1 2007-12-14 Dorit Nuzman <dorit@il.ibm.com>
3 PR tree-optimization/30771
4 * tree-vect-analyze.c (vect_determine_vectorization_factor): Traverse
6 (vect_analyze_operations): Induction phis can now be marked as
8 (vect_mark_stmts_to_be_vectorized): No special treatment for phis.
9 Update documentation accordingly.
11 2007-02-14 Nick Clifton <nickc@redhat.com>
13 * builtin-types.def (DEF_FUNCTION_TYPE_x): Do not imply that at
14 most 3 arguments are supported.
15 (DEF_FUNCTION_TYPE_VAR_5): Fix typo in its description.
17 2007-02-13 Seongbae Park <seongbae.park@gmail.com>
19 * bitmap.c (bitmap_and, bitmap_and_compl, bitmap_xor):
20 Ensure dst->current is valid.
22 2007-02-13 Paul Brook <paul@codesourcery.com>
24 * config.gcc: Add arm*-*-uclinux-*eabi.
25 * config/arm/uclinux-elf.h (TARGET_OS_CPP_BUILTINS): Define.
26 (SUBTARGET_EXTRA_LINK_SPEC): Define.
27 (STARTFILE_SPEC, ENDFILE_SPEC): Remove broken -shared handling.
28 (LINK_GCC_C_SEQUENCE_SPEC): Undef.
31 * config/arm/arm.c (arm_override_options): Use r9 as EABI PIC
33 * config/arm/uclinux-eabi.h: New file.
34 * config/arm/linux-eabi.h (WCHAR_TYPE): Remove.
35 * config/arm/linux-gas.h (WCHAR_TYPE): Use unsigned long on AAPCS
38 2007-02-13 Ian Lance Taylor <iant@google.com>
40 * common.opt: Add Wstrict-overflow and Wstrict-overflow=.
41 * flags.h (warn_strict_overflow): Declare.
42 (enum warn_strict_overflow_code): Define.
43 (issue_strict_overflow_warning): New static inline function.
44 * opts.c (warn_strict_overflow): New variable.
45 (common_handle_option): Handle OPT_Wstrict_overflow and
46 OPT_Wstrict_overflow_.
47 * c-opts.c (c_common_handle_option): Set warn_strict_overflow for
49 * fold-const.c: Include intl.h.
50 (fold_deferring_overflow_warnings): New static variable.
51 (fold_deferred_overflow_warning): New static variable.
52 (fold_deferred_overflow_code): New static variable.
53 (fold_defer_overflow_warnings): New function.
54 (fold_undefer_overflow_warnings): New function.
55 (fold_undefer_and_ignore_overflow_warnings): New function.
56 (fold_deferring_overflow_warnings_p): New function.
57 (fold_overflow_warning): New static function.
58 (make_range): Add strict_overflow_p parameter. Change all
60 (extract_muldiv, extract_muldiv_1): Likewise.
61 (fold_unary) [ABS_EXPR]: Check ABS_EXPR before calling
62 tree_expr_nonnegative_p.
63 (fold_negate_expr): Call fold_overflow_warning.
64 (fold_range_test): Likewise.
65 (fold_comparison): Likewise.
66 (fold_binary): Likewise. Call tree_expr_nonnegative_warnv_p
67 instead of tree_expr_nonnegative_p.
68 (tree_expr_nonnegative_warnv_p): Rename from
69 tree_expr_nonnegative_p, add strict_overflow_p parameter.
70 (tree_expr_nonnegative_p): New function.
71 (tree_expr_nonzero_warnv_p): Rename from tree_expr_nonzero_p, add
72 strict_overflow_p parameter.
73 (tree_expr_nonzero_p): New function.
74 * passes.c (verify_interpass_invariants): New static function.
75 (execute_one_pass): Call it.
76 * tree-ssa-loop-niter.c (expand_simple_operations): Ignore fold
78 (number_of_iterations_exit, loop_niter_by_eval): Likewise.
79 (estimate_numbers_of_iterations): Likewise.
80 (scev_probably_wraps_p): Likewise.
81 * tree-ssa-ccp.c: Include "toplev.h".
82 (evaluate_stmt): Defer fold overflow warnings until we know we are
84 (struct fold_stmt_r_data): Add stmt field.
85 (fold_stmt_r): Defer fold overflow warnings until we know we
87 (fold_stmt): Initialize stmt field of fold_stmt_r_data.
88 (fold_stmt_inplace): Likewise.
89 * tree-cfgcleanup.c: Include "toplev.h" rather than "errors.h".
90 (cleanup_control_expr_graph): Defer fold overflow warnings until
91 we know we are going to optimize.
92 * tree-cfg.c (fold_cond_expr_cond): Likewise.
93 * tree-ssa-threadedge.c (simplify_control_stmt_condition):
95 * tree-vrp.c (vrp_expr_computes_nonnegative): Call
96 tree_expr_nonnegative_warnv_p instead of tree_expr_nonnegative_p.
97 * tree-ssa-loop-manip.c (create_iv): Likewise.
98 * c-typeck.c (build_conditional_expr): Likewise.
99 (build_binary_op): Likewise.
100 * tree-vrp.c (vrp_expr_computes_nonzero): Call
101 tree_expr_nonzero_warnv_p instead of tree_expr_nonzero_p.
102 (extract_range_from_unary_expr): Likewise.
103 * simplify-rtx.c (simplify_const_relational_operation): Warn when
104 assuming that signed overflow does not occur.
105 * c-common.c (pointer_int_sum): Ignore fold overflow warnings.
106 * tree.h (tree_expr_nonnegative_warnv_p): Declare.
107 (fold_defer_overflow_warnings): Declare.
108 (fold_undefer_overflow_warnings): Declare.
109 (fold_undefer_and_ignore_overflow_warnings): Declare.
110 (fold_deferring_overflow_warnings_p): Declare.
111 (tree_expr_nonzero_warnv_p): Declare.
112 * doc/invoke.texi (Option Summary): Add -Wstrict-overflow to list
114 (Warning Options): Document -Wstrict-overflow.
115 * Makefile.in (tree-ssa-threadedge.o): Depend on toplev.h.
116 (tree-ssa-ccp.o): Likewise.
117 (tree-cfgcleanup.o): Change errors.h dependency to toplev.h.
118 (fold-const.o): Depend on intl.h.
120 2007-02-13 Ian Lance Taylor <iant@google.com>
123 * lower-subreg.c (resolve_simple_move): Decompose subregs in
126 2007-02-13 Stuart Hastings <stuart@apple.com>
128 * gcc/config/i386/i386.md (fixuns_truncdfhi2): Require SSE2.
130 2007-02-13 Richard Henderson <rth@redhat.com>
132 * config/alpha/alpha.c (alpha_stdarg_optimize_hook): Strip
133 handled_component_p before looking for the indirect_ref.
135 2007-02-13 Richard Henderson <rth@redhat.com>
137 * config/i386/i386.md (bswapsi_1): Rename from bswapsi2,
138 remove flags clobber.
139 (bswapsi2): New expander, emit code for !TARGET_BSWAP.
140 (bswaphi_lowpart): New.
141 (bswapdi2): Rename from bswapdi2_rex, remove flags clobber,
142 remove TARGET_BSWAP test. Delete expander of the same name.
144 * optabs.c (widen_bswap, expand_doubleword_bswap): New.
145 (expand_unop): Use them.
147 2007-02-13 Uros Bizjak <ubizjak@gmail.com>
149 * config/i386/i386.md (cmpdi_ccno_1_rex64, *cmpsi_ccno_1,
150 *cmphi_ccno_1, *cmpqi_ccno_1, *movsi_xor, *movstricthi_xor,
151 *movstrictqi_xor, *movdi_xor_rex64, *ashldi3_1_rex64,
152 *ashldi3_cmp_rex64, *ashldi3_cconly_rex64, ashlsi3, *ashlsi3_1_zext,
153 *ashlsi3_cmp, *ashlsi3_cconly, *ashlsi3_cmp_zext, *ashlhi3_1_lea,
154 *ashlhi3_1, *ashlhi3_cmp, *ashlhi3_cconly, *ashlqi3_1_lea,
155 *ashlqi3_1, *ashlqi3_cmp, *ashlqi3_cconly): Remove equivalent
156 assembler dialect choice from asm templates.
158 2007-02-12 Richard Henderson <rth@redhat.com>
160 * config/i386/i386.md (fixuns_trunc<SSEMODEF>si_1): New insn.
161 (fixuns_trunc<SSEMODEF>si2): Use it.
162 * config/i386/sse.md (vec_setv4sf_0): Export.
163 * config/i386/i386.c (ix86_build_const_vector): Export.
164 (ix86_split_convert_uns_si_sse): Rename from
165 ix86_expand_convert_uns_si_sse and rewrite as a splitter.
166 * config/i386/i386-protos.h: Update.
168 2007-02-13 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
171 * c-typeck.c (c_finish_return): Improve warning message.
173 2007-02-12 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
175 * alias.c (find_symbolic_term): Delete unused function.
177 2007-02-12 Uros Bizjak <ubizjak@gmail.com>
179 * config/i386/i386.md (paritydi2, paritysi2): New expanders.
180 (paritydi2_cmp, paritydi2_cmp): New insn and split patterns.
181 (*parityhi2_cmp, *parityqi2_cmp): New insn patterns.
183 2007-02-12 Eric Botcazou <ebotcazou@adacore.com>
185 * tree.h (DECL_IGNORED_P): Document further effect for FUNCTION_DECL.
186 * cgraphunit.c (cgraph_expand_function): If DECL_IGNORED_P is set on
187 the function, temporarily point the debug interface to the null one.
189 2007-02-12 Eric Botcazou <ebotcazou@adacore.com>
191 * dwarf2out.c (round_up_to_align): New static function.
192 (field_byte_offset): Use it to round the offset.
194 2007-02-12 Richard Henderson <rth@redhat.com>
196 * config/alpha/alpha.md (bswapsi2, bswapdi2): New.
197 (inswl_const): Export.
199 2007-02-12 Richard Henderson <rth@redhat.com>
201 * calls.c (emit_library_call_value_1): If PROMOTE_MODE modifed the
202 result mode of the libcall, convert back to outmode.
204 2007-02-12 Roger Sayle <roger@eyesopen.com>
206 * config/i386/i386.md (*bswapdi2_rex): Renamed from bswapdi2.
207 (bswapdi2): New define_expand to implement 32-bit implementation.
209 2007-02-12 Nick Clifton <nickc@redhat.com>
211 * doc/invoke.texi (Overall Options): Document --help=.
212 * gcc.c (target_help_flag): Rename to print_subprocess_flag.
213 (cc1_options): Pass --help= on to cc1.
214 (display_help): Add description of --help=.
215 (process_command): Add code to handle --help=. Allow translated
216 --help and --target-help switches to be passed on to compiler
218 (main): Remove unused if statement.
219 * opts.c (columns): Remove.
220 (LEFT_COLUMN): Define.
221 (wrap_help): Add columns argument.
222 (print_filtered_help): Change parameters to be an include bitmask,
223 an exclude bitmask, an any bitmask and the column width. Move the
224 code to display the params list here. Add code to display the
225 status of options rather than their descriptions if the quiet flag
227 (print_specific_help): Change parameters to be an include bitmask,
228 an exclude bitmask and an any bitmask. Move code to look up the
229 column width here. Decide upon the title for an options listing.
230 (common_handle_options): Add code to handle --help=. Adapt code
231 for --help and --target-help to use the revised form of the
232 print_specific_help function.
233 (print_help): Delete.
234 (print_param_help): Delete.
235 (print_switch): Delete.
236 * opts.h (cl_lang_count): Add prototype.
237 (CL_PARAMS, CL_WARNING, CL_OPTIMIZATION, CL_MIN_OPTION_CLASS,
238 CL_MAX_OPTION_CLASS): New defines.
239 * optc-gen.awk: Add construction of cl_lang_count.
240 * c.opt: Add Warning attribute to warning options and Optimization
241 attribute to optimization options.
242 * common.opt: Likewise.
244 Add -fhelp and -ftarget-help as aliases for the transformed --help
245 and --target-help options.
246 * opt-functions.awk: Add code to handle Warning and Optimization
249 2007-02-12 Richard Henderson <rth@redhat.com>
251 * config/alpha/constraints.md: New file.
252 * config/alpha/alpha.c: Include tm-constrs.h.
253 (alpha_const_ok_for_letter_p, alpha_const_double_ok_for_letter_p,
254 alpha_extra_constraint): Remove.
255 (alpha_emit_conditional_branch): Use satisfies_constraint_*.
256 * config/alpha/alpha-protos.h: Update.
257 * config/alpha/alpha.h (REG_CLASS_FROM_LETTER): Remove.
258 (CONST_OK_FOR_LETTER_P, CONST_DOUBLE_OK_FOR_LETTER_P): Remove.
259 (EXTRA_CONSTRAINT): Remove.
260 * config/alpha/alpha.md: Include constraints.md.
261 (adddi splitter): Use satisfies_constraint_*.
262 * config/alpha/predicates.md (add_operand): Likewise.
263 (sext_add_operand, addition_operation): Likewise.
265 2007-02-12 Dorit Nuzman <dorit@il.ibm.com>
267 PR tree-optimization/29145
268 * tree-data-ref.c (base_addr_differ_p): Make us more conservative
269 in our handling of restrict qualified pointers.
271 2007-02-12 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
274 * doc/invoke.texi (Wunused-value): Update description.
275 (Wextra): Delete item.
276 * opts.c (set_Wextra): Don't use the value of Wextra to set the
277 value of Wunused-value.
278 * c-typeck.c (c_process_expr_stmt): Don't check extra_warnings.
279 (c_finish_stmt_expr): Don't check extra_warnings.
280 (emit_side_effect_warnings): The caller is responsible to check
283 2007-02-11 Roger Sayle <roger@eyesopen.com>
284 Matt Thomas <matt@3am-software.com>
286 * simplify-rtx.c (simplify_relational_operation_1): Correct typo.
288 2007-02-11 Roger Sayle <roger@eyesopen.com>
290 * simplify-rtx.c (simplify_relational_operation_1): Optimize
291 comparisons of POPCOUNT against zero.
292 (simplify_const_relational_operation): Likewise.
294 2007-02-11 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
296 * doc/invoke.texi (Wextra): Delete outdated paragraph.
298 2007-02-11 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
300 * dwarf2out.c (root_type): Delete unused function.
302 2007-02-11 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
304 * genattrtab.c (contained_in_p): Delete unused function.
305 (write_expr_attr_cache): Likewise.
307 2007-02-11 Jan Hubicka <jh@suse.cz>
309 * ipa-inline.c (cgraph_edge_badness): Add "else" missing in the
312 2007-02-11 Steven Bosscher <steven@gcc.gnu.org>
314 * fwprop.c (try_fwprop_subst): Use set_unique_reg_note
315 to add the REG_EQ* note.
316 * see.c (see_merge_one_use_extension): Likewise.
317 * local-alloc.c (update_equiv_regs): Likewise. Also don't
318 turn REG_EQUAL notes into REG_EQUIV notes if the target
319 register may have more than one set.
320 * function.c (assign_parm_setup_reg): Use set_unique_reg_note.
321 * gcse.c (try_replace_reg): Likewise.
322 * alias.c (init_alias_analysis): Use find_reg_equal_equiv_note.
323 * calls.c (fixup_tail_calls): Likewise. Abort if there is
324 more than one REG_EQUIV note.
325 * reload1.c (gen_reload): Use set_unique_reg_note.
327 2007-02-11 Uros Bizjak <ubizjak@gmail.com>
329 * config/i386/i386.c (TARGET_VECTORIZE_BUILTIN_CONVERSION): Define.
330 (ix86_builtin_conversion): New function.
332 2007-02-06 Mark Mitchell <mark@codesourcery.com>
335 * tree.h (DECL_REPLACEABLE_P): New macro.
336 * except.c (set_nothrow_function_flags): Likewise.
338 2007-02-11 Tehila Meyzels <tehila@il.ibm.com>
339 Ira Rosen <irar@il.ibm.com>
340 Dorit Nuzman <dorit@il.ibm.com>
342 * doc/tm.texi (TARGET_VECTORIZE_BUILTIN_CONVERSION): New target hook.
343 * targhooks.c (default_builtin_vectorized_conversion): New.
344 * targhooks.h (default_builtin_vectorized_function): New declaration.
345 * target.h (struct vectorize): Add builtin_conversion field.
346 * tree-vectorizer.h (type_conversion_vec_info_type): New enum
347 stmt_vec_info_type value.
348 (vectorizable_conversion): New declaration.
349 * tree-vect-analyze.c (vect_analyze_operations): Add
350 vectorizable_conversion call.
351 * target-def.h (TARGET_VECTORIZE_BUILTIN_CONVERSION): New.
352 * tree-vect-transform.c (vectorizable_conversion): New function.
353 (vect_transform_stmt): Add case for type_conversion_vec_info_type.
354 * tree-vect-generic.c (expand_vector_operations_1): Consider correct
356 * config/rs6000/rs6000.c (rs6000_builtin_conversion): New.
357 (TARGET_VECTORIZE_BUILTIN_CONVERSION): Defined.
358 (rs6000_expand_builtin): Add handling a case of ALTIVEC_BUILTIN_VCFUX
359 or ALTIVEC_BUILTIN_VCFSX.
361 2007-02-10 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
364 * pa.md (movdf): For 64-bit target, fail if operand 1 is a non-zero
365 CONST_DOUBLE and operand 0 is a hard register.
366 (movdi): For 64-bit target, remove code to force CONST_DOUBLE to
367 memory. Fail if operand 1 is a non-zero CONST_INT and operand 0
368 is a hard floating-point register.
370 2007-02-10 Richard Henderson <rth@redhat.com>
371 Jakub Jelinek <jakub@redhat.com>
372 Alexandre Oliva <aoliva@redhat.com>
374 * Makefile.in (libgcc-support, libgcc.mvars): Add emutls.c.
375 * builtin-types.def (BT_WORD): Make unsigned.
376 (BT_FN_VOID_PTR_WORD_WORD_PTR): New.
377 * builtins.def (BUILT_IN_EMUTLS_GET_ADDRESS): New.
378 (BUILT_IN_EMUTLS_REGISTER_COMMON): New.
379 * c-decl.c (grokdeclarator): Don't error if !have_tls.
380 * c-parser.c (c_parser_omp_threadprivate): Likewise.
381 * dwarf2out.c (loc_descriptor_from_tree_1): Don't do anything for
383 * expr.c (emutls_var_address): New.
384 (expand_expr_real_1): Expand emulated tls.
385 (expand_expr_addr_expr_1): Likewise.
386 * libgcc-std.ver: Add __emutls_get_address, __emutls_register_common.
387 * output.h (emutls_finish): Declare.
388 * toplev.c (compile_file): Call it.
389 * tree-ssa-address.c (gen_addr_rtx): Check for const-ness of the
390 address before wrapping in CONST.
391 * varasm.c (emutls_htab, emutls_object_type): New.
392 (EMUTLS_VAR_PREFIX, EMUTLS_TMPL_PREFIX): New.
393 (get_emutls_object_name, get_emutls_object_type): New.
394 (get_emutls_init_templ_addr, emutls_decl): New.
395 (emutls_common_1, emutls_finish): New.
396 (assemble_variable): When emulating tls, swap decls; generate
397 constructor for the emutls objects.
398 (do_assemble_alias): When emulating tls, swap decl and target name.
399 (default_encode_section_info): Don't add SYMBOL_FLAG_TLS_SHIFT
401 * varpool.c (decide_is_variable_needed): Look at force_output.
402 Recurse for emulated tls.
403 (cgraph_varpool_remove_unreferenced_decls): Remove checks redundant
404 with decide_is_variable_needed.
405 * emutls.c: New file.
406 * config/sparc/sol2.h (ASM_DECLARE_OBJECT_NAME): Only emit
407 tls_object for real tls.
409 2007-02-10 Kaz Kojima <kkojima@gcc.gnu.org>
411 PR rtl-optimization/29599
412 * reload1.c (eliminate_regs_in_insn): Take the destination
413 mode into account when computing the offset.
415 2007-02-09 Stuart Hastings <stuart@apple.com>
416 Richard Henderson <rth@redhat.com>
418 * gcc/config/i386/i386.h (TARGET_KEEPS_VECTOR_ALIGNED_STACK): New.
419 * gcc/config/i386/darwin.h: (TARGET_KEEPS_VECTOR_ALIGNED_STACK): New.
420 * gcc/config/i386/i386.md (fixuns_trunc<mode>si2, fixuns_truncsfhi2,
421 fixuns_truncdfhi2): New.
422 (fix_truncsfdi_sse): Call ix86_expand_convert_sign_didf_sse.
423 (floatunsdidf2): Call ix86_expand_convert_uns_didf_sse.
424 (floatunssisf2): Add call to ix86_expand_convert_uns_sisf_sse.
425 (floatunssidf2): Allow nonimmediate source.
426 * gcc/config/i386/sse.md (movdi_to_sse): New.
427 (vec_concatv2di): Drop '*'.
428 * gcc/config/i386/i386-protos.h (ix86_expand_convert_uns_si_sse,
429 ix86_expand_convert_uns_didf_sse, ix86_expand_convert_uns_sidf_sse,
430 ix86_expand_convert_uns_sisf_sse, ix86_expand_convert_sign_didf_sse):
432 * gcc/config/i386/i386.c (ix86_expand_convert_uns_si_sse,
433 ix86_expand_convert_uns_didf_sse, ix86_expand_convert_uns_sidf_sse,
434 ix86_expand_convert_uns_sisf_sse, ix86_expand_convert_sign_didf_sse,
435 ix86_build_const_vector, ix86_expand_vector_init_one_nonzero): New.
436 (ix86_build_signbit_mask): Fix decl of v, refactor to call
437 ix86_build_const_vector.
438 (x86_emit_floatuns): Rewrite.
440 2007-02-10 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
442 * genautomata.c (longest_path_length): Delete unused function.
443 (struct state): Delete unused longest_path_length.
444 (UNDEFINED_LONGEST_PATH_LENGTH): Delete unused macro.
445 (get_free_state): Delete unused.
447 2007-02-09 Jan Hubicka <jh@suse.cz>
449 * params.def (PARAM_INLINE_UNIT_GROWTH): Set to 30.
450 * doc/invoke.texi (inline-unit-growth): Update default value.
452 * Makefile.in (passes.o, ipa-inline.o): Add dependencies.
453 * cgraphbuild.c (build_cgraph_edges): Compute frequencies.
454 (rebuild_cgraph_edges): Likewise.
455 * cgraph.c (cgraph_set_call_stmt): Add new argument frequency.
456 (dump_cgraph_node): Dump frequencies.
457 (cgraph_clone_edge): Add frequency scales.
458 (cgraph_clone_node): Add freuqnecy.
459 * cgraph.h (cgraph_edge): Add freuqnecy argument.
460 (CGRAPH_FREQ_BASE, CGRAPH_FREQ_MAX): New constants.
461 (cgraph_create_edge, cgraph_clone_edge, cgraph_clone_node): Update.
462 * tree-pass.h (TODO_rebuild_frequencies): New constant.
463 * cgraphunit.c (verify_cgraph_node): Verify frequencies.
464 (cgraph_copy_node_for_versioning): Update call of cgraph_clone_edge.
465 (save_inline_function_body): Likewise.
466 * ipa-inline.c: inluce rtl.h
467 (cgraph_clone_inlined_nods): Update call of cgraph_clone_node.
468 (cgraph_edge_badness): Use frequencies.
469 (cgraph_decide_recursive_inlining): Update clonning.
470 (cgraph_decide_inlining_of_small_function): Dump frequency.
471 * predict.c (estimate_bb_frequencies): Export.
472 * predict.h (estimate_bb_frequencies): Declare.
473 * tree-inline.c (copy_bb): Watch overflows.
474 (expand_call_inline): Update call of cgraph_create_edge.
475 (optimize_inline_calls): Use TODO flags to update frequnecies.
476 * passes.h: Include predict.h
477 (init_optimization_passes): Move profile ahead.
478 (execute_function_todo): Handle TODO_rebuild_frequencies.
480 2007-02-09 Roger Sayle <roger@eyesopen.com>
482 * config/alpha/alpha.c (emit_insxl): Force the first operand of
483 the insbl or inswl pattern into a register.
485 2007-02-09 Roger Sayle <roger@eyesopen.com>
487 * config/ia64/ia64.md (bswapdi2): New define_insn.
489 2007-02-09 Richard Henderson <rth@redhat.com>
491 * config/i386/constraints.md (Ym): New constraint.
492 * config/i386/i386.md (movsi_1): Change Y2 to Yi constraints.
493 (movdi_1_rex64): Split sse and xmm general register moves from
494 memory move alternatives. Use conditional register constraints.
495 (movsf_1, movdf_integer): Likewise.
496 (zero_extendsidi2_32, zero_extendsidi2_rex64): Likewise.
497 (movdf_integer_rex64): New.
498 (pushsf_rex64): Fix output constraints.
499 * config/i386/sse.md (sse2_loadld): Split rm alternative, use Yi.
500 (sse2_stored): Likewise.
501 (sse2_storeq_rex64): New.
502 * config/i386/i386.c (x86_inter_unit_moves): Enable for not
504 (ix86_secondary_memory_needed): Don't bypass TARGET_INTER_UNIT_MOVES
505 for optimize_size. Remove SF/DFmode hack.
507 2007-02-09 Dwarakanath Rajagopal <dwarak.rajagopal@amd.com>
509 * config/i386/driver-i386.c: Turn on -mtune=native for AMDFAM10.
512 2007-02-09 Nathan Sidwell <nathan@codesourcery.com>
513 Richard Sandiford <richard@codesourcery.com>
515 * config.gcc (m68010-*-netbsdelf*, m68k*-*-netbsdelf*)
516 (m68k*-*-openbsd*, m68k-*-linux*): Set default_cf_cpu.
517 (m68k-*-aout*, m68k-*-coff*, m68k-*-uclinux*, m68k-*-rtems*): Add
518 m68k/t-mlib to tmake_file.
519 (m68020-*-elf*, m68k-*-elf*): Likewise. Add t-m68kbare as well.
520 (m68k*-*-*): Use --with-arch to pick a default for --with-cpu.
521 (m680[012]0-*-*, m68k*-*-*): Add support for --with-arch.
522 Allow it to be cf or m68k. Set m68k_arch_family. If that
523 variable is not empty, add t-$m68k_arch_family to tmake_file.
524 Add t-mlibs to tmake_file.
525 * doc/install.texi: Document --with-arch=m68k and --with-arch=cf.
526 * config/m68k/t-cf: New file.
527 * config/m68k/t-m68k: Likewise.
528 * config/m68k/t-mlibs: Likewise.
529 * config/m68k/t-m68kbare (MULTILIB_OPTIONS, MULTILIB_DIRNAMES)
530 (MULTILIB_MATCHES, MULTILIB_EXCEPTIONS): Delete.
531 (M68K_MLIB_DIRNAMES, M68K_MLIB_OPTIONS): Define.
532 * config/m68k/t-m68kelf (MULTILIB_OPTIONS, MULTILIB_DIRNAMES)
533 (MULTILIB_MATCHES, MULTILIB_EXCEPTIONS, LIBGCC, INSTALL_LIBGCC):
535 * config/m68k/t-openbsd (MULTILIB_OPTIONS, LIBGCC): Delete.
536 (INSTALL_LIBGCC): Delete.
537 (M68K_MLIB_DIRNAMES, M68K_MLIB_OPTIONS): Define.
538 * config/m68k/t-rtems (MULTILIB_OPTIONS, MULTILIB_DIRNAMES)
539 (MULTILIB_MATCHES, MULTILIB_EXCEPTIONS): Delete.
540 (M68K_MLIB_CPU): Define.
541 * config/m68k/t-uclinux (MULTILIB_OPTIONS, MULTILIB_DIRNAMES)
542 (MULTILIB_MATCHES, MULTILIB_EXCEPTIONS): Delete.
543 (M68K_MLIB_CPU, M68K_MLIB_OPTIONS, M68K_MLIB_DIRNAMES): Define.
545 2007-02-09 Zdenek Dvorak <dvorakz@suse.cz>
546 Richard Guenther <rguenther@suse.de>
549 * fold-const.c (fold_comparison): Handle obfuscated comparisons
550 against INT_MIN/INT_MAX.
551 * tree-ssa-loop-ivcanon.c (remove_empty_loop): Print to dump
552 file if a loop is removed.
554 2007-02-09 Joseph Myers <joseph@codesourcery.com>
556 * calls.c (store_one_arg): Pass correct alignment to
557 emit_push_insn for non-BLKmode values.
558 * expr.c (emit_push_insn): If STRICT_ALIGNMENT, copy to an
559 unaligned stack slot via a suitably aligned slot.
561 2007-02-08 DJ Delorie <dj@redhat.com>
563 * config/m32c/m32c.c (m32c_unpend_compare): Add default to silence
565 (legal_subregs): Use unsigned char, make const.
566 (m32c_illegal_subreg_p): Use ARRAY_SIZE. Delete unused variables.
568 2007-02-08 Paul Brook <paul@codesourcery.com>
570 * config/arm/lib1funcs.asm (RETLDM): Pop directly into PC when no
571 special interworking needed.
573 2007-02-08 Harsha Jagasia <harsha.jagasia@amd.com>
575 * config/i386/xmmintrin.h: Make inclusion of emmintrin.h
576 conditional to __SSE2__.
577 (Entries below should have been added to first ChangeLog
578 entry for amdfam10 dated 2007-02-05)
579 * config/i386/emmintrin.h: Generate #error if __SSE2__ is not
581 * config/i386/pmmintrin.h: Generate #error if __SSE3__ is not
583 * config/i386/tmmintrin.h: Generate #error if __SSSE3__ is not
586 2007-02-08 DJ Delorie <dj@redhat.com>
588 * config/m32c/m32c-protos.h (m32c_illegal_subreg_p): New.
589 * config/m32c/m32c.c (legal_subregs): New.
590 (m32c_illegal_subreg_p): New.
591 * config/m32c/predicates.md (m32c_any_operand): Use it to reject
592 unsupported subregs of hard regs.
594 2007-02-08 Jan Hubicka <jh@suse.cz>
596 * tree-cfg.c (bsi_replace): Shortcut when replacing the statement with
597 the same one; always update histograms.
599 2007-02-08 Diego Novillo <dnovillo@redhat.com>
601 * passes.c (init_optimization_passes): Tidy comment.
603 2007-02-08 Roger Sayle <roger@eyesopen.com>
605 * simplify-rtx.c (simplify_unary_operation_1) <POPCOUNT>: We can
606 strip zero_extend, bswap and rotates from POCOUNT's argument.
607 <PARITY>: Likewise, we can strip not, bswap, sign_extend,
608 zero_extend and rotates from PARITY's argument.
609 <BSWAP>: A byte-swap followed by a byte-swap is an identity.
610 (simplify_const_unary_operation) <BSWAP>: Evaluate the byte-swap
611 of an integer constant at compile-time.
613 2007-02-08 Diego Novillo <dnovillo@redhat.com>
616 * tree-flow.h (struct var_ann_d): Remove field 'is_used'.
618 * tree-ssa-alias.c (compute_is_aliased): Remove. Update all
621 * tree-ssa-live.c (remove_unused_locals): Do not remove
622 TREE_ADDRESSABLE variables.
623 * tree-ssa-structalias.c (compute_points_to_sets): Tidy.
624 * tree-ssa-operands.c (add_virtual_operand): Remove argument
625 FOR_CLOBBER. Update all users.
626 If VAR has an associated alias set, add a virtual operand for
627 it if no alias is found to conflict with the memory reference.
629 2007-02-07 Jan Hubicka <jh@suse.cz>
630 Robert Kidd <rkidd@crhc.uiuc.edu>
632 * value-prof.c (visit_hist, free_hist): Return 1 instead of 0.
634 2007-02-07 Ian Lance Taylor <iant@google.com>
636 * lower-subreg.c (simple_move): Reject PARTIAL_INT modes.
638 2007-02-07 Roger Sayle <roger@eyesopen.com>
640 * config/rs6000/rs6000.md (ctz<mode>2, ffs<mode>2, popcount<mode>2,
641 parity<mode>2, smulsi3_highpart, abstf2_internal, allocate_stack,
642 tablejumpdi, movsi_to_cr_one): Remove constraints from
643 define_expand's match_operands.
645 2007-02-07 Roger Sayle <roger@eyesopen.com>
647 * global.c (compute_regsets): Move declatation of "i" inside of
648 #ifdef ELIMINABLE_REGS to avoid unused variable bootstrap failure.
650 2007-02-07 Jakub Jelinek <jakub@redhat.com>
653 * gimplify.c (gimplify_scan_omp_clauses): Remove special casing
654 of INDIRECT_REF <RESULT_DECL>.
656 * config/i386/i386.c (override_options): Set PTA_SSSE3 for core2.
658 2007-02-06 J"orn Rennecke <joern.rennecke@arc.com>
659 Kaz Kojima <kkojima@gcc.gnu.org>
662 * config/sh/sh.c (expand_cbranchdi4): Use scratch register
664 (sh_initialize_trampoline): Add parentheses to avoid a warning.
666 2007-02-06 Zdenek Dvorak <dvorakz@suse.cz>
668 * doc/loop.texi: Document possibility not to perform disambiguation
669 of loops with multiple latches.
670 * cfgloopmanip.c (alp_enum_p): Removed.
671 (add_loop): Handle subloops. Use get_loop_body_with_size.
672 (create_preheader): Do not allow ENTRY_BLOCK_PTR to be preheader.
673 * cfghooks.c (redirect_edge_and_branch_force): Set dominator for
674 the new forwarder block.
675 (make_forwarder_block): Only call new_bb_cbk if it is not NULL.
676 Handle the case latch is NULL.
677 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Avoid cfg modifications
678 when marking loop exits.
679 * ifcvt.c (if_convert): Ditto. Mark loop exits even if cfg cannot
681 * loop-init.c (loop_optimizer_init): Do not modify cfg. Call
682 disambiguate_loops_with_multiple_latches.
683 * tree-cfgcleanup.c (cleanup_tree_cfg_loop): Calculate dominators
684 before fix_loop_structure.
685 * cfgloop.c: Include pointer-set.h and output.h.
686 (canonicalize_loop_headers, HEADER_BLOCK, LATCH_EDGE,
687 update_latch_info, mfb_keep_just, mfb_keep_nonlatch): Removed.
688 (get_loop_latch_edges, find_subloop_latch_edge_by_profile,
689 find_subloop_latch_edge_by_ivs, find_subloop_latch_edge,
690 mfb_redirect_edges_in_set, form_subloop, merge_latch_edges,
691 disambiguate_multiple_latches, get_loop_body_with_size,
692 disambiguate_loops_with_multiple_latches): New functions.
693 (flow_loop_dump): Dump multiple latch edges.
694 (flow_loop_nodes_find): Handle loops with multiple latches.
695 (flow_loops_find): Ditto. Do not call canonicalize_loop_headers.
696 (glb_enum_p): Modified.
697 (get_loop_body): Use get_loop_body_with_size.
698 * cfgloop.h (LOOPS_HAVE_RECORDED_EXITS): New flag.
699 (AVOID_CFG_MODIFICATIONS): New constant.
700 (disambiguate_loops_with_multiple_latches, add_loop,
701 get_loop_body_with_size): Declare.
702 * Makefile.in (cfgloop.o): Add pointer-set.h and output.h.
704 2007-02-06 Seongbae Park <seongbae.park@gmail.com>
707 * global.c (compute_regsets): New function.
708 (global_alloc): Refactored ELIMINABLE_REGSET
709 and NO_GLOBAL_ALLOC_REGS computation out.
710 (rest_of_handle_global_alloc): Call compute_regsets()
711 for non-optimizing case.
713 2007-02-06 Richard Henderson <rth@redhat.com>
715 * config/i386/constraints.md (Y2): Rename from Y.
716 (Yi): New constraint.
717 * config/i386/i386.md (movsi_1, movdi_2, pushdf_nointeger,
718 pushdf_integer, movdf_nointeger, movdf_integer, zero_extendsidi2_32,
719 zero_extendsidi2_rex64, truncxfdf2_mixed): Change Y constraints to Y2.
720 (extendsfdf2_mixed, extendsfdf2_sse, truncdfsf_fast_mixed,
721 truncdfsf_fast_sse, truncdfsf_mixed, fix_truncdfdi_sse,
722 fix_truncdfsi_sse, floatsidf2_mixed, floatsidf2_sse,
723 floatdidf2_mixed, floatdidf2_sse, absnegdf2_mixed,
724 absnegdf2_sse, sse_setccdf, fop_df_comm_mixed, fop_df_comm_sse,
725 fop_df_1_mixed, fop_df_1_sse): Change Y constraints to x.
726 * config/i386/mmx.md (mov<MMXMODEI>_internal_rex64,
727 mov<MMXMODEI>_internal, movv2sf_internal_rex64, movv2sf_internal,
728 vec_extractv2si_1): Change Y constraints to Y2.
729 * config/i386/sse.md (vec_setv4sf_0, vec_concatv2df, vec_dupv4si,
730 vec_dupv2di, sse2_concatv2si, vec_concatv4si_1, vec_concatv2di):
731 Change Y constraints to Y2.
732 (sse2_loadld): Change Y constraints to x.
734 2007-02-06 Roger Sayle <roger@eyesopen.com>
736 * config/rs6000/rs6000.md (popcount<mode>2): Rewrite.
737 (parity<mode>2): New define_expand using rs6000_emit_parity.
738 * config/rs6000/rs6000.c (rs6000_emit_popcount,
739 rs6000_emit_parity): New functions.
740 * config/rs6000/rs6000-protos.h (rs6000_emit_popcount,
741 rs6000_emit_parity): Prototype here.
743 2007-02-06 Ian Lance Taylor <iant@google.com>
745 * lower-subreg.c (simple_move_operand): Reject CONST.
746 (resolve_clobber): Call validate_change rather than directly
747 assigning to XEXP (pat, 0).
749 2006-02-06 Paolo Bonzini <bonzini@gnu.org>
751 * Makefile.in (tree-ssa-loop-ivopts.o): Add pointer-set.h dependency.
752 (tree-ssa-reassoc.o): Add pointer-set.h dependency.
753 (tree-cfg.o): Remove hashtab.h dependency.
755 * tree-ssa-loop-ivopts.c: Include pointer-set.h.
756 (struct ivopts_data): Change niters to pointer_map_t.
757 (struct nfe_cache_elt, nfe_hash, nfe_eq): Delete.
758 (niter_for_exit): Create pointer_map on demand. Change for
760 (tree_ssa_iv_optimize_init): Initialize data->niters to NULL.
761 (free_loop_data): Destroy data->niters if created and reset field.
762 (tree_ssa_iv_optimize_finalize): Don't delete data->niters here.
763 (tree_ssa_iv_optimize_loop): Check for presence of stale data.
765 * tree-ssa-reassoc.c: Include pointer-set.h.
766 (bb_rank): Change to long *.
767 (operand_rank): Change to pointer_map_t.
768 (find_operand_rank): Return long, -1 if not found. Declare as inline.
769 (insert_operand_rank): Accept long.
770 (operand_entry_hash, operand_entry_eq): Remove.
771 (get_rank): Return long. Adjust for changes above.
772 (init_reassoc): Change rank type to long. Adjust creation of bb_rank
774 (fini_reassoc): Delete operand_rank with pointer_map_destroy.
776 * tree-ssa-structalias.c (vi_for_tree): Change to pointer_map.
777 (struct tree_vi, tree_vi_t, tree_vi_hash, tree_vi_eq): Delete.
778 (insert_vi_for_tree): Rewrite for pointer_map API. Assert argument
780 (lookup_vi_for_tree): Rewrite for pointer_map API. Return varinfo_t
781 directly since it cannot be NULL.
782 (get_vi_for_tree): Rewrite for pointer_map API.
783 (find_what_p_points_to): Adjust for change to lookup_vi_for_tree.
784 (init_alias_vars): Create vi_for_tree as pointer_map.
785 (delete_points_to_sets): Delete vi_for_tree using pointer_map_destroy.
787 * tree-cfg.c: Don't include hashtab.h.
788 (edge_to_cases): Declare as pointer_map.
789 (struct edge_to_cases_elt, edge_to_cases_hash, edge_to_cases_eq):
791 (edge_to_cases_cleanup): Rewrite as pointer_map_traverse callback.
792 (start_recording_case_labels): Create edge_to_cases as pointer_map.
793 (end_recoding_case_labels): Cleanup edge_to_cases manually before
795 (record_switch_edge): Delete.
796 (get_cases_for_edge): Adjust for pointer_map API, inline
797 record_switch_edge (rewritten for new API), remove goto.
799 2006-02-06 Paolo Bonzini <bonzini@gnu.org>
801 * Makefile.in (tree-nested.o): Add pointer-set.h dependency.
802 * tree-nested.c: Include pointer-set.h.
803 (var_map_elt, var_map_eq, var_map_hash): Delete.
804 (struct nesting_info): Remove GTY marker. Change the two htab_t's
806 (nesting_info_bitmap_obstack): New.
807 (lookup_field_for_decl): Adjust for pointer_map API.
808 (lookup_tramp_for_decl): Adjust for pointer_map API.
809 (get_nonlocal_debug_decl): Adjust for pointer_map API.
810 (get_local_debug_decl): Adjust for pointer_map API.
811 (convert_nl_goto_reference): Adjust for pointer_map API.
812 (convert_nl_goto_receiver): Adjust for pointer_map API.
813 (create_nesting_tree): Create outside GGC space. Create bitmap on
814 the new obstack. Create field_map and var_map as pointer_maps.
815 (free_nesting_tree): Adjust for changes to create_nesting_tree.
817 (lower_nested_functions): Move root here, no need to NULL it.
818 Initialize and release the obstack.
820 2007-02-06 Paolo Bonzini <bonzini@gnu.org>
822 * tree.c (tree_int_map_hash, tree_int_map_eq, tree_int_map_marked_p):
823 Remove prototypes and make them non-static.
824 (struct tree_int_map): Remove.
825 * tree.h (struct tree_int_map): Move here, turning TO into an
827 (tree_int_map_hash, tree_int_map_eq, tree_int_map_marked_p): Declare.
829 * tree.h (TREE_COMPLEXITY): Remove.
830 (struct tree_exp): Remove complexity field.
831 * tree.c (build1_stat): Don't set it.
833 2007-02-06 Dorit Nuzman <dorit@il.ibm.com>
834 Victor Kaplansky <victork@il.ibm.com>
836 * tree-vectorizer.c (vect_is_simple_use): Support induction.
837 (vect_is_simple_reduction): Support reduction with induction as
839 (vect_is_simple_iv_evolution): Fix formatting.
840 * tree-vect-analyze.c (vect_mark_stmts_to_be_vectorized): Fix
841 formatting. Don't mark induction phis for vectorization.
842 (vect_analyze_scalar_cycles): Analyze all inductions, then reductions.
843 * tree-vect-transform.c (get_initial_def_for_induction): New function.
844 (vect_get_vec_def_for_operand): Support induction.
845 (vect_get_vec_def_for_stmt_copy): Fix formatting and add check for
847 (vectorizable_reduction): Support reduction with induction as one of
849 (vectorizable_type_demotion): Use def-type of stmt argument rather
852 * tree-ssa-loop.c (gate_scev_const_prop): Return the value of
853 flag_tree_scev_cprop.
854 * common.opt (tree-scev-cprop): New flag.
856 * tree-vect-transform.c (vect_create_destination_var): Use 'kind' in
857 call to vect_get_new_vect_var.
859 2007-02-06 Ira Rosen <irar@il.ibm.com>
861 * tree-vect-patterns.c (vect_recog_widen_mult_pattern): Check that
863 (vect_pattern_recog_1): Likewise.
865 2007-02-05 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
867 * fold-const.c (negate_expr_p): Handle CONJ_EXPR.
868 (fold_negate_expr): Likewise.
870 2007-02-05 Alexandre Oliva <aoliva@redhat.com>
873 * dwarf2out.c (modified_type_die): Follow DECL_ORIGINAL_TYPE
874 even if cv-qualification is the same.
876 2007-02-05 Geoffrey Keating <geoffk@apple.com>
878 * config/rs6000/darwin-tramp.asm (__trampoline_setup): Call
879 __enable_execute_stack on completion.
881 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
883 * config/i386/athlon.md (athlon_fldxf_k8, athlon_fld_k8,
884 athlon_fstxf_k8, athlon_fst_k8, athlon_fist, athlon_fmov,
885 athlon_fadd_load, athlon_fadd_load_k8, athlon_fadd, athlon_fmul,
886 athlon_fmul_load, athlon_fmul_load_k8, athlon_fsgn,
887 athlon_fdiv_load, athlon_fdiv_load_k8, athlon_fdiv_k8,
888 athlon_fpspc_load, athlon_fpspc, athlon_fcmov_load,
889 athlon_fcmov_load_k8, athlon_fcmov_k8, athlon_fcomi_load_k8,
890 athlon_fcomi, athlon_fcom_load_k8, athlon_fcom): Added amdfam10.
892 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
894 * config/i386/i386.md (x86_sahf_1, cmpfp_i_mixed, cmpfp_i_sse,
895 cmpfp_i_i387, cmpfp_iu_mixed, cmpfp_iu_sse, cmpfp_iu_387,
896 swapsi, swaphi_1, swapqi_1, swapdi_rex64, fix_truncsfdi_sse,
897 fix_truncdfdi_sse, fix_truncsfsi_sse, fix_truncdfsi_sse,
898 x86_fldcw_1, floatsisf2_mixed, floatsisf2_sse, floatdisf2_mixed,
899 floatdisf2_sse, floatsidf2_mixed, floatsidf2_sse,
900 floatdidf2_mixed, floatdidf2_sse, muldi3_1_rex64, mulsi3_1,
901 mulsi3_1_zext, mulhi3_1, mulqi3_1, umulqihi3_1, mulqihi3_insn,
902 umulditi3_insn, umulsidi3_insn, mulditi3_insn, mulsidi3_insn,
903 umuldi3_highpart_rex64, umulsi3_highpart_insn,
904 umulsi3_highpart_zext, smuldi3_highpart_rex64,
905 smulsi3_highpart_insn, smulsi3_highpart_zext, x86_64_shld,
906 x86_shld_1, x86_64_shrd, sqrtsf2_mixed, sqrtsf2_sse,
907 sqrtsf2_i387, sqrtdf2_mixed, sqrtdf2_sse, sqrtdf2_i387,
908 sqrtextendsfdf2_i387, sqrtxf2, sqrtextendsfxf2_i387,
909 sqrtextenddfxf2_i387): Added amdfam10_decode.
911 * config/i386/athlon.md (athlon_idirect_amdfam10,
912 athlon_ivector_amdfam10, athlon_idirect_load_amdfam10,
913 athlon_ivector_load_amdfam10, athlon_idirect_both_amdfam10,
914 athlon_ivector_both_amdfam10, athlon_idirect_store_amdfam10,
915 athlon_ivector_store_amdfam10): New define_insn_reservation.
916 (athlon_idirect_loadmov, athlon_idirect_movstore): Added
919 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
921 * config/i386/athlon.md (athlon_call_amdfam10,
922 athlon_pop_amdfam10, athlon_lea_amdfam10): New
923 define_insn_reservation.
924 (athlon_branch, athlon_push, athlon_leave_k8, athlon_imul_k8,
925 athlon_imul_k8_DI, athlon_imul_mem_k8, athlon_imul_mem_k8_DI,
926 athlon_idiv, athlon_idiv_mem, athlon_str): Added amdfam10.
928 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
930 * config/i386/athlon.md (athlon_sseld_amdfam10,
931 athlon_mmxld_amdfam10, athlon_ssest_amdfam10,
932 athlon_mmxssest_short_amdfam10): New define_insn_reservation.
934 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
936 * config/i386/athlon.md (athlon_sseins_amdfam10): New
937 define_insn_reservation.
938 * config/i386/i386.md (sseins): Added sseins to define_attr type
939 and define_attr unit.
940 * config/i386/sse.md: Set type attribute to sseins for insertq
943 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
945 * config/i386/athlon.md (sselog_load_amdfam10, sselog_amdfam10,
946 ssecmpvector_load_amdfam10, ssecmpvector_amdfam10,
947 ssecomi_load_amdfam10, ssecomi_amdfam10,
948 sseaddvector_load_amdfam10, sseaddvector_amdfam10): New
949 define_insn_reservation.
950 (ssecmp_load_k8, ssecmp, sseadd_load_k8, seadd): Added amdfam10.
952 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
954 * config/i386/athlon.md (cvtss2sd_load_amdfam10,
955 cvtss2sd_amdfam10, cvtps2pd_load_amdfam10, cvtps2pd_amdfam10,
956 cvtsi2sd_load_amdfam10, cvtsi2ss_load_amdfam10,
957 cvtsi2sd_amdfam10, cvtsi2ss_amdfam10, cvtsd2ss_load_amdfam10,
958 cvtsd2ss_amdfam10, cvtpd2ps_load_amdfam10, cvtpd2ps_amdfam10,
959 cvtsX2si_load_amdfam10, cvtsX2si_amdfam10): New
960 define_insn_reservation.
962 * config/i386/sse.md (cvtsi2ss, cvtsi2ssq, cvtss2si,
963 cvtss2siq, cvttss2si, cvttss2siq, cvtsi2sd, cvtsi2sdq,
964 cvtsd2si, cvtsd2siq, cvttsd2si, cvttsd2siq,
965 cvtpd2dq, cvttpd2dq, cvtsd2ss, cvtss2sd,
966 cvtpd2ps, cvtps2pd): Added amdfam10_decode attribute.
968 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
970 * config/i386/athlon.md (athlon_ssedivvector_amdfam10,
971 athlon_ssedivvector_load_amdfam10, athlon_ssemulvector_amdfam10,
972 athlon_ssemulvector_load_amdfam10): New define_insn_reservation.
973 (athlon_ssediv, athlon_ssediv_load_k8, athlon_ssemul,
974 athlon_ssemul_load_k8): Added amdfam10.
976 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
978 * config/i386/i386.h (TARGET_SSE_UNALIGNED_MOVE_OPTIMAL): New macro.
979 (x86_sse_unaligned_move_optimal): New variable.
981 * config/i386/i386.c (x86_sse_unaligned_move_optimal): Enable for
983 (ix86_expand_vector_move_misalign): Add code to generate movupd/movups
984 for unaligned vector SSE double/single precision loads for AMDFAM10.
986 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
988 * config/i386/i386.h (TARGET_AMDFAM10): New macro.
989 (TARGET_CPU_CPP_BUILTINS): Add code for amdfam10.
990 Define TARGET_CPU_DEFAULT_amdfam10.
991 (TARGET_CPU_DEFAULT_NAMES): Add amdfam10.
992 (processor_type): Add PROCESSOR_AMDFAM10.
994 * config/i386/i386.md: Add amdfam10 as a new cpu attribute to match
995 processor_type in config/i386/i386.h.
996 Enable imul peepholes for TARGET_AMDFAM10.
998 * config.gcc: Add support for --with-cpu option for amdfam10.
1000 * config/i386/i386.c (amdfam10_cost): New variable.
1001 (m_AMDFAM10): New macro.
1002 (m_ATHLON_K8_AMDFAM10): New macro.
1003 (x86_use_leave, x86_push_memory, x86_movx, x86_unroll_strlen,
1004 x86_cmove, x86_3dnow_a, x86_deep_branch, x86_use_simode_fiop,
1005 x86_promote_QImode, x86_integer_DFmode_moves,
1006 x86_partial_reg_dependency, x86_memory_mismatch_stall,
1007 x86_accumulate_outgoing_args, x86_arch_always_fancy_math_387,
1008 x86_sse_partial_reg_dependency, x86_sse_typeless_stores,
1009 x86_use_ffreep, x86_use_incdec, x86_four_jump_limit,
1010 x86_schedule, x86_use_bt, x86_cmpxchg16b, x86_pad_returns):
1011 Enable/disable for amdfam10.
1012 (override_options): Add amdfam10_cost to processor_target_table.
1013 Set up PROCESSOR_AMDFAM10 for amdfam10 entry in
1014 processor_alias_table.
1015 (ix86_issue_rate): Add PROCESSOR_AMDFAM10.
1016 (ix86_adjust_cost): Add code for amdfam10.
1018 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
1020 * config/i386/i386.opt: Add new Advanced Bit Manipulation (-mabm)
1021 instruction set feature flag. Add new (-mpopcnt) flag for popcnt
1022 instruction. Add new SSE4A (-msse4a) instruction set feature flag.
1023 * config/i386/i386.h: Add builtin definition for SSE4A.
1024 * config/i386/i386.md: Add support for ABM instructions
1026 * config/i386/sse.md: Add support for SSE4A instructions
1027 (movntss, movntsd, extrq, insertq).
1028 * config/i386/i386.c: Add support for ABM and SSE4A builtins.
1029 Add -march=amdfam10 flag.
1030 * config/i386/ammintrin.h: Add support for SSE4A intrinsics.
1031 * doc/invoke.texi: Add documentation on flags for sse4a, abm, popcnt
1033 * doc/extend.texi: Add documentation for SSE4A builtins.
1035 2007-02-05 Bob Wilson <bob.wilson@acm.org>
1037 * config/xtensa/xtensa.c (constantpool_mem_p): Skip over SUBREGs.
1039 2007-02-05 Richard Guenther <rguenther@suse.de>
1041 * tree-vectorizer.h (vectorizable_function): Add argument type
1042 argument, change return type.
1043 * tree-vect-patterns.c (vect_recog_pow_pattern): Adjust caller.
1044 * tree-vect-transform.c (vectorizable_function): Handle extra
1045 argument, return vectorized function decl.
1046 (build_vectorized_function_call): Remove.
1047 (vectorizable_call): Handle calls with result and argument types
1048 differing. Handle loop vectorization factor correctly.
1049 * targhooks.c (default_builtin_vectorized_function): Adjust for
1051 * targhooks.h (default_builtin_vectorized_function): Likewise.
1052 * target.h (builtin_vectorized_function): Add argument type
1054 * config/i386/i386.c (ix86_builtin_vectorized_function): Handle
1055 extra argument, allow vectorizing of lrintf.
1056 * doc/tm.texi (TARGET_VECTORIZE_BUILTIN_VECTORIZED_FUNCTION): Adjust
1057 documentation of target hook.
1059 2007-02-05 Hans-Peter Nilsson <hp@axis.com>
1062 * config/cris/cris.md ("*andsi_movu", "*andsi_clear", "*andhi_movu")
1063 ("*andhi_clear", andu (casesi+45)): For size-changed operand where
1064 memory is allowed, require !side_effects_p, not just !MEM_VOLATILE_P.
1066 2007-02-05 Roger Sayle <roger@eyesopen.com>
1068 * fold-const.c (fold_unary) <REAL_PART>: Test for availability of
1069 BUILT_IN_COS before simplifying REAL_PART(CEXPI)) to COS.
1070 <IMAG_PART>: Likewise, check for availability of BUILT_IN_SIN.
1071 * builtins.c (fold_builtin_sincos): Check for TARGET_C99_FUNCTIONS
1072 before canonicalizing sincos to cexpi.
1073 (fold_builtin_cexp): Likewise, for canonicalizing cexp to cexpi.
1075 2007-02-05 Roger Sayle <roger@eyesopen.com>
1077 * config/alpha/alpha.c (alpha_add_builtins): New Helper function.
1078 Set TREE_READONLY and TREE_NOTHROW directly, not via attributes.
1079 (alpha_init_builtins): Use alpha_add_builtins to process tables.
1081 2007-02-05 Roger Sayle <roger@eyesopen.com>
1083 * mips-tfile.c (initialize_init_file): Correct endianness test.
1085 2007-02-05 Kazu Hirata <kazu@codesourcery.com>
1087 * config/m68k/m68k.md (pushdi-1, pushdi, movsi+1): Don't use
1090 2007-02-05 Richard Sandiford <richard@codesourcery.com>
1092 * dwarf2out.c (dwarf2out_frame_debug_expr): Record the register
1093 saves in a PARALLEL before the register assignments.
1095 2007-02-05 Richard Sandiford <richard@codesourcery.com>
1097 * doc/tm.texi (DWARF_ALT_FRAME_RETURN_COLUMN): Do not require
1098 DWARF_FRAME_RETURN_COLUMN to be a general register.
1099 * dwarf2out.c (init_return_column_size): New function, split from...
1100 (expand_builtin_init_dwarf_reg_sizes): ...here. Allow both
1101 DWARF_FRAME_RETURN_COLUMN and DWARF_ALT_FRAME_RETURN_COLUMN
1102 to be nongeneral registers.
1103 * config/m68k/m68k.h (DWARF_FRAME_REGNUM): Only map FP and
1105 (DWARF_FRAME_REGISTERS, DWARF_FRAME_RETURN_COLUMN): Define.
1106 (DWARF_ALT_FRAME_RETURN_COLUMN): Define.
1108 2007-02-04 Zdenek Dvorak <dvorakz@suse.cz>
1110 * cfgcleanup.c (try_optimize_cfg): Avoid removing ENTRY_BLOCK_PTR.
1112 2007-02-04 Zdenek Dvorak <dvorakz@suse.cz>
1114 * cfgloopmanip.c (loop_delete_branch_edge): Removed.
1115 (remove_path): Use can_remove_branch_p and remove_branch instead
1116 of loop_delete_branch_edge.
1117 * tree-ssa-loop-manip.c (scale_dominated_blocks_in_loop): New function.
1118 (tree_transform_and_unroll_loop): Remove dead branches immediately.
1119 Update profile using scale_dominated_blocks_in_loop.
1120 * cfghooks.c (can_remove_branch_p, remove_branch): New functions.
1121 * cfghooks.h (struct cfg_hooks): Add can_remove_branch_p.
1122 (can_remove_branch_p, remove_branch): Declare.
1123 * tree-cfg.c (tree_can_remove_branch_p): New function.
1124 (tree_cfg_hooks): Add tree_can_remove_branch_p.
1125 * cfgrtl.c (rtl_can_remove_branch_p): New function.
1126 (rtl_cfg_hooks, cfg_layout_rtl_cfg_hook): Add rtl_can_remove_branch_p.
1128 2007-02-05 Jan Hubicka <jh@suse.cz>
1131 * ipa-inline.c (cgraph_clone_inlined_nodes): When there are unanalyzed
1132 nodes in cgraph, don't remove offline copy of the function.
1134 2007-02-04 Jan Hubicka <jh@suse.cz>
1136 * tree-sra.c (sra_walk_expr): Add linebreaks. BITFIELD_REFs into
1137 vectors might cause maybe_lookup_element_for_expr to be called
1138 on non-sra-candidate.
1140 2007-02-04 Kazu Hirata <kazu@codesourcery.com>
1142 * config/bfin/bfin-modes.def, config/bfin/bfin.c,
1143 config/bfin/bfin.md, config/bfin/predicates.md: Follow
1144 spelling conventions.
1146 2007-02-04 Richard Guenther <rguenther@suse.de>
1149 * fold-const.c (try_move_mult_to_index): Make sure to not
1150 overflow one dimension of a multi-dimensional array access.
1152 2007-02-04 Jan Hubicka <jh@suse.cz>
1154 * passes.c (init_optimization_passes): Reindent.
1156 2007-02-04 Jan Hubicka <jh@suse.cz>
1157 Eric Botcazou <ebotcazou@adacore.com>
1159 * tree-optimize.c (has_abnormal_outgoing_edge_p): Move to...
1160 (execute_fixup_cfg): Break out the abnormal goto code.
1161 * tree-inline.c (has_abnormal_outgoing_edge_p): ...here.
1162 (make_nonlocal_label_edges): Move here from execute_fixup_cfg.
1163 (optimize_inline_calls): Call make_nonlocal_label_edges.
1165 2007-02-04 Jan Hubicka <jh@suse.cz>
1167 * tree-ssa-copyrename.c (copy_rename_partition_coalesce): Return
1168 true when something was changed.
1169 (rename_ssa_copies): When something was changed, do
1170 TODO_remove_unused_locals.
1171 * tree-ssa-forwprop.c (tree_ssa_forward_propagate_single_use_value):
1172 add TODO_remove_unused_locals when instruction was removed.
1174 2007-02-04 Jan Hubicka <jh@suse.cz>
1176 * ipa-inline.c (try_inline): Improve debug output; work on already
1178 (cgraph_decide_inlining_incrementally): Indent; improve debug output;
1179 call try_inline for already inlined edges too when flattening;
1180 inline also functions that make callee growth but overall unit size
1183 2007-02-04 Kazu Hirata <kazu@codesourcery.com>
1185 * config/m32c/bitops.md, config/m32c/jump.md,
1186 config/m32c/m32c.c, config/m32c/m32c.h, config/m32r/m32r.c,
1187 config/m32r/m32r.h, config/m32r/m32r.md,
1188 config/m32r/predicates.md, config/m68hc11/larith.asm,
1189 config/m68hc11/m68hc11.c, config/m68hc11/m68hc11.h,
1190 config/m68k/m68k.h, config/mcore/mcore.md, config/mips/4k.md,
1191 config/mips/mips-protos.h, config/mips/mips.c,
1192 config/mips/mips.h, config/mips/mips.md, config/mips/mips16.S,
1193 config/mn10300/mn10300.h, config/mn10300/predicates.md,
1194 config/mt/mt.c, config/mt/mt.h, config/mt/mt.md: Follow
1195 spelling conventions.
1197 * config/v850/v850.c, config/v850/v850.h, config/v850/v850.md:
1198 Follow spelling conventions.
1200 2007-02-03 Douglas Gregor <doug.gregor@gmail.com>
1202 * c-opts.c (c_common_post_options): If C++0x mode is enabled, don't
1203 warn about C++0x compatibility.
1205 2007-02-04 Kazu Hirata <kazu@codesourcery.com>
1207 * config/h8300/h8300.c, config/h8300/h8300.h,
1208 config/h8300/h8300.md: Follow spelling conventions.
1210 2007-02-03 Uros Bizjak <ubizjak@gmail.com>
1213 * combine.c (try_combine): Do not substitute source operand
1214 with constants wider than 2 * HOST_BITS_PER_WIDE_INT.
1216 2007-02-03 Jan Hubicka <jh@suse.cz>
1218 PR gcov-profile/30650
1219 * value-prof.c (stringop_block_profile): Fix handling of size counter;
1220 do not divide by zero for never executed counters.
1221 (tree_find_values_to_profile): Fix counters.
1222 * gcov-ui.h (GCOV_COUNTER_AVERAGE, GCOV_COUNTER_IOR): Fix comments.
1224 2007-02-03 Ian Lance Taylor <iant@google.com>
1226 * lower-subreg.c (simple_move_operand): New static function,
1227 broken out of simple_move. Reject LABEL_REF, SYMBOL_REF, and HIGH
1229 (simple_move): Call simple_move_operand.
1230 (find_decomposable_subregs): Add special handling of MEMs.
1231 (can_decompose_p): Rename from cannot_decompose_p. Reverse
1232 meaning of return value. If we see a hard register, test whether
1233 it can store a word_mode value. Change all callers.
1235 2007-02-03 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
1237 * pa.md (addvdi3, addvsi3, subvdi3, subvsi3, negvdi2, negvsi2): New
1238 ftrapv insns and expanders.
1239 (subdi3): Change define_expand operand 1 to arith11_operand, and
1240 operand 2 to reg_or_0_operand. Change constraints of 64-bit insn
1241 pattern to handle reg_or_0 operands. Revise 32-bit insn pattern to
1242 handle 11-bit constants and reg_or_0 operands in operands 1 and 2,
1246 * varasm.c (notice_global_symbol): Treat global objects as weak when
1249 2007-02-03 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1251 * emit-rtl.c (dconstpi): Delete.
1253 (init_emit_once): Delete dconstpi and init dconstsqrt2.
1254 * real.h (dconstpi): Delete.
1256 * builtins.c (fold_builtin_cabs): Use dconstsqrt2.
1257 (fold_builtin_hypot): Likewise.
1259 2007-02-03 Tom Tromey <tromey@redhat.com>
1262 * gcc.c (cpp_unique_options): Any of -ggdb3, -gstabs3,
1263 -gcoff3, -gxcoff3, -gvms3 implies -dD.
1265 2007-02-03 Kazu Hirata <kazu@codesourcery.com>
1267 * c-decl.c, config/avr/avr.c, config/avr/avr.h,
1268 config/m68k/m68k.c, config/m68k/netbsd-elf.h,
1269 config/mn10300/mn10300.c, config/pdp11/pdp11.h,
1270 config/rs6000/cell.md, config/rs6000/darwin.h,
1271 config/sh/sh.md, config/sh/sh4-300.md, config/spu/spu.c,
1272 config/spu/spu.md, cselib.c, expr.c, haifa-sched.c, hwint.h,
1273 jump.c, reload.c, sched-deps.c, sched-int.h, tree-inline.c,
1274 tree-profile.c, tree-ssa-live.h, tree-vrp.c: Fix comment
1275 typos. Follow spelling conventions.
1276 * doc/invoke.texi: Follow spelling conventions.
1278 2007-02-03 Roger Sayle <roger@eyesopen.com>
1280 * simplify-rtx.c (simplify_relational_operation_1): Implement some
1281 canonicalization transformations that attempt to simplify integer
1282 constant comparisons to become comparisons against zero.
1284 2007-02-03 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1286 * builtins.c (fold_builtin_cabs): Fold cabs(x+xi) into
1288 * fold-const.c (fold_binary): Fix comment typos. Fold complex
1289 (x,0)-(0,y) into (x,-y). Likewise (0,y)-(x,0) into (-x,y).
1291 2007-02-02 Mike Stump <mrs@apple.com>
1293 * config/darwin9.h (DARWIN_LINKER_GENERATES_ISLANDS): Add.
1294 * config/rs6000/rs6000.c (DARWIN_GENERATE_ISLANDS): Add.
1295 (output_call): Use DARWIN_GENERATE_ISLANDS to decide when to
1296 generate a branch island.
1298 2007-02-02 Bob Wilson <bob.wilson@acm.org>
1300 * config/xtensa/xtensa.c (smalloffset_mem_p): Use BASE_REG_P.
1301 (xtensa_legitimate_address_p): New.
1302 (xtensa_legitimize_address): New.
1303 (xtensa_output_addr_const_extra): New.
1304 * config/xtensa/xtensa.h (REG_OK_STRICT_FLAG): Define.
1306 (REG_OK_FOR_BASE_P): Use BASE_REG_P.
1307 (GO_IF_LEGITIMATE_ADDRESS): Move code to xtensa_legitimate_address_p.
1308 (LEGITIMIZE_ADDRESS): Move code to xtensa_legitimize_address.
1309 (OUTPUT_ADDR_CONST_EXTRA): Move code to xtensa_output_addr_const_extra.
1310 * config/xtensa/xtensa-protos.h (xtensa_legitimate_address_p): New.
1311 (xtensa_legitimize_address): New.
1312 (xtensa_output_addr_const_extra): New.
1314 2007-02-02 Steve Ellcey <sje@cup.hp.com>
1316 * config/ia64/ia64.c (ia64_print_operand): Fix compare strings.
1318 2007-02-02 Ian Lance Taylor <iant@google.com>
1320 * expmed.c (expand_divmod): Add comment.
1322 2007-02-02 Kazu Hirata <kazu@codesourcery.com>
1324 * emit-rtl.c (renumber_insns): Remove.
1325 * flags.h: Remove the extern for flag_renumber_insns.
1326 * rtl.h: Remove the prototype for renumber_insns.
1327 * toplev.c (flag_renumber_insns): Remove.
1329 2007-02-02 Hui-May Chang <hm.chang@apple.com>
1331 Revert for x86 darwin:
1332 2005-06-19 Uros Bizjak <uros@kss-loka.si>
1334 * config/i386/i386.c (ix86_function_arg_regno_p): Put back the
1335 code before the following patch under TARGET_MACHO.
1336 (ix86_function_value_regno_p): Likewise.
1338 2007-02-02 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1340 * fold-const.c (negate_expr_p, fold_negate_expr): Handle
1343 2007-02-02 Maxim Kuvyrkov <mkuvyrkov@ispras.ru>
1346 * config/ia64/ia64.c (ia64_speculate_insn): Restrict to memory loads to
1347 general or fp registers. Add comments.
1348 * config/ia64/ia64.md (reg_pred_prefix): Add comment.
1350 2007-02-02 Paolo Bonzini <bonzini@gnu.org>
1352 * pointer-set.c (insert_aux): Only return insertion slot.
1353 (pointer_set_insert): Adjust.
1354 (pointer_set_traverse, struct pointer_map_t, pointer_map_create,
1355 pointer_map_destroy, pointer_map_insert, pointer_map_contains,
1356 pointer_map_traverse): New.
1357 * pointer-set.h (pointer_set_traverse, struct pointer_map_t,
1358 pointer_map_create, pointer_map_destroy, pointer_map_insert,
1359 pointer_map_contains, pointer_map_traverse): Declare.
1361 2007-02-02 Jakub Jelinek <jakub@redhat.com>
1364 * builtins.c (fold_builtin_sprintf): Do not attempt to optimize
1365 sprintf (str, "%s"). Do not optimize sprintf (str, "nopercent", p++).
1367 2007-02-02 Maxim Kuvyrkov <mkuvyrkov@ispras.ru>
1369 * sched-int.h (ds_to_dk, dk_to_ds): Declare functions.
1371 (struct _dep): New type.
1372 (dep_t): New typedef.
1373 (DEP_PRO, DEP_CON, DEP_KIND): New access macros.
1374 (DEP_STATUS): New access macro. The macro with the same name was
1375 renamed to DEP_LINK_STATUS.
1376 (dep_init): Declare function
1378 (struct _dep_link): New type.
1379 (dep_link_t): New typedef.
1380 (DEP_LINK_NODE, DEP_LINK_NEXT, DEP_LINK_PREV_NEXTP): New access macros.
1381 (DEP_LINK_DEP, DEP_LINK_PRO, DEP_LINK_CON, DEP_LINK_KIND): New macros.
1382 (DEP_LINK_STATUS): New macro.
1383 (debug_dep_links): New debug function.
1385 (struct _deps_list): New type.
1386 (deps_list_t): New typedef.
1387 (DEPS_LIST_FIRST): New access macro.
1388 (FOR_EACH_DEP_LINK): New cycle macro.
1389 (create_deps_list, free_deps_list, delete_deps_list): Declare
1391 (deps_list_empty_p, debug_deps_list, add_back_dep_to_deps_list): Ditto.
1392 (find_link_by_pro_in_deps_list, find_link_by_con_in_deps_list): Ditto.
1393 (copy_deps_list_change_con): Ditto.
1395 (move_dep_link): Declare function.
1397 (struct _dep_node): New type.
1398 (dep_node_t): New typedef.
1399 (DEP_NODE_BACK, DEP_NODE_DEP, DEP_NODE_FORW): New access macros.
1401 (struct haifa_insn_data.back_deps): New field to hold backward
1402 dependencies of the insn.
1403 (struct haifa_insn_data.depend): Rename to forw_deps. Change its type
1405 (struct haifa_insn_data.resolved_deps): Rename to resolved_back_deps.
1406 Change its type to deps_list_t.
1407 (INSN_BACK_DEPS): New access macro to use instead of LOG_LINKS.
1408 (INSN_DEPEND): Rename to INSN_FORW_DEPS.
1409 (RESOLVED_DEPS): Rename to INSN_RESOLVED_BACK_DEPS.
1411 (INSN_COST): Move to haifa-sched.c. Use insn_cost () instead.
1413 (DEP_STATUS): Rename to DEP_LINK_STATUS. Fix typo in the comment.
1415 (add_forw_dep, delete_back_forw_dep, insn_cost): Update declaration and
1417 (dep_cost): Declare.
1419 * sched-deps.c (CHECK): New macro to (en/dis)able sanity checks.
1420 (ds_to_dk, dk_to_ds): New functions.
1422 (init_dep_1): New static function.
1423 (init_dep): New function.
1424 (copy_dep): New static function.
1426 (dep_link_consistent_p, attach_dep_link, add_to_deps_list): New static
1428 (detach_dep_link): New static function.
1429 (move_dep_link): New function.
1431 (dep_links_consistent_p, dump_dep_links): New static functions.
1432 (debug_dep_links): New debugging function.
1434 (deps_obstack, dl_obstack, dn_obstack): New static variables.
1436 (alloc_deps_list, init_deps_list): New static functions.
1437 (create_deps_list): New function.
1438 (clear_deps_list): New static function.
1439 (free_deps_list, delete_deps_list, deps_list_empty_p): New functions.
1440 (deps_list_consistent_p, dump_deps_list): New static functions.
1441 (debug_deps_list): New function.
1442 (add_back_dep_to_deps_list, find_link_by_pro_in_deps_list): New
1444 (find_link_by_con_in_deps_list, copy_deps_list_change_con): Ditto.
1446 (maybe_add_or_update_back_dep_1, add_or_update_back_dep_1): Update to
1447 use new scheduler dependencies lists.
1448 (add_back_dep, delete_all_dependences, fixup_sched_groups): Ditto.
1449 (sched_analyze): Ditto. Initialize dependencies lists.
1450 (add_forw_dep, compute_forward_dependences): Update to use new
1451 scheduler dependencies lists.
1453 (init_dependency_caches): Init deps_obstack.
1454 (free_dependency_caches): Free deps_obstack.
1456 (adjust_add_sorted_back_dep, adjust_back_add_forw_dep): Update to use
1457 new scheduler dependencies lists.
1458 (delete_forw_dep, add_or_update_back_forw_dep): Ditto.
1459 (add_back_forw_dep, delete_back_forw_dep): Ditto.
1461 * sched-rgn.c (set_spec_fed, find_conditional_protection, is_pfree):
1462 Update to use new scheduler dependencies lists.
1463 (is_conditionally_protected, is_prisky, add_branch_dependences): Ditto.
1464 (debug_dependencies): Ditto.
1465 (schedule_region): Update comments.
1467 * sched-ebb.c (earliest_block_with_similiar_load): Update to use new
1468 scheduler dependencies lists.
1469 (schedule_ebb): Update comments.
1471 * rtl.def (DEPS_LIST): Remove.
1473 * lists.c (unused_deps_list): Remove.
1474 (free_list): Update assertions.
1476 (alloc_DEPS_LIST, free_DEPS_LIST_list, free_DEPS_LIST_node): Remove.
1477 (remove_free_DEPS_LIST_elem, copy_DEPS_LIST_list): Ditto.
1479 * rtl.h (free_DEPS_LIST_list, alloc_DEPS_LIST): Remove declarations.
1480 (remove_free_DEPS_LIST_elem, copy_DEPS_LIST_list): Ditto.
1482 * haifa-sched.c (comments): Update.
1483 (insn_cost1): Remove. Inline the code into insn_cost ().
1484 (insn_cost): Update to use new scheduler dependencies lists. Move
1485 processing of the dependency cost to dep_cost ().
1486 (dep_cost): New function. Use it instead of insn_cost () when
1487 evaluating cost of the dependency. Use compatible interface to
1488 interact with the target.
1489 (priority): Update to use new scheduler dependencies lists.
1490 (rank_for_schedule): Ditto. Optimize heuristic that prefers the insn
1491 with greater number of insns that depend on the insn.
1492 (schedule_insn): Update to use new scheduler dependencies lists. Add
1493 code to free backward dependencies lists. Inline and optimize code
1494 from resolve_dep () - see PR28071.
1495 (ok_for_early_queue_removal): Update to use new scheduler dependencies
1496 lists. Update call to targetm.sched.is_costly_dependence hook.
1498 (fix_inter_tick, try_ready, fix_tick_ready): Update to use new
1499 scheduler dependencies lists.
1501 (resolve_dep): Remove. Move the logic to schedule_insn ().
1502 (init_h_i_d): Initialize dependencies lists.
1504 (process_insn_depend_be_in_spec): Rename to
1505 process_insn_forw_deps_be_in_spec. Update to use new scheduler
1507 (add_to_speculative_block, create_check_block_twin, fix_recovery_deps):
1508 Update to use new scheduler dependencies lists.
1509 (clear_priorities, calc_priorities, add_jump_dependencies): Ditto.
1511 * ddg.c (create_ddg_dependence, create_ddg_dep_no_link): Update to use
1512 new scheduler dependencies lists.
1513 (build_intra_loop_deps): Ditto.
1515 * target.h (struct _dep): Declare to use in
1516 gcc_target.sched.is_costly_dependence.
1517 (struct gcc_target.sched.adjust_cost): Fix typo.
1518 (struct gcc_target.sched.is_costly_dependence): Change signature to use
1519 single dep_t parameter instead of an equivalent triad.
1520 (struct gcc_target.sched.adjust_cost_2): Remove.
1522 * target-def.h (TARGET_SCHED_ADJUST_COST_2): Remove.
1524 * reg-notes.def (DEP_TRUE, DEP_OUTPUT, DEP_ANTI): Update comments.
1526 * doc/tm.texi (TARGET_SCHED_IS_COSTLY_DEPENDENCE): Update
1528 (TARGET_SCHED_ADJUST_COST_2): Remove documentation.
1530 * doc/rtl.texi (LOG_LINKS): Remove part about instruction scheduler.
1531 (REG_DEP_TRUE): Document.
1533 * config/ia64/ia64.c (ia64_adjust_cost_2): Rename to ia64_adjust_cost.
1534 Change signature to correspond to the targetm.sched.adjust_cost hook.
1535 Update use in TARGET_SCHED_ADJUST_COST_2.
1536 (TARGET_SCHED_ADJUST_COST_2): Rename to TARGET_SCHED_ADJUST_COST.
1537 (ia64_dependencies_evaluation_hook, ia64_dfa_new_cycle): Update to use
1538 new scheduler dependencies lists.
1539 (ia64_gen_check): Ditto.
1541 * config/mips/mips.c (vr4130_swap_insns_p): Update to use new scheduler
1544 * config/rs6000/rs6000.c (rs6000_is_costly_dependence): Change
1545 signature to correspond to the targetm.sched.is_costly_dependence hook.
1546 (is_costly_group): Update to use new scheduler dependencies lists.
1548 * config/spu/spu.c (spu_sched_adjust_cost): Use insn_cost () function
1549 instead of INSN_COST () macro.
1551 2007-02-01 Ian Lance Taylor <iant@google.com>
1553 * lower-subreg.c (resolve_clobber): Handle a subreg of a concatn.
1555 2007-02-01 Guy Martin <gmsoft@gentoo.org>
1557 * pa.md (tp_load): Correct mfctl instruction syntax.
1559 2007-02-01 Geoffrey Keating <geoffk@apple.com>
1561 * config/rs6000/rs6000.c (rs6000_stack_info): Correct
1562 altivec_padding_size calculation on AIX. Improve comment, add
1563 assert to verify that it's right.
1565 * config/rs6000/darwin.h (MD_UNWIND_SUPPORT): Don't define for
1568 2007-2-01 Seongbae Park <seongbae.park@gmail.com>
1571 * global.c (global_alloc): Add mising initialization of
1574 2007-02-01 Roger Sayle <roger@eyesopen.com>
1576 * alias.c (init_alias_analysis): Correct whitespace.
1577 * bb-reorder.c (fix_edges_for_rarely_executed_code,
1578 partition_hot_cold_basic_blocks): Likewise.
1579 * builtins.c (expand_builtin_printf, expand_builtin_fprintf,
1580 expand_builtin_sprintf, fold_builtin_carg, fold_builtin_sprintf,
1581 maybe_emit_sprintf_chk_warning, fold_builtin_sprintf_chk,
1582 fold_builtin_snprintf_chk, fold_builtin_printf,
1583 fold_builtin_fprintf, do_mpfr_ckconv, do_mpfr_arg1, do_mpfr_arg2,
1584 do_mpfr_arg3, do_mpfr_sincos): Likewise.
1585 * cfgcleanup.c (cleanup_cfg): Likewise.
1586 * cfgexpand.c (tree_expand_cfg): Likewise.
1587 * fold-const.c (fold_binary) <RDIV_EXPR>: Likewise.
1588 * function.c (get_next_funcdef_no): Likewise.
1589 * gengtype.c (main): Likewise.
1590 * genmodes.c (main): Likewise.
1591 * gcse.c (bypass_conditional_jumps, print_ldst_list): Likewise.
1592 * haifa-sched.c (schedule_block, extend_h_i_d): Likewise.
1593 * ifcvt.c (noce_emit_move_insn): Likewise.
1594 * modulo-sched.c (generate_prolog_epilog, sms_schedule_by_order):
1596 * stor-layout.c (get_best_mode): Likewise.
1597 * tree-ssa-loop-niter.c (get_val_for): Likewise.
1598 * tree-ssa-structalias.c (get_varinfo, get_varinfo_fc,
1599 scc_visit, do_ds_constraint, do_complex_constraint, label_visit,
1600 perform_var_substitution, solve_graph): Likewise.
1601 * tree-vrp.c (vrp_finalize): Likewise.
1603 2007-02-01 Ian Lance Taylor <iant@google.com>
1605 * lower-subreg.c (simplify_gen_subreg_concatn): If we ask for the
1606 high part of a paradoxical subreg, return a constant zero.
1608 2007-02-01 Zdenek Dvorak <dvorakz@suse.cz>
1610 * toplev.c (lang_dependent_init): Call init_set_costs.
1611 * loop-init.c (loop_optimizer_init): Do not call init_set_costs.
1613 2007-02-01 Richard Guenther <rguenther@suse.de>
1616 * fold-const.c (fold_negate_expr): Allow negating a
1617 constant if overflow does not change.
1619 2007-02-01 Ralf Wildenhues <Ralf.Wildenhues@gmx.de>
1621 * doc/c-tree.texi (Expression trees): Improve markup.
1622 * doc/tm.texi (Register Classes, Addressing Modes)
1623 (Floating Point): Fix spacing after abbreviations. Fix some
1626 2007-02-01 Ben Elliston <bje@au.ibm.com>
1628 * doc/invoke.texi: Replace "bugfix" with "bug fix" throughout.
1629 * doc/contrib.texi: Likewise.
1630 * doc/install.texi: Likewise.
1632 2007-01-31 Richard Henderson <rth@redhat.com>
1633 Ian Lance Taylor <iant@google.com>
1635 * lower-subreg.c: New file.
1636 * rtl.def (CONCATN): Define.
1637 * passes.c (init_optimization_passes): Add pass_lower_subreg and
1639 * emit-rtl.c (update_reg_offset): New static function, broken out
1640 of gen_rtx_REG_offset.
1641 (gen_rtx_REG_offset): Call update_reg_offset.
1642 (gen_reg_rtx_offset): New function.
1643 * regclass.c: Revert patch of 2006-03-05, restoring
1645 (clear_reg_info_regno): New function.
1646 * dwarf2out.c (concatn_loc_descriptor): New static function.
1647 (loc_descriptor): Handle CONCATN.
1648 * common.opt (fsplit_wide_types): New option.
1649 * opts.c (decode_options): Set flag_split_wide_types when
1651 * timevar.def (TV_LOWER_SUBREG): Define.
1652 * rtl.h (gen_reg_rtx_offset): Declare.
1653 (reg_scan_update): Declare.
1654 * regs.h (clear_reg_info_regno): Declare.
1655 * tree-pass.h (pass_lower_subreg): Declare.
1656 (pass_lower_subreg2): Declare.
1657 * doc/invoke.texi (Option Summary): List -fno-split-wide-types.
1658 (Optimize Options): Add -fsplit-wide-types to -O1 list. Document
1660 * doc/rtl.texi (Regs and Memory): Document concat and concatn.
1661 * Makefile.in (OBJS-common): Add lower-subreg.o.
1662 (lower-subreg.o): New target.
1664 2007-01-31 Kazu Hirata <kazu@codesourcery.com>
1666 * config/sh/sh.h (HAVE_SECONDARY_RELOADS): Remove.
1668 2007-01-31 Anatoly Sokolov <aesok@post.ru>
1671 * config/avr/avr.c (DWARF2_ADDR_SIZE): Define.
1673 2007-01-31 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1676 * builtins.c (fold_builtin_sqrt): Use MPFR for constant args.
1678 2007-01-31 Zdenek Dvorak <dvorakz@suse.cz>
1680 * cfgloop.h: Include vec-prim.h.
1681 (enum li_flags): Remove LI_ONLY_OLD.
1682 (loop_iterator): Changed.
1683 (fel_next, fel_init): Iterate over loop tree.
1684 (FOR_EACH_LOOP_BREAK): New macro.
1685 * loop-unswitch.c (unswitch_loops): Do not pass LI_ONLY_OLD to
1687 * tree-ssa-loop-unswitch.c (tree_ssa_unswitch_loops): Ditto.
1688 * modulo-sched.c (sms_schedule): Ditto.
1689 * tree-vectorizer.c (vectorize_loops): Ditto.
1690 * doc/loop.texi: Update information on loop numbering and behavior of
1691 FOR_EACH_LOOP wrto new loops.
1692 * tree-scalar-evolution.c (compute_overall_effect_of_inner_loop,
1693 add_to_evolution_1): Test nestedness of loops instead of comparing
1695 * tree-chrec.c (chrec_fold_plus_poly_poly,
1696 chrec_fold_multiply_poly_poly, chrec_evaluate,
1697 hide_evolution_in_other_loops_than_loop, chrec_component_in_loop_num,
1698 reset_evolution_in_loop): Ditto.
1699 * Makefile.in (CFGLOOP_H): Add vecprim.h dependency.
1701 2007-01-31 Dirk Mueller <dmueller@suse.de>
1703 * c-common.c (warn_about_parentheses): Separate warning about
1704 un-parenthized sequence of comparison operators from the one
1705 which is supposed to warn about x <= y <= z.
1707 2007-01-31 Uros Bizjak <ubizjak@gmail.com>
1709 * optabs.h (enum optab_index): Add new OTI_isinf.
1710 (isinf_optab): Define corresponding macro.
1711 * optabs.c (init_optabs): Initialize isinf_optab.
1712 * genopinit.c (optabs): Implement isinf_optab using isinf?f2
1714 * builtins.c (mathfn_built_in): Handle BUILT_IN_ISINF{,F,L}.
1715 (expand_builtin_interclass_mathfn): Expand BUILT_IN_ISINF{,F,L}
1717 (expand_builtin): Expand BUILT_IN_ISINF{,F,L} using
1718 expand_builtin_interclass_mathfn.
1719 * reg_stack.c (subst_stack_regs_pat): Handle UNSPEC_FXAM.
1720 * config/i386/i386.md (UNSPEC_FXAM): New constant.
1721 (fxam<mode>2_i387): New insn pattern.
1722 (isinf<mode>2) New expander to implement isinf, isinff and isinfl
1723 built-in functions as x87 inline asm.
1725 2007-01-31 Kazu Hirata <kazu@codesourcery.com>
1727 * gcc/config/arm/unwind-arm.h (_sleb128_t, _uleb128_t): New.
1729 2007-01-30 Eric Christopher <echristo@apple.com>
1731 * config.gcc: Add geode.
1733 2007-01-31 Kazu Hirata <kazu@codesourcery.com>
1735 * cgraphunit.c, config/arm/arm.c, config/m68k/m68k.c,
1736 ipa-inline.c, tree-profile.c, tree-ssa-live.c,
1737 tree-ssa-math-opts.c, tree-ssanames.c, tree-vect-analyze.c,
1738 value-prof.c: Fix comment typos.
1740 2007-01-30 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
1743 * doc/invoke.texi (Wpointer-arith): Document warning.
1745 2007-01-30 Janis Johnson <janis187@us.ibm.com>
1747 * doc/extend.texi (Decimal Floating Types): Remove decfloat.h from
1748 the list of discrepancies from the draft TR.
1750 2007-01-30 Dirk Mueller <dmueller@suse.de>
1753 * doc/invoke.texi (-Wreturn-type): Update description to
1756 2007-01-30 Richard Sandiford <richard@codesourcery.com>
1758 * cfgrtl.c (try_redirect_by_replacing_jump): Check only_sets_cc0_p.
1760 2007-01-30 Uros Bizjak <ubizjak@gmail.com>
1762 * builtins.c (expand_builtin_int_interclass_roundingfn): New function
1763 to handle optabs that operate on floating point input argument and
1764 output to integer output.
1765 (expand_builtin_mathfn) [BUILT_IN_ILOGB]: Move from here ...
1766 (expand_builtin_interclass_mathfn) [BUILT_IN_ILOGB]: ... to here.
1767 (expand_builtin): Expand BUILT_IN_ILOGB{,F,L} using
1768 expand_builtin_interclass_mathfn ().
1769 * config/i386/i386.md (fxtractxf3_i387): Rename from *fxtractxf3_i387.
1771 (ilogbxf2, ilogb<mode>2): New expanders to implement ilogb, ilogbf and
1772 ilogbl built-in functions as x87 intrinsics.
1774 2007-01-30 Richard Guenther <rguenther@suse.de>
1777 * dwarf2out.c (reference_to_unused): Query varpool if the
1778 variable was output.
1780 2007-01-30 Richard Guenther <rguenther@suse.de>
1783 * passes.c (execute_one_pass): Reset in_gimple_form to not
1784 confuse non-unit-at-a-time mode.
1786 2007-01-29 Roger Sayle <roger@eyesopen.com>
1787 Richard Guenther <rguenther@suse.de>
1789 * fold-const.c (round_up): Make HIGH an unsigned HOST_WIDE_INT to
1790 avoid undefined behaviour on overflow. Use force_fit_type_double
1791 to construct the constant with the specified TREE_OVERFLOW.
1793 2007-01-29 Janis Johnson <janis187@us.ibm.com>
1795 * config/dfp-bit.c: Add parameterized support for fp exceptions.
1796 * config/dfp-bit.h: Ditto.
1798 2007-01-29 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
1800 * c-decl.c (pop_scope): Replace warnings with call to
1801 warn_for_unused_label.
1802 * c-common.h (warn_for_unused_label): Declare.
1803 * c-common.c (warn_for_unused_label): Define.
1805 2007-01-29 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
1807 * tree-optimize.c (update_inlined_to_pointers): Delete unused
1810 2007-01-29 Janis Johnson <janis187@us.ibm.com>
1812 * Makefile.in (USER_H): Remove decfloat.h.
1813 * ginclude/decfloat.h: Delete, moving contents to ...
1814 * ginclude/float.h: Add support for decimal floating point,
1815 guarded by __STDC_WANT_DEC_FP__.
1817 2007-01-29 Mike Stump <mrs@apple.com>
1819 * doc/gccint.texi (Top): Rename Loop Representation to Loop
1820 Analysis and Representation to resolve case insensitive conflict.
1821 * doc/loop.texi (Loop Analysis and Representation): Likewise.
1823 2007-01-28 Daniel Berlin <dberlin@dberlin.org>
1825 * tree.h (struct tree_memory_tag): Add aliases member.
1826 (MTAG_ALIASES): New macro.
1827 * tree-ssa-alias.c (alias_bitmap_obstack): New variable.
1828 (add_may_alias): Remove pointer-set. Update for may_aliases being
1830 (mark_aliases_call_clobbered): Update for may_aliases being a
1832 (compute_tag_properties): Ditto.
1833 (create_partition_for): Ditto.
1834 (compute_memory_partitions): Ditto.
1835 (dump_may_aliases_for): Ditto.
1836 (is_aliased_with): Ditto.
1837 (add_may_alias_for_new_tag): Ditto.
1838 (rewrite_alias_set_for): Rewrite for may_aliases being a bitmap.
1839 (compute_is_aliased): New function.
1840 (compute_may_aliases): Call compute_is_aliased).
1841 (init_alias_info): Initialize alias_bitmap_obstack.
1842 (union_alias_set_into): New function.
1843 (compute_flow_sensitive_aliasing): Use union_aliases_into.
1844 (have_common_aliases_p): Rewrite to take two bitmaps and use
1846 (compute_flow_insensitive_aliasing): Stop using pointer-sets.
1848 (finalize_ref_all_pointers): Update for add_may_alias changes.
1849 (new_type_alias): Ditto.
1850 * tree-flow-inline.h (may_aliases): Return a bitmap.
1851 * tree-dfa.c (dump_variable): Check for MTAG_P'ness.
1852 * tree-ssa.c (verify_flow_insensitive_alias_info): Update for
1853 may_aliases being a bitmap.
1854 * tree-flow.h (struct var_ann_d): Remove may_aliases member.
1855 may_aliases now returns a bitmap.
1856 * tree-ssa-structalias.c (merge_smts_into): Update for may_aliases
1858 * tree-ssa-operands.c (add_virtual_operand): Update for
1859 may_aliases being a bitmap.
1861 2007-01-29 Daniel Berlin <dberlin@dberlin.org>
1863 PR tree-optimization/30630
1864 * tree-ssa-structalias.c (do_complex_constraint): Mark correct
1865 variable as changed.
1867 2007-01-29 Simon Martin <simartin@users.sourceforge.net>
1870 * gimplify.c (gimplify_target_expr): Make sure that the TARGET_EXPR is
1871 expanded only once even if an error occurs.
1873 2007-01-29 Ben Elliston <bje@au.ibm.com>
1875 * gcov-io.h (__gcov_indirect_call_profiler): Declare.
1876 (__gcov_average_profiler): Likewise.
1877 (__gcov_ior_profiler): Likewise.
1878 (__gcov_merge_ior): Likewise.
1880 2007-01-28 Jan Hubicka <jh@suse.cz>
1882 * builtins.c (expand_builtin_memset): Fix typo in my last patch.
1883 * value-prof.c (stringop_block_profile): Likewise.
1885 2007-01-28 Jan Hubicka <jh@suse.cz>
1887 * expr.c (emit_block_move_via_movmem, emit_block_move_via_libcall): Add
1888 variant handling histograms; add wrapper.
1889 (clear_storage_via_libcall): Export.
1890 (emit_block_move_hints): Break out from ...; add histograms.
1891 (emit_block_move): ... this one.
1892 (clear_storage_hints): Break out from ...; add histograms.
1893 (clear_storage): ... this one.
1894 (set_storage_via_memset): Handle histogram.
1895 * expr.h (emit_block_move_via_libcall, emit_block_move_hints): Declare.
1896 (clear_storage_hints, clear_storage_via_libcall): Declare.
1897 (set_storage_via_setmem): Update prototype.
1898 * doc/md.texi (movmem, setmem): Document new arguments.
1900 * value-prof.c (dump_histogram_value, tree_find_values_to_profile): Add
1902 (stringop_block_profile): New global function.
1903 (tree_stringops_values_to_profile): Profile block size and alignment.
1904 * value-prof.h (enum hist_type): add HIST_TYPE_AVERAGE and
1906 (struct profile_hooks): Add gen_average_profiler and gen_ior_profiler.
1907 (stringop_block_profile): Declare.
1908 * builtins.c: Include value-prof.h.
1909 (expand_builtin_memcpy, expand_builtin_memset): Pass block profile.
1910 * gcov-ui.h (GCOV_COUNTER_NAMES): Add new counter.
1911 (GCOV_COUNTER_AVERAGE, GCOV_COUNTER_IOR): New constants.
1912 (GCOV_COUNTERS, GCOV_LAST_VALUE_COUNTER): Update.
1913 * profile.c (instrument_values): Add new counters.
1914 * cfgexpand.c (expand_gimple_basic_block): Propagate histograms to
1916 * tree-profile.c (tree_average_profiler_fn, tree_ior_profiler_fn): New.
1917 (tree_init_edge_profiler): Build new profilers.
1918 (tree_gen_average_profiler, tree_gen_ior_profiler): New.
1919 (pass_tree_profile): Add dump.
1920 (tree_profile_hooks): Update.
1921 * Makefile.in (LIBGCOV): Add new constants.
1922 * libgcov.c (__gcov_merge_ior, __gcov_average_profiler,
1923 __gcov_ior_profiler): New.
1924 * i386.md (movmem/setmem expanders): Add new optional arguments.
1926 2007-01-28 David Edelsohn <edelsohn@gnu.org>
1928 * doc/md.texi (Standard Pattern Names): Document blockage pattern.
1930 2007-01-28 Zdenek Dvorak <dvorakz@suse.cz>
1932 * tree-ssa-loop-unswitch.c: Include tree-inline.h.
1933 (tree_unswitch_single_loop): Pass eni_size_weights to
1934 tree_num_loop_insns.
1935 * tree-ssa-loop-manip.c: Include tree-inline.h.
1936 (can_unroll_loop_p): Pass eni_size_weights to
1937 tree_num_loop_insns.
1938 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p):
1939 Pass eni_size_weights to estimate_num_insns.
1940 * tree.h (init_inline_once): Export.
1941 * toplev.c (backend_init): Call init_inline_once.
1942 * cgraphunit.c (cgraph_process_new_functions,
1943 cgraph_analyze_function): Pass eni_inlining_weights to
1945 * ipa-inline.c (compute_inline_parameters): Ditto.
1946 * tree-ssa-loop-ivcanon.c (tree_num_loop_insns): Pass weights
1947 to estimate_num_insns.
1948 (try_unroll_loop_completely): Pass eni_size_weights to
1949 tree_num_loop_insns.
1950 * tree-eh.c (decide_copy_try_finally): Pass eni_size_weights
1951 ot estimate_num_insns.
1952 * tree-ssa-loop-prefetch.c: Include tree-inline.h.
1953 (loop_prefetch_arrays): Pass eni_time_weights to tree_num_loop_insns.
1954 * tree-inline.c (eni_inlining_weights, eni_size_weights,
1955 eni_time_weights): New variables.
1956 (init_inline_once): Initialize them.
1957 (struct eni_data): Mew.
1958 (estimate_num_insns_1, estimate_num_insns): Use weights.
1959 * tree-inline.h (struct eni_weights_d): New.
1960 (eni_inlining_weights, eni_size_weights, eni_time_weights): Declare.
1961 (estimate_num_insns): Declaration changed.
1962 * cfgloop.h (tree_num_loop_insns): Declaration changed.
1963 * Makefile.in (tree-ssa-loop-unswitch.o, tree-ssa-loop-prefetch.o,
1964 tree-ssa-loop-manip.o): Add TREE_INLINE_H dependency.
1966 2007-01-28 Zdenek Dvorak <dvorakz@suse.cz>
1968 * tree-data-ref.c (conflict_fn): Assert that the number of affine
1969 relations in the conflict function is valid.
1971 2007-01-27 Ian Lance Taylor <iant@google.com>
1973 * common.opt: Add fstrict-overflow.
1974 * opts.c (decode_options): Set flag_strict_overflow if -O2.
1975 * flags.h (TYPE_OVERFLOW_WRAPS): Define.
1976 (TYPE_OVERFLOW_UNDEFINED): Define.
1977 (TYPE_OVERFLOW_TRAPS): Define. This replaces TYPE_TRAP_SIGNED.
1979 * tree.h (TYPE_TRAP_SIGNED): Don't define.
1980 * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_UNDEFINED.
1981 (fold_negate_expr): Likewise.
1982 (make_range): Likewise.
1983 (extract_muldiv_1): Likewise.
1984 (maybe_canonicalize_comparison): Likewise.
1985 (fold_comparison): Likewise.
1986 (fold_binary): Likewise.
1987 (tree_expr_nonnegative_p): Likewise.
1988 (tree_expr_nonzero_p): Likewise.
1989 * tree-vrp.c (compare_values): Likewise.
1990 (extract_range_from_binary_expr): Likewise.
1991 (extract_range_from_unary_expr): Likewise.
1992 * tree-ssa-loop-niter.c (infer_loop_bounds_from_signedness):
1994 (nowrap_type_p): Likewise.
1995 * tree-scalar-evolution.c (simple_iv): Likewise.
1996 * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_WRAPS.
1997 (build_range_check): Likewise.
1998 (extract_muldiv_1): Likewise.
1999 (fold_comparison): Likewise.
2000 * tree-vrp.c (vrp_int_const_binop): Likewise.
2001 (extract_range_from_unary_expr): Likewise.
2002 * convert.c (convert_to_integer): Likewise.
2003 * fold-const.c (fold_negate_expr): Use TYPE_OVERFLOW_TRAPS.
2004 (fold_comparison): Likewise.
2005 (fold_binary): Likewise.
2006 * optabs.c (optab_for_tree_code): Likewise.
2007 * tree-vectorizer.c (vect_is_simple_reduction): Likewise.
2008 * simplify-rtx.c (simplify_const_relational_operation): Check
2009 flag_strict_overflow and flag_trapv.
2010 (simplify_const_relational_operation): Likewise.
2011 * doc/invoke.texi (Option Summary): Mention -fstrict-overflow.
2012 (Optimize Options): Add -fstrict-overflow to -O2 list. Document
2015 2007-01-27 Roger Sayle <roger@eyesopen.com>
2017 * tree.c (tree_fold_gcd): Delete.
2018 * tree.h (tree_fold_gcd): Remove prototype.
2019 * tree-data-ref.c (tree_fold_divides_p): Don't use tree_fold_gcd to
2020 test whether one constant integer is a multiple of another. Instead
2021 call int_const_binop with TRUNC_MOD_EXPR and test for a zero result.
2022 * fold-const.c (multiple_of_p): We've determined both TOP and
2023 BOTTOM are integer constants so we can call int_const_binop directly
2024 instead of the more generic const_binop.
2026 2007-01-27 Roger Sayle <roger@eyesopen.com>
2028 * fold-const.c (size_binop): In the fast-paths for X+0, 0+X, X-0 and
2029 1*X check that the constant hasn't overflowed, to preserve the
2031 (round_up): Provide an efficient implementation when rouding-up an
2032 INTEGER_CST to a power-of-two.
2034 2007-01-28 Ralf Wildenhues <Ralf.Wildenhues@gmx.de>
2036 * doc/sourcebuild.texi: Add comma for clarity.
2037 * doc/extend.texi: Fix some typos.
2038 * doc/passes.texi: Likewise.
2039 * doc/cppinternals.texi: Likewise.
2040 * doc/c-tree.texi: Likewise.
2041 * doc/tree-ssa.texi: Likewise.
2042 * doc/install.texi: Likewise.
2044 2007-01-27 Jan Hubicka <jh@suse.cz>
2046 * tree-sra.c (sra_walk_function): Don't rely on aliases being build.
2047 (pass_sra): Do not require alias information.
2048 * passes.c (init_optimization_passes): Add SRA
2050 2007-01-27 Steven Bosscher <steven@gcc.gnu.org>
2052 * tracer.c (rest_of_handle_tracer): We already cleaned
2053 up the CFG in tracer() so don't do it here again.
2054 * cfgcleanup.c (rest_of_handle_jump2): Don't repeat
2055 cleanup_cfg here, either. And don't call renumber_insns.
2057 * cfgrtl.c (rtl_verify_flow_info_1): Don't verify that BB_END
2058 and BB_HEAD are in the insn stream here. Instead make sure
2059 that BB_INSN is valid on all insns. Also, do check here that
2060 there are no pending branch predictions...
2061 (rtl_verify_flow_info): ...instead of doing it here. Checks
2062 for BB_END and BB_HEAD moved from rtl_verify_flow_info_1 to
2065 2007-01-26 Roger Sayle <roger@eyesopen.com>
2067 * config/i386/i386.c (ix86_swap_binary_operands_p): New helper
2068 function to simplify/factorize operand order canonicalization.
2069 (ix86_fixup_binary_operands): Reorganize using the above function.
2070 (ix86_binary_operator_ok): Likewise.
2072 2007-01-27 Jakub Jelinek <jakub@redhat.com>
2074 * genattrtab.c (struct attr_value_list, insn_code_values): Move to
2075 file scope from optimize_attrs.
2076 (simplify_test_exp): If insn_code_values is not NULL, use it to speed
2078 (optimize_attrs): Clear insn_code_values after freeing it.
2080 2007-01-26 Zdenek Dvorak <dvorakz@suse.cz>
2082 * tree-ssa-address.c (create_mem_ref): Remove ", bsi" from
2083 a parts.base assignment.
2085 2007-01-26 Zdenek Dvorak <dvorakz@suse.cz>
2087 * tree-data-ref.c (dump_subscript): Use dump_conflict_function.
2088 (compute_subscript_distance, initialize_data_dependence_relation,
2089 finalize_ddr_dependent, analyze_ziv_subscript,
2090 analyze_siv_subscript_cst_affine,
2091 compute_overlap_steps_for_affine_univar,
2092 compute_overlap_steps_for_affine_1_2, analyze_subscript_affine_affine,
2093 analyze_siv_subscript, analyze_miv_subscript,
2094 analyze_overlapping_iterations, subscript_dependence_tester_1,
2095 compute_self_dependence, free_dependence_relation): Work
2096 with affine_fn instead of chrecs.
2097 (dump_affine_function, dump_conflict_function, affine_function_equal_p,
2098 common_affine_function, affine_function_base,
2099 affine_function_constant_p, affine_fn_op, affine_fn_plus,
2100 affine_fn_minus, affine_fn_free, conflict_fn_not_known,
2101 conflict_fn_no_dependence, free_conflict_function, free_subscripts,
2102 conflict_fn, affine_fn_cst, affine_fn_univar): New functions.
2103 (all_chrecs_equal_p): Removed.
2104 * tree-data-ref.h (affine_fn, conflict_function): New types.
2105 (struct subscript): Change type of conflicting_iterations_in_a
2106 and conflicting_iterations_in_b.
2108 2007-01-26 Steve Ellcey <sje@cup.hp.com>
2111 * config/pa/pa.h (TARGET_HPUX_11): New.
2112 * config/pa/pa-hpux11.h (TARGET_HPUX_11): New.
2113 * config/pa/pa.c (pa_init_builtins): Use TARGET_HPUX_11.
2115 2007-01-26 Daniel Berlin <dberlin@dberlin.org>
2116 Richard Guenther <rguenther@suse.de>
2118 * tree-ssa-structalias.c (solve_graph): Handle case
2119 we merged the variable to another.
2121 2007-01-25 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
2123 * builtins.c (fold_builtin_1): Treat ccos and ccosh as 'even'
2126 * fold-const.c (negate_mathfn_p): Treat casin, casinh, catan,
2127 catanh, cproj, csin, csinh, ctan and ctanh as 'odd' functions.
2129 2007-01-25 DJ Delorie <dj@redhat.com>
2131 * config/m32c/m32c.c (m32c_cannot_change_mode_class): We don't
2132 allow changes to modes which don't fit in those registers.
2134 * reload1.c (choose_reload_regs): Check for invalid subregs before
2135 computing their locations, not after.
2137 2007-01-25 Geoffrey Keating <geoffk@apple.com>
2140 * config/rs6000/rs6000.c (first_altivec_reg_to_save): On Darwin,
2141 save Altivec registers in an eh_return function.
2142 (compute_vrsave_mask): Likewise.
2143 (rs6000_stack_info): Correct AIX/Darwin stack alignment computation
2144 for saving Altivec registers.
2145 (rs6000_emit_prologue): Don't allocate stack twice in
2146 eh_return function. Correct expected value of altivec_save_offset
2147 when using save_world. Describe save of R0 to stack when using
2148 save_world. Describe stack pointer adjustment when using
2149 save_world. Remove duplicated eh_return parameter register saving.
2150 Update sp_offset variable after save_world.
2151 * config/rs6000/t-darwin (LIB2FUNCS_STATIC_EXTRA): Remove
2153 (LIB2FUNCS_EXTRA): Add darwin-world.asm.
2154 * config/rs6000/darwin.h (SUBTARGET_OVERRIDE_OPTIONS): -m64
2157 2007-01-25 Steve Ellcey <sje@cup.hp.com>
2159 * config.gcc (ia64*-*-hpux*): Make posix threads the default.
2161 2007-01-25 Steve Ellcey <sje@cup.hp.com>
2164 * config/pa/pa.c (pa_init_builtins): Set asm names for finite routines.
2165 * config/ia64/ia64.c (ia64_init_builtins): Ditto.
2167 2007-01-25 Richard Guenther <rguenther@suse.de>
2169 * doc/invoke.texi (-Wcoverage-mismatch): Document.
2170 * common.opt (-Wcoverage-mismatch): New warning option.
2171 * coverage.c (get_coverage_counts): Ignore coverage mismatch
2172 if -Wcoverage-mismatch is given.
2174 2007-01-25 Razya Ladelsky <razya@il.ibm.com>
2176 * ipa-cp.c (ipcp_insert_stage, ipcp_driver): Support for SSA.
2177 (ipcp_driver): Change to static definition.
2178 Add dumping of the ifunctions.
2179 (constant_val_insert): Remove unused parameter. Support for SSA.
2180 (ipcp_propagate_const): Support for SSA.
2181 (ipcp_profile_bb_print): Print only analyzed nodes.
2182 (ipcp_replace_map_create): Remove support for Fortran constant
2184 * ipa-prop.c (ipa_method_modify_stmt,
2185 ipa_callsite_compute_param): Support for SSA.
2186 * ipa-prop.h (ipcp_driver): Remove declaration.
2187 (IS_VALID_TREE_MAP_INDEX): Add define.
2189 2007-01-24 Geoffrey Keating <geoffk@apple.com>
2191 * unwind-dw2.c (execute_stack_op): Handle DW_OP_swap.
2193 2007-01-24 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
2196 * builtins.c (fold_builtin_cabs): Use MPFR to evaluate a
2197 constant argument to cabs and do it without checking for
2198 -funsafe-math-optimizations.
2200 2007-01-24 Douglas Gregor <dgregor@osl.iu.edu>
2202 * c-common.h (RID_FIRST_CXX0X): New.
2203 (RID_LAST_CXX0X): New.
2204 * c-opts.c (c_common_handle_option): -Wc++0x-compat is triggered
2206 * c.opt (Wc++0x-compat): New.
2207 * doc/invoke.texi (-Wc++0x-compat): Document.
2209 2007-01-24 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
2211 * builtins.c (fold_builtin_carg): New.
2212 (fold_builtin_1): Use it.
2214 2007-01-24 Jan Hubicka <jh@suse.cz>
2216 * ipa-inline.c (cgraph_decide_inlining): Initialize initial_insns.
2218 * ipa-inline.c (initial_insns, max_insns): Delete.
2219 (compute_max_insns): New function.
2220 (cgraph_decide_inlining_of_small_function): Use it; take minimal amount
2221 of insns as base for code growth.
2222 (cgraph_decide_inlining): Make initial_insns local; do not compute
2224 * params.def (PARAM_INLINE_UNIT_GROWTH): Set to 60.
2225 * doc/invoke.texi (inline-unit-growth): Update docs.
2227 2007-01-24 Jakub Jelinek <jakub@redhat.com>
2229 * config/i386/i386.h (x86_cmpxchg16b): Remove const.
2230 (TARGET_CMPXCHG16B): Define to x86_cmpxchg16b.
2231 * config/i386/i386.c (x86_cmpxchg16b): Remove const.
2232 (override_options): Add PTA_CX16 flag. Set x86_cmpxchg16b
2233 for CPUs that have PTA_CX16 set.
2236 * gimplify.c (omp_check_private): New function.
2237 (gimplify_scan_omp_clauses): Use it for
2238 firstprivate/lastprivate/reduction.
2241 * gimplify.c (omp_add_variable): Don't call omp_notice_variable
2242 on TYPE_SIZE_UNIT for GOVD_LOCAL VLAs.
2245 * omp-low.c (lower_omp_for_lastprivate): Add dlist argument.
2246 If lower_lastprivate_clauses emits some statements, append them
2247 to dlist rather than body_p and to body_p append an initializer.
2248 (lower_omp_for): Adjust caller.
2250 2007-01-24 Steve Ellcey <sje@cup.hp.com>
2252 * target.h (globalize_decl_name): New.
2253 * target-def.h (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
2254 * output.h (default_globalize_decl_name): New.
2255 * varasm.c (asm_output_bss): Use globalize_decl_name instead of
2257 (globalize_decl): Ditto.
2258 (default_globalize_decl_name): New.
2259 * config/ia64/ia64.c (ia64_globalize_decl_name): New.
2260 (ia64_handle_version_id_attribute): New.
2261 (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
2262 (ia64_asm_output_external): Use globalize_decl_name instead
2264 * doc/extend.texi (version_id): New pragma.
2265 * doc/tm.texi (ARGET_ASM_GLOBALIZE_DECL_NAME): New target hook.
2267 2007-01-24 Andreas Krebbel <krebbel1@de.ibm.com>
2269 * unwind-dw2-fde.c (get_cie_encoding): Replaced _Unwind_Word with
2270 _uleb128_t and _Unwind_SWord with _sleb128_t.
2271 * unwind-dw2.c (extract_cie_info, execute_stack_op, execute_cfa_program,
2272 uw_frame_state_for, uw_update_context_1): Likewise.
2273 * unwind-c.c (parse_lsda_header, PERSONALITY_FUNCTION): Likewise.
2274 * unwind-pe.h (read_uleb128, read_sleb128,
2275 read_encoded_value_with_base): Likewise.
2276 * unwind-generic.h: Define _sleb128_t and _uleb128_t types.
2278 2007-01-24 Richard Guenther <rguenther@suse.de>
2280 * builtins.c (expand_builtin_cexpi): Get the fndecl
2281 for cexp in the correct way.
2283 2007-01-24 Jan Hubicka <jh@suse.cz>
2285 * tree-ssa-dce.c (eliminate_unnecesary_stmts): Remove dead LHS of calls.
2287 2007-01-24 Andreas Krebbel <krebbel1@de.ibm.com>
2289 * c-cppbuiltin.c (builtin_define_type_sizeof): New function.
2290 (c_cpp_builtins): New builtin macros: __SIZEOF_INT__, __SIZEOF_LONG__,
2291 __SIZEOF_LONG_LONG__, __SIZEOF_SHORT__, __SIZEOF_POINTER__,
2292 __SIZEOF_FLOAT__, __SIZEOF_DOUBLE__, __SIZEOF_LONG_DOUBLE__,
2293 __SIZEOF_SIZE_T__, __SIZEOF_WCHAR_T__, __SIZEOF_WINT_T__ and
2294 __SIZEOF_PTRDIFF_T__.
2295 * doc/cpp.texi: Documentation for the new builtin macros added.
2297 2007-01-24 Uros Bizjak <ubizjak@gmail.com>
2299 * config/i386/i386.md (tanxf2, tan<mode>2, atan<mode>2, log<mode>2,
2300 log10<mode>2, log2<mode>2, expxf2, exp10xf2, exp2xf2): Use op2
2301 instead of operands[2] to avoid access past the end of array.
2303 2007-01-24 Richard Sandiford <richard@codesourcery.com>
2305 * reload1.c (emit_reload_insns): Pass the reload register
2306 for a non-spill output reload through forget_old_reloads_1.
2308 2007-01-23 Joseph Myers <joseph@codesourcery.com>
2310 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Define
2311 _SOFT_DOUBLE if doubles use software floating-point.
2312 * config/rs6000/libgcc-ppc-glibc.ver: Export additional long
2313 double functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
2314 * config/rs6000/darwin-ldouble.c: Also compile functions for
2315 hard-float without FPRs. Use fmsub function for all __NO_FPRS__
2316 cases. Compile extra functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
2317 * config/rs6000/linuxspe.h (SUBSUBTARGET_OVERRIDE_OPTIONS): Remove
2318 commented-out long double override.
2319 (CPP_LONGDOUBLE_DEFAULT_SPEC): Likewise.
2320 * config/rs6000/eabispe.h: Likewise.
2321 * config/rs6000/rs6000.c (rs6000_override_options): Don't override
2322 long double for non-SPE.
2323 (rs6000_handle_option): Likewise.
2324 (invalid_e500_subreg): Disallow more subregs involding DImode,
2325 DFmode, TImode or TFmode.
2326 (rs6000_legitimate_offset_address_p): Check TFmode offsets for
2328 (legitimate_lo_sum_address_p): Also check for TFmode for E500
2330 (rs6000_legitimize_address): Also handle TFmode for E500 double.
2331 (rs6000_legitimize_reload_address): Also handle TFmode for E500
2333 (rs6000_legitimate_address): Also check for TFmode for E500
2335 (rs6000_emit_move): Use DFmode subregs of TFmode for E500 double.
2336 (spe_build_register_parallel): Handle TFmode and TCmode.
2337 (rs6000_spe_function_arg): Handle TFmode and TCmode for E500
2339 (function_arg): Handle TFmode and TCmode for E500 double.
2340 (rs6000_init_libfuncs): Initialize extra libfuncs for soft double
2342 (print_operand): Handle TFmode and TImode for %y.
2343 (rs6000_generate_compare): Handle TFmode comparisons for E500
2345 (spe_func_has_64bit_regs_p): Check for TFmode for E500 double.
2346 (rs6000_function_value): Handle TFmode and TCmode for E500 double.
2347 (rs6000_libcall_value): Handle TFmode and TCmode for E500 double.
2348 * config/rs6000/rs6000.h (CANNOT_CHANGE_MODE_CLASS): Check for
2349 TFmode for E500 double.
2350 * config/rs6000/rs6000.md (FP): Allow TF for E500 double.
2351 (floatsidf2): Enable for E500 double.
2352 (movtf_softfloat): Use rs6000_nonimmediate_operand.
2353 (extenddftf2): Change to extenddftf2_fprs.
2354 (extenddftf2): Call gen_spe_extenddftf2 or gen_extenddftf2_fprs
2355 depending on TARGET_E500_DOUBLE.
2356 (extendsftf2): Enable for E500 double.
2357 (trunctfdf2): Enable for E500 double.
2358 (trunctfsf2): Change to trunctfsf2_fprs.
2359 (trunctfsf2): Call gen_spe_trunctfsf2 or gen_trunctfsf2_fprs
2360 depending on TARGET_E500_DOUBLE.
2361 (floatsitf2): Enable for E500 double.
2362 (fix_trunctfsi2): Change to fix_trunctfsi2_fprs.
2363 (fix_trunctfsi2): Call gen_spe_fix_trunctfsi2 or
2364 gen_fix_trunctfsi2_fprs depending on TARGET_E500_DOUBLE.
2365 (negtf2): Change to negtf2_internal.
2366 (negtf2): New expander.
2367 (abstf2): Enable for E500 double. Call gen_spe_abstf2_tst,
2368 gen_spe_abstf2_cmp or gen_abstf2_internal depending on
2369 TARGET_E500_DOUBLE and flag_unsafe_math_optimizations.
2370 (movdi_internal32): Use rs6000_nonimmediate_operand.
2371 (unnamed splitter): Likewise.
2372 * config/rs6000/spe.md (CMPTFEQ_GPR, TSTTFEQ_GPR, CMPTFGT_GPR,
2373 TSTTFGT_GPR, CMPTFLT_GPR, TSTTFLT_GPR): New unspecs.
2374 (SPE64TF, DITI): New mode macros.
2375 (frob_df_di): Change to frob_<SPE64:mode>_<DITI:mode>; allow more
2378 (frob_<mode>_di_2): New.
2379 (frob_tf_di_8_2): New.
2380 (frob_di_df): Change to frob_di_<mode>; allow more modes.
2382 (frob_di_df_2): Change to frob_<DITI:mode>_<SPE64:mode>_2; allow
2384 (frob_ti_<mode>_8_2): New.
2385 (frob_ti_tf_2): New.
2386 (mov_si<mode>_e500_subreg0, mov_si<mode>_e500_subreg0_2,
2387 mov_si<mode>_e500_subreg4, mov_si<mode>_e500_subreg4_2): Allow
2389 (mov_sitf_e500_subreg8, mov_sitf_e500_subreg8_2,
2390 mov_sitf_e500_subreg12, mov_sitf_e500_subreg12_2): New.
2391 (spe_trunctfdf2_internal1, spe_trunctfsf2, spe_extenddftf2,
2392 spe_fix_trunctfsi2, spe_fix_trunctfsi2_internal,
2393 spe_negtf2_internal, spe_abstf2_cmp, spe_abstf2_tst): New.
2394 (cmptfeq_gpr, tsttfeq_gpr, cmptfgt_gpr, tsttfgt_gpr, cmptflt_gpr,
2397 2007-01-23 Ian Lance Taylor <iant@google.com>
2399 * Makefile.in (OBJS-common): Reformat, alphabetize, but put
2401 (OBJS-archive): Reformat, alphabetize.
2402 (OBJS): Change out_object_file to OBJS-md.
2404 2007-01-23 Uros Bizjak <ubizjak@gmail.com>
2406 * config/i386/i386.md: Use REG_P, MEM_P, CONST_INT_P, LABEL_P,
2407 JUMP_P and CALL_P predicates where applicable.
2408 * config/i386/i386.c: Ditto.
2409 * config/i386/i386.md: Ditto.
2410 * config/i386/mmx.md: Ditto.
2411 * config/i386/predicates.md: Ditto.
2413 2007-01-22 Andreas Schwab <schwab@suse.de>
2415 * config/m68k/m68k.h: Fix comment.
2417 2007-01-22 Jan Hubicka <jh@suse.cz>
2419 * passes.c (init_optimization_passes): Do not rerun
2420 pass_early_warn_uninitialized.
2422 2007-01-22 Richard Guenther <rguenther@suse.de>
2424 PR tree-optimization/30038
2425 * tree-ssa-math-opts.c (maybe_record_sincos): New static helper
2427 (execute_cse_sincos_1): Likewise.
2428 (execute_cse_sincos): Likewise.
2429 (gate_cse_sincos): Likewise.
2430 (pass_cse_sincos): New pass CSEing sin() and cos() calls using
2431 the cexpi() canonicalization of sincos().
2432 * tree-pass.h (pass_cse_sincos): Declare.
2433 * passes.c (init_optimization_passes): New pass pas_cse_sincos.
2435 2007-01-21 Eric Botcazou <ebotcazou@libertysurf.fr>
2437 PR rtl-optimization/29329
2438 * combine.c (replaced_rhs_insn): Rename to i2mod.
2439 (replaced_rhs_value): Rename to i2mod_new_rhs.
2440 (i2mod_old_rhs): New global variable.
2441 (combine_instructions): Adjust for above change. Save a copy of
2442 the old RHS into i2mod_old_rhs when the contents of a REG_EQUAL
2443 note are substituted in the second instruction.
2444 (distribute_notes) <REG_DEAD>: Adjust for above change. Do not
2445 ditch the note if it pertains to the second eliminated register
2446 and this register is mentioned in i2mod_old_rhs.
2449 2006-09-12 Eric Botcazou <ebotcazou@libertysurf.fr>
2451 * combine.c (distribute_notes) <REG_DEAD>: Do not consider SETs past
2452 the insn to which the note was originally attached.
2454 2007-01-21 Jan Hubicka <jh@suse.cz>
2456 * ipa-inline.c (inlining_mode): Comment, move up.
2457 (cgraph_decide_inlining_incrementally): Do not perform inlining
2458 itself; fix handling of flattening of self recursive functions.
2459 (cgraph_find_cycles): Remove.
2460 (cgraph_flatten_node): Remove.
2461 (cgraph_decide_inlining): Use incremental inliner to handle flattening.
2462 (try_inline): New function.
2463 (cgraph_early_inlining): Update call of
2464 cgraph_decide_inlining_incrementally. Apply inlining here.
2465 (apply_inline): Update call of cgraph_decide_inlining_incrementally.
2467 2007-01-21 Dirk Mueller <dmueller@suse.de>
2470 * tree-vrp.c (check_array_bounds): do not warn
2471 about ADDR_EXPR's of ARRAY_REF's which are immediately
2472 used in binary expressions.
2474 2007-01-21 Ira Rosen <irar@il.ibm.com>
2476 * tree-vectorizer.h (struct _stmt_vec_info): Add new field
2477 read_write_dep and macros for its access.
2478 * tree-vectorizer.c (new_stmt_vec_info): Initialize the new field.
2479 * tree-vect-analyze.c (vect_analyze_data_ref_dependence): Remove
2480 argument, call vect_check_interleaving for every independent pair of
2481 data-refs. Mark loads that access the same memory location as a store
2483 (vect_check_dependences): Remove.
2484 (vect_analyze_data_ref_dependences): Remove vect_check_dependences
2485 call, fix the call to vect_analyze_data_ref_dependence.
2486 (vect_analyze_data_ref_access): For statements that access the same
2487 data-ref, check that they are not stores; for loads, check that there
2488 is no store that access the same location.
2490 2007-01-20 Roger Sayle <roger@eyesopen.com>
2491 Joseph Myers <joseph@codesourcery.com>
2493 * doc/invoke.texi (-fdump-rtl-combine): Fix under/overfull hbox.
2494 (-fdump-rtl-stack): Likewise.
2495 (-fno-signed-zeros): Use @minus{} for a minus sign. Correct typo.
2496 (-mcheck-zero-division, -mcpu): Fix under/overfull hbox.
2497 (-mpt-fixed): Use @minus{} for minus sign.
2498 (Using Precompiled Headers): Fix under/overfull hbox.
2500 2007-01-20 Jan Hubicka <jh@suse.cz>
2502 * tree-flow.h (struct stmt_ann_d): Move references_memory to proper
2503 place within annotation.
2505 2007-01-20 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
2507 * pa.c (output_move_double): Change array size of xoperands to 4.
2509 2007-01-20 Jan Hubicka <jh@suse.cz>
2511 * tree-tailcall.c (adjust_return_value): Do not use RESULT_DECL
2514 2007-01-19 Ian Lance Taylor <iant@google.com>
2516 * expmed.c (expand_divmod) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]: Cast
2517 constant to unsigned HOST_WIDE_INT before negating.
2519 2007-01-19 Ian Lance Taylor <iant@google.com>
2521 * tree-ssa-operands.h (struct vuse_vec_d): Change num_vuse field
2523 (VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Use unsigned comparison.
2524 (VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
2525 (SET_VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Likewise.
2526 (SET_VUSE_ELEMENT_VAR) [ENABLE_CHECKING]: Likewise.
2527 (SET_VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
2528 (realloc_vdef, realloc_vuse): Change second parameter to
2530 (ssa_operand_iterator_d): Change vuse_index and mayuse_index
2532 * tree-ssa-operands.c (realloc_vop): Change num_elem parameter to
2533 unsigned. Change x and lim locals to unsigned.
2534 (realloc_vdef, realloc_vuse): Change num_elem parameter to
2536 (finalize_ssa_vuse_ops): Change old_i local to unsigned.
2537 (copy_virtual_operands): Change i and n locals to unsigned.
2539 2007-01-19 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
2542 * builtins.c (fold_builtin_1): Handle builtin fdim.
2544 2007-01-20 Jan Hubicka <jh@suse.cz>
2546 * tree-ssa.c (init_tree_ssa): Do not call init_alias_heapvars.
2547 * tree-ssa-structalias.c (compute_points_to_sets): Do call
2548 init_alias_heapvars.
2549 (init_alias_heapvars): Initialize only when not already initialized.
2550 (delete_alias_heapvars): Set heapvar_for_stmt to NULL.
2552 2007-01-19 Roger Sayle <roger@eyesopen.com>
2554 * common.opt (fsigned-zeros): New command line option.
2555 * flags.h (HONOR_SIGNED_ZEROS): Control via flag_signed_zeros instead
2556 of flag_unsafe_math_optimizations.
2557 * opts.c (set_fast_math_flags): The -ffast-math command line option
2558 implies -fno-signed-zeros.
2559 (fast_math_flags_set_p): Likewise.
2561 * doc/invoke.texi: Document new -fno-signed-zeros option, and update
2562 the documentation of -ffast-math appropriately. Wrap long lines.
2564 2007-01-19 Steve Ellcey <sje@cup.hp.com>
2566 * system.h (ASM_MAKE_LABEL_LINKONCE): Poison.
2567 * varasm.c (globalize_decl): Remove ASM_MAKE_LABEL_LINKONCE ifdef.
2569 2007-01-19 Tomas Bily <tbily@suse.cz>
2571 * cgraphunit.c (cgraph_finalize_function): Updating of pid
2573 (tree_init_ic_make_global_vars): New function
2574 (tree_init_edge_profiler): call of tree_init_ic_make_global_vars
2575 (tree_gen_ic_profiler): New function
2576 (tree_gen_ic_func_profiler): New function
2577 (tree_profiling): Added calling of tree_gen_ic_func_profiler
2578 (tree_profile_hooks): Added hook for indirec/virtual calls
2579 * value-prof.c (tree_find_values_to_profile): New case for
2581 (tree_values_to_profile): Call for determining indirect/virtual
2583 (tree_indirect_call_to_profile): New function
2584 (tree_ic_transform): New function
2585 (tree_ic): New function
2586 (find_func_by_pid): New function
2587 (init_pid_map): New function
2588 (tree_value_profile_transformations): Added check for
2589 indirect/virtual call transformation
2590 * value-prof.h (enum hist_type): New counter type for
2591 indirect/virtual calls
2592 (profile_hooks): Added new hook for profiling indirect/virtual
2594 * profile.c (instrument_values): New case for indirect/virtual
2596 * gcov-io.h (GCOV_LAST_VALUE_COUNTER): Changed to 6
2597 (GCOV_COUNTER_V_INDIR): New counter type
2598 (GCOV_COUNTER_NAMES): New name of counter "indirect" added
2599 (GCOV_MERGE_FUNCTIONS): New merge function for indirect/virtual
2601 * cgraph.c: Definition of cgraph_max_pid
2602 (cgraph_create_node): Default init of pid attribute
2603 * cgraph.h: Declaration of cgraph_max_pid
2604 (struct cgraph_node): Added pid attribute
2605 * libgcov.c (__gcov_indirect_call_profiler): New function
2606 (__gcov_one_value_profiler_body): New function
2607 (__gcov_one_value_profiler): Body was moved to
2608 __gcov_one_value_profiler_body and calls it
2610 2007-01-19 Basile Starynkevitch <basile@starynkevitch.net>
2612 * doc/gty.texi (Options): Document the mark_hook option to GTY.
2613 * gengtype.c (write_types_data, write_func_for_structure,
2614 write_types, ggc_wtd, pch_wtd): Add skip_hooks to
2615 write_types_data, ggc_wtd, pch_wtd for processing mark_hook.
2616 (walk_type, write_func_for_structure): Generate the mark_hook if
2619 2007-01-19 Jan Hubicka <jh@suse.cz>
2621 * ipa-inline.c (cgraph_decide_inlining_incrementally): Instead of
2622 'early' argument take inlining mode argument specifying whether to
2623 inline for size/speeed or all functions; add support for flattening;
2625 (cgraph_early_inlining): Update call of decide_inlining_incrementally.
2627 2007-01-19 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
2630 * toplev.c (warn_deprecated_use): Use %qD instead of %qs to print
2631 the name of the declared identifier.
2633 2007-01-19 Dirk Mueller <dmueller@suse.de>
2635 * config/i386.h (CONDITIONAL_REGISTER_USAGE): Store
2636 result of PIC_OFFSET_TABLE_REGNUM in temporary variable to avoid
2637 duplicate evaluation.
2639 2007-01-19 Uros Bizjak <ubizjak@gmail.com>
2641 * config/i386/i386.md (acos<mode>2): Rename from acossf2 and acosdf2.
2642 Macroize expander using X87MODEF12 mode macro. Extend operand 1
2643 to XFMode, use acosxf2 and truncate result to requested mode.
2644 Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
2645 (asin<mode>2): Similarly, with asin expanders.
2646 (*fscalexf4_i387): Rename from *fscalexf4.
2647 (expNcorexf3): New expander.
2648 (expxf2, exp10xf2, exp2xf2): Use expNcorexf3 expander.
2649 (exp<mode>2): Rename from expsf2 and expdf2. Macroize expander using
2650 X87MODEF12 mode macro. Extend operand 1 to XFMode, use expxf2 and
2651 truncate result to requested mode. Use SSE_FLOAT_MODE_P to disable
2652 patterns for SSE math.
2653 (exp10<mode>2): Similarly, with exp10 expanders.
2654 (exp2<mode>2): Similarly, with exp2 expanders.
2655 (expm1<mode>2): Similarly, with expm1 expanders.
2656 (ldexp<mode>3): Similarly, with ldexp expanders.
2657 (log<mode>2, log10<mode>2, log2<mode>2, log1p<mode>2, logb<mode>2):
2658 Use gen_truncxf<mode>2_i387_noop to truncate result.
2660 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2662 * config/m68k/m68k.h (PREFERRED_STACK_BOUNDARY): Define to 32
2663 for ColdFire targets.
2665 2007-01-19 Nathan Sidwell <nathan@codesourcery.com>
2666 Richard Sandiford <richard@codesourcery.com>
2668 * config/m68k/m68k.h (M68K_STATIC_CHAIN_REG_NAME): New macro.
2669 (INITIALIZE_TRAMPOLINE): Use STATIC_CHAIN_REGNUM.
2670 (__transfer_from_trampoline): Use M68K_STATIC_CHAIN_REG_NAME.
2671 * config/m68k/m68kelf.h (STATIC_CHAIN_REGNUM): Override.
2672 (M68K_STATIC_CHAIN_REG_NAME): Likewise.
2673 * config/m68k/netbsd-elf.h (M68K_STATIC_CHAIN_REG_NAME): Likewise.
2675 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2677 * config/m68k/m68k.md (adddi_dilshr32): Rename to...
2678 (*adddi_dilshr32): ...this. Fix formatting. Remove commented-out
2679 non-canonical pattern. Restrict to !TARGET_COLDFIRE.
2680 (*adddi_dilshr32_cf): New pattern.
2681 (adddi3, subdi3): Remove first alternatives.
2683 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2685 * config/m68k/m68k.c (notice_update_cc): If an SFmode move is
2686 implemented using move.l, do not use its cc result for floating-point
2689 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2691 * config/m68k/m68k.h (EXTRA_CONSTRAINT): Stop the 'T' constraint
2692 from accepting 's' constraints if flag_pic.
2694 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2696 * config/m68k/m68k.md (bordered, bunordered, buneq, bunge, bungt)
2697 (bunle, bunlt, bltgt, bordered_rev, bunordered_rev, buneq_rev)
2698 (bunge_rev, bungt_rev, bunle_rev, bunlt_rev, bltgt_rev): Change
2699 condition from TARGET_68881 to TARGET_HARD_FLOAT.
2701 2007-01-19 Sandra Loosemore <sandra@codesourcery.com>
2703 * longlong.h (count_leading_zeros, COUNT_LEADING_ZEROS_0): Add
2704 ColdFire alternatives.
2705 * config/m68k/m68k.h (CLZ_DEFINED_VALUE_AT_ZERO): New macro.
2706 * config/m68k/m68k.md (clzsi2): Define for ColdFire
2707 architectures that support ff1 instruction.
2709 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2710 Julian Brown <julian@codesourcery.com>
2712 * config/m68k/m68k.h (CONST_OK_FOR_LETTER_P): Add an 'R' case.
2713 * config/m68k/m68k.md (*movsi_cfv4): Fold into...
2714 (*movsi_cf): ...here. Remove unnecessary 'R' from 'Rg'.
2717 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2719 * config/m68k/m68k-protos.h (valid_mov3q_const): Take a HOST_WIDE_INT
2721 (output_move_const_into_data_reg, output_move_simode_const): Delete.
2722 * config/m68k/m68k.c (const_method, const_int_cost): Take a
2723 HOST_WIDE_INT instead of an rtx.
2724 (m68k_rtx_costs): Update call accordingly.
2725 (output_move_const_into_data_reg): Likewise. Fix formatting.
2726 (valid_mov3q_const): Take a HOST_WIDE_INT instead of an rtx.
2728 (output_move_simode_const): Update calls after above changes.
2729 Rework to use automatic variables and predicates like MEM_P.
2730 * config/m68k/m68k.md (pushexthisi_const): Update call to
2733 2007-01-19 Dirk Mueller <dmueller@suse.de>
2735 * tree-ssa-alias.c (perform_var_substitution): Fix typo
2738 2007-01-19 Richard Guenther <rguenther@suse.de>
2740 * builtins.c (expand_builtin_cexpi): Fall back to expanding
2741 via cexp in case sincos is not available.
2743 2007-01-19 Richard Guenther <rguenther@suse.de>
2745 * doc/tm.texi (TARGET_HAS_SINCOS): Document new target macro.
2746 * defaults.h (TARGET_HAS_SINCOS): Default to off.
2747 * config/linux.h (TARGET_HAS_SINCOS): Set to on if we have glibc.
2748 * config/alpha/linux.h (TARGET_HAS_SINCOS): Likewise.
2749 * config/sparc/linux.h (TARGET_HAS_SINCOS): Likewise.
2750 * config/sparc/linux64.h (TARGET_HAS_SINCOS): Likewise.
2751 * config/rs6000/linux.h (TARGET_HAS_SINCOS): Likewise.
2752 * config/rs6000/linux64.h (TARGET_HAS_SINCOS): Likewise.
2754 2007-01-19 Uros Bizjak <ubizjak@gmail.com>
2756 * config/i386/i386.md (*fpatanxf3_i387, fpatan_extend<mode>xf3_i387):
2758 (atan2sf3_1, atan2df3_1, atan2xf3_1): Remove insn patterns.
2759 (atan2xf3): Directly generate RTL pattern.
2760 (atan2<mode>3): Rename from atan2sf3 and atan2df3 and macroize insn
2761 patterns using X87MODEF12 mode macro. Use fpatan_extend<mode>xf3_i387
2762 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
2763 disable patterns for SSE math.
2764 (atan<mode>2): Rename from atansf2 and atandf2 and macroize insn
2765 patterns using X87MODEF12 mode macro. Use fpatan_extend<mode>xf3_i387
2766 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
2767 disable patterns for SSE math.
2769 2007-01-19 Alexandre Oliva <aoliva@redhat.com>
2771 * libgcc-std.ver: Fix typo in %inherit for GCC_4.3.0.
2773 2007-01-18 Roger Sayle <roger@eyesopen.com>
2775 * fold-const.c (fold_unary) <VIEW_CONVERT_EXPR>: Optimize away a
2776 VIEW_CONVERT_EXPR to the same type as it's operand.
2778 2007-01-18 David Edelsohn <edelsohn@gnu.org>
2780 * config/rs6000/darwin-ldouble.c: Only build _SOFT_FLOAT if
2781 configured for long double 128.
2783 2007-01-18 Mike Stump <mrs@apple.com>
2785 * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Fix build
2788 2007-01-18 Michael Meissner <michael.meissner@amd.com>
2790 * i386.c (ix86_compute_frame_layout): Make fprintf's in #if 0 code
2793 2007-01-18 Jan Hubicka <jh@suse.cz>
2795 * tree-ssa-operands.c (vop_free_bucket_size): Never return value
2796 greater than NUM_VOP_FREE_BUCKETS.
2798 2007-01-18 Daniel Berlin <dberlin@dberlin.org>
2800 * tree-ssa-structalias.c: Update comments.
2801 (ptabitmap_obstack): Removed.
2803 (oldpta_obstack): Ditto.
2804 (stats): Add a few members.
2805 (struct variable_info): Remove node, complex, address_taken, and
2806 indirect_target members. Add oldsolution member.
2807 (new_var_info): Do not initialize removed members.
2808 (constraint_expr_type): Remove INCLUDES.
2809 (constraint_graph): Add size, implicit_preds, rep,
2810 indirect_cycles, eq_rep, label, direct_nodes, and complex members.
2811 (FIRST_REF_NODE): New macro.
2812 (LAST_REF_NODE): Ditto.
2813 (FIRST_ADDR_NODE): Ditto.
2814 (find): New function.
2816 (dump_constraint): Do not handle INCLUDES.
2817 (insert_into_complex): Do not insert duplicate constraints.
2818 (condense_varmap_nodes): Renamed and rewritten into ...
2819 (merge_node_constraints): This. Also fix bug in handling of
2820 offseted copy constraints.
2821 (clear_edges_for_node): No longer need to deal with preds at all,
2822 or removing associated preds/succs.
2823 (merge_graph_nodes): Deal with indirect_cycles.
2824 Don't deal with predecessors.
2825 (add_implicit_graph_edge): New function.
2826 (add_pred_graph_edge): Ditto.
2827 (add_graph_edge): Don't deal with predecessors.
2828 (build_constraint_graph): Removed.
2829 (build_pred_graph): New function.
2830 (build_succ_graph): Ditto.
2831 (struct scc_info): Removed in_component. Added roots, dfs, and
2832 node_mapping. Remove visited_index, unification_queue.
2833 (scc_visit): Deal with union-find we do now.
2834 Deal with cycles with REF nodes.
2835 (collapse_nodes): Renamed and rewritten to ...
2836 (unify_nodes): This.
2837 (process_unification_queue): Removed.
2838 (topo_visit): Cleanup
2839 (do_da_constraint): Use find.
2840 (do_sd_constraint): Ditto.
2841 (do_ds_constraint): Ditto.
2842 (do_complex_constraint): Ditto.
2843 (init_scc_info): Update for removed and added members.
2844 (find_and_collapse_graph_cycles): Renamed and rewritten into ...
2845 (find_indirect_cycles): This.
2846 (equivalence_class): New variable.
2847 (label_visit): New function.
2848 (perform_variable_substitution): Rewritten.
2849 (free_var_substitution_info): New function.
2850 (find_equivalent_node): Ditto.
2851 (move_complex_constraints): Ditto.
2852 (eliminate_indirect_cycles): Ditto.
2853 (solve_graph): Only propagate changed bits.
2854 Use indirect cycle elimination.
2856 (tree_id_t): Rename to tree_vi_t, delete id member, add vi member.
2857 (tree_id_eq): Renamed to ...
2858 (tree_vi_eq): This. Update for member change
2859 (insert_id_for_tree): Renamed and rewritten to ...
2860 (insert_vi_for_tree): This.
2861 (lookup_id_for_tree): Renamed and rewritten to ...
2862 (lookup_vi_for_tree): This.
2863 (get_id_for_tree): Renamed and rewritten to ...
2864 (get_vi_for_tree): Ditto.
2865 (get_constraint_exp_from_ssa_var): Update to use get_vi_for_tree.
2866 (process_constraint): Don't handle INCLUDES.
2867 Remove special ADDRESSOF case.
2868 (find_func_aliases): Rewrite to use vi functions instead of id
2870 (create_function_info_for): Ditto.
2871 (create_variable_info_for): Ditto.
2872 (intra_create_variable_infos): Ditto.
2873 (merge_smts_into): Ditto.
2874 (find_what_p_points_to): Ditto.
2875 (init_base_vars): Ditto.
2876 (init_alias_vars): Ditto.
2877 (remove_preds_and_fake_succs): New function.
2878 (dump_sa_points_to_info): Dump new stats.
2879 (dump_solution_for_var): Use find.
2880 (set_used_smts): Fix formatting.
2881 (compute_points_to_sets): Updated for new functions.
2882 (ipa_pta_execute): Ditto.
2884 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
2885 Richard Sandiford <richard@codesourcery.com>
2887 * doc/tm.texi (TARGET_FUNCTION_VALUE): Expand documentation of
2889 * calls.c (expand_call): If the return value is a PARALLEL,
2890 extract its first member.
2891 * config/m68k/linux.h (FUNCTION_EXTRA_EPILOGUE): Remove.
2892 * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
2893 use FUNCTION_EXTRA_EPILOGUE.
2894 (m68k_function_value): Return a PARALLEL if the return value
2895 is of a pointer type.
2896 * config/m68k/netbsd-elf.h (current_function_returns_pointer)
2897 (FUNCTION_EXTRA_EPILOGUE): Remove.
2898 * config/m68k/m68k.md (D0_REG): New constant.
2900 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
2902 * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
2903 output a NOP for empty epilogues.
2905 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2907 * config/m68k/m68k.c (m68k_use_return_insn): Update comments
2908 before function. Extend register save check to include all
2909 registers, not just integer ones.
2911 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
2913 * config/m68k/m68k-protos.h (use_return_insn): Rename to...
2914 (m68k_use_return_insn): ...this.
2915 * config/m68k/m68k.h (USE_RETURN_INSN): Delete.
2916 * config/m68k/m68k.c (use_return_insn): Rename to...
2917 (m68k_use_return_insn): ...this.
2918 * config/m68k/m68k.md (return): Use m68k_use_return_insn instead
2921 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
2923 * config/m68k/fpgnulib.c (__truncdfsf2): Implement round to
2924 nearest even, fix denormal rounding overflow.
2926 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2928 * config/m68k/m68k.md (movsf_cf_hard): Use fsmove instead of
2929 f%$smove and f%$move.
2930 (movdf_cf_hard): Use fdmove for cases 0 and 3 and fmove for case 1.
2931 (extendsfdf2_cf): Use fdmove instead of f%&move.
2932 (truncdfsf2_cf): Use fsmove instead of f%$smove.
2933 (add<mode>3_cf, sub<mode>3_cf): Use <FP:prec> instead of <FP:round>.
2935 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2937 * config/m68k/m68k.md (movdf_cf_hard): Use output_move_double for
2940 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2942 * real.h (real_format): Add a canonical_nan_lsbs_set field.
2943 (coldfire_single_format): Declare.
2944 (coldfire_double_format): Likewise.
2945 * real.c (encode_ieee_single): Use canonical_nan_lsbs_set instead
2946 of qnan_msb_set to determine the lower bits of a canonical
2948 (encode_ieee_double): Likewise.
2949 (encode_ieee_quad): Likewise.
2950 (ieee_single_format): Initialize canonical_nan_lsbs_set.
2951 (mips_single_format): Likewise.
2952 (ieee_double_format): Likewise.
2953 (mips_double_format): Likewise.
2954 (ieee_extended_motorola_format): Likewise.
2955 (ieee_extended_intel_96_format): Likewise.
2956 (ieee_extended_intel_128_format): Likewise.
2957 (ieee_extended_intel_96_round_53_format): Likewise.
2958 (ibm_extended_format): Likewise.
2959 (mips_extended_format): Likewise.
2960 (ieee_quad_format): Likewise.
2961 (mips_quad_format): Likewise.
2962 (vax_f_format): Likewise.
2963 (vax_d_format): Likewise.
2964 (vax_g_format): Likewise.
2965 (i370_single_format): Likewise.
2966 (i370_double_format): Likewise.
2967 (decimal_single_format): Likewise.
2968 (decimal_double_format): Likewise.
2969 (decimal_quad_format): Likewise.
2970 (c4x_single_format): Likewise.
2971 (c4x_extended_format): Likewise.
2972 (real_internal_format): Likewise.
2973 (coldfire_single_format): New real_format.
2974 (coldfire_double_format): Likewise.
2975 * config/pdp11/pdp11.c (pdp11_f_format): Initialize
2976 canonical_nan_lsbs_set.
2977 (pdp11_d_format): Likewise.
2978 * config/m68k/m68k.c (override_options): Override REAL_FORMAT_MODE
2979 if TARGET_COLDFIRE_CPU.
2981 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2983 * config/m68k/m68k-protos.h (m68k_output_pic_call): Delete.
2984 (output_call, m68k_legitimize_call_address): Declare.
2985 * config/m68k/m68k.h (EXTRA_CONSTRAINT): Remove unnecessary
2986 parenthesees. Add support for a 'W' constraint.
2987 (LEGITIMATE_PIC_OPERAND_P): Remove SYMBOL_REF_FLAG handling.
2988 (PRINT_OPERAND_PUNCT_VALID_P): Remove comment about 'o'.
2989 (m68k_symbolic_call, m68k_symbolic_jump): Declare.
2990 * config/m68k/m68k.c (m68k_symbolic_call, m68k_symbolic_jump): New
2992 (override_options): Initialize them. Do not set flag_no_function_cse
2993 for TARGET_ID_SHARED_LIBRARY.
2994 (m68k_output_pic_call): Delete.
2995 (m68k_legitimize_call_address): New function.
2996 (print_operand): Remove the %o prefix. Handle the %p prefix.
2997 (output_call): New function.
2998 (m68k_output_mi_thunk): Use m68k_symbolic_jump. Always load the
2999 target address from the GOT if symbolic jumps are not allowed.
3000 * config/m68k/m68k.md (call, general_operand): Do not set
3001 SYMBOL_REF_FLAG. Use m68k_legitimize_call_address instead.
3002 Merge separate flag_pic and !flag_pic define_insns into...
3003 (*call, *call_value): ...these new patterns. Match the address
3004 rather than the containing MEM and require it to be a call_operand.
3005 Use output_call to generate the asm template.
3006 * config/m68k/predicates.md (const_call_operand): New predicate.
3007 (call_operand): Likewise.
3009 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
3011 * config/m68k/m68k.h (REGISTER_MOVE_COST): Simplify definition.
3012 (STACK_GROWS_DOWNWARD): Define to 1.
3013 (FUNCTION_VALUE, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P): Equivocate
3014 comments, emphasizing that these values are only defaults.
3015 * config/m68k/linux.h (LINK_SPEC): Fix formatting in #undef.
3016 * config/m68k/m68k.c (const_method): Remove trailing whitespace.
3018 2007-01-18 Richard Sandiford <richard@codesourcery.com>
3020 * config/m68k/m68k.md (cmpsi): Remove outdated flag_pic handling.
3022 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
3023 Richard Sandiford <richard@codesourcery.com>
3025 * config/m68k/m68k.h (DATA_REGNO_P, ADDRESS_REGNO_P, INT_REGNO_P)
3026 (FP_REGNO_P): New macros.
3027 (REGNO_OK_FOR_INDEX_P, REGNO_OK_FOR_BASE_P, REGNO_OK_FOR_DATA_P)
3028 (REGNO_OK_FOR_FP_P, REG_OK_FOR_INDEX_P, REG_OK_FOR_BASE_P): Use them.
3029 (EH_RETURN_STACKADJ_RTX): Use A0_REG.
3030 * config/m68k/m68k.c (m68k_regno_mode_ok): Use the new REGNO macros.
3032 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
3034 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
3035 (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-rtems*): Add t-floatlib
3037 * config/m68k/t-floatlib: New file, extracting common code from...
3038 * config/m68k/t-m68kbare, config/m68k/t-m68kelf,
3039 * config/m68k/t-uclinux: Here.
3040 * config/m68k/fpgnulib.c: Do not compile extendeed precision
3041 routines on ColdFire targets.
3043 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
3045 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
3046 (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*)
3047 (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-linux*)
3048 (m68k-*-rtems*): Use tm_file rather than m68k/m68k.h and
3049 explicitly set MOTOROLA to 1.
3050 * config/m68k/m68k.h (MOTOROLA): Simplify definition accordingly.
3052 2007-01-18 Richard Sandiford <richard@codesourcery.com>
3053 Nathan Sidwell <nathan@codesourcery.com>
3055 * config/m68k/m68k.h (PCC_STATIC_STRUCT_RETURN): Delete.
3056 (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Add MOTOROLA cases.
3057 * config/m68k/coff.h (REGISTER_PREFIX_MD): Delete.
3058 * config/m68k/m68020-elf.h (LIB_SPEC): Delete.
3059 * config/m68k/m68k-none.h (CC1_SPEC, CPP_SUBTARGET_SPEC): Delete.
3060 * config/m68k/m68kelf.h (IMMEDIATE_PREFIX, REGISTER_PREFIX_MD)
3061 (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Delete.
3062 (NO_DOLLAR_IN_LABEL, PCC_STATIC_STRUCT_RETURN): Don't undefine.
3063 (BSS_ASM_OP, ASM_OUTPUT_SKIP, ASM_OUTPUT_ASCII): Delete.
3064 * config/m68k/m68kemb.h (PCC_STATIC_STRUCT_RETURN): Don't undefine.
3065 * config/m68k/linux.h (SIZE_TYPE, PTRDIFF_TYPE, WCHAR_TYPE)
3066 (WCHAR_TYPE_SIZE, TARGET_OBJFMT_CPP_BUILTINS): Delete.
3067 (TARGET_OS_CPP_BUILTINS): Don't define mc68000 and mc68020 here.
3068 (DBX_REGISTER_NUMBER): Delete.
3069 * config/m68k/netbsd-elf.h (IMMEDIATE_PREFIX): Delete.
3070 (PCC_STATIC_STRUCT_RETURN): Don't undefine.
3071 * config/m68k/openbsd.h (PCC_STATIC_STRUCT_RETURN): Define.
3073 2007-01-18 Richard Sandiford <richard@codesourcery.com>
3075 * config.gcc (m68k-*-uclinux*): Add flat.h to $tm_file.
3076 * config/flat.h: New file.
3077 * crtstuff.c (USE_PT_GNU_EH_FRAME): Don't define if
3079 * config/m68k/m68k.h (ASM_PREFERRED_EH_DATA_FORMAT): Do not use
3080 indirect references for -msep-data or -mid-shared-library.
3081 Do not use PC-relative code addresses either.
3082 * config/m68k/m68k.c (override_options): Restrict -fPIC error
3084 * config/m68k/uclinux.h (STARTFILE_SPEC): Define. Use Scrt1.o
3085 for shared libraries and crt1.o for executables. Use crti.o and
3087 (ENDFILE_SPEC): Use crtend.o and crtn.o.
3088 (LIB_SPEC): Suppress -Rlibc.gdb if -static-libc is given.
3089 Do not add -elf2flt or -shared-lib-id options here.
3090 (LINK_SPEC): Define. Pass -elf2flt if no -elf2flt option is given.
3091 Pass -shared-lib-id if -mid-shared-library, taking the library
3092 identifier from -mshared-library-id if given, otherwise
3094 (EH_FRAME_IN_DATA_SECTION): Do not undefine.
3095 (INIT_SECTION_ASM_OP, FINI_SECTION_ASM_OP): Likewise.
3096 (TARGET_OS_CPP_BUILTINS): Define __GXX_MERGED_TYPEINFO_NAMES=0
3097 and __GXX_TYPEINFO_EQUALITY_INLINE=0 if -mid-shared-library.
3098 (DRIVER_SELF_SPECS): Map unadorned PIC options to -msep-data.
3099 * config/m68k/t-uclinux (EXTRA_MULTILIB_PARTS): Add crtbegin.o
3101 * config/m68k/lb1sf68.asm (PICCALL): Use an lea and pc-relative
3102 jump sequence for ISA A and ISA A+.
3103 (PICJUMP): Likewise.
3105 2007-01-18 Richard Sandiford <richard@codesourcery.com>
3107 * config.gcc (m68k-*-uclinux*): Base the port on the common
3108 and m68k GNU/Linux files rather than on the generic ELF ones.
3109 * config/m68k/uclinux.h (TARGET_VERSION): Override.
3110 (TARGET_OS_CPP_BUILTINS): Use LINUX_TARGET_OS_CPP_BUILTINS.
3112 2007-01-18 Julian Brown <julian@codesourcery.com>
3113 Richard Sandiford <richard@codesourcery.com>
3115 * config/m68k/m68k.h (LONG_DOUBLE_TYPE_SIZE): Make 64-bit on ColdFire.
3116 (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
3117 * config/m68k/netbsd-elf.h (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Undefine
3119 * config/m68k/uclinux-oldabi.h (LONG_DOUBLE_TYPE_SIZE): Redefine to
3121 (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
3123 2007-01-18 Richard Sandiford <richard@codesourcery.com>
3125 * doc/install.texi: Document m68k-uclinuxoldabi.
3126 * config.gcc (m68k-*-uclinuxoldabi): New configuration.
3127 * config/m68k/uclinux-oldabi.h: New file, copied from
3128 config/m68k/uclinux.h.
3130 2007-01-18 Kaz Kojima <kkojima@rr.iij4u.or.jp>
3132 * config/m32r/m32r-protos.h (m32r_expand_epilogue): Declare it.
3133 * config/m32r/m32r.c (m32r_setup_incoming_varargs): Use gen_frame_mem.
3134 (m32r_compute_frame_size): Use unsigned for regno.
3135 (m32r_reload_lr): Use gen_frame_mem.
3137 (m32r_output_function_epilogue): Don't output the function epilogue
3139 (m32r_expand_epilogue): New.
3140 (direct_return): Return false if the function has the interrupt
3142 (m32r_hard_regno_rename_ok): Remove code for the textual epilogue.
3143 * config/m32r/m32r.md (epilogue): New expander.
3144 (return_lr, return_rte): New insns.
3145 (return): Make it expander.
3146 (return_normal): New expander.
3148 2007-01-18 Josh Conner <jconner@apple.com>
3151 * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Add
3152 support for UNLE, UNLT, UNGE, and UNGT.
3154 2007-01-18 Jan Hubicka <jh@suse.cz>
3156 * tree-vrp.c (finalize_jump_threads): Do not call cleanup_cfg by hand.
3157 * tree-tailcall (add_virtual_phis): Likewise.
3158 (optimize_tail_call): Return TODOs.
3159 (execute_tail_calls): Return TODOs.
3160 * tree-ssa-ccp (execute_fold_all_builtins): Do cleanup_cfg via TODO.
3161 * tree-cfgcleanup.c (cleanup_tree_cfg_loop): Return if something
3163 * tree-ssa-forwprop.c (tree_ssa_forward_propagate_single_use_value):
3164 Cleanup cfg using TODO.
3165 * tree-flow.h (cleanup_tree_cfg_loop): Update prototype.
3166 * passes.c (execute_function_todo): When cleanup did something, remove
3168 * tree-cfg.c (pass_build_cfg): Add cleanup_cfg TODO.
3169 (make_edges): Don't cleanup_cfg.
3171 2007-01-18 Uros Bizjak <ubizjak@gmail.com>
3173 * reg-stack.c (subst_stack_regs_pat) [UNSPEC_SINCOS_COS,
3174 UNSPEC_XTRACT_FRACT]: Use generic code for instructions that
3175 operate on the top of stack.
3176 [UNSPEC_SINCOS_SIN, UNSPEC_XTRACT_EXP, UNSPEC_TAN]: Rewrite
3177 register handling of instructions that output to the second
3179 [UNSPEC_TAN_ONE, UNSPEC_TAN_TAN]: Remove.
3180 (move_for_stack_reg): Special-case check for dead destination
3181 stack slot for constant load of 1.0 inside UNSPEC_TAN.
3183 * config/i386/i386.md (UNSPEC_TAN): New constant.
3184 (UNSPEC_TAN_ONE, UNSPEC_TAN_TAN): Remove.
3185 (fptanxf4_i387, fptan_extend<mode>xf4_i387): New patterns
3186 to correctly model move of constant 1.0 to top stack slot.
3187 (*tandf3_1, *tansf3_1, *tanxf3_1): Remove insn patterns.
3188 (unnamed peephole2 pattern): Remove corresponding peephole2
3189 pattern that optimizes tan insn and loading of constant 1.0.
3190 (tanxf2): Use fptanxf4_i387.
3191 (tan<mode>2): Rename from tansf2 and tandf2 and macroize insn
3192 patterns using X87MODEF12 mode macro. Use fptan_extend<mode>xf4_i387
3193 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
3194 disable patterns for SSE math.
3195 (sincos<mode>3): Use truncxf<mode>2_i387_noop for truncation.
3196 (fyl2x_extend<mode>xf3_i387): Use X87MODEF12 for operand 1.
3198 2007-01-18 Dirk Mueller <dmueller@suse.de>
3199 Richard Guenther <rguenther@suse.de>
3202 * doc/invoke.texi (Warray-bounds): Document -Warray-bounds.
3203 * common.opt (Warray-bounds): Add new warning option.
3204 * c-opts.c (c_common_handle_option): Define -Warray-bounds
3206 * Makefile.in: make tree-vrp.o depend on toplev.h
3207 * tree-vrp.c (vrp_finalize): Call check_array_refs if -Warray-bounds
3209 (check_array_refs, check_array_bounds, check_array_ref): New.
3211 2007-01-18 Jan Hubicka <jh@suse.cz>
3213 * tree-ssa-ccp.c (ccp_finalize): Return if something changed.
3214 (execute_ssa_ccp): Return flags conditionally.
3215 * tree-ssa-propagate.c (substitue_and_fold): Return if something was
3217 * tree-ssa-propagate.h (substitute_and_fold): Update prototype.
3219 2007-01-18 Steven Bosscher <steven@gcc.gnu.org>
3221 * cfgcleanup.c (cleanup_cfg): Detect cfglayout mode and set
3222 the CLEANUP_CFGLAYOUT flag when in cfglayout mode.
3224 * Makefile.c (GTFILES): Add cfglayout.h.
3225 * gengtype.c (open_base_files): Likewise.
3226 * cfglayout.c (cfg_layout_function_footer,
3227 cfg_layout_function_header) Reindent to make gengtype happy.
3228 * cfglayout.h (cfg_layout_function_footer,
3229 cfg_layout_function_header): Add GTY(()) marker.
3231 * ifcvt.c (noce_try_sign_mask): Make sure INSN_B is non-null.
3233 2007-01-18 Ben Elliston <bje@au.ibm.com>
3235 * genautomata.c (write_automata): Include xstrerror output in the
3236 error message if writing the DFA description file fails.
3238 2007-01-17 H.J. Lu <hongjiu.lu@intel.com>
3240 * config/mips/mips-protos.h (mips_output_external): Make it
3242 * config/mips/iris.h (TARGET_ASM_EXTERNAL_LIBCALL): Removed.
3243 * config/mips/mips.c (irix_output_external_libcall): Likewise.
3244 (extern_list): Likewise.
3245 (extern_head): Likewise.
3246 (TARGET_ASM_FILE_END): Likewise.
3247 (mips_file_end): Likewise.
3248 (mips_output_external): Rewritten.
3250 2007-01-18 Ben Elliston <bje@au.ibm.com>
3252 * genpreds.c (write_insn_preds_c): Only write out the function
3253 body for regclass_for_constraint if we have register constraints.
3255 2007-01-17 Tom Tromey <tromey@redhat.com>
3257 * doc/sourcebuild.texi (libgcj Tests): Use sourceware.org.
3258 * doc/install.texi (Testing): Use sourceware.org.
3259 (Binaries): Likewise.
3260 (Specific): Likewise.
3261 * doc/contrib.texi (Contributors): Use sourceware.org.
3263 2007-01-17 Anatoly Sokolov <aesok@post.ru>
3265 * config/avr/avr.h (AVR_HAVE_LPMX): New macro.
3266 (AVR_ENHANCED): Rename to ...
3267 (AVR_HAVE_MUL): ... new.
3268 (avr_enhanced_p): Rename to ...
3269 (avr_have_mul_p): ... new.
3270 (TARGET_CPU_CPP_BUILTINS): Use 'avr_have_mul_p' instead of
3271 'avr_enhanced_p' for "__AVR_ENHANCED__". Define "__AVR_HAVE_MUL__".
3272 * config/avr/avr.c (avr_enhanced_p): Rename to ...
3273 (avr_have_mul_p): ... new.
3274 (base_arch_s): Rename 'enhanced' to 'have_mul'.
3275 (avr_override_options): Use 'avr_have_mul_p' and 'have_mul' instead of
3276 'avr_enhanced_p' and 'enhanced'.
3277 (ashlhi3_out, ashrhi3_out, lshrhi3_out, avr_rtx_costs): Use
3278 AVR_HAVE_MUL instead of AVR_ENHANCED.
3279 * avr.md (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
3280 (mulqi3, *mulqi3_enh, *mulqi3_call, mulqihi3, umulqihi3, mulhi3,
3281 *mulhi3_enh, *mulhi3_call, mulsi3, *mulsi3_call): Use AVR_HAVE_MUL
3282 instead of AVR_ENHANCED.
3283 (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
3284 * libgcc.S: Use __AVR_HAVE_MUL__ instead of __AVR_ENHANCED__.
3285 (__tablejump__): Use __AVR_HAVE_LPMX__ instead of __AVR_ENHANCED__.
3287 2007-01-17 Ian Lance Taylor <iant@google.com>
3289 * vec.h (VEC_reserve_exact): Define.
3290 (vec_gc_p_reserve_exact): Declare.
3291 (vec_gc_o_reserve_exact): Declare.
3292 (vec_heap_p_reserve_exact): Declare.
3293 (vec_heap_o_reserve_exact): Declare.
3294 (VEC_OP (T,A,reserve_exact)): New static inline function, three
3296 (VEC_OP (T,A,reserve)) [all versions]: Remove handling of
3298 (VEC_OP (T,A,alloc)) [all versions]: Call ...reserve_exact.
3299 (VEC_OP (T,A,copy)) [all versions]: Likewise.
3300 (VEC_OP (T,a,safe_grow)) [all versions]: Likewise.
3301 * vec.c (calculate_allocation): Add exact parameter. Change all
3303 (vec_gc_o_reserve_1): New static function, from vec_gc_o_reserve.
3304 (vec_gc_p_reserve, vec_gc_o_reserve): Call vec_gc_o_reserve_1.
3305 (vec_gc_p_reserve_exact, vec_gc_o_reserve_exact): New functions.
3306 (vec_heap_o_reserve_1): New static function, from vec_heap_o_reserve.
3307 (vec_heap_p_reserve, vec_heap_o_reserve): Call vec_heap_o_reserve_1.
3308 (vec_heap_p_reserve_exact): New function.
3309 (vec_heap_o_reserve_exact): New function.
3311 2007-01-17 Jan Hubicka <jh@suse.cz>
3313 * ipa-type-escape.c (look_for_casts): Revamp using handled_component_p.
3315 2007-01-17 Eric Christopher <echristo@apple.com>
3317 * config.gcc: Support core2 processor.
3319 2007-01-16 Jan Hubicka <jh@suse.cz>
3321 * tree-ssanames.c (release_dead_ssa_names): Instead of ggc_freeing
3322 the names, just unlink the chain so we don't crash on dangling pointers
3325 2007-01-16 Jan Hubicka <jh@suse.cz>
3327 * cgraph.h (cgraph_decide_inlining_incrementally): Kill.
3328 * tree-pass.h: Reorder to make IPA passes appear toegher.
3329 (pass_early_inline, pass_inline_parameters, pass_apply_inline): Declare.
3330 * cgraphunit.c (cgraph_finalize_function): Do not compute inling
3331 parameters, do not call early inliner.
3332 * ipa-inline.c: Update comments. Include tree-flow.h
3333 (cgraph_decide_inlining): Do not compute inlining parameters.
3334 (cgraph_decide_inlining_incrementally): Return TODOs; assume to
3335 be called with function context set up.
3336 (pass_ipa_inline): Remove unreachable functions before pass.
3337 (cgraph_early_inlining): Simplify assuming to be called from the
3339 (pass_early_inline): New pass.
3340 (cgraph_gate_ipa_early_inlining): New gate.
3341 (pass_ipa_early_inline): Turn into simple wrapper.
3342 (compute_inline_parameters): New function.
3343 (gate_inline_passes): New gate.
3344 (pass_inline_parameters): New pass.
3345 (apply_inline): Move here from tree-optimize.c
3346 (pass_apply_inline): New pass.
3347 * ipa.c (cgraph_remove_unreachable_nodes): Verify cgraph after
3349 * tree-inline.c (optimize_inline_calls): Return TODOs rather than
3351 (tree_function_versioning): Do not allocate dummy struct function.
3352 * tree-inline.h (optimize_inline_calls): Update prototype.
3353 * tree-optimize.c (execute_fixup_cfg): Export.
3354 (pass_fixup_cfg): Remove
3355 (tree_rest_of_compilation): Do not apply inlines.
3356 * tree-flow.h (execute_fixup_cfg): Declare.
3357 * Makefile.in (gt-passes.c): New.
3358 * passes.c: Include gt-passes.h
3359 (init_optimization_passes): New passes.
3360 (nnodes, order): New static vars.
3361 (do_per_function_toporder): New function.
3362 (execute_one_pass): Dump current pass here.
3363 (execute_ipa_pass_list): Don't dump current pass here.
3365 2007-01-16 Janis Johnson <janis187@us.ibm.com>
3367 * config/dfp-bit.c (dfp_compare_op): Return separate value for NaN.
3368 (DFP_NE, DFP_LE, DFP_GE): Return false for NaN.
3370 2007-01-16 David Edelsohn <edelsohn@gnu.org>
3372 * config/rs6000/darwin-ldouble.c: Build file for SOFT_FLOAT.
3373 (strong_alias): Define.
3374 (__gcc_qmul): Provide non-FMA for soft-float.
3380 (__gcc_qunord): New.
3390 * config/rs6000/rs6000.c (rs6000_init_libfuncs): Initialize
3391 soft-float functions.
3392 * config/rs6000/libgcc-ppc-glibc.ver: Version soft-float symbols.
3393 * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Do not warn
3394 about long double soft float.
3396 2007-01-16 Dorit Nuzman <dorit@il.ibm.com>
3397 Tehila Meyzels <tehila@il.ibm.com>
3399 * tree-vectorizer.h (is_pattern_stmt_p): New.
3400 * tree-vect-analyze.c (vect_determine_vectorization_factor): Fix
3401 formatting (tabs instead of spaces). Cleanup and clarify setting
3402 of STMT_VINFO_VECTYPE. Call is_pattern_stmt_p.
3403 * tree-vect-transform.c (vect_get_vec_def_for_operand): Fix typo.
3404 (vectorizable_type_demotion): Check that types are integral.
3405 (vectorizable_type_promotion): Likewise.
3406 (vectorizable_store): Fix typo. Eliminate new-line at end of
3409 2007-01-16 Jan Hubicka <jh@suse.cz>
3411 * tree-ssanames.c (release_dead_ssa_names): Remove invalidated
3414 2007-01-15 Eric Christopher <echristo@apple.com>
3416 * ifcvt.c: Include vec.h, vecprim.h.
3417 (check_cond_move_block): New argument regs.
3418 Reorganize. Add registers used to regs.
3419 (cond_move_process_if_block): Use regs set above as
3422 2007-01-15 Eric Christopher <echristo@apple.com>
3424 * config/darwin.h: Update copyright.
3425 (TARGET_OPTION_TRANSLATE_TABLE): Add umbrella.
3426 (LINK_COMMAND_SPEC): Add -u.
3427 (LINK_SPEC): Fix umbrella for above.
3429 2007-01-15 Joseph S. Myers <joseph@codesourcery.com>
3431 * config/soft-fp/op-common.h, config/soft-fp/op-4.h: Update from
3434 2007-01-15 Tom Tromey <tromey@redhat.com>
3436 * doc/sourcebuild.texi (libgcj Tests): Don't mention jacks.
3437 * doc/install.texi (Testing): Don't mention jacks.
3438 (Configuration): Document --enable-java-maintainer-mode. Move
3439 --with-java-home to libgcj-specific section. Document
3441 (Prerequisites): Mention --enable-java-maintainer-mode, ecj1.
3443 2007-01-15 Jan Hubicka <jh@suse.cz>
3445 * tree-ssa-dce.c (DCE_TODOs): New.
3446 (propagate_necessity): Return if something changed.
3447 (eliminate_unnecessary_stmts): Likewise.
3448 (perform_tree_ssa_dce): Return TODO flags when needed.
3449 (pass_dce, pass_dce_loop, pass_cd_dce): Remove TODO flags.
3451 2007-01-15 Uros Bizjak <ubizjak@gmail.com>
3453 * config/i386/i386.md (fyl2xxf3_i387): Rename from fyl2x_xf3.
3454 (fyl2x_extend<mode>xf3_i387): New insn pattern.
3455 (log<mode>2): Rename from logsf2 and logdf2 and macroize insn
3456 patterns using X87MODEF12 mode macro. Extend operand 1
3457 to XFmode. Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
3458 (log10<mode>2): Ditto.
3459 (log2<mode>2): Ditto.
3460 (log1p<mode>2): Ditto.
3461 (logb<mode>2): Ditto.
3462 (fyl2xp1xf3_i387): Rename from fyl2xp1_xf3.
3463 (fyl2xp1_extend<mode>xf3_i387): New insn pattern.
3464 (*fxtractxf3_i387): Rename from *fxtractxf3.
3465 (fxtract_extend<mode>xf3_i387): New insn pattern.
3466 (ilogbsi2): Use match_dup 3, not match_operand:XF 3.
3467 * config/i386/i386.c (ix86_emit_i387_log1p): Use gen_fyl2xp1xf3_i387()
3468 and gen_fyl2xxf3_i387().
3470 2007-01-14 Zdenek Dvorak <dvorakz@suse.cz>
3472 * loop-unswitch.c (unswitch_loop): Do not call fix_loop_placement.
3473 * cfgloopmanip.c (fix_loop_placement): Made static. Use
3474 get_loop_exit_edges. Changed return type to bool.
3475 * cfgloop.h (fix_loop_placement): Declaration removed.
3477 2007-01-14 Dorit Nuzman <dorit@il.ibm.com>
3479 * param.h (MIN_VECT_LOOP_BOUND): New.
3480 * params.def (MIN_VECT_LOOP_BOUND): New.
3481 * tree-vectorizer.c (slpeel_tree_peel_loop_to_edge): Takes another
3482 argument - minimum threshold for number of iterations.
3483 * tree-vectorizer.h (slpeel_tree_peel_loop_to_edge): Add another
3484 argument to declaration.
3485 * tree-vect-analyze.c (vect_analyze_operations): Check value of
3486 MIN_VECT_LOOP_BOUND.
3487 * tree-vect-transform.c (vect_do_peeling_for_loop_bound): Call
3488 slpeel_tree_peel_loop_to_edge with additional argument.
3489 (vect_do_peeling_for_alignment): Likewise.
3490 * doc/invoke.texi (min-vect-loop-bound): Document new param option.
3492 2007-01-14 Uros Bizjak <ubizjak@gmail.com>
3495 * config/i386/i386.c (print_operand) ['z']: Output 'b' for
3498 2007-01-14 Jan Hubicka <jh@suse.cz>
3500 * tree-dfa.c (remove_referenced_var): New function.
3501 * tree-ssa-live.c (remove_unused_locals): Walk referenced vars and
3502 prune referenced vars list too.
3503 * tree-flow.h (remove_referenced_var): Declare.
3505 2007-01-14 Jan Hubicka <jh@suse.cz>
3507 * tree-eh.c (add_stmt_to_eh_region_fn): Do not add call_exprs
3509 (remove_stmt_from_eh_region_fn): Do not remove call_exprs.
3510 (verify_eh_throw_stmt_node, verify_eh_throw_table_statements): Kill.
3511 * except.h (verify_eh_throw_table_statements): Kill prototype.
3512 * cfgexpand.c (expand_gimple_basic_block): Propagate Eh regions
3513 into call exrepssions.
3514 * tree-optimize.c (execute_free_cfg_annotatiosn): Do not call
3516 * tree-cfg.c: Include pointer-set.h.
3517 (verify_node_sharing): Work on pointer set.
3518 (verify_eh_throw_stmt_node): New.
3519 (verify_stmts): Use pointers sets, verify throw_stmt.
3521 2007-01-13 Zdenek Dvorak <dvorakz@suse.cz>
3523 * ipa-reference.c (analyze_function): Consider also addresses taken
3526 2007-01-12 Roger Sayle <roger@eyesopen.com>
3528 * c-typeck.c (null_pointer_constant_p): Replace use of
3529 TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
3530 (build_c_cast): Likewise.
3532 2007-01-12 Roger Sayle <roger@eyesopen.com>
3534 * tree.h (force_fit_type_double): Remove unused final argument.
3535 * c-common.c (constant_expression_warning): Replace use of
3536 TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
3537 (convert_and_check): Likewise.
3538 (shorten_compare): Update call to force_fit_type_double.
3539 (c_common_truthvalue_conversion) <INTEGER_CST>: Use integer_zerop.
3540 * convert.c (convert_to_pointer): Update call to
3541 force_fit_type_double.
3542 * fold-const.c (force_fit_type_double): Remove overflowed_const
3544 (int_const_binop, fold_convert_const_int_from_int,
3545 fold_convert_const_int_from_real, fold_div_compare,
3546 fold_sign_changed_comparison, fold_unary, fold_negate_const,
3547 fold_abs_const, fold_not_const): Remove the final argument from
3548 calls to force_fit_type_double.
3550 2007-01-12 Andrew Pinski <andrew_pinski@playstation.sony.com>
3552 * configure.ac: Set insn to "nop" for spu-*-* also.
3553 * configure: Regenerate.
3555 2007-01-12 Olga Golovanevsky <olga@il.ibm.com>
3557 * builtins.def : Add BUILT_IN_FREE.
3559 2007-01-12 Jan Hubicka <jh@suse.cz>
3561 PR tree-optimization/30443
3562 * tree-inline.c (tree_function_versioning): Do not optimize when
3563 cloning for inlining.
3565 2007-01-12 Zdenek Dvorak <dvorakz@suse.cz>
3567 * doc/loop.texi: Document recording of loop exits.
3568 * cfgloopmanip.c (loopify, duplicate_loop): Use alloc_loop.
3569 (update_single_exits_after_duplication,
3570 update_single_exit_for_duplicated_loop,
3571 update_single_exit_for_duplicated_loops): Removed.
3572 (duplicate_loop_to_header_edge): Do not call
3573 update_single_exits_after_duplication and
3574 update_single_exit_for_duplicated_loops.
3575 (loop_version): Do not update single_exit information.
3576 (fix_loop_structure): Use record_loop_exits instead of
3577 mark_single_exit_loops.
3578 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update
3579 the lists of loop exits.
3580 * cfghooks.c (redirect_edge_and_branch, redirect_edge_and_branch_force,
3581 split_edge, merge_blocks): Update the lists of loop exits.
3582 * modulo-sched.c (sms_schedule): Pass LOOPS_HAVE_RECORDED_EXITS to
3583 loop_optimizer_init.
3584 * loop-init.c (loop_optimizer_init): Call record_loop_exits instead
3585 of mark_single_exit_loops.
3586 (loop_optimizer_finalize): Call release_recorded_exits.
3587 * tree-ssa-loop.c (tree_loop_optimizer_init): Pass
3588 LOOPS_HAVE_RECORDED_EXITS to loop_optimizer_init.
3589 * tree-vectorizer.c (slpeel_tree_duplicate_loop_to_edge_cfg): Do not
3590 update single exit information.
3591 * lambda-code.c (perfect_nestify): Ditto.
3592 * cfgloop.c (flow_loop_free): Destroy the list of exits of the loop.
3593 (mark_single_exit_loops): Removed.
3594 (alloc_loop, loop_exit_hash, loop_exit_eq, loop_exit_free,
3595 get_exit_descriptions, rescan_loop_exit, record_loop_exits,
3596 dump_recorded_exit, dump_recorded_exits, release_recorded_exits): New
3598 (get_loop_exit_edges, single_exit): Use recorded exit lists.
3599 (add_bb_to_loop, remove_bb_from_loops): Update the lists of loop exits.
3600 (verify_loop_structure): Verify consistency of the exit lists.
3601 (flow_loops_find): Use alloc_loop. Initialize exits hash.
3602 (set_single_exit): Removed.
3603 * cfgloop.h (struct loop_exit): New function.
3604 (struct loop): single_exit_ field replaced by exits field.
3605 (LOOPS_HAVE_MARKED_SINGLE_EXITS): Replaced by LOOPS_HAVE_RECORDED_EXITS.
3606 (struct loops): Added exits hash.
3607 (mark_single_exit_loops, set_single_exit): Declaration removed.
3608 (release_recorded_exits, record_loop_exits, rescan_loop_exit): Declare.
3610 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3612 * doc/invoke.texi: Avoid use of @headitem.
3614 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3616 * cse.c (cse_insn): Move HAVE_CC0 code after declarations.
3618 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3620 * doc/install.texi: Fix m68k-*-* anchor and add m68k-*-* to the
3623 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
3624 Richard Sandiford <richard@codesourcery.com>
3626 * doc/invoke.texi: Document -mno-bitfield, -mno-rtd and -mno-short.
3627 * config/m68k/m68k.opt: Resort options.
3628 (mbitfield, mrtd, mshort): Remove RejectNegative properties.
3630 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
3631 Richard Sandiford <richard@codesourcery.com>
3633 * doc/invoke.texi: Document the macros that are defined by
3634 m68k's -mtune and -mhard-float options.
3635 * config/m68k/m68k-protos.h (m68k_cpp_cpu_ident) Declare.
3636 (m68k_cpp_cpu_family): Likewise.
3637 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Add a full set
3638 of __ucfv*__ macros. Define __mcffpu__ if generating code for
3639 ColdFire FPUs. Define __mcf_cpu_* and __mcf_family_* macros.
3640 * config/m68k/m68k.c (m68k_cpp_cpu_ident): New function.
3641 (m68k_cpp_cpu_family): Likewise.
3643 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3645 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Treat all mc68*
3646 macros besides mc68000 as tuning macros. Use a switch statement
3647 to set them and mcpu32.
3649 2007-01-12 Julian Brown <julian@codesourcery.com>
3651 * config/m68k/m68k.h: Use TARGET_68040 instead of TARGET_68040_ONLY.
3652 (TARGET_68040_ONLY): Rename to...
3653 (TARGET_68040): ...this.
3654 * config/m68k/m68k.c: Use TARGET_68040 instead of TARGET_68040_ONLY.
3655 * config/m68k/m68k.md: Likewise.
3657 2007-01-12 Julian Brown <julian@codesourcery.com>
3658 Nathan Sidwell <nathan@codesourcery.com>
3659 Richard Sandiford <richard@codesourcery.com>
3661 * config.gcc (m680[012]0-*-*, m68k*-*-*): Set m68k_cpu_ident to
3662 the -mcpu= argument associated with the --with-cpu setting.
3663 Define M68K_DEFAULT_TUNE to the default -mtune= option,
3664 if different from the one implied by the -mcpu setting.
3665 Accept --with-cpu=FOO if FOO is listed in m68k-devices.def,
3666 using mcpu=FOO as the default CPU option. Set target_cpu_default2.
3667 * doc/invoke.texi: Mention ColdFire in the introduction to the
3668 m68k options. Document the new -march, -mcpu, -mtune, -mdiv,
3669 -mno-div and -mhard-float options. Make -m68881 a synonym for
3670 -mhard-float. Document the previously-undocumented -m5206e,
3671 -m528x, -m5307 and -m5407 options. Tweak the existing option
3672 documentation for consistency.
3673 * doc/install.texi: Mention new --with-cpu arguments.
3674 * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Only use the
3675 default CPU if neither -mcpu nor -march are specified.
3676 (ASM_CPU_SPEC): Pass down -mcpu and -march options.
3677 (TARGET_CPU_CPP_BUILTINS): Set __mcfisa*__ macros from
3678 TARGET_ISA*. Set the legacy __mcf*__ cpu macros in the same way,
3679 using m68k_tune to decide between families that implement the
3680 same ISA. Use m68k_tune to set __mcfv4e__.
3681 (FL_BITFIELD, FL_68881, FL_COLDFIRE, FL_CF_HWDIV, FL_CF_MAC)
3682 (FL_CF_EMAC, FL_CF_EMAC_B, FL_CF_USP, FL_CF_FPU, FL_ISA_68000)
3683 (FL_ISA_68010, FL_ISA_68020, FL_ISA_68040, FL_ISA_A, FL_ISA_B)
3684 (FL_ISA_C, FL_ISA_MMU): New macros.
3685 (MASK_COLDFIRE): Delete.
3686 (TARGET_68010, TARGET_68020, TARGET_68040_ONLY, TARGET_COLDFIRE)
3687 (TARGET_ISAB): Redefine in terms of m68k_cpu_flags.
3688 (TARGET_68881, TARGET_COLDFIRE_FPU): Redefine in terms of m68k_fpu.
3689 (TARGET_HARD_FLOAT): Do not define here.
3690 (TARGET_ISAAPLUS, TARGET_ISAC): New macros.
3691 (TUNE_68000): New macro.
3692 (TUNE_68000_10): Redefine in terms of TUNE_68000 and TUNE_68010.
3693 (TUNE_68010, TUNE_68030, TUNE_68040, TUNE_68060, TUNE_CPU32)
3694 (TUNE_CFV2): Redefine in terms of m68k_tune.
3695 (uarch_type, target_device, fpu_type): New enums.
3696 (m68k_cpu, m68k_tune, m68k_fpu, m68k_cpu_flags): Declare.
3697 * config/m68k/m68k.c (TARGET_DEFAULT): Remove MASK_68881.
3698 (FL_FOR_isa_00, FL_FOR_isa_10, FL_FOR_isa_20, FL_FOR_isa_40)
3699 (FL_FOR_isa_cpu32, FL_FOR_isa_a, FL_FOR_isa_aplus, FL_FOR_isa_b)
3700 (FL_FOR_isa_c): New macros.
3701 (m68k_isa): New enum.
3702 (m68k_target_selection): New structure.
3703 (all_devices, all_isas, all_microarchs): New tables.
3704 (m68k_cpu_entry, m68k_arch_entry, m68k_tune_entry, m68k_cpu)
3705 (m68k_tune, m68k_fpu, m68k_cpu_flags): New variables.
3706 (MASK_ALL_CPU_BITS): Delete.
3707 (m68k_find_selection): New function.
3708 (m68k_handle_option): Handle -mcpu=, -march= and -mtune=.
3709 Map the legacy target options to a combination of the new ones.
3710 (override_options): Set m68k_cpu, m68k_tune, m68k_fpu and
3711 m68k_cpu_flags. Handle M68K_DEFAULT_TUNE. Use m68k_cpu_flags
3712 to derive default MASK_BITFIELD, MASK_CF_HWDIV and MASK_HARD_FLOAT
3714 * config/m68k/m68k.opt (m5200, m5206e, m528x, m5307, m5407, mcfv4e)
3715 (m68010, m68020, m68020-40, m68020-60, m68030, m68040): Remove Mask
3717 (m68881, msoft-float): Change mask from 68881 to HARD_FLOAT.
3718 (march=, mcpu=, mdiv, mhard-float, mtune=): New options.
3719 * config/m68k/m68k-devices.def: New file.
3721 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3722 Nathan Sidwell <nathan@codesourcery.com>
3724 * config/m68k/m68k.h (ASM_CPU_SPEC, ASM_SPEC, EXTRA_SPECS)
3725 (SUBTARGET_EXTRA_SPECS): New macros.
3726 * config/m68k/linux.h (ASM_SPEC): Remove CPU flags;
3727 use %(asm_cpu_spec) instead.
3728 * config/m68k/m68k-none.h (ASM_SPEC): Likewise.
3729 * config/m68k/openbsd.h (ASM_SPEC): Likewise.
3730 * config/m68k/netbsd-elf.h (ASM_SPEC): Likewise.
3731 (EXTRA_SPECS): Rename to...
3732 (SUBTARGET_EXTRA_SPECS): ...this.
3734 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
3735 Richard Sandiford <richard@codesourcery.com>
3736 Julian Brown <julian@codesourcery.com>
3738 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
3739 (m68k-*-uclinux*, m68k-*-linux*, m68k-*-rtems*): Set default_m68k_cpu
3740 to the configuration's default CPU.
3741 (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*): Likewise.
3742 Remove default masks.
3743 (m680[012]0-*-*): Set the default with_cpu to the first part of
3745 (m68k*-*-*): Set the default with_cpu to m$default_m68k_cpu.
3746 (m68k*-*-linux): Extend the --with-cpu handling to...
3747 (m680[012]0-*-*, m68k*-*-*): ...these configurations. Allow m68000
3748 and m68010. Don't set target_cpu_default2.
3749 * doc/install.texi: Document --with-cpu for m68k.
3750 * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Define.
3751 * config/m68k/m68k-none.h (TARGET_CPU_DEFAULT, M68K_CPU_m68k)
3752 (M68K_CPU_m68000, M68K_CPU_m68010, M68K_CPU_m68020, M68K_CPU_m68030)
3753 (M68K_CPU_m68040, M68K_CPU_m68302, M68K_CPU_m68332, TARGET_DEFAULT)
3754 (ASM_CPU_DEFAULT_SPEC, CC1_CPU_DEFAULT_SPEC): Delete.
3755 (ASM_SPEC): Remove use of %(asm_cpu_default).
3756 (EXTRA_SPECS, SUBTARGET_EXTRA_SPECS, MULTILIB_DEFAULTS): Delete.
3757 * config/m68k/linux.h (TARGET_DEFAULT): Delete.
3758 (CPP_SPEC): Merge definitions. Do not handle __HAVE_68881__ here.
3759 * config/m68k/netbsd-elf.h (TARGET_OS_CPP_BUILTINS): Define
3760 __HAVE_FPU__ if TARGET_HARD_FLOAT.
3761 (TARGET_DEFAULT): Delete.
3762 (EXTRA_SPECS): Delete cpp_cpu_default_spec, cpp_cpu_spec,
3763 cpp_fpu_spec, asm_default_spec and netbsd_cpp_spec.
3764 (CPP_CPU_SPEC): Delete.
3765 (TARGET_VERSION): Merge definitions, using TARGET_68010 to pick
3766 the appropriate string.
3767 (CPP_CPU_DEFAULT_SPEC, ASM_DEFAULT_SPEC, CPP_FPU_SPEC): Delete.
3768 (CPP_SPEC): Define to NETBSD_CPP_SPEC.
3769 (ASM_SPEC): Don't use %(asm_default_spec).
3770 * config/m68k/m68k.c (TARGET_DEFAULT_TARGET_FLAGS): Remove
3771 TARGET_DEFAULT and add MASK_68881.
3772 * config/m68k/m68k.md: Remove mention of TARGET_DEFAULT from comments.
3774 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3776 * config.gcc (m68010-*-netbsdelf*): Add MASK_68010.
3777 (m68k*-*-netbsdelf*, m68k*-*-openbsd*, m68k*-linux*): Add
3778 MASK_68010 alongside MASK_68020.
3779 * doc/invoke.texi: Document -m68010.
3780 * config/m68k/m68k.opt (m68010): New.
3781 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Define mc68010
3783 (TUNE_68010): New macro.
3784 * config/m68k/m68k-none.h (M68K_CPU_m68k, M68K_CPU_m68010)
3785 (M68K_CPU_m68020, M68K_CPU_m68030, M68K_CPU_m68040)
3786 (M68K_CPU_m68332): Add MASK_68010.
3787 * config/m68k/linux.h (TARGET_DEFAULT): Add MASK_68010 to
3788 fallback definition.
3789 * config/m68k/netbsd-elf.h (CPP_CPU_SPEC): Remove now-redundant
3791 * config/m68k/m68k.c (MASK_ALL_CPU_BITS): Add MASK_68010.
3792 (m68k_handle_option): Handle OPT_m68010. Add MASK_68010
3793 to all entries that use MASK_68020.
3794 (output_move_simode_const, output_move_himode, output_move_qimode)
3795 (output_move_stricthi, output_move_strictqi): Use TARGET_68010
3796 instead of TARGET_68020 to select clr behavior. Remove comment
3797 about there being no TARGET_68010.
3798 * config/m68k/m68k.md: Likewise throughout.
3800 2007-01-12 Julian Brown <julian@codesourcery.com>
3802 * config/m68k/m68k.h (TARGET_ISAB): New macro.
3803 * config/m68k/m68k.c: Use TARGET_ISAB rather than TARGET_CFV4.
3804 * config/m68k/m68k.md: Likewise.
3806 2007-01-12 Julian Brown <julian@codesourcery.com>
3808 * config/m68k/m68k.h (LEGITIMATE_INDEX_P, LEGITIMIZE_ADDRESS): Use
3809 TARGET_COLDFIRE_FPU instead of TARGET_CFV4E.
3811 2007-01-12 Julian Brown <julian@codesourcery.com>
3813 * config/m68k/m68k.h (TUNE_68040_60): New macro.
3814 * config/m68k/m68k.c (standard_68881_constant_p): Use it.
3815 * config/m68k/m68k.md: Likewise.
3817 2007-01-12 Julian Brown <julian@codesourcery.com>
3818 Richard Sandiford <richard@codesourcery.com>
3820 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Use TUNE_68030
3821 instead of TARGET_68030, TUNE_68040 instead of TARGET_68040,
3822 TUNE_68060 instead of TARGET_68060 and TUNE_CPU32 instead of
3824 (TARGET_CPU32): Rename to...
3825 (TUNE_CPU32): ...this.
3826 (TUNE_68000_10, TUNE_68030, TUNE_68040, TUNE_68060)
3827 (TUNE_CFV2): New macros.
3828 * config/m68k/netbsd-elf.h (LONG_DOUBLE_TYPE_SIZE): Simplify;
3829 remove conditions that are implied by TARGET_68020.
3830 * config/m68k/m68k.c (m68k_output_function_prologue): Use TUNE_68040
3831 instead of TARGET_68040 and TUNE_CPU32 instead of TARGET_CPU32.
3832 (m68k_output_function_epilogue): Likewise.
3833 (m68k_rtx_costs): Likewise. Use TUNE_68060 instead of TARGET_68060
3834 and TUNE_CFV2 instead of TARGET_5200. Use TUNE_68000_10 instead of
3835 "!TARGET_68020 && !TARGET_COLDFIRE" to choose between 68000 and
3836 non-68000 timings. Refactor multiplication and division costs.
3837 (output_addsi3): Use TUNE_68040 instead of TARGET_68040 and
3838 TUNE_CPU32 instead of TARGET_CPU32.
3839 (standard_68881_constant_p): Use TUNE_68040 instead of TARGET_68040
3840 and TUNE_68060 instead of TARGET_68060.
3841 * config/m68k/m68k.md: Use TUNE_68040 instead of TARGET_68040,
3842 TUNE_68060 instead of TARGET_68060, and TUNE_CPU32 instead of
3844 (movsi_const0): Use TUNE_68000_10 rather than "!TARGET_68020
3845 && !TARGET_COLDFIRE" to choose between moveq and clr.
3846 Likewise in the unnamed movsf pattern.
3847 (ashlsi_17_24, lshrsi_17_24): Guard with TUNE_68000_10 rather than
3848 "!TARGET_68020 && !TARGET_COLDFIRE". Likewise the unnamed
3851 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3853 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Increase amount
3854 of tabbing before backslashes.
3856 2007-01-11 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
3858 * pa-linux.h (ASM_OUTPUT_INTERNAL_LABEL): Undefine.
3859 * pa.h (ASM_OUTPUT_LABEL): Output colon when using GAS.
3860 (ASM_OUTPUT_INTERNAL_LABEL): Define.
3862 2007-01-11 Zdenek Dvorak <dvorakz@suse.cz>
3864 * tree-ssa-loop-ivopts.c (extract_cond_operands): Split from
3865 find_interesting_uses_cond.
3866 (find_interesting_uses_cond): Use extract_cond_operands.
3867 (rewrite_use_compare): Use extract_cond_operands and
3868 force_gimple_operand_bsi. Do not call update_stmt.
3869 (determine_use_iv_cost_condition): Use extract_cond_operands.
3870 Return cheaper of using original bound and changing the exit bound.
3872 2007-01-11 Zdenek Dvorak <dvorakz@suse.cz>
3874 PR tree-optimization/29516
3875 * tree-ssa-address.c (tree_mem_ref_addr, add_to_parts,
3876 most_expensive_mult_to_index, addr_to_parts,
3877 create_mem_ref, maybe_fold_tmr): Make the type of
3878 fields of TARGET_MEM_REF sizetype.
3879 (move_fixed_address_to_symbol, move_pointer_to_base):
3881 * tree.def (TARGET_MEM_REF): Add comment on types of
3884 2007-01-11 Joseph Myers <joseph@codesourcery.com>
3886 * c-common.c (vector_types_convertible_p): Treat opaque types as
3887 always convertible if they have the same size, but not otherwise.
3889 2007-01-11 Steven Bosscher <steven@gcc.gnu.org>
3891 * ifcvt.c (struct noce_if_info): Add comments to the fields.
3892 Remove the b_unconditional field.
3893 (noce_try_sign_mask): Do not look at b_unconditional.
3894 (noce_process_if_block): Do not use merge_if_blocks. Update
3895 the CFG here. Do not set b_unconditional.
3896 (cond_move_process_if_block): Likewise.
3897 (find_cond_trap): Likewise.
3898 (check_cond_move_block): Require simple jump insns at the end
3901 2007-01-11 Jan Hubicka <jh@suse.cz>
3903 PR tree-optimization/1046
3904 * tree-tailcall.c (suitable_for_tail_call_opt_p): Use TREE_ADDRESSABLE
3905 when alias info is not ready.
3906 (pass_tail_recursion): Do not require aliasing.
3907 * tree-ssa-copyrename.c (pass_rename_ssa_cop): Likewise.
3908 * tree-ssa-ccp.c (pass_ccp, pass_fold_builtins): Likewise.
3909 * tree-ssa-copy.c (pass_copy_prop): Likewise.
3910 * tree-ssa-forwprop.c (pass_forwprop): Likewise.
3911 * tree-ssa-dce.c (pass_dce, pass_dce_loop, pass_cd_dce): Likewise.
3912 * passes.c (init_optimization_passes): Execute rename_ssa_copies,
3913 ccp, forwprop, copy_prop, merge_phi, copy_prop, dce and tail recursion
3915 * tree-ssa-operands.c (add_virtual_operand, get_indirect_ref_operand):
3916 When aliasing is not build, mark statement as volatile.
3918 2007-01-11 Tom Tromey <tromey@redhat.com>
3920 PR preprocessor/15185, PR preprocessor/20989:
3921 * doc/cppopts.texi <-MT>: Update description of algorithm for
3922 computing default target.
3923 <-M, -MD>: Reword "basename" text.
3925 2007-01-11 Roger Sayle <roger@eyesopen.com>
3927 * builtins.c (expand_builtin_pow, expand_builtin_powi,
3928 fold_builtin_cabs, fold_builtin_sqrt, fold_builtin_trunc,
3929 fold_builtin_floor, fold_builtin_ceil, fold_builtin_round,
3930 fold_builtin_int_int_roundingfn, fold_builtin_bitop,
3931 fold_builtin_bswap, real_constp, fold_builtin_pow,
3932 fold_builtin_powi, fold_builtin_signbit, fold_builtin_copysign,
3933 do_mpfr_arg1, do_mpfr_arg2, do_mpfr_arg3, do_mpfr_sincos): Replace
3934 uses of the macro TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
3935 * convert.c (convert_to_pointer): Likewise.
3936 * expr.c (highest_pow2_factor, expand_expr_real_1): Likewise.
3937 * fold-const.c (force_fit_type, fold_negate_expr, int_const_binop,
3938 const_binop, fold_convert_const_int_from_int,
3939 fold_convert_const_int_from_real,
3940 fold_convert_const_real_from_real, sign_bit_p,
3941 optimize_minmax_comparison, extract_muldiv_1, fold_div_compare,
3942 fold_sign_changed_comparison, fold_unary, fold_comparison,
3943 fold_binary, multiple_of_p, tree_Expr_non_zero_p,
3944 fold_negate_const, fold_abs_const, fold_not_const): Likewise.
3945 * print-tree.c (print_node_brief, print_node): Likewise.
3946 * stor-layout.c (place_field, layout_type): Likewise.
3947 * tree-chrec.c (keep_cast): Likewise.
3948 * tree.c (build_vector, build_real, build_real_from_int_cst,
3949 build_complex): Likewise.
3951 2007-01-11 Roger Sayle <roger@eyesopen.com>
3953 * tree.h (TREE_CONSTANT_OVERFLOW): Obsolete. For the time being,
3954 treat TREE_CONSTANT_OVERFLOW as a synonym of TREE_OVERFLOW.
3956 2007-01-11 Paolo Bonzini <bonzini@gnu.org>
3958 * configure.ac (strict1_warn): Rename to strict_warn.
3959 (WERROR, --enable-werror, symlink hacks, stage1_cflags,
3960 cc_set_by_configure, quoted_cc_set_by_configure,
3961 stage_prefix_set_by_configure, quoted_stage_prefix_set_by_configure,
3962 all_boot_languages, all_stagestuff): Remove.
3963 (target_list): Remove bootstrap targets.
3964 * Makefile.in (quickstrap): Unconditionally make a synonym of all.
3965 (BOOT_LANGUAGES, STAGE1_CFLAGS, STAGE1_CHECKING,
3966 REMAKEFLAGS, FLAGS_TO_PASS, PREPEND_DOTDOT_TO_RELATIVE_PATHS,
3967 SUBDIR_FLAGS_TO_PASS, WERROR_FLAGS, STRICT2_WARN, LANG_STAGESTUFF,
3968 VOL_FILES, POSTSTAGE1_FLAGS_TO_PASS, STAGE2_FLAGS_TO_PASS,
3969 STAGEPROFILE_FLAGS_TO_PASS, STAGEFEEDBACK_FLAGS_TO_PASS, stage1_build,
3970 stage1_copy, stage2_build, stage2_copy, stageprofile_build,
3971 stageprofile_copy, stage3_build, stage3_copy, stagefeedback_build,
3972 stagefeedback_copy, stage4_build, clean_s1, clean_sw, bootstrap,
3973 bootstrap-lean, bootstrap2, bootstrap2-lean, bootstrap3,
3974 bootstrap3-lean, bootstrap4, bootstrap4-lean, unstage1, unstage2,
3975 unstage3, unstage4, unstageprofile, unstagefeedback, restage, restage2,
3976 restage3, restage4, restageprofile, restagefeedback, bubbleestrap,
3977 cleanstrap, unstrap, restrap, *compare, *compare3, *compare4,
3978 *compare-lean, *compare3-lean, *compare4-lean, stage1-start, stage1,
3979 stage2-start, stage2, stage3-start, stage3, stage4-start, stage4,
3980 stageprofile-start, stageprofile, stagefeedback-start, stagefeedback,
3981 risky-stage1, risky-stage2, risky-stage3, risky-stage4): Remove.
3982 (ORDINARY_FLAGS_TO_PASS): Rename to FLAGS_TO_PASS.
3983 (STAGECOPYSTUFF, STAGEMOVESTUFF): Consolidate into MOSTLYCLEANFILES.
3984 (mostlyclean): Adjust.
3985 (clean, distclean): Don't mention bootstrap stuff.
3986 * configure: Regenerate.
3987 * ada/config-lang.in, cp/config-lang.in, forttran/config-lang.in,
3988 java/config-lang.in, objc/config-lang.in, objcp/config-lang.in,
3989 treelang/config-lang.in (stagestuff): Remove.
3990 * doc/sourcebuild.texi (stage1, stage2, stage3, stage4,
3991 stageprofile, stagefeedback, stagestuff): Remove mention.
3993 2007-01-11 Nick Clifton <nickc@redhat.com>
3995 * config/mcore/predicates.md (mcore_general_movesrc_operand):
3997 (mcore_general_movdst_operand): Do not accept CONST_INTs.
3998 (mcore_arith_K_S_operand): Run the test for the S constraint not
3999 the test for the M constraint.
4000 (mcore_addsub_operand): Do not accept integer values that are
4001 larger than 32 bits.
4002 * config/mcore/mcore.md: Remove unused constraints from split.
4003 (andsi3): Use HOST_WIDE_INT instead of int to hold an INTVAL.
4005 (allocate_stack): Likewise.
4006 * config/mcore/mcore.c (mcore_print_operand): Restrict output of P
4007 operands to 32 bits.
4008 (mcore_const_costs): Use HOST_WIDE_INT instead of int to hold an
4010 (mcore_and_cost, mcore_modify_comparison, const_ok_for_mcore,
4011 mcore_const_ok_for_inline, mcore_const_trick_uses_not,
4012 try_constant_tricks, mcore_num_ones, mcore_num_zeros,
4013 mcore_output_bclri, mcore_output_andn, output_inline_const,
4014 mcore_output_move, mcore_output_movedouble): Likewise.
4015 (mcore_output_cmov): Use CONST_OK_FOR_M and CONST_OK_FOR_N.
4016 (output_inline_const): Likewise.
4017 (output_inline_const): Fix format strings used in sprintf
4019 * config/mcore/mcore-protos.h: Update prototypes for changed
4020 functions in mcore.c.
4021 * config/mcore/mcore.h (CONST_OK_FOR_I): Cast values to
4022 HOST_WIDE_INT and not int.
4023 (CONST_OK_FOR_J, CONST_OK_FOR_K, CONST_OK_FOR_L, CONST_OK_FOR_M,
4024 CONST_OK_FOR_N): Likewise.
4025 (LEGITIMATE_CONSTANT_P): Also check CONSTANT_P.
4026 (GO_IF_LEGITIMATE_INDEX): Use HOST_WIDE_INT instead of int to hold
4029 2007-01-10 Jan Hubicka <jh@suse.cz>
4031 * tree-vrp.c (remove_range_assertions): Release defs.
4032 * tree-ssa-loop-ivopts.c (rmeove_statement): Likewise.
4033 * tree-ssa-dom.c (remove_stmt_or_phi): Likewise.
4035 2007-01-10 Paul Brook <paul@codesourcery.com>
4037 * config/arm/arm.c (arm_rtx_costs_1): Handle mutiply-subtract.
4038 * config/arm/arm.md (mulsi3subsi): New insn.
4040 2007-01-10 Zdenek Dvorak <dvorakz@suse.cz>
4042 * tree-ssa-loop-manip.c (tree_unroll_loop): Make it a wrapper over ...
4043 (tree_transform_and_unroll_loop): New.
4044 * tree-flow.h (transform_callback, tree_transform_and_unroll_loop):
4047 2007-01-10 Robert Kennedy <jimbob@google.com>
4049 * fold-const.c (fold_comparison): Fold comparisons like (x *
4050 1000 < 0) to (x < 0).
4052 2007-01-10 Ian Lance Taylor <iant@google.com>
4054 * tree-pretty-print.c (dump_generic_node): Print parentheses when
4055 operands have the same priority.
4057 2007-01-10 Tom Tromey <tromey@redhat.com>
4059 * fold-const.c (fold_truthop): Don't check can_use_bit_fields_p.
4060 (fold_binary): Likewise.
4061 * langhooks.c (lhd_can_use_bit_fields_p): Removed.
4062 * langhooks-def.h (lhd_can_use_bit_fields_p): Removed.
4063 (LANG_HOOKS_CAN_USE_BIT_FIELDS_P): Removed.
4064 (LANG_HOOKS_INITIALIZER): Remove LANG_HOOKS_CAN_USE_BIT_FIELDS_P.
4065 * langhooks.h (struct lang_hooks): Removed field
4066 'can_use_bit_fields_p'.
4068 2007-01-10 Ralf Corsépius <ralf.corsepius@rtems.org>
4070 * config/bfin/t-bfin, config/bfin/t-bfin-elf: Remove GCC_CFLAGS.
4072 2007-01-10 Razya Ladelsky <razya@il.ibm.com>
4074 * function.c (get_last_funcdef_no): New function.
4075 * function.h (get_last_funcdef_no): Declare.
4076 * tree-inline.c (initialize_cfun): Add initialization.
4077 (tree_function_versioning): Cleanup.
4079 2007-01-10 Jan Hubicka <jh@suse.cz>
4081 * tree-inline.c (setup_one_parameter): Do not propagate into abnormal
4084 2007-01-10 Sa Liu <saliu@de.ibm.com>
4085 Ben Elliston <bje@au.ibm.com>
4087 * spu.h (STACK_SAVE_AREA): Use VOIDmode for SAVE_FUNCTION, SImode
4088 for SAVE_NONLOCAL and Pmode for any other save level.
4089 * spu-protos.h (spu_restore_stack_block): Declare.
4090 * spu.md (save_stack_block): Remove.
4091 (restore_stack_block): Call spu_restore_stack_block.
4092 * spu.c (spu_restore_stack_block): New function.
4093 (spu_expand_epilogue): Remove old comment.
4095 2007-01-09 Zdenek Dvorak <dvorakz@suse.cz>
4097 PR tree-optimization/30322
4098 * tree-ssa-loop-ivopts.c (fold_affine_expr, iv_value): Removed.
4099 (cand_value_at): Return the value as aff_tree.
4100 (may_eliminate_iv): Convert the bound from aff_tree to tree.
4101 * tree-affine.c (aff_combination_add_cst, aff_combination_add_product,
4102 aff_combination_mult): New functions.
4103 (aff_combination_add): Use aff_combination_add_cst.
4104 (aff_combination_convert): Allow conversions to a wider type.
4105 (tree_to_aff_combination): Handle BIT_NOT_EXPR.
4106 * tree-affine.h (aff_combination_mult): Declare.
4108 2007-01-09 Carlos O'Donell <carlos@codesourcery.com>
4110 * doc/tm.texi: Update documentation to reflect reality of exec
4111 and start file search behaviours. Update copyright year.
4112 * doc/invoke.texi: Explain how GCC_EXEC_PREFIX is used to find
4113 header file directories.
4115 2007-01-09 Uros Bizjak <ubizjak@gmail.com>
4117 * config/i386/i386.md (*sinxf2): Rename to *sinxf2_i387.
4118 (*cosxf2): Rename to cosxf2_i387.
4119 (*sindf2, *sinsf2): Extend operand 1 to XFmode. Macroize patterns
4120 using X87MODEF12 mode macro. Rename patterns to
4121 *sin_extend<mode>xf2_i387. Use SSE_FLOAT_MODE_P to disable patterns
4123 (*cosdf2, *cossf2): Ditto.
4124 (sincosdf3, sincossf3): Ditto. Rewrite corresponding splitters
4125 to match extended input operands.
4126 (sincos<mode>3): New expander.
4127 (*sinextendsfdf2, *cosextendsfdf2, *sincosextendsfdf3): Remove
4128 insn patterns and corresponding splitters.
4130 2007-01-09 Kaz Kojima <kkojima@gcc.gnu.org>
4132 * config/sh/t-linux (TARGET_LIBGCC2_CFLAGS): Delete.
4133 (SHLIB_MAPFILES, SHLIB_LINK, SHLIB_INSTALL): Likewise.
4135 2007-01-09 Nicolas Pitre <nico@cam.org>
4138 * arm/ieee754-df.S (Lad_s): Also test the low word of X for zero.
4140 2007-01-08 Geoffrey Keating <geoffk@apple.com>
4142 * target.h (struct gcc_target): New field library_rtti_comdat.
4143 * target-def.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): New.
4144 (TARGET_CXX): Add TARGET_CXX_LIBRARY_RTTI_COMDAT.
4145 * doc/tm.texi (C++ ABI): Document TARGET_CXX_LIBRARY_RTTI_COMDAT.
4146 * config/darwin.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
4148 2007-01-08 Geoffrey Keating <geoffk@apple.com>
4150 * doc/invoke.texi (Optimize Options): Correct description of -O0.
4152 2007-01-08 Richard Guenther <rguenther@suse.de>
4154 * tree.h (force_fit_type_double): Export.
4155 (force_fit_type): Remove.
4156 * fold-const.c (force_fit_type_double): New function.
4157 (force_fit_type): Remove.
4158 (int_const_binop): Use it.
4159 (fold_convert_const_int_from_int): Likewise.
4160 (fold_convert_const_int_from_real): Likewise.
4161 (fold_div_compare): Likewise.
4162 (fold_sign_changed_comparison): Likewise.
4163 (fold_unary): Likewise.
4164 (fold_negate_const): Likewise.
4165 (fold_abs_const): Likewise.
4166 (fold_not_const): Likewise.
4167 * c-common.c (shorten_compare): Use force_fit_type_double.
4168 * convert.c (convert_to_pointer): Likewise.
4170 2007-01-08 Richard Guenther <rguenther@suse.de>
4172 * tree.h (build_int_cst_wide_type): Export.
4173 * tree.c (build_int_cst_wide_type): New function.
4174 (build_int_cst_wide): Fix comment.
4175 * builtins.c (fold_builtin_object_size): Use build_int_cst
4176 to build -1 or 0 of the correct type. Use fit_double_type
4177 to check for overflow.
4178 * fold-const.c (optimize_bit_field_compare): Use build_int_cst_type
4180 (decode_field_reference): Likewise.
4181 (all_ones_mask_p): Likewise.
4182 (native_interpret_int): Use build_int_cst_wide_type.
4183 (fold_binary): Use build_int_cst_type to build an all-ones
4185 * stor-layout.c (set_sizetype): Use build_int_cst_wide_type.
4187 2007-01-08 Daniel Jacobowitz <dan@codesourcery.com>
4189 * config/pa/t-pa64 (libgcc_stub.a): Use $(T).
4191 2007-01-09 Ben Elliston <bje@au.ibm.com>
4193 * genautomata.c (STATS_OPTION): New option.
4194 (stats_flag): New flag.
4195 (gen_automata_option): Handle it.
4196 (initiate_automaton_gen): Ditto.
4197 (write_automata): Output statistics only if stats_flag is
4198 set. Likewise, output time statistics only if time_flag is set.
4199 * doc/md.texi (Processor pipeline description): Document new flag.
4201 2007-01-08 Richard Guenther <rguenther@suse.de>
4203 * builtins.c (fold_builtin_int_roundingfn): Use fit_double_type.
4204 * tree.c (build_int_cst_type): Likewise.
4205 (size_in_bytes): Don't call force_fit_type on the result.
4206 (int_fits_type_p): Use fit_double_type.
4207 * fold-const.c (fit_double_type): New function.
4208 (force_fit_type): Use it.
4209 * tree.h (fit_double_type): Export.
4211 2007-01-08 Jan Hubicka <jh@suse.cz>
4213 * tree-vectorizer.c (gate_increase_alignment): Fix return type.
4214 * ipa.c (function_and_variable_visibility): Fix return type.
4216 2007-01-08 Richard Guenther <rguenther@suse.de>
4218 * tree-ssa-ccp.c (maybe_fold_offset_to_array_ref): Use type
4219 of offset to build the index.
4220 * tree-pretty-print.c (dump_generic_node): Don't build negated
4221 const just for printing.
4222 * c-pretty-print.c (pp_c_integer_constant): Likewise.
4223 * builtins.c (fold_builtin_int_roundingfn): Check if result
4224 fits the type by using force_fit_type and comparing the result.
4225 * predict.c (predict_loops): Use compare_tree_int for comparison.
4226 * tree.c (build_int_cst): Fall back to integer_type_node for
4228 (build_int_cst_wide): Assert type is non-null.
4230 2007-01-08 Roberto Costa <roberto.costa@st.com>
4232 * tree-vrp.c (extract_range_from_cond_expr): New.
4233 (extract_range_from_expr): Handle COND_EXPR nodes used as expressions.
4234 * tree-ssa-ccp.c (get_maxval_strlen): Handle COND_EXPR nodes used
4236 (fold_stmt): Bug fix, avoid infinite recursion when folding COND_EXPRs.
4237 * tree-ssa-forwprop.c (simplify_cond, forward_propagate_into_cond,
4238 tree_ssa_forward_propagate_single_use_vars): Handle COND_EXPR nodes
4239 used as expressions.
4240 * tree-object-size.c (cond_expr_object_size): New.
4241 (collect_object_sizes_for): Handle COND_EXPR nodes used as expressions.
4243 2007-01-08 Jan Hubicka <jh@suse.cz>
4245 * tree-ssa-forwprop.c (forward_propagate_into_cond,
4246 tree_ssa_forward_propagate_single_use_va): Release defs of propagated
4249 2007-01-08 Richard Guenther <rguenther@suse.de>
4251 PR tree-optimization/23603
4252 * tree-vrp.c (set_value_range_to_truthvalue): New function.
4253 (extract_range_from_binary): Fall back to truthvalue instead of
4254 varying for TRUTH_*_EXPR.
4255 (extract_range_from_comparison): Fall back to truthvalue instead of
4257 (vrp_visit_phi_node): Don't adjust new range bounds to +INF/-INF
4258 if all visited PHI values were constant.
4260 2007-01-08 Jan Hubicka <jh@suse.cz>
4262 * cgraphunit.c (cgraph_process_new_functions): Reset reachable flag.
4263 (cgraph_analyze_function): break out from ...
4264 (cgraph_finalize_compilation_unit): ... here.
4265 (cgraph_expand_function): Remove forgoten commented out line.
4266 (cgraph_optimize): Analyze functions.
4268 2007-01-08 Jan Hubicka <jh@suse.cz>
4270 * tree-pas.h (TODO_remove_function): New flag.
4271 (TODO_update*): Renumber.
4272 (pass_ipa_increase_alignment,
4273 pass_ipa_function_and_variable_visibility): New passes.
4274 * cgraphunit.c (cgraph_increase_alignment): Move to tree-vectorizer.c
4275 (cgraph_function_and_variable_visibility): Move to ipa.c
4276 (cgraph_optimize): Don't call cgraph_function_and_variable_visibility,
4277 cgraph_increase_alignment.
4278 * ipa-inline.c (cgraph_decide_inlining): Don't push timevar.
4279 (cgraph_decide_inlining_incrementally): Push TV_INTEGRATION before
4280 calling tree-inline.
4281 (cgraph_early_inlining): Do not call cgraph_remove_unreachable_nodes.
4282 (pass_ipa_inline, pass_early_ipa_inlining): Set TODO_remove_functions
4283 * tree-vectorizer.c (increase_alignment): Move here from cgraphunit.c
4284 (gate_increase_alignment): New function.
4285 (pass_ipa_increase_alignment): New pass.
4286 * ipa.c: Inline tree-pass.h and timevar.h
4287 (function_and_variable_visibility): Move here from cgraphunit.c
4288 * tree-optimize.c (pass_early_local_passes): Add TODO_remove_functions.
4289 * passes.c (init_optimization_passes): Add the two new passes.
4290 (execute_todo): Handle cgraph_remove_functions.
4292 2007-01-08 Nick Clifton <nickc@redhat.com>
4294 * config/frv/predicates.md (reg_or_0_operand): Accept
4297 2007-01-08 Ralf Corsépius <ralf.corsepius@rtems.org>
4299 * config/bfin/rtems.h, config/bfin/t-rtems: New.
4300 * config.gcc: Add bfin*-rtems*.
4302 2007-01-08 Mark Shinwell <shinwell@codesourcery.com>
4304 * c.opt: Add -flax-vector-conversions.
4305 * c-typeck.c (convert_for_assignment): Pass flag to
4306 vector_types_convertible_p to allow emission of note.
4307 (digest_init): Likewise.
4308 * c-opts.c: Handle -flax-vector-conversions.
4309 * c-common.c (flag_lax_vector_conversions): New.
4310 (vector_types_convertible_p): Unless -flax-vector conversions
4311 has been passed, disallow conversions between vectors with
4312 differing numbers of subparts and/or element types. If such
4313 a conversion is disallowed, possibly emit a note on the first
4314 occasion only to inform the user of -flax-vector-conversions.
4315 The new last argument specifies this.
4316 * c-common.h (flag_lax_vector_conversions): New.
4317 (vector_types_convertible_p): Add extra argument.
4318 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Use
4319 char_type_node for V*QI type vectors.
4320 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins):
4321 Update to satisfy new typechecking rules.
4322 * config/rs6000/altivec.h (vec_cmple): Use vec_cmpge, for both
4324 * doc/invoke.texi (C Dialect Options): Document
4325 -flax-vector-conversions.
4327 2007-01-08 Mark Shinwell <shinwell@codesourcery.com>
4329 PR tree-optimization/29877
4330 * tree-ssa-ter.c (is_replaceable_p): Deem assignments with
4331 a register variable on the RHS to not be replaceable.
4333 2007-01-08 Chen Liqin <liqin@sunnorth.com.cn>
4334 * config/score/t-score-elf (MULTILIB_OPTIONS): Change.
4335 * config/score/predicates.md (const_uimm5, sr0_operand, const_simm12,
4336 const_simm15, const_pow2, const_npow2): Added.
4337 * config/score/misc.md (insv, extv, extzv, movmemsi,
4338 move_lbu_a/b, mov_lhu_a/b etc): Added and fix some bug.
4339 * config/score/score.c (score_address_cost, score_select_cc_mode):
4341 Change CONST_OK_FOR_LETTER_P/EXTRA_CONSTRAINT define.
4342 Update score_rtx_costs for MACRO TARGET_RTX_COSTS.
4343 Update score_print_operand.
4344 * config/score/score.h (DATA_ALIGNMENT, SELECT_CC_MODE): Added.
4345 Adjust register allocate order and update some macro define.
4346 * config/score/score-mdaux.c (mdx_unaligned_load, mdx_unsigned_store,
4347 mdx_block_move_straight, mdx_block_move_loop_head,
4348 mdx_block_move_loop_body, mdx_block_move_loop_foot, mdx_block_move_loop,
4349 mdx_block_move): Added.
4350 (mdx_movsicc, mdp_select_add_imm, mdp_select, mds_zero_extract_andi,
4351 mdp_limm): Updated and fix some bug and typo.
4352 * config/score/score.md (movqi/hi/si, add/sub/zero/ext): Updated.
4353 (movsf, movdf, doloop_end): Added.
4355 2007-01-08 Kazu Hirata <kazu@codesourcery.com>
4357 * config/arm/arm.c, config/arm/arm.h, config/arm/arm.md,
4358 config/arm/thumb2.md: Fix comment typos.
4359 * doc/extend.texi: Fix a typo.
4361 2007-01-07 Eric Christopher <echristo@apple.com>
4363 * configure.ac: Check for __stack_chk_fail for darwin.
4364 * configure: Regenerate.
4366 2007-01-07 Richard Guenther <rguenther@suse.de>
4368 * tree-vrp.c (extract_range_from_assert): CSE calls to
4369 compare_values where possible.
4370 (extract_range_from_unary_expr): Likewise.
4372 2007-01-07 Anatoly Sokolov <aesok@post.ru>
4374 * config/avr/avr-protos.h (call_insn_operand): Delete prototype.
4375 * config/avr/avr.c (call_insn_operand): Delete function.
4376 * config/avr/avr.md (*pushqi, *pushhi, *pushsi, *pushsf): Use REG_SP
4377 instead of register number. Use predicates.
4378 * config/avr/predicates.md (const0_operand, reg_or_0_operand,
4379 call_insn_operand): Add.
4381 2007-01-06 Jan Hubicka <jh@suse.cz>
4383 * tree-pass.h (pass_build_cgraph_edges): Declare.
4384 * cgraphunit.c (record_refernece): Move to cgraphbuild.c
4385 (visited_nodes): Remove.
4386 (cgraph_create_edges): Move to cgraphbuild.c; rename to
4387 build_cgrpah_edges; make visited_nodes local.
4388 (cgraph_process_new_functions): DO not call initialize_inline_failed.
4389 (record_references_in_initializer): Move to cgraphbuild.c
4390 (initialize_inline_failed, rebuild_cgraph_edges,
4391 pass_rebuild_cgraph_edges): Move to cgraphbuild.c.
4392 (verify_cgraph_node): Make visited_nodes local.
4393 (cgraph_analyze_function): Do not call cgraph_create_edges and
4394 initialize_inline_failed.
4395 (cgraph_expand_function): Do not call cgraph_lower_function;
4396 assert that function is already lowered.
4397 * Makefile.in (cgraphbuild.o): New.
4398 * passes.c (init_optimization_passes): Add pass_build_cgraph_edges
4399 at the end of lowering passes.
4401 2007-01-06 Steven Bosscher <steven@gcc.gnu.org>
4403 * ifcvt.c (cond_move_convert_if_block): New function, code
4404 factored out from...
4405 (cond_move_process_if_block): ...here. Call the new function
4406 on the THEN and ELSE blocks.
4407 (merge_if_block): Do not copy global_live_at_end, merge_blocks
4408 already takes care of this.
4410 2007-01-05 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
4413 * tree.h (TREE_OVERFLOW_P): New.
4414 * c-typeck.c (parser_build_unary_op): Warn only if result
4415 overflowed and operands did not.
4416 (parser_build_binary_op): Likewise.
4417 (convert_for_assignment): Remove redundant overflow_warning.
4418 * c-common.c (overflow_warning): Don't check or set TREE_OVERFLOW.
4420 2007-01-05 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
4422 * c-typeck.c (store_init_value): Split over two lines to follow
4423 the GNU coding style.
4425 2007-01-05 Benjamin Kosnik <bkoz@redhat.com>
4427 * c-cppbuiltin.c (c_cpp_builtins): __GXX_EXPERIMENTAL_CPP0X__ to
4428 __GXX_EXPERIMENTAL_CXX0X__.
4429 * doc/cpp.texi: Same.
4431 2007-01-05 Richard Guenther <rguenther@suse.de>
4434 * tree.c (get_narrower): Do not construct COMPONENT_REFs
4435 with mismatched types. Instead explicitly build a
4436 conversion NOP_EXPR.
4438 2007-01-05 Ian Lance Taylor <iant@google.com>
4440 * c-common.c (decl_with_nonnull_addr_p): New function.
4441 (c_common_truthvalue_conversion): Call it.
4442 * c-typeck.c (build_binary_op): Likewise.
4443 * c-common.h (decl_with_nonnull_addr_p): Declare.
4445 2007-01-05 Jakub Jelinek <jakub@redhat.com>
4448 * libgcc2.c (__divdc3): Compare c and d against 0.0 instead of
4451 2007-01-05 Joel Brobecker <brobecker@adacore.com>
4453 * doc/install.texi (Final install): Document the fact that
4454 the GNAT runtime should not be stripped.
4456 2007-01-04 Jan Hubicka <jh@suse.cz>
4458 * tree-inline.c (fold_marked_statements): Update operand caches
4459 and EH after folding
4461 2007-01-04 Ian Lance Taylor <iant@google.com>
4463 * c-common.c (check_function_nonnull): Whitespace fix.
4465 2007-01-04 Jan Hubicka <jh@suse.cz>
4467 * tree-optimize.c (execute_fixup_cfg): Correct previously mistakely
4468 comitted older version of patch.
4469 (pass_fixup_cfg): Add TODOs to verify flow and statements, dump
4470 function, celanup cfg and collect garbage.
4472 2007-01-04 Mike Stump <mrs@apple.com>
4474 * Makefile.in (mostlyclean): Don't remove libgcc anymore.
4477 2007-01-04 Eric Christopher <echristo@apple.com>
4479 * libgcc2.c (__bswapsi2): Use SItype.
4480 (__bswapdi2): Use DItype.
4481 * libgcc2.h: Update for above.
4483 2007-01-04 Paul Brook <paul@codesourcery.com>
4485 * config/arm/arm.md (arm_mulsi3, thumb_mulsi3, mulsi3_compare0,
4486 mulsi_compare0_scratch, mulsi3addsi, mulsi3addsi_compare0,
4487 mulsi3addsi_compare0_scratch, mulsidi3adddi, mulsidi3,
4488 umulsidi3, umulsidi3adddi, smulsi3_highpart,
4489 umulsi3_highpart): Make conditional on !arm_arch6.
4490 (arm_mulsi3_v6, thumb_mulsi3_v6, mulsi3_compare0_v6,
4491 mulsi_compare0_scratch_v6, mulsi3addsi_v6, mulsi3addsi_compare0_v6,
4492 mulsi3addsi_compare0_scratch_v6, mulsidi3adddi_v6, mulsidi3_v6,
4493 umulsidi3_v6, umulsidi3adddi_v6, smulsi3_highpart_v6,
4494 umulsi3_highpart_v6): New insns.
4496 2007-01-04 Roger Sayle <roger@eyesopen.com>
4498 * fold-const.c (fold_convert): When casting an expression to void,
4499 fold_ignored_result may discover a GIMPLE_MODIFY_STMT which doesn't
4500 have a type. Instead of attempting to build a NOP_EXPR, return
4501 these "special" trees directly.
4503 2007-01-04 Joseph Myers <joseph@codesourcery.com>
4505 * config/rs6000/rs6000.c (rs6000_rtx_costs): Make adjustment for
4506 MULT inside MINUS as either argument. Use rs6000_cost->dmul -
4507 rs6000_cost->fp not 0 as adjustment for outer NEG.
4509 2007-01-04 Jan Hubicka <jh@suse.cz>
4511 * cgraph.c (cgraph_release_function_body): New function.
4512 (cgraph_remove_node): Use it.
4513 * cgraph.h (cgraph_release_function_body): Declare.
4514 * cgraphunit.c (cgraph_expand_function): Use it.
4515 * ipa.c (cgraph_remove_unreahchable_nodes): Use it.
4516 * tree-ssa.c (delete_tree_ssa): Allow to be called before aliasing
4517 is initialized and while compilation of other function is running.
4518 * tree-optimize.c (execute_free_cfg_annotations): Move code to clear
4519 statement CFG annotations from here to ...
4520 * tree-cfg.c (delete_tree_cfg_annotations): ... here.
4522 2007-01-04 Zdenek Dvorak <dvorakz@suse.cz>
4524 * cfgloop.h (enum li_flags): Make the constants powers of two.
4526 2007-01-04 Jan Hubicka <jh@suse.cz>
4528 * tree-inline.c (copy_bb): Insert new statements to statements_to_fold
4530 (fold_marked_statements): New function.
4531 (optimize_inline_calls, tree_function_versioning): Fold new statements.
4532 * tree-inline.h (copy_body_data): Add statements_to_fold.
4534 2007-01-03 Daniel Jacobowitz <dan@codesourcery.com>
4536 * config.gcc: Mention libgcc/config.host.
4537 * Makefile.in: Update comments mentioning libgcc.
4538 (LIBGCC, INSTALL_LIBGCC, GCC_PARTS, mklibgcc): Delete.
4539 (all.cross, start.encap, rest.encap, rest.cross): Update
4540 dependencies for libgcc move.
4541 (libgcc.mk, LIBGCC_DEPS, libgcov.a, libgcc.a, stmp-multilib)
4542 (clean-target, clean-target-libgcc): Delete.
4543 (srcdirify, GCC_EXTRA_PARTS): New macros.
4544 (libgcc-support, libgcc.mvars): New rules.
4545 (distclean): Remove mention of mklibgcc.
4546 (install): Don't reference INSTALL_LIBGCC.
4547 (install-common): Don't reference EXTRA_PARTS.
4548 (install-libgcc, install-multilib): Delete rules.
4549 * mklibgcc.in: Delete file.
4550 * doc/configfiles.texi: Don't mention mklibgcc.
4552 * config/i386/t-darwin (SHLIB_VERPFX): Delete (moved to libgcc).
4553 * config/i386/t-darwin64 (SHLIB_VERPFX): Likewise.
4554 * config/rs6000/t-darwin (SHLIB_VERPFX): Likewise.
4555 * config/rs6000/t-ppccomm (TARGET_LIBGCC2_CFLAGS, SHLIB_MAPFILES)
4556 (mklibgcc, ldblspecs): Likewise.
4558 * config/i386/t-nwld (libgcc.def, libc.def, libpcre.def)
4559 (posixpre.def): Use $(T).
4560 (SHLIB_EXT, SHLIB_NAME, SHLIB_SLIBDIR_QUAL, SHLIB_DEF, SHLIB_MAP)
4561 (SHLIB_SRC, SHLIB_INSTALL): Delete.
4562 (SHLIB_LINK): Make dummy.
4563 * config/t-slibgcc-darwin: Delete contents except for dummy SHLIB_LINK.
4565 * config/frv/t-linux (EXTRA_MULTILIB_PARTS): Clear.
4567 * config/alpha/t-crtfm: Use $(T) in rules for EXTRA_PARTS.
4568 * config/alpha/t-vms, config/alpha/t-vms64, config/fr30/t-fr30,
4569 config/i386/t-rtems-i386, config/ia64/t-ia64, config/rs6000/t-beos,
4570 config/rs6000/t-newas, config/sparc/t-elf: Likewise.
4572 * configure.ac (all_outputs): Remove mklibgcc.
4573 * configure: Regenerated.
4575 2007-01-03 Josh Conner <jconner@apple.com>
4578 * calls.c (compute_argument_addresses): Set stack and stack_slot
4579 for partial args, too.
4580 (store_one_arg): Use locate.size.constant for the size when
4581 generating a save_area.
4583 2007-01-03 Robert Kennedy <jimbob@google.com>
4585 * tree-cfg.c (tree_merge_blocks): Release SSA_NAME phi results
4586 whose definitions are deleted due to basic block merging.
4588 2007-01-03 Paul Brook <paul@codesourcery.com>
4591 * config/arm/arm.c (output_return_instruction): Pop PC in interrupt
4593 (use_return_insn): Return 0 for Thumb interrupt functions.
4594 (print_multi_reg): Add rfe argument for IRQ returns.
4595 (arm_output_epilogue): Pop interrupt return address directly into PC.
4596 (arm_expand_prologue): Only adjust IRQ return address in Arm mode.
4598 2007-01-03 Paul Brook <paul@codesourcery.com>
4600 Merge from sourcerygxx-4_1.
4601 * config/arm/thumb2.md: New file.
4602 * config/arm/elf.h (JUMP_TABLES_IN_TEXT_SECTION): Return True for
4604 * config/arm/coff.h (JUMP_TABLES_IN_TEXT_SECTION): Ditto.
4605 * config/arm/aout.h (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
4606 (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump tables.
4607 * config/arm/aof.h (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump
4609 (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
4610 * config/arm/ieee754-df.S: Use macros for Thumb-2/Unified asm
4612 * config/arm/ieee754-sf.S: Ditto.
4613 * config/arm/arm.c (thumb_base_register_rtx_p): Rename...
4614 (thumb1_base_register_rtx_p): ... to this.
4615 (thumb_index_register_rtx_p): Rename...
4616 (thumb1_index_register_rtx_p): ... to this.
4617 (thumb_output_function_prologue): Rename...
4618 (thumb1_output_function_prologue): ... to this.
4619 (thumb_legitimate_address_p): Rename...
4620 (thumb1_legitimate_address_p): ... to this.
4621 (thumb_rtx_costs): Rename...
4622 (thumb1_rtx_costs): ... to this.
4623 (thumb_compute_save_reg_mask): Rename...
4624 (thumb1_compute_save_reg_mask): ... to this.
4625 (thumb_final_prescan_insn): Rename...
4626 (thumb1_final_prescan_insn): ... to this.
4627 (thumb_expand_epilogue): Rename...
4628 (thumb1_expand_epilogue): ... to this.
4629 (arm_unwind_emit_stm): Rename...
4630 (arm_unwind_emit_sequence): ... to this.
4631 (thumb2_legitimate_index_p, thumb2_legitimate_address_p,
4632 thumb1_compute_save_reg_mask, arm_dwarf_handle_frame_unspec,
4633 thumb2_index_mul_operand, output_move_vfp, arm_shift_nmem,
4634 arm_save_coproc_regs, thumb_set_frame_pointer, arm_print_condition,
4635 thumb2_final_prescan_insn, thumb2_asm_output_opcode, arm_output_shift,
4636 thumb2_output_casesi): New functions.
4637 (TARGET_DWARF_HANDLE_FRAME_UNSPEC): Define.
4638 (FL_THUMB2, FL_NOTM, FL_DIV, FL_FOR_ARCH6T2, FL_FOR_ARCH7,
4639 FL_FOR_ARCH7A, FL_FOR_ARCH7R, FL_FOR_ARCH7M, ARM_LSL_NAME,
4640 THUMB2_WORK_REGS): Define.
4641 (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv, arm_condexec_count,
4642 arm_condexec_mask, arm_condexec_masklen)): New variables.
4643 (all_architectures): Add armv6t2, armv7, armv7a, armv7r and armv7m.
4644 (arm_override_options): Check new CPU capabilities.
4645 Set new architecture flag variables.
4646 (arm_isr_value): Handle v7m interrupt functions.
4647 (user_return_insn): Return 0 for v7m interrupt functions. Handle
4649 (const_ok_for_arm): Handle Thumb-2 constants.
4650 (arm_gen_constant): Ditto. Use movw when available.
4651 (arm_function_ok_for_sibcall): Return false for v7m interrupt
4653 (legitimize_pic_address, arm_call_tls_get_addr): Handle Thumb-2.
4654 (thumb_find_work_register, arm_load_pic_register,
4655 legitimize_tls_address, arm_address_cost, load_multiple_sequence,
4656 emit_ldm_seq, emit_stm_seq, arm_select_cc_mode, get_jump_table_size,
4657 print_multi_reg, output_mov_long_double_fpa_from_arm,
4658 output_mov_long_double_arm_from_fpa, output_mov_double_fpa_from_arm,
4659 output_mov_double_fpa_from_arm, output_move_double,
4660 arm_compute_save_reg_mask, arm_compute_save_reg0_reg12_mask,
4661 output_return_instruction, arm_output_function_prologue,
4662 arm_output_epilogue, arm_get_frame_offsets, arm_regno_class,
4663 arm_output_mi_thunk, thumb_set_return_address): Ditto.
4664 (arm_expand_prologue): Handle Thumb-2. Use arm_save_coproc_regs.
4665 (arm_coproc_mem_operand): Allow POST_INC/PRE_DEC.
4666 (arithmetic_instr, shift_op): Use arm_shift_nmem.
4667 (arm_print_operand): Use arm_print_condition. Handle '(', ')', '.',
4669 (arm_final_prescan_insn): Use extract_constrain_insn_cached.
4670 (thumb_expand_prologue): Use thumb_set_frame_pointer.
4671 (arm_file_start): Output directive for unified syntax.
4672 (arm_unwind_emit_set): Handle stack alignment instruction.
4673 * config/arm/lib1funcs.asm: Remove default for __ARM_ARCH__.
4674 Add v6t2, v7, v7a, v7r and v7m.
4675 (RETLDM): Add Thumb-2 code.
4676 (do_it, shift1, do_push, do_pop, COND, THUMB_SYNTAX): New macros.
4677 * config/arm/arm.h (TARGET_CPU_CPP_BUILTINS): Define __thumb2__.
4678 (TARGET_THUMB1, TARGET_32BIT, TARGET_THUMB2, TARGET_DSP_MULTIPLY,
4679 TARGET_INT_SIMD, TARGET_UNIFIED_ASM, ARM_FT_STACKALIGN, IS_STACKALIGN,
4680 THUMB2_TRAMPOLINE_TEMPLATE, TRAMPOLINE_ADJUST_ADDRESS,
4681 ASM_OUTPUT_OPCODE, THUMB2_GO_IF_LEGITIMATE_ADDRESS,
4682 THUMB2_LEGITIMIZE_ADDRESS, CASE_VECTOR_PC_RELATIVE,
4683 CASE_VECTOR_SHORTEN_MODE, ADDR_VEC_ALIGN, ASM_OUTPUT_CASE_END,
4684 ADJUST_INSN_LENGTH): Define.
4685 (TARGET_REALLY_IWMMXT, TARGET_IWMMXT_ABI, CONDITIONAL_REGISTER_USAGE,
4686 STATIC_CHAIN_REGNUM, HARD_REGNO_NREGS, INDEX_REG_CLASS,
4687 BASE_REG_CLASS, MODE_BASE_REG_CLASS, SMALL_REGISTER_CLASSES,
4688 PREFERRED_RELOAD_CLASS, SECONDARY_OUTPUT_RELOAD_CLASS,
4689 SECONDARY_INPUT_RELOAD_CLASS, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P,
4690 TRAMPOLINE_SIZE, INITIALIZE_TRAMPOLINE, HAVE_PRE_INCREMENT,
4691 HAVE_POST_DECREMENT, HAVE_PRE_DECREMENT, HAVE_PRE_MODIFY_DISP,
4692 HAVE_POST_MODIFY_DISP, HAVE_PRE_MODIFY_REG, HAVE_POST_MODIFY_REG,
4693 REGNO_MODE_OK_FOR_BASE_P, LEGITIMATE_CONSTANT_P,
4694 REG_MODE_OK_FOR_BASE_P, REG_OK_FOR_INDEX_P, GO_IF_LEGITIMATE_ADDRESS,
4695 LEGITIMIZE_ADDRESS, THUMB2_LEGITIMIZE_ADDRESS,
4696 GO_IF_MODE_DEPENDENT_ADDRESS, MEMORY_MOVE_COST, BRANCH_COST,
4697 ASM_APP_OFF, ASM_OUTPUT_CASE_LABEL, ARM_DECLARE_FUNCTION_NAME,
4698 FINAL_PRESCAN_INSN, PRINT_OPERAND_PUNCT_VALID_P,
4699 PRINT_OPERAND_ADDRESS): Adjust for Thumb-2.
4700 (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv): New declarations.
4701 * config/arm/arm-cores.def: Add arm1156t2-s, cortex-a8, cortex-r4 and
4703 * config/arm/arm-tune.md: Regenerate.
4704 * config/arm/arm-protos.h: Update prototypes.
4705 * config/arm/vfp.md: Enable patterns for Thumb-2.
4706 (arm_movsi_vfp): Add movw alternative. Use output_move_vfp.
4707 (arm_movdi_vfp, movsf_vfp, movdf_vfp): Use output_move_vfp.
4708 (thumb2_movsi_vfp, thumb2_movdi_vfp, thumb2_movsf_vfp,
4709 thumb2_movdf_vfp, thumb2_movsfcc_vfp, thumb2_movdfcc_vfp): New.
4710 * config/arm/libunwind.S: Add Thumb-2 code.
4711 * config/arm/constraints.md: Update include Thumb-2.
4712 * config/arm/ieee754-sf.S: Add Thumb-2/Unified asm support.
4713 * config/arm/ieee754-df.S: Ditto.
4714 * config/arm/bpabi.S: Ditto.
4715 * config/arm/t-arm (MD_INCLUDES): Add thumb2.md.
4716 * config/arm/predicates.md (low_register_operand,
4717 low_reg_or_int_operand, thumb_16bit_operator): New.
4718 (thumb_cmp_operand, thumb_cmpneg_operand): Rename...
4719 (thumb1_cmp_operand, thumb1_cmpneg_operand): ... to this.
4720 * config/arm/t-arm-elf: Add armv7 multilib.
4721 * config/arm/arm.md: Update patterns for Thumb-2 and Unified asm.
4723 (UNSPEC_STACK_ALIGN, ce_count): New.
4724 (arm_incscc, arm_decscc, arm_umaxsi3, arm_uminsi3,
4725 arm_zero_extendsidi2, arm_zero_extendqidi2): New
4727 * config/arm/fpa.md: Update patterns for Thumb-2 and Unified asm.
4728 (thumb2_movsf_fpa, thumb2_movdf_fpa, thumb2_movxf_fpa,
4729 thumb2_movsfcc_fpa, thumb2_movdfcc_fpa): New insns.
4730 * config/arm/cirrus.md: Update patterns for Thumb-2 and Unified asm.
4731 (cirrus_thumb2_movdi, cirrus_thumb2_movsi_insn,
4732 thumb2_cirrus_movsf_hard_insn, thumb2_cirrus_movdf_hard_insn): New
4734 * doc/extend.texi: Document ARMv7-M interrupt functions.
4735 * doc/invoke.texi: Document Thumb-2 new cores+architectures.
4737 2007-01-03 Jakub Jelinek <jakub@redhat.com>
4739 * unwind-dw2.c (SIGNAL_FRAME_BIT, EXTENDED_CONTEXT_BIT): Define.
4740 (struct _Unwind_Context): Rename args_size to flags, remove
4741 signal_frame field, add a new args_size field and version field.
4742 (_Unwind_IsSignalFrame, _Unwind_SetSignalFrame,
4743 _Unwind_IsExtendedContext): New inline functions.
4744 (_Unwind_GetGR, _Unwind_SetGR, _Unwind_GetGRPtr, _Unwind_SetGRPtr):
4745 Assume by_value array is only present if _Unwind_IsExtendedContext.
4746 (_Unwind_GetIPInfo, execute_cfa_program, uw_frame_state_for): Use
4747 _Unwind_IsSignalFrame.
4748 (__frame_state_for): Initialize context.flags to EXTENDED_CONTEXT_BIT.
4749 (uw_update_context_1): Use _Unwind_SetSignalFrame.
4750 (uw_init_context_1): Initialize context->flags to
4751 EXTENDED_CONTEXT_BIT.
4752 * config/rs6000/linux-unwind.h (frob_update_context): Use
4753 _Unwind_SetSignalFrame.
4755 2007-01-03 Andrew Pinski <andrew_pinski@playstation.sony.com>
4758 * gimplify.c (gimplify_modify_expr_complex_part): Move below
4759 tree_to_gimple_tuple. Call tree_to_gimple_tuple when we need
4762 2007-01-03 Kazu Hirata <kazu@codesourcery.com>
4764 * config/i386/i386.h (NON_STACK_REG_P, REGNO_OK_FOR_SIREG_P,
4765 REGNO_OK_FOR_DIREG_P, REWRITE_ADDRESS, ASM_OPERAND_LETTER,
4766 RET, AT_SP): Remove.
4767 * config/i386/i386.md (*sse_prologue_save_insn): Use return
4770 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
4771 * cfgbuild.c (find_basic_blocks): Likewise.
4772 * cfgrtl.c (rtl_create_basic_block): Likewise.
4773 * function.c (temp_slots_at_level): Likewise.
4774 * reg-stack.c (stack_regs_mentioned): Likewise.
4775 * regclass.c (allocate_reg_info): Likewise.
4776 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
4777 set_bb_for_stmt, move_block_to_fn): Likewise.
4778 * tree-complex.c (tree_lower_complex): Likewise.
4779 * vec.h (VEC_safe_grow_cleared): New.
4781 * cgraphunit.c, tree-ssa-alias.c: Fix comment typos.
4783 2007-01-03 Zdenek Dvorak <dvorakz@suse.cz>
4785 * loop-unswitch.c (unswitch_loop): Pass probabilities to loopify.
4786 * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Pass probabilities
4788 * cfgloopmanip.c (scale_loop_frequencies): Export.
4789 (loopify): Scale the frequencies by prescribed coefficients.
4790 (set_zero_probability): New function.
4791 (duplicate_loop_to_header_edge): Improve updating of frequencies.
4792 (lv_adjust_loop_entry_edge, loop_version): Set probabilities
4793 and frequencies according to arguments.
4794 * tree-ssa-loop-manip.c (tree_unroll_loop): Set probabilities
4796 * cfg.c (scale_bbs_frequencies_int): Allow scaling the frequencies up.
4797 * modulo-sched.c (sms_schedule): Set probabilities for entering
4798 versioned loop correctly.
4799 * tree-vect-transform.c (vect_transform_loop): Ditto.
4800 * cfgloop.h (loopify, loop_version): Declaration changed.
4801 (scale_loop_frequencies): Declared.
4803 2007-01-02 Jan Hubicka <jh@suse.cz>
4805 * cgraph.c: Include tree-flow.h
4806 (cgraph_add_new-function): Handle IPA_SSA mode; execute
4808 * cgraph.h (enum cgraph_state): Add CGRAPH_STATE_IPA_SSA.
4809 * tree-pass.h (pass_all_early_optimizations): Declare.
4810 * cgraphunit.c (cgraph_process_new_functions): Add IPA_SSA; execute
4812 (cgraph_analyze_function): Do early_local_passes.
4813 * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
4814 Do not add referenced vars.
4815 * tree-optimize.c (gate_all_optimizations): Do not execute when not in
4817 (gate_all_early_local_passes): New gate.
4818 (pass_early_local_passes): Use new gate.
4819 (execute_early_local_optimizations): New functions.
4820 (gate_all_early_optimizations): New gate.
4821 (pass_all_early_optimizations): New pass.
4822 (execute_free_datastructures): Free SSA only when initialized.
4823 (gate_init_datastructures): Init only when optimizing.
4824 (tree_lowering_passes): Do early local passes when called late.
4825 * tree-profile.c (do_tree_profiling): Don't profile functions added
4827 (do_early_tree_profiling, pass_early_tree_profile): Kill.
4828 * tree-cfg.c (update_modified_stmts): Do not update when operands are
4830 * passes.c (init_optimizations_passes): Reorder so we go into SSA
4831 during early_local_passes.
4832 * Makefile.in (cgraph.o): Add dependency on tree-flow.h.
4835 2007-01-02 Carlos O'Donell <carlos@codesourcery.com>
4837 * Makefile.in: Update copyright year.
4839 2007-01-02 Carlos O'Donell <carlos@codesourcery.com>
4841 * Makefile.in: Export GCC_EXEC_PREFIX before calling $(RUNTEST)
4842 in $(lang_checks) and check-consistency targets.
4844 2007-01-02 Jan Hubicka <jh@suse.cz>
4846 * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
4847 Do not add referenced vars.
4848 * tree-cfg.c (update_modified_stmts): Do not update when SSA operands
4850 * passes.c (init_optimization_passes): Put mudflap_2 after
4851 free_datastructures.
4853 2007-01-02 Jan Hubicka <jh@suse.cz>
4855 * tree-optimize (execute_fixup_cfg): Set after_inlining flag.
4856 Set NOTHROW flag on call statements proved to be nothrow.
4857 Update statement of local calls so new pure/const functions are
4858 updated. Update_ssa when in ssa form. Mark PHI nodes of nonlocal
4860 (tree_rest_of_compilation): Register hooks and initialize bitmap
4861 early. Do not set after_inlining flag.
4863 2007-01-02 Steve Ellcey <sje@cup.hp.com>
4865 * sbitmap.c (HOST_BITS_PER_LONG_LONG): Change to
4866 HOST_BITS_PER_LONGLONG
4868 2007-01-02 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
4871 * c-typeck.c (store_init_value): Don't emit pedantic overflow
4872 warning for non-static initializers.
4874 2007-01-02 Steven Bosscher <steven@gcc.gnu.org>
4876 * config/alpha/alpha.md, arm/arm.c, darwin.c, frv/frv.md,
4877 m32r/m32r.c, m32r/m32r.c, mn10300/mn10300.md, pa/pa.c,
4878 rs6000/rs6000.c, s390/s390.md, sh/sh.md, sparc/sparc.c:
4879 Always use set_unique_reg_note to add REG_EQUAL notes.
4881 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
4884 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
4886 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
4887 * cfgbuild.c (find_basic_blocks): Likewise.
4888 * cfgrtl.c (rtl_create_basic_block): Likewise.
4889 * function.c (temp_slots_at_level): Likewise.
4890 * reg-stack.c (stack_regs_mentioned): Likewise.
4891 * regclass.c (allocate_reg_info): Likewise.
4892 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
4893 set_bb_for_stmt, move_block_to_fn): Likewise.
4894 * tree-complex.c (tree_lower_complex): Likewise.
4895 * vec.h (VEC_safe_grow_cleared): New.
4897 2007-01-02 Ian Lance Taylor <iant@google.com>
4899 * c-common.c (c_common_truthvalue_conversion): When warning about
4900 using an assignment as a truth value, set TREE_NO_WARNING.
4902 2007-01-02 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
4905 * c.opt (Wold-style-declaration): New.
4906 * doc/invoke.texi (C-only Warning Options): New.
4907 (Wold-style-declaration): Document it.
4908 (Wextra): Enabled by -Wextra.
4909 * c-opts.c (c_common_post_options): Enabled by -Wextra.
4910 * c-decl.c (declspecs_add_scspec): Replace -Wextra with
4911 -Wold-style-declaration.
4913 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
4915 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
4916 * cfgbuild.c (find_basic_blocks): Likewise.
4917 * cfgrtl.c (rtl_create_basic_block): Likewise.
4918 * function.c (temp_slots_at_level): Likewise.
4919 * reg-stack.c (stack_regs_mentioned): Likewise.
4920 * regclass.c (allocate_reg_info): Likewise.
4921 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
4922 set_bb_for_stmt, move_block_to_fn): Likewise.
4923 * tree-complex.c (tree_lower_complex): Likewise.
4924 * vec.h (VEC_safe_grow_cleared): New.
4926 2007-01-02 Douglas Gregor <doug.gregor@gmail.com>
4928 * c-common.c (c_common_nodes_and_builtins): Since variants of
4929 void_type_node get built before it is given a name, we need to
4930 give those variants the name, too.
4931 (complete_array_type): We need to work with the canonical main
4932 type of the array, from which we will build the qualified version.
4933 * params.def (PARAM_VERIFY_CANONICAL_TYPES): New.
4934 * print-tree.c (print_node): Display canonical type information
4936 * stor-layout.c (layout_type): When we don't know the
4937 alignment of a type for which we're building an array, we end up
4938 guessing wrong, so make the type require structural equality.
4939 * tree.c (make_node_stat): When we build a new type, it is its
4941 (build_type_attribute_qual_variant): When building an attribute
4942 variant, its canonical type is the non-attribute variant. However,
4943 if the attributes are target-dependent and they differ, we need to
4944 use structural equality checks for this type.
4945 (build_qualified_type): A qualified type is not equivalent to its
4946 unqualified variant; set the canonical type appropriately.
4947 (build_distinct_type_copy): When building a distinct type from
4948 another type, the new type is its own canonical type.
4949 (build_variant_type_copy): When building a new type variant, we
4950 assume that it is equivalent to the original type.
4951 (build_pointer_type_for_mode): When building a pointer type, also
4952 build a canonical type pointer.
4953 (build_reference_type_for_mode): When building a reference type,
4954 also build a canonical type reference.
4955 (build_index_type): When we can't hash an index type (e.g.,
4956 because its maximum value is negative), the index type requires
4957 structural equality tests.
4958 (build_array_type): Build the canonical form of an array type.
4959 (build_function_type): Function types require structural equality,
4960 because they contain default arguments, attributes, etc.
4961 (build_method_type_directly): Ditto for method types.
4962 (build_offset_type): Build the canonical offset type.
4963 (build_complex_type): Build the canonical vector type.
4964 (make_vector_type): Build the canonical vector type.
4965 * tree.h (TYPE_CANONICAL): New.
4966 (TYPE_STRUCTURAL_EQUALITY_P): New.
4967 (SET_TYPE_STRUCTURAL_EQUALITY): New.
4968 (struct tree_type): Added "canonical" field.
4969 * params.h (VERIFY_CANONICAL_TYPES): New.
4970 * doc/c-tree.texi (TYPE_CANONICAL): Document.
4971 (TYPE_STRUCTURAL_EQUALITY_P): Document.
4972 (SET_TYPE_STRUCTURAL_EQUALITY): Document.
4973 * doc/invoke.texi (verify-canonical-types): Document --param
4974 parameter for verifying canonical types.
4976 2007-01-02 Joseph Myers <joseph@codesourcery.com>
4978 * config.gcc (powerpc-*-eabispe*, powerpc-*-eabisimaltivec*,
4979 powerpc-*-eabisim*, powerpc-*-eabialtivec*, powerpc-*-eabi*,
4980 powerpc-*-rtems*, powerpc-wrs-vxworks, powerpc-wrs-vxworksae,
4981 powerpcle-*-eabisim*, powerpcle-*-eabi*): Add rs6000/e500.h to
4983 * config/rs6000/e500.h: New.
4984 * config/rs6000/eabi.h (TARGET_SPE_ABI, TARGET_SPE, TARGET_E500,
4985 TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE, TARGET_E500_DOUBLE):
4987 * config/rs6000/linuxspe.h (TARGET_SPE_ABI, TARGET_SPE,
4988 TARGET_E500, TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE,
4989 TARGET_E500_DOUBLE): Remove.
4990 * config/rs6000/vxworks.h (TARGET_SPE_ABI, TARGET_SPE,
4991 TARGET_E500, TARGET_ISEL, TARGET_FPRS): Remove.
4992 * config/rs6000/rs6000.h (CHECK_E500_OPTIONS): Define.
4993 * config/rs6000/rs6000.c (rs6000_override_options): Use
4996 2007-01-02 Joseph Myers <joseph@codesourcery.com>
4998 * config/rs6000/rs6000.c (print_operand): Check (TARGET_SPE ||
4999 TARGET_E500_DOUBLE), not TARGET_E500, for %y.
5000 (rs6000_generate_compare, rs6000_emit_sCOND, output_cbranch,
5001 rs6000_emit_cmove): Don't check TARGET_E500.
5002 * config/rs6000/rs6000.md (bunordered, bordered, sunordered,
5003 sordered): Don't check TARGET_E500.
5005 2007-01-01 Eric Christopher <echristo@apple.com>
5007 * config/mips/mips.c (mips_regno_mode_ok_for_base_p): Use
5008 HARD_REGISTER_NUM_P.
5010 2007-01-01 Roger Sayle <roger@eyesopen.com>
5012 * fold-const.c (fold_binary) <EQ_EXPR>: Fold "(X^C1) eq/ne C2" into
5013 "X eq/ne (C1^C2)". Fold "(X^Z) eq/ne (Y^Z)" as "X eq/ne Y" when Z
5014 has no side-effects. Fold "(X^C1) eq/ne (Y^C2)" as "(X^(C1^C2))
5017 2007-01-01 Mike Stump <mrs@apple.com>
5019 * configure.ac: Remove support for building with Apple's gcc-3.1.
5021 2007-01-02 Joseph Myers <joseph@codesourcery.com>
5024 * caller-save.c (add_stored_regs): Only handle SUBREGs if inner
5025 REG is a hard register. Do not modify REG before calling
5027 * rtlanal.c (subreg_get_info): Don't assert size of XMODE is a
5028 multiple of the size of YMODE for certain lowpart cases.
5030 2007-01-01 Andrew Pinski <pinskia@gmail.com>
5033 * gimplify (voidify_wrapper_expr): Update for
5036 2007-01-01 Andreas Schwab <schwab@suse.de>
5039 * config/ia64/ia64.c (ia64_compute_frame_size): Account space for
5040 save of BR0 in extra_spill_size instead of spill_size.
5041 (ia64_expand_prologue): Save BR0 outside of the gr/br/fr spill
5043 (ia64_expand_epilogue): Restore BR0 from its new location.
5045 2007-01-01 Andrew Pinski <pinskia@gmail.com>
5047 * gimplify.c (gimplify_init_constructor <case VECTOR_TYPE>):
5048 Use a temporary variable if the left hand side is not a gimple
5051 2007-01-01 Andrew Pinski <pinskia@gmail.com>
5053 * gimplify.c (gimplify_return_expr): Make the temporary variable
5054 for the return expression, a gimple register variable.
5056 2007-01-01 Jan Hubicka <jh@suse.cz>
5058 * emit-rtl.c (emit_copy_of_insn_after): Do not call copy_insn_1 for
5061 2007-01-01 Mike Stump <mrs@apple.com>
5063 * configure.ac (HAVE_GAS_LITERAL16): Add autoconf check for
5065 * config/darwin.c (machopic_select_rtx_section): Use
5067 (darwin_mergeable_constant_section): Likewise.
5068 * configure: Regenerate.
5069 * config.in: Regenerate.
5071 2007-01-01 Jan Hubicka <jh@suse.cz>
5072 Andrew Pinski <pinskia@gmail.com>
5074 * cgraphunit.c (cgraph_optimize): Call cgraph_add_new_functions
5075 before starting IPA passes.