1502 (ia64_gen_check): Ditto.
1504 * config/mips/mips.c (vr4130_swap_insns_p): Update to use new scheduler
1507 * config/rs6000/rs6000.c (rs6000_is_costly_dependence): Change
1508 signature to correspond to the targetm.sched.is_costly_dependence hook.
1509 (is_costly_group): Update to use new scheduler dependencies lists.
1511 * config/spu/spu.c (spu_sched_adjust_cost): Use insn_cost () function
1512 instead of INSN_COST () macro.
1514 2007-02-01 Ian Lance Taylor <iant@google.com>
1516 * lower-subreg.c (resolve_clobber): Handle a subreg of a concatn.
1518 2007-02-01 Guy Martin <gmsoft@gentoo.org>
1520 * pa.md (tp_load): Correct mfctl instruction syntax.
1522 2007-02-01 Geoffrey Keating <geoffk@apple.com>
1524 * config/rs6000/rs6000.c (rs6000_stack_info): Correct
1525 altivec_padding_size calculation on AIX. Improve comment, add
1526 assert to verify that it's right.
1528 * config/rs6000/darwin.h (MD_UNWIND_SUPPORT): Don't define for
1531 2007-2-01 Seongbae Park <seongbae.park@gmail.com>
1534 * global.c (global_alloc): Add mising initialization of
1537 2007-02-01 Roger Sayle <roger@eyesopen.com>
1539 * alias.c (init_alias_analysis): Correct whitespace.
1540 * bb-reorder.c (fix_edges_for_rarely_executed_code,
1541 partition_hot_cold_basic_blocks): Likewise.
1542 * builtins.c (expand_builtin_printf, expand_builtin_fprintf,
1543 expand_builtin_sprintf, fold_builtin_carg, fold_builtin_sprintf,
1544 maybe_emit_sprintf_chk_warning, fold_builtin_sprintf_chk,
1545 fold_builtin_snprintf_chk, fold_builtin_printf,
1546 fold_builtin_fprintf, do_mpfr_ckconv, do_mpfr_arg1, do_mpfr_arg2,
1547 do_mpfr_arg3, do_mpfr_sincos): Likewise.
1548 * cfgcleanup.c (cleanup_cfg): Likewise.
1549 * cfgexpand.c (tree_expand_cfg): Likewise.
1550 * fold-const.c (fold_binary) <RDIV_EXPR>: Likewise.
1551 * function.c (get_next_funcdef_no): Likewise.
1552 * gengtype.c (main): Likewise.
1553 * genmodes.c (main): Likewise.
1554 * gcse.c (bypass_conditional_jumps, print_ldst_list): Likewise.
1555 * haifa-sched.c (schedule_block, extend_h_i_d): Likewise.
1556 * ifcvt.c (noce_emit_move_insn): Likewise.
1557 * modulo-sched.c (generate_prolog_epilog, sms_schedule_by_order):
1559 * stor-layout.c (get_best_mode): Likewise.
1560 * tree-ssa-loop-niter.c (get_val_for): Likewise.
1561 * tree-ssa-structalias.c (get_varinfo, get_varinfo_fc,
1562 scc_visit, do_ds_constraint, do_complex_constraint, label_visit,
1563 perform_var_substitution, solve_graph): Likewise.
1564 * tree-vrp.c (vrp_finalize): Likewise.
1566 2007-02-01 Ian Lance Taylor <iant@google.com>
1568 * lower-subreg.c (simplify_gen_subreg_concatn): If we ask for the
1569 high part of a paradoxical subreg, return a constant zero.
1571 2007-02-01 Zdenek Dvorak <dvorakz@suse.cz>
1573 * toplev.c (lang_dependent_init): Call init_set_costs.
1574 * loop-init.c (loop_optimizer_init): Do not call init_set_costs.
1576 2007-02-01 Richard Guenther <rguenther@suse.de>
1579 * fold-const.c (fold_negate_expr): Allow negating a
1580 constant if overflow does not change.
1582 2007-02-01 Ralf Wildenhues <Ralf.Wildenhues@gmx.de>
1584 * doc/c-tree.texi (Expression trees): Improve markup.
1585 * doc/tm.texi (Register Classes, Addressing Modes)
1586 (Floating Point): Fix spacing after abbreviations. Fix some
1589 2007-02-01 Ben Elliston <bje@au.ibm.com>
1591 * doc/invoke.texi: Replace "bugfix" with "bug fix" throughout.
1592 * doc/contrib.texi: Likewise.
1593 * doc/install.texi: Likewise.
1595 2007-01-31 Richard Henderson <rth@redhat.com>
1596 Ian Lance Taylor <iant@google.com>
1598 * lower-subreg.c: New file.
1599 * rtl.def (CONCATN): Define.
1600 * passes.c (init_optimization_passes): Add pass_lower_subreg and
1602 * emit-rtl.c (update_reg_offset): New static function, broken out
1603 of gen_rtx_REG_offset.
1604 (gen_rtx_REG_offset): Call update_reg_offset.
1605 (gen_reg_rtx_offset): New function.
1606 * regclass.c: Revert patch of 2006-03-05, restoring
1608 (clear_reg_info_regno): New function.
1609 * dwarf2out.c (concatn_loc_descriptor): New static function.
1610 (loc_descriptor): Handle CONCATN.
1611 * common.opt (fsplit_wide_types): New option.
1612 * opts.c (decode_options): Set flag_split_wide_types when
1614 * timevar.def (TV_LOWER_SUBREG): Define.
1615 * rtl.h (gen_reg_rtx_offset): Declare.
1616 (reg_scan_update): Declare.
1617 * regs.h (clear_reg_info_regno): Declare.
1618 * tree-pass.h (pass_lower_subreg): Declare.
1619 (pass_lower_subreg2): Declare.
1620 * doc/invoke.texi (Option Summary): List -fno-split-wide-types.
1621 (Optimize Options): Add -fsplit-wide-types to -O1 list. Document
1623 * doc/rtl.texi (Regs and Memory): Document concat and concatn.
1624 * Makefile.in (OBJS-common): Add lower-subreg.o.
1625 (lower-subreg.o): New target.
1627 2007-01-31 Kazu Hirata <kazu@codesourcery.com>
1629 * config/sh/sh.h (HAVE_SECONDARY_RELOADS): Remove.
1631 2007-01-31 Anatoly Sokolov <aesok@post.ru>
1634 * config/avr/avr.c (DWARF2_ADDR_SIZE): Define.
1636 2007-01-31 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1639 * builtins.c (fold_builtin_sqrt): Use MPFR for constant args.
1641 2007-01-31 Zdenek Dvorak <dvorakz@suse.cz>
1643 * cfgloop.h: Include vec-prim.h.
1644 (enum li_flags): Remove LI_ONLY_OLD.
1645 (loop_iterator): Changed.
1646 (fel_next, fel_init): Iterate over loop tree.
1647 (FOR_EACH_LOOP_BREAK): New macro.
1648 * loop-unswitch.c (unswitch_loops): Do not pass LI_ONLY_OLD to
1650 * tree-ssa-loop-unswitch.c (tree_ssa_unswitch_loops): Ditto.
1651 * modulo-sched.c (sms_schedule): Ditto.
1652 * tree-vectorizer.c (vectorize_loops): Ditto.
1653 * doc/loop.texi: Update information on loop numbering and behavior of
1654 FOR_EACH_LOOP wrto new loops.
1655 * tree-scalar-evolution.c (compute_overall_effect_of_inner_loop,
1656 add_to_evolution_1): Test nestedness of loops instead of comparing
1658 * tree-chrec.c (chrec_fold_plus_poly_poly,
1659 chrec_fold_multiply_poly_poly, chrec_evaluate,
1660 hide_evolution_in_other_loops_than_loop, chrec_component_in_loop_num,
1661 reset_evolution_in_loop): Ditto.
1662 * Makefile.in (CFGLOOP_H): Add vecprim.h dependency.
1664 2007-01-31 Dirk Mueller <dmueller@suse.de>
1666 * c-common.c (warn_about_parentheses): Separate warning about
1667 un-parenthized sequence of comparison operators from the one
1668 which is supposed to warn about x <= y <= z.
1670 2007-01-31 Uros Bizjak <ubizjak@gmail.com>
1672 * optabs.h (enum optab_index): Add new OTI_isinf.
1673 (isinf_optab): Define corresponding macro.
1674 * optabs.c (init_optabs): Initialize isinf_optab.
1675 * genopinit.c (optabs): Implement isinf_optab using isinf?f2
1677 * builtins.c (mathfn_built_in): Handle BUILT_IN_ISINF{,F,L}.
1678 (expand_builtin_interclass_mathfn): Expand BUILT_IN_ISINF{,F,L}
1680 (expand_builtin): Expand BUILT_IN_ISINF{,F,L} using
1681 expand_builtin_interclass_mathfn.
1682 * reg_stack.c (subst_stack_regs_pat): Handle UNSPEC_FXAM.
1683 * config/i386/i386.md (UNSPEC_FXAM): New constant.
1684 (fxam<mode>2_i387): New insn pattern.
1685 (isinf<mode>2) New expander to implement isinf, isinff and isinfl
1686 built-in functions as x87 inline asm.
1688 2007-01-31 Kazu Hirata <kazu@codesourcery.com>
1690 * gcc/config/arm/unwind-arm.h (_sleb128_t, _uleb128_t): New.
1692 2007-01-30 Eric Christopher <echristo@apple.com>
1694 * config.gcc: Add geode.
1696 2007-01-31 Kazu Hirata <kazu@codesourcery.com>
1698 * cgraphunit.c, config/arm/arm.c, config/m68k/m68k.c,
1699 ipa-inline.c, tree-profile.c, tree-ssa-live.c,
1700 tree-ssa-math-opts.c, tree-ssanames.c, tree-vect-analyze.c,
1701 value-prof.c: Fix comment typos.
1703 2007-01-30 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
1706 * doc/invoke.texi (Wpointer-arith): Document warning.
1708 2007-01-30 Janis Johnson <janis187@us.ibm.com>
1710 * doc/extend.texi (Decimal Floating Types): Remove decfloat.h from
1711 the list of discrepancies from the draft TR.
1713 2007-01-30 Dirk Mueller <dmueller@suse.de>
1716 * doc/invoke.texi (-Wreturn-type): Update description to
1719 2007-01-30 Richard Sandiford <richard@codesourcery.com>
1721 * cfgrtl.c (try_redirect_by_replacing_jump): Check only_sets_cc0_p.
1723 2007-01-30 Uros Bizjak <ubizjak@gmail.com>
1725 * builtins.c (expand_builtin_int_interclass_roundingfn): New function
1726 to handle optabs that operate on floating point input argument and
1727 output to integer output.
1728 (expand_builtin_mathfn) [BUILT_IN_ILOGB]: Move from here ...
1729 (expand_builtin_interclass_mathfn) [BUILT_IN_ILOGB]: ... to here.
1730 (expand_builtin): Expand BUILT_IN_ILOGB{,F,L} using
1731 expand_builtin_interclass_mathfn ().
1732 * config/i386/i386.md (fxtractxf3_i387): Rename from *fxtractxf3_i387.
1734 (ilogbxf2, ilogb<mode>2): New expanders to implement ilogb, ilogbf and
1735 ilogbl built-in functions as x87 intrinsics.
1737 2007-01-30 Richard Guenther <rguenther@suse.de>
1740 * dwarf2out.c (reference_to_unused): Query varpool if the
1741 variable was output.
1743 2007-01-30 Richard Guenther <rguenther@suse.de>
1746 * passes.c (execute_one_pass): Reset in_gimple_form to not
1747 confuse non-unit-at-a-time mode.
1749 2007-01-29 Roger Sayle <roger@eyesopen.com>
1750 Richard Guenther <rguenther@suse.de>
1752 * fold-const.c (round_up): Make HIGH an unsigned HOST_WIDE_INT to
1753 avoid undefined behaviour on overflow. Use force_fit_type_double
1754 to construct the constant with the specified TREE_OVERFLOW.
1756 2007-01-29 Janis Johnson <janis187@us.ibm.com>
1758 * config/dfp-bit.c: Add parameterized support for fp exceptions.
1759 * config/dfp-bit.h: Ditto.
1761 2007-01-29 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
1763 * c-decl.c (pop_scope): Replace warnings with call to
1764 warn_for_unused_label.
1765 * c-common.h (warn_for_unused_label): Declare.
1766 * c-common.c (warn_for_unused_label): Define.
1768 2007-01-29 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
1770 * tree-optimize.c (update_inlined_to_pointers): Delete unused
1773 2007-01-29 Janis Johnson <janis187@us.ibm.com>
1775 * Makefile.in (USER_H): Remove decfloat.h.
1776 * ginclude/decfloat.h: Delete, moving contents to ...
1777 * ginclude/float.h: Add support for decimal floating point,
1778 guarded by __STDC_WANT_DEC_FP__.
1780 2007-01-29 Mike Stump <mrs@apple.com>
1782 * doc/gccint.texi (Top): Rename Loop Representation to Loop
1783 Analysis and Representation to resolve case insensitive conflict.
1784 * doc/loop.texi (Loop Analysis and Representation): Likewise.
1786 2007-01-28 Daniel Berlin <dberlin@dberlin.org>
1788 * tree.h (struct tree_memory_tag): Add aliases member.
1789 (MTAG_ALIASES): New macro.
1790 * tree-ssa-alias.c (alias_bitmap_obstack): New variable.
1791 (add_may_alias): Remove pointer-set. Update for may_aliases being
1793 (mark_aliases_call_clobbered): Update for may_aliases being a
1795 (compute_tag_properties): Ditto.
1796 (create_partition_for): Ditto.
1797 (compute_memory_partitions): Ditto.
1798 (dump_may_aliases_for): Ditto.
1799 (is_aliased_with): Ditto.
1800 (add_may_alias_for_new_tag): Ditto.
1801 (rewrite_alias_set_for): Rewrite for may_aliases being a bitmap.
1802 (compute_is_aliased): New function.
1803 (compute_may_aliases): Call compute_is_aliased).
1804 (init_alias_info): Initialize alias_bitmap_obstack.
1805 (union_alias_set_into): New function.
1806 (compute_flow_sensitive_aliasing): Use union_aliases_into.
1807 (have_common_aliases_p): Rewrite to take two bitmaps and use
1809 (compute_flow_insensitive_aliasing): Stop using pointer-sets.
1811 (finalize_ref_all_pointers): Update for add_may_alias changes.
1812 (new_type_alias): Ditto.
1813 * tree-flow-inline.h (may_aliases): Return a bitmap.
1814 * tree-dfa.c (dump_variable): Check for MTAG_P'ness.
1815 * tree-ssa.c (verify_flow_insensitive_alias_info): Update for
1816 may_aliases being a bitmap.
1817 * tree-flow.h (struct var_ann_d): Remove may_aliases member.
1818 may_aliases now returns a bitmap.
1819 * tree-ssa-structalias.c (merge_smts_into): Update for may_aliases
1821 * tree-ssa-operands.c (add_virtual_operand): Update for
1822 may_aliases being a bitmap.
1824 2007-01-29 Daniel Berlin <dberlin@dberlin.org>
1826 PR tree-optimization/30630
1827 * tree-ssa-structalias.c (do_complex_constraint): Mark correct
1828 variable as changed.
1830 2007-01-29 Simon Martin <simartin@users.sourceforge.net>
1833 * gimplify.c (gimplify_target_expr): Make sure that the TARGET_EXPR is
1834 expanded only once even if an error occurs.
1836 2007-01-29 Ben Elliston <bje@au.ibm.com>
1838 * gcov-io.h (__gcov_indirect_call_profiler): Declare.
1839 (__gcov_average_profiler): Likewise.
1840 (__gcov_ior_profiler): Likewise.
1841 (__gcov_merge_ior): Likewise.
1843 2007-01-28 Jan Hubicka <jh@suse.cz>
1845 * builtins.c (expand_builtin_memset): Fix typo in my last patch.
1846 * value-prof.c (stringop_block_profile): Likewise.
1848 2007-01-28 Jan Hubicka <jh@suse.cz>
1850 * expr.c (emit_block_move_via_movmem, emit_block_move_via_libcall): Add
1851 variant handling histograms; add wrapper.
1852 (clear_storage_via_libcall): Export.
1853 (emit_block_move_hints): Break out from ...; add histograms.
1854 (emit_block_move): ... this one.
1855 (clear_storage_hints): Break out from ...; add histograms.
1856 (clear_storage): ... this one.
1857 (set_storage_via_memset): Handle histogram.
1858 * expr.h (emit_block_move_via_libcall, emit_block_move_hints): Declare.
1859 (clear_storage_hints, clear_storage_via_libcall): Declare.
1860 (set_storage_via_setmem): Update prototype.
1861 * doc/md.texi (movmem, setmem): Document new arguments.
1863 * value-prof.c (dump_histogram_value, tree_find_values_to_profile): Add
1865 (stringop_block_profile): New global function.
1866 (tree_stringops_values_to_profile): Profile block size and alignment.
1867 * value-prof.h (enum hist_type): add HIST_TYPE_AVERAGE and
1869 (struct profile_hooks): Add gen_average_profiler and gen_ior_profiler.
1870 (stringop_block_profile): Declare.
1871 * builtins.c: Include value-prof.h.
1872 (expand_builtin_memcpy, expand_builtin_memset): Pass block profile.
1873 * gcov-ui.h (GCOV_COUNTER_NAMES): Add new counter.
1874 (GCOV_COUNTER_AVERAGE, GCOV_COUNTER_IOR): New constants.
1875 (GCOV_COUNTERS, GCOV_LAST_VALUE_COUNTER): Update.
1876 * profile.c (instrument_values): Add new counters.
1877 * cfgexpand.c (expand_gimple_basic_block): Propagate histograms to
1879 * tree-profile.c (tree_average_profiler_fn, tree_ior_profiler_fn): New.
1880 (tree_init_edge_profiler): Build new profilers.
1881 (tree_gen_average_profiler, tree_gen_ior_profiler): New.
1882 (pass_tree_profile): Add dump.
1883 (tree_profile_hooks): Update.
1884 * Makefile.in (LIBGCOV): Add new constants.
1885 * libgcov.c (__gcov_merge_ior, __gcov_average_profiler,
1886 __gcov_ior_profiler): New.
1887 * i386.md (movmem/setmem expanders): Add new optional arguments.
1889 2007-01-28 David Edelsohn <edelsohn@gnu.org>
1891 * doc/md.texi (Standard Pattern Names): Document blockage pattern.
1893 2007-01-28 Zdenek Dvorak <dvorakz@suse.cz>
1895 * tree-ssa-loop-unswitch.c: Include tree-inline.h.
1896 (tree_unswitch_single_loop): Pass eni_size_weights to
1897 tree_num_loop_insns.
1898 * tree-ssa-loop-manip.c: Include tree-inline.h.
1899 (can_unroll_loop_p): Pass eni_size_weights to
1900 tree_num_loop_insns.
1901 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p):
1902 Pass eni_size_weights to estimate_num_insns.
1903 * tree.h (init_inline_once): Export.
1904 * toplev.c (backend_init): Call init_inline_once.
1905 * cgraphunit.c (cgraph_process_new_functions,
1906 cgraph_analyze_function): Pass eni_inlining_weights to
1908 * ipa-inline.c (compute_inline_parameters): Ditto.
1909 * tree-ssa-loop-ivcanon.c (tree_num_loop_insns): Pass weights
1910 to estimate_num_insns.
1911 (try_unroll_loop_completely): Pass eni_size_weights to
1912 tree_num_loop_insns.
1913 * tree-eh.c (decide_copy_try_finally): Pass eni_size_weights
1914 ot estimate_num_insns.
1915 * tree-ssa-loop-prefetch.c: Include tree-inline.h.
1916 (loop_prefetch_arrays): Pass eni_time_weights to tree_num_loop_insns.
1917 * tree-inline.c (eni_inlining_weights, eni_size_weights,
1918 eni_time_weights): New variables.
1919 (init_inline_once): Initialize them.
1920 (struct eni_data): Mew.
1921 (estimate_num_insns_1, estimate_num_insns): Use weights.
1922 * tree-inline.h (struct eni_weights_d): New.
1923 (eni_inlining_weights, eni_size_weights, eni_time_weights): Declare.
1924 (estimate_num_insns): Declaration changed.
1925 * cfgloop.h (tree_num_loop_insns): Declaration changed.
1926 * Makefile.in (tree-ssa-loop-unswitch.o, tree-ssa-loop-prefetch.o,
1927 tree-ssa-loop-manip.o): Add TREE_INLINE_H dependency.
1929 2007-01-28 Zdenek Dvorak <dvorakz@suse.cz>
1931 * tree-data-ref.c (conflict_fn): Assert that the number of affine
1932 relations in the conflict function is valid.
1934 2007-01-27 Ian Lance Taylor <iant@google.com>
1936 * common.opt: Add fstrict-overflow.
1937 * opts.c (decode_options): Set flag_strict_overflow if -O2.
1938 * flags.h (TYPE_OVERFLOW_WRAPS): Define.
1939 (TYPE_OVERFLOW_UNDEFINED): Define.
1940 (TYPE_OVERFLOW_TRAPS): Define. This replaces TYPE_TRAP_SIGNED.
1942 * tree.h (TYPE_TRAP_SIGNED): Don't define.
1943 * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_UNDEFINED.
1944 (fold_negate_expr): Likewise.
1945 (make_range): Likewise.
1946 (extract_muldiv_1): Likewise.
1947 (maybe_canonicalize_comparison): Likewise.
1948 (fold_comparison): Likewise.
1949 (fold_binary): Likewise.
1950 (tree_expr_nonnegative_p): Likewise.
1951 (tree_expr_nonzero_p): Likewise.
1952 * tree-vrp.c (compare_values): Likewise.
1953 (extract_range_from_binary_expr): Likewise.
1954 (extract_range_from_unary_expr): Likewise.
1955 * tree-ssa-loop-niter.c (infer_loop_bounds_from_signedness):
1957 (nowrap_type_p): Likewise.
1958 * tree-scalar-evolution.c (simple_iv): Likewise.
1959 * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_WRAPS.
1960 (build_range_check): Likewise.
1961 (extract_muldiv_1): Likewise.
1962 (fold_comparison): Likewise.
1963 * tree-vrp.c (vrp_int_const_binop): Likewise.
1964 (extract_range_from_unary_expr): Likewise.
1965 * convert.c (convert_to_integer): Likewise.
1966 * fold-const.c (fold_negate_expr): Use TYPE_OVERFLOW_TRAPS.
1967 (fold_comparison): Likewise.
1968 (fold_binary): Likewise.
1969 * optabs.c (optab_for_tree_code): Likewise.
1970 * tree-vectorizer.c (vect_is_simple_reduction): Likewise.
1971 * simplify-rtx.c (simplify_const_relational_operation): Check
1972 flag_strict_overflow and flag_trapv.
1973 (simplify_const_relational_operation): Likewise.
1974 * doc/invoke.texi (Option Summary): Mention -fstrict-overflow.
1975 (Optimize Options): Add -fstrict-overflow to -O2 list. Document
1978 2007-01-27 Roger Sayle <roger@eyesopen.com>
1980 * tree.c (tree_fold_gcd): Delete.
1981 * tree.h (tree_fold_gcd): Remove prototype.
1982 * tree-data-ref.c (tree_fold_divides_p): Don't use tree_fold_gcd to
1983 test whether one constant integer is a multiple of another. Instead
1984 call int_const_binop with TRUNC_MOD_EXPR and test for a zero result.
1985 * fold-const.c (multiple_of_p): We've determined both TOP and
1986 BOTTOM are integer constants so we can call int_const_binop directly
1987 instead of the more generic const_binop.
1989 2007-01-27 Roger Sayle <roger@eyesopen.com>
1991 * fold-const.c (size_binop): In the fast-paths for X+0, 0+X, X-0 and
1992 1*X check that the constant hasn't overflowed, to preserve the
1994 (round_up): Provide an efficient implementation when rouding-up an
1995 INTEGER_CST to a power-of-two.
1997 2007-01-28 Ralf Wildenhues <Ralf.Wildenhues@gmx.de>
1999 * doc/sourcebuild.texi: Add comma for clarity.
2000 * doc/extend.texi: Fix some typos.
2001 * doc/passes.texi: Likewise.
2002 * doc/cppinternals.texi: Likewise.
2003 * doc/c-tree.texi: Likewise.
2004 * doc/tree-ssa.texi: Likewise.
2005 * doc/install.texi: Likewise.
2007 2007-01-27 Jan Hubicka <jh@suse.cz>
2009 * tree-sra.c (sra_walk_function): Don't rely on aliases being build.
2010 (pass_sra): Do not require alias information.
2011 * passes.c (init_optimization_passes): Add SRA
2013 2007-01-27 Steven Bosscher <steven@gcc.gnu.org>
2015 * tracer.c (rest_of_handle_tracer): We already cleaned
2016 up the CFG in tracer() so don't do it here again.
2017 * cfgcleanup.c (rest_of_handle_jump2): Don't repeat
2018 cleanup_cfg here, either. And don't call renumber_insns.
2020 * cfgrtl.c (rtl_verify_flow_info_1): Don't verify that BB_END
2021 and BB_HEAD are in the insn stream here. Instead make sure
2022 that BB_INSN is valid on all insns. Also, do check here that
2023 there are no pending branch predictions...
2024 (rtl_verify_flow_info): ...instead of doing it here. Checks
2025 for BB_END and BB_HEAD moved from rtl_verify_flow_info_1 to
2028 2007-01-26 Roger Sayle <roger@eyesopen.com>
2030 * config/i386/i386.c (ix86_swap_binary_operands_p): New helper
2031 function to simplify/factorize operand order canonicalization.
2032 (ix86_fixup_binary_operands): Reorganize using the above function.
2033 (ix86_binary_operator_ok): Likewise.
2035 2007-01-27 Jakub Jelinek <jakub@redhat.com>
2037 * genattrtab.c (struct attr_value_list, insn_code_values): Move to
2038 file scope from optimize_attrs.
2039 (simplify_test_exp): If insn_code_values is not NULL, use it to speed
2041 (optimize_attrs): Clear insn_code_values after freeing it.
2043 2007-01-26 Zdenek Dvorak <dvorakz@suse.cz>
2045 * tree-ssa-address.c (create_mem_ref): Remove ", bsi" from
2046 a parts.base assignment.
2048 2007-01-26 Zdenek Dvorak <dvorakz@suse.cz>
2050 * tree-data-ref.c (dump_subscript): Use dump_conflict_function.
2051 (compute_subscript_distance, initialize_data_dependence_relation,
2052 finalize_ddr_dependent, analyze_ziv_subscript,
2053 analyze_siv_subscript_cst_affine,
2054 compute_overlap_steps_for_affine_univar,
2055 compute_overlap_steps_for_affine_1_2, analyze_subscript_affine_affine,
2056 analyze_siv_subscript, analyze_miv_subscript,
2057 analyze_overlapping_iterations, subscript_dependence_tester_1,
2058 compute_self_dependence, free_dependence_relation): Work
2059 with affine_fn instead of chrecs.
2060 (dump_affine_function, dump_conflict_function, affine_function_equal_p,
2061 common_affine_function, affine_function_base,
2062 affine_function_constant_p, affine_fn_op, affine_fn_plus,
2063 affine_fn_minus, affine_fn_free, conflict_fn_not_known,
2064 conflict_fn_no_dependence, free_conflict_function, free_subscripts,
2065 conflict_fn, affine_fn_cst, affine_fn_univar): New functions.
2066 (all_chrecs_equal_p): Removed.
2067 * tree-data-ref.h (affine_fn, conflict_function): New types.
2068 (struct subscript): Change type of conflicting_iterations_in_a
2069 and conflicting_iterations_in_b.
2071 2007-01-26 Steve Ellcey <sje@cup.hp.com>
2074 * config/pa/pa.h (TARGET_HPUX_11): New.
2075 * config/pa/pa-hpux11.h (TARGET_HPUX_11): New.
2076 * config/pa/pa.c (pa_init_builtins): Use TARGET_HPUX_11.
2078 2007-01-26 Daniel Berlin <dberlin@dberlin.org>
2079 Richard Guenther <rguenther@suse.de>
2081 * tree-ssa-structalias.c (solve_graph): Handle case
2082 we merged the variable to another.
2084 2007-01-25 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
2086 * builtins.c (fold_builtin_1): Treat ccos and ccosh as 'even'
2089 * fold-const.c (negate_mathfn_p): Treat casin, casinh, catan,
2090 catanh, cproj, csin, csinh, ctan and ctanh as 'odd' functions.
2092 2007-01-25 DJ Delorie <dj@redhat.com>
2094 * config/m32c/m32c.c (m32c_cannot_change_mode_class): We don't
2095 allow changes to modes which don't fit in those registers.
2097 * reload1.c (choose_reload_regs): Check for invalid subregs before
2098 computing their locations, not after.
2100 2007-01-25 Geoffrey Keating <geoffk@apple.com>
2103 * config/rs6000/rs6000.c (first_altivec_reg_to_save): On Darwin,
2104 save Altivec registers in an eh_return function.
2105 (compute_vrsave_mask): Likewise.
2106 (rs6000_stack_info): Correct AIX/Darwin stack alignment computation
2107 for saving Altivec registers.
2108 (rs6000_emit_prologue): Don't allocate stack twice in
2109 eh_return function. Correct expected value of altivec_save_offset
2110 when using save_world. Describe save of R0 to stack when using
2111 save_world. Describe stack pointer adjustment when using
2112 save_world. Remove duplicated eh_return parameter register saving.
2113 Update sp_offset variable after save_world.
2114 * config/rs6000/t-darwin (LIB2FUNCS_STATIC_EXTRA): Remove
2116 (LIB2FUNCS_EXTRA): Add darwin-world.asm.
2117 * config/rs6000/darwin.h (SUBTARGET_OVERRIDE_OPTIONS): -m64
2120 2007-01-25 Steve Ellcey <sje@cup.hp.com>
2122 * config.gcc (ia64*-*-hpux*): Make posix threads the default.
2124 2007-01-25 Steve Ellcey <sje@cup.hp.com>
2127 * config/pa/pa.c (pa_init_builtins): Set asm names for finite routines.
2128 * config/ia64/ia64.c (ia64_init_builtins): Ditto.
2130 2007-01-25 Richard Guenther <rguenther@suse.de>
2132 * doc/invoke.texi (-Wcoverage-mismatch): Document.
2133 * common.opt (-Wcoverage-mismatch): New warning option.
2134 * coverage.c (get_coverage_counts): Ignore coverage mismatch
2135 if -Wcoverage-mismatch is given.
2137 2007-01-25 Razya Ladelsky <razya@il.ibm.com>
2139 * ipa-cp.c (ipcp_insert_stage, ipcp_driver): Support for SSA.
2140 (ipcp_driver): Change to static definition.
2141 Add dumping of the ifunctions.
2142 (constant_val_insert): Remove unused parameter. Support for SSA.
2143 (ipcp_propagate_const): Support for SSA.
2144 (ipcp_profile_bb_print): Print only analyzed nodes.
2145 (ipcp_replace_map_create): Remove support for Fortran constant
2147 * ipa-prop.c (ipa_method_modify_stmt,
2148 ipa_callsite_compute_param): Support for SSA.
2149 * ipa-prop.h (ipcp_driver): Remove declaration.
2150 (IS_VALID_TREE_MAP_INDEX): Add define.
2152 2007-01-24 Geoffrey Keating <geoffk@apple.com>
2154 * unwind-dw2.c (execute_stack_op): Handle DW_OP_swap.
2156 2007-01-24 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
2159 * builtins.c (fold_builtin_cabs): Use MPFR to evaluate a
2160 constant argument to cabs and do it without checking for
2161 -funsafe-math-optimizations.
2163 2007-01-24 Douglas Gregor <dgregor@osl.iu.edu>
2165 * c-common.h (RID_FIRST_CXX0X): New.
2166 (RID_LAST_CXX0X): New.
2167 * c-opts.c (c_common_handle_option): -Wc++0x-compat is triggered
2169 * c.opt (Wc++0x-compat): New.
2170 * doc/invoke.texi (-Wc++0x-compat): Document.
2172 2007-01-24 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
2174 * builtins.c (fold_builtin_carg): New.
2175 (fold_builtin_1): Use it.
2177 2007-01-24 Jan Hubicka <jh@suse.cz>
2179 * ipa-inline.c (cgraph_decide_inlining): Initialize initial_insns.
2181 * ipa-inline.c (initial_insns, max_insns): Delete.
2182 (compute_max_insns): New function.
2183 (cgraph_decide_inlining_of_small_function): Use it; take minimal amount
2184 of insns as base for code growth.
2185 (cgraph_decide_inlining): Make initial_insns local; do not compute
2187 * params.def (PARAM_INLINE_UNIT_GROWTH): Set to 60.
2188 * doc/invoke.texi (inline-unit-growth): Update docs.
2190 2007-01-24 Jakub Jelinek <jakub@redhat.com>
2192 * config/i386/i386.h (x86_cmpxchg16b): Remove const.
2193 (TARGET_CMPXCHG16B): Define to x86_cmpxchg16b.
2194 * config/i386/i386.c (x86_cmpxchg16b): Remove const.
2195 (override_options): Add PTA_CX16 flag. Set x86_cmpxchg16b
2196 for CPUs that have PTA_CX16 set.
2199 * gimplify.c (omp_check_private): New function.
2200 (gimplify_scan_omp_clauses): Use it for
2201 firstprivate/lastprivate/reduction.
2204 * gimplify.c (omp_add_variable): Don't call omp_notice_variable
2205 on TYPE_SIZE_UNIT for GOVD_LOCAL VLAs.
2208 * omp-low.c (lower_omp_for_lastprivate): Add dlist argument.
2209 If lower_lastprivate_clauses emits some statements, append them
2210 to dlist rather than body_p and to body_p append an initializer.
2211 (lower_omp_for): Adjust caller.
2213 2007-01-24 Steve Ellcey <sje@cup.hp.com>
2215 * target.h (globalize_decl_name): New.
2216 * target-def.h (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
2217 * output.h (default_globalize_decl_name): New.
2218 * varasm.c (asm_output_bss): Use globalize_decl_name instead of
2220 (globalize_decl): Ditto.
2221 (default_globalize_decl_name): New.
2222 * config/ia64/ia64.c (ia64_globalize_decl_name): New.
2223 (ia64_handle_version_id_attribute): New.
2224 (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
2225 (ia64_asm_output_external): Use globalize_decl_name instead
2227 * doc/extend.texi (version_id): New pragma.
2228 * doc/tm.texi (ARGET_ASM_GLOBALIZE_DECL_NAME): New target hook.
2230 2007-01-24 Andreas Krebbel <krebbel1@de.ibm.com>
2232 * unwind-dw2-fde.c (get_cie_encoding): Replaced _Unwind_Word with
2233 _uleb128_t and _Unwind_SWord with _sleb128_t.
2234 * unwind-dw2.c (extract_cie_info, execute_stack_op, execute_cfa_program,
2235 uw_frame_state_for, uw_update_context_1): Likewise.
2236 * unwind-c.c (parse_lsda_header, PERSONALITY_FUNCTION): Likewise.
2237 * unwind-pe.h (read_uleb128, read_sleb128,
2238 read_encoded_value_with_base): Likewise.
2239 * unwind-generic.h: Define _sleb128_t and _uleb128_t types.
2241 2007-01-24 Richard Guenther <rguenther@suse.de>
2243 * builtins.c (expand_builtin_cexpi): Get the fndecl
2244 for cexp in the correct way.
2246 2007-01-24 Jan Hubicka <jh@suse.cz>
2248 * tree-ssa-dce.c (eliminate_unnecesary_stmts): Remove dead LHS of calls.
2250 2007-01-24 Andreas Krebbel <krebbel1@de.ibm.com>
2252 * c-cppbuiltin.c (builtin_define_type_sizeof): New function.
2253 (c_cpp_builtins): New builtin macros: __SIZEOF_INT__, __SIZEOF_LONG__,
2254 __SIZEOF_LONG_LONG__, __SIZEOF_SHORT__, __SIZEOF_POINTER__,
2255 __SIZEOF_FLOAT__, __SIZEOF_DOUBLE__, __SIZEOF_LONG_DOUBLE__,
2256 __SIZEOF_SIZE_T__, __SIZEOF_WCHAR_T__, __SIZEOF_WINT_T__ and
2257 __SIZEOF_PTRDIFF_T__.
2258 * doc/cpp.texi: Documentation for the new builtin macros added.
2260 2007-01-24 Uros Bizjak <ubizjak@gmail.com>
2262 * config/i386/i386.md (tanxf2, tan<mode>2, atan<mode>2, log<mode>2,
2263 log10<mode>2, log2<mode>2, expxf2, exp10xf2, exp2xf2): Use op2
2264 instead of operands[2] to avoid access past the end of array.
2266 2007-01-24 Richard Sandiford <richard@codesourcery.com>
2268 * reload1.c (emit_reload_insns): Pass the reload register
2269 for a non-spill output reload through forget_old_reloads_1.
2271 2007-01-23 Joseph Myers <joseph@codesourcery.com>
2273 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Define
2274 _SOFT_DOUBLE if doubles use software floating-point.
2275 * config/rs6000/libgcc-ppc-glibc.ver: Export additional long
2276 double functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
2277 * config/rs6000/darwin-ldouble.c: Also compile functions for
2278 hard-float without FPRs. Use fmsub function for all __NO_FPRS__
2279 cases. Compile extra functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
2280 * config/rs6000/linuxspe.h (SUBSUBTARGET_OVERRIDE_OPTIONS): Remove
2281 commented-out long double override.
2282 (CPP_LONGDOUBLE_DEFAULT_SPEC): Likewise.
2283 * config/rs6000/eabispe.h: Likewise.
2284 * config/rs6000/rs6000.c (rs6000_override_options): Don't override
2285 long double for non-SPE.
2286 (rs6000_handle_option): Likewise.
2287 (invalid_e500_subreg): Disallow more subregs involding DImode,
2288 DFmode, TImode or TFmode.
2289 (rs6000_legitimate_offset_address_p): Check TFmode offsets for
2291 (legitimate_lo_sum_address_p): Also check for TFmode for E500
2293 (rs6000_legitimize_address): Also handle TFmode for E500 double.
2294 (rs6000_legitimize_reload_address): Also handle TFmode for E500
2296 (rs6000_legitimate_address): Also check for TFmode for E500
2298 (rs6000_emit_move): Use DFmode subregs of TFmode for E500 double.
2299 (spe_build_register_parallel): Handle TFmode and TCmode.
2300 (rs6000_spe_function_arg): Handle TFmode and TCmode for E500
2302 (function_arg): Handle TFmode and TCmode for E500 double.
2303 (rs6000_init_libfuncs): Initialize extra libfuncs for soft double
2305 (print_operand): Handle TFmode and TImode for %y.
2306 (rs6000_generate_compare): Handle TFmode comparisons for E500
2308 (spe_func_has_64bit_regs_p): Check for TFmode for E500 double.
2309 (rs6000_function_value): Handle TFmode and TCmode for E500 double.
2310 (rs6000_libcall_value): Handle TFmode and TCmode for E500 double.
2311 * config/rs6000/rs6000.h (CANNOT_CHANGE_MODE_CLASS): Check for
2312 TFmode for E500 double.
2313 * config/rs6000/rs6000.md (FP): Allow TF for E500 double.
2314 (floatsidf2): Enable for E500 double.
2315 (movtf_softfloat): Use rs6000_nonimmediate_operand.
2316 (extenddftf2): Change to extenddftf2_fprs.
2317 (extenddftf2): Call gen_spe_extenddftf2 or gen_extenddftf2_fprs
2318 depending on TARGET_E500_DOUBLE.
2319 (extendsftf2): Enable for E500 double.
2320 (trunctfdf2): Enable for E500 double.
2321 (trunctfsf2): Change to trunctfsf2_fprs.
2322 (trunctfsf2): Call gen_spe_trunctfsf2 or gen_trunctfsf2_fprs
2323 depending on TARGET_E500_DOUBLE.
2324 (floatsitf2): Enable for E500 double.
2325 (fix_trunctfsi2): Change to fix_trunctfsi2_fprs.
2326 (fix_trunctfsi2): Call gen_spe_fix_trunctfsi2 or
2327 gen_fix_trunctfsi2_fprs depending on TARGET_E500_DOUBLE.
2328 (negtf2): Change to negtf2_internal.
2329 (negtf2): New expander.
2330 (abstf2): Enable for E500 double. Call gen_spe_abstf2_tst,
2331 gen_spe_abstf2_cmp or gen_abstf2_internal depending on
2332 TARGET_E500_DOUBLE and flag_unsafe_math_optimizations.
2333 (movdi_internal32): Use rs6000_nonimmediate_operand.
2334 (unnamed splitter): Likewise.
2335 * config/rs6000/spe.md (CMPTFEQ_GPR, TSTTFEQ_GPR, CMPTFGT_GPR,
2336 TSTTFGT_GPR, CMPTFLT_GPR, TSTTFLT_GPR): New unspecs.
2337 (SPE64TF, DITI): New mode macros.
2338 (frob_df_di): Change to frob_<SPE64:mode>_<DITI:mode>; allow more
2341 (frob_<mode>_di_2): New.
2342 (frob_tf_di_8_2): New.
2343 (frob_di_df): Change to frob_di_<mode>; allow more modes.
2345 (frob_di_df_2): Change to frob_<DITI:mode>_<SPE64:mode>_2; allow
2347 (frob_ti_<mode>_8_2): New.
2348 (frob_ti_tf_2): New.
2349 (mov_si<mode>_e500_subreg0, mov_si<mode>_e500_subreg0_2,
2350 mov_si<mode>_e500_subreg4, mov_si<mode>_e500_subreg4_2): Allow
2352 (mov_sitf_e500_subreg8, mov_sitf_e500_subreg8_2,
2353 mov_sitf_e500_subreg12, mov_sitf_e500_subreg12_2): New.
2354 (spe_trunctfdf2_internal1, spe_trunctfsf2, spe_extenddftf2,
2355 spe_fix_trunctfsi2, spe_fix_trunctfsi2_internal,
2356 spe_negtf2_internal, spe_abstf2_cmp, spe_abstf2_tst): New.
2357 (cmptfeq_gpr, tsttfeq_gpr, cmptfgt_gpr, tsttfgt_gpr, cmptflt_gpr,
2360 2007-01-23 Ian Lance Taylor <iant@google.com>
2362 * Makefile.in (OBJS-common): Reformat, alphabetize, but put
2364 (OBJS-archive): Reformat, alphabetize.
2365 (OBJS): Change out_object_file to OBJS-md.
2367 2007-01-23 Uros Bizjak <ubizjak@gmail.com>
2369 * config/i386/i386.md: Use REG_P, MEM_P, CONST_INT_P, LABEL_P,
2370 JUMP_P and CALL_P predicates where applicable.
2371 * config/i386/i386.c: Ditto.
2372 * config/i386/i386.md: Ditto.
2373 * config/i386/mmx.md: Ditto.
2374 * config/i386/predicates.md: Ditto.
2376 2007-01-22 Andreas Schwab <schwab@suse.de>
2378 * config/m68k/m68k.h: Fix comment.
2380 2007-01-22 Jan Hubicka <jh@suse.cz>
2382 * passes.c (init_optimization_passes): Do not rerun
2383 pass_early_warn_uninitialized.
2385 2007-01-22 Richard Guenther <rguenther@suse.de>
2387 PR tree-optimization/30038
2388 * tree-ssa-math-opts.c (maybe_record_sincos): New static helper
2390 (execute_cse_sincos_1): Likewise.
2391 (execute_cse_sincos): Likewise.
2392 (gate_cse_sincos): Likewise.
2393 (pass_cse_sincos): New pass CSEing sin() and cos() calls using
2394 the cexpi() canonicalization of sincos().
2395 * tree-pass.h (pass_cse_sincos): Declare.
2396 * passes.c (init_optimization_passes): New pass pas_cse_sincos.
2398 2007-01-21 Eric Botcazou <ebotcazou@libertysurf.fr>
2400 PR rtl-optimization/29329
2401 * combine.c (replaced_rhs_insn): Rename to i2mod.
2402 (replaced_rhs_value): Rename to i2mod_new_rhs.
2403 (i2mod_old_rhs): New global variable.
2404 (combine_instructions): Adjust for above change. Save a copy of
2405 the old RHS into i2mod_old_rhs when the contents of a REG_EQUAL
2406 note are substituted in the second instruction.
2407 (distribute_notes) <REG_DEAD>: Adjust for above change. Do not
2408 ditch the note if it pertains to the second eliminated register
2409 and this register is mentioned in i2mod_old_rhs.
2412 2006-09-12 Eric Botcazou <ebotcazou@libertysurf.fr>
2414 * combine.c (distribute_notes) <REG_DEAD>: Do not consider SETs past
2415 the insn to which the note was originally attached.
2417 2007-01-21 Jan Hubicka <jh@suse.cz>
2419 * ipa-inline.c (inlining_mode): Comment, move up.
2420 (cgraph_decide_inlining_incrementally): Do not perform inlining
2421 itself; fix handling of flattening of self recursive functions.
2422 (cgraph_find_cycles): Remove.
2423 (cgraph_flatten_node): Remove.
2424 (cgraph_decide_inlining): Use incremental inliner to handle flattening.
2425 (try_inline): New function.
2426 (cgraph_early_inlining): Update call of
2427 cgraph_decide_inlining_incrementally. Apply inlining here.
2428 (apply_inline): Update call of cgraph_decide_inlining_incrementally.
2430 2007-01-21 Dirk Mueller <dmueller@suse.de>
2433 * tree-vrp.c (check_array_bounds): do not warn
2434 about ADDR_EXPR's of ARRAY_REF's which are immediately
2435 used in binary expressions.
2437 2007-01-21 Ira Rosen <irar@il.ibm.com>
2439 * tree-vectorizer.h (struct _stmt_vec_info): Add new field
2440 read_write_dep and macros for its access.
2441 * tree-vectorizer.c (new_stmt_vec_info): Initialize the new field.
2442 * tree-vect-analyze.c (vect_analyze_data_ref_dependence): Remove
2443 argument, call vect_check_interleaving for every independent pair of
2444 data-refs. Mark loads that access the same memory location as a store
2446 (vect_check_dependences): Remove.
2447 (vect_analyze_data_ref_dependences): Remove vect_check_dependences
2448 call, fix the call to vect_analyze_data_ref_dependence.
2449 (vect_analyze_data_ref_access): For statements that access the same
2450 data-ref, check that they are not stores; for loads, check that there
2451 is no store that access the same location.
2453 2007-01-20 Roger Sayle <roger@eyesopen.com>
2454 Joseph Myers <joseph@codesourcery.com>
2456 * doc/invoke.texi (-fdump-rtl-combine): Fix under/overfull hbox.
2457 (-fdump-rtl-stack): Likewise.
2458 (-fno-signed-zeros): Use @minus{} for a minus sign. Correct typo.
2459 (-mcheck-zero-division, -mcpu): Fix under/overfull hbox.
2460 (-mpt-fixed): Use @minus{} for minus sign.
2461 (Using Precompiled Headers): Fix under/overfull hbox.
2463 2007-01-20 Jan Hubicka <jh@suse.cz>
2465 * tree-flow.h (struct stmt_ann_d): Move references_memory to proper
2466 place within annotation.
2468 2007-01-20 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
2470 * pa.c (output_move_double): Change array size of xoperands to 4.
2472 2007-01-20 Jan Hubicka <jh@suse.cz>
2474 * tree-tailcall.c (adjust_return_value): Do not use RESULT_DECL
2477 2007-01-19 Ian Lance Taylor <iant@google.com>
2479 * expmed.c (expand_divmod) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]: Cast
2480 constant to unsigned HOST_WIDE_INT before negating.
2482 2007-01-19 Ian Lance Taylor <iant@google.com>
2484 * tree-ssa-operands.h (struct vuse_vec_d): Change num_vuse field
2486 (VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Use unsigned comparison.
2487 (VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
2488 (SET_VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Likewise.
2489 (SET_VUSE_ELEMENT_VAR) [ENABLE_CHECKING]: Likewise.
2490 (SET_VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
2491 (realloc_vdef, realloc_vuse): Change second parameter to
2493 (ssa_operand_iterator_d): Change vuse_index and mayuse_index
2495 * tree-ssa-operands.c (realloc_vop): Change num_elem parameter to
2496 unsigned. Change x and lim locals to unsigned.
2497 (realloc_vdef, realloc_vuse): Change num_elem parameter to
2499 (finalize_ssa_vuse_ops): Change old_i local to unsigned.
2500 (copy_virtual_operands): Change i and n locals to unsigned.
2502 2007-01-19 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
2505 * builtins.c (fold_builtin_1): Handle builtin fdim.
2507 2007-01-20 Jan Hubicka <jh@suse.cz>
2509 * tree-ssa.c (init_tree_ssa): Do not call init_alias_heapvars.
2510 * tree-ssa-structalias.c (compute_points_to_sets): Do call
2511 init_alias_heapvars.
2512 (init_alias_heapvars): Initialize only when not already initialized.
2513 (delete_alias_heapvars): Set heapvar_for_stmt to NULL.
2515 2007-01-19 Roger Sayle <roger@eyesopen.com>
2517 * common.opt (fsigned-zeros): New command line option.
2518 * flags.h (HONOR_SIGNED_ZEROS): Control via flag_signed_zeros instead
2519 of flag_unsafe_math_optimizations.
2520 * opts.c (set_fast_math_flags): The -ffast-math command line option
2521 implies -fno-signed-zeros.
2522 (fast_math_flags_set_p): Likewise.
2524 * doc/invoke.texi: Document new -fno-signed-zeros option, and update
2525 the documentation of -ffast-math appropriately. Wrap long lines.
2527 2007-01-19 Steve Ellcey <sje@cup.hp.com>
2529 * system.h (ASM_MAKE_LABEL_LINKONCE): Poison.
2530 * varasm.c (globalize_decl): Remove ASM_MAKE_LABEL_LINKONCE ifdef.
2532 2007-01-19 Tomas Bily <tbily@suse.cz>
2534 * cgraphunit.c (cgraph_finalize_function): Updating of pid
2536 (tree_init_ic_make_global_vars): New function
2537 (tree_init_edge_profiler): call of tree_init_ic_make_global_vars
2538 (tree_gen_ic_profiler): New function
2539 (tree_gen_ic_func_profiler): New function
2540 (tree_profiling): Added calling of tree_gen_ic_func_profiler
2541 (tree_profile_hooks): Added hook for indirec/virtual calls
2542 * value-prof.c (tree_find_values_to_profile): New case for
2544 (tree_values_to_profile): Call for determining indirect/virtual
2546 (tree_indirect_call_to_profile): New function
2547 (tree_ic_transform): New function
2548 (tree_ic): New function
2549 (find_func_by_pid): New function
2550 (init_pid_map): New function
2551 (tree_value_profile_transformations): Added check for
2552 indirect/virtual call transformation
2553 * value-prof.h (enum hist_type): New counter type for
2554 indirect/virtual calls
2555 (profile_hooks): Added new hook for profiling indirect/virtual
2557 * profile.c (instrument_values): New case for indirect/virtual
2559 * gcov-io.h (GCOV_LAST_VALUE_COUNTER): Changed to 6
2560 (GCOV_COUNTER_V_INDIR): New counter type
2561 (GCOV_COUNTER_NAMES): New name of counter "indirect" added
2562 (GCOV_MERGE_FUNCTIONS): New merge function for indirect/virtual
2564 * cgraph.c: Definition of cgraph_max_pid
2565 (cgraph_create_node): Default init of pid attribute
2566 * cgraph.h: Declaration of cgraph_max_pid
2567 (struct cgraph_node): Added pid attribute
2568 * libgcov.c (__gcov_indirect_call_profiler): New function
2569 (__gcov_one_value_profiler_body): New function
2570 (__gcov_one_value_profiler): Body was moved to
2571 __gcov_one_value_profiler_body and calls it
2573 2007-01-19 Basile Starynkevitch <basile@starynkevitch.net>
2575 * doc/gty.texi (Options): Document the mark_hook option to GTY.
2576 * gengtype.c (write_types_data, write_func_for_structure,
2577 write_types, ggc_wtd, pch_wtd): Add skip_hooks to
2578 write_types_data, ggc_wtd, pch_wtd for processing mark_hook.
2579 (walk_type, write_func_for_structure): Generate the mark_hook if
2582 2007-01-19 Jan Hubicka <jh@suse.cz>
2584 * ipa-inline.c (cgraph_decide_inlining_incrementally): Instead of
2585 'early' argument take inlining mode argument specifying whether to
2586 inline for size/speeed or all functions; add support for flattening;
2588 (cgraph_early_inlining): Update call of decide_inlining_incrementally.
2590 2007-01-19 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
2593 * toplev.c (warn_deprecated_use): Use %qD instead of %qs to print
2594 the name of the declared identifier.
2596 2007-01-19 Dirk Mueller <dmueller@suse.de>
2598 * config/i386.h (CONDITIONAL_REGISTER_USAGE): Store
2599 result of PIC_OFFSET_TABLE_REGNUM in temporary variable to avoid
2600 duplicate evaluation.
2602 2007-01-19 Uros Bizjak <ubizjak@gmail.com>
2604 * config/i386/i386.md (acos<mode>2): Rename from acossf2 and acosdf2.
2605 Macroize expander using X87MODEF12 mode macro. Extend operand 1
2606 to XFMode, use acosxf2 and truncate result to requested mode.
2607 Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
2608 (asin<mode>2): Similarly, with asin expanders.
2609 (*fscalexf4_i387): Rename from *fscalexf4.
2610 (expNcorexf3): New expander.
2611 (expxf2, exp10xf2, exp2xf2): Use expNcorexf3 expander.
2612 (exp<mode>2): Rename from expsf2 and expdf2. Macroize expander using
2613 X87MODEF12 mode macro. Extend operand 1 to XFMode, use expxf2 and
2614 truncate result to requested mode. Use SSE_FLOAT_MODE_P to disable
2615 patterns for SSE math.
2616 (exp10<mode>2): Similarly, with exp10 expanders.
2617 (exp2<mode>2): Similarly, with exp2 expanders.
2618 (expm1<mode>2): Similarly, with expm1 expanders.
2619 (ldexp<mode>3): Similarly, with ldexp expanders.
2620 (log<mode>2, log10<mode>2, log2<mode>2, log1p<mode>2, logb<mode>2):
2621 Use gen_truncxf<mode>2_i387_noop to truncate result.
2623 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2625 * config/m68k/m68k.h (PREFERRED_STACK_BOUNDARY): Define to 32
2626 for ColdFire targets.
2628 2007-01-19 Nathan Sidwell <nathan@codesourcery.com>
2629 Richard Sandiford <richard@codesourcery.com>
2631 * config/m68k/m68k.h (M68K_STATIC_CHAIN_REG_NAME): New macro.
2632 (INITIALIZE_TRAMPOLINE): Use STATIC_CHAIN_REGNUM.
2633 (__transfer_from_trampoline): Use M68K_STATIC_CHAIN_REG_NAME.
2634 * config/m68k/m68kelf.h (STATIC_CHAIN_REGNUM): Override.
2635 (M68K_STATIC_CHAIN_REG_NAME): Likewise.
2636 * config/m68k/netbsd-elf.h (M68K_STATIC_CHAIN_REG_NAME): Likewise.
2638 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2640 * config/m68k/m68k.md (adddi_dilshr32): Rename to...
2641 (*adddi_dilshr32): ...this. Fix formatting. Remove commented-out
2642 non-canonical pattern. Restrict to !TARGET_COLDFIRE.
2643 (*adddi_dilshr32_cf): New pattern.
2644 (adddi3, subdi3): Remove first alternatives.
2646 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2648 * config/m68k/m68k.c (notice_update_cc): If an SFmode move is
2649 implemented using move.l, do not use its cc result for floating-point
2652 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2654 * config/m68k/m68k.h (EXTRA_CONSTRAINT): Stop the 'T' constraint
2655 from accepting 's' constraints if flag_pic.
2657 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2659 * config/m68k/m68k.md (bordered, bunordered, buneq, bunge, bungt)
2660 (bunle, bunlt, bltgt, bordered_rev, bunordered_rev, buneq_rev)
2661 (bunge_rev, bungt_rev, bunle_rev, bunlt_rev, bltgt_rev): Change
2662 condition from TARGET_68881 to TARGET_HARD_FLOAT.
2664 2007-01-19 Sandra Loosemore <sandra@codesourcery.com>
2666 * longlong.h (count_leading_zeros, COUNT_LEADING_ZEROS_0): Add
2667 ColdFire alternatives.
2668 * config/m68k/m68k.h (CLZ_DEFINED_VALUE_AT_ZERO): New macro.
2669 * config/m68k/m68k.md (clzsi2): Define for ColdFire
2670 architectures that support ff1 instruction.
2672 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2673 Julian Brown <julian@codesourcery.com>
2675 * config/m68k/m68k.h (CONST_OK_FOR_LETTER_P): Add an 'R' case.
2676 * config/m68k/m68k.md (*movsi_cfv4): Fold into...
2677 (*movsi_cf): ...here. Remove unnecessary 'R' from 'Rg'.
2680 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2682 * config/m68k/m68k-protos.h (valid_mov3q_const): Take a HOST_WIDE_INT
2684 (output_move_const_into_data_reg, output_move_simode_const): Delete.
2685 * config/m68k/m68k.c (const_method, const_int_cost): Take a
2686 HOST_WIDE_INT instead of an rtx.
2687 (m68k_rtx_costs): Update call accordingly.
2688 (output_move_const_into_data_reg): Likewise. Fix formatting.
2689 (valid_mov3q_const): Take a HOST_WIDE_INT instead of an rtx.
2691 (output_move_simode_const): Update calls after above changes.
2692 Rework to use automatic variables and predicates like MEM_P.
2693 * config/m68k/m68k.md (pushexthisi_const): Update call to
2696 2007-01-19 Dirk Mueller <dmueller@suse.de>
2698 * tree-ssa-alias.c (perform_var_substitution): Fix typo
2701 2007-01-19 Richard Guenther <rguenther@suse.de>
2703 * builtins.c (expand_builtin_cexpi): Fall back to expanding
2704 via cexp in case sincos is not available.
2706 2007-01-19 Richard Guenther <rguenther@suse.de>
2708 * doc/tm.texi (TARGET_HAS_SINCOS): Document new target macro.
2709 * defaults.h (TARGET_HAS_SINCOS): Default to off.
2710 * config/linux.h (TARGET_HAS_SINCOS): Set to on if we have glibc.
2711 * config/alpha/linux.h (TARGET_HAS_SINCOS): Likewise.
2712 * config/sparc/linux.h (TARGET_HAS_SINCOS): Likewise.
2713 * config/sparc/linux64.h (TARGET_HAS_SINCOS): Likewise.
2714 * config/rs6000/linux.h (TARGET_HAS_SINCOS): Likewise.
2715 * config/rs6000/linux64.h (TARGET_HAS_SINCOS): Likewise.
2717 2007-01-19 Uros Bizjak <ubizjak@gmail.com>
2719 * config/i386/i386.md (*fpatanxf3_i387, fpatan_extend<mode>xf3_i387):
2721 (atan2sf3_1, atan2df3_1, atan2xf3_1): Remove insn patterns.
2722 (atan2xf3): Directly generate RTL pattern.
2723 (atan2<mode>3): Rename from atan2sf3 and atan2df3 and macroize insn
2724 patterns using X87MODEF12 mode macro. Use fpatan_extend<mode>xf3_i387
2725 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
2726 disable patterns for SSE math.
2727 (atan<mode>2): Rename from atansf2 and atandf2 and macroize insn
2728 patterns using X87MODEF12 mode macro. Use fpatan_extend<mode>xf3_i387
2729 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
2730 disable patterns for SSE math.
2732 2007-01-19 Alexandre Oliva <aoliva@redhat.com>
2734 * libgcc-std.ver: Fix typo in %inherit for GCC_4.3.0.
2736 2007-01-18 Roger Sayle <roger@eyesopen.com>
2738 * fold-const.c (fold_unary) <VIEW_CONVERT_EXPR>: Optimize away a
2739 VIEW_CONVERT_EXPR to the same type as it's operand.
2741 2007-01-18 David Edelsohn <edelsohn@gnu.org>
2743 * config/rs6000/darwin-ldouble.c: Only build _SOFT_FLOAT if
2744 configured for long double 128.
2746 2007-01-18 Mike Stump <mrs@apple.com>
2748 * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Fix build
2751 2007-01-18 Michael Meissner <michael.meissner@amd.com>
2753 * i386.c (ix86_compute_frame_layout): Make fprintf's in #if 0 code
2756 2007-01-18 Jan Hubicka <jh@suse.cz>
2758 * tree-ssa-operands.c (vop_free_bucket_size): Never return value
2759 greater than NUM_VOP_FREE_BUCKETS.
2761 2007-01-18 Daniel Berlin <dberlin@dberlin.org>
2763 * tree-ssa-structalias.c: Update comments.
2764 (ptabitmap_obstack): Removed.
2766 (oldpta_obstack): Ditto.
2767 (stats): Add a few members.
2768 (struct variable_info): Remove node, complex, address_taken, and
2769 indirect_target members. Add oldsolution member.
2770 (new_var_info): Do not initialize removed members.
2771 (constraint_expr_type): Remove INCLUDES.
2772 (constraint_graph): Add size, implicit_preds, rep,
2773 indirect_cycles, eq_rep, label, direct_nodes, and complex members.
2774 (FIRST_REF_NODE): New macro.
2775 (LAST_REF_NODE): Ditto.
2776 (FIRST_ADDR_NODE): Ditto.
2777 (find): New function.
2779 (dump_constraint): Do not handle INCLUDES.
2780 (insert_into_complex): Do not insert duplicate constraints.
2781 (condense_varmap_nodes): Renamed and rewritten into ...
2782 (merge_node_constraints): This. Also fix bug in handling of
2783 offseted copy constraints.
2784 (clear_edges_for_node): No longer need to deal with preds at all,
2785 or removing associated preds/succs.
2786 (merge_graph_nodes): Deal with indirect_cycles.
2787 Don't deal with predecessors.
2788 (add_implicit_graph_edge): New function.
2789 (add_pred_graph_edge): Ditto.
2790 (add_graph_edge): Don't deal with predecessors.
2791 (build_constraint_graph): Removed.
2792 (build_pred_graph): New function.
2793 (build_succ_graph): Ditto.
2794 (struct scc_info): Removed in_component. Added roots, dfs, and
2795 node_mapping. Remove visited_index, unification_queue.
2796 (scc_visit): Deal with union-find we do now.
2797 Deal with cycles with REF nodes.
2798 (collapse_nodes): Renamed and rewritten to ...
2799 (unify_nodes): This.
2800 (process_unification_queue): Removed.
2801 (topo_visit): Cleanup
2802 (do_da_constraint): Use find.
2803 (do_sd_constraint): Ditto.
2804 (do_ds_constraint): Ditto.
2805 (do_complex_constraint): Ditto.
2806 (init_scc_info): Update for removed and added members.
2807 (find_and_collapse_graph_cycles): Renamed and rewritten into ...
2808 (find_indirect_cycles): This.
2809 (equivalence_class): New variable.
2810 (label_visit): New function.
2811 (perform_variable_substitution): Rewritten.
2812 (free_var_substitution_info): New function.
2813 (find_equivalent_node): Ditto.
2814 (move_complex_constraints): Ditto.
2815 (eliminate_indirect_cycles): Ditto.
2816 (solve_graph): Only propagate changed bits.
2817 Use indirect cycle elimination.
2819 (tree_id_t): Rename to tree_vi_t, delete id member, add vi member.
2820 (tree_id_eq): Renamed to ...
2821 (tree_vi_eq): This. Update for member change
2822 (insert_id_for_tree): Renamed and rewritten to ...
2823 (insert_vi_for_tree): This.
2824 (lookup_id_for_tree): Renamed and rewritten to ...
2825 (lookup_vi_for_tree): This.
2826 (get_id_for_tree): Renamed and rewritten to ...
2827 (get_vi_for_tree): Ditto.
2828 (get_constraint_exp_from_ssa_var): Update to use get_vi_for_tree.
2829 (process_constraint): Don't handle INCLUDES.
2830 Remove special ADDRESSOF case.
2831 (find_func_aliases): Rewrite to use vi functions instead of id
2833 (create_function_info_for): Ditto.
2834 (create_variable_info_for): Ditto.
2835 (intra_create_variable_infos): Ditto.
2836 (merge_smts_into): Ditto.
2837 (find_what_p_points_to): Ditto.
2838 (init_base_vars): Ditto.
2839 (init_alias_vars): Ditto.
2840 (remove_preds_and_fake_succs): New function.
2841 (dump_sa_points_to_info): Dump new stats.
2842 (dump_solution_for_var): Use find.
2843 (set_used_smts): Fix formatting.
2844 (compute_points_to_sets): Updated for new functions.
2845 (ipa_pta_execute): Ditto.
2847 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
2848 Richard Sandiford <richard@codesourcery.com>
2850 * doc/tm.texi (TARGET_FUNCTION_VALUE): Expand documentation of
2852 * calls.c (expand_call): If the return value is a PARALLEL,
2853 extract its first member.
2854 * config/m68k/linux.h (FUNCTION_EXTRA_EPILOGUE): Remove.
2855 * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
2856 use FUNCTION_EXTRA_EPILOGUE.
2857 (m68k_function_value): Return a PARALLEL if the return value
2858 is of a pointer type.
2859 * config/m68k/netbsd-elf.h (current_function_returns_pointer)
2860 (FUNCTION_EXTRA_EPILOGUE): Remove.
2861 * config/m68k/m68k.md (D0_REG): New constant.
2863 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
2865 * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
2866 output a NOP for empty epilogues.
2868 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2870 * config/m68k/m68k.c (m68k_use_return_insn): Update comments
2871 before function. Extend register save check to include all
2872 registers, not just integer ones.
2874 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
2876 * config/m68k/m68k-protos.h (use_return_insn): Rename to...
2877 (m68k_use_return_insn): ...this.
2878 * config/m68k/m68k.h (USE_RETURN_INSN): Delete.
2879 * config/m68k/m68k.c (use_return_insn): Rename to...
2880 (m68k_use_return_insn): ...this.
2881 * config/m68k/m68k.md (return): Use m68k_use_return_insn instead
2884 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
2886 * config/m68k/fpgnulib.c (__truncdfsf2): Implement round to
2887 nearest even, fix denormal rounding overflow.
2889 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2891 * config/m68k/m68k.md (movsf_cf_hard): Use fsmove instead of
2892 f%$smove and f%$move.
2893 (movdf_cf_hard): Use fdmove for cases 0 and 3 and fmove for case 1.
2894 (extendsfdf2_cf): Use fdmove instead of f%&move.
2895 (truncdfsf2_cf): Use fsmove instead of f%$smove.
2896 (add<mode>3_cf, sub<mode>3_cf): Use <FP:prec> instead of <FP:round>.
2898 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2900 * config/m68k/m68k.md (movdf_cf_hard): Use output_move_double for
2903 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2905 * real.h (real_format): Add a canonical_nan_lsbs_set field.
2906 (coldfire_single_format): Declare.
2907 (coldfire_double_format): Likewise.
2908 * real.c (encode_ieee_single): Use canonical_nan_lsbs_set instead
2909 of qnan_msb_set to determine the lower bits of a canonical
2911 (encode_ieee_double): Likewise.
2912 (encode_ieee_quad): Likewise.
2913 (ieee_single_format): Initialize canonical_nan_lsbs_set.
2914 (mips_single_format): Likewise.
2915 (ieee_double_format): Likewise.
2916 (mips_double_format): Likewise.
2917 (ieee_extended_motorola_format): Likewise.
2918 (ieee_extended_intel_96_format): Likewise.
2919 (ieee_extended_intel_128_format): Likewise.
2920 (ieee_extended_intel_96_round_53_format): Likewise.
2921 (ibm_extended_format): Likewise.
2922 (mips_extended_format): Likewise.
2923 (ieee_quad_format): Likewise.
2924 (mips_quad_format): Likewise.
2925 (vax_f_format): Likewise.
2926 (vax_d_format): Likewise.
2927 (vax_g_format): Likewise.
2928 (i370_single_format): Likewise.
2929 (i370_double_format): Likewise.
2930 (decimal_single_format): Likewise.
2931 (decimal_double_format): Likewise.
2932 (decimal_quad_format): Likewise.
2933 (c4x_single_format): Likewise.
2934 (c4x_extended_format): Likewise.
2935 (real_internal_format): Likewise.
2936 (coldfire_single_format): New real_format.
2937 (coldfire_double_format): Likewise.
2938 * config/pdp11/pdp11.c (pdp11_f_format): Initialize
2939 canonical_nan_lsbs_set.
2940 (pdp11_d_format): Likewise.
2941 * config/m68k/m68k.c (override_options): Override REAL_FORMAT_MODE
2942 if TARGET_COLDFIRE_CPU.
2944 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2946 * config/m68k/m68k-protos.h (m68k_output_pic_call): Delete.
2947 (output_call, m68k_legitimize_call_address): Declare.
2948 * config/m68k/m68k.h (EXTRA_CONSTRAINT): Remove unnecessary
2949 parenthesees. Add support for a 'W' constraint.
2950 (LEGITIMATE_PIC_OPERAND_P): Remove SYMBOL_REF_FLAG handling.
2951 (PRINT_OPERAND_PUNCT_VALID_P): Remove comment about 'o'.
2952 (m68k_symbolic_call, m68k_symbolic_jump): Declare.
2953 * config/m68k/m68k.c (m68k_symbolic_call, m68k_symbolic_jump): New
2955 (override_options): Initialize them. Do not set flag_no_function_cse
2956 for TARGET_ID_SHARED_LIBRARY.
2957 (m68k_output_pic_call): Delete.
2958 (m68k_legitimize_call_address): New function.
2959 (print_operand): Remove the %o prefix. Handle the %p prefix.
2960 (output_call): New function.
2961 (m68k_output_mi_thunk): Use m68k_symbolic_jump. Always load the
2962 target address from the GOT if symbolic jumps are not allowed.
2963 * config/m68k/m68k.md (call, general_operand): Do not set
2964 SYMBOL_REF_FLAG. Use m68k_legitimize_call_address instead.
2965 Merge separate flag_pic and !flag_pic define_insns into...
2966 (*call, *call_value): ...these new patterns. Match the address
2967 rather than the containing MEM and require it to be a call_operand.
2968 Use output_call to generate the asm template.
2969 * config/m68k/predicates.md (const_call_operand): New predicate.
2970 (call_operand): Likewise.
2972 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
2974 * config/m68k/m68k.h (REGISTER_MOVE_COST): Simplify definition.
2975 (STACK_GROWS_DOWNWARD): Define to 1.
2976 (FUNCTION_VALUE, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P): Equivocate
2977 comments, emphasizing that these values are only defaults.
2978 * config/m68k/linux.h (LINK_SPEC): Fix formatting in #undef.
2979 * config/m68k/m68k.c (const_method): Remove trailing whitespace.
2981 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2983 * config/m68k/m68k.md (cmpsi): Remove outdated flag_pic handling.
2985 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
2986 Richard Sandiford <richard@codesourcery.com>
2988 * config/m68k/m68k.h (DATA_REGNO_P, ADDRESS_REGNO_P, INT_REGNO_P)
2989 (FP_REGNO_P): New macros.
2990 (REGNO_OK_FOR_INDEX_P, REGNO_OK_FOR_BASE_P, REGNO_OK_FOR_DATA_P)
2991 (REGNO_OK_FOR_FP_P, REG_OK_FOR_INDEX_P, REG_OK_FOR_BASE_P): Use them.
2992 (EH_RETURN_STACKADJ_RTX): Use A0_REG.
2993 * config/m68k/m68k.c (m68k_regno_mode_ok): Use the new REGNO macros.
2995 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
2997 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
2998 (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-rtems*): Add t-floatlib
3000 * config/m68k/t-floatlib: New file, extracting common code from...
3001 * config/m68k/t-m68kbare, config/m68k/t-m68kelf,
3002 * config/m68k/t-uclinux: Here.
3003 * config/m68k/fpgnulib.c: Do not compile extendeed precision
3004 routines on ColdFire targets.
3006 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
3008 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
3009 (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*)
3010 (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-linux*)
3011 (m68k-*-rtems*): Use tm_file rather than m68k/m68k.h and
3012 explicitly set MOTOROLA to 1.
3013 * config/m68k/m68k.h (MOTOROLA): Simplify definition accordingly.
3015 2007-01-18 Richard Sandiford <richard@codesourcery.com>
3016 Nathan Sidwell <nathan@codesourcery.com>
3018 * config/m68k/m68k.h (PCC_STATIC_STRUCT_RETURN): Delete.
3019 (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Add MOTOROLA cases.
3020 * config/m68k/coff.h (REGISTER_PREFIX_MD): Delete.
3021 * config/m68k/m68020-elf.h (LIB_SPEC): Delete.
3022 * config/m68k/m68k-none.h (CC1_SPEC, CPP_SUBTARGET_SPEC): Delete.
3023 * config/m68k/m68kelf.h (IMMEDIATE_PREFIX, REGISTER_PREFIX_MD)
3024 (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Delete.
3025 (NO_DOLLAR_IN_LABEL, PCC_STATIC_STRUCT_RETURN): Don't undefine.
3026 (BSS_ASM_OP, ASM_OUTPUT_SKIP, ASM_OUTPUT_ASCII): Delete.
3027 * config/m68k/m68kemb.h (PCC_STATIC_STRUCT_RETURN): Don't undefine.
3028 * config/m68k/linux.h (SIZE_TYPE, PTRDIFF_TYPE, WCHAR_TYPE)
3029 (WCHAR_TYPE_SIZE, TARGET_OBJFMT_CPP_BUILTINS): Delete.
3030 (TARGET_OS_CPP_BUILTINS): Don't define mc68000 and mc68020 here.
3031 (DBX_REGISTER_NUMBER): Delete.
3032 * config/m68k/netbsd-elf.h (IMMEDIATE_PREFIX): Delete.
3033 (PCC_STATIC_STRUCT_RETURN): Don't undefine.
3034 * config/m68k/openbsd.h (PCC_STATIC_STRUCT_RETURN): Define.
3036 2007-01-18 Richard Sandiford <richard@codesourcery.com>
3038 * config.gcc (m68k-*-uclinux*): Add flat.h to $tm_file.
3039 * config/flat.h: New file.
3040 * crtstuff.c (USE_PT_GNU_EH_FRAME): Don't define if
3042 * config/m68k/m68k.h (ASM_PREFERRED_EH_DATA_FORMAT): Do not use
3043 indirect references for -msep-data or -mid-shared-library.
3044 Do not use PC-relative code addresses either.
3045 * config/m68k/m68k.c (override_options): Restrict -fPIC error
3047 * config/m68k/uclinux.h (STARTFILE_SPEC): Define. Use Scrt1.o
3048 for shared libraries and crt1.o for executables. Use crti.o and
3050 (ENDFILE_SPEC): Use crtend.o and crtn.o.
3051 (LIB_SPEC): Suppress -Rlibc.gdb if -static-libc is given.
3052 Do not add -elf2flt or -shared-lib-id options here.
3053 (LINK_SPEC): Define. Pass -elf2flt if no -elf2flt option is given.
3054 Pass -shared-lib-id if -mid-shared-library, taking the library
3055 identifier from -mshared-library-id if given, otherwise
3057 (EH_FRAME_IN_DATA_SECTION): Do not undefine.
3058 (INIT_SECTION_ASM_OP, FINI_SECTION_ASM_OP): Likewise.
3059 (TARGET_OS_CPP_BUILTINS): Define __GXX_MERGED_TYPEINFO_NAMES=0
3060 and __GXX_TYPEINFO_EQUALITY_INLINE=0 if -mid-shared-library.
3061 (DRIVER_SELF_SPECS): Map unadorned PIC options to -msep-data.
3062 * config/m68k/t-uclinux (EXTRA_MULTILIB_PARTS): Add crtbegin.o
3064 * config/m68k/lb1sf68.asm (PICCALL): Use an lea and pc-relative
3065 jump sequence for ISA A and ISA A+.
3066 (PICJUMP): Likewise.
3068 2007-01-18 Richard Sandiford <richard@codesourcery.com>
3070 * config.gcc (m68k-*-uclinux*): Base the port on the common
3071 and m68k GNU/Linux files rather than on the generic ELF ones.
3072 * config/m68k/uclinux.h (TARGET_VERSION): Override.
3073 (TARGET_OS_CPP_BUILTINS): Use LINUX_TARGET_OS_CPP_BUILTINS.
3075 2007-01-18 Julian Brown <julian@codesourcery.com>
3076 Richard Sandiford <richard@codesourcery.com>
3078 * config/m68k/m68k.h (LONG_DOUBLE_TYPE_SIZE): Make 64-bit on ColdFire.
3079 (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
3080 * config/m68k/netbsd-elf.h (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Undefine
3082 * config/m68k/uclinux-oldabi.h (LONG_DOUBLE_TYPE_SIZE): Redefine to
3084 (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
3086 2007-01-18 Richard Sandiford <richard@codesourcery.com>
3088 * doc/install.texi: Document m68k-uclinuxoldabi.
3089 * config.gcc (m68k-*-uclinuxoldabi): New configuration.
3090 * config/m68k/uclinux-oldabi.h: New file, copied from
3091 config/m68k/uclinux.h.
3093 2007-01-18 Kaz Kojima <kkojima@rr.iij4u.or.jp>
3095 * config/m32r/m32r-protos.h (m32r_expand_epilogue): Declare it.
3096 * config/m32r/m32r.c (m32r_setup_incoming_varargs): Use gen_frame_mem.
3097 (m32r_compute_frame_size): Use unsigned for regno.
3098 (m32r_reload_lr): Use gen_frame_mem.
3100 (m32r_output_function_epilogue): Don't output the function epilogue
3102 (m32r_expand_epilogue): New.
3103 (direct_return): Return false if the function has the interrupt
3105 (m32r_hard_regno_rename_ok): Remove code for the textual epilogue.
3106 * config/m32r/m32r.md (epilogue): New expander.
3107 (return_lr, return_rte): New insns.
3108 (return): Make it expander.
3109 (return_normal): New expander.
3111 2007-01-18 Josh Conner <jconner@apple.com>
3114 * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Add
3115 support for UNLE, UNLT, UNGE, and UNGT.
3117 2007-01-18 Jan Hubicka <jh@suse.cz>
3119 * tree-vrp.c (finalize_jump_threads): Do not call cleanup_cfg by hand.
3120 * tree-tailcall (add_virtual_phis): Likewise.
3121 (optimize_tail_call): Return TODOs.
3122 (execute_tail_calls): Return TODOs.
3123 * tree-ssa-ccp (execute_fold_all_builtins): Do cleanup_cfg via TODO.
3124 * tree-cfgcleanup.c (cleanup_tree_cfg_loop): Return if something
3126 * tree-ssa-forwprop.c (tree_ssa_forward_propagate_single_use_value):
3127 Cleanup cfg using TODO.
3128 * tree-flow.h (cleanup_tree_cfg_loop): Update prototype.
3129 * passes.c (execute_function_todo): When cleanup did something, remove
3131 * tree-cfg.c (pass_build_cfg): Add cleanup_cfg TODO.
3132 (make_edges): Don't cleanup_cfg.
3134 2007-01-18 Uros Bizjak <ubizjak@gmail.com>
3136 * reg-stack.c (subst_stack_regs_pat) [UNSPEC_SINCOS_COS,
3137 UNSPEC_XTRACT_FRACT]: Use generic code for instructions that
3138 operate on the top of stack.
3139 [UNSPEC_SINCOS_SIN, UNSPEC_XTRACT_EXP, UNSPEC_TAN]: Rewrite
3140 register handling of instructions that output to the second
3142 [UNSPEC_TAN_ONE, UNSPEC_TAN_TAN]: Remove.
3143 (move_for_stack_reg): Special-case check for dead destination
3144 stack slot for constant load of 1.0 inside UNSPEC_TAN.
3146 * config/i386/i386.md (UNSPEC_TAN): New constant.
3147 (UNSPEC_TAN_ONE, UNSPEC_TAN_TAN): Remove.
3148 (fptanxf4_i387, fptan_extend<mode>xf4_i387): New patterns
3149 to correctly model move of constant 1.0 to top stack slot.
3150 (*tandf3_1, *tansf3_1, *tanxf3_1): Remove insn patterns.
3151 (unnamed peephole2 pattern): Remove corresponding peephole2
3152 pattern that optimizes tan insn and loading of constant 1.0.
3153 (tanxf2): Use fptanxf4_i387.
3154 (tan<mode>2): Rename from tansf2 and tandf2 and macroize insn
3155 patterns using X87MODEF12 mode macro. Use fptan_extend<mode>xf4_i387
3156 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
3157 disable patterns for SSE math.
3158 (sincos<mode>3): Use truncxf<mode>2_i387_noop for truncation.
3159 (fyl2x_extend<mode>xf3_i387): Use X87MODEF12 for operand 1.
3161 2007-01-18 Dirk Mueller <dmueller@suse.de>
3162 Richard Guenther <rguenther@suse.de>
3165 * doc/invoke.texi (Warray-bounds): Document -Warray-bounds.
3166 * common.opt (Warray-bounds): Add new warning option.
3167 * c-opts.c (c_common_handle_option): Define -Warray-bounds
3169 * Makefile.in: make tree-vrp.o depend on toplev.h
3170 * tree-vrp.c (vrp_finalize): Call check_array_refs if -Warray-bounds
3172 (check_array_refs, check_array_bounds, check_array_ref): New.
3174 2007-01-18 Jan Hubicka <jh@suse.cz>
3176 * tree-ssa-ccp.c (ccp_finalize): Return if something changed.
3177 (execute_ssa_ccp): Return flags conditionally.
3178 * tree-ssa-propagate.c (substitue_and_fold): Return if something was
3180 * tree-ssa-propagate.h (substitute_and_fold): Update prototype.
3182 2007-01-18 Steven Bosscher <steven@gcc.gnu.org>
3184 * cfgcleanup.c (cleanup_cfg): Detect cfglayout mode and set
3185 the CLEANUP_CFGLAYOUT flag when in cfglayout mode.
3187 * Makefile.c (GTFILES): Add cfglayout.h.
3188 * gengtype.c (open_base_files): Likewise.
3189 * cfglayout.c (cfg_layout_function_footer,
3190 cfg_layout_function_header) Reindent to make gengtype happy.
3191 * cfglayout.h (cfg_layout_function_footer,
3192 cfg_layout_function_header): Add GTY(()) marker.
3194 * ifcvt.c (noce_try_sign_mask): Make sure INSN_B is non-null.
3196 2007-01-18 Ben Elliston <bje@au.ibm.com>
3198 * genautomata.c (write_automata): Include xstrerror output in the
3199 error message if writing the DFA description file fails.
3201 2007-01-17 H.J. Lu <hongjiu.lu@intel.com>
3203 * config/mips/mips-protos.h (mips_output_external): Make it
3205 * config/mips/iris.h (TARGET_ASM_EXTERNAL_LIBCALL): Removed.
3206 * config/mips/mips.c (irix_output_external_libcall): Likewise.
3207 (extern_list): Likewise.
3208 (extern_head): Likewise.
3209 (TARGET_ASM_FILE_END): Likewise.
3210 (mips_file_end): Likewise.
3211 (mips_output_external): Rewritten.
3213 2007-01-18 Ben Elliston <bje@au.ibm.com>
3215 * genpreds.c (write_insn_preds_c): Only write out the function
3216 body for regclass_for_constraint if we have register constraints.
3218 2007-01-17 Tom Tromey <tromey@redhat.com>
3220 * doc/sourcebuild.texi (libgcj Tests): Use sourceware.org.
3221 * doc/install.texi (Testing): Use sourceware.org.
3222 (Binaries): Likewise.
3223 (Specific): Likewise.
3224 * doc/contrib.texi (Contributors): Use sourceware.org.
3226 2007-01-17 Anatoly Sokolov <aesok@post.ru>
3228 * config/avr/avr.h (AVR_HAVE_LPMX): New macro.
3229 (AVR_ENHANCED): Rename to ...
3230 (AVR_HAVE_MUL): ... new.
3231 (avr_enhanced_p): Rename to ...
3232 (avr_have_mul_p): ... new.
3233 (TARGET_CPU_CPP_BUILTINS): Use 'avr_have_mul_p' instead of
3234 'avr_enhanced_p' for "__AVR_ENHANCED__". Define "__AVR_HAVE_MUL__".
3235 * config/avr/avr.c (avr_enhanced_p): Rename to ...
3236 (avr_have_mul_p): ... new.
3237 (base_arch_s): Rename 'enhanced' to 'have_mul'.
3238 (avr_override_options): Use 'avr_have_mul_p' and 'have_mul' instead of
3239 'avr_enhanced_p' and 'enhanced'.
3240 (ashlhi3_out, ashrhi3_out, lshrhi3_out, avr_rtx_costs): Use
3241 AVR_HAVE_MUL instead of AVR_ENHANCED.
3242 * avr.md (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
3243 (mulqi3, *mulqi3_enh, *mulqi3_call, mulqihi3, umulqihi3, mulhi3,
3244 *mulhi3_enh, *mulhi3_call, mulsi3, *mulsi3_call): Use AVR_HAVE_MUL
3245 instead of AVR_ENHANCED.
3246 (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
3247 * libgcc.S: Use __AVR_HAVE_MUL__ instead of __AVR_ENHANCED__.
3248 (__tablejump__): Use __AVR_HAVE_LPMX__ instead of __AVR_ENHANCED__.
3250 2007-01-17 Ian Lance Taylor <iant@google.com>
3252 * vec.h (VEC_reserve_exact): Define.
3253 (vec_gc_p_reserve_exact): Declare.
3254 (vec_gc_o_reserve_exact): Declare.
3255 (vec_heap_p_reserve_exact): Declare.
3256 (vec_heap_o_reserve_exact): Declare.
3257 (VEC_OP (T,A,reserve_exact)): New static inline function, three
3259 (VEC_OP (T,A,reserve)) [all versions]: Remove handling of
3261 (VEC_OP (T,A,alloc)) [all versions]: Call ...reserve_exact.
3262 (VEC_OP (T,A,copy)) [all versions]: Likewise.
3263 (VEC_OP (T,a,safe_grow)) [all versions]: Likewise.
3264 * vec.c (calculate_allocation): Add exact parameter. Change all
3266 (vec_gc_o_reserve_1): New static function, from vec_gc_o_reserve.
3267 (vec_gc_p_reserve, vec_gc_o_reserve): Call vec_gc_o_reserve_1.
3268 (vec_gc_p_reserve_exact, vec_gc_o_reserve_exact): New functions.
3269 (vec_heap_o_reserve_1): New static function, from vec_heap_o_reserve.
3270 (vec_heap_p_reserve, vec_heap_o_reserve): Call vec_heap_o_reserve_1.
3271 (vec_heap_p_reserve_exact): New function.
3272 (vec_heap_o_reserve_exact): New function.
3274 2007-01-17 Jan Hubicka <jh@suse.cz>
3276 * ipa-type-escape.c (look_for_casts): Revamp using handled_component_p.
3278 2007-01-17 Eric Christopher <echristo@apple.com>
3280 * config.gcc: Support core2 processor.
3282 2007-01-16 Jan Hubicka <jh@suse.cz>
3284 * tree-ssanames.c (release_dead_ssa_names): Instead of ggc_freeing
3285 the names, just unlink the chain so we don't crash on dangling pointers
3288 2007-01-16 Jan Hubicka <jh@suse.cz>
3290 * cgraph.h (cgraph_decide_inlining_incrementally): Kill.
3291 * tree-pass.h: Reorder to make IPA passes appear toegher.
3292 (pass_early_inline, pass_inline_parameters, pass_apply_inline): Declare.
3293 * cgraphunit.c (cgraph_finalize_function): Do not compute inling
3294 parameters, do not call early inliner.
3295 * ipa-inline.c: Update comments. Include tree-flow.h
3296 (cgraph_decide_inlining): Do not compute inlining parameters.
3297 (cgraph_decide_inlining_incrementally): Return TODOs; assume to
3298 be called with function context set up.
3299 (pass_ipa_inline): Remove unreachable functions before pass.
3300 (cgraph_early_inlining): Simplify assuming to be called from the
3302 (pass_early_inline): New pass.
3303 (cgraph_gate_ipa_early_inlining): New gate.
3304 (pass_ipa_early_inline): Turn into simple wrapper.
3305 (compute_inline_parameters): New function.
3306 (gate_inline_passes): New gate.
3307 (pass_inline_parameters): New pass.
3308 (apply_inline): Move here from tree-optimize.c
3309 (pass_apply_inline): New pass.
3310 * ipa.c (cgraph_remove_unreachable_nodes): Verify cgraph after
3312 * tree-inline.c (optimize_inline_calls): Return TODOs rather than
3314 (tree_function_versioning): Do not allocate dummy struct function.
3315 * tree-inline.h (optimize_inline_calls): Update prototype.
3316 * tree-optimize.c (execute_fixup_cfg): Export.
3317 (pass_fixup_cfg): Remove
3318 (tree_rest_of_compilation): Do not apply inlines.
3319 * tree-flow.h (execute_fixup_cfg): Declare.
3320 * Makefile.in (gt-passes.c): New.
3321 * passes.c: Include gt-passes.h
3322 (init_optimization_passes): New passes.
3323 (nnodes, order): New static vars.
3324 (do_per_function_toporder): New function.
3325 (execute_one_pass): Dump current pass here.
3326 (execute_ipa_pass_list): Don't dump current pass here.
3328 2007-01-16 Janis Johnson <janis187@us.ibm.com>
3330 * config/dfp-bit.c (dfp_compare_op): Return separate value for NaN.
3331 (DFP_NE, DFP_LE, DFP_GE): Return false for NaN.
3333 2007-01-16 David Edelsohn <edelsohn@gnu.org>
3335 * config/rs6000/darwin-ldouble.c: Build file for SOFT_FLOAT.
3336 (strong_alias): Define.
3337 (__gcc_qmul): Provide non-FMA for soft-float.
3343 (__gcc_qunord): New.
3353 * config/rs6000/rs6000.c (rs6000_init_libfuncs): Initialize
3354 soft-float functions.
3355 * config/rs6000/libgcc-ppc-glibc.ver: Version soft-float symbols.
3356 * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Do not warn
3357 about long double soft float.
3359 2007-01-16 Dorit Nuzman <dorit@il.ibm.com>
3360 Tehila Meyzels <tehila@il.ibm.com>
3362 * tree-vectorizer.h (is_pattern_stmt_p): New.
3363 * tree-vect-analyze.c (vect_determine_vectorization_factor): Fix
3364 formatting (tabs instead of spaces). Cleanup and clarify setting
3365 of STMT_VINFO_VECTYPE. Call is_pattern_stmt_p.
3366 * tree-vect-transform.c (vect_get_vec_def_for_operand): Fix typo.
3367 (vectorizable_type_demotion): Check that types are integral.
3368 (vectorizable_type_promotion): Likewise.
3369 (vectorizable_store): Fix typo. Eliminate new-line at end of
3372 2007-01-16 Jan Hubicka <jh@suse.cz>
3374 * tree-ssanames.c (release_dead_ssa_names): Remove invalidated
3377 2007-01-15 Eric Christopher <echristo@apple.com>
3379 * ifcvt.c: Include vec.h, vecprim.h.
3380 (check_cond_move_block): New argument regs.
3381 Reorganize. Add registers used to regs.
3382 (cond_move_process_if_block): Use regs set above as
3385 2007-01-15 Eric Christopher <echristo@apple.com>
3387 * config/darwin.h: Update copyright.
3388 (TARGET_OPTION_TRANSLATE_TABLE): Add umbrella.
3389 (LINK_COMMAND_SPEC): Add -u.
3390 (LINK_SPEC): Fix umbrella for above.
3392 2007-01-15 Joseph S. Myers <joseph@codesourcery.com>
3394 * config/soft-fp/op-common.h, config/soft-fp/op-4.h: Update from
3397 2007-01-15 Tom Tromey <tromey@redhat.com>
3399 * doc/sourcebuild.texi (libgcj Tests): Don't mention jacks.
3400 * doc/install.texi (Testing): Don't mention jacks.
3401 (Configuration): Document --enable-java-maintainer-mode. Move
3402 --with-java-home to libgcj-specific section. Document
3404 (Prerequisites): Mention --enable-java-maintainer-mode, ecj1.
3406 2007-01-15 Jan Hubicka <jh@suse.cz>
3408 * tree-ssa-dce.c (DCE_TODOs): New.
3409 (propagate_necessity): Return if something changed.
3410 (eliminate_unnecessary_stmts): Likewise.
3411 (perform_tree_ssa_dce): Return TODO flags when needed.
3412 (pass_dce, pass_dce_loop, pass_cd_dce): Remove TODO flags.
3414 2007-01-15 Uros Bizjak <ubizjak@gmail.com>
3416 * config/i386/i386.md (fyl2xxf3_i387): Rename from fyl2x_xf3.
3417 (fyl2x_extend<mode>xf3_i387): New insn pattern.
3418 (log<mode>2): Rename from logsf2 and logdf2 and macroize insn
3419 patterns using X87MODEF12 mode macro. Extend operand 1
3420 to XFmode. Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
3421 (log10<mode>2): Ditto.
3422 (log2<mode>2): Ditto.
3423 (log1p<mode>2): Ditto.
3424 (logb<mode>2): Ditto.
3425 (fyl2xp1xf3_i387): Rename from fyl2xp1_xf3.
3426 (fyl2xp1_extend<mode>xf3_i387): New insn pattern.
3427 (*fxtractxf3_i387): Rename from *fxtractxf3.
3428 (fxtract_extend<mode>xf3_i387): New insn pattern.
3429 (ilogbsi2): Use match_dup 3, not match_operand:XF 3.
3430 * config/i386/i386.c (ix86_emit_i387_log1p): Use gen_fyl2xp1xf3_i387()
3431 and gen_fyl2xxf3_i387().
3433 2007-01-14 Zdenek Dvorak <dvorakz@suse.cz>
3435 * loop-unswitch.c (unswitch_loop): Do not call fix_loop_placement.
3436 * cfgloopmanip.c (fix_loop_placement): Made static. Use
3437 get_loop_exit_edges. Changed return type to bool.
3438 * cfgloop.h (fix_loop_placement): Declaration removed.
3440 2007-01-14 Dorit Nuzman <dorit@il.ibm.com>
3442 * param.h (MIN_VECT_LOOP_BOUND): New.
3443 * params.def (MIN_VECT_LOOP_BOUND): New.
3444 * tree-vectorizer.c (slpeel_tree_peel_loop_to_edge): Takes another
3445 argument - minimum threshold for number of iterations.
3446 * tree-vectorizer.h (slpeel_tree_peel_loop_to_edge): Add another
3447 argument to declaration.
3448 * tree-vect-analyze.c (vect_analyze_operations): Check value of
3449 MIN_VECT_LOOP_BOUND.
3450 * tree-vect-transform.c (vect_do_peeling_for_loop_bound): Call
3451 slpeel_tree_peel_loop_to_edge with additional argument.
3452 (vect_do_peeling_for_alignment): Likewise.
3453 * doc/invoke.texi (min-vect-loop-bound): Document new param option.
3455 2007-01-14 Uros Bizjak <ubizjak@gmail.com>
3458 * config/i386/i386.c (print_operand) ['z']: Output 'b' for
3461 2007-01-14 Jan Hubicka <jh@suse.cz>
3463 * tree-dfa.c (remove_referenced_var): New function.
3464 * tree-ssa-live.c (remove_unused_locals): Walk referenced vars and
3465 prune referenced vars list too.
3466 * tree-flow.h (remove_referenced_var): Declare.
3468 2007-01-14 Jan Hubicka <jh@suse.cz>
3470 * tree-eh.c (add_stmt_to_eh_region_fn): Do not add call_exprs
3472 (remove_stmt_from_eh_region_fn): Do not remove call_exprs.
3473 (verify_eh_throw_stmt_node, verify_eh_throw_table_statements): Kill.
3474 * except.h (verify_eh_throw_table_statements): Kill prototype.
3475 * cfgexpand.c (expand_gimple_basic_block): Propagate Eh regions
3476 into call exrepssions.
3477 * tree-optimize.c (execute_free_cfg_annotatiosn): Do not call
3479 * tree-cfg.c: Include pointer-set.h.
3480 (verify_node_sharing): Work on pointer set.
3481 (verify_eh_throw_stmt_node): New.
3482 (verify_stmts): Use pointers sets, verify throw_stmt.
3484 2007-01-13 Zdenek Dvorak <dvorakz@suse.cz>
3486 * ipa-reference.c (analyze_function): Consider also addresses taken
3489 2007-01-12 Roger Sayle <roger@eyesopen.com>
3491 * c-typeck.c (null_pointer_constant_p): Replace use of
3492 TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
3493 (build_c_cast): Likewise.
3495 2007-01-12 Roger Sayle <roger@eyesopen.com>
3497 * tree.h (force_fit_type_double): Remove unused final argument.
3498 * c-common.c (constant_expression_warning): Replace use of
3499 TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
3500 (convert_and_check): Likewise.
3501 (shorten_compare): Update call to force_fit_type_double.
3502 (c_common_truthvalue_conversion) <INTEGER_CST>: Use integer_zerop.
3503 * convert.c (convert_to_pointer): Update call to
3504 force_fit_type_double.
3505 * fold-const.c (force_fit_type_double): Remove overflowed_const
3507 (int_const_binop, fold_convert_const_int_from_int,
3508 fold_convert_const_int_from_real, fold_div_compare,
3509 fold_sign_changed_comparison, fold_unary, fold_negate_const,
3510 fold_abs_const, fold_not_const): Remove the final argument from
3511 calls to force_fit_type_double.
3513 2007-01-12 Andrew Pinski <andrew_pinski@playstation.sony.com>
3515 * configure.ac: Set insn to "nop" for spu-*-* also.
3516 * configure: Regenerate.
3518 2007-01-12 Olga Golovanevsky <olga@il.ibm.com>
3520 * builtins.def : Add BUILT_IN_FREE.
3522 2007-01-12 Jan Hubicka <jh@suse.cz>
3524 PR tree-optimization/30443
3525 * tree-inline.c (tree_function_versioning): Do not optimize when
3526 cloning for inlining.
3528 2007-01-12 Zdenek Dvorak <dvorakz@suse.cz>
3530 * doc/loop.texi: Document recording of loop exits.
3531 * cfgloopmanip.c (loopify, duplicate_loop): Use alloc_loop.
3532 (update_single_exits_after_duplication,
3533 update_single_exit_for_duplicated_loop,
3534 update_single_exit_for_duplicated_loops): Removed.
3535 (duplicate_loop_to_header_edge): Do not call
3536 update_single_exits_after_duplication and
3537 update_single_exit_for_duplicated_loops.
3538 (loop_version): Do not update single_exit information.
3539 (fix_loop_structure): Use record_loop_exits instead of
3540 mark_single_exit_loops.
3541 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update
3542 the lists of loop exits.
3543 * cfghooks.c (redirect_edge_and_branch, redirect_edge_and_branch_force,
3544 split_edge, merge_blocks): Update the lists of loop exits.
3545 * modulo-sched.c (sms_schedule): Pass LOOPS_HAVE_RECORDED_EXITS to
3546 loop_optimizer_init.
3547 * loop-init.c (loop_optimizer_init): Call record_loop_exits instead
3548 of mark_single_exit_loops.
3549 (loop_optimizer_finalize): Call release_recorded_exits.
3550 * tree-ssa-loop.c (tree_loop_optimizer_init): Pass
3551 LOOPS_HAVE_RECORDED_EXITS to loop_optimizer_init.
3552 * tree-vectorizer.c (slpeel_tree_duplicate_loop_to_edge_cfg): Do not
3553 update single exit information.
3554 * lambda-code.c (perfect_nestify): Ditto.
3555 * cfgloop.c (flow_loop_free): Destroy the list of exits of the loop.
3556 (mark_single_exit_loops): Removed.
3557 (alloc_loop, loop_exit_hash, loop_exit_eq, loop_exit_free,
3558 get_exit_descriptions, rescan_loop_exit, record_loop_exits,
3559 dump_recorded_exit, dump_recorded_exits, release_recorded_exits): New
3561 (get_loop_exit_edges, single_exit): Use recorded exit lists.
3562 (add_bb_to_loop, remove_bb_from_loops): Update the lists of loop exits.
3563 (verify_loop_structure): Verify consistency of the exit lists.
3564 (flow_loops_find): Use alloc_loop. Initialize exits hash.
3565 (set_single_exit): Removed.
3566 * cfgloop.h (struct loop_exit): New function.
3567 (struct loop): single_exit_ field replaced by exits field.
3568 (LOOPS_HAVE_MARKED_SINGLE_EXITS): Replaced by LOOPS_HAVE_RECORDED_EXITS.
3569 (struct loops): Added exits hash.
3570 (mark_single_exit_loops, set_single_exit): Declaration removed.
3571 (release_recorded_exits, record_loop_exits, rescan_loop_exit): Declare.
3573 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3575 * doc/invoke.texi: Avoid use of @headitem.
3577 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3579 * cse.c (cse_insn): Move HAVE_CC0 code after declarations.
3581 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3583 * doc/install.texi: Fix m68k-*-* anchor and add m68k-*-* to the
3586 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
3587 Richard Sandiford <richard@codesourcery.com>
3589 * doc/invoke.texi: Document -mno-bitfield, -mno-rtd and -mno-short.
3590 * config/m68k/m68k.opt: Resort options.
3591 (mbitfield, mrtd, mshort): Remove RejectNegative properties.
3593 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
3594 Richard Sandiford <richard@codesourcery.com>
3596 * doc/invoke.texi: Document the macros that are defined by
3597 m68k's -mtune and -mhard-float options.
3598 * config/m68k/m68k-protos.h (m68k_cpp_cpu_ident) Declare.
3599 (m68k_cpp_cpu_family): Likewise.
3600 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Add a full set
3601 of __ucfv*__ macros. Define __mcffpu__ if generating code for
3602 ColdFire FPUs. Define __mcf_cpu_* and __mcf_family_* macros.
3603 * config/m68k/m68k.c (m68k_cpp_cpu_ident): New function.
3604 (m68k_cpp_cpu_family): Likewise.
3606 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3608 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Treat all mc68*
3609 macros besides mc68000 as tuning macros. Use a switch statement
3610 to set them and mcpu32.
3612 2007-01-12 Julian Brown <julian@codesourcery.com>
3614 * config/m68k/m68k.h: Use TARGET_68040 instead of TARGET_68040_ONLY.
3615 (TARGET_68040_ONLY): Rename to...
3616 (TARGET_68040): ...this.
3617 * config/m68k/m68k.c: Use TARGET_68040 instead of TARGET_68040_ONLY.
3618 * config/m68k/m68k.md: Likewise.
3620 2007-01-12 Julian Brown <julian@codesourcery.com>
3621 Nathan Sidwell <nathan@codesourcery.com>
3622 Richard Sandiford <richard@codesourcery.com>
3624 * config.gcc (m680[012]0-*-*, m68k*-*-*): Set m68k_cpu_ident to
3625 the -mcpu= argument associated with the --with-cpu setting.
3626 Define M68K_DEFAULT_TUNE to the default -mtune= option,
3627 if different from the one implied by the -mcpu setting.
3628 Accept --with-cpu=FOO if FOO is listed in m68k-devices.def,
3629 using mcpu=FOO as the default CPU option. Set target_cpu_default2.
3630 * doc/invoke.texi: Mention ColdFire in the introduction to the
3631 m68k options. Document the new -march, -mcpu, -mtune, -mdiv,
3632 -mno-div and -mhard-float options. Make -m68881 a synonym for
3633 -mhard-float. Document the previously-undocumented -m5206e,
3634 -m528x, -m5307 and -m5407 options. Tweak the existing option
3635 documentation for consistency.
3636 * doc/install.texi: Mention new --with-cpu arguments.
3637 * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Only use the
3638 default CPU if neither -mcpu nor -march are specified.
3639 (ASM_CPU_SPEC): Pass down -mcpu and -march options.
3640 (TARGET_CPU_CPP_BUILTINS): Set __mcfisa*__ macros from
3641 TARGET_ISA*. Set the legacy __mcf*__ cpu macros in the same way,
3642 using m68k_tune to decide between families that implement the
3643 same ISA. Use m68k_tune to set __mcfv4e__.
3644 (FL_BITFIELD, FL_68881, FL_COLDFIRE, FL_CF_HWDIV, FL_CF_MAC)
3645 (FL_CF_EMAC, FL_CF_EMAC_B, FL_CF_USP, FL_CF_FPU, FL_ISA_68000)
3646 (FL_ISA_68010, FL_ISA_68020, FL_ISA_68040, FL_ISA_A, FL_ISA_B)
3647 (FL_ISA_C, FL_ISA_MMU): New macros.
3648 (MASK_COLDFIRE): Delete.
3649 (TARGET_68010, TARGET_68020, TARGET_68040_ONLY, TARGET_COLDFIRE)
3650 (TARGET_ISAB): Redefine in terms of m68k_cpu_flags.
3651 (TARGET_68881, TARGET_COLDFIRE_FPU): Redefine in terms of m68k_fpu.
3652 (TARGET_HARD_FLOAT): Do not define here.
3653 (TARGET_ISAAPLUS, TARGET_ISAC): New macros.
3654 (TUNE_68000): New macro.
3655 (TUNE_68000_10): Redefine in terms of TUNE_68000 and TUNE_68010.
3656 (TUNE_68010, TUNE_68030, TUNE_68040, TUNE_68060, TUNE_CPU32)
3657 (TUNE_CFV2): Redefine in terms of m68k_tune.
3658 (uarch_type, target_device, fpu_type): New enums.
3659 (m68k_cpu, m68k_tune, m68k_fpu, m68k_cpu_flags): Declare.
3660 * config/m68k/m68k.c (TARGET_DEFAULT): Remove MASK_68881.
3661 (FL_FOR_isa_00, FL_FOR_isa_10, FL_FOR_isa_20, FL_FOR_isa_40)
3662 (FL_FOR_isa_cpu32, FL_FOR_isa_a, FL_FOR_isa_aplus, FL_FOR_isa_b)
3663 (FL_FOR_isa_c): New macros.
3664 (m68k_isa): New enum.
3665 (m68k_target_selection): New structure.
3666 (all_devices, all_isas, all_microarchs): New tables.
3667 (m68k_cpu_entry, m68k_arch_entry, m68k_tune_entry, m68k_cpu)
3668 (m68k_tune, m68k_fpu, m68k_cpu_flags): New variables.
3669 (MASK_ALL_CPU_BITS): Delete.
3670 (m68k_find_selection): New function.
3671 (m68k_handle_option): Handle -mcpu=, -march= and -mtune=.
3672 Map the legacy target options to a combination of the new ones.
3673 (override_options): Set m68k_cpu, m68k_tune, m68k_fpu and
3674 m68k_cpu_flags. Handle M68K_DEFAULT_TUNE. Use m68k_cpu_flags
3675 to derive default MASK_BITFIELD, MASK_CF_HWDIV and MASK_HARD_FLOAT
3677 * config/m68k/m68k.opt (m5200, m5206e, m528x, m5307, m5407, mcfv4e)
3678 (m68010, m68020, m68020-40, m68020-60, m68030, m68040): Remove Mask
3680 (m68881, msoft-float): Change mask from 68881 to HARD_FLOAT.
3681 (march=, mcpu=, mdiv, mhard-float, mtune=): New options.
3682 * config/m68k/m68k-devices.def: New file.
3684 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3685 Nathan Sidwell <nathan@codesourcery.com>
3687 * config/m68k/m68k.h (ASM_CPU_SPEC, ASM_SPEC, EXTRA_SPECS)
3688 (SUBTARGET_EXTRA_SPECS): New macros.
3689 * config/m68k/linux.h (ASM_SPEC): Remove CPU flags;
3690 use %(asm_cpu_spec) instead.
3691 * config/m68k/m68k-none.h (ASM_SPEC): Likewise.
3692 * config/m68k/openbsd.h (ASM_SPEC): Likewise.
3693 * config/m68k/netbsd-elf.h (ASM_SPEC): Likewise.
3694 (EXTRA_SPECS): Rename to...
3695 (SUBTARGET_EXTRA_SPECS): ...this.
3697 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
3698 Richard Sandiford <richard@codesourcery.com>
3699 Julian Brown <julian@codesourcery.com>
3701 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
3702 (m68k-*-uclinux*, m68k-*-linux*, m68k-*-rtems*): Set default_m68k_cpu
3703 to the configuration's default CPU.
3704 (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*): Likewise.
3705 Remove default masks.
3706 (m680[012]0-*-*): Set the default with_cpu to the first part of
3708 (m68k*-*-*): Set the default with_cpu to m$default_m68k_cpu.
3709 (m68k*-*-linux): Extend the --with-cpu handling to...
3710 (m680[012]0-*-*, m68k*-*-*): ...these configurations. Allow m68000
3711 and m68010. Don't set target_cpu_default2.
3712 * doc/install.texi: Document --with-cpu for m68k.
3713 * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Define.
3714 * config/m68k/m68k-none.h (TARGET_CPU_DEFAULT, M68K_CPU_m68k)
3715 (M68K_CPU_m68000, M68K_CPU_m68010, M68K_CPU_m68020, M68K_CPU_m68030)
3716 (M68K_CPU_m68040, M68K_CPU_m68302, M68K_CPU_m68332, TARGET_DEFAULT)
3717 (ASM_CPU_DEFAULT_SPEC, CC1_CPU_DEFAULT_SPEC): Delete.
3718 (ASM_SPEC): Remove use of %(asm_cpu_default).
3719 (EXTRA_SPECS, SUBTARGET_EXTRA_SPECS, MULTILIB_DEFAULTS): Delete.
3720 * config/m68k/linux.h (TARGET_DEFAULT): Delete.
3721 (CPP_SPEC): Merge definitions. Do not handle __HAVE_68881__ here.
3722 * config/m68k/netbsd-elf.h (TARGET_OS_CPP_BUILTINS): Define
3723 __HAVE_FPU__ if TARGET_HARD_FLOAT.
3724 (TARGET_DEFAULT): Delete.
3725 (EXTRA_SPECS): Delete cpp_cpu_default_spec, cpp_cpu_spec,
3726 cpp_fpu_spec, asm_default_spec and netbsd_cpp_spec.
3727 (CPP_CPU_SPEC): Delete.
3728 (TARGET_VERSION): Merge definitions, using TARGET_68010 to pick
3729 the appropriate string.
3730 (CPP_CPU_DEFAULT_SPEC, ASM_DEFAULT_SPEC, CPP_FPU_SPEC): Delete.
3731 (CPP_SPEC): Define to NETBSD_CPP_SPEC.
3732 (ASM_SPEC): Don't use %(asm_default_spec).
3733 * config/m68k/m68k.c (TARGET_DEFAULT_TARGET_FLAGS): Remove
3734 TARGET_DEFAULT and add MASK_68881.
3735 * config/m68k/m68k.md: Remove mention of TARGET_DEFAULT from comments.
3737 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3739 * config.gcc (m68010-*-netbsdelf*): Add MASK_68010.
3740 (m68k*-*-netbsdelf*, m68k*-*-openbsd*, m68k*-linux*): Add
3741 MASK_68010 alongside MASK_68020.
3742 * doc/invoke.texi: Document -m68010.
3743 * config/m68k/m68k.opt (m68010): New.
3744 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Define mc68010
3746 (TUNE_68010): New macro.
3747 * config/m68k/m68k-none.h (M68K_CPU_m68k, M68K_CPU_m68010)
3748 (M68K_CPU_m68020, M68K_CPU_m68030, M68K_CPU_m68040)
3749 (M68K_CPU_m68332): Add MASK_68010.
3750 * config/m68k/linux.h (TARGET_DEFAULT): Add MASK_68010 to
3751 fallback definition.
3752 * config/m68k/netbsd-elf.h (CPP_CPU_SPEC): Remove now-redundant
3754 * config/m68k/m68k.c (MASK_ALL_CPU_BITS): Add MASK_68010.
3755 (m68k_handle_option): Handle OPT_m68010. Add MASK_68010
3756 to all entries that use MASK_68020.
3757 (output_move_simode_const, output_move_himode, output_move_qimode)
3758 (output_move_stricthi, output_move_strictqi): Use TARGET_68010
3759 instead of TARGET_68020 to select clr behavior. Remove comment
3760 about there being no TARGET_68010.
3761 * config/m68k/m68k.md: Likewise throughout.
3763 2007-01-12 Julian Brown <julian@codesourcery.com>
3765 * config/m68k/m68k.h (TARGET_ISAB): New macro.
3766 * config/m68k/m68k.c: Use TARGET_ISAB rather than TARGET_CFV4.
3767 * config/m68k/m68k.md: Likewise.
3769 2007-01-12 Julian Brown <julian@codesourcery.com>
3771 * config/m68k/m68k.h (LEGITIMATE_INDEX_P, LEGITIMIZE_ADDRESS): Use
3772 TARGET_COLDFIRE_FPU instead of TARGET_CFV4E.
3774 2007-01-12 Julian Brown <julian@codesourcery.com>
3776 * config/m68k/m68k.h (TUNE_68040_60): New macro.
3777 * config/m68k/m68k.c (standard_68881_constant_p): Use it.
3778 * config/m68k/m68k.md: Likewise.
3780 2007-01-12 Julian Brown <julian@codesourcery.com>
3781 Richard Sandiford <richard@codesourcery.com>
3783 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Use TUNE_68030
3784 instead of TARGET_68030, TUNE_68040 instead of TARGET_68040,
3785 TUNE_68060 instead of TARGET_68060 and TUNE_CPU32 instead of
3787 (TARGET_CPU32): Rename to...
3788 (TUNE_CPU32): ...this.
3789 (TUNE_68000_10, TUNE_68030, TUNE_68040, TUNE_68060)
3790 (TUNE_CFV2): New macros.
3791 * config/m68k/netbsd-elf.h (LONG_DOUBLE_TYPE_SIZE): Simplify;
3792 remove conditions that are implied by TARGET_68020.
3793 * config/m68k/m68k.c (m68k_output_function_prologue): Use TUNE_68040
3794 instead of TARGET_68040 and TUNE_CPU32 instead of TARGET_CPU32.
3795 (m68k_output_function_epilogue): Likewise.
3796 (m68k_rtx_costs): Likewise. Use TUNE_68060 instead of TARGET_68060
3797 and TUNE_CFV2 instead of TARGET_5200. Use TUNE_68000_10 instead of
3798 "!TARGET_68020 && !TARGET_COLDFIRE" to choose between 68000 and
3799 non-68000 timings. Refactor multiplication and division costs.
3800 (output_addsi3): Use TUNE_68040 instead of TARGET_68040 and
3801 TUNE_CPU32 instead of TARGET_CPU32.
3802 (standard_68881_constant_p): Use TUNE_68040 instead of TARGET_68040
3803 and TUNE_68060 instead of TARGET_68060.
3804 * config/m68k/m68k.md: Use TUNE_68040 instead of TARGET_68040,
3805 TUNE_68060 instead of TARGET_68060, and TUNE_CPU32 instead of
3807 (movsi_const0): Use TUNE_68000_10 rather than "!TARGET_68020
3808 && !TARGET_COLDFIRE" to choose between moveq and clr.
3809 Likewise in the unnamed movsf pattern.
3810 (ashlsi_17_24, lshrsi_17_24): Guard with TUNE_68000_10 rather than
3811 "!TARGET_68020 && !TARGET_COLDFIRE". Likewise the unnamed
3814 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3816 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Increase amount
3817 of tabbing before backslashes.
3819 2007-01-11 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
3821 * pa-linux.h (ASM_OUTPUT_INTERNAL_LABEL): Undefine.
3822 * pa.h (ASM_OUTPUT_LABEL): Output colon when using GAS.
3823 (ASM_OUTPUT_INTERNAL_LABEL): Define.
3825 2007-01-11 Zdenek Dvorak <dvorakz@suse.cz>
3827 * tree-ssa-loop-ivopts.c (extract_cond_operands): Split from
3828 find_interesting_uses_cond.
3829 (find_interesting_uses_cond): Use extract_cond_operands.
3830 (rewrite_use_compare): Use extract_cond_operands and
3831 force_gimple_operand_bsi. Do not call update_stmt.
3832 (determine_use_iv_cost_condition): Use extract_cond_operands.
3833 Return cheaper of using original bound and changing the exit bound.
3835 2007-01-11 Zdenek Dvorak <dvorakz@suse.cz>
3837 PR tree-optimization/29516
3838 * tree-ssa-address.c (tree_mem_ref_addr, add_to_parts,
3839 most_expensive_mult_to_index, addr_to_parts,
3840 create_mem_ref, maybe_fold_tmr): Make the type of
3841 fields of TARGET_MEM_REF sizetype.
3842 (move_fixed_address_to_symbol, move_pointer_to_base):
3844 * tree.def (TARGET_MEM_REF): Add comment on types of
3847 2007-01-11 Joseph Myers <joseph@codesourcery.com>
3849 * c-common.c (vector_types_convertible_p): Treat opaque types as
3850 always convertible if they have the same size, but not otherwise.
3852 2007-01-11 Steven Bosscher <steven@gcc.gnu.org>
3854 * ifcvt.c (struct noce_if_info): Add comments to the fields.
3855 Remove the b_unconditional field.
3856 (noce_try_sign_mask): Do not look at b_unconditional.
3857 (noce_process_if_block): Do not use merge_if_blocks. Update
3858 the CFG here. Do not set b_unconditional.
3859 (cond_move_process_if_block): Likewise.
3860 (find_cond_trap): Likewise.
3861 (check_cond_move_block): Require simple jump insns at the end
3864 2007-01-11 Jan Hubicka <jh@suse.cz>
3866 PR tree-optimization/1046
3867 * tree-tailcall.c (suitable_for_tail_call_opt_p): Use TREE_ADDRESSABLE
3868 when alias info is not ready.
3869 (pass_tail_recursion): Do not require aliasing.
3870 * tree-ssa-copyrename.c (pass_rename_ssa_cop): Likewise.
3871 * tree-ssa-ccp.c (pass_ccp, pass_fold_builtins): Likewise.
3872 * tree-ssa-copy.c (pass_copy_prop): Likewise.
3873 * tree-ssa-forwprop.c (pass_forwprop): Likewise.
3874 * tree-ssa-dce.c (pass_dce, pass_dce_loop, pass_cd_dce): Likewise.
3875 * passes.c (init_optimization_passes): Execute rename_ssa_copies,
3876 ccp, forwprop, copy_prop, merge_phi, copy_prop, dce and tail recursion
3878 * tree-ssa-operands.c (add_virtual_operand, get_indirect_ref_operand):
3879 When aliasing is not build, mark statement as volatile.
3881 2007-01-11 Tom Tromey <tromey@redhat.com>
3883 PR preprocessor/15185, PR preprocessor/20989:
3884 * doc/cppopts.texi <-MT>: Update description of algorithm for
3885 computing default target.
3886 <-M, -MD>: Reword "basename" text.
3888 2007-01-11 Roger Sayle <roger@eyesopen.com>
3890 * builtins.c (expand_builtin_pow, expand_builtin_powi,
3891 fold_builtin_cabs, fold_builtin_sqrt, fold_builtin_trunc,
3892 fold_builtin_floor, fold_builtin_ceil, fold_builtin_round,
3893 fold_builtin_int_int_roundingfn, fold_builtin_bitop,
3894 fold_builtin_bswap, real_constp, fold_builtin_pow,
3895 fold_builtin_powi, fold_builtin_signbit, fold_builtin_copysign,
3896 do_mpfr_arg1, do_mpfr_arg2, do_mpfr_arg3, do_mpfr_sincos): Replace
3897 uses of the macro TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
3898 * convert.c (convert_to_pointer): Likewise.
3899 * expr.c (highest_pow2_factor, expand_expr_real_1): Likewise.
3900 * fold-const.c (force_fit_type, fold_negate_expr, int_const_binop,
3901 const_binop, fold_convert_const_int_from_int,
3902 fold_convert_const_int_from_real,
3903 fold_convert_const_real_from_real, sign_bit_p,
3904 optimize_minmax_comparison, extract_muldiv_1, fold_div_compare,
3905 fold_sign_changed_comparison, fold_unary, fold_comparison,
3906 fold_binary, multiple_of_p, tree_Expr_non_zero_p,
3907 fold_negate_const, fold_abs_const, fold_not_const): Likewise.
3908 * print-tree.c (print_node_brief, print_node): Likewise.
3909 * stor-layout.c (place_field, layout_type): Likewise.
3910 * tree-chrec.c (keep_cast): Likewise.
3911 * tree.c (build_vector, build_real, build_real_from_int_cst,
3912 build_complex): Likewise.
3914 2007-01-11 Roger Sayle <roger@eyesopen.com>
3916 * tree.h (TREE_CONSTANT_OVERFLOW): Obsolete. For the time being,
3917 treat TREE_CONSTANT_OVERFLOW as a synonym of TREE_OVERFLOW.
3919 2007-01-11 Paolo Bonzini <bonzini@gnu.org>
3921 * configure.ac (strict1_warn): Rename to strict_warn.
3922 (WERROR, --enable-werror, symlink hacks, stage1_cflags,
3923 cc_set_by_configure, quoted_cc_set_by_configure,
3924 stage_prefix_set_by_configure, quoted_stage_prefix_set_by_configure,
3925 all_boot_languages, all_stagestuff): Remove.
3926 (target_list): Remove bootstrap targets.
3927 * Makefile.in (quickstrap): Unconditionally make a synonym of all.
3928 (BOOT_LANGUAGES, STAGE1_CFLAGS, STAGE1_CHECKING,
3929 REMAKEFLAGS, FLAGS_TO_PASS, PREPEND_DOTDOT_TO_RELATIVE_PATHS,
3930 SUBDIR_FLAGS_TO_PASS, WERROR_FLAGS, STRICT2_WARN, LANG_STAGESTUFF,
3931 VOL_FILES, POSTSTAGE1_FLAGS_TO_PASS, STAGE2_FLAGS_TO_PASS,
3932 STAGEPROFILE_FLAGS_TO_PASS, STAGEFEEDBACK_FLAGS_TO_PASS, stage1_build,
3933 stage1_copy, stage2_build, stage2_copy, stageprofile_build,
3934 stageprofile_copy, stage3_build, stage3_copy, stagefeedback_build,
3935 stagefeedback_copy, stage4_build, clean_s1, clean_sw, bootstrap,
3936 bootstrap-lean, bootstrap2, bootstrap2-lean, bootstrap3,
3937 bootstrap3-lean, bootstrap4, bootstrap4-lean, unstage1, unstage2,
3938 unstage3, unstage4, unstageprofile, unstagefeedback, restage, restage2,
3939 restage3, restage4, restageprofile, restagefeedback, bubbleestrap,
3940 cleanstrap, unstrap, restrap, *compare, *compare3, *compare4,
3941 *compare-lean, *compare3-lean, *compare4-lean, stage1-start, stage1,
3942 stage2-start, stage2, stage3-start, stage3, stage4-start, stage4,
3943 stageprofile-start, stageprofile, stagefeedback-start, stagefeedback,
3944 risky-stage1, risky-stage2, risky-stage3, risky-stage4): Remove.
3945 (ORDINARY_FLAGS_TO_PASS): Rename to FLAGS_TO_PASS.
3946 (STAGECOPYSTUFF, STAGEMOVESTUFF): Consolidate into MOSTLYCLEANFILES.
3947 (mostlyclean): Adjust.
3948 (clean, distclean): Don't mention bootstrap stuff.
3949 * configure: Regenerate.
3950 * ada/config-lang.in, cp/config-lang.in, forttran/config-lang.in,
3951 java/config-lang.in, objc/config-lang.in, objcp/config-lang.in,
3952 treelang/config-lang.in (stagestuff): Remove.
3953 * doc/sourcebuild.texi (stage1, stage2, stage3, stage4,
3954 stageprofile, stagefeedback, stagestuff): Remove mention.
3956 2007-01-11 Nick Clifton <nickc@redhat.com>
3958 * config/mcore/predicates.md (mcore_general_movesrc_operand):
3960 (mcore_general_movdst_operand): Do not accept CONST_INTs.
3961 (mcore_arith_K_S_operand): Run the test for the S constraint not
3962 the test for the M constraint.
3963 (mcore_addsub_operand): Do not accept integer values that are
3964 larger than 32 bits.
3965 * config/mcore/mcore.md: Remove unused constraints from split.
3966 (andsi3): Use HOST_WIDE_INT instead of int to hold an INTVAL.
3968 (allocate_stack): Likewise.
3969 * config/mcore/mcore.c (mcore_print_operand): Restrict output of P
3970 operands to 32 bits.
3971 (mcore_const_costs): Use HOST_WIDE_INT instead of int to hold an
3973 (mcore_and_cost, mcore_modify_comparison, const_ok_for_mcore,
3974 mcore_const_ok_for_inline, mcore_const_trick_uses_not,
3975 try_constant_tricks, mcore_num_ones, mcore_num_zeros,
3976 mcore_output_bclri, mcore_output_andn, output_inline_const,
3977 mcore_output_move, mcore_output_movedouble): Likewise.
3978 (mcore_output_cmov): Use CONST_OK_FOR_M and CONST_OK_FOR_N.
3979 (output_inline_const): Likewise.
3980 (output_inline_const): Fix format strings used in sprintf
3982 * config/mcore/mcore-protos.h: Update prototypes for changed
3983 functions in mcore.c.
3984 * config/mcore/mcore.h (CONST_OK_FOR_I): Cast values to
3985 HOST_WIDE_INT and not int.
3986 (CONST_OK_FOR_J, CONST_OK_FOR_K, CONST_OK_FOR_L, CONST_OK_FOR_M,
3987 CONST_OK_FOR_N): Likewise.
3988 (LEGITIMATE_CONSTANT_P): Also check CONSTANT_P.
3989 (GO_IF_LEGITIMATE_INDEX): Use HOST_WIDE_INT instead of int to hold
3992 2007-01-10 Jan Hubicka <jh@suse.cz>
3994 * tree-vrp.c (remove_range_assertions): Release defs.
3995 * tree-ssa-loop-ivopts.c (rmeove_statement): Likewise.
3996 * tree-ssa-dom.c (remove_stmt_or_phi): Likewise.
3998 2007-01-10 Paul Brook <paul@codesourcery.com>
4000 * config/arm/arm.c (arm_rtx_costs_1): Handle mutiply-subtract.
4001 * config/arm/arm.md (mulsi3subsi): New insn.
4003 2007-01-10 Zdenek Dvorak <dvorakz@suse.cz>
4005 * tree-ssa-loop-manip.c (tree_unroll_loop): Make it a wrapper over ...
4006 (tree_transform_and_unroll_loop): New.
4007 * tree-flow.h (transform_callback, tree_transform_and_unroll_loop):
4010 2007-01-10 Robert Kennedy <jimbob@google.com>
4012 * fold-const.c (fold_comparison): Fold comparisons like (x *
4013 1000 < 0) to (x < 0).
4015 2007-01-10 Ian Lance Taylor <iant@google.com>
4017 * tree-pretty-print.c (dump_generic_node): Print parentheses when
4018 operands have the same priority.
4020 2007-01-10 Tom Tromey <tromey@redhat.com>
4022 * fold-const.c (fold_truthop): Don't check can_use_bit_fields_p.
4023 (fold_binary): Likewise.
4024 * langhooks.c (lhd_can_use_bit_fields_p): Removed.
4025 * langhooks-def.h (lhd_can_use_bit_fields_p): Removed.
4026 (LANG_HOOKS_CAN_USE_BIT_FIELDS_P): Removed.
4027 (LANG_HOOKS_INITIALIZER): Remove LANG_HOOKS_CAN_USE_BIT_FIELDS_P.
4028 * langhooks.h (struct lang_hooks): Removed field
4029 'can_use_bit_fields_p'.
4031 2007-01-10 Ralf Corsépius <ralf.corsepius@rtems.org>
4033 * config/bfin/t-bfin, config/bfin/t-bfin-elf: Remove GCC_CFLAGS.
4035 2007-01-10 Razya Ladelsky <razya@il.ibm.com>
4037 * function.c (get_last_funcdef_no): New function.
4038 * function.h (get_last_funcdef_no): Declare.
4039 * tree-inline.c (initialize_cfun): Add initialization.
4040 (tree_function_versioning): Cleanup.
4042 2007-01-10 Jan Hubicka <jh@suse.cz>
4044 * tree-inline.c (setup_one_parameter): Do not propagate into abnormal
4047 2007-01-10 Sa Liu <saliu@de.ibm.com>
4048 Ben Elliston <bje@au.ibm.com>
4050 * spu.h (STACK_SAVE_AREA): Use VOIDmode for SAVE_FUNCTION, SImode
4051 for SAVE_NONLOCAL and Pmode for any other save level.
4052 * spu-protos.h (spu_restore_stack_block): Declare.
4053 * spu.md (save_stack_block): Remove.
4054 (restore_stack_block): Call spu_restore_stack_block.
4055 * spu.c (spu_restore_stack_block): New function.
4056 (spu_expand_epilogue): Remove old comment.
4058 2007-01-09 Zdenek Dvorak <dvorakz@suse.cz>
4060 PR tree-optimization/30322
4061 * tree-ssa-loop-ivopts.c (fold_affine_expr, iv_value): Removed.
4062 (cand_value_at): Return the value as aff_tree.
4063 (may_eliminate_iv): Convert the bound from aff_tree to tree.
4064 * tree-affine.c (aff_combination_add_cst, aff_combination_add_product,
4065 aff_combination_mult): New functions.
4066 (aff_combination_add): Use aff_combination_add_cst.
4067 (aff_combination_convert): Allow conversions to a wider type.
4068 (tree_to_aff_combination): Handle BIT_NOT_EXPR.
4069 * tree-affine.h (aff_combination_mult): Declare.
4071 2007-01-09 Carlos O'Donell <carlos@codesourcery.com>
4073 * doc/tm.texi: Update documentation to reflect reality of exec
4074 and start file search behaviours. Update copyright year.
4075 * doc/invoke.texi: Explain how GCC_EXEC_PREFIX is used to find
4076 header file directories.
4078 2007-01-09 Uros Bizjak <ubizjak@gmail.com>
4080 * config/i386/i386.md (*sinxf2): Rename to *sinxf2_i387.
4081 (*cosxf2): Rename to cosxf2_i387.
4082 (*sindf2, *sinsf2): Extend operand 1 to XFmode. Macroize patterns
4083 using X87MODEF12 mode macro. Rename patterns to
4084 *sin_extend<mode>xf2_i387. Use SSE_FLOAT_MODE_P to disable patterns
4086 (*cosdf2, *cossf2): Ditto.
4087 (sincosdf3, sincossf3): Ditto. Rewrite corresponding splitters
4088 to match extended input operands.
4089 (sincos<mode>3): New expander.
4090 (*sinextendsfdf2, *cosextendsfdf2, *sincosextendsfdf3): Remove
4091 insn patterns and corresponding splitters.
4093 2007-01-09 Kaz Kojima <kkojima@gcc.gnu.org>
4095 * config/sh/t-linux (TARGET_LIBGCC2_CFLAGS): Delete.
4096 (SHLIB_MAPFILES, SHLIB_LINK, SHLIB_INSTALL): Likewise.
4098 2007-01-09 Nicolas Pitre <nico@cam.org>
4101 * arm/ieee754-df.S (Lad_s): Also test the low word of X for zero.
4103 2007-01-08 Geoffrey Keating <geoffk@apple.com>
4105 * target.h (struct gcc_target): New field library_rtti_comdat.
4106 * target-def.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): New.
4107 (TARGET_CXX): Add TARGET_CXX_LIBRARY_RTTI_COMDAT.
4108 * doc/tm.texi (C++ ABI): Document TARGET_CXX_LIBRARY_RTTI_COMDAT.
4109 * config/darwin.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
4111 2007-01-08 Geoffrey Keating <geoffk@apple.com>
4113 * doc/invoke.texi (Optimize Options): Correct description of -O0.
4115 2007-01-08 Richard Guenther <rguenther@suse.de>
4117 * tree.h (force_fit_type_double): Export.
4118 (force_fit_type): Remove.
4119 * fold-const.c (force_fit_type_double): New function.
4120 (force_fit_type): Remove.
4121 (int_const_binop): Use it.
4122 (fold_convert_const_int_from_int): Likewise.
4123 (fold_convert_const_int_from_real): Likewise.
4124 (fold_div_compare): Likewise.
4125 (fold_sign_changed_comparison): Likewise.
4126 (fold_unary): Likewise.
4127 (fold_negate_const): Likewise.
4128 (fold_abs_const): Likewise.
4129 (fold_not_const): Likewise.
4130 * c-common.c (shorten_compare): Use force_fit_type_double.
4131 * convert.c (convert_to_pointer): Likewise.
4133 2007-01-08 Richard Guenther <rguenther@suse.de>
4135 * tree.h (build_int_cst_wide_type): Export.
4136 * tree.c (build_int_cst_wide_type): New function.
4137 (build_int_cst_wide): Fix comment.
4138 * builtins.c (fold_builtin_object_size): Use build_int_cst
4139 to build -1 or 0 of the correct type. Use fit_double_type
4140 to check for overflow.
4141 * fold-const.c (optimize_bit_field_compare): Use build_int_cst_type
4143 (decode_field_reference): Likewise.
4144 (all_ones_mask_p): Likewise.
4145 (native_interpret_int): Use build_int_cst_wide_type.
4146 (fold_binary): Use build_int_cst_type to build an all-ones
4148 * stor-layout.c (set_sizetype): Use build_int_cst_wide_type.
4150 2007-01-08 Daniel Jacobowitz <dan@codesourcery.com>
4152 * config/pa/t-pa64 (libgcc_stub.a): Use $(T).
4154 2007-01-09 Ben Elliston <bje@au.ibm.com>
4156 * genautomata.c (STATS_OPTION): New option.
4157 (stats_flag): New flag.
4158 (gen_automata_option): Handle it.
4159 (initiate_automaton_gen): Ditto.
4160 (write_automata): Output statistics only if stats_flag is
4161 set. Likewise, output time statistics only if time_flag is set.
4162 * doc/md.texi (Processor pipeline description): Document new flag.
4164 2007-01-08 Richard Guenther <rguenther@suse.de>
4166 * builtins.c (fold_builtin_int_roundingfn): Use fit_double_type.
4167 * tree.c (build_int_cst_type): Likewise.
4168 (size_in_bytes): Don't call force_fit_type on the result.
4169 (int_fits_type_p): Use fit_double_type.
4170 * fold-const.c (fit_double_type): New function.
4171 (force_fit_type): Use it.
4172 * tree.h (fit_double_type): Export.
4174 2007-01-08 Jan Hubicka <jh@suse.cz>
4176 * tree-vectorizer.c (gate_increase_alignment): Fix return type.
4177 * ipa.c (function_and_variable_visibility): Fix return type.
4179 2007-01-08 Richard Guenther <rguenther@suse.de>
4181 * tree-ssa-ccp.c (maybe_fold_offset_to_array_ref): Use type
4182 of offset to build the index.
4183 * tree-pretty-print.c (dump_generic_node): Don't build negated
4184 const just for printing.
4185 * c-pretty-print.c (pp_c_integer_constant): Likewise.
4186 * builtins.c (fold_builtin_int_roundingfn): Check if result
4187 fits the type by using force_fit_type and comparing the result.
4188 * predict.c (predict_loops): Use compare_tree_int for comparison.
4189 * tree.c (build_int_cst): Fall back to integer_type_node for
4191 (build_int_cst_wide): Assert type is non-null.
4193 2007-01-08 Roberto Costa <roberto.costa@st.com>
4195 * tree-vrp.c (extract_range_from_cond_expr): New.
4196 (extract_range_from_expr): Handle COND_EXPR nodes used as expressions.
4197 * tree-ssa-ccp.c (get_maxval_strlen): Handle COND_EXPR nodes used
4199 (fold_stmt): Bug fix, avoid infinite recursion when folding COND_EXPRs.
4200 * tree-ssa-forwprop.c (simplify_cond, forward_propagate_into_cond,
4201 tree_ssa_forward_propagate_single_use_vars): Handle COND_EXPR nodes
4202 used as expressions.
4203 * tree-object-size.c (cond_expr_object_size): New.
4204 (collect_object_sizes_for): Handle COND_EXPR nodes used as expressions.
4206 2007-01-08 Jan Hubicka <jh@suse.cz>
4208 * tree-ssa-forwprop.c (forward_propagate_into_cond,
4209 tree_ssa_forward_propagate_single_use_va): Release defs of propagated
4212 2007-01-08 Richard Guenther <rguenther@suse.de>
4214 PR tree-optimization/23603
4215 * tree-vrp.c (set_value_range_to_truthvalue): New function.
4216 (extract_range_from_binary): Fall back to truthvalue instead of
4217 varying for TRUTH_*_EXPR.
4218 (extract_range_from_comparison): Fall back to truthvalue instead of
4220 (vrp_visit_phi_node): Don't adjust new range bounds to +INF/-INF
4221 if all visited PHI values were constant.
4223 2007-01-08 Jan Hubicka <jh@suse.cz>
4225 * cgraphunit.c (cgraph_process_new_functions): Reset reachable flag.
4226 (cgraph_analyze_function): break out from ...
4227 (cgraph_finalize_compilation_unit): ... here.
4228 (cgraph_expand_function): Remove forgoten commented out line.
4229 (cgraph_optimize): Analyze functions.
4231 2007-01-08 Jan Hubicka <jh@suse.cz>
4233 * tree-pas.h (TODO_remove_function): New flag.
4234 (TODO_update*): Renumber.
4235 (pass_ipa_increase_alignment,
4236 pass_ipa_function_and_variable_visibility): New passes.
4237 * cgraphunit.c (cgraph_increase_alignment): Move to tree-vectorizer.c
4238 (cgraph_function_and_variable_visibility): Move to ipa.c
4239 (cgraph_optimize): Don't call cgraph_function_and_variable_visibility,
4240 cgraph_increase_alignment.
4241 * ipa-inline.c (cgraph_decide_inlining): Don't push timevar.
4242 (cgraph_decide_inlining_incrementally): Push TV_INTEGRATION before
4243 calling tree-inline.
4244 (cgraph_early_inlining): Do not call cgraph_remove_unreachable_nodes.
4245 (pass_ipa_inline, pass_early_ipa_inlining): Set TODO_remove_functions
4246 * tree-vectorizer.c (increase_alignment): Move here from cgraphunit.c
4247 (gate_increase_alignment): New function.
4248 (pass_ipa_increase_alignment): New pass.
4249 * ipa.c: Inline tree-pass.h and timevar.h
4250 (function_and_variable_visibility): Move here from cgraphunit.c
4251 * tree-optimize.c (pass_early_local_passes): Add TODO_remove_functions.
4252 * passes.c (init_optimization_passes): Add the two new passes.
4253 (execute_todo): Handle cgraph_remove_functions.
4255 2007-01-08 Nick Clifton <nickc@redhat.com>
4257 * config/frv/predicates.md (reg_or_0_operand): Accept
4260 2007-01-08 Ralf Corsépius <ralf.corsepius@rtems.org>
4262 * config/bfin/rtems.h, config/bfin/t-rtems: New.
4263 * config.gcc: Add bfin*-rtems*.
4265 2007-01-08 Mark Shinwell <shinwell@codesourcery.com>
4267 * c.opt: Add -flax-vector-conversions.
4268 * c-typeck.c (convert_for_assignment): Pass flag to
4269 vector_types_convertible_p to allow emission of note.
4270 (digest_init): Likewise.
4271 * c-opts.c: Handle -flax-vector-conversions.
4272 * c-common.c (flag_lax_vector_conversions): New.
4273 (vector_types_convertible_p): Unless -flax-vector conversions
4274 has been passed, disallow conversions between vectors with
4275 differing numbers of subparts and/or element types. If such
4276 a conversion is disallowed, possibly emit a note on the first
4277 occasion only to inform the user of -flax-vector-conversions.
4278 The new last argument specifies this.
4279 * c-common.h (flag_lax_vector_conversions): New.
4280 (vector_types_convertible_p): Add extra argument.
4281 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Use
4282 char_type_node for V*QI type vectors.
4283 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins):
4284 Update to satisfy new typechecking rules.
4285 * config/rs6000/altivec.h (vec_cmple): Use vec_cmpge, for both
4287 * doc/invoke.texi (C Dialect Options): Document
4288 -flax-vector-conversions.
4290 2007-01-08 Mark Shinwell <shinwell@codesourcery.com>
4292 PR tree-optimization/29877
4293 * tree-ssa-ter.c (is_replaceable_p): Deem assignments with
4294 a register variable on the RHS to not be replaceable.
4296 2007-01-08 Chen Liqin <liqin@sunnorth.com.cn>
4297 * config/score/t-score-elf (MULTILIB_OPTIONS): Change.
4298 * config/score/predicates.md (const_uimm5, sr0_operand, const_simm12,
4299 const_simm15, const_pow2, const_npow2): Added.
4300 * config/score/misc.md (insv, extv, extzv, movmemsi,
4301 move_lbu_a/b, mov_lhu_a/b etc): Added and fix some bug.
4302 * config/score/score.c (score_address_cost, score_select_cc_mode):
4304 Change CONST_OK_FOR_LETTER_P/EXTRA_CONSTRAINT define.
4305 Update score_rtx_costs for MACRO TARGET_RTX_COSTS.
4306 Update score_print_operand.
4307 * config/score/score.h (DATA_ALIGNMENT, SELECT_CC_MODE): Added.
4308 Adjust register allocate order and update some macro define.
4309 * config/score/score-mdaux.c (mdx_unaligned_load, mdx_unsigned_store,
4310 mdx_block_move_straight, mdx_block_move_loop_head,
4311 mdx_block_move_loop_body, mdx_block_move_loop_foot, mdx_block_move_loop,
4312 mdx_block_move): Added.
4313 (mdx_movsicc, mdp_select_add_imm, mdp_select, mds_zero_extract_andi,
4314 mdp_limm): Updated and fix some bug and typo.
4315 * config/score/score.md (movqi/hi/si, add/sub/zero/ext): Updated.
4316 (movsf, movdf, doloop_end): Added.
4318 2007-01-08 Kazu Hirata <kazu@codesourcery.com>
4320 * config/arm/arm.c, config/arm/arm.h, config/arm/arm.md,
4321 config/arm/thumb2.md: Fix comment typos.
4322 * doc/extend.texi: Fix a typo.
4324 2007-01-07 Eric Christopher <echristo@apple.com>
4326 * configure.ac: Check for __stack_chk_fail for darwin.
4327 * configure: Regenerate.
4329 2007-01-07 Richard Guenther <rguenther@suse.de>
4331 * tree-vrp.c (extract_range_from_assert): CSE calls to
4332 compare_values where possible.
4333 (extract_range_from_unary_expr): Likewise.
4335 2007-01-07 Anatoly Sokolov <aesok@post.ru>
4337 * config/avr/avr-protos.h (call_insn_operand): Delete prototype.
4338 * config/avr/avr.c (call_insn_operand): Delete function.
4339 * config/avr/avr.md (*pushqi, *pushhi, *pushsi, *pushsf): Use REG_SP
4340 instead of register number. Use predicates.
4341 * config/avr/predicates.md (const0_operand, reg_or_0_operand,
4342 call_insn_operand): Add.
4344 2007-01-06 Jan Hubicka <jh@suse.cz>
4346 * tree-pass.h (pass_build_cgraph_edges): Declare.
4347 * cgraphunit.c (record_refernece): Move to cgraphbuild.c
4348 (visited_nodes): Remove.
4349 (cgraph_create_edges): Move to cgraphbuild.c; rename to
4350 build_cgrpah_edges; make visited_nodes local.
4351 (cgraph_process_new_functions): DO not call initialize_inline_failed.
4352 (record_references_in_initializer): Move to cgraphbuild.c
4353 (initialize_inline_failed, rebuild_cgraph_edges,
4354 pass_rebuild_cgraph_edges): Move to cgraphbuild.c.
4355 (verify_cgraph_node): Make visited_nodes local.
4356 (cgraph_analyze_function): Do not call cgraph_create_edges and
4357 initialize_inline_failed.
4358 (cgraph_expand_function): Do not call cgraph_lower_function;
4359 assert that function is already lowered.
4360 * Makefile.in (cgraphbuild.o): New.
4361 * passes.c (init_optimization_passes): Add pass_build_cgraph_edges
4362 at the end of lowering passes.
4364 2007-01-06 Steven Bosscher <steven@gcc.gnu.org>
4366 * ifcvt.c (cond_move_convert_if_block): New function, code
4367 factored out from...
4368 (cond_move_process_if_block): ...here. Call the new function
4369 on the THEN and ELSE blocks.
4370 (merge_if_block): Do not copy global_live_at_end, merge_blocks
4371 already takes care of this.
4373 2007-01-05 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
4376 * tree.h (TREE_OVERFLOW_P): New.
4377 * c-typeck.c (parser_build_unary_op): Warn only if result
4378 overflowed and operands did not.
4379 (parser_build_binary_op): Likewise.
4380 (convert_for_assignment): Remove redundant overflow_warning.
4381 * c-common.c (overflow_warning): Don't check or set TREE_OVERFLOW.
4383 2007-01-05 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
4385 * c-typeck.c (store_init_value): Split over two lines to follow
4386 the GNU coding style.
4388 2007-01-05 Benjamin Kosnik <bkoz@redhat.com>
4390 * c-cppbuiltin.c (c_cpp_builtins): __GXX_EXPERIMENTAL_CPP0X__ to
4391 __GXX_EXPERIMENTAL_CXX0X__.
4392 * doc/cpp.texi: Same.
4394 2007-01-05 Richard Guenther <rguenther@suse.de>
4397 * tree.c (get_narrower): Do not construct COMPONENT_REFs
4398 with mismatched types. Instead explicitly build a
4399 conversion NOP_EXPR.
4401 2007-01-05 Ian Lance Taylor <iant@google.com>
4403 * c-common.c (decl_with_nonnull_addr_p): New function.
4404 (c_common_truthvalue_conversion): Call it.
4405 * c-typeck.c (build_binary_op): Likewise.
4406 * c-common.h (decl_with_nonnull_addr_p): Declare.
4408 2007-01-05 Jakub Jelinek <jakub@redhat.com>
4411 * libgcc2.c (__divdc3): Compare c and d against 0.0 instead of
4414 2007-01-05 Joel Brobecker <brobecker@adacore.com>
4416 * doc/install.texi (Final install): Document the fact that
4417 the GNAT runtime should not be stripped.
4419 2007-01-04 Jan Hubicka <jh@suse.cz>
4421 * tree-inline.c (fold_marked_statements): Update operand caches
4422 and EH after folding
4424 2007-01-04 Ian Lance Taylor <iant@google.com>
4426 * c-common.c (check_function_nonnull): Whitespace fix.
4428 2007-01-04 Jan Hubicka <jh@suse.cz>
4430 * tree-optimize.c (execute_fixup_cfg): Correct previously mistakely
4431 comitted older version of patch.
4432 (pass_fixup_cfg): Add TODOs to verify flow and statements, dump
4433 function, celanup cfg and collect garbage.
4435 2007-01-04 Mike Stump <mrs@apple.com>
4437 * Makefile.in (mostlyclean): Don't remove libgcc anymore.
4440 2007-01-04 Eric Christopher <echristo@apple.com>
4442 * libgcc2.c (__bswapsi2): Use SItype.
4443 (__bswapdi2): Use DItype.
4444 * libgcc2.h: Update for above.
4446 2007-01-04 Paul Brook <paul@codesourcery.com>
4448 * config/arm/arm.md (arm_mulsi3, thumb_mulsi3, mulsi3_compare0,
4449 mulsi_compare0_scratch, mulsi3addsi, mulsi3addsi_compare0,
4450 mulsi3addsi_compare0_scratch, mulsidi3adddi, mulsidi3,
4451 umulsidi3, umulsidi3adddi, smulsi3_highpart,
4452 umulsi3_highpart): Make conditional on !arm_arch6.
4453 (arm_mulsi3_v6, thumb_mulsi3_v6, mulsi3_compare0_v6,
4454 mulsi_compare0_scratch_v6, mulsi3addsi_v6, mulsi3addsi_compare0_v6,
4455 mulsi3addsi_compare0_scratch_v6, mulsidi3adddi_v6, mulsidi3_v6,
4456 umulsidi3_v6, umulsidi3adddi_v6, smulsi3_highpart_v6,
4457 umulsi3_highpart_v6): New insns.
4459 2007-01-04 Roger Sayle <roger@eyesopen.com>
4461 * fold-const.c (fold_convert): When casting an expression to void,
4462 fold_ignored_result may discover a GIMPLE_MODIFY_STMT which doesn't
4463 have a type. Instead of attempting to build a NOP_EXPR, return
4464 these "special" trees directly.
4466 2007-01-04 Joseph Myers <joseph@codesourcery.com>
4468 * config/rs6000/rs6000.c (rs6000_rtx_costs): Make adjustment for
4469 MULT inside MINUS as either argument. Use rs6000_cost->dmul -
4470 rs6000_cost->fp not 0 as adjustment for outer NEG.
4472 2007-01-04 Jan Hubicka <jh@suse.cz>
4474 * cgraph.c (cgraph_release_function_body): New function.
4475 (cgraph_remove_node): Use it.
4476 * cgraph.h (cgraph_release_function_body): Declare.
4477 * cgraphunit.c (cgraph_expand_function): Use it.
4478 * ipa.c (cgraph_remove_unreahchable_nodes): Use it.
4479 * tree-ssa.c (delete_tree_ssa): Allow to be called before aliasing
4480 is initialized and while compilation of other function is running.
4481 * tree-optimize.c (execute_free_cfg_annotations): Move code to clear
4482 statement CFG annotations from here to ...
4483 * tree-cfg.c (delete_tree_cfg_annotations): ... here.
4485 2007-01-04 Zdenek Dvorak <dvorakz@suse.cz>
4487 * cfgloop.h (enum li_flags): Make the constants powers of two.
4489 2007-01-04 Jan Hubicka <jh@suse.cz>
4491 * tree-inline.c (copy_bb): Insert new statements to statements_to_fold
4493 (fold_marked_statements): New function.
4494 (optimize_inline_calls, tree_function_versioning): Fold new statements.
4495 * tree-inline.h (copy_body_data): Add statements_to_fold.
4497 2007-01-03 Daniel Jacobowitz <dan@codesourcery.com>
4499 * config.gcc: Mention libgcc/config.host.
4500 * Makefile.in: Update comments mentioning libgcc.
4501 (LIBGCC, INSTALL_LIBGCC, GCC_PARTS, mklibgcc): Delete.
4502 (all.cross, start.encap, rest.encap, rest.cross): Update
4503 dependencies for libgcc move.
4504 (libgcc.mk, LIBGCC_DEPS, libgcov.a, libgcc.a, stmp-multilib)
4505 (clean-target, clean-target-libgcc): Delete.
4506 (srcdirify, GCC_EXTRA_PARTS): New macros.
4507 (libgcc-support, libgcc.mvars): New rules.
4508 (distclean): Remove mention of mklibgcc.
4509 (install): Don't reference INSTALL_LIBGCC.
4510 (install-common): Don't reference EXTRA_PARTS.
4511 (install-libgcc, install-multilib): Delete rules.
4512 * mklibgcc.in: Delete file.
4513 * doc/configfiles.texi: Don't mention mklibgcc.
4515 * config/i386/t-darwin (SHLIB_VERPFX): Delete (moved to libgcc).
4516 * config/i386/t-darwin64 (SHLIB_VERPFX): Likewise.
4517 * config/rs6000/t-darwin (SHLIB_VERPFX): Likewise.
4518 * config/rs6000/t-ppccomm (TARGET_LIBGCC2_CFLAGS, SHLIB_MAPFILES)
4519 (mklibgcc, ldblspecs): Likewise.
4521 * config/i386/t-nwld (libgcc.def, libc.def, libpcre.def)
4522 (posixpre.def): Use $(T).
4523 (SHLIB_EXT, SHLIB_NAME, SHLIB_SLIBDIR_QUAL, SHLIB_DEF, SHLIB_MAP)
4524 (SHLIB_SRC, SHLIB_INSTALL): Delete.
4525 (SHLIB_LINK): Make dummy.
4526 * config/t-slibgcc-darwin: Delete contents except for dummy SHLIB_LINK.
4528 * config/frv/t-linux (EXTRA_MULTILIB_PARTS): Clear.
4530 * config/alpha/t-crtfm: Use $(T) in rules for EXTRA_PARTS.
4531 * config/alpha/t-vms, config/alpha/t-vms64, config/fr30/t-fr30,
4532 config/i386/t-rtems-i386, config/ia64/t-ia64, config/rs6000/t-beos,
4533 config/rs6000/t-newas, config/sparc/t-elf: Likewise.
4535 * configure.ac (all_outputs): Remove mklibgcc.
4536 * configure: Regenerated.
4538 2007-01-03 Josh Conner <jconner@apple.com>
4541 * calls.c (compute_argument_addresses): Set stack and stack_slot
4542 for partial args, too.
4543 (store_one_arg): Use locate.size.constant for the size when
4544 generating a save_area.
4546 2007-01-03 Robert Kennedy <jimbob@google.com>
4548 * tree-cfg.c (tree_merge_blocks): Release SSA_NAME phi results
4549 whose definitions are deleted due to basic block merging.
4551 2007-01-03 Paul Brook <paul@codesourcery.com>
4554 * config/arm/arm.c (output_return_instruction): Pop PC in interrupt
4556 (use_return_insn): Return 0 for Thumb interrupt functions.
4557 (print_multi_reg): Add rfe argument for IRQ returns.
4558 (arm_output_epilogue): Pop interrupt return address directly into PC.
4559 (arm_expand_prologue): Only adjust IRQ return address in Arm mode.
4561 2007-01-03 Paul Brook <paul@codesourcery.com>
4563 Merge from sourcerygxx-4_1.
4564 * config/arm/thumb2.md: New file.
4565 * config/arm/elf.h (JUMP_TABLES_IN_TEXT_SECTION): Return True for
4567 * config/arm/coff.h (JUMP_TABLES_IN_TEXT_SECTION): Ditto.
4568 * config/arm/aout.h (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
4569 (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump tables.
4570 * config/arm/aof.h (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump
4572 (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
4573 * config/arm/ieee754-df.S: Use macros for Thumb-2/Unified asm
4575 * config/arm/ieee754-sf.S: Ditto.
4576 * config/arm/arm.c (thumb_base_register_rtx_p): Rename...
4577 (thumb1_base_register_rtx_p): ... to this.
4578 (thumb_index_register_rtx_p): Rename...
4579 (thumb1_index_register_rtx_p): ... to this.
4580 (thumb_output_function_prologue): Rename...
4581 (thumb1_output_function_prologue): ... to this.
4582 (thumb_legitimate_address_p): Rename...
4583 (thumb1_legitimate_address_p): ... to this.
4584 (thumb_rtx_costs): Rename...
4585 (thumb1_rtx_costs): ... to this.
4586 (thumb_compute_save_reg_mask): Rename...
4587 (thumb1_compute_save_reg_mask): ... to this.
4588 (thumb_final_prescan_insn): Rename...
4589 (thumb1_final_prescan_insn): ... to this.
4590 (thumb_expand_epilogue): Rename...
4591 (thumb1_expand_epilogue): ... to this.
4592 (arm_unwind_emit_stm): Rename...
4593 (arm_unwind_emit_sequence): ... to this.
4594 (thumb2_legitimate_index_p, thumb2_legitimate_address_p,
4595 thumb1_compute_save_reg_mask, arm_dwarf_handle_frame_unspec,
4596 thumb2_index_mul_operand, output_move_vfp, arm_shift_nmem,
4597 arm_save_coproc_regs, thumb_set_frame_pointer, arm_print_condition,
4598 thumb2_final_prescan_insn, thumb2_asm_output_opcode, arm_output_shift,
4599 thumb2_output_casesi): New functions.
4600 (TARGET_DWARF_HANDLE_FRAME_UNSPEC): Define.
4601 (FL_THUMB2, FL_NOTM, FL_DIV, FL_FOR_ARCH6T2, FL_FOR_ARCH7,
4602 FL_FOR_ARCH7A, FL_FOR_ARCH7R, FL_FOR_ARCH7M, ARM_LSL_NAME,
4603 THUMB2_WORK_REGS): Define.
4604 (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv, arm_condexec_count,
4605 arm_condexec_mask, arm_condexec_masklen)): New variables.
4606 (all_architectures): Add armv6t2, armv7, armv7a, armv7r and armv7m.
4607 (arm_override_options): Check new CPU capabilities.
4608 Set new architecture flag variables.
4609 (arm_isr_value): Handle v7m interrupt functions.
4610 (user_return_insn): Return 0 for v7m interrupt functions. Handle
4612 (const_ok_for_arm): Handle Thumb-2 constants.
4613 (arm_gen_constant): Ditto. Use movw when available.
4614 (arm_function_ok_for_sibcall): Return false for v7m interrupt
4616 (legitimize_pic_address, arm_call_tls_get_addr): Handle Thumb-2.
4617 (thumb_find_work_register, arm_load_pic_register,
4618 legitimize_tls_address, arm_address_cost, load_multiple_sequence,
4619 emit_ldm_seq, emit_stm_seq, arm_select_cc_mode, get_jump_table_size,
4620 print_multi_reg, output_mov_long_double_fpa_from_arm,
4621 output_mov_long_double_arm_from_fpa, output_mov_double_fpa_from_arm,
4622 output_mov_double_fpa_from_arm, output_move_double,
4623 arm_compute_save_reg_mask, arm_compute_save_reg0_reg12_mask,
4624 output_return_instruction, arm_output_function_prologue,
4625 arm_output_epilogue, arm_get_frame_offsets, arm_regno_class,
4626 arm_output_mi_thunk, thumb_set_return_address): Ditto.
4627 (arm_expand_prologue): Handle Thumb-2. Use arm_save_coproc_regs.
4628 (arm_coproc_mem_operand): Allow POST_INC/PRE_DEC.
4629 (arithmetic_instr, shift_op): Use arm_shift_nmem.
4630 (arm_print_operand): Use arm_print_condition. Handle '(', ')', '.',
4632 (arm_final_prescan_insn): Use extract_constrain_insn_cached.
4633 (thumb_expand_prologue): Use thumb_set_frame_pointer.
4634 (arm_file_start): Output directive for unified syntax.
4635 (arm_unwind_emit_set): Handle stack alignment instruction.
4636 * config/arm/lib1funcs.asm: Remove default for __ARM_ARCH__.
4637 Add v6t2, v7, v7a, v7r and v7m.
4638 (RETLDM): Add Thumb-2 code.
4639 (do_it, shift1, do_push, do_pop, COND, THUMB_SYNTAX): New macros.
4640 * config/arm/arm.h (TARGET_CPU_CPP_BUILTINS): Define __thumb2__.
4641 (TARGET_THUMB1, TARGET_32BIT, TARGET_THUMB2, TARGET_DSP_MULTIPLY,
4642 TARGET_INT_SIMD, TARGET_UNIFIED_ASM, ARM_FT_STACKALIGN, IS_STACKALIGN,
4643 THUMB2_TRAMPOLINE_TEMPLATE, TRAMPOLINE_ADJUST_ADDRESS,
4644 ASM_OUTPUT_OPCODE, THUMB2_GO_IF_LEGITIMATE_ADDRESS,
4645 THUMB2_LEGITIMIZE_ADDRESS, CASE_VECTOR_PC_RELATIVE,
4646 CASE_VECTOR_SHORTEN_MODE, ADDR_VEC_ALIGN, ASM_OUTPUT_CASE_END,
4647 ADJUST_INSN_LENGTH): Define.
4648 (TARGET_REALLY_IWMMXT, TARGET_IWMMXT_ABI, CONDITIONAL_REGISTER_USAGE,
4649 STATIC_CHAIN_REGNUM, HARD_REGNO_NREGS, INDEX_REG_CLASS,
4650 BASE_REG_CLASS, MODE_BASE_REG_CLASS, SMALL_REGISTER_CLASSES,
4651 PREFERRED_RELOAD_CLASS, SECONDARY_OUTPUT_RELOAD_CLASS,
4652 SECONDARY_INPUT_RELOAD_CLASS, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P,
4653 TRAMPOLINE_SIZE, INITIALIZE_TRAMPOLINE, HAVE_PRE_INCREMENT,
4654 HAVE_POST_DECREMENT, HAVE_PRE_DECREMENT, HAVE_PRE_MODIFY_DISP,
4655 HAVE_POST_MODIFY_DISP, HAVE_PRE_MODIFY_REG, HAVE_POST_MODIFY_REG,
4656 REGNO_MODE_OK_FOR_BASE_P, LEGITIMATE_CONSTANT_P,
4657 REG_MODE_OK_FOR_BASE_P, REG_OK_FOR_INDEX_P, GO_IF_LEGITIMATE_ADDRESS,
4658 LEGITIMIZE_ADDRESS, THUMB2_LEGITIMIZE_ADDRESS,
4659 GO_IF_MODE_DEPENDENT_ADDRESS, MEMORY_MOVE_COST, BRANCH_COST,
4660 ASM_APP_OFF, ASM_OUTPUT_CASE_LABEL, ARM_DECLARE_FUNCTION_NAME,
4661 FINAL_PRESCAN_INSN, PRINT_OPERAND_PUNCT_VALID_P,
4662 PRINT_OPERAND_ADDRESS): Adjust for Thumb-2.
4663 (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv): New declarations.
4664 * config/arm/arm-cores.def: Add arm1156t2-s, cortex-a8, cortex-r4 and
4666 * config/arm/arm-tune.md: Regenerate.
4667 * config/arm/arm-protos.h: Update prototypes.
4668 * config/arm/vfp.md: Enable patterns for Thumb-2.
4669 (arm_movsi_vfp): Add movw alternative. Use output_move_vfp.
4670 (arm_movdi_vfp, movsf_vfp, movdf_vfp): Use output_move_vfp.
4671 (thumb2_movsi_vfp, thumb2_movdi_vfp, thumb2_movsf_vfp,
4672 thumb2_movdf_vfp, thumb2_movsfcc_vfp, thumb2_movdfcc_vfp): New.
4673 * config/arm/libunwind.S: Add Thumb-2 code.
4674 * config/arm/constraints.md: Update include Thumb-2.
4675 * config/arm/ieee754-sf.S: Add Thumb-2/Unified asm support.
4676 * config/arm/ieee754-df.S: Ditto.
4677 * config/arm/bpabi.S: Ditto.
4678 * config/arm/t-arm (MD_INCLUDES): Add thumb2.md.
4679 * config/arm/predicates.md (low_register_operand,
4680 low_reg_or_int_operand, thumb_16bit_operator): New.
4681 (thumb_cmp_operand, thumb_cmpneg_operand): Rename...
4682 (thumb1_cmp_operand, thumb1_cmpneg_operand): ... to this.
4683 * config/arm/t-arm-elf: Add armv7 multilib.
4684 * config/arm/arm.md: Update patterns for Thumb-2 and Unified asm.
4686 (UNSPEC_STACK_ALIGN, ce_count): New.
4687 (arm_incscc, arm_decscc, arm_umaxsi3, arm_uminsi3,
4688 arm_zero_extendsidi2, arm_zero_extendqidi2): New
4690 * config/arm/fpa.md: Update patterns for Thumb-2 and Unified asm.
4691 (thumb2_movsf_fpa, thumb2_movdf_fpa, thumb2_movxf_fpa,
4692 thumb2_movsfcc_fpa, thumb2_movdfcc_fpa): New insns.
4693 * config/arm/cirrus.md: Update patterns for Thumb-2 and Unified asm.
4694 (cirrus_thumb2_movdi, cirrus_thumb2_movsi_insn,
4695 thumb2_cirrus_movsf_hard_insn, thumb2_cirrus_movdf_hard_insn): New
4697 * doc/extend.texi: Document ARMv7-M interrupt functions.
4698 * doc/invoke.texi: Document Thumb-2 new cores+architectures.
4700 2007-01-03 Jakub Jelinek <jakub@redhat.com>
4702 * unwind-dw2.c (SIGNAL_FRAME_BIT, EXTENDED_CONTEXT_BIT): Define.
4703 (struct _Unwind_Context): Rename args_size to flags, remove
4704 signal_frame field, add a new args_size field and version field.
4705 (_Unwind_IsSignalFrame, _Unwind_SetSignalFrame,
4706 _Unwind_IsExtendedContext): New inline functions.
4707 (_Unwind_GetGR, _Unwind_SetGR, _Unwind_GetGRPtr, _Unwind_SetGRPtr):
4708 Assume by_value array is only present if _Unwind_IsExtendedContext.
4709 (_Unwind_GetIPInfo, execute_cfa_program, uw_frame_state_for): Use
4710 _Unwind_IsSignalFrame.
4711 (__frame_state_for): Initialize context.flags to EXTENDED_CONTEXT_BIT.
4712 (uw_update_context_1): Use _Unwind_SetSignalFrame.
4713 (uw_init_context_1): Initialize context->flags to
4714 EXTENDED_CONTEXT_BIT.
4715 * config/rs6000/linux-unwind.h (frob_update_context): Use
4716 _Unwind_SetSignalFrame.
4718 2007-01-03 Andrew Pinski <andrew_pinski@playstation.sony.com>
4721 * gimplify.c (gimplify_modify_expr_complex_part): Move below
4722 tree_to_gimple_tuple. Call tree_to_gimple_tuple when we need
4725 2007-01-03 Kazu Hirata <kazu@codesourcery.com>
4727 * config/i386/i386.h (NON_STACK_REG_P, REGNO_OK_FOR_SIREG_P,
4728 REGNO_OK_FOR_DIREG_P, REWRITE_ADDRESS, ASM_OPERAND_LETTER,
4729 RET, AT_SP): Remove.
4730 * config/i386/i386.md (*sse_prologue_save_insn): Use return
4733 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
4734 * cfgbuild.c (find_basic_blocks): Likewise.
4735 * cfgrtl.c (rtl_create_basic_block): Likewise.
4736 * function.c (temp_slots_at_level): Likewise.
4737 * reg-stack.c (stack_regs_mentioned): Likewise.
4738 * regclass.c (allocate_reg_info): Likewise.
4739 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
4740 set_bb_for_stmt, move_block_to_fn): Likewise.
4741 * tree-complex.c (tree_lower_complex): Likewise.
4742 * vec.h (VEC_safe_grow_cleared): New.
4744 * cgraphunit.c, tree-ssa-alias.c: Fix comment typos.
4746 2007-01-03 Zdenek Dvorak <dvorakz@suse.cz>
4748 * loop-unswitch.c (unswitch_loop): Pass probabilities to loopify.
4749 * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Pass probabilities
4751 * cfgloopmanip.c (scale_loop_frequencies): Export.
4752 (loopify): Scale the frequencies by prescribed coefficients.
4753 (set_zero_probability): New function.
4754 (duplicate_loop_to_header_edge): Improve updating of frequencies.
4755 (lv_adjust_loop_entry_edge, loop_version): Set probabilities
4756 and frequencies according to arguments.
4757 * tree-ssa-loop-manip.c (tree_unroll_loop): Set probabilities
4759 * cfg.c (scale_bbs_frequencies_int): Allow scaling the frequencies up.
4760 * modulo-sched.c (sms_schedule): Set probabilities for entering
4761 versioned loop correctly.
4762 * tree-vect-transform.c (vect_transform_loop): Ditto.
4763 * cfgloop.h (loopify, loop_version): Declaration changed.
4764 (scale_loop_frequencies): Declared.
4766 2007-01-02 Jan Hubicka <jh@suse.cz>
4768 * cgraph.c: Include tree-flow.h
4769 (cgraph_add_new-function): Handle IPA_SSA mode; execute
4771 * cgraph.h (enum cgraph_state): Add CGRAPH_STATE_IPA_SSA.
4772 * tree-pass.h (pass_all_early_optimizations): Declare.
4773 * cgraphunit.c (cgraph_process_new_functions): Add IPA_SSA; execute
4775 (cgraph_analyze_function): Do early_local_passes.
4776 * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
4777 Do not add referenced vars.
4778 * tree-optimize.c (gate_all_optimizations): Do not execute when not in
4780 (gate_all_early_local_passes): New gate.
4781 (pass_early_local_passes): Use new gate.
4782 (execute_early_local_optimizations): New functions.
4783 (gate_all_early_optimizations): New gate.
4784 (pass_all_early_optimizations): New pass.
4785 (execute_free_datastructures): Free SSA only when initialized.
4786 (gate_init_datastructures): Init only when optimizing.
4787 (tree_lowering_passes): Do early local passes when called late.
4788 * tree-profile.c (do_tree_profiling): Don't profile functions added
4790 (do_early_tree_profiling, pass_early_tree_profile): Kill.
4791 * tree-cfg.c (update_modified_stmts): Do not update when operands are
4793 * passes.c (init_optimizations_passes): Reorder so we go into SSA
4794 during early_local_passes.
4795 * Makefile.in (cgraph.o): Add dependency on tree-flow.h.
4798 2007-01-02 Carlos O'Donell <carlos@codesourcery.com>
4800 * Makefile.in: Update copyright year.
4802 2007-01-02 Carlos O'Donell <carlos@codesourcery.com>
4804 * Makefile.in: Export GCC_EXEC_PREFIX before calling $(RUNTEST)
4805 in $(lang_checks) and check-consistency targets.
4807 2007-01-02 Jan Hubicka <jh@suse.cz>
4809 * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
4810 Do not add referenced vars.
4811 * tree-cfg.c (update_modified_stmts): Do not update when SSA operands
4813 * passes.c (init_optimization_passes): Put mudflap_2 after
4814 free_datastructures.
4816 2007-01-02 Jan Hubicka <jh@suse.cz>
4818 * tree-optimize (execute_fixup_cfg): Set after_inlining flag.
4819 Set NOTHROW flag on call statements proved to be nothrow.
4820 Update statement of local calls so new pure/const functions are
4821 updated. Update_ssa when in ssa form. Mark PHI nodes of nonlocal
4823 (tree_rest_of_compilation): Register hooks and initialize bitmap
4824 early. Do not set after_inlining flag.
4826 2007-01-02 Steve Ellcey <sje@cup.hp.com>
4828 * sbitmap.c (HOST_BITS_PER_LONG_LONG): Change to
4829 HOST_BITS_PER_LONGLONG
4831 2007-01-02 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
4834 * c-typeck.c (store_init_value): Don't emit pedantic overflow
4835 warning for non-static initializers.
4837 2007-01-02 Steven Bosscher <steven@gcc.gnu.org>
4839 * config/alpha/alpha.md, arm/arm.c, darwin.c, frv/frv.md,
4840 m32r/m32r.c, m32r/m32r.c, mn10300/mn10300.md, pa/pa.c,
4841 rs6000/rs6000.c, s390/s390.md, sh/sh.md, sparc/sparc.c:
4842 Always use set_unique_reg_note to add REG_EQUAL notes.
4844 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
4847 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
4849 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
4850 * cfgbuild.c (find_basic_blocks): Likewise.
4851 * cfgrtl.c (rtl_create_basic_block): Likewise.
4852 * function.c (temp_slots_at_level): Likewise.
4853 * reg-stack.c (stack_regs_mentioned): Likewise.
4854 * regclass.c (allocate_reg_info): Likewise.
4855 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
4856 set_bb_for_stmt, move_block_to_fn): Likewise.
4857 * tree-complex.c (tree_lower_complex): Likewise.
4858 * vec.h (VEC_safe_grow_cleared): New.
4860 2007-01-02 Ian Lance Taylor <iant@google.com>
4862 * c-common.c (c_common_truthvalue_conversion): When warning about
4863 using an assignment as a truth value, set TREE_NO_WARNING.
4865 2007-01-02 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
4868 * c.opt (Wold-style-declaration): New.
4869 * doc/invoke.texi (C-only Warning Options): New.
4870 (Wold-style-declaration): Document it.
4871 (Wextra): Enabled by -Wextra.
4872 * c-opts.c (c_common_post_options): Enabled by -Wextra.
4873 * c-decl.c (declspecs_add_scspec): Replace -Wextra with
4874 -Wold-style-declaration.
4876 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
4878 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
4879 * cfgbuild.c (find_basic_blocks): Likewise.
4880 * cfgrtl.c (rtl_create_basic_block): Likewise.
4881 * function.c (temp_slots_at_level): Likewise.
4882 * reg-stack.c (stack_regs_mentioned): Likewise.
4883 * regclass.c (allocate_reg_info): Likewise.
4884 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
4885 set_bb_for_stmt, move_block_to_fn): Likewise.
4886 * tree-complex.c (tree_lower_complex): Likewise.
4887 * vec.h (VEC_safe_grow_cleared): New.
4889 2007-01-02 Douglas Gregor <doug.gregor@gmail.com>
4891 * c-common.c (c_common_nodes_and_builtins): Since variants of
4892 void_type_node get built before it is given a name, we need to
4893 give those variants the name, too.
4894 (complete_array_type): We need to work with the canonical main
4895 type of the array, from which we will build the qualified version.
4896 * params.def (PARAM_VERIFY_CANONICAL_TYPES): New.
4897 * print-tree.c (print_node): Display canonical type information
4899 * stor-layout.c (layout_type): When we don't know the
4900 alignment of a type for which we're building an array, we end up
4901 guessing wrong, so make the type require structural equality.
4902 * tree.c (make_node_stat): When we build a new type, it is its
4904 (build_type_attribute_qual_variant): When building an attribute
4905 variant, its canonical type is the non-attribute variant. However,
4906 if the attributes are target-dependent and they differ, we need to
4907 use structural equality checks for this type.
4908 (build_qualified_type): A qualified type is not equivalent to its
4909 unqualified variant; set the canonical type appropriately.
4910 (build_distinct_type_copy): When building a distinct type from
4911 another type, the new type is its own canonical type.
4912 (build_variant_type_copy): When building a new type variant, we
4913 assume that it is equivalent to the original type.
4914 (build_pointer_type_for_mode): When building a pointer type, also
4915 build a canonical type pointer.
4916 (build_reference_type_for_mode): When building a reference type,
4917 also build a canonical type reference.
4918 (build_index_type): When we can't hash an index type (e.g.,
4919 because its maximum value is negative), the index type requires
4920 structural equality tests.
4921 (build_array_type): Build the canonical form of an array type.
4922 (build_function_type): Function types require structural equality,
4923 because they contain default arguments, attributes, etc.
4924 (build_method_type_directly): Ditto for method types.
4925 (build_offset_type): Build the canonical offset type.
4926 (build_complex_type): Build the canonical vector type.
4927 (make_vector_type): Build the canonical vector type.
4928 * tree.h (TYPE_CANONICAL): New.
4929 (TYPE_STRUCTURAL_EQUALITY_P): New.
4930 (SET_TYPE_STRUCTURAL_EQUALITY): New.
4931 (struct tree_type): Added "canonical" field.
4932 * params.h (VERIFY_CANONICAL_TYPES): New.
4933 * doc/c-tree.texi (TYPE_CANONICAL): Document.
4934 (TYPE_STRUCTURAL_EQUALITY_P): Document.
4935 (SET_TYPE_STRUCTURAL_EQUALITY): Document.
4936 * doc/invoke.texi (verify-canonical-types): Document --param
4937 parameter for verifying canonical types.
4939 2007-01-02 Joseph Myers <joseph@codesourcery.com>
4941 * config.gcc (powerpc-*-eabispe*, powerpc-*-eabisimaltivec*,
4942 powerpc-*-eabisim*, powerpc-*-eabialtivec*, powerpc-*-eabi*,
4943 powerpc-*-rtems*, powerpc-wrs-vxworks, powerpc-wrs-vxworksae,
4944 powerpcle-*-eabisim*, powerpcle-*-eabi*): Add rs6000/e500.h to
4946 * config/rs6000/e500.h: New.
4947 * config/rs6000/eabi.h (TARGET_SPE_ABI, TARGET_SPE, TARGET_E500,
4948 TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE, TARGET_E500_DOUBLE):
4950 * config/rs6000/linuxspe.h (TARGET_SPE_ABI, TARGET_SPE,
4951 TARGET_E500, TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE,
4952 TARGET_E500_DOUBLE): Remove.
4953 * config/rs6000/vxworks.h (TARGET_SPE_ABI, TARGET_SPE,
4954 TARGET_E500, TARGET_ISEL, TARGET_FPRS): Remove.
4955 * config/rs6000/rs6000.h (CHECK_E500_OPTIONS): Define.
4956 * config/rs6000/rs6000.c (rs6000_override_options): Use
4959 2007-01-02 Joseph Myers <joseph@codesourcery.com>
4961 * config/rs6000/rs6000.c (print_operand): Check (TARGET_SPE ||
4962 TARGET_E500_DOUBLE), not TARGET_E500, for %y.
4963 (rs6000_generate_compare, rs6000_emit_sCOND, output_cbranch,
4964 rs6000_emit_cmove): Don't check TARGET_E500.
4965 * config/rs6000/rs6000.md (bunordered, bordered, sunordered,
4966 sordered): Don't check TARGET_E500.
4968 2007-01-01 Eric Christopher <echristo@apple.com>
4970 * config/mips/mips.c (mips_regno_mode_ok_for_base_p): Use
4971 HARD_REGISTER_NUM_P.
4973 2007-01-01 Roger Sayle <roger@eyesopen.com>
4975 * fold-const.c (fold_binary) <EQ_EXPR>: Fold "(X^C1) eq/ne C2" into
4976 "X eq/ne (C1^C2)". Fold "(X^Z) eq/ne (Y^Z)" as "X eq/ne Y" when Z
4977 has no side-effects. Fold "(X^C1) eq/ne (Y^C2)" as "(X^(C1^C2))
4980 2007-01-01 Mike Stump <mrs@apple.com>
4982 * configure.ac: Remove support for building with Apple's gcc-3.1.
4984 2007-01-02 Joseph Myers <joseph@codesourcery.com>
4987 * caller-save.c (add_stored_regs): Only handle SUBREGs if inner
4988 REG is a hard register. Do not modify REG before calling
4990 * rtlanal.c (subreg_get_info): Don't assert size of XMODE is a
4991 multiple of the size of YMODE for certain lowpart cases.
4993 2007-01-01 Andrew Pinski <pinskia@gmail.com>
4996 * gimplify (voidify_wrapper_expr): Update for
4999 2007-01-01 Andreas Schwab <schwab@suse.de>
5002 * config/ia64/ia64.c (ia64_compute_frame_size): Account space for
5003 save of BR0 in extra_spill_size instead of spill_size.
5004 (ia64_expand_prologue): Save BR0 outside of the gr/br/fr spill
5006 (ia64_expand_epilogue): Restore BR0 from its new location.
5008 2007-01-01 Andrew Pinski <pinskia@gmail.com>
5010 * gimplify.c (gimplify_init_constructor <case VECTOR_TYPE>):
5011 Use a temporary variable if the left hand side is not a gimple
5014 2007-01-01 Andrew Pinski <pinskia@gmail.com>
5016 * gimplify.c (gimplify_return_expr): Make the temporary variable
5017 for the return expression, a gimple register variable.
5019 2007-01-01 Jan Hubicka <jh@suse.cz>
5021 * emit-rtl.c (emit_copy_of_insn_after): Do not call copy_insn_1 for
5024 2007-01-01 Mike Stump <mrs@apple.com>
5026 * configure.ac (HAVE_GAS_LITERAL16): Add autoconf check for
5028 * config/darwin.c (machopic_select_rtx_section): Use
5030 (darwin_mergeable_constant_section): Likewise.
5031 * configure: Regenerate.
5032 * config.in: Regenerate.
5034 2007-01-01 Jan Hubicka <jh@suse.cz>
5035 Andrew Pinski <pinskia@gmail.com>
5037 * cgraphunit.c (cgraph_optimize): Call cgraph_add_new_functions
5038 before starting IPA passes.
OSDN Git Service
