1 2007-02-14 Nick Clifton <nickc@redhat.com>
3 * builtin-types.def (DEF_FUNCTION_TYPE_x): Do not imply that at
4 most 3 arguments are supported.
5 (DEF_FUNCTION_TYPE_VAR_5): Fix typo in its description.
7 2007-02-13 Seongbae Park <seongbae.park@gmail.com>
9 * bitmap.c (bitmap_and, bitmap_and_compl, bitmap_xor):
10 Ensure dst->current is valid.
12 2007-02-13 Paul Brook <paul@codesourcery.com>
14 * config.gcc: Add arm*-*-uclinux-*eabi.
15 * config/arm/uclinux-elf.h (TARGET_OS_CPP_BUILTINS): Define.
16 (SUBTARGET_EXTRA_LINK_SPEC): Define.
17 (STARTFILE_SPEC, ENDFILE_SPEC): Remove broken -shared handling.
18 (LINK_GCC_C_SEQUENCE_SPEC): Undef.
21 * config/arm/arm.c (arm_override_options): Use r9 as EABI PIC
23 * config/arm/uclinux-eabi.h: New file.
24 * config/arm/linux-eabi.h (WCHAR_TYPE): Remove.
25 * config/arm/linux-gas.h (WCHAR_TYPE): Use unsigned long on AAPCS
28 2007-02-13 Ian Lance Taylor <iant@google.com>
30 * common.opt: Add Wstrict-overflow and Wstrict-overflow=.
31 * flags.h (warn_strict_overflow): Declare.
32 (enum warn_strict_overflow_code): Define.
33 (issue_strict_overflow_warning): New static inline function.
34 * opts.c (warn_strict_overflow): New variable.
35 (common_handle_option): Handle OPT_Wstrict_overflow and
36 OPT_Wstrict_overflow_.
37 * c-opts.c (c_common_handle_option): Set warn_strict_overflow for
39 * fold-const.c: Include intl.h.
40 (fold_deferring_overflow_warnings): New static variable.
41 (fold_deferred_overflow_warning): New static variable.
42 (fold_deferred_overflow_code): New static variable.
43 (fold_defer_overflow_warnings): New function.
44 (fold_undefer_overflow_warnings): New function.
45 (fold_undefer_and_ignore_overflow_warnings): New function.
46 (fold_deferring_overflow_warnings_p): New function.
47 (fold_overflow_warning): New static function.
48 (make_range): Add strict_overflow_p parameter. Change all
50 (extract_muldiv, extract_muldiv_1): Likewise.
51 (fold_unary) [ABS_EXPR]: Check ABS_EXPR before calling
52 tree_expr_nonnegative_p.
53 (fold_negate_expr): Call fold_overflow_warning.
54 (fold_range_test): Likewise.
55 (fold_comparison): Likewise.
56 (fold_binary): Likewise. Call tree_expr_nonnegative_warnv_p
57 instead of tree_expr_nonnegative_p.
58 (tree_expr_nonnegative_warnv_p): Rename from
59 tree_expr_nonnegative_p, add strict_overflow_p parameter.
60 (tree_expr_nonnegative_p): New function.
61 (tree_expr_nonzero_warnv_p): Rename from tree_expr_nonzero_p, add
62 strict_overflow_p parameter.
63 (tree_expr_nonzero_p): New function.
64 * passes.c (verify_interpass_invariants): New static function.
65 (execute_one_pass): Call it.
66 * tree-ssa-loop-niter.c (expand_simple_operations): Ignore fold
68 (number_of_iterations_exit, loop_niter_by_eval): Likewise.
69 (estimate_numbers_of_iterations): Likewise.
70 (scev_probably_wraps_p): Likewise.
71 * tree-ssa-ccp.c: Include "toplev.h".
72 (evaluate_stmt): Defer fold overflow warnings until we know we are
74 (struct fold_stmt_r_data): Add stmt field.
75 (fold_stmt_r): Defer fold overflow warnings until we know we
77 (fold_stmt): Initialize stmt field of fold_stmt_r_data.
78 (fold_stmt_inplace): Likewise.
79 * tree-cfgcleanup.c: Include "toplev.h" rather than "errors.h".
80 (cleanup_control_expr_graph): Defer fold overflow warnings until
81 we know we are going to optimize.
82 * tree-cfg.c (fold_cond_expr_cond): Likewise.
83 * tree-ssa-threadedge.c (simplify_control_stmt_condition):
85 * tree-vrp.c (vrp_expr_computes_nonnegative): Call
86 tree_expr_nonnegative_warnv_p instead of tree_expr_nonnegative_p.
87 * tree-ssa-loop-manip.c (create_iv): Likewise.
88 * c-typeck.c (build_conditional_expr): Likewise.
89 (build_binary_op): Likewise.
90 * tree-vrp.c (vrp_expr_computes_nonzero): Call
91 tree_expr_nonzero_warnv_p instead of tree_expr_nonzero_p.
92 (extract_range_from_unary_expr): Likewise.
93 * simplify-rtx.c (simplify_const_relational_operation): Warn when
94 assuming that signed overflow does not occur.
95 * c-common.c (pointer_int_sum): Ignore fold overflow warnings.
96 * tree.h (tree_expr_nonnegative_warnv_p): Declare.
97 (fold_defer_overflow_warnings): Declare.
98 (fold_undefer_overflow_warnings): Declare.
99 (fold_undefer_and_ignore_overflow_warnings): Declare.
100 (fold_deferring_overflow_warnings_p): Declare.
101 (tree_expr_nonzero_warnv_p): Declare.
102 * doc/invoke.texi (Option Summary): Add -Wstrict-overflow to list
104 (Warning Options): Document -Wstrict-overflow.
105 * Makefile.in (tree-ssa-threadedge.o): Depend on toplev.h.
106 (tree-ssa-ccp.o): Likewise.
107 (tree-cfgcleanup.o): Change errors.h dependency to toplev.h.
108 (fold-const.o): Depend on intl.h.
110 2007-02-13 Ian Lance Taylor <iant@google.com>
113 * lower-subreg.c (resolve_simple_move): Decompose subregs in
116 2007-02-13 Stuart Hastings <stuart@apple.com>
118 * gcc/config/i386/i386.md (fixuns_truncdfhi2): Require SSE2.
120 2007-02-13 Richard Henderson <rth@redhat.com>
122 * config/alpha/alpha.c (alpha_stdarg_optimize_hook): Strip
123 handled_component_p before looking for the indirect_ref.
125 2007-02-13 Richard Henderson <rth@redhat.com>
127 * config/i386/i386.md (bswapsi_1): Rename from bswapsi2,
128 remove flags clobber.
129 (bswapsi2): New expander, emit code for !TARGET_BSWAP.
130 (bswaphi_lowpart): New.
131 (bswapdi2): Rename from bswapdi2_rex, remove flags clobber,
132 remove TARGET_BSWAP test. Delete expander of the same name.
134 * optabs.c (widen_bswap, expand_doubleword_bswap): New.
135 (expand_unop): Use them.
137 2007-02-13 Uros Bizjak <ubizjak@gmail.com>
139 * config/i386/i386.md (cmpdi_ccno_1_rex64, *cmpsi_ccno_1,
140 *cmphi_ccno_1, *cmpqi_ccno_1, *movsi_xor, *movstricthi_xor,
141 *movstrictqi_xor, *movdi_xor_rex64, *ashldi3_1_rex64,
142 *ashldi3_cmp_rex64, *ashldi3_cconly_rex64, ashlsi3, *ashlsi3_1_zext,
143 *ashlsi3_cmp, *ashlsi3_cconly, *ashlsi3_cmp_zext, *ashlhi3_1_lea,
144 *ashlhi3_1, *ashlhi3_cmp, *ashlhi3_cconly, *ashlqi3_1_lea,
145 *ashlqi3_1, *ashlqi3_cmp, *ashlqi3_cconly): Remove equivalent
146 assembler dialect choice from asm templates.
148 2007-02-12 Richard Henderson <rth@redhat.com>
150 * config/i386/i386.md (fixuns_trunc<SSEMODEF>si_1): New insn.
151 (fixuns_trunc<SSEMODEF>si2): Use it.
152 * config/i386/sse.md (vec_setv4sf_0): Export.
153 * config/i386/i386.c (ix86_build_const_vector): Export.
154 (ix86_split_convert_uns_si_sse): Rename from
155 ix86_expand_convert_uns_si_sse and rewrite as a splitter.
156 * config/i386/i386-protos.h: Update.
158 2007-02-13 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
161 * c-typeck.c (c_finish_return): Improve warning message.
163 2007-02-12 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
165 * alias.c (find_symbolic_term): Delete unused function.
167 2007-02-12 Uros Bizjak <ubizjak@gmail.com>
169 * config/i386/i386.md (paritydi2, paritysi2): New expanders.
170 (paritydi2_cmp, paritydi2_cmp): New insn and split patterns.
171 (*parityhi2_cmp, *parityqi2_cmp): New insn patterns.
173 2007-02-12 Eric Botcazou <ebotcazou@adacore.com>
175 * tree.h (DECL_IGNORED_P): Document further effect for FUNCTION_DECL.
176 * cgraphunit.c (cgraph_expand_function): If DECL_IGNORED_P is set on
177 the function, temporarily point the debug interface to the null one.
179 2007-02-12 Eric Botcazou <ebotcazou@adacore.com>
181 * dwarf2out.c (round_up_to_align): New static function.
182 (field_byte_offset): Use it to round the offset.
184 2007-02-12 Richard Henderson <rth@redhat.com>
186 * config/alpha/alpha.md (bswapsi2, bswapdi2): New.
187 (inswl_const): Export.
189 2007-02-12 Richard Henderson <rth@redhat.com>
191 * calls.c (emit_library_call_value_1): If PROMOTE_MODE modifed the
192 result mode of the libcall, convert back to outmode.
194 2007-02-12 Roger Sayle <roger@eyesopen.com>
196 * config/i386/i386.md (*bswapdi2_rex): Renamed from bswapdi2.
197 (bswapdi2): New define_expand to implement 32-bit implementation.
199 2007-02-12 Nick Clifton <nickc@redhat.com>
201 * doc/invoke.texi (Overall Options): Document --help=.
202 * gcc.c (target_help_flag): Rename to print_subprocess_flag.
203 (cc1_options): Pass --help= on to cc1.
204 (display_help): Add description of --help=.
205 (process_command): Add code to handle --help=. Allow translated
206 --help and --target-help switches to be passed on to compiler
208 (main): Remove unused if statement.
209 * opts.c (columns): Remove.
210 (LEFT_COLUMN): Define.
211 (wrap_help): Add columns argument.
212 (print_filtered_help): Change parameters to be an include bitmask,
213 an exclude bitmask, an any bitmask and the column width. Move the
214 code to display the params list here. Add code to display the
215 status of options rather than their descriptions if the quiet flag
217 (print_specific_help): Change parameters to be an include bitmask,
218 an exclude bitmask and an any bitmask. Move code to look up the
219 column width here. Decide upon the title for an options listing.
220 (common_handle_options): Add code to handle --help=. Adapt code
221 for --help and --target-help to use the revised form of the
222 print_specific_help function.
223 (print_help): Delete.
224 (print_param_help): Delete.
225 (print_switch): Delete.
226 * opts.h (cl_lang_count): Add prototype.
227 (CL_PARAMS, CL_WARNING, CL_OPTIMIZATION, CL_MIN_OPTION_CLASS,
228 CL_MAX_OPTION_CLASS): New defines.
229 * optc-gen.awk: Add construction of cl_lang_count.
230 * c.opt: Add Warning attribute to warning options and Optimization
231 attribute to optimization options.
232 * common.opt: Likewise.
234 Add -fhelp and -ftarget-help as aliases for the transformed --help
235 and --target-help options.
236 * opt-functions.awk: Add code to handle Warning and Optimization
239 2007-02-12 Richard Henderson <rth@redhat.com>
241 * config/alpha/constraints.md: New file.
242 * config/alpha/alpha.c: Include tm-constrs.h.
243 (alpha_const_ok_for_letter_p, alpha_const_double_ok_for_letter_p,
244 alpha_extra_constraint): Remove.
245 (alpha_emit_conditional_branch): Use satisfies_constraint_*.
246 * config/alpha/alpha-protos.h: Update.
247 * config/alpha/alpha.h (REG_CLASS_FROM_LETTER): Remove.
248 (CONST_OK_FOR_LETTER_P, CONST_DOUBLE_OK_FOR_LETTER_P): Remove.
249 (EXTRA_CONSTRAINT): Remove.
250 * config/alpha/alpha.md: Include constraints.md.
251 (adddi splitter): Use satisfies_constraint_*.
252 * config/alpha/predicates.md (add_operand): Likewise.
253 (sext_add_operand, addition_operation): Likewise.
255 2007-02-12 Dorit Nuzman <dorit@il.ibm.com>
257 PR tree-optimization/29145
258 * tree-data-ref.c (base_addr_differ_p): Make us more conservative
259 in our handling of restrict qualified pointers.
261 2007-02-12 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
264 * doc/invoke.texi (Wunused-value): Update description.
265 (Wextra): Delete item.
266 * opts.c (set_Wextra): Don't use the value of Wextra to set the
267 value of Wunused-value.
268 * c-typeck.c (c_process_expr_stmt): Don't check extra_warnings.
269 (c_finish_stmt_expr): Don't check extra_warnings.
270 (emit_side_effect_warnings): The caller is responsible to check
273 2007-02-11 Roger Sayle <roger@eyesopen.com>
274 Matt Thomas <matt@3am-software.com>
276 * simplify-rtx.c (simplify_relational_operation_1): Correct typo.
278 2007-02-11 Roger Sayle <roger@eyesopen.com>
280 * simplify-rtx.c (simplify_relational_operation_1): Optimize
281 comparisons of POPCOUNT against zero.
282 (simplify_const_relational_operation): Likewise.
284 2007-02-11 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
286 * doc/invoke.texi (Wextra): Delete outdated paragraph.
288 2007-02-11 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
290 * dwarf2out.c (root_type): Delete unused function.
292 2007-02-11 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
294 * genattrtab.c (contained_in_p): Delete unused function.
295 (write_expr_attr_cache): Likewise.
297 2007-02-11 Jan Hubicka <jh@suse.cz>
299 * ipa-inline.c (cgraph_edge_badness): Add "else" missing in the
302 2007-02-11 Steven Bosscher <steven@gcc.gnu.org>
304 * fwprop.c (try_fwprop_subst): Use set_unique_reg_note
305 to add the REG_EQ* note.
306 * see.c (see_merge_one_use_extension): Likewise.
307 * local-alloc.c (update_equiv_regs): Likewise. Also don't
308 turn REG_EQUAL notes into REG_EQUIV notes if the target
309 register may have more than one set.
310 * function.c (assign_parm_setup_reg): Use set_unique_reg_note.
311 * gcse.c (try_replace_reg): Likewise.
312 * alias.c (init_alias_analysis): Use find_reg_equal_equiv_note.
313 * calls.c (fixup_tail_calls): Likewise. Abort if there is
314 more than one REG_EQUIV note.
315 * reload1.c (gen_reload): Use set_unique_reg_note.
317 2007-02-11 Uros Bizjak <ubizjak@gmail.com>
319 * config/i386/i386.c (TARGET_VECTORIZE_BUILTIN_CONVERSION): Define.
320 (ix86_builtin_conversion): New function.
322 2007-02-06 Mark Mitchell <mark@codesourcery.com>
325 * tree.h (DECL_REPLACEABLE_P): New macro.
326 * except.c (set_nothrow_function_flags): Likewise.
328 2007-02-11 Tehila Meyzels <tehila@il.ibm.com>
329 Ira Rosen <irar@il.ibm.com>
330 Dorit Nuzman <dorit@il.ibm.com>
332 * doc/tm.texi (TARGET_VECTORIZE_BUILTIN_CONVERSION): New target hook.
333 * targhooks.c (default_builtin_vectorized_conversion): New.
334 * targhooks.h (default_builtin_vectorized_function): New declaration.
335 * target.h (struct vectorize): Add builtin_conversion field.
336 * tree-vectorizer.h (type_conversion_vec_info_type): New enum
337 stmt_vec_info_type value.
338 (vectorizable_conversion): New declaration.
339 * tree-vect-analyze.c (vect_analyze_operations): Add
340 vectorizable_conversion call.
341 * target-def.h (TARGET_VECTORIZE_BUILTIN_CONVERSION): New.
342 * tree-vect-transform.c (vectorizable_conversion): New function.
343 (vect_transform_stmt): Add case for type_conversion_vec_info_type.
344 * tree-vect-generic.c (expand_vector_operations_1): Consider correct
346 * config/rs6000/rs6000.c (rs6000_builtin_conversion): New.
347 (TARGET_VECTORIZE_BUILTIN_CONVERSION): Defined.
348 (rs6000_expand_builtin): Add handling a case of ALTIVEC_BUILTIN_VCFUX
349 or ALTIVEC_BUILTIN_VCFSX.
351 2007-02-10 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
354 * pa.md (movdf): For 64-bit target, fail if operand 1 is a non-zero
355 CONST_DOUBLE and operand 0 is a hard register.
356 (movdi): For 64-bit target, remove code to force CONST_DOUBLE to
357 memory. Fail if operand 1 is a non-zero CONST_INT and operand 0
358 is a hard floating-point register.
360 2007-02-10 Richard Henderson <rth@redhat.com>
361 Jakub Jelinek <jakub@redhat.com>
362 Alexandre Oliva <aoliva@redhat.com>
364 * Makefile.in (libgcc-support, libgcc.mvars): Add emutls.c.
365 * builtin-types.def (BT_WORD): Make unsigned.
366 (BT_FN_VOID_PTR_WORD_WORD_PTR): New.
367 * builtins.def (BUILT_IN_EMUTLS_GET_ADDRESS): New.
368 (BUILT_IN_EMUTLS_REGISTER_COMMON): New.
369 * c-decl.c (grokdeclarator): Don't error if !have_tls.
370 * c-parser.c (c_parser_omp_threadprivate): Likewise.
371 * dwarf2out.c (loc_descriptor_from_tree_1): Don't do anything for
373 * expr.c (emutls_var_address): New.
374 (expand_expr_real_1): Expand emulated tls.
375 (expand_expr_addr_expr_1): Likewise.
376 * libgcc-std.ver: Add __emutls_get_address, __emutls_register_common.
377 * output.h (emutls_finish): Declare.
378 * toplev.c (compile_file): Call it.
379 * tree-ssa-address.c (gen_addr_rtx): Check for const-ness of the
380 address before wrapping in CONST.
381 * varasm.c (emutls_htab, emutls_object_type): New.
382 (EMUTLS_VAR_PREFIX, EMUTLS_TMPL_PREFIX): New.
383 (get_emutls_object_name, get_emutls_object_type): New.
384 (get_emutls_init_templ_addr, emutls_decl): New.
385 (emutls_common_1, emutls_finish): New.
386 (assemble_variable): When emulating tls, swap decls; generate
387 constructor for the emutls objects.
388 (do_assemble_alias): When emulating tls, swap decl and target name.
389 (default_encode_section_info): Don't add SYMBOL_FLAG_TLS_SHIFT
391 * varpool.c (decide_is_variable_needed): Look at force_output.
392 Recurse for emulated tls.
393 (cgraph_varpool_remove_unreferenced_decls): Remove checks redundant
394 with decide_is_variable_needed.
395 * emutls.c: New file.
396 * config/sparc/sol2.h (ASM_DECLARE_OBJECT_NAME): Only emit
397 tls_object for real tls.
399 2007-02-10 Kaz Kojima <kkojima@gcc.gnu.org>
401 PR rtl-optimization/29599
402 * reload1.c (eliminate_regs_in_insn): Take the destination
403 mode into account when computing the offset.
405 2007-02-09 Stuart Hastings <stuart@apple.com>
406 Richard Henderson <rth@redhat.com>
408 * gcc/config/i386/i386.h (TARGET_KEEPS_VECTOR_ALIGNED_STACK): New.
409 * gcc/config/i386/darwin.h: (TARGET_KEEPS_VECTOR_ALIGNED_STACK): New.
410 * gcc/config/i386/i386.md (fixuns_trunc<mode>si2, fixuns_truncsfhi2,
411 fixuns_truncdfhi2): New.
412 (fix_truncsfdi_sse): Call ix86_expand_convert_sign_didf_sse.
413 (floatunsdidf2): Call ix86_expand_convert_uns_didf_sse.
414 (floatunssisf2): Add call to ix86_expand_convert_uns_sisf_sse.
415 (floatunssidf2): Allow nonimmediate source.
416 * gcc/config/i386/sse.md (movdi_to_sse): New.
417 (vec_concatv2di): Drop '*'.
418 * gcc/config/i386/i386-protos.h (ix86_expand_convert_uns_si_sse,
419 ix86_expand_convert_uns_didf_sse, ix86_expand_convert_uns_sidf_sse,
420 ix86_expand_convert_uns_sisf_sse, ix86_expand_convert_sign_didf_sse):
422 * gcc/config/i386/i386.c (ix86_expand_convert_uns_si_sse,
423 ix86_expand_convert_uns_didf_sse, ix86_expand_convert_uns_sidf_sse,
424 ix86_expand_convert_uns_sisf_sse, ix86_expand_convert_sign_didf_sse,
425 ix86_build_const_vector, ix86_expand_vector_init_one_nonzero): New.
426 (ix86_build_signbit_mask): Fix decl of v, refactor to call
427 ix86_build_const_vector.
428 (x86_emit_floatuns): Rewrite.
430 2007-02-10 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
432 * genautomata.c (longest_path_length): Delete unused function.
433 (struct state): Delete unused longest_path_length.
434 (UNDEFINED_LONGEST_PATH_LENGTH): Delete unused macro.
435 (get_free_state): Delete unused.
437 2007-02-09 Jan Hubicka <jh@suse.cz>
439 * params.def (PARAM_INLINE_UNIT_GROWTH): Set to 30.
440 * doc/invoke.texi (inline-unit-growth): Update default value.
442 * Makefile.in (passes.o, ipa-inline.o): Add dependencies.
443 * cgraphbuild.c (build_cgraph_edges): Compute frequencies.
444 (rebuild_cgraph_edges): Likewise.
445 * cgraph.c (cgraph_set_call_stmt): Add new argument frequency.
446 (dump_cgraph_node): Dump frequencies.
447 (cgraph_clone_edge): Add frequency scales.
448 (cgraph_clone_node): Add freuqnecy.
449 * cgraph.h (cgraph_edge): Add freuqnecy argument.
450 (CGRAPH_FREQ_BASE, CGRAPH_FREQ_MAX): New constants.
451 (cgraph_create_edge, cgraph_clone_edge, cgraph_clone_node): Update.
452 * tree-pass.h (TODO_rebuild_frequencies): New constant.
453 * cgraphunit.c (verify_cgraph_node): Verify frequencies.
454 (cgraph_copy_node_for_versioning): Update call of cgraph_clone_edge.
455 (save_inline_function_body): Likewise.
456 * ipa-inline.c: inluce rtl.h
457 (cgraph_clone_inlined_nods): Update call of cgraph_clone_node.
458 (cgraph_edge_badness): Use frequencies.
459 (cgraph_decide_recursive_inlining): Update clonning.
460 (cgraph_decide_inlining_of_small_function): Dump frequency.
461 * predict.c (estimate_bb_frequencies): Export.
462 * predict.h (estimate_bb_frequencies): Declare.
463 * tree-inline.c (copy_bb): Watch overflows.
464 (expand_call_inline): Update call of cgraph_create_edge.
465 (optimize_inline_calls): Use TODO flags to update frequnecies.
466 * passes.h: Include predict.h
467 (init_optimization_passes): Move profile ahead.
468 (execute_function_todo): Handle TODO_rebuild_frequencies.
470 2007-02-09 Roger Sayle <roger@eyesopen.com>
472 * config/alpha/alpha.c (emit_insxl): Force the first operand of
473 the insbl or inswl pattern into a register.
475 2007-02-09 Roger Sayle <roger@eyesopen.com>
477 * config/ia64/ia64.md (bswapdi2): New define_insn.
479 2007-02-09 Richard Henderson <rth@redhat.com>
481 * config/i386/constraints.md (Ym): New constraint.
482 * config/i386/i386.md (movsi_1): Change Y2 to Yi constraints.
483 (movdi_1_rex64): Split sse and xmm general register moves from
484 memory move alternatives. Use conditional register constraints.
485 (movsf_1, movdf_integer): Likewise.
486 (zero_extendsidi2_32, zero_extendsidi2_rex64): Likewise.
487 (movdf_integer_rex64): New.
488 (pushsf_rex64): Fix output constraints.
489 * config/i386/sse.md (sse2_loadld): Split rm alternative, use Yi.
490 (sse2_stored): Likewise.
491 (sse2_storeq_rex64): New.
492 * config/i386/i386.c (x86_inter_unit_moves): Enable for not
494 (ix86_secondary_memory_needed): Don't bypass TARGET_INTER_UNIT_MOVES
495 for optimize_size. Remove SF/DFmode hack.
497 2007-02-09 Dwarakanath Rajagopal <dwarak.rajagopal@amd.com>
499 * config/i386/driver-i386.c: Turn on -mtune=native for AMDFAM10.
502 2007-02-09 Nathan Sidwell <nathan@codesourcery.com>
503 Richard Sandiford <richard@codesourcery.com>
505 * config.gcc (m68010-*-netbsdelf*, m68k*-*-netbsdelf*)
506 (m68k*-*-openbsd*, m68k-*-linux*): Set default_cf_cpu.
507 (m68k-*-aout*, m68k-*-coff*, m68k-*-uclinux*, m68k-*-rtems*): Add
508 m68k/t-mlib to tmake_file.
509 (m68020-*-elf*, m68k-*-elf*): Likewise. Add t-m68kbare as well.
510 (m68k*-*-*): Use --with-arch to pick a default for --with-cpu.
511 (m680[012]0-*-*, m68k*-*-*): Add support for --with-arch.
512 Allow it to be cf or m68k. Set m68k_arch_family. If that
513 variable is not empty, add t-$m68k_arch_family to tmake_file.
514 Add t-mlibs to tmake_file.
515 * doc/install.texi: Document --with-arch=m68k and --with-arch=cf.
516 * config/m68k/t-cf: New file.
517 * config/m68k/t-m68k: Likewise.
518 * config/m68k/t-mlibs: Likewise.
519 * config/m68k/t-m68kbare (MULTILIB_OPTIONS, MULTILIB_DIRNAMES)
520 (MULTILIB_MATCHES, MULTILIB_EXCEPTIONS): Delete.
521 (M68K_MLIB_DIRNAMES, M68K_MLIB_OPTIONS): Define.
522 * config/m68k/t-m68kelf (MULTILIB_OPTIONS, MULTILIB_DIRNAMES)
523 (MULTILIB_MATCHES, MULTILIB_EXCEPTIONS, LIBGCC, INSTALL_LIBGCC):
525 * config/m68k/t-openbsd (MULTILIB_OPTIONS, LIBGCC): Delete.
526 (INSTALL_LIBGCC): Delete.
527 (M68K_MLIB_DIRNAMES, M68K_MLIB_OPTIONS): Define.
528 * config/m68k/t-rtems (MULTILIB_OPTIONS, MULTILIB_DIRNAMES)
529 (MULTILIB_MATCHES, MULTILIB_EXCEPTIONS): Delete.
530 (M68K_MLIB_CPU): Define.
531 * config/m68k/t-uclinux (MULTILIB_OPTIONS, MULTILIB_DIRNAMES)
532 (MULTILIB_MATCHES, MULTILIB_EXCEPTIONS): Delete.
533 (M68K_MLIB_CPU, M68K_MLIB_OPTIONS, M68K_MLIB_DIRNAMES): Define.
535 2007-02-09 Zdenek Dvorak <dvorakz@suse.cz>
536 Richard Guenther <rguenther@suse.de>
539 * fold-const.c (fold_comparison): Handle obfuscated comparisons
540 against INT_MIN/INT_MAX.
541 * tree-ssa-loop-ivcanon.c (remove_empty_loop): Print to dump
542 file if a loop is removed.
544 2007-02-09 Joseph Myers <joseph@codesourcery.com>
546 * calls.c (store_one_arg): Pass correct alignment to
547 emit_push_insn for non-BLKmode values.
548 * expr.c (emit_push_insn): If STRICT_ALIGNMENT, copy to an
549 unaligned stack slot via a suitably aligned slot.
551 2007-02-08 DJ Delorie <dj@redhat.com>
553 * config/m32c/m32c.c (m32c_unpend_compare): Add default to silence
555 (legal_subregs): Use unsigned char, make const.
556 (m32c_illegal_subreg_p): Use ARRAY_SIZE. Delete unused variables.
558 2007-02-08 Paul Brook <paul@codesourcery.com>
560 * config/arm/lib1funcs.asm (RETLDM): Pop directly into PC when no
561 special interworking needed.
563 2007-02-08 Harsha Jagasia <harsha.jagasia@amd.com>
565 * config/i386/xmmintrin.h: Make inclusion of emmintrin.h
566 conditional to __SSE2__.
567 (Entries below should have been added to first ChangeLog
568 entry for amdfam10 dated 2007-02-05)
569 * config/i386/emmintrin.h: Generate #error if __SSE2__ is not
571 * config/i386/pmmintrin.h: Generate #error if __SSE3__ is not
573 * config/i386/tmmintrin.h: Generate #error if __SSSE3__ is not
576 2007-02-08 DJ Delorie <dj@redhat.com>
578 * config/m32c/m32c-protos.h (m32c_illegal_subreg_p): New.
579 * config/m32c/m32c.c (legal_subregs): New.
580 (m32c_illegal_subreg_p): New.
581 * config/m32c/predicates.md (m32c_any_operand): Use it to reject
582 unsupported subregs of hard regs.
584 2007-02-08 Jan Hubicka <jh@suse.cz>
586 * tree-cfg.c (bsi_replace): Shortcut when replacing the statement with
587 the same one; always update histograms.
589 2007-02-08 Diego Novillo <dnovillo@redhat.com>
591 * passes.c (init_optimization_passes): Tidy comment.
593 2007-02-08 Roger Sayle <roger@eyesopen.com>
595 * simplify-rtx.c (simplify_unary_operation_1) <POPCOUNT>: We can
596 strip zero_extend, bswap and rotates from POCOUNT's argument.
597 <PARITY>: Likewise, we can strip not, bswap, sign_extend,
598 zero_extend and rotates from PARITY's argument.
599 <BSWAP>: A byte-swap followed by a byte-swap is an identity.
600 (simplify_const_unary_operation) <BSWAP>: Evaluate the byte-swap
601 of an integer constant at compile-time.
603 2007-02-08 Diego Novillo <dnovillo@redhat.com>
606 * tree-flow.h (struct var_ann_d): Remove field 'is_used'.
608 * tree-ssa-alias.c (compute_is_aliased): Remove. Update all
611 * tree-ssa-live.c (remove_unused_locals): Do not remove
612 TREE_ADDRESSABLE variables.
613 * tree-ssa-structalias.c (compute_points_to_sets): Tidy.
614 * tree-ssa-operands.c (add_virtual_operand): Remove argument
615 FOR_CLOBBER. Update all users.
616 If VAR has an associated alias set, add a virtual operand for
617 it if no alias is found to conflict with the memory reference.
619 2007-02-07 Jan Hubicka <jh@suse.cz>
620 Robert Kidd <rkidd@crhc.uiuc.edu>
622 * value-prof.c (visit_hist, free_hist): Return 1 instead of 0.
624 2007-02-07 Ian Lance Taylor <iant@google.com>
626 * lower-subreg.c (simple_move): Reject PARTIAL_INT modes.
628 2007-02-07 Roger Sayle <roger@eyesopen.com>
630 * config/rs6000/rs6000.md (ctz<mode>2, ffs<mode>2, popcount<mode>2,
631 parity<mode>2, smulsi3_highpart, abstf2_internal, allocate_stack,
632 tablejumpdi, movsi_to_cr_one): Remove constraints from
633 define_expand's match_operands.
635 2007-02-07 Roger Sayle <roger@eyesopen.com>
637 * global.c (compute_regsets): Move declatation of "i" inside of
638 #ifdef ELIMINABLE_REGS to avoid unused variable bootstrap failure.
640 2007-02-07 Jakub Jelinek <jakub@redhat.com>
643 * gimplify.c (gimplify_scan_omp_clauses): Remove special casing
644 of INDIRECT_REF <RESULT_DECL>.
646 * config/i386/i386.c (override_options): Set PTA_SSSE3 for core2.
648 2007-02-06 J"orn Rennecke <joern.rennecke@arc.com>
649 Kaz Kojima <kkojima@gcc.gnu.org>
652 * config/sh/sh.c (expand_cbranchdi4): Use scratch register
654 (sh_initialize_trampoline): Add parentheses to avoid a warning.
656 2007-02-06 Zdenek Dvorak <dvorakz@suse.cz>
658 * doc/loop.texi: Document possibility not to perform disambiguation
659 of loops with multiple latches.
660 * cfgloopmanip.c (alp_enum_p): Removed.
661 (add_loop): Handle subloops. Use get_loop_body_with_size.
662 (create_preheader): Do not allow ENTRY_BLOCK_PTR to be preheader.
663 * cfghooks.c (redirect_edge_and_branch_force): Set dominator for
664 the new forwarder block.
665 (make_forwarder_block): Only call new_bb_cbk if it is not NULL.
666 Handle the case latch is NULL.
667 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Avoid cfg modifications
668 when marking loop exits.
669 * ifcvt.c (if_convert): Ditto. Mark loop exits even if cfg cannot
671 * loop-init.c (loop_optimizer_init): Do not modify cfg. Call
672 disambiguate_loops_with_multiple_latches.
673 * tree-cfgcleanup.c (cleanup_tree_cfg_loop): Calculate dominators
674 before fix_loop_structure.
675 * cfgloop.c: Include pointer-set.h and output.h.
676 (canonicalize_loop_headers, HEADER_BLOCK, LATCH_EDGE,
677 update_latch_info, mfb_keep_just, mfb_keep_nonlatch): Removed.
678 (get_loop_latch_edges, find_subloop_latch_edge_by_profile,
679 find_subloop_latch_edge_by_ivs, find_subloop_latch_edge,
680 mfb_redirect_edges_in_set, form_subloop, merge_latch_edges,
681 disambiguate_multiple_latches, get_loop_body_with_size,
682 disambiguate_loops_with_multiple_latches): New functions.
683 (flow_loop_dump): Dump multiple latch edges.
684 (flow_loop_nodes_find): Handle loops with multiple latches.
685 (flow_loops_find): Ditto. Do not call canonicalize_loop_headers.
686 (glb_enum_p): Modified.
687 (get_loop_body): Use get_loop_body_with_size.
688 * cfgloop.h (LOOPS_HAVE_RECORDED_EXITS): New flag.
689 (AVOID_CFG_MODIFICATIONS): New constant.
690 (disambiguate_loops_with_multiple_latches, add_loop,
691 get_loop_body_with_size): Declare.
692 * Makefile.in (cfgloop.o): Add pointer-set.h and output.h.
694 2007-02-06 Seongbae Park <seongbae.park@gmail.com>
697 * global.c (compute_regsets): New function.
698 (global_alloc): Refactored ELIMINABLE_REGSET
699 and NO_GLOBAL_ALLOC_REGS computation out.
700 (rest_of_handle_global_alloc): Call compute_regsets()
701 for non-optimizing case.
703 2007-02-06 Richard Henderson <rth@redhat.com>
705 * config/i386/constraints.md (Y2): Rename from Y.
706 (Yi): New constraint.
707 * config/i386/i386.md (movsi_1, movdi_2, pushdf_nointeger,
708 pushdf_integer, movdf_nointeger, movdf_integer, zero_extendsidi2_32,
709 zero_extendsidi2_rex64, truncxfdf2_mixed): Change Y constraints to Y2.
710 (extendsfdf2_mixed, extendsfdf2_sse, truncdfsf_fast_mixed,
711 truncdfsf_fast_sse, truncdfsf_mixed, fix_truncdfdi_sse,
712 fix_truncdfsi_sse, floatsidf2_mixed, floatsidf2_sse,
713 floatdidf2_mixed, floatdidf2_sse, absnegdf2_mixed,
714 absnegdf2_sse, sse_setccdf, fop_df_comm_mixed, fop_df_comm_sse,
715 fop_df_1_mixed, fop_df_1_sse): Change Y constraints to x.
716 * config/i386/mmx.md (mov<MMXMODEI>_internal_rex64,
717 mov<MMXMODEI>_internal, movv2sf_internal_rex64, movv2sf_internal,
718 vec_extractv2si_1): Change Y constraints to Y2.
719 * config/i386/sse.md (vec_setv4sf_0, vec_concatv2df, vec_dupv4si,
720 vec_dupv2di, sse2_concatv2si, vec_concatv4si_1, vec_concatv2di):
721 Change Y constraints to Y2.
722 (sse2_loadld): Change Y constraints to x.
724 2007-02-06 Roger Sayle <roger@eyesopen.com>
726 * config/rs6000/rs6000.md (popcount<mode>2): Rewrite.
727 (parity<mode>2): New define_expand using rs6000_emit_parity.
728 * config/rs6000/rs6000.c (rs6000_emit_popcount,
729 rs6000_emit_parity): New functions.
730 * config/rs6000/rs6000-protos.h (rs6000_emit_popcount,
731 rs6000_emit_parity): Prototype here.
733 2007-02-06 Ian Lance Taylor <iant@google.com>
735 * lower-subreg.c (simple_move_operand): Reject CONST.
736 (resolve_clobber): Call validate_change rather than directly
737 assigning to XEXP (pat, 0).
739 2006-02-06 Paolo Bonzini <bonzini@gnu.org>
741 * Makefile.in (tree-ssa-loop-ivopts.o): Add pointer-set.h dependency.
742 (tree-ssa-reassoc.o): Add pointer-set.h dependency.
743 (tree-cfg.o): Remove hashtab.h dependency.
745 * tree-ssa-loop-ivopts.c: Include pointer-set.h.
746 (struct ivopts_data): Change niters to pointer_map_t.
747 (struct nfe_cache_elt, nfe_hash, nfe_eq): Delete.
748 (niter_for_exit): Create pointer_map on demand. Change for
750 (tree_ssa_iv_optimize_init): Initialize data->niters to NULL.
751 (free_loop_data): Destroy data->niters if created and reset field.
752 (tree_ssa_iv_optimize_finalize): Don't delete data->niters here.
753 (tree_ssa_iv_optimize_loop): Check for presence of stale data.
755 * tree-ssa-reassoc.c: Include pointer-set.h.
756 (bb_rank): Change to long *.
757 (operand_rank): Change to pointer_map_t.
758 (find_operand_rank): Return long, -1 if not found. Declare as inline.
759 (insert_operand_rank): Accept long.
760 (operand_entry_hash, operand_entry_eq): Remove.
761 (get_rank): Return long. Adjust for changes above.
762 (init_reassoc): Change rank type to long. Adjust creation of bb_rank
764 (fini_reassoc): Delete operand_rank with pointer_map_destroy.
766 * tree-ssa-structalias.c (vi_for_tree): Change to pointer_map.
767 (struct tree_vi, tree_vi_t, tree_vi_hash, tree_vi_eq): Delete.
768 (insert_vi_for_tree): Rewrite for pointer_map API. Assert argument
770 (lookup_vi_for_tree): Rewrite for pointer_map API. Return varinfo_t
771 directly since it cannot be NULL.
772 (get_vi_for_tree): Rewrite for pointer_map API.
773 (find_what_p_points_to): Adjust for change to lookup_vi_for_tree.
774 (init_alias_vars): Create vi_for_tree as pointer_map.
775 (delete_points_to_sets): Delete vi_for_tree using pointer_map_destroy.
777 * tree-cfg.c: Don't include hashtab.h.
778 (edge_to_cases): Declare as pointer_map.
779 (struct edge_to_cases_elt, edge_to_cases_hash, edge_to_cases_eq):
781 (edge_to_cases_cleanup): Rewrite as pointer_map_traverse callback.
782 (start_recording_case_labels): Create edge_to_cases as pointer_map.
783 (end_recoding_case_labels): Cleanup edge_to_cases manually before
785 (record_switch_edge): Delete.
786 (get_cases_for_edge): Adjust for pointer_map API, inline
787 record_switch_edge (rewritten for new API), remove goto.
789 2006-02-06 Paolo Bonzini <bonzini@gnu.org>
791 * Makefile.in (tree-nested.o): Add pointer-set.h dependency.
792 * tree-nested.c: Include pointer-set.h.
793 (var_map_elt, var_map_eq, var_map_hash): Delete.
794 (struct nesting_info): Remove GTY marker. Change the two htab_t's
796 (nesting_info_bitmap_obstack): New.
797 (lookup_field_for_decl): Adjust for pointer_map API.
798 (lookup_tramp_for_decl): Adjust for pointer_map API.
799 (get_nonlocal_debug_decl): Adjust for pointer_map API.
800 (get_local_debug_decl): Adjust for pointer_map API.
801 (convert_nl_goto_reference): Adjust for pointer_map API.
802 (convert_nl_goto_receiver): Adjust for pointer_map API.
803 (create_nesting_tree): Create outside GGC space. Create bitmap on
804 the new obstack. Create field_map and var_map as pointer_maps.
805 (free_nesting_tree): Adjust for changes to create_nesting_tree.
807 (lower_nested_functions): Move root here, no need to NULL it.
808 Initialize and release the obstack.
810 2007-02-06 Paolo Bonzini <bonzini@gnu.org>
812 * tree.c (tree_int_map_hash, tree_int_map_eq, tree_int_map_marked_p):
813 Remove prototypes and make them non-static.
814 (struct tree_int_map): Remove.
815 * tree.h (struct tree_int_map): Move here, turning TO into an
817 (tree_int_map_hash, tree_int_map_eq, tree_int_map_marked_p): Declare.
819 * tree.h (TREE_COMPLEXITY): Remove.
820 (struct tree_exp): Remove complexity field.
821 * tree.c (build1_stat): Don't set it.
823 2007-02-06 Dorit Nuzman <dorit@il.ibm.com>
824 Victor Kaplansky <victork@il.ibm.com>
826 * tree-vectorizer.c (vect_is_simple_use): Support induction.
827 (vect_is_simple_reduction): Support reduction with induction as
829 (vect_is_simple_iv_evolution): Fix formatting.
830 * tree-vect-analyze.c (vect_mark_stmts_to_be_vectorized): Fix
831 formatting. Don't mark induction phis for vectorization.
832 (vect_analyze_scalar_cycles): Analyze all inductions, then reductions.
833 * tree-vect-transform.c (get_initial_def_for_induction): New function.
834 (vect_get_vec_def_for_operand): Support induction.
835 (vect_get_vec_def_for_stmt_copy): Fix formatting and add check for
837 (vectorizable_reduction): Support reduction with induction as one of
839 (vectorizable_type_demotion): Use def-type of stmt argument rather
842 * tree-ssa-loop.c (gate_scev_const_prop): Return the value of
843 flag_tree_scev_cprop.
844 * common.opt (tree-scev-cprop): New flag.
846 * tree-vect-transform.c (vect_create_destination_var): Use 'kind' in
847 call to vect_get_new_vect_var.
849 2007-02-06 Ira Rosen <irar@il.ibm.com>
851 * tree-vect-patterns.c (vect_recog_widen_mult_pattern): Check that
853 (vect_pattern_recog_1): Likewise.
855 2007-02-05 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
857 * fold-const.c (negate_expr_p): Handle CONJ_EXPR.
858 (fold_negate_expr): Likewise.
860 2007-02-05 Alexandre Oliva <aoliva@redhat.com>
863 * dwarf2out.c (modified_type_die): Follow DECL_ORIGINAL_TYPE
864 even if cv-qualification is the same.
866 2007-02-05 Geoffrey Keating <geoffk@apple.com>
868 * config/rs6000/darwin-tramp.asm (__trampoline_setup): Call
869 __enable_execute_stack on completion.
871 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
873 * config/i386/athlon.md (athlon_fldxf_k8, athlon_fld_k8,
874 athlon_fstxf_k8, athlon_fst_k8, athlon_fist, athlon_fmov,
875 athlon_fadd_load, athlon_fadd_load_k8, athlon_fadd, athlon_fmul,
876 athlon_fmul_load, athlon_fmul_load_k8, athlon_fsgn,
877 athlon_fdiv_load, athlon_fdiv_load_k8, athlon_fdiv_k8,
878 athlon_fpspc_load, athlon_fpspc, athlon_fcmov_load,
879 athlon_fcmov_load_k8, athlon_fcmov_k8, athlon_fcomi_load_k8,
880 athlon_fcomi, athlon_fcom_load_k8, athlon_fcom): Added amdfam10.
882 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
884 * config/i386/i386.md (x86_sahf_1, cmpfp_i_mixed, cmpfp_i_sse,
885 cmpfp_i_i387, cmpfp_iu_mixed, cmpfp_iu_sse, cmpfp_iu_387,
886 swapsi, swaphi_1, swapqi_1, swapdi_rex64, fix_truncsfdi_sse,
887 fix_truncdfdi_sse, fix_truncsfsi_sse, fix_truncdfsi_sse,
888 x86_fldcw_1, floatsisf2_mixed, floatsisf2_sse, floatdisf2_mixed,
889 floatdisf2_sse, floatsidf2_mixed, floatsidf2_sse,
890 floatdidf2_mixed, floatdidf2_sse, muldi3_1_rex64, mulsi3_1,
891 mulsi3_1_zext, mulhi3_1, mulqi3_1, umulqihi3_1, mulqihi3_insn,
892 umulditi3_insn, umulsidi3_insn, mulditi3_insn, mulsidi3_insn,
893 umuldi3_highpart_rex64, umulsi3_highpart_insn,
894 umulsi3_highpart_zext, smuldi3_highpart_rex64,
895 smulsi3_highpart_insn, smulsi3_highpart_zext, x86_64_shld,
896 x86_shld_1, x86_64_shrd, sqrtsf2_mixed, sqrtsf2_sse,
897 sqrtsf2_i387, sqrtdf2_mixed, sqrtdf2_sse, sqrtdf2_i387,
898 sqrtextendsfdf2_i387, sqrtxf2, sqrtextendsfxf2_i387,
899 sqrtextenddfxf2_i387): Added amdfam10_decode.
901 * config/i386/athlon.md (athlon_idirect_amdfam10,
902 athlon_ivector_amdfam10, athlon_idirect_load_amdfam10,
903 athlon_ivector_load_amdfam10, athlon_idirect_both_amdfam10,
904 athlon_ivector_both_amdfam10, athlon_idirect_store_amdfam10,
905 athlon_ivector_store_amdfam10): New define_insn_reservation.
906 (athlon_idirect_loadmov, athlon_idirect_movstore): Added
909 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
911 * config/i386/athlon.md (athlon_call_amdfam10,
912 athlon_pop_amdfam10, athlon_lea_amdfam10): New
913 define_insn_reservation.
914 (athlon_branch, athlon_push, athlon_leave_k8, athlon_imul_k8,
915 athlon_imul_k8_DI, athlon_imul_mem_k8, athlon_imul_mem_k8_DI,
916 athlon_idiv, athlon_idiv_mem, athlon_str): Added amdfam10.
918 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
920 * config/i386/athlon.md (athlon_sseld_amdfam10,
921 athlon_mmxld_amdfam10, athlon_ssest_amdfam10,
922 athlon_mmxssest_short_amdfam10): New define_insn_reservation.
924 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
926 * config/i386/athlon.md (athlon_sseins_amdfam10): New
927 define_insn_reservation.
928 * config/i386/i386.md (sseins): Added sseins to define_attr type
929 and define_attr unit.
930 * config/i386/sse.md: Set type attribute to sseins for insertq
933 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
935 * config/i386/athlon.md (sselog_load_amdfam10, sselog_amdfam10,
936 ssecmpvector_load_amdfam10, ssecmpvector_amdfam10,
937 ssecomi_load_amdfam10, ssecomi_amdfam10,
938 sseaddvector_load_amdfam10, sseaddvector_amdfam10): New
939 define_insn_reservation.
940 (ssecmp_load_k8, ssecmp, sseadd_load_k8, seadd): Added amdfam10.
942 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
944 * config/i386/athlon.md (cvtss2sd_load_amdfam10,
945 cvtss2sd_amdfam10, cvtps2pd_load_amdfam10, cvtps2pd_amdfam10,
946 cvtsi2sd_load_amdfam10, cvtsi2ss_load_amdfam10,
947 cvtsi2sd_amdfam10, cvtsi2ss_amdfam10, cvtsd2ss_load_amdfam10,
948 cvtsd2ss_amdfam10, cvtpd2ps_load_amdfam10, cvtpd2ps_amdfam10,
949 cvtsX2si_load_amdfam10, cvtsX2si_amdfam10): New
950 define_insn_reservation.
952 * config/i386/sse.md (cvtsi2ss, cvtsi2ssq, cvtss2si,
953 cvtss2siq, cvttss2si, cvttss2siq, cvtsi2sd, cvtsi2sdq,
954 cvtsd2si, cvtsd2siq, cvttsd2si, cvttsd2siq,
955 cvtpd2dq, cvttpd2dq, cvtsd2ss, cvtss2sd,
956 cvtpd2ps, cvtps2pd): Added amdfam10_decode attribute.
958 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
960 * config/i386/athlon.md (athlon_ssedivvector_amdfam10,
961 athlon_ssedivvector_load_amdfam10, athlon_ssemulvector_amdfam10,
962 athlon_ssemulvector_load_amdfam10): New define_insn_reservation.
963 (athlon_ssediv, athlon_ssediv_load_k8, athlon_ssemul,
964 athlon_ssemul_load_k8): Added amdfam10.
966 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
968 * config/i386/i386.h (TARGET_SSE_UNALIGNED_MOVE_OPTIMAL): New macro.
969 (x86_sse_unaligned_move_optimal): New variable.
971 * config/i386/i386.c (x86_sse_unaligned_move_optimal): Enable for
973 (ix86_expand_vector_move_misalign): Add code to generate movupd/movups
974 for unaligned vector SSE double/single precision loads for AMDFAM10.
976 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
978 * config/i386/i386.h (TARGET_AMDFAM10): New macro.
979 (TARGET_CPU_CPP_BUILTINS): Add code for amdfam10.
980 Define TARGET_CPU_DEFAULT_amdfam10.
981 (TARGET_CPU_DEFAULT_NAMES): Add amdfam10.
982 (processor_type): Add PROCESSOR_AMDFAM10.
984 * config/i386/i386.md: Add amdfam10 as a new cpu attribute to match
985 processor_type in config/i386/i386.h.
986 Enable imul peepholes for TARGET_AMDFAM10.
988 * config.gcc: Add support for --with-cpu option for amdfam10.
990 * config/i386/i386.c (amdfam10_cost): New variable.
991 (m_AMDFAM10): New macro.
992 (m_ATHLON_K8_AMDFAM10): New macro.
993 (x86_use_leave, x86_push_memory, x86_movx, x86_unroll_strlen,
994 x86_cmove, x86_3dnow_a, x86_deep_branch, x86_use_simode_fiop,
995 x86_promote_QImode, x86_integer_DFmode_moves,
996 x86_partial_reg_dependency, x86_memory_mismatch_stall,
997 x86_accumulate_outgoing_args, x86_arch_always_fancy_math_387,
998 x86_sse_partial_reg_dependency, x86_sse_typeless_stores,
999 x86_use_ffreep, x86_use_incdec, x86_four_jump_limit,
1000 x86_schedule, x86_use_bt, x86_cmpxchg16b, x86_pad_returns):
1001 Enable/disable for amdfam10.
1002 (override_options): Add amdfam10_cost to processor_target_table.
1003 Set up PROCESSOR_AMDFAM10 for amdfam10 entry in
1004 processor_alias_table.
1005 (ix86_issue_rate): Add PROCESSOR_AMDFAM10.
1006 (ix86_adjust_cost): Add code for amdfam10.
1008 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
1010 * config/i386/i386.opt: Add new Advanced Bit Manipulation (-mabm)
1011 instruction set feature flag. Add new (-mpopcnt) flag for popcnt
1012 instruction. Add new SSE4A (-msse4a) instruction set feature flag.
1013 * config/i386/i386.h: Add builtin definition for SSE4A.
1014 * config/i386/i386.md: Add support for ABM instructions
1016 * config/i386/sse.md: Add support for SSE4A instructions
1017 (movntss, movntsd, extrq, insertq).
1018 * config/i386/i386.c: Add support for ABM and SSE4A builtins.
1019 Add -march=amdfam10 flag.
1020 * config/i386/ammintrin.h: Add support for SSE4A intrinsics.
1021 * doc/invoke.texi: Add documentation on flags for sse4a, abm, popcnt
1023 * doc/extend.texi: Add documentation for SSE4A builtins.
1025 2007-02-05 Bob Wilson <bob.wilson@acm.org>
1027 * config/xtensa/xtensa.c (constantpool_mem_p): Skip over SUBREGs.
1029 2007-02-05 Richard Guenther <rguenther@suse.de>
1031 * tree-vectorizer.h (vectorizable_function): Add argument type
1032 argument, change return type.
1033 * tree-vect-patterns.c (vect_recog_pow_pattern): Adjust caller.
1034 * tree-vect-transform.c (vectorizable_function): Handle extra
1035 argument, return vectorized function decl.
1036 (build_vectorized_function_call): Remove.
1037 (vectorizable_call): Handle calls with result and argument types
1038 differing. Handle loop vectorization factor correctly.
1039 * targhooks.c (default_builtin_vectorized_function): Adjust for
1041 * targhooks.h (default_builtin_vectorized_function): Likewise.
1042 * target.h (builtin_vectorized_function): Add argument type
1044 * config/i386/i386.c (ix86_builtin_vectorized_function): Handle
1045 extra argument, allow vectorizing of lrintf.
1046 * doc/tm.texi (TARGET_VECTORIZE_BUILTIN_VECTORIZED_FUNCTION): Adjust
1047 documentation of target hook.
1049 2007-02-05 Hans-Peter Nilsson <hp@axis.com>
1052 * config/cris/cris.md ("*andsi_movu", "*andsi_clear", "*andhi_movu")
1053 ("*andhi_clear", andu (casesi+45)): For size-changed operand where
1054 memory is allowed, require !side_effects_p, not just !MEM_VOLATILE_P.
1056 2007-02-05 Roger Sayle <roger@eyesopen.com>
1058 * fold-const.c (fold_unary) <REAL_PART>: Test for availability of
1059 BUILT_IN_COS before simplifying REAL_PART(CEXPI)) to COS.
1060 <IMAG_PART>: Likewise, check for availability of BUILT_IN_SIN.
1061 * builtins.c (fold_builtin_sincos): Check for TARGET_C99_FUNCTIONS
1062 before canonicalizing sincos to cexpi.
1063 (fold_builtin_cexp): Likewise, for canonicalizing cexp to cexpi.
1065 2007-02-05 Roger Sayle <roger@eyesopen.com>
1067 * config/alpha/alpha.c (alpha_add_builtins): New Helper function.
1068 Set TREE_READONLY and TREE_NOTHROW directly, not via attributes.
1069 (alpha_init_builtins): Use alpha_add_builtins to process tables.
1071 2007-02-05 Roger Sayle <roger@eyesopen.com>
1073 * mips-tfile.c (initialize_init_file): Correct endianness test.
1075 2007-02-05 Kazu Hirata <kazu@codesourcery.com>
1077 * config/m68k/m68k.md (pushdi-1, pushdi, movsi+1): Don't use
1080 2007-02-05 Richard Sandiford <richard@codesourcery.com>
1082 * dwarf2out.c (dwarf2out_frame_debug_expr): Record the register
1083 saves in a PARALLEL before the register assignments.
1085 2007-02-05 Richard Sandiford <richard@codesourcery.com>
1087 * doc/tm.texi (DWARF_ALT_FRAME_RETURN_COLUMN): Do not require
1088 DWARF_FRAME_RETURN_COLUMN to be a general register.
1089 * dwarf2out.c (init_return_column_size): New function, split from...
1090 (expand_builtin_init_dwarf_reg_sizes): ...here. Allow both
1091 DWARF_FRAME_RETURN_COLUMN and DWARF_ALT_FRAME_RETURN_COLUMN
1092 to be nongeneral registers.
1093 * config/m68k/m68k.h (DWARF_FRAME_REGNUM): Only map FP and
1095 (DWARF_FRAME_REGISTERS, DWARF_FRAME_RETURN_COLUMN): Define.
1096 (DWARF_ALT_FRAME_RETURN_COLUMN): Define.
1098 2007-02-04 Zdenek Dvorak <dvorakz@suse.cz>
1100 * cfgcleanup.c (try_optimize_cfg): Avoid removing ENTRY_BLOCK_PTR.
1102 2007-02-04 Zdenek Dvorak <dvorakz@suse.cz>
1104 * cfgloopmanip.c (loop_delete_branch_edge): Removed.
1105 (remove_path): Use can_remove_branch_p and remove_branch instead
1106 of loop_delete_branch_edge.
1107 * tree-ssa-loop-manip.c (scale_dominated_blocks_in_loop): New function.
1108 (tree_transform_and_unroll_loop): Remove dead branches immediately.
1109 Update profile using scale_dominated_blocks_in_loop.
1110 * cfghooks.c (can_remove_branch_p, remove_branch): New functions.
1111 * cfghooks.h (struct cfg_hooks): Add can_remove_branch_p.
1112 (can_remove_branch_p, remove_branch): Declare.
1113 * tree-cfg.c (tree_can_remove_branch_p): New function.
1114 (tree_cfg_hooks): Add tree_can_remove_branch_p.
1115 * cfgrtl.c (rtl_can_remove_branch_p): New function.
1116 (rtl_cfg_hooks, cfg_layout_rtl_cfg_hook): Add rtl_can_remove_branch_p.
1118 2007-02-05 Jan Hubicka <jh@suse.cz>
1121 * ipa-inline.c (cgraph_clone_inlined_nodes): When there are unanalyzed
1122 nodes in cgraph, don't remove offline copy of the function.
1124 2007-02-04 Jan Hubicka <jh@suse.cz>
1126 * tree-sra.c (sra_walk_expr): Add linebreaks. BITFIELD_REFs into
1127 vectors might cause maybe_lookup_element_for_expr to be called
1128 on non-sra-candidate.
1130 2007-02-04 Kazu Hirata <kazu@codesourcery.com>
1132 * config/bfin/bfin-modes.def, config/bfin/bfin.c,
1133 config/bfin/bfin.md, config/bfin/predicates.md: Follow
1134 spelling conventions.
1136 2007-02-04 Richard Guenther <rguenther@suse.de>
1139 * fold-const.c (try_move_mult_to_index): Make sure to not
1140 overflow one dimension of a multi-dimensional array access.
1142 2007-02-04 Jan Hubicka <jh@suse.cz>
1144 * passes.c (init_optimization_passes): Reindent.
1146 2007-02-04 Jan Hubicka <jh@suse.cz>
1147 Eric Botcazou <ebotcazou@adacore.com>
1149 * tree-optimize.c (has_abnormal_outgoing_edge_p): Move to...
1150 (execute_fixup_cfg): Break out the abnormal goto code.
1151 * tree-inline.c (has_abnormal_outgoing_edge_p): ...here.
1152 (make_nonlocal_label_edges): Move here from execute_fixup_cfg.
1153 (optimize_inline_calls): Call make_nonlocal_label_edges.
1155 2007-02-04 Jan Hubicka <jh@suse.cz>
1157 * tree-ssa-copyrename.c (copy_rename_partition_coalesce): Return
1158 true when something was changed.
1159 (rename_ssa_copies): When something was changed, do
1160 TODO_remove_unused_locals.
1161 * tree-ssa-forwprop.c (tree_ssa_forward_propagate_single_use_value):
1162 add TODO_remove_unused_locals when instruction was removed.
1164 2007-02-04 Jan Hubicka <jh@suse.cz>
1166 * ipa-inline.c (try_inline): Improve debug output; work on already
1168 (cgraph_decide_inlining_incrementally): Indent; improve debug output;
1169 call try_inline for already inlined edges too when flattening;
1170 inline also functions that make callee growth but overall unit size
1173 2007-02-04 Kazu Hirata <kazu@codesourcery.com>
1175 * config/m32c/bitops.md, config/m32c/jump.md,
1176 config/m32c/m32c.c, config/m32c/m32c.h, config/m32r/m32r.c,
1177 config/m32r/m32r.h, config/m32r/m32r.md,
1178 config/m32r/predicates.md, config/m68hc11/larith.asm,
1179 config/m68hc11/m68hc11.c, config/m68hc11/m68hc11.h,
1180 config/m68k/m68k.h, config/mcore/mcore.md, config/mips/4k.md,
1181 config/mips/mips-protos.h, config/mips/mips.c,
1182 config/mips/mips.h, config/mips/mips.md, config/mips/mips16.S,
1183 config/mn10300/mn10300.h, config/mn10300/predicates.md,
1184 config/mt/mt.c, config/mt/mt.h, config/mt/mt.md: Follow
1185 spelling conventions.
1187 * config/v850/v850.c, config/v850/v850.h, config/v850/v850.md:
1188 Follow spelling conventions.
1190 2007-02-03 Douglas Gregor <doug.gregor@gmail.com>
1192 * c-opts.c (c_common_post_options): If C++0x mode is enabled, don't
1193 warn about C++0x compatibility.
1195 2007-02-04 Kazu Hirata <kazu@codesourcery.com>
1197 * config/h8300/h8300.c, config/h8300/h8300.h,
1198 config/h8300/h8300.md: Follow spelling conventions.
1200 2007-02-03 Uros Bizjak <ubizjak@gmail.com>
1203 * combine.c (try_combine): Do not substitute source operand
1204 with constants wider than 2 * HOST_BITS_PER_WIDE_INT.
1206 2007-02-03 Jan Hubicka <jh@suse.cz>
1208 PR gcov-profile/30650
1209 * value-prof.c (stringop_block_profile): Fix handling of size counter;
1210 do not divide by zero for never executed counters.
1211 (tree_find_values_to_profile): Fix counters.
1212 * gcov-ui.h (GCOV_COUNTER_AVERAGE, GCOV_COUNTER_IOR): Fix comments.
1214 2007-02-03 Ian Lance Taylor <iant@google.com>
1216 * lower-subreg.c (simple_move_operand): New static function,
1217 broken out of simple_move. Reject LABEL_REF, SYMBOL_REF, and HIGH
1219 (simple_move): Call simple_move_operand.
1220 (find_decomposable_subregs): Add special handling of MEMs.
1221 (can_decompose_p): Rename from cannot_decompose_p. Reverse
1222 meaning of return value. If we see a hard register, test whether
1223 it can store a word_mode value. Change all callers.
1225 2007-02-03 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
1227 * pa.md (addvdi3, addvsi3, subvdi3, subvsi3, negvdi2, negvsi2): New
1228 ftrapv insns and expanders.
1229 (subdi3): Change define_expand operand 1 to arith11_operand, and
1230 operand 2 to reg_or_0_operand. Change constraints of 64-bit insn
1231 pattern to handle reg_or_0 operands. Revise 32-bit insn pattern to
1232 handle 11-bit constants and reg_or_0 operands in operands 1 and 2,
1236 * varasm.c (notice_global_symbol): Treat global objects as weak when
1239 2007-02-03 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1241 * emit-rtl.c (dconstpi): Delete.
1243 (init_emit_once): Delete dconstpi and init dconstsqrt2.
1244 * real.h (dconstpi): Delete.
1246 * builtins.c (fold_builtin_cabs): Use dconstsqrt2.
1247 (fold_builtin_hypot): Likewise.
1249 2007-02-03 Tom Tromey <tromey@redhat.com>
1252 * gcc.c (cpp_unique_options): Any of -ggdb3, -gstabs3,
1253 -gcoff3, -gxcoff3, -gvms3 implies -dD.
1255 2007-02-03 Kazu Hirata <kazu@codesourcery.com>
1257 * c-decl.c, config/avr/avr.c, config/avr/avr.h,
1258 config/m68k/m68k.c, config/m68k/netbsd-elf.h,
1259 config/mn10300/mn10300.c, config/pdp11/pdp11.h,
1260 config/rs6000/cell.md, config/rs6000/darwin.h,
1261 config/sh/sh.md, config/sh/sh4-300.md, config/spu/spu.c,
1262 config/spu/spu.md, cselib.c, expr.c, haifa-sched.c, hwint.h,
1263 jump.c, reload.c, sched-deps.c, sched-int.h, tree-inline.c,
1264 tree-profile.c, tree-ssa-live.h, tree-vrp.c: Fix comment
1265 typos. Follow spelling conventions.
1266 * doc/invoke.texi: Follow spelling conventions.
1268 2007-02-03 Roger Sayle <roger@eyesopen.com>
1270 * simplify-rtx.c (simplify_relational_operation_1): Implement some
1271 canonicalization transformations that attempt to simplify integer
1272 constant comparisons to become comparisons against zero.
1274 2007-02-03 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1276 * builtins.c (fold_builtin_cabs): Fold cabs(x+xi) into
1278 * fold-const.c (fold_binary): Fix comment typos. Fold complex
1279 (x,0)-(0,y) into (x,-y). Likewise (0,y)-(x,0) into (-x,y).
1281 2007-02-02 Mike Stump <mrs@apple.com>
1283 * config/darwin9.h (DARWIN_LINKER_GENERATES_ISLANDS): Add.
1284 * config/rs6000/rs6000.c (DARWIN_GENERATE_ISLANDS): Add.
1285 (output_call): Use DARWIN_GENERATE_ISLANDS to decide when to
1286 generate a branch island.
1288 2007-02-02 Bob Wilson <bob.wilson@acm.org>
1290 * config/xtensa/xtensa.c (smalloffset_mem_p): Use BASE_REG_P.
1291 (xtensa_legitimate_address_p): New.
1292 (xtensa_legitimize_address): New.
1293 (xtensa_output_addr_const_extra): New.
1294 * config/xtensa/xtensa.h (REG_OK_STRICT_FLAG): Define.
1296 (REG_OK_FOR_BASE_P): Use BASE_REG_P.
1297 (GO_IF_LEGITIMATE_ADDRESS): Move code to xtensa_legitimate_address_p.
1298 (LEGITIMIZE_ADDRESS): Move code to xtensa_legitimize_address.
1299 (OUTPUT_ADDR_CONST_EXTRA): Move code to xtensa_output_addr_const_extra.
1300 * config/xtensa/xtensa-protos.h (xtensa_legitimate_address_p): New.
1301 (xtensa_legitimize_address): New.
1302 (xtensa_output_addr_const_extra): New.
1304 2007-02-02 Steve Ellcey <sje@cup.hp.com>
1306 * config/ia64/ia64.c (ia64_print_operand): Fix compare strings.
1308 2007-02-02 Ian Lance Taylor <iant@google.com>
1310 * expmed.c (expand_divmod): Add comment.
1312 2007-02-02 Kazu Hirata <kazu@codesourcery.com>
1314 * emit-rtl.c (renumber_insns): Remove.
1315 * flags.h: Remove the extern for flag_renumber_insns.
1316 * rtl.h: Remove the prototype for renumber_insns.
1317 * toplev.c (flag_renumber_insns): Remove.
1319 2007-02-02 Hui-May Chang <hm.chang@apple.com>
1321 Revert for x86 darwin:
1322 2005-06-19 Uros Bizjak <uros@kss-loka.si>
1324 * config/i386/i386.c (ix86_function_arg_regno_p): Put back the
1325 code before the following patch under TARGET_MACHO.
1326 (ix86_function_value_regno_p): Likewise.
1328 2007-02-02 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1330 * fold-const.c (negate_expr_p, fold_negate_expr): Handle
1333 2007-02-02 Maxim Kuvyrkov <mkuvyrkov@ispras.ru>
1336 * config/ia64/ia64.c (ia64_speculate_insn): Restrict to memory loads to
1337 general or fp registers. Add comments.
1338 * config/ia64/ia64.md (reg_pred_prefix): Add comment.
1340 2007-02-02 Paolo Bonzini <bonzini@gnu.org>
1342 * pointer-set.c (insert_aux): Only return insertion slot.
1343 (pointer_set_insert): Adjust.
1344 (pointer_set_traverse, struct pointer_map_t, pointer_map_create,
1345 pointer_map_destroy, pointer_map_insert, pointer_map_contains,
1346 pointer_map_traverse): New.
1347 * pointer-set.h (pointer_set_traverse, struct pointer_map_t,
1348 pointer_map_create, pointer_map_destroy, pointer_map_insert,
1349 pointer_map_contains, pointer_map_traverse): Declare.
1351 2007-02-02 Jakub Jelinek <jakub@redhat.com>
1354 * builtins.c (fold_builtin_sprintf): Do not attempt to optimize
1355 sprintf (str, "%s"). Do not optimize sprintf (str, "nopercent", p++).
1357 2007-02-02 Maxim Kuvyrkov <mkuvyrkov@ispras.ru>
1359 * sched-int.h (ds_to_dk, dk_to_ds): Declare functions.
1361 (struct _dep): New type.
1362 (dep_t): New typedef.
1363 (DEP_PRO, DEP_CON, DEP_KIND): New access macros.
1364 (DEP_STATUS): New access macro. The macro with the same name was
1365 renamed to DEP_LINK_STATUS.
1366 (dep_init): Declare function
1368 (struct _dep_link): New type.
1369 (dep_link_t): New typedef.
1370 (DEP_LINK_NODE, DEP_LINK_NEXT, DEP_LINK_PREV_NEXTP): New access macros.
1371 (DEP_LINK_DEP, DEP_LINK_PRO, DEP_LINK_CON, DEP_LINK_KIND): New macros.
1372 (DEP_LINK_STATUS): New macro.
1373 (debug_dep_links): New debug function.
1375 (struct _deps_list): New type.
1376 (deps_list_t): New typedef.
1377 (DEPS_LIST_FIRST): New access macro.
1378 (FOR_EACH_DEP_LINK): New cycle macro.
1379 (create_deps_list, free_deps_list, delete_deps_list): Declare
1381 (deps_list_empty_p, debug_deps_list, add_back_dep_to_deps_list): Ditto.
1382 (find_link_by_pro_in_deps_list, find_link_by_con_in_deps_list): Ditto.
1383 (copy_deps_list_change_con): Ditto.
1385 (move_dep_link): Declare function.
1387 (struct _dep_node): New type.
1388 (dep_node_t): New typedef.
1389 (DEP_NODE_BACK, DEP_NODE_DEP, DEP_NODE_FORW): New access macros.
1391 (struct haifa_insn_data.back_deps): New field to hold backward
1392 dependencies of the insn.
1393 (struct haifa_insn_data.depend): Rename to forw_deps. Change its type
1395 (struct haifa_insn_data.resolved_deps): Rename to resolved_back_deps.
1396 Change its type to deps_list_t.
1397 (INSN_BACK_DEPS): New access macro to use instead of LOG_LINKS.
1398 (INSN_DEPEND): Rename to INSN_FORW_DEPS.
1399 (RESOLVED_DEPS): Rename to INSN_RESOLVED_BACK_DEPS.
1401 (INSN_COST): Move to haifa-sched.c. Use insn_cost () instead.
1403 (DEP_STATUS): Rename to DEP_LINK_STATUS. Fix typo in the comment.
1405 (add_forw_dep, delete_back_forw_dep, insn_cost): Update declaration and
1407 (dep_cost): Declare.
1409 * sched-deps.c (CHECK): New macro to (en/dis)able sanity checks.
1410 (ds_to_dk, dk_to_ds): New functions.
1412 (init_dep_1): New static function.
1413 (init_dep): New function.
1414 (copy_dep): New static function.
1416 (dep_link_consistent_p, attach_dep_link, add_to_deps_list): New static
1418 (detach_dep_link): New static function.
1419 (move_dep_link): New function.
1421 (dep_links_consistent_p, dump_dep_links): New static functions.
1422 (debug_dep_links): New debugging function.
1424 (deps_obstack, dl_obstack, dn_obstack): New static variables.
1426 (alloc_deps_list, init_deps_list): New static functions.
1427 (create_deps_list): New function.
1428 (clear_deps_list): New static function.
1429 (free_deps_list, delete_deps_list, deps_list_empty_p): New functions.
1430 (deps_list_consistent_p, dump_deps_list): New static functions.
1431 (debug_deps_list): New function.
1432 (add_back_dep_to_deps_list, find_link_by_pro_in_deps_list): New
1434 (find_link_by_con_in_deps_list, copy_deps_list_change_con): Ditto.
1436 (maybe_add_or_update_back_dep_1, add_or_update_back_dep_1): Update to
1437 use new scheduler dependencies lists.
1438 (add_back_dep, delete_all_dependences, fixup_sched_groups): Ditto.
1439 (sched_analyze): Ditto. Initialize dependencies lists.
1440 (add_forw_dep, compute_forward_dependences): Update to use new
1441 scheduler dependencies lists.
1443 (init_dependency_caches): Init deps_obstack.
1444 (free_dependency_caches): Free deps_obstack.
1446 (adjust_add_sorted_back_dep, adjust_back_add_forw_dep): Update to use
1447 new scheduler dependencies lists.
1448 (delete_forw_dep, add_or_update_back_forw_dep): Ditto.
1449 (add_back_forw_dep, delete_back_forw_dep): Ditto.
1451 * sched-rgn.c (set_spec_fed, find_conditional_protection, is_pfree):
1452 Update to use new scheduler dependencies lists.
1453 (is_conditionally_protected, is_prisky, add_branch_dependences): Ditto.
1454 (debug_dependencies): Ditto.
1455 (schedule_region): Update comments.
1457 * sched-ebb.c (earliest_block_with_similiar_load): Update to use new
1458 scheduler dependencies lists.
1459 (schedule_ebb): Update comments.
1461 * rtl.def (DEPS_LIST): Remove.
1463 * lists.c (unused_deps_list): Remove.
1464 (free_list): Update assertions.
1466 (alloc_DEPS_LIST, free_DEPS_LIST_list, free_DEPS_LIST_node): Remove.
1467 (remove_free_DEPS_LIST_elem, copy_DEPS_LIST_list): Ditto.
1469 * rtl.h (free_DEPS_LIST_list, alloc_DEPS_LIST): Remove declarations.
1470 (remove_free_DEPS_LIST_elem, copy_DEPS_LIST_list): Ditto.
1472 * haifa-sched.c (comments): Update.
1473 (insn_cost1): Remove. Inline the code into insn_cost ().
1474 (insn_cost): Update to use new scheduler dependencies lists. Move
1475 processing of the dependency cost to dep_cost ().
1476 (dep_cost): New function. Use it instead of insn_cost () when
1477 evaluating cost of the dependency. Use compatible interface to
1478 interact with the target.
1479 (priority): Update to use new scheduler dependencies lists.
1480 (rank_for_schedule): Ditto. Optimize heuristic that prefers the insn
1481 with greater number of insns that depend on the insn.
1482 (schedule_insn): Update to use new scheduler dependencies lists. Add
1483 code to free backward dependencies lists. Inline and optimize code
1484 from resolve_dep () - see PR28071.
1485 (ok_for_early_queue_removal): Update to use new scheduler dependencies
1486 lists. Update call to targetm.sched.is_costly_dependence hook.
1488 (fix_inter_tick, try_ready, fix_tick_ready): Update to use new
1489 scheduler dependencies lists.
1491 (resolve_dep): Remove. Move the logic to schedule_insn ().
1492 (init_h_i_d): Initialize dependencies lists.
1494 (process_insn_depend_be_in_spec): Rename to
1495 process_insn_forw_deps_be_in_spec. Update to use new scheduler
1497 (add_to_speculative_block, create_check_block_twin, fix_recovery_deps):
1498 Update to use new scheduler dependencies lists.
1499 (clear_priorities, calc_priorities, add_jump_dependencies): Ditto.
1501 * ddg.c (create_ddg_dependence, create_ddg_dep_no_link): Update to use
1502 new scheduler dependencies lists.
1503 (build_intra_loop_deps): Ditto.
1505 * target.h (struct _dep): Declare to use in
1506 gcc_target.sched.is_costly_dependence.
1507 (struct gcc_target.sched.adjust_cost): Fix typo.
1508 (struct gcc_target.sched.is_costly_dependence): Change signature to use
1509 single dep_t parameter instead of an equivalent triad.
1510 (struct gcc_target.sched.adjust_cost_2): Remove.
1512 * target-def.h (TARGET_SCHED_ADJUST_COST_2): Remove.
1514 * reg-notes.def (DEP_TRUE, DEP_OUTPUT, DEP_ANTI): Update comments.
1516 * doc/tm.texi (TARGET_SCHED_IS_COSTLY_DEPENDENCE): Update
1518 (TARGET_SCHED_ADJUST_COST_2): Remove documentation.
1520 * doc/rtl.texi (LOG_LINKS): Remove part about instruction scheduler.
1521 (REG_DEP_TRUE): Document.
1523 * config/ia64/ia64.c (ia64_adjust_cost_2): Rename to ia64_adjust_cost.
1524 Change signature to correspond to the targetm.sched.adjust_cost hook.
1525 Update use in TARGET_SCHED_ADJUST_COST_2.
1526 (TARGET_SCHED_ADJUST_COST_2): Rename to TARGET_SCHED_ADJUST_COST.
1527 (ia64_dependencies_evaluation_hook, ia64_dfa_new_cycle): Update to use
1528 new scheduler dependencies lists.
1529 (ia64_gen_check): Ditto.
1531 * config/mips/mips.c (vr4130_swap_insns_p): Update to use new scheduler
1534 * config/rs6000/rs6000.c (rs6000_is_costly_dependence): Change
1535 signature to correspond to the targetm.sched.is_costly_dependence hook.
1536 (is_costly_group): Update to use new scheduler dependencies lists.
1538 * config/spu/spu.c (spu_sched_adjust_cost): Use insn_cost () function
1539 instead of INSN_COST () macro.
1541 2007-02-01 Ian Lance Taylor <iant@google.com>
1543 * lower-subreg.c (resolve_clobber): Handle a subreg of a concatn.
1545 2007-02-01 Guy Martin <gmsoft@gentoo.org>
1547 * pa.md (tp_load): Correct mfctl instruction syntax.
1549 2007-02-01 Geoffrey Keating <geoffk@apple.com>
1551 * config/rs6000/rs6000.c (rs6000_stack_info): Correct
1552 altivec_padding_size calculation on AIX. Improve comment, add
1553 assert to verify that it's right.
1555 * config/rs6000/darwin.h (MD_UNWIND_SUPPORT): Don't define for
1558 2007-2-01 Seongbae Park <seongbae.park@gmail.com>
1561 * global.c (global_alloc): Add mising initialization of
1564 2007-02-01 Roger Sayle <roger@eyesopen.com>
1566 * alias.c (init_alias_analysis): Correct whitespace.
1567 * bb-reorder.c (fix_edges_for_rarely_executed_code,
1568 partition_hot_cold_basic_blocks): Likewise.
1569 * builtins.c (expand_builtin_printf, expand_builtin_fprintf,
1570 expand_builtin_sprintf, fold_builtin_carg, fold_builtin_sprintf,
1571 maybe_emit_sprintf_chk_warning, fold_builtin_sprintf_chk,
1572 fold_builtin_snprintf_chk, fold_builtin_printf,
1573 fold_builtin_fprintf, do_mpfr_ckconv, do_mpfr_arg1, do_mpfr_arg2,
1574 do_mpfr_arg3, do_mpfr_sincos): Likewise.
1575 * cfgcleanup.c (cleanup_cfg): Likewise.
1576 * cfgexpand.c (tree_expand_cfg): Likewise.
1577 * fold-const.c (fold_binary) <RDIV_EXPR>: Likewise.
1578 * function.c (get_next_funcdef_no): Likewise.
1579 * gengtype.c (main): Likewise.
1580 * genmodes.c (main): Likewise.
1581 * gcse.c (bypass_conditional_jumps, print_ldst_list): Likewise.
1582 * haifa-sched.c (schedule_block, extend_h_i_d): Likewise.
1583 * ifcvt.c (noce_emit_move_insn): Likewise.
1584 * modulo-sched.c (generate_prolog_epilog, sms_schedule_by_order):
1586 * stor-layout.c (get_best_mode): Likewise.
1587 * tree-ssa-loop-niter.c (get_val_for): Likewise.
1588 * tree-ssa-structalias.c (get_varinfo, get_varinfo_fc,
1589 scc_visit, do_ds_constraint, do_complex_constraint, label_visit,
1590 perform_var_substitution, solve_graph): Likewise.
1591 * tree-vrp.c (vrp_finalize): Likewise.
1593 2007-02-01 Ian Lance Taylor <iant@google.com>
1595 * lower-subreg.c (simplify_gen_subreg_concatn): If we ask for the
1596 high part of a paradoxical subreg, return a constant zero.
1598 2007-02-01 Zdenek Dvorak <dvorakz@suse.cz>
1600 * toplev.c (lang_dependent_init): Call init_set_costs.
1601 * loop-init.c (loop_optimizer_init): Do not call init_set_costs.
1603 2007-02-01 Richard Guenther <rguenther@suse.de>
1606 * fold-const.c (fold_negate_expr): Allow negating a
1607 constant if overflow does not change.
1609 2007-02-01 Ralf Wildenhues <Ralf.Wildenhues@gmx.de>
1611 * doc/c-tree.texi (Expression trees): Improve markup.
1612 * doc/tm.texi (Register Classes, Addressing Modes)
1613 (Floating Point): Fix spacing after abbreviations. Fix some
1616 2007-02-01 Ben Elliston <bje@au.ibm.com>
1618 * doc/invoke.texi: Replace "bugfix" with "bug fix" throughout.
1619 * doc/contrib.texi: Likewise.
1620 * doc/install.texi: Likewise.
1622 2007-01-31 Richard Henderson <rth@redhat.com>
1623 Ian Lance Taylor <iant@google.com>
1625 * lower-subreg.c: New file.
1626 * rtl.def (CONCATN): Define.
1627 * passes.c (init_optimization_passes): Add pass_lower_subreg and
1629 * emit-rtl.c (update_reg_offset): New static function, broken out
1630 of gen_rtx_REG_offset.
1631 (gen_rtx_REG_offset): Call update_reg_offset.
1632 (gen_reg_rtx_offset): New function.
1633 * regclass.c: Revert patch of 2006-03-05, restoring
1635 (clear_reg_info_regno): New function.
1636 * dwarf2out.c (concatn_loc_descriptor): New static function.
1637 (loc_descriptor): Handle CONCATN.
1638 * common.opt (fsplit_wide_types): New option.
1639 * opts.c (decode_options): Set flag_split_wide_types when
1641 * timevar.def (TV_LOWER_SUBREG): Define.
1642 * rtl.h (gen_reg_rtx_offset): Declare.
1643 (reg_scan_update): Declare.
1644 * regs.h (clear_reg_info_regno): Declare.
1645 * tree-pass.h (pass_lower_subreg): Declare.
1646 (pass_lower_subreg2): Declare.
1647 * doc/invoke.texi (Option Summary): List -fno-split-wide-types.
1648 (Optimize Options): Add -fsplit-wide-types to -O1 list. Document
1650 * doc/rtl.texi (Regs and Memory): Document concat and concatn.
1651 * Makefile.in (OBJS-common): Add lower-subreg.o.
1652 (lower-subreg.o): New target.
1654 2007-01-31 Kazu Hirata <kazu@codesourcery.com>
1656 * config/sh/sh.h (HAVE_SECONDARY_RELOADS): Remove.
1658 2007-01-31 Anatoly Sokolov <aesok@post.ru>
1661 * config/avr/avr.c (DWARF2_ADDR_SIZE): Define.
1663 2007-01-31 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1666 * builtins.c (fold_builtin_sqrt): Use MPFR for constant args.
1668 2007-01-31 Zdenek Dvorak <dvorakz@suse.cz>
1670 * cfgloop.h: Include vec-prim.h.
1671 (enum li_flags): Remove LI_ONLY_OLD.
1672 (loop_iterator): Changed.
1673 (fel_next, fel_init): Iterate over loop tree.
1674 (FOR_EACH_LOOP_BREAK): New macro.
1675 * loop-unswitch.c (unswitch_loops): Do not pass LI_ONLY_OLD to
1677 * tree-ssa-loop-unswitch.c (tree_ssa_unswitch_loops): Ditto.
1678 * modulo-sched.c (sms_schedule): Ditto.
1679 * tree-vectorizer.c (vectorize_loops): Ditto.
1680 * doc/loop.texi: Update information on loop numbering and behavior of
1681 FOR_EACH_LOOP wrto new loops.
1682 * tree-scalar-evolution.c (compute_overall_effect_of_inner_loop,
1683 add_to_evolution_1): Test nestedness of loops instead of comparing
1685 * tree-chrec.c (chrec_fold_plus_poly_poly,
1686 chrec_fold_multiply_poly_poly, chrec_evaluate,
1687 hide_evolution_in_other_loops_than_loop, chrec_component_in_loop_num,
1688 reset_evolution_in_loop): Ditto.
1689 * Makefile.in (CFGLOOP_H): Add vecprim.h dependency.
1691 2007-01-31 Dirk Mueller <dmueller@suse.de>
1693 * c-common.c (warn_about_parentheses): Separate warning about
1694 un-parenthized sequence of comparison operators from the one
1695 which is supposed to warn about x <= y <= z.
1697 2007-01-31 Uros Bizjak <ubizjak@gmail.com>
1699 * optabs.h (enum optab_index): Add new OTI_isinf.
1700 (isinf_optab): Define corresponding macro.
1701 * optabs.c (init_optabs): Initialize isinf_optab.
1702 * genopinit.c (optabs): Implement isinf_optab using isinf?f2
1704 * builtins.c (mathfn_built_in): Handle BUILT_IN_ISINF{,F,L}.
1705 (expand_builtin_interclass_mathfn): Expand BUILT_IN_ISINF{,F,L}
1707 (expand_builtin): Expand BUILT_IN_ISINF{,F,L} using
1708 expand_builtin_interclass_mathfn.
1709 * reg_stack.c (subst_stack_regs_pat): Handle UNSPEC_FXAM.
1710 * config/i386/i386.md (UNSPEC_FXAM): New constant.
1711 (fxam<mode>2_i387): New insn pattern.
1712 (isinf<mode>2) New expander to implement isinf, isinff and isinfl
1713 built-in functions as x87 inline asm.
1715 2007-01-31 Kazu Hirata <kazu@codesourcery.com>
1717 * gcc/config/arm/unwind-arm.h (_sleb128_t, _uleb128_t): New.
1719 2007-01-30 Eric Christopher <echristo@apple.com>
1721 * config.gcc: Add geode.
1723 2007-01-31 Kazu Hirata <kazu@codesourcery.com>
1725 * cgraphunit.c, config/arm/arm.c, config/m68k/m68k.c,
1726 ipa-inline.c, tree-profile.c, tree-ssa-live.c,
1727 tree-ssa-math-opts.c, tree-ssanames.c, tree-vect-analyze.c,
1728 value-prof.c: Fix comment typos.
1730 2007-01-30 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
1733 * doc/invoke.texi (Wpointer-arith): Document warning.
1735 2007-01-30 Janis Johnson <janis187@us.ibm.com>
1737 * doc/extend.texi (Decimal Floating Types): Remove decfloat.h from
1738 the list of discrepancies from the draft TR.
1740 2007-01-30 Dirk Mueller <dmueller@suse.de>
1743 * doc/invoke.texi (-Wreturn-type): Update description to
1746 2007-01-30 Richard Sandiford <richard@codesourcery.com>
1748 * cfgrtl.c (try_redirect_by_replacing_jump): Check only_sets_cc0_p.
1750 2007-01-30 Uros Bizjak <ubizjak@gmail.com>
1752 * builtins.c (expand_builtin_int_interclass_roundingfn): New function
1753 to handle optabs that operate on floating point input argument and
1754 output to integer output.
1755 (expand_builtin_mathfn) [BUILT_IN_ILOGB]: Move from here ...
1756 (expand_builtin_interclass_mathfn) [BUILT_IN_ILOGB]: ... to here.
1757 (expand_builtin): Expand BUILT_IN_ILOGB{,F,L} using
1758 expand_builtin_interclass_mathfn ().
1759 * config/i386/i386.md (fxtractxf3_i387): Rename from *fxtractxf3_i387.
1761 (ilogbxf2, ilogb<mode>2): New expanders to implement ilogb, ilogbf and
1762 ilogbl built-in functions as x87 intrinsics.
1764 2007-01-30 Richard Guenther <rguenther@suse.de>
1767 * dwarf2out.c (reference_to_unused): Query varpool if the
1768 variable was output.
1770 2007-01-30 Richard Guenther <rguenther@suse.de>
1773 * passes.c (execute_one_pass): Reset in_gimple_form to not
1774 confuse non-unit-at-a-time mode.
1776 2007-01-29 Roger Sayle <roger@eyesopen.com>
1777 Richard Guenther <rguenther@suse.de>
1779 * fold-const.c (round_up): Make HIGH an unsigned HOST_WIDE_INT to
1780 avoid undefined behaviour on overflow. Use force_fit_type_double
1781 to construct the constant with the specified TREE_OVERFLOW.
1783 2007-01-29 Janis Johnson <janis187@us.ibm.com>
1785 * config/dfp-bit.c: Add parameterized support for fp exceptions.
1786 * config/dfp-bit.h: Ditto.
1788 2007-01-29 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
1790 * c-decl.c (pop_scope): Replace warnings with call to
1791 warn_for_unused_label.
1792 * c-common.h (warn_for_unused_label): Declare.
1793 * c-common.c (warn_for_unused_label): Define.
1795 2007-01-29 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
1797 * tree-optimize.c (update_inlined_to_pointers): Delete unused
1800 2007-01-29 Janis Johnson <janis187@us.ibm.com>
1802 * Makefile.in (USER_H): Remove decfloat.h.
1803 * ginclude/decfloat.h: Delete, moving contents to ...
1804 * ginclude/float.h: Add support for decimal floating point,
1805 guarded by __STDC_WANT_DEC_FP__.
1807 2007-01-29 Mike Stump <mrs@apple.com>
1809 * doc/gccint.texi (Top): Rename Loop Representation to Loop
1810 Analysis and Representation to resolve case insensitive conflict.
1811 * doc/loop.texi (Loop Analysis and Representation): Likewise.
1813 2007-01-28 Daniel Berlin <dberlin@dberlin.org>
1815 * tree.h (struct tree_memory_tag): Add aliases member.
1816 (MTAG_ALIASES): New macro.
1817 * tree-ssa-alias.c (alias_bitmap_obstack): New variable.
1818 (add_may_alias): Remove pointer-set. Update for may_aliases being
1820 (mark_aliases_call_clobbered): Update for may_aliases being a
1822 (compute_tag_properties): Ditto.
1823 (create_partition_for): Ditto.
1824 (compute_memory_partitions): Ditto.
1825 (dump_may_aliases_for): Ditto.
1826 (is_aliased_with): Ditto.
1827 (add_may_alias_for_new_tag): Ditto.
1828 (rewrite_alias_set_for): Rewrite for may_aliases being a bitmap.
1829 (compute_is_aliased): New function.
1830 (compute_may_aliases): Call compute_is_aliased).
1831 (init_alias_info): Initialize alias_bitmap_obstack.
1832 (union_alias_set_into): New function.
1833 (compute_flow_sensitive_aliasing): Use union_aliases_into.
1834 (have_common_aliases_p): Rewrite to take two bitmaps and use
1836 (compute_flow_insensitive_aliasing): Stop using pointer-sets.
1838 (finalize_ref_all_pointers): Update for add_may_alias changes.
1839 (new_type_alias): Ditto.
1840 * tree-flow-inline.h (may_aliases): Return a bitmap.
1841 * tree-dfa.c (dump_variable): Check for MTAG_P'ness.
1842 * tree-ssa.c (verify_flow_insensitive_alias_info): Update for
1843 may_aliases being a bitmap.
1844 * tree-flow.h (struct var_ann_d): Remove may_aliases member.
1845 may_aliases now returns a bitmap.
1846 * tree-ssa-structalias.c (merge_smts_into): Update for may_aliases
1848 * tree-ssa-operands.c (add_virtual_operand): Update for
1849 may_aliases being a bitmap.
1851 2007-01-29 Daniel Berlin <dberlin@dberlin.org>
1853 PR tree-optimization/30630
1854 * tree-ssa-structalias.c (do_complex_constraint): Mark correct
1855 variable as changed.
1857 2007-01-29 Simon Martin <simartin@users.sourceforge.net>
1860 * gimplify.c (gimplify_target_expr): Make sure that the TARGET_EXPR is
1861 expanded only once even if an error occurs.
1863 2007-01-29 Ben Elliston <bje@au.ibm.com>
1865 * gcov-io.h (__gcov_indirect_call_profiler): Declare.
1866 (__gcov_average_profiler): Likewise.
1867 (__gcov_ior_profiler): Likewise.
1868 (__gcov_merge_ior): Likewise.
1870 2007-01-28 Jan Hubicka <jh@suse.cz>
1872 * builtins.c (expand_builtin_memset): Fix typo in my last patch.
1873 * value-prof.c (stringop_block_profile): Likewise.
1875 2007-01-28 Jan Hubicka <jh@suse.cz>
1877 * expr.c (emit_block_move_via_movmem, emit_block_move_via_libcall): Add
1878 variant handling histograms; add wrapper.
1879 (clear_storage_via_libcall): Export.
1880 (emit_block_move_hints): Break out from ...; add histograms.
1881 (emit_block_move): ... this one.
1882 (clear_storage_hints): Break out from ...; add histograms.
1883 (clear_storage): ... this one.
1884 (set_storage_via_memset): Handle histogram.
1885 * expr.h (emit_block_move_via_libcall, emit_block_move_hints): Declare.
1886 (clear_storage_hints, clear_storage_via_libcall): Declare.
1887 (set_storage_via_setmem): Update prototype.
1888 * doc/md.texi (movmem, setmem): Document new arguments.
1890 * value-prof.c (dump_histogram_value, tree_find_values_to_profile): Add
1892 (stringop_block_profile): New global function.
1893 (tree_stringops_values_to_profile): Profile block size and alignment.
1894 * value-prof.h (enum hist_type): add HIST_TYPE_AVERAGE and
1896 (struct profile_hooks): Add gen_average_profiler and gen_ior_profiler.
1897 (stringop_block_profile): Declare.
1898 * builtins.c: Include value-prof.h.
1899 (expand_builtin_memcpy, expand_builtin_memset): Pass block profile.
1900 * gcov-ui.h (GCOV_COUNTER_NAMES): Add new counter.
1901 (GCOV_COUNTER_AVERAGE, GCOV_COUNTER_IOR): New constants.
1902 (GCOV_COUNTERS, GCOV_LAST_VALUE_COUNTER): Update.
1903 * profile.c (instrument_values): Add new counters.
1904 * cfgexpand.c (expand_gimple_basic_block): Propagate histograms to
1906 * tree-profile.c (tree_average_profiler_fn, tree_ior_profiler_fn): New.
1907 (tree_init_edge_profiler): Build new profilers.
1908 (tree_gen_average_profiler, tree_gen_ior_profiler): New.
1909 (pass_tree_profile): Add dump.
1910 (tree_profile_hooks): Update.
1911 * Makefile.in (LIBGCOV): Add new constants.
1912 * libgcov.c (__gcov_merge_ior, __gcov_average_profiler,
1913 __gcov_ior_profiler): New.
1914 * i386.md (movmem/setmem expanders): Add new optional arguments.
1916 2007-01-28 David Edelsohn <edelsohn@gnu.org>
1918 * doc/md.texi (Standard Pattern Names): Document blockage pattern.
1920 2007-01-28 Zdenek Dvorak <dvorakz@suse.cz>
1922 * tree-ssa-loop-unswitch.c: Include tree-inline.h.
1923 (tree_unswitch_single_loop): Pass eni_size_weights to
1924 tree_num_loop_insns.
1925 * tree-ssa-loop-manip.c: Include tree-inline.h.
1926 (can_unroll_loop_p): Pass eni_size_weights to
1927 tree_num_loop_insns.
1928 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p):
1929 Pass eni_size_weights to estimate_num_insns.
1930 * tree.h (init_inline_once): Export.
1931 * toplev.c (backend_init): Call init_inline_once.
1932 * cgraphunit.c (cgraph_process_new_functions,
1933 cgraph_analyze_function): Pass eni_inlining_weights to
1935 * ipa-inline.c (compute_inline_parameters): Ditto.
1936 * tree-ssa-loop-ivcanon.c (tree_num_loop_insns): Pass weights
1937 to estimate_num_insns.
1938 (try_unroll_loop_completely): Pass eni_size_weights to
1939 tree_num_loop_insns.
1940 * tree-eh.c (decide_copy_try_finally): Pass eni_size_weights
1941 ot estimate_num_insns.
1942 * tree-ssa-loop-prefetch.c: Include tree-inline.h.
1943 (loop_prefetch_arrays): Pass eni_time_weights to tree_num_loop_insns.
1944 * tree-inline.c (eni_inlining_weights, eni_size_weights,
1945 eni_time_weights): New variables.
1946 (init_inline_once): Initialize them.
1947 (struct eni_data): Mew.
1948 (estimate_num_insns_1, estimate_num_insns): Use weights.
1949 * tree-inline.h (struct eni_weights_d): New.
1950 (eni_inlining_weights, eni_size_weights, eni_time_weights): Declare.
1951 (estimate_num_insns): Declaration changed.
1952 * cfgloop.h (tree_num_loop_insns): Declaration changed.
1953 * Makefile.in (tree-ssa-loop-unswitch.o, tree-ssa-loop-prefetch.o,
1954 tree-ssa-loop-manip.o): Add TREE_INLINE_H dependency.
1956 2007-01-28 Zdenek Dvorak <dvorakz@suse.cz>
1958 * tree-data-ref.c (conflict_fn): Assert that the number of affine
1959 relations in the conflict function is valid.
1961 2007-01-27 Ian Lance Taylor <iant@google.com>
1963 * common.opt: Add fstrict-overflow.
1964 * opts.c (decode_options): Set flag_strict_overflow if -O2.
1965 * flags.h (TYPE_OVERFLOW_WRAPS): Define.
1966 (TYPE_OVERFLOW_UNDEFINED): Define.
1967 (TYPE_OVERFLOW_TRAPS): Define. This replaces TYPE_TRAP_SIGNED.
1969 * tree.h (TYPE_TRAP_SIGNED): Don't define.
1970 * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_UNDEFINED.
1971 (fold_negate_expr): Likewise.
1972 (make_range): Likewise.
1973 (extract_muldiv_1): Likewise.
1974 (maybe_canonicalize_comparison): Likewise.
1975 (fold_comparison): Likewise.
1976 (fold_binary): Likewise.
1977 (tree_expr_nonnegative_p): Likewise.
1978 (tree_expr_nonzero_p): Likewise.
1979 * tree-vrp.c (compare_values): Likewise.
1980 (extract_range_from_binary_expr): Likewise.
1981 (extract_range_from_unary_expr): Likewise.
1982 * tree-ssa-loop-niter.c (infer_loop_bounds_from_signedness):
1984 (nowrap_type_p): Likewise.
1985 * tree-scalar-evolution.c (simple_iv): Likewise.
1986 * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_WRAPS.
1987 (build_range_check): Likewise.
1988 (extract_muldiv_1): Likewise.
1989 (fold_comparison): Likewise.
1990 * tree-vrp.c (vrp_int_const_binop): Likewise.
1991 (extract_range_from_unary_expr): Likewise.
1992 * convert.c (convert_to_integer): Likewise.
1993 * fold-const.c (fold_negate_expr): Use TYPE_OVERFLOW_TRAPS.
1994 (fold_comparison): Likewise.
1995 (fold_binary): Likewise.
1996 * optabs.c (optab_for_tree_code): Likewise.
1997 * tree-vectorizer.c (vect_is_simple_reduction): Likewise.
1998 * simplify-rtx.c (simplify_const_relational_operation): Check
1999 flag_strict_overflow and flag_trapv.
2000 (simplify_const_relational_operation): Likewise.
2001 * doc/invoke.texi (Option Summary): Mention -fstrict-overflow.
2002 (Optimize Options): Add -fstrict-overflow to -O2 list. Document
2005 2007-01-27 Roger Sayle <roger@eyesopen.com>
2007 * tree.c (tree_fold_gcd): Delete.
2008 * tree.h (tree_fold_gcd): Remove prototype.
2009 * tree-data-ref.c (tree_fold_divides_p): Don't use tree_fold_gcd to
2010 test whether one constant integer is a multiple of another. Instead
2011 call int_const_binop with TRUNC_MOD_EXPR and test for a zero result.
2012 * fold-const.c (multiple_of_p): We've determined both TOP and
2013 BOTTOM are integer constants so we can call int_const_binop directly
2014 instead of the more generic const_binop.
2016 2007-01-27 Roger Sayle <roger@eyesopen.com>
2018 * fold-const.c (size_binop): In the fast-paths for X+0, 0+X, X-0 and
2019 1*X check that the constant hasn't overflowed, to preserve the
2021 (round_up): Provide an efficient implementation when rouding-up an
2022 INTEGER_CST to a power-of-two.
2024 2007-01-28 Ralf Wildenhues <Ralf.Wildenhues@gmx.de>
2026 * doc/sourcebuild.texi: Add comma for clarity.
2027 * doc/extend.texi: Fix some typos.
2028 * doc/passes.texi: Likewise.
2029 * doc/cppinternals.texi: Likewise.
2030 * doc/c-tree.texi: Likewise.
2031 * doc/tree-ssa.texi: Likewise.
2032 * doc/install.texi: Likewise.
2034 2007-01-27 Jan Hubicka <jh@suse.cz>
2036 * tree-sra.c (sra_walk_function): Don't rely on aliases being build.
2037 (pass_sra): Do not require alias information.
2038 * passes.c (init_optimization_passes): Add SRA
2040 2007-01-27 Steven Bosscher <steven@gcc.gnu.org>
2042 * tracer.c (rest_of_handle_tracer): We already cleaned
2043 up the CFG in tracer() so don't do it here again.
2044 * cfgcleanup.c (rest_of_handle_jump2): Don't repeat
2045 cleanup_cfg here, either. And don't call renumber_insns.
2047 * cfgrtl.c (rtl_verify_flow_info_1): Don't verify that BB_END
2048 and BB_HEAD are in the insn stream here. Instead make sure
2049 that BB_INSN is valid on all insns. Also, do check here that
2050 there are no pending branch predictions...
2051 (rtl_verify_flow_info): ...instead of doing it here. Checks
2052 for BB_END and BB_HEAD moved from rtl_verify_flow_info_1 to
2055 2007-01-26 Roger Sayle <roger@eyesopen.com>
2057 * config/i386/i386.c (ix86_swap_binary_operands_p): New helper
2058 function to simplify/factorize operand order canonicalization.
2059 (ix86_fixup_binary_operands): Reorganize using the above function.
2060 (ix86_binary_operator_ok): Likewise.
2062 2007-01-27 Jakub Jelinek <jakub@redhat.com>
2064 * genattrtab.c (struct attr_value_list, insn_code_values): Move to
2065 file scope from optimize_attrs.
2066 (simplify_test_exp): If insn_code_values is not NULL, use it to speed
2068 (optimize_attrs): Clear insn_code_values after freeing it.
2070 2007-01-26 Zdenek Dvorak <dvorakz@suse.cz>
2072 * tree-ssa-address.c (create_mem_ref): Remove ", bsi" from
2073 a parts.base assignment.
2075 2007-01-26 Zdenek Dvorak <dvorakz@suse.cz>
2077 * tree-data-ref.c (dump_subscript): Use dump_conflict_function.
2078 (compute_subscript_distance, initialize_data_dependence_relation,
2079 finalize_ddr_dependent, analyze_ziv_subscript,
2080 analyze_siv_subscript_cst_affine,
2081 compute_overlap_steps_for_affine_univar,
2082 compute_overlap_steps_for_affine_1_2, analyze_subscript_affine_affine,
2083 analyze_siv_subscript, analyze_miv_subscript,
2084 analyze_overlapping_iterations, subscript_dependence_tester_1,
2085 compute_self_dependence, free_dependence_relation): Work
2086 with affine_fn instead of chrecs.
2087 (dump_affine_function, dump_conflict_function, affine_function_equal_p,
2088 common_affine_function, affine_function_base,
2089 affine_function_constant_p, affine_fn_op, affine_fn_plus,
2090 affine_fn_minus, affine_fn_free, conflict_fn_not_known,
2091 conflict_fn_no_dependence, free_conflict_function, free_subscripts,
2092 conflict_fn, affine_fn_cst, affine_fn_univar): New functions.
2093 (all_chrecs_equal_p): Removed.
2094 * tree-data-ref.h (affine_fn, conflict_function): New types.
2095 (struct subscript): Change type of conflicting_iterations_in_a
2096 and conflicting_iterations_in_b.
2098 2007-01-26 Steve Ellcey <sje@cup.hp.com>
2101 * config/pa/pa.h (TARGET_HPUX_11): New.
2102 * config/pa/pa-hpux11.h (TARGET_HPUX_11): New.
2103 * config/pa/pa.c (pa_init_builtins): Use TARGET_HPUX_11.
2105 2007-01-26 Daniel Berlin <dberlin@dberlin.org>
2106 Richard Guenther <rguenther@suse.de>
2108 * tree-ssa-structalias.c (solve_graph): Handle case
2109 we merged the variable to another.
2111 2007-01-25 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
2113 * builtins.c (fold_builtin_1): Treat ccos and ccosh as 'even'
2116 * fold-const.c (negate_mathfn_p): Treat casin, casinh, catan,
2117 catanh, cproj, csin, csinh, ctan and ctanh as 'odd' functions.
2119 2007-01-25 DJ Delorie <dj@redhat.com>
2121 * config/m32c/m32c.c (m32c_cannot_change_mode_class): We don't
2122 allow changes to modes which don't fit in those registers.
2124 * reload1.c (choose_reload_regs): Check for invalid subregs before
2125 computing their locations, not after.
2127 2007-01-25 Geoffrey Keating <geoffk@apple.com>
2130 * config/rs6000/rs6000.c (first_altivec_reg_to_save): On Darwin,
2131 save Altivec registers in an eh_return function.
2132 (compute_vrsave_mask): Likewise.
2133 (rs6000_stack_info): Correct AIX/Darwin stack alignment computation
2134 for saving Altivec registers.
2135 (rs6000_emit_prologue): Don't allocate stack twice in
2136 eh_return function. Correct expected value of altivec_save_offset
2137 when using save_world. Describe save of R0 to stack when using
2138 save_world. Describe stack pointer adjustment when using
2139 save_world. Remove duplicated eh_return parameter register saving.
2140 Update sp_offset variable after save_world.
2141 * config/rs6000/t-darwin (LIB2FUNCS_STATIC_EXTRA): Remove
2143 (LIB2FUNCS_EXTRA): Add darwin-world.asm.
2144 * config/rs6000/darwin.h (SUBTARGET_OVERRIDE_OPTIONS): -m64
2147 2007-01-25 Steve Ellcey <sje@cup.hp.com>
2149 * config.gcc (ia64*-*-hpux*): Make posix threads the default.
2151 2007-01-25 Steve Ellcey <sje@cup.hp.com>
2154 * config/pa/pa.c (pa_init_builtins): Set asm names for finite routines.
2155 * config/ia64/ia64.c (ia64_init_builtins): Ditto.
2157 2007-01-25 Richard Guenther <rguenther@suse.de>
2159 * doc/invoke.texi (-Wcoverage-mismatch): Document.
2160 * common.opt (-Wcoverage-mismatch): New warning option.
2161 * coverage.c (get_coverage_counts): Ignore coverage mismatch
2162 if -Wcoverage-mismatch is given.
2164 2007-01-25 Razya Ladelsky <razya@il.ibm.com>
2166 * ipa-cp.c (ipcp_insert_stage, ipcp_driver): Support for SSA.
2167 (ipcp_driver): Change to static definition.
2168 Add dumping of the ifunctions.
2169 (constant_val_insert): Remove unused parameter. Support for SSA.
2170 (ipcp_propagate_const): Support for SSA.
2171 (ipcp_profile_bb_print): Print only analyzed nodes.
2172 (ipcp_replace_map_create): Remove support for Fortran constant
2174 * ipa-prop.c (ipa_method_modify_stmt,
2175 ipa_callsite_compute_param): Support for SSA.
2176 * ipa-prop.h (ipcp_driver): Remove declaration.
2177 (IS_VALID_TREE_MAP_INDEX): Add define.
2179 2007-01-24 Geoffrey Keating <geoffk@apple.com>
2181 * unwind-dw2.c (execute_stack_op): Handle DW_OP_swap.
2183 2007-01-24 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
2186 * builtins.c (fold_builtin_cabs): Use MPFR to evaluate a
2187 constant argument to cabs and do it without checking for
2188 -funsafe-math-optimizations.
2190 2007-01-24 Douglas Gregor <dgregor@osl.iu.edu>
2192 * c-common.h (RID_FIRST_CXX0X): New.
2193 (RID_LAST_CXX0X): New.
2194 * c-opts.c (c_common_handle_option): -Wc++0x-compat is triggered
2196 * c.opt (Wc++0x-compat): New.
2197 * doc/invoke.texi (-Wc++0x-compat): Document.
2199 2007-01-24 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
2201 * builtins.c (fold_builtin_carg): New.
2202 (fold_builtin_1): Use it.
2204 2007-01-24 Jan Hubicka <jh@suse.cz>
2206 * ipa-inline.c (cgraph_decide_inlining): Initialize initial_insns.
2208 * ipa-inline.c (initial_insns, max_insns): Delete.
2209 (compute_max_insns): New function.
2210 (cgraph_decide_inlining_of_small_function): Use it; take minimal amount
2211 of insns as base for code growth.
2212 (cgraph_decide_inlining): Make initial_insns local; do not compute
2214 * params.def (PARAM_INLINE_UNIT_GROWTH): Set to 60.
2215 * doc/invoke.texi (inline-unit-growth): Update docs.
2217 2007-01-24 Jakub Jelinek <jakub@redhat.com>
2219 * config/i386/i386.h (x86_cmpxchg16b): Remove const.
2220 (TARGET_CMPXCHG16B): Define to x86_cmpxchg16b.
2221 * config/i386/i386.c (x86_cmpxchg16b): Remove const.
2222 (override_options): Add PTA_CX16 flag. Set x86_cmpxchg16b
2223 for CPUs that have PTA_CX16 set.
2226 * gimplify.c (omp_check_private): New function.
2227 (gimplify_scan_omp_clauses): Use it for
2228 firstprivate/lastprivate/reduction.
2231 * gimplify.c (omp_add_variable): Don't call omp_notice_variable
2232 on TYPE_SIZE_UNIT for GOVD_LOCAL VLAs.
2235 * omp-low.c (lower_omp_for_lastprivate): Add dlist argument.
2236 If lower_lastprivate_clauses emits some statements, append them
2237 to dlist rather than body_p and to body_p append an initializer.
2238 (lower_omp_for): Adjust caller.
2240 2007-01-24 Steve Ellcey <sje@cup.hp.com>
2242 * target.h (globalize_decl_name): New.
2243 * target-def.h (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
2244 * output.h (default_globalize_decl_name): New.
2245 * varasm.c (asm_output_bss): Use globalize_decl_name instead of
2247 (globalize_decl): Ditto.
2248 (default_globalize_decl_name): New.
2249 * config/ia64/ia64.c (ia64_globalize_decl_name): New.
2250 (ia64_handle_version_id_attribute): New.
2251 (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
2252 (ia64_asm_output_external): Use globalize_decl_name instead
2254 * doc/extend.texi (version_id): New pragma.
2255 * doc/tm.texi (ARGET_ASM_GLOBALIZE_DECL_NAME): New target hook.
2257 2007-01-24 Andreas Krebbel <krebbel1@de.ibm.com>
2259 * unwind-dw2-fde.c (get_cie_encoding): Replaced _Unwind_Word with
2260 _uleb128_t and _Unwind_SWord with _sleb128_t.
2261 * unwind-dw2.c (extract_cie_info, execute_stack_op, execute_cfa_program,
2262 uw_frame_state_for, uw_update_context_1): Likewise.
2263 * unwind-c.c (parse_lsda_header, PERSONALITY_FUNCTION): Likewise.
2264 * unwind-pe.h (read_uleb128, read_sleb128,
2265 read_encoded_value_with_base): Likewise.
2266 * unwind-generic.h: Define _sleb128_t and _uleb128_t types.
2268 2007-01-24 Richard Guenther <rguenther@suse.de>
2270 * builtins.c (expand_builtin_cexpi): Get the fndecl
2271 for cexp in the correct way.
2273 2007-01-24 Jan Hubicka <jh@suse.cz>
2275 * tree-ssa-dce.c (eliminate_unnecesary_stmts): Remove dead LHS of calls.
2277 2007-01-24 Andreas Krebbel <krebbel1@de.ibm.com>
2279 * c-cppbuiltin.c (builtin_define_type_sizeof): New function.
2280 (c_cpp_builtins): New builtin macros: __SIZEOF_INT__, __SIZEOF_LONG__,
2281 __SIZEOF_LONG_LONG__, __SIZEOF_SHORT__, __SIZEOF_POINTER__,
2282 __SIZEOF_FLOAT__, __SIZEOF_DOUBLE__, __SIZEOF_LONG_DOUBLE__,
2283 __SIZEOF_SIZE_T__, __SIZEOF_WCHAR_T__, __SIZEOF_WINT_T__ and
2284 __SIZEOF_PTRDIFF_T__.
2285 * doc/cpp.texi: Documentation for the new builtin macros added.
2287 2007-01-24 Uros Bizjak <ubizjak@gmail.com>
2289 * config/i386/i386.md (tanxf2, tan<mode>2, atan<mode>2, log<mode>2,
2290 log10<mode>2, log2<mode>2, expxf2, exp10xf2, exp2xf2): Use op2
2291 instead of operands[2] to avoid access past the end of array.
2293 2007-01-24 Richard Sandiford <richard@codesourcery.com>
2295 * reload1.c (emit_reload_insns): Pass the reload register
2296 for a non-spill output reload through forget_old_reloads_1.
2298 2007-01-23 Joseph Myers <joseph@codesourcery.com>
2300 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Define
2301 _SOFT_DOUBLE if doubles use software floating-point.
2302 * config/rs6000/libgcc-ppc-glibc.ver: Export additional long
2303 double functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
2304 * config/rs6000/darwin-ldouble.c: Also compile functions for
2305 hard-float without FPRs. Use fmsub function for all __NO_FPRS__
2306 cases. Compile extra functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
2307 * config/rs6000/linuxspe.h (SUBSUBTARGET_OVERRIDE_OPTIONS): Remove
2308 commented-out long double override.
2309 (CPP_LONGDOUBLE_DEFAULT_SPEC): Likewise.
2310 * config/rs6000/eabispe.h: Likewise.
2311 * config/rs6000/rs6000.c (rs6000_override_options): Don't override
2312 long double for non-SPE.
2313 (rs6000_handle_option): Likewise.
2314 (invalid_e500_subreg): Disallow more subregs involding DImode,
2315 DFmode, TImode or TFmode.
2316 (rs6000_legitimate_offset_address_p): Check TFmode offsets for
2318 (legitimate_lo_sum_address_p): Also check for TFmode for E500
2320 (rs6000_legitimize_address): Also handle TFmode for E500 double.
2321 (rs6000_legitimize_reload_address): Also handle TFmode for E500
2323 (rs6000_legitimate_address): Also check for TFmode for E500
2325 (rs6000_emit_move): Use DFmode subregs of TFmode for E500 double.
2326 (spe_build_register_parallel): Handle TFmode and TCmode.
2327 (rs6000_spe_function_arg): Handle TFmode and TCmode for E500
2329 (function_arg): Handle TFmode and TCmode for E500 double.
2330 (rs6000_init_libfuncs): Initialize extra libfuncs for soft double
2332 (print_operand): Handle TFmode and TImode for %y.
2333 (rs6000_generate_compare): Handle TFmode comparisons for E500
2335 (spe_func_has_64bit_regs_p): Check for TFmode for E500 double.
2336 (rs6000_function_value): Handle TFmode and TCmode for E500 double.
2337 (rs6000_libcall_value): Handle TFmode and TCmode for E500 double.
2338 * config/rs6000/rs6000.h (CANNOT_CHANGE_MODE_CLASS): Check for
2339 TFmode for E500 double.
2340 * config/rs6000/rs6000.md (FP): Allow TF for E500 double.
2341 (floatsidf2): Enable for E500 double.
2342 (movtf_softfloat): Use rs6000_nonimmediate_operand.
2343 (extenddftf2): Change to extenddftf2_fprs.
2344 (extenddftf2): Call gen_spe_extenddftf2 or gen_extenddftf2_fprs
2345 depending on TARGET_E500_DOUBLE.
2346 (extendsftf2): Enable for E500 double.
2347 (trunctfdf2): Enable for E500 double.
2348 (trunctfsf2): Change to trunctfsf2_fprs.
2349 (trunctfsf2): Call gen_spe_trunctfsf2 or gen_trunctfsf2_fprs
2350 depending on TARGET_E500_DOUBLE.
2351 (floatsitf2): Enable for E500 double.
2352 (fix_trunctfsi2): Change to fix_trunctfsi2_fprs.
2353 (fix_trunctfsi2): Call gen_spe_fix_trunctfsi2 or
2354 gen_fix_trunctfsi2_fprs depending on TARGET_E500_DOUBLE.
2355 (negtf2): Change to negtf2_internal.
2356 (negtf2): New expander.
2357 (abstf2): Enable for E500 double. Call gen_spe_abstf2_tst,
2358 gen_spe_abstf2_cmp or gen_abstf2_internal depending on
2359 TARGET_E500_DOUBLE and flag_unsafe_math_optimizations.
2360 (movdi_internal32): Use rs6000_nonimmediate_operand.
2361 (unnamed splitter): Likewise.
2362 * config/rs6000/spe.md (CMPTFEQ_GPR, TSTTFEQ_GPR, CMPTFGT_GPR,
2363 TSTTFGT_GPR, CMPTFLT_GPR, TSTTFLT_GPR): New unspecs.
2364 (SPE64TF, DITI): New mode macros.
2365 (frob_df_di): Change to frob_<SPE64:mode>_<DITI:mode>; allow more
2368 (frob_<mode>_di_2): New.
2369 (frob_tf_di_8_2): New.
2370 (frob_di_df): Change to frob_di_<mode>; allow more modes.
2372 (frob_di_df_2): Change to frob_<DITI:mode>_<SPE64:mode>_2; allow
2374 (frob_ti_<mode>_8_2): New.
2375 (frob_ti_tf_2): New.
2376 (mov_si<mode>_e500_subreg0, mov_si<mode>_e500_subreg0_2,
2377 mov_si<mode>_e500_subreg4, mov_si<mode>_e500_subreg4_2): Allow
2379 (mov_sitf_e500_subreg8, mov_sitf_e500_subreg8_2,
2380 mov_sitf_e500_subreg12, mov_sitf_e500_subreg12_2): New.
2381 (spe_trunctfdf2_internal1, spe_trunctfsf2, spe_extenddftf2,
2382 spe_fix_trunctfsi2, spe_fix_trunctfsi2_internal,
2383 spe_negtf2_internal, spe_abstf2_cmp, spe_abstf2_tst): New.
2384 (cmptfeq_gpr, tsttfeq_gpr, cmptfgt_gpr, tsttfgt_gpr, cmptflt_gpr,
2387 2007-01-23 Ian Lance Taylor <iant@google.com>
2389 * Makefile.in (OBJS-common): Reformat, alphabetize, but put
2391 (OBJS-archive): Reformat, alphabetize.
2392 (OBJS): Change out_object_file to OBJS-md.
2394 2007-01-23 Uros Bizjak <ubizjak@gmail.com>
2396 * config/i386/i386.md: Use REG_P, MEM_P, CONST_INT_P, LABEL_P,
2397 JUMP_P and CALL_P predicates where applicable.
2398 * config/i386/i386.c: Ditto.
2399 * config/i386/i386.md: Ditto.
2400 * config/i386/mmx.md: Ditto.
2401 * config/i386/predicates.md: Ditto.
2403 2007-01-22 Andreas Schwab <schwab@suse.de>
2405 * config/m68k/m68k.h: Fix comment.
2407 2007-01-22 Jan Hubicka <jh@suse.cz>
2409 * passes.c (init_optimization_passes): Do not rerun
2410 pass_early_warn_uninitialized.
2412 2007-01-22 Richard Guenther <rguenther@suse.de>
2414 PR tree-optimization/30038
2415 * tree-ssa-math-opts.c (maybe_record_sincos): New static helper
2417 (execute_cse_sincos_1): Likewise.
2418 (execute_cse_sincos): Likewise.
2419 (gate_cse_sincos): Likewise.
2420 (pass_cse_sincos): New pass CSEing sin() and cos() calls using
2421 the cexpi() canonicalization of sincos().
2422 * tree-pass.h (pass_cse_sincos): Declare.
2423 * passes.c (init_optimization_passes): New pass pas_cse_sincos.
2425 2007-01-21 Eric Botcazou <ebotcazou@libertysurf.fr>
2427 PR rtl-optimization/29329
2428 * combine.c (replaced_rhs_insn): Rename to i2mod.
2429 (replaced_rhs_value): Rename to i2mod_new_rhs.
2430 (i2mod_old_rhs): New global variable.
2431 (combine_instructions): Adjust for above change. Save a copy of
2432 the old RHS into i2mod_old_rhs when the contents of a REG_EQUAL
2433 note are substituted in the second instruction.
2434 (distribute_notes) <REG_DEAD>: Adjust for above change. Do not
2435 ditch the note if it pertains to the second eliminated register
2436 and this register is mentioned in i2mod_old_rhs.
2439 2006-09-12 Eric Botcazou <ebotcazou@libertysurf.fr>
2441 * combine.c (distribute_notes) <REG_DEAD>: Do not consider SETs past
2442 the insn to which the note was originally attached.
2444 2007-01-21 Jan Hubicka <jh@suse.cz>
2446 * ipa-inline.c (inlining_mode): Comment, move up.
2447 (cgraph_decide_inlining_incrementally): Do not perform inlining
2448 itself; fix handling of flattening of self recursive functions.
2449 (cgraph_find_cycles): Remove.
2450 (cgraph_flatten_node): Remove.
2451 (cgraph_decide_inlining): Use incremental inliner to handle flattening.
2452 (try_inline): New function.
2453 (cgraph_early_inlining): Update call of
2454 cgraph_decide_inlining_incrementally. Apply inlining here.
2455 (apply_inline): Update call of cgraph_decide_inlining_incrementally.
2457 2007-01-21 Dirk Mueller <dmueller@suse.de>
2460 * tree-vrp.c (check_array_bounds): do not warn
2461 about ADDR_EXPR's of ARRAY_REF's which are immediately
2462 used in binary expressions.
2464 2007-01-21 Ira Rosen <irar@il.ibm.com>
2466 * tree-vectorizer.h (struct _stmt_vec_info): Add new field
2467 read_write_dep and macros for its access.
2468 * tree-vectorizer.c (new_stmt_vec_info): Initialize the new field.
2469 * tree-vect-analyze.c (vect_analyze_data_ref_dependence): Remove
2470 argument, call vect_check_interleaving for every independent pair of
2471 data-refs. Mark loads that access the same memory location as a store
2473 (vect_check_dependences): Remove.
2474 (vect_analyze_data_ref_dependences): Remove vect_check_dependences
2475 call, fix the call to vect_analyze_data_ref_dependence.
2476 (vect_analyze_data_ref_access): For statements that access the same
2477 data-ref, check that they are not stores; for loads, check that there
2478 is no store that access the same location.
2480 2007-01-20 Roger Sayle <roger@eyesopen.com>
2481 Joseph Myers <joseph@codesourcery.com>
2483 * doc/invoke.texi (-fdump-rtl-combine): Fix under/overfull hbox.
2484 (-fdump-rtl-stack): Likewise.
2485 (-fno-signed-zeros): Use @minus{} for a minus sign. Correct typo.
2486 (-mcheck-zero-division, -mcpu): Fix under/overfull hbox.
2487 (-mpt-fixed): Use @minus{} for minus sign.
2488 (Using Precompiled Headers): Fix under/overfull hbox.
2490 2007-01-20 Jan Hubicka <jh@suse.cz>
2492 * tree-flow.h (struct stmt_ann_d): Move references_memory to proper
2493 place within annotation.
2495 2007-01-20 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
2497 * pa.c (output_move_double): Change array size of xoperands to 4.
2499 2007-01-20 Jan Hubicka <jh@suse.cz>
2501 * tree-tailcall.c (adjust_return_value): Do not use RESULT_DECL
2504 2007-01-19 Ian Lance Taylor <iant@google.com>
2506 * expmed.c (expand_divmod) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]: Cast
2507 constant to unsigned HOST_WIDE_INT before negating.
2509 2007-01-19 Ian Lance Taylor <iant@google.com>
2511 * tree-ssa-operands.h (struct vuse_vec_d): Change num_vuse field
2513 (VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Use unsigned comparison.
2514 (VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
2515 (SET_VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Likewise.
2516 (SET_VUSE_ELEMENT_VAR) [ENABLE_CHECKING]: Likewise.
2517 (SET_VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
2518 (realloc_vdef, realloc_vuse): Change second parameter to
2520 (ssa_operand_iterator_d): Change vuse_index and mayuse_index
2522 * tree-ssa-operands.c (realloc_vop): Change num_elem parameter to
2523 unsigned. Change x and lim locals to unsigned.
2524 (realloc_vdef, realloc_vuse): Change num_elem parameter to
2526 (finalize_ssa_vuse_ops): Change old_i local to unsigned.
2527 (copy_virtual_operands): Change i and n locals to unsigned.
2529 2007-01-19 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
2532 * builtins.c (fold_builtin_1): Handle builtin fdim.
2534 2007-01-20 Jan Hubicka <jh@suse.cz>
2536 * tree-ssa.c (init_tree_ssa): Do not call init_alias_heapvars.
2537 * tree-ssa-structalias.c (compute_points_to_sets): Do call
2538 init_alias_heapvars.
2539 (init_alias_heapvars): Initialize only when not already initialized.
2540 (delete_alias_heapvars): Set heapvar_for_stmt to NULL.
2542 2007-01-19 Roger Sayle <roger@eyesopen.com>
2544 * common.opt (fsigned-zeros): New command line option.
2545 * flags.h (HONOR_SIGNED_ZEROS): Control via flag_signed_zeros instead
2546 of flag_unsafe_math_optimizations.
2547 * opts.c (set_fast_math_flags): The -ffast-math command line option
2548 implies -fno-signed-zeros.
2549 (fast_math_flags_set_p): Likewise.
2551 * doc/invoke.texi: Document new -fno-signed-zeros option, and update
2552 the documentation of -ffast-math appropriately. Wrap long lines.
2554 2007-01-19 Steve Ellcey <sje@cup.hp.com>
2556 * system.h (ASM_MAKE_LABEL_LINKONCE): Poison.
2557 * varasm.c (globalize_decl): Remove ASM_MAKE_LABEL_LINKONCE ifdef.
2559 2007-01-19 Tomas Bily <tbily@suse.cz>
2561 * cgraphunit.c (cgraph_finalize_function): Updating of pid
2563 (tree_init_ic_make_global_vars): New function
2564 (tree_init_edge_profiler): call of tree_init_ic_make_global_vars
2565 (tree_gen_ic_profiler): New function
2566 (tree_gen_ic_func_profiler): New function
2567 (tree_profiling): Added calling of tree_gen_ic_func_profiler
2568 (tree_profile_hooks): Added hook for indirec/virtual calls
2569 * value-prof.c (tree_find_values_to_profile): New case for
2571 (tree_values_to_profile): Call for determining indirect/virtual
2573 (tree_indirect_call_to_profile): New function
2574 (tree_ic_transform): New function
2575 (tree_ic): New function
2576 (find_func_by_pid): New function
2577 (init_pid_map): New function
2578 (tree_value_profile_transformations): Added check for
2579 indirect/virtual call transformation
2580 * value-prof.h (enum hist_type): New counter type for
2581 indirect/virtual calls
2582 (profile_hooks): Added new hook for profiling indirect/virtual
2584 * profile.c (instrument_values): New case for indirect/virtual
2586 * gcov-io.h (GCOV_LAST_VALUE_COUNTER): Changed to 6
2587 (GCOV_COUNTER_V_INDIR): New counter type
2588 (GCOV_COUNTER_NAMES): New name of counter "indirect" added
2589 (GCOV_MERGE_FUNCTIONS): New merge function for indirect/virtual
2591 * cgraph.c: Definition of cgraph_max_pid
2592 (cgraph_create_node): Default init of pid attribute
2593 * cgraph.h: Declaration of cgraph_max_pid
2594 (struct cgraph_node): Added pid attribute
2595 * libgcov.c (__gcov_indirect_call_profiler): New function
2596 (__gcov_one_value_profiler_body): New function
2597 (__gcov_one_value_profiler): Body was moved to
2598 __gcov_one_value_profiler_body and calls it
2600 2007-01-19 Basile Starynkevitch <basile@starynkevitch.net>
2602 * doc/gty.texi (Options): Document the mark_hook option to GTY.
2603 * gengtype.c (write_types_data, write_func_for_structure,
2604 write_types, ggc_wtd, pch_wtd): Add skip_hooks to
2605 write_types_data, ggc_wtd, pch_wtd for processing mark_hook.
2606 (walk_type, write_func_for_structure): Generate the mark_hook if
2609 2007-01-19 Jan Hubicka <jh@suse.cz>
2611 * ipa-inline.c (cgraph_decide_inlining_incrementally): Instead of
2612 'early' argument take inlining mode argument specifying whether to
2613 inline for size/speeed or all functions; add support for flattening;
2615 (cgraph_early_inlining): Update call of decide_inlining_incrementally.
2617 2007-01-19 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
2620 * toplev.c (warn_deprecated_use): Use %qD instead of %qs to print
2621 the name of the declared identifier.
2623 2007-01-19 Dirk Mueller <dmueller@suse.de>
2625 * config/i386.h (CONDITIONAL_REGISTER_USAGE): Store
2626 result of PIC_OFFSET_TABLE_REGNUM in temporary variable to avoid
2627 duplicate evaluation.
2629 2007-01-19 Uros Bizjak <ubizjak@gmail.com>
2631 * config/i386/i386.md (acos<mode>2): Rename from acossf2 and acosdf2.
2632 Macroize expander using X87MODEF12 mode macro. Extend operand 1
2633 to XFMode, use acosxf2 and truncate result to requested mode.
2634 Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
2635 (asin<mode>2): Similarly, with asin expanders.
2636 (*fscalexf4_i387): Rename from *fscalexf4.
2637 (expNcorexf3): New expander.
2638 (expxf2, exp10xf2, exp2xf2): Use expNcorexf3 expander.
2639 (exp<mode>2): Rename from expsf2 and expdf2. Macroize expander using
2640 X87MODEF12 mode macro. Extend operand 1 to XFMode, use expxf2 and
2641 truncate result to requested mode. Use SSE_FLOAT_MODE_P to disable
2642 patterns for SSE math.
2643 (exp10<mode>2): Similarly, with exp10 expanders.
2644 (exp2<mode>2): Similarly, with exp2 expanders.
2645 (expm1<mode>2): Similarly, with expm1 expanders.
2646 (ldexp<mode>3): Similarly, with ldexp expanders.
2647 (log<mode>2, log10<mode>2, log2<mode>2, log1p<mode>2, logb<mode>2):
2648 Use gen_truncxf<mode>2_i387_noop to truncate result.
2650 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2652 * config/m68k/m68k.h (PREFERRED_STACK_BOUNDARY): Define to 32
2653 for ColdFire targets.
2655 2007-01-19 Nathan Sidwell <nathan@codesourcery.com>
2656 Richard Sandiford <richard@codesourcery.com>
2658 * config/m68k/m68k.h (M68K_STATIC_CHAIN_REG_NAME): New macro.
2659 (INITIALIZE_TRAMPOLINE): Use STATIC_CHAIN_REGNUM.
2660 (__transfer_from_trampoline): Use M68K_STATIC_CHAIN_REG_NAME.
2661 * config/m68k/m68kelf.h (STATIC_CHAIN_REGNUM): Override.
2662 (M68K_STATIC_CHAIN_REG_NAME): Likewise.
2663 * config/m68k/netbsd-elf.h (M68K_STATIC_CHAIN_REG_NAME): Likewise.
2665 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2667 * config/m68k/m68k.md (adddi_dilshr32): Rename to...
2668 (*adddi_dilshr32): ...this. Fix formatting. Remove commented-out
2669 non-canonical pattern. Restrict to !TARGET_COLDFIRE.
2670 (*adddi_dilshr32_cf): New pattern.
2671 (adddi3, subdi3): Remove first alternatives.
2673 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2675 * config/m68k/m68k.c (notice_update_cc): If an SFmode move is
2676 implemented using move.l, do not use its cc result for floating-point
2679 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2681 * config/m68k/m68k.h (EXTRA_CONSTRAINT): Stop the 'T' constraint
2682 from accepting 's' constraints if flag_pic.
2684 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2686 * config/m68k/m68k.md (bordered, bunordered, buneq, bunge, bungt)
2687 (bunle, bunlt, bltgt, bordered_rev, bunordered_rev, buneq_rev)
2688 (bunge_rev, bungt_rev, bunle_rev, bunlt_rev, bltgt_rev): Change
2689 condition from TARGET_68881 to TARGET_HARD_FLOAT.
2691 2007-01-19 Sandra Loosemore <sandra@codesourcery.com>
2693 * longlong.h (count_leading_zeros, COUNT_LEADING_ZEROS_0): Add
2694 ColdFire alternatives.
2695 * config/m68k/m68k.h (CLZ_DEFINED_VALUE_AT_ZERO): New macro.
2696 * config/m68k/m68k.md (clzsi2): Define for ColdFire
2697 architectures that support ff1 instruction.
2699 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2700 Julian Brown <julian@codesourcery.com>
2702 * config/m68k/m68k.h (CONST_OK_FOR_LETTER_P): Add an 'R' case.
2703 * config/m68k/m68k.md (*movsi_cfv4): Fold into...
2704 (*movsi_cf): ...here. Remove unnecessary 'R' from 'Rg'.
2707 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2709 * config/m68k/m68k-protos.h (valid_mov3q_const): Take a HOST_WIDE_INT
2711 (output_move_const_into_data_reg, output_move_simode_const): Delete.
2712 * config/m68k/m68k.c (const_method, const_int_cost): Take a
2713 HOST_WIDE_INT instead of an rtx.
2714 (m68k_rtx_costs): Update call accordingly.
2715 (output_move_const_into_data_reg): Likewise. Fix formatting.
2716 (valid_mov3q_const): Take a HOST_WIDE_INT instead of an rtx.
2718 (output_move_simode_const): Update calls after above changes.
2719 Rework to use automatic variables and predicates like MEM_P.
2720 * config/m68k/m68k.md (pushexthisi_const): Update call to
2723 2007-01-19 Dirk Mueller <dmueller@suse.de>
2725 * tree-ssa-alias.c (perform_var_substitution): Fix typo
2728 2007-01-19 Richard Guenther <rguenther@suse.de>
2730 * builtins.c (expand_builtin_cexpi): Fall back to expanding
2731 via cexp in case sincos is not available.
2733 2007-01-19 Richard Guenther <rguenther@suse.de>
2735 * doc/tm.texi (TARGET_HAS_SINCOS): Document new target macro.
2736 * defaults.h (TARGET_HAS_SINCOS): Default to off.
2737 * config/linux.h (TARGET_HAS_SINCOS): Set to on if we have glibc.
2738 * config/alpha/linux.h (TARGET_HAS_SINCOS): Likewise.
2739 * config/sparc/linux.h (TARGET_HAS_SINCOS): Likewise.
2740 * config/sparc/linux64.h (TARGET_HAS_SINCOS): Likewise.
2741 * config/rs6000/linux.h (TARGET_HAS_SINCOS): Likewise.
2742 * config/rs6000/linux64.h (TARGET_HAS_SINCOS): Likewise.
2744 2007-01-19 Uros Bizjak <ubizjak@gmail.com>
2746 * config/i386/i386.md (*fpatanxf3_i387, fpatan_extend<mode>xf3_i387):
2748 (atan2sf3_1, atan2df3_1, atan2xf3_1): Remove insn patterns.
2749 (atan2xf3): Directly generate RTL pattern.
2750 (atan2<mode>3): Rename from atan2sf3 and atan2df3 and macroize insn
2751 patterns using X87MODEF12 mode macro. Use fpatan_extend<mode>xf3_i387
2752 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
2753 disable patterns for SSE math.
2754 (atan<mode>2): Rename from atansf2 and atandf2 and macroize insn
2755 patterns using X87MODEF12 mode macro. Use fpatan_extend<mode>xf3_i387
2756 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
2757 disable patterns for SSE math.
2759 2007-01-19 Alexandre Oliva <aoliva@redhat.com>
2761 * libgcc-std.ver: Fix typo in %inherit for GCC_4.3.0.
2763 2007-01-18 Roger Sayle <roger@eyesopen.com>
2765 * fold-const.c (fold_unary) <VIEW_CONVERT_EXPR>: Optimize away a
2766 VIEW_CONVERT_EXPR to the same type as it's operand.
2768 2007-01-18 David Edelsohn <edelsohn@gnu.org>
2770 * config/rs6000/darwin-ldouble.c: Only build _SOFT_FLOAT if
2771 configured for long double 128.
2773 2007-01-18 Mike Stump <mrs@apple.com>
2775 * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Fix build
2778 2007-01-18 Michael Meissner <michael.meissner@amd.com>
2780 * i386.c (ix86_compute_frame_layout): Make fprintf's in #if 0 code
2783 2007-01-18 Jan Hubicka <jh@suse.cz>
2785 * tree-ssa-operands.c (vop_free_bucket_size): Never return value
2786 greater than NUM_VOP_FREE_BUCKETS.
2788 2007-01-18 Daniel Berlin <dberlin@dberlin.org>
2790 * tree-ssa-structalias.c: Update comments.
2791 (ptabitmap_obstack): Removed.
2793 (oldpta_obstack): Ditto.
2794 (stats): Add a few members.
2795 (struct variable_info): Remove node, complex, address_taken, and
2796 indirect_target members. Add oldsolution member.
2797 (new_var_info): Do not initialize removed members.
2798 (constraint_expr_type): Remove INCLUDES.
2799 (constraint_graph): Add size, implicit_preds, rep,
2800 indirect_cycles, eq_rep, label, direct_nodes, and complex members.
2801 (FIRST_REF_NODE): New macro.
2802 (LAST_REF_NODE): Ditto.
2803 (FIRST_ADDR_NODE): Ditto.
2804 (find): New function.
2806 (dump_constraint): Do not handle INCLUDES.
2807 (insert_into_complex): Do not insert duplicate constraints.
2808 (condense_varmap_nodes): Renamed and rewritten into ...
2809 (merge_node_constraints): This. Also fix bug in handling of
2810 offseted copy constraints.
2811 (clear_edges_for_node): No longer need to deal with preds at all,
2812 or removing associated preds/succs.
2813 (merge_graph_nodes): Deal with indirect_cycles.
2814 Don't deal with predecessors.
2815 (add_implicit_graph_edge): New function.
2816 (add_pred_graph_edge): Ditto.
2817 (add_graph_edge): Don't deal with predecessors.
2818 (build_constraint_graph): Removed.
2819 (build_pred_graph): New function.
2820 (build_succ_graph): Ditto.
2821 (struct scc_info): Removed in_component. Added roots, dfs, and
2822 node_mapping. Remove visited_index, unification_queue.
2823 (scc_visit): Deal with union-find we do now.
2824 Deal with cycles with REF nodes.
2825 (collapse_nodes): Renamed and rewritten to ...
2826 (unify_nodes): This.
2827 (process_unification_queue): Removed.
2828 (topo_visit): Cleanup
2829 (do_da_constraint): Use find.
2830 (do_sd_constraint): Ditto.
2831 (do_ds_constraint): Ditto.
2832 (do_complex_constraint): Ditto.
2833 (init_scc_info): Update for removed and added members.
2834 (find_and_collapse_graph_cycles): Renamed and rewritten into ...
2835 (find_indirect_cycles): This.
2836 (equivalence_class): New variable.
2837 (label_visit): New function.
2838 (perform_variable_substitution): Rewritten.
2839 (free_var_substitution_info): New function.
2840 (find_equivalent_node): Ditto.
2841 (move_complex_constraints): Ditto.
2842 (eliminate_indirect_cycles): Ditto.
2843 (solve_graph): Only propagate changed bits.
2844 Use indirect cycle elimination.
2846 (tree_id_t): Rename to tree_vi_t, delete id member, add vi member.
2847 (tree_id_eq): Renamed to ...
2848 (tree_vi_eq): This. Update for member change
2849 (insert_id_for_tree): Renamed and rewritten to ...
2850 (insert_vi_for_tree): This.
2851 (lookup_id_for_tree): Renamed and rewritten to ...
2852 (lookup_vi_for_tree): This.
2853 (get_id_for_tree): Renamed and rewritten to ...
2854 (get_vi_for_tree): Ditto.
2855 (get_constraint_exp_from_ssa_var): Update to use get_vi_for_tree.
2856 (process_constraint): Don't handle INCLUDES.
2857 Remove special ADDRESSOF case.
2858 (find_func_aliases): Rewrite to use vi functions instead of id
2860 (create_function_info_for): Ditto.
2861 (create_variable_info_for): Ditto.
2862 (intra_create_variable_infos): Ditto.
2863 (merge_smts_into): Ditto.
2864 (find_what_p_points_to): Ditto.
2865 (init_base_vars): Ditto.
2866 (init_alias_vars): Ditto.
2867 (remove_preds_and_fake_succs): New function.
2868 (dump_sa_points_to_info): Dump new stats.
2869 (dump_solution_for_var): Use find.
2870 (set_used_smts): Fix formatting.
2871 (compute_points_to_sets): Updated for new functions.
2872 (ipa_pta_execute): Ditto.
2874 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
2875 Richard Sandiford <richard@codesourcery.com>
2877 * doc/tm.texi (TARGET_FUNCTION_VALUE): Expand documentation of
2879 * calls.c (expand_call): If the return value is a PARALLEL,
2880 extract its first member.
2881 * config/m68k/linux.h (FUNCTION_EXTRA_EPILOGUE): Remove.
2882 * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
2883 use FUNCTION_EXTRA_EPILOGUE.
2884 (m68k_function_value): Return a PARALLEL if the return value
2885 is of a pointer type.
2886 * config/m68k/netbsd-elf.h (current_function_returns_pointer)
2887 (FUNCTION_EXTRA_EPILOGUE): Remove.
2888 * config/m68k/m68k.md (D0_REG): New constant.
2890 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
2892 * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
2893 output a NOP for empty epilogues.
2895 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2897 * config/m68k/m68k.c (m68k_use_return_insn): Update comments
2898 before function. Extend register save check to include all
2899 registers, not just integer ones.
2901 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
2903 * config/m68k/m68k-protos.h (use_return_insn): Rename to...
2904 (m68k_use_return_insn): ...this.
2905 * config/m68k/m68k.h (USE_RETURN_INSN): Delete.
2906 * config/m68k/m68k.c (use_return_insn): Rename to...
2907 (m68k_use_return_insn): ...this.
2908 * config/m68k/m68k.md (return): Use m68k_use_return_insn instead
2911 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
2913 * config/m68k/fpgnulib.c (__truncdfsf2): Implement round to
2914 nearest even, fix denormal rounding overflow.
2916 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2918 * config/m68k/m68k.md (movsf_cf_hard): Use fsmove instead of
2919 f%$smove and f%$move.
2920 (movdf_cf_hard): Use fdmove for cases 0 and 3 and fmove for case 1.
2921 (extendsfdf2_cf): Use fdmove instead of f%&move.
2922 (truncdfsf2_cf): Use fsmove instead of f%$smove.
2923 (add<mode>3_cf, sub<mode>3_cf): Use <FP:prec> instead of <FP:round>.
2925 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2927 * config/m68k/m68k.md (movdf_cf_hard): Use output_move_double for
2930 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2932 * real.h (real_format): Add a canonical_nan_lsbs_set field.
2933 (coldfire_single_format): Declare.
2934 (coldfire_double_format): Likewise.
2935 * real.c (encode_ieee_single): Use canonical_nan_lsbs_set instead
2936 of qnan_msb_set to determine the lower bits of a canonical
2938 (encode_ieee_double): Likewise.
2939 (encode_ieee_quad): Likewise.
2940 (ieee_single_format): Initialize canonical_nan_lsbs_set.
2941 (mips_single_format): Likewise.
2942 (ieee_double_format): Likewise.
2943 (mips_double_format): Likewise.
2944 (ieee_extended_motorola_format): Likewise.
2945 (ieee_extended_intel_96_format): Likewise.
2946 (ieee_extended_intel_128_format): Likewise.
2947 (ieee_extended_intel_96_round_53_format): Likewise.
2948 (ibm_extended_format): Likewise.
2949 (mips_extended_format): Likewise.
2950 (ieee_quad_format): Likewise.
2951 (mips_quad_format): Likewise.
2952 (vax_f_format): Likewise.
2953 (vax_d_format): Likewise.
2954 (vax_g_format): Likewise.
2955 (i370_single_format): Likewise.
2956 (i370_double_format): Likewise.
2957 (decimal_single_format): Likewise.
2958 (decimal_double_format): Likewise.
2959 (decimal_quad_format): Likewise.
2960 (c4x_single_format): Likewise.
2961 (c4x_extended_format): Likewise.
2962 (real_internal_format): Likewise.
2963 (coldfire_single_format): New real_format.
2964 (coldfire_double_format): Likewise.
2965 * config/pdp11/pdp11.c (pdp11_f_format): Initialize
2966 canonical_nan_lsbs_set.
2967 (pdp11_d_format): Likewise.
2968 * config/m68k/m68k.c (override_options): Override REAL_FORMAT_MODE
2969 if TARGET_COLDFIRE_CPU.
2971 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2973 * config/m68k/m68k-protos.h (m68k_output_pic_call): Delete.
2974 (output_call, m68k_legitimize_call_address): Declare.
2975 * config/m68k/m68k.h (EXTRA_CONSTRAINT): Remove unnecessary
2976 parenthesees. Add support for a 'W' constraint.
2977 (LEGITIMATE_PIC_OPERAND_P): Remove SYMBOL_REF_FLAG handling.
2978 (PRINT_OPERAND_PUNCT_VALID_P): Remove comment about 'o'.
2979 (m68k_symbolic_call, m68k_symbolic_jump): Declare.
2980 * config/m68k/m68k.c (m68k_symbolic_call, m68k_symbolic_jump): New
2982 (override_options): Initialize them. Do not set flag_no_function_cse
2983 for TARGET_ID_SHARED_LIBRARY.
2984 (m68k_output_pic_call): Delete.
2985 (m68k_legitimize_call_address): New function.
2986 (print_operand): Remove the %o prefix. Handle the %p prefix.
2987 (output_call): New function.
2988 (m68k_output_mi_thunk): Use m68k_symbolic_jump. Always load the
2989 target address from the GOT if symbolic jumps are not allowed.
2990 * config/m68k/m68k.md (call, general_operand): Do not set
2991 SYMBOL_REF_FLAG. Use m68k_legitimize_call_address instead.
2992 Merge separate flag_pic and !flag_pic define_insns into...
2993 (*call, *call_value): ...these new patterns. Match the address
2994 rather than the containing MEM and require it to be a call_operand.
2995 Use output_call to generate the asm template.
2996 * config/m68k/predicates.md (const_call_operand): New predicate.
2997 (call_operand): Likewise.
2999 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
3001 * config/m68k/m68k.h (REGISTER_MOVE_COST): Simplify definition.
3002 (STACK_GROWS_DOWNWARD): Define to 1.
3003 (FUNCTION_VALUE, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P): Equivocate
3004 comments, emphasizing that these values are only defaults.
3005 * config/m68k/linux.h (LINK_SPEC): Fix formatting in #undef.
3006 * config/m68k/m68k.c (const_method): Remove trailing whitespace.
3008 2007-01-18 Richard Sandiford <richard@codesourcery.com>
3010 * config/m68k/m68k.md (cmpsi): Remove outdated flag_pic handling.
3012 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
3013 Richard Sandiford <richard@codesourcery.com>
3015 * config/m68k/m68k.h (DATA_REGNO_P, ADDRESS_REGNO_P, INT_REGNO_P)
3016 (FP_REGNO_P): New macros.
3017 (REGNO_OK_FOR_INDEX_P, REGNO_OK_FOR_BASE_P, REGNO_OK_FOR_DATA_P)
3018 (REGNO_OK_FOR_FP_P, REG_OK_FOR_INDEX_P, REG_OK_FOR_BASE_P): Use them.
3019 (EH_RETURN_STACKADJ_RTX): Use A0_REG.
3020 * config/m68k/m68k.c (m68k_regno_mode_ok): Use the new REGNO macros.
3022 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
3024 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
3025 (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-rtems*): Add t-floatlib
3027 * config/m68k/t-floatlib: New file, extracting common code from...
3028 * config/m68k/t-m68kbare, config/m68k/t-m68kelf,
3029 * config/m68k/t-uclinux: Here.
3030 * config/m68k/fpgnulib.c: Do not compile extendeed precision
3031 routines on ColdFire targets.
3033 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
3035 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
3036 (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*)
3037 (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-linux*)
3038 (m68k-*-rtems*): Use tm_file rather than m68k/m68k.h and
3039 explicitly set MOTOROLA to 1.
3040 * config/m68k/m68k.h (MOTOROLA): Simplify definition accordingly.
3042 2007-01-18 Richard Sandiford <richard@codesourcery.com>
3043 Nathan Sidwell <nathan@codesourcery.com>
3045 * config/m68k/m68k.h (PCC_STATIC_STRUCT_RETURN): Delete.
3046 (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Add MOTOROLA cases.
3047 * config/m68k/coff.h (REGISTER_PREFIX_MD): Delete.
3048 * config/m68k/m68020-elf.h (LIB_SPEC): Delete.
3049 * config/m68k/m68k-none.h (CC1_SPEC, CPP_SUBTARGET_SPEC): Delete.
3050 * config/m68k/m68kelf.h (IMMEDIATE_PREFIX, REGISTER_PREFIX_MD)
3051 (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Delete.
3052 (NO_DOLLAR_IN_LABEL, PCC_STATIC_STRUCT_RETURN): Don't undefine.
3053 (BSS_ASM_OP, ASM_OUTPUT_SKIP, ASM_OUTPUT_ASCII): Delete.
3054 * config/m68k/m68kemb.h (PCC_STATIC_STRUCT_RETURN): Don't undefine.
3055 * config/m68k/linux.h (SIZE_TYPE, PTRDIFF_TYPE, WCHAR_TYPE)
3056 (WCHAR_TYPE_SIZE, TARGET_OBJFMT_CPP_BUILTINS): Delete.
3057 (TARGET_OS_CPP_BUILTINS): Don't define mc68000 and mc68020 here.
3058 (DBX_REGISTER_NUMBER): Delete.
3059 * config/m68k/netbsd-elf.h (IMMEDIATE_PREFIX): Delete.
3060 (PCC_STATIC_STRUCT_RETURN): Don't undefine.
3061 * config/m68k/openbsd.h (PCC_STATIC_STRUCT_RETURN): Define.
3063 2007-01-18 Richard Sandiford <richard@codesourcery.com>
3065 * config.gcc (m68k-*-uclinux*): Add flat.h to $tm_file.
3066 * config/flat.h: New file.
3067 * crtstuff.c (USE_PT_GNU_EH_FRAME): Don't define if
3069 * config/m68k/m68k.h (ASM_PREFERRED_EH_DATA_FORMAT): Do not use
3070 indirect references for -msep-data or -mid-shared-library.
3071 Do not use PC-relative code addresses either.
3072 * config/m68k/m68k.c (override_options): Restrict -fPIC error
3074 * config/m68k/uclinux.h (STARTFILE_SPEC): Define. Use Scrt1.o
3075 for shared libraries and crt1.o for executables. Use crti.o and
3077 (ENDFILE_SPEC): Use crtend.o and crtn.o.
3078 (LIB_SPEC): Suppress -Rlibc.gdb if -static-libc is given.
3079 Do not add -elf2flt or -shared-lib-id options here.
3080 (LINK_SPEC): Define. Pass -elf2flt if no -elf2flt option is given.
3081 Pass -shared-lib-id if -mid-shared-library, taking the library
3082 identifier from -mshared-library-id if given, otherwise
3084 (EH_FRAME_IN_DATA_SECTION): Do not undefine.
3085 (INIT_SECTION_ASM_OP, FINI_SECTION_ASM_OP): Likewise.
3086 (TARGET_OS_CPP_BUILTINS): Define __GXX_MERGED_TYPEINFO_NAMES=0
3087 and __GXX_TYPEINFO_EQUALITY_INLINE=0 if -mid-shared-library.
3088 (DRIVER_SELF_SPECS): Map unadorned PIC options to -msep-data.
3089 * config/m68k/t-uclinux (EXTRA_MULTILIB_PARTS): Add crtbegin.o
3091 * config/m68k/lb1sf68.asm (PICCALL): Use an lea and pc-relative
3092 jump sequence for ISA A and ISA A+.
3093 (PICJUMP): Likewise.
3095 2007-01-18 Richard Sandiford <richard@codesourcery.com>
3097 * config.gcc (m68k-*-uclinux*): Base the port on the common
3098 and m68k GNU/Linux files rather than on the generic ELF ones.
3099 * config/m68k/uclinux.h (TARGET_VERSION): Override.
3100 (TARGET_OS_CPP_BUILTINS): Use LINUX_TARGET_OS_CPP_BUILTINS.
3102 2007-01-18 Julian Brown <julian@codesourcery.com>
3103 Richard Sandiford <richard@codesourcery.com>
3105 * config/m68k/m68k.h (LONG_DOUBLE_TYPE_SIZE): Make 64-bit on ColdFire.
3106 (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
3107 * config/m68k/netbsd-elf.h (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Undefine
3109 * config/m68k/uclinux-oldabi.h (LONG_DOUBLE_TYPE_SIZE): Redefine to
3111 (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
3113 2007-01-18 Richard Sandiford <richard@codesourcery.com>
3115 * doc/install.texi: Document m68k-uclinuxoldabi.
3116 * config.gcc (m68k-*-uclinuxoldabi): New configuration.
3117 * config/m68k/uclinux-oldabi.h: New file, copied from
3118 config/m68k/uclinux.h.
3120 2007-01-18 Kaz Kojima <kkojima@rr.iij4u.or.jp>
3122 * config/m32r/m32r-protos.h (m32r_expand_epilogue): Declare it.
3123 * config/m32r/m32r.c (m32r_setup_incoming_varargs): Use gen_frame_mem.
3124 (m32r_compute_frame_size): Use unsigned for regno.
3125 (m32r_reload_lr): Use gen_frame_mem.
3127 (m32r_output_function_epilogue): Don't output the function epilogue
3129 (m32r_expand_epilogue): New.
3130 (direct_return): Return false if the function has the interrupt
3132 (m32r_hard_regno_rename_ok): Remove code for the textual epilogue.
3133 * config/m32r/m32r.md (epilogue): New expander.
3134 (return_lr, return_rte): New insns.
3135 (return): Make it expander.
3136 (return_normal): New expander.
3138 2007-01-18 Josh Conner <jconner@apple.com>
3141 * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Add
3142 support for UNLE, UNLT, UNGE, and UNGT.
3144 2007-01-18 Jan Hubicka <jh@suse.cz>
3146 * tree-vrp.c (finalize_jump_threads): Do not call cleanup_cfg by hand.
3147 * tree-tailcall (add_virtual_phis): Likewise.
3148 (optimize_tail_call): Return TODOs.
3149 (execute_tail_calls): Return TODOs.
3150 * tree-ssa-ccp (execute_fold_all_builtins): Do cleanup_cfg via TODO.
3151 * tree-cfgcleanup.c (cleanup_tree_cfg_loop): Return if something
3153 * tree-ssa-forwprop.c (tree_ssa_forward_propagate_single_use_value):
3154 Cleanup cfg using TODO.
3155 * tree-flow.h (cleanup_tree_cfg_loop): Update prototype.
3156 * passes.c (execute_function_todo): When cleanup did something, remove
3158 * tree-cfg.c (pass_build_cfg): Add cleanup_cfg TODO.
3159 (make_edges): Don't cleanup_cfg.
3161 2007-01-18 Uros Bizjak <ubizjak@gmail.com>
3163 * reg-stack.c (subst_stack_regs_pat) [UNSPEC_SINCOS_COS,
3164 UNSPEC_XTRACT_FRACT]: Use generic code for instructions that
3165 operate on the top of stack.
3166 [UNSPEC_SINCOS_SIN, UNSPEC_XTRACT_EXP, UNSPEC_TAN]: Rewrite
3167 register handling of instructions that output to the second
3169 [UNSPEC_TAN_ONE, UNSPEC_TAN_TAN]: Remove.
3170 (move_for_stack_reg): Special-case check for dead destination
3171 stack slot for constant load of 1.0 inside UNSPEC_TAN.
3173 * config/i386/i386.md (UNSPEC_TAN): New constant.
3174 (UNSPEC_TAN_ONE, UNSPEC_TAN_TAN): Remove.
3175 (fptanxf4_i387, fptan_extend<mode>xf4_i387): New patterns
3176 to correctly model move of constant 1.0 to top stack slot.
3177 (*tandf3_1, *tansf3_1, *tanxf3_1): Remove insn patterns.
3178 (unnamed peephole2 pattern): Remove corresponding peephole2
3179 pattern that optimizes tan insn and loading of constant 1.0.
3180 (tanxf2): Use fptanxf4_i387.
3181 (tan<mode>2): Rename from tansf2 and tandf2 and macroize insn
3182 patterns using X87MODEF12 mode macro. Use fptan_extend<mode>xf4_i387
3183 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
3184 disable patterns for SSE math.
3185 (sincos<mode>3): Use truncxf<mode>2_i387_noop for truncation.
3186 (fyl2x_extend<mode>xf3_i387): Use X87MODEF12 for operand 1.
3188 2007-01-18 Dirk Mueller <dmueller@suse.de>
3189 Richard Guenther <rguenther@suse.de>
3192 * doc/invoke.texi (Warray-bounds): Document -Warray-bounds.
3193 * common.opt (Warray-bounds): Add new warning option.
3194 * c-opts.c (c_common_handle_option): Define -Warray-bounds
3196 * Makefile.in: make tree-vrp.o depend on toplev.h
3197 * tree-vrp.c (vrp_finalize): Call check_array_refs if -Warray-bounds
3199 (check_array_refs, check_array_bounds, check_array_ref): New.
3201 2007-01-18 Jan Hubicka <jh@suse.cz>
3203 * tree-ssa-ccp.c (ccp_finalize): Return if something changed.
3204 (execute_ssa_ccp): Return flags conditionally.
3205 * tree-ssa-propagate.c (substitue_and_fold): Return if something was
3207 * tree-ssa-propagate.h (substitute_and_fold): Update prototype.
3209 2007-01-18 Steven Bosscher <steven@gcc.gnu.org>
3211 * cfgcleanup.c (cleanup_cfg): Detect cfglayout mode and set
3212 the CLEANUP_CFGLAYOUT flag when in cfglayout mode.
3214 * Makefile.c (GTFILES): Add cfglayout.h.
3215 * gengtype.c (open_base_files): Likewise.
3216 * cfglayout.c (cfg_layout_function_footer,
3217 cfg_layout_function_header) Reindent to make gengtype happy.
3218 * cfglayout.h (cfg_layout_function_footer,
3219 cfg_layout_function_header): Add GTY(()) marker.
3221 * ifcvt.c (noce_try_sign_mask): Make sure INSN_B is non-null.
3223 2007-01-18 Ben Elliston <bje@au.ibm.com>
3225 * genautomata.c (write_automata): Include xstrerror output in the
3226 error message if writing the DFA description file fails.
3228 2007-01-17 H.J. Lu <hongjiu.lu@intel.com>
3230 * config/mips/mips-protos.h (mips_output_external): Make it
3232 * config/mips/iris.h (TARGET_ASM_EXTERNAL_LIBCALL): Removed.
3233 * config/mips/mips.c (irix_output_external_libcall): Likewise.
3234 (extern_list): Likewise.
3235 (extern_head): Likewise.
3236 (TARGET_ASM_FILE_END): Likewise.
3237 (mips_file_end): Likewise.
3238 (mips_output_external): Rewritten.
3240 2007-01-18 Ben Elliston <bje@au.ibm.com>
3242 * genpreds.c (write_insn_preds_c): Only write out the function
3243 body for regclass_for_constraint if we have register constraints.
3245 2007-01-17 Tom Tromey <tromey@redhat.com>
3247 * doc/sourcebuild.texi (libgcj Tests): Use sourceware.org.
3248 * doc/install.texi (Testing): Use sourceware.org.
3249 (Binaries): Likewise.
3250 (Specific): Likewise.
3251 * doc/contrib.texi (Contributors): Use sourceware.org.
3253 2007-01-17 Anatoly Sokolov <aesok@post.ru>
3255 * config/avr/avr.h (AVR_HAVE_LPMX): New macro.
3256 (AVR_ENHANCED): Rename to ...
3257 (AVR_HAVE_MUL): ... new.
3258 (avr_enhanced_p): Rename to ...
3259 (avr_have_mul_p): ... new.
3260 (TARGET_CPU_CPP_BUILTINS): Use 'avr_have_mul_p' instead of
3261 'avr_enhanced_p' for "__AVR_ENHANCED__". Define "__AVR_HAVE_MUL__".
3262 * config/avr/avr.c (avr_enhanced_p): Rename to ...
3263 (avr_have_mul_p): ... new.
3264 (base_arch_s): Rename 'enhanced' to 'have_mul'.
3265 (avr_override_options): Use 'avr_have_mul_p' and 'have_mul' instead of
3266 'avr_enhanced_p' and 'enhanced'.
3267 (ashlhi3_out, ashrhi3_out, lshrhi3_out, avr_rtx_costs): Use
3268 AVR_HAVE_MUL instead of AVR_ENHANCED.
3269 * avr.md (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
3270 (mulqi3, *mulqi3_enh, *mulqi3_call, mulqihi3, umulqihi3, mulhi3,
3271 *mulhi3_enh, *mulhi3_call, mulsi3, *mulsi3_call): Use AVR_HAVE_MUL
3272 instead of AVR_ENHANCED.
3273 (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
3274 * libgcc.S: Use __AVR_HAVE_MUL__ instead of __AVR_ENHANCED__.
3275 (__tablejump__): Use __AVR_HAVE_LPMX__ instead of __AVR_ENHANCED__.
3277 2007-01-17 Ian Lance Taylor <iant@google.com>
3279 * vec.h (VEC_reserve_exact): Define.
3280 (vec_gc_p_reserve_exact): Declare.
3281 (vec_gc_o_reserve_exact): Declare.
3282 (vec_heap_p_reserve_exact): Declare.
3283 (vec_heap_o_reserve_exact): Declare.
3284 (VEC_OP (T,A,reserve_exact)): New static inline function, three
3286 (VEC_OP (T,A,reserve)) [all versions]: Remove handling of
3288 (VEC_OP (T,A,alloc)) [all versions]: Call ...reserve_exact.
3289 (VEC_OP (T,A,copy)) [all versions]: Likewise.
3290 (VEC_OP (T,a,safe_grow)) [all versions]: Likewise.
3291 * vec.c (calculate_allocation): Add exact parameter. Change all
3293 (vec_gc_o_reserve_1): New static function, from vec_gc_o_reserve.
3294 (vec_gc_p_reserve, vec_gc_o_reserve): Call vec_gc_o_reserve_1.
3295 (vec_gc_p_reserve_exact, vec_gc_o_reserve_exact): New functions.
3296 (vec_heap_o_reserve_1): New static function, from vec_heap_o_reserve.
3297 (vec_heap_p_reserve, vec_heap_o_reserve): Call vec_heap_o_reserve_1.
3298 (vec_heap_p_reserve_exact): New function.
3299 (vec_heap_o_reserve_exact): New function.
3301 2007-01-17 Jan Hubicka <jh@suse.cz>
3303 * ipa-type-escape.c (look_for_casts): Revamp using handled_component_p.
3305 2007-01-17 Eric Christopher <echristo@apple.com>
3307 * config.gcc: Support core2 processor.
3309 2007-01-16 Jan Hubicka <jh@suse.cz>
3311 * tree-ssanames.c (release_dead_ssa_names): Instead of ggc_freeing
3312 the names, just unlink the chain so we don't crash on dangling pointers
3315 2007-01-16 Jan Hubicka <jh@suse.cz>
3317 * cgraph.h (cgraph_decide_inlining_incrementally): Kill.
3318 * tree-pass.h: Reorder to make IPA passes appear toegher.
3319 (pass_early_inline, pass_inline_parameters, pass_apply_inline): Declare.
3320 * cgraphunit.c (cgraph_finalize_function): Do not compute inling
3321 parameters, do not call early inliner.
3322 * ipa-inline.c: Update comments. Include tree-flow.h
3323 (cgraph_decide_inlining): Do not compute inlining parameters.
3324 (cgraph_decide_inlining_incrementally): Return TODOs; assume to
3325 be called with function context set up.
3326 (pass_ipa_inline): Remove unreachable functions before pass.
3327 (cgraph_early_inlining): Simplify assuming to be called from the
3329 (pass_early_inline): New pass.
3330 (cgraph_gate_ipa_early_inlining): New gate.
3331 (pass_ipa_early_inline): Turn into simple wrapper.
3332 (compute_inline_parameters): New function.
3333 (gate_inline_passes): New gate.
3334 (pass_inline_parameters): New pass.
3335 (apply_inline): Move here from tree-optimize.c
3336 (pass_apply_inline): New pass.
3337 * ipa.c (cgraph_remove_unreachable_nodes): Verify cgraph after
3339 * tree-inline.c (optimize_inline_calls): Return TODOs rather than
3341 (tree_function_versioning): Do not allocate dummy struct function.
3342 * tree-inline.h (optimize_inline_calls): Update prototype.
3343 * tree-optimize.c (execute_fixup_cfg): Export.
3344 (pass_fixup_cfg): Remove
3345 (tree_rest_of_compilation): Do not apply inlines.
3346 * tree-flow.h (execute_fixup_cfg): Declare.
3347 * Makefile.in (gt-passes.c): New.
3348 * passes.c: Include gt-passes.h
3349 (init_optimization_passes): New passes.
3350 (nnodes, order): New static vars.
3351 (do_per_function_toporder): New function.
3352 (execute_one_pass): Dump current pass here.
3353 (execute_ipa_pass_list): Don't dump current pass here.
3355 2007-01-16 Janis Johnson <janis187@us.ibm.com>
3357 * config/dfp-bit.c (dfp_compare_op): Return separate value for NaN.
3358 (DFP_NE, DFP_LE, DFP_GE): Return false for NaN.
3360 2007-01-16 David Edelsohn <edelsohn@gnu.org>
3362 * config/rs6000/darwin-ldouble.c: Build file for SOFT_FLOAT.
3363 (strong_alias): Define.
3364 (__gcc_qmul): Provide non-FMA for soft-float.
3370 (__gcc_qunord): New.
3380 * config/rs6000/rs6000.c (rs6000_init_libfuncs): Initialize
3381 soft-float functions.
3382 * config/rs6000/libgcc-ppc-glibc.ver: Version soft-float symbols.
3383 * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Do not warn
3384 about long double soft float.
3386 2007-01-16 Dorit Nuzman <dorit@il.ibm.com>
3387 Tehila Meyzels <tehila@il.ibm.com>
3389 * tree-vectorizer.h (is_pattern_stmt_p): New.
3390 * tree-vect-analyze.c (vect_determine_vectorization_factor): Fix
3391 formatting (tabs instead of spaces). Cleanup and clarify setting
3392 of STMT_VINFO_VECTYPE. Call is_pattern_stmt_p.
3393 * tree-vect-transform.c (vect_get_vec_def_for_operand): Fix typo.
3394 (vectorizable_type_demotion): Check that types are integral.
3395 (vectorizable_type_promotion): Likewise.
3396 (vectorizable_store): Fix typo. Eliminate new-line at end of
3399 2007-01-16 Jan Hubicka <jh@suse.cz>
3401 * tree-ssanames.c (release_dead_ssa_names): Remove invalidated
3404 2007-01-15 Eric Christopher <echristo@apple.com>
3406 * ifcvt.c: Include vec.h, vecprim.h.
3407 (check_cond_move_block): New argument regs.
3408 Reorganize. Add registers used to regs.
3409 (cond_move_process_if_block): Use regs set above as
3412 2007-01-15 Eric Christopher <echristo@apple.com>
3414 * config/darwin.h: Update copyright.
3415 (TARGET_OPTION_TRANSLATE_TABLE): Add umbrella.
3416 (LINK_COMMAND_SPEC): Add -u.
3417 (LINK_SPEC): Fix umbrella for above.
3419 2007-01-15 Joseph S. Myers <joseph@codesourcery.com>
3421 * config/soft-fp/op-common.h, config/soft-fp/op-4.h: Update from
3424 2007-01-15 Tom Tromey <tromey@redhat.com>
3426 * doc/sourcebuild.texi (libgcj Tests): Don't mention jacks.
3427 * doc/install.texi (Testing): Don't mention jacks.
3428 (Configuration): Document --enable-java-maintainer-mode. Move
3429 --with-java-home to libgcj-specific section. Document
3431 (Prerequisites): Mention --enable-java-maintainer-mode, ecj1.
3433 2007-01-15 Jan Hubicka <jh@suse.cz>
3435 * tree-ssa-dce.c (DCE_TODOs): New.
3436 (propagate_necessity): Return if something changed.
3437 (eliminate_unnecessary_stmts): Likewise.
3438 (perform_tree_ssa_dce): Return TODO flags when needed.
3439 (pass_dce, pass_dce_loop, pass_cd_dce): Remove TODO flags.
3441 2007-01-15 Uros Bizjak <ubizjak@gmail.com>
3443 * config/i386/i386.md (fyl2xxf3_i387): Rename from fyl2x_xf3.
3444 (fyl2x_extend<mode>xf3_i387): New insn pattern.
3445 (log<mode>2): Rename from logsf2 and logdf2 and macroize insn
3446 patterns using X87MODEF12 mode macro. Extend operand 1
3447 to XFmode. Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
3448 (log10<mode>2): Ditto.
3449 (log2<mode>2): Ditto.
3450 (log1p<mode>2): Ditto.
3451 (logb<mode>2): Ditto.
3452 (fyl2xp1xf3_i387): Rename from fyl2xp1_xf3.
3453 (fyl2xp1_extend<mode>xf3_i387): New insn pattern.
3454 (*fxtractxf3_i387): Rename from *fxtractxf3.
3455 (fxtract_extend<mode>xf3_i387): New insn pattern.
3456 (ilogbsi2): Use match_dup 3, not match_operand:XF 3.
3457 * config/i386/i386.c (ix86_emit_i387_log1p): Use gen_fyl2xp1xf3_i387()
3458 and gen_fyl2xxf3_i387().
3460 2007-01-14 Zdenek Dvorak <dvorakz@suse.cz>
3462 * loop-unswitch.c (unswitch_loop): Do not call fix_loop_placement.
3463 * cfgloopmanip.c (fix_loop_placement): Made static. Use
3464 get_loop_exit_edges. Changed return type to bool.
3465 * cfgloop.h (fix_loop_placement): Declaration removed.
3467 2007-01-14 Dorit Nuzman <dorit@il.ibm.com>
3469 * param.h (MIN_VECT_LOOP_BOUND): New.
3470 * params.def (MIN_VECT_LOOP_BOUND): New.
3471 * tree-vectorizer.c (slpeel_tree_peel_loop_to_edge): Takes another
3472 argument - minimum threshold for number of iterations.
3473 * tree-vectorizer.h (slpeel_tree_peel_loop_to_edge): Add another
3474 argument to declaration.
3475 * tree-vect-analyze.c (vect_analyze_operations): Check value of
3476 MIN_VECT_LOOP_BOUND.
3477 * tree-vect-transform.c (vect_do_peeling_for_loop_bound): Call
3478 slpeel_tree_peel_loop_to_edge with additional argument.
3479 (vect_do_peeling_for_alignment): Likewise.
3480 * doc/invoke.texi (min-vect-loop-bound): Document new param option.
3482 2007-01-14 Uros Bizjak <ubizjak@gmail.com>
3485 * config/i386/i386.c (print_operand) ['z']: Output 'b' for
3488 2007-01-14 Jan Hubicka <jh@suse.cz>
3490 * tree-dfa.c (remove_referenced_var): New function.
3491 * tree-ssa-live.c (remove_unused_locals): Walk referenced vars and
3492 prune referenced vars list too.
3493 * tree-flow.h (remove_referenced_var): Declare.
3495 2007-01-14 Jan Hubicka <jh@suse.cz>
3497 * tree-eh.c (add_stmt_to_eh_region_fn): Do not add call_exprs
3499 (remove_stmt_from_eh_region_fn): Do not remove call_exprs.
3500 (verify_eh_throw_stmt_node, verify_eh_throw_table_statements): Kill.
3501 * except.h (verify_eh_throw_table_statements): Kill prototype.
3502 * cfgexpand.c (expand_gimple_basic_block): Propagate Eh regions
3503 into call exrepssions.
3504 * tree-optimize.c (execute_free_cfg_annotatiosn): Do not call
3506 * tree-cfg.c: Include pointer-set.h.
3507 (verify_node_sharing): Work on pointer set.
3508 (verify_eh_throw_stmt_node): New.
3509 (verify_stmts): Use pointers sets, verify throw_stmt.
3511 2007-01-13 Zdenek Dvorak <dvorakz@suse.cz>
3513 * ipa-reference.c (analyze_function): Consider also addresses taken
3516 2007-01-12 Roger Sayle <roger@eyesopen.com>
3518 * c-typeck.c (null_pointer_constant_p): Replace use of
3519 TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
3520 (build_c_cast): Likewise.
3522 2007-01-12 Roger Sayle <roger@eyesopen.com>
3524 * tree.h (force_fit_type_double): Remove unused final argument.
3525 * c-common.c (constant_expression_warning): Replace use of
3526 TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
3527 (convert_and_check): Likewise.
3528 (shorten_compare): Update call to force_fit_type_double.
3529 (c_common_truthvalue_conversion) <INTEGER_CST>: Use integer_zerop.
3530 * convert.c (convert_to_pointer): Update call to
3531 force_fit_type_double.
3532 * fold-const.c (force_fit_type_double): Remove overflowed_const
3534 (int_const_binop, fold_convert_const_int_from_int,
3535 fold_convert_const_int_from_real, fold_div_compare,
3536 fold_sign_changed_comparison, fold_unary, fold_negate_const,
3537 fold_abs_const, fold_not_const): Remove the final argument from
3538 calls to force_fit_type_double.
3540 2007-01-12 Andrew Pinski <andrew_pinski@playstation.sony.com>
3542 * configure.ac: Set insn to "nop" for spu-*-* also.
3543 * configure: Regenerate.
3545 2007-01-12 Olga Golovanevsky <olga@il.ibm.com>
3547 * builtins.def : Add BUILT_IN_FREE.
3549 2007-01-12 Jan Hubicka <jh@suse.cz>
3551 PR tree-optimization/30443
3552 * tree-inline.c (tree_function_versioning): Do not optimize when
3553 cloning for inlining.
3555 2007-01-12 Zdenek Dvorak <dvorakz@suse.cz>
3557 * doc/loop.texi: Document recording of loop exits.
3558 * cfgloopmanip.c (loopify, duplicate_loop): Use alloc_loop.
3559 (update_single_exits_after_duplication,
3560 update_single_exit_for_duplicated_loop,
3561 update_single_exit_for_duplicated_loops): Removed.
3562 (duplicate_loop_to_header_edge): Do not call
3563 update_single_exits_after_duplication and
3564 update_single_exit_for_duplicated_loops.
3565 (loop_version): Do not update single_exit information.
3566 (fix_loop_structure): Use record_loop_exits instead of
3567 mark_single_exit_loops.
3568 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update
3569 the lists of loop exits.
3570 * cfghooks.c (redirect_edge_and_branch, redirect_edge_and_branch_force,
3571 split_edge, merge_blocks): Update the lists of loop exits.
3572 * modulo-sched.c (sms_schedule): Pass LOOPS_HAVE_RECORDED_EXITS to
3573 loop_optimizer_init.
3574 * loop-init.c (loop_optimizer_init): Call record_loop_exits instead
3575 of mark_single_exit_loops.
3576 (loop_optimizer_finalize): Call release_recorded_exits.
3577 * tree-ssa-loop.c (tree_loop_optimizer_init): Pass
3578 LOOPS_HAVE_RECORDED_EXITS to loop_optimizer_init.
3579 * tree-vectorizer.c (slpeel_tree_duplicate_loop_to_edge_cfg): Do not
3580 update single exit information.
3581 * lambda-code.c (perfect_nestify): Ditto.
3582 * cfgloop.c (flow_loop_free): Destroy the list of exits of the loop.
3583 (mark_single_exit_loops): Removed.
3584 (alloc_loop, loop_exit_hash, loop_exit_eq, loop_exit_free,
3585 get_exit_descriptions, rescan_loop_exit, record_loop_exits,
3586 dump_recorded_exit, dump_recorded_exits, release_recorded_exits): New
3588 (get_loop_exit_edges, single_exit): Use recorded exit lists.
3589 (add_bb_to_loop, remove_bb_from_loops): Update the lists of loop exits.
3590 (verify_loop_structure): Verify consistency of the exit lists.
3591 (flow_loops_find): Use alloc_loop. Initialize exits hash.
3592 (set_single_exit): Removed.
3593 * cfgloop.h (struct loop_exit): New function.
3594 (struct loop): single_exit_ field replaced by exits field.
3595 (LOOPS_HAVE_MARKED_SINGLE_EXITS): Replaced by LOOPS_HAVE_RECORDED_EXITS.
3596 (struct loops): Added exits hash.
3597 (mark_single_exit_loops, set_single_exit): Declaration removed.
3598 (release_recorded_exits, record_loop_exits, rescan_loop_exit): Declare.
3600 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3602 * doc/invoke.texi: Avoid use of @headitem.
3604 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3606 * cse.c (cse_insn): Move HAVE_CC0 code after declarations.
3608 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3610 * doc/install.texi: Fix m68k-*-* anchor and add m68k-*-* to the
3613 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
3614 Richard Sandiford <richard@codesourcery.com>
3616 * doc/invoke.texi: Document -mno-bitfield, -mno-rtd and -mno-short.
3617 * config/m68k/m68k.opt: Resort options.
3618 (mbitfield, mrtd, mshort): Remove RejectNegative properties.
3620 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
3621 Richard Sandiford <richard@codesourcery.com>
3623 * doc/invoke.texi: Document the macros that are defined by
3624 m68k's -mtune and -mhard-float options.
3625 * config/m68k/m68k-protos.h (m68k_cpp_cpu_ident) Declare.
3626 (m68k_cpp_cpu_family): Likewise.
3627 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Add a full set
3628 of __ucfv*__ macros. Define __mcffpu__ if generating code for
3629 ColdFire FPUs. Define __mcf_cpu_* and __mcf_family_* macros.
3630 * config/m68k/m68k.c (m68k_cpp_cpu_ident): New function.
3631 (m68k_cpp_cpu_family): Likewise.
3633 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3635 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Treat all mc68*
3636 macros besides mc68000 as tuning macros. Use a switch statement
3637 to set them and mcpu32.
3639 2007-01-12 Julian Brown <julian@codesourcery.com>
3641 * config/m68k/m68k.h: Use TARGET_68040 instead of TARGET_68040_ONLY.
3642 (TARGET_68040_ONLY): Rename to...
3643 (TARGET_68040): ...this.
3644 * config/m68k/m68k.c: Use TARGET_68040 instead of TARGET_68040_ONLY.
3645 * config/m68k/m68k.md: Likewise.
3647 2007-01-12 Julian Brown <julian@codesourcery.com>
3648 Nathan Sidwell <nathan@codesourcery.com>
3649 Richard Sandiford <richard@codesourcery.com>
3651 * config.gcc (m680[012]0-*-*, m68k*-*-*): Set m68k_cpu_ident to
3652 the -mcpu= argument associated with the --with-cpu setting.
3653 Define M68K_DEFAULT_TUNE to the default -mtune= option,
3654 if different from the one implied by the -mcpu setting.
3655 Accept --with-cpu=FOO if FOO is listed in m68k-devices.def,
3656 using mcpu=FOO as the default CPU option. Set target_cpu_default2.
3657 * doc/invoke.texi: Mention ColdFire in the introduction to the
3658 m68k options. Document the new -march, -mcpu, -mtune, -mdiv,
3659 -mno-div and -mhard-float options. Make -m68881 a synonym for
3660 -mhard-float. Document the previously-undocumented -m5206e,
3661 -m528x, -m5307 and -m5407 options. Tweak the existing option
3662 documentation for consistency.
3663 * doc/install.texi: Mention new --with-cpu arguments.
3664 * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Only use the
3665 default CPU if neither -mcpu nor -march are specified.
3666 (ASM_CPU_SPEC): Pass down -mcpu and -march options.
3667 (TARGET_CPU_CPP_BUILTINS): Set __mcfisa*__ macros from
3668 TARGET_ISA*. Set the legacy __mcf*__ cpu macros in the same way,
3669 using m68k_tune to decide between families that implement the
3670 same ISA. Use m68k_tune to set __mcfv4e__.
3671 (FL_BITFIELD, FL_68881, FL_COLDFIRE, FL_CF_HWDIV, FL_CF_MAC)
3672 (FL_CF_EMAC, FL_CF_EMAC_B, FL_CF_USP, FL_CF_FPU, FL_ISA_68000)
3673 (FL_ISA_68010, FL_ISA_68020, FL_ISA_68040, FL_ISA_A, FL_ISA_B)
3674 (FL_ISA_C, FL_ISA_MMU): New macros.
3675 (MASK_COLDFIRE): Delete.
3676 (TARGET_68010, TARGET_68020, TARGET_68040_ONLY, TARGET_COLDFIRE)
3677 (TARGET_ISAB): Redefine in terms of m68k_cpu_flags.
3678 (TARGET_68881, TARGET_COLDFIRE_FPU): Redefine in terms of m68k_fpu.
3679 (TARGET_HARD_FLOAT): Do not define here.
3680 (TARGET_ISAAPLUS, TARGET_ISAC): New macros.
3681 (TUNE_68000): New macro.
3682 (TUNE_68000_10): Redefine in terms of TUNE_68000 and TUNE_68010.
3683 (TUNE_68010, TUNE_68030, TUNE_68040, TUNE_68060, TUNE_CPU32)
3684 (TUNE_CFV2): Redefine in terms of m68k_tune.
3685 (uarch_type, target_device, fpu_type): New enums.
3686 (m68k_cpu, m68k_tune, m68k_fpu, m68k_cpu_flags): Declare.
3687 * config/m68k/m68k.c (TARGET_DEFAULT): Remove MASK_68881.
3688 (FL_FOR_isa_00, FL_FOR_isa_10, FL_FOR_isa_20, FL_FOR_isa_40)
3689 (FL_FOR_isa_cpu32, FL_FOR_isa_a, FL_FOR_isa_aplus, FL_FOR_isa_b)
3690 (FL_FOR_isa_c): New macros.
3691 (m68k_isa): New enum.
3692 (m68k_target_selection): New structure.
3693 (all_devices, all_isas, all_microarchs): New tables.
3694 (m68k_cpu_entry, m68k_arch_entry, m68k_tune_entry, m68k_cpu)
3695 (m68k_tune, m68k_fpu, m68k_cpu_flags): New variables.
3696 (MASK_ALL_CPU_BITS): Delete.
3697 (m68k_find_selection): New function.
3698 (m68k_handle_option): Handle -mcpu=, -march= and -mtune=.
3699 Map the legacy target options to a combination of the new ones.
3700 (override_options): Set m68k_cpu, m68k_tune, m68k_fpu and
3701 m68k_cpu_flags. Handle M68K_DEFAULT_TUNE. Use m68k_cpu_flags
3702 to derive default MASK_BITFIELD, MASK_CF_HWDIV and MASK_HARD_FLOAT
3704 * config/m68k/m68k.opt (m5200, m5206e, m528x, m5307, m5407, mcfv4e)
3705 (m68010, m68020, m68020-40, m68020-60, m68030, m68040): Remove Mask
3707 (m68881, msoft-float): Change mask from 68881 to HARD_FLOAT.
3708 (march=, mcpu=, mdiv, mhard-float, mtune=): New options.
3709 * config/m68k/m68k-devices.def: New file.
3711 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3712 Nathan Sidwell <nathan@codesourcery.com>
3714 * config/m68k/m68k.h (ASM_CPU_SPEC, ASM_SPEC, EXTRA_SPECS)
3715 (SUBTARGET_EXTRA_SPECS): New macros.
3716 * config/m68k/linux.h (ASM_SPEC): Remove CPU flags;
3717 use %(asm_cpu_spec) instead.
3718 * config/m68k/m68k-none.h (ASM_SPEC): Likewise.
3719 * config/m68k/openbsd.h (ASM_SPEC): Likewise.
3720 * config/m68k/netbsd-elf.h (ASM_SPEC): Likewise.
3721 (EXTRA_SPECS): Rename to...
3722 (SUBTARGET_EXTRA_SPECS): ...this.
3724 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
3725 Richard Sandiford <richard@codesourcery.com>
3726 Julian Brown <julian@codesourcery.com>
3728 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
3729 (m68k-*-uclinux*, m68k-*-linux*, m68k-*-rtems*): Set default_m68k_cpu
3730 to the configuration's default CPU.
3731 (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*): Likewise.
3732 Remove default masks.
3733 (m680[012]0-*-*): Set the default with_cpu to the first part of
3735 (m68k*-*-*): Set the default with_cpu to m$default_m68k_cpu.
3736 (m68k*-*-linux): Extend the --with-cpu handling to...
3737 (m680[012]0-*-*, m68k*-*-*): ...these configurations. Allow m68000
3738 and m68010. Don't set target_cpu_default2.
3739 * doc/install.texi: Document --with-cpu for m68k.
3740 * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Define.
3741 * config/m68k/m68k-none.h (TARGET_CPU_DEFAULT, M68K_CPU_m68k)
3742 (M68K_CPU_m68000, M68K_CPU_m68010, M68K_CPU_m68020, M68K_CPU_m68030)
3743 (M68K_CPU_m68040, M68K_CPU_m68302, M68K_CPU_m68332, TARGET_DEFAULT)
3744 (ASM_CPU_DEFAULT_SPEC, CC1_CPU_DEFAULT_SPEC): Delete.
3745 (ASM_SPEC): Remove use of %(asm_cpu_default).
3746 (EXTRA_SPECS, SUBTARGET_EXTRA_SPECS, MULTILIB_DEFAULTS): Delete.
3747 * config/m68k/linux.h (TARGET_DEFAULT): Delete.
3748 (CPP_SPEC): Merge definitions. Do not handle __HAVE_68881__ here.
3749 * config/m68k/netbsd-elf.h (TARGET_OS_CPP_BUILTINS): Define
3750 __HAVE_FPU__ if TARGET_HARD_FLOAT.
3751 (TARGET_DEFAULT): Delete.
3752 (EXTRA_SPECS): Delete cpp_cpu_default_spec, cpp_cpu_spec,
3753 cpp_fpu_spec, asm_default_spec and netbsd_cpp_spec.
3754 (CPP_CPU_SPEC): Delete.
3755 (TARGET_VERSION): Merge definitions, using TARGET_68010 to pick
3756 the appropriate string.
3757 (CPP_CPU_DEFAULT_SPEC, ASM_DEFAULT_SPEC, CPP_FPU_SPEC): Delete.
3758 (CPP_SPEC): Define to NETBSD_CPP_SPEC.
3759 (ASM_SPEC): Don't use %(asm_default_spec).
3760 * config/m68k/m68k.c (TARGET_DEFAULT_TARGET_FLAGS): Remove
3761 TARGET_DEFAULT and add MASK_68881.
3762 * config/m68k/m68k.md: Remove mention of TARGET_DEFAULT from comments.
3764 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3766 * config.gcc (m68010-*-netbsdelf*): Add MASK_68010.
3767 (m68k*-*-netbsdelf*, m68k*-*-openbsd*, m68k*-linux*): Add
3768 MASK_68010 alongside MASK_68020.
3769 * doc/invoke.texi: Document -m68010.
3770 * config/m68k/m68k.opt (m68010): New.
3771 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Define mc68010
3773 (TUNE_68010): New macro.
3774 * config/m68k/m68k-none.h (M68K_CPU_m68k, M68K_CPU_m68010)
3775 (M68K_CPU_m68020, M68K_CPU_m68030, M68K_CPU_m68040)
3776 (M68K_CPU_m68332): Add MASK_68010.
3777 * config/m68k/linux.h (TARGET_DEFAULT): Add MASK_68010 to
3778 fallback definition.
3779 * config/m68k/netbsd-elf.h (CPP_CPU_SPEC): Remove now-redundant
3781 * config/m68k/m68k.c (MASK_ALL_CPU_BITS): Add MASK_68010.
3782 (m68k_handle_option): Handle OPT_m68010. Add MASK_68010
3783 to all entries that use MASK_68020.
3784 (output_move_simode_const, output_move_himode, output_move_qimode)
3785 (output_move_stricthi, output_move_strictqi): Use TARGET_68010
3786 instead of TARGET_68020 to select clr behavior. Remove comment
3787 about there being no TARGET_68010.
3788 * config/m68k/m68k.md: Likewise throughout.
3790 2007-01-12 Julian Brown <julian@codesourcery.com>
3792 * config/m68k/m68k.h (TARGET_ISAB): New macro.
3793 * config/m68k/m68k.c: Use TARGET_ISAB rather than TARGET_CFV4.
3794 * config/m68k/m68k.md: Likewise.
3796 2007-01-12 Julian Brown <julian@codesourcery.com>
3798 * config/m68k/m68k.h (LEGITIMATE_INDEX_P, LEGITIMIZE_ADDRESS): Use
3799 TARGET_COLDFIRE_FPU instead of TARGET_CFV4E.
3801 2007-01-12 Julian Brown <julian@codesourcery.com>
3803 * config/m68k/m68k.h (TUNE_68040_60): New macro.
3804 * config/m68k/m68k.c (standard_68881_constant_p): Use it.
3805 * config/m68k/m68k.md: Likewise.
3807 2007-01-12 Julian Brown <julian@codesourcery.com>
3808 Richard Sandiford <richard@codesourcery.com>
3810 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Use TUNE_68030
3811 instead of TARGET_68030, TUNE_68040 instead of TARGET_68040,
3812 TUNE_68060 instead of TARGET_68060 and TUNE_CPU32 instead of
3814 (TARGET_CPU32): Rename to...
3815 (TUNE_CPU32): ...this.
3816 (TUNE_68000_10, TUNE_68030, TUNE_68040, TUNE_68060)
3817 (TUNE_CFV2): New macros.
3818 * config/m68k/netbsd-elf.h (LONG_DOUBLE_TYPE_SIZE): Simplify;
3819 remove conditions that are implied by TARGET_68020.
3820 * config/m68k/m68k.c (m68k_output_function_prologue): Use TUNE_68040
3821 instead of TARGET_68040 and TUNE_CPU32 instead of TARGET_CPU32.
3822 (m68k_output_function_epilogue): Likewise.
3823 (m68k_rtx_costs): Likewise. Use TUNE_68060 instead of TARGET_68060
3824 and TUNE_CFV2 instead of TARGET_5200. Use TUNE_68000_10 instead of
3825 "!TARGET_68020 && !TARGET_COLDFIRE" to choose between 68000 and
3826 non-68000 timings. Refactor multiplication and division costs.
3827 (output_addsi3): Use TUNE_68040 instead of TARGET_68040 and
3828 TUNE_CPU32 instead of TARGET_CPU32.
3829 (standard_68881_constant_p): Use TUNE_68040 instead of TARGET_68040
3830 and TUNE_68060 instead of TARGET_68060.
3831 * config/m68k/m68k.md: Use TUNE_68040 instead of TARGET_68040,
3832 TUNE_68060 instead of TARGET_68060, and TUNE_CPU32 instead of
3834 (movsi_const0): Use TUNE_68000_10 rather than "!TARGET_68020
3835 && !TARGET_COLDFIRE" to choose between moveq and clr.
3836 Likewise in the unnamed movsf pattern.
3837 (ashlsi_17_24, lshrsi_17_24): Guard with TUNE_68000_10 rather than
3838 "!TARGET_68020 && !TARGET_COLDFIRE". Likewise the unnamed
3841 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3843 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Increase amount
3844 of tabbing before backslashes.
3846 2007-01-11 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
3848 * pa-linux.h (ASM_OUTPUT_INTERNAL_LABEL): Undefine.
3849 * pa.h (ASM_OUTPUT_LABEL): Output colon when using GAS.
3850 (ASM_OUTPUT_INTERNAL_LABEL): Define.
3852 2007-01-11 Zdenek Dvorak <dvorakz@suse.cz>
3854 * tree-ssa-loop-ivopts.c (extract_cond_operands): Split from
3855 find_interesting_uses_cond.
3856 (find_interesting_uses_cond): Use extract_cond_operands.
3857 (rewrite_use_compare): Use extract_cond_operands and
3858 force_gimple_operand_bsi. Do not call update_stmt.
3859 (determine_use_iv_cost_condition): Use extract_cond_operands.
3860 Return cheaper of using original bound and changing the exit bound.
3862 2007-01-11 Zdenek Dvorak <dvorakz@suse.cz>
3864 PR tree-optimization/29516
3865 * tree-ssa-address.c (tree_mem_ref_addr, add_to_parts,
3866 most_expensive_mult_to_index, addr_to_parts,
3867 create_mem_ref, maybe_fold_tmr): Make the type of
3868 fields of TARGET_MEM_REF sizetype.
3869 (move_fixed_address_to_symbol, move_pointer_to_base):
3871 * tree.def (TARGET_MEM_REF): Add comment on types of
3874 2007-01-11 Joseph Myers <joseph@codesourcery.com>
3876 * c-common.c (vector_types_convertible_p): Treat opaque types as
3877 always convertible if they have the same size, but not otherwise.
3879 2007-01-11 Steven Bosscher <steven@gcc.gnu.org>
3881 * ifcvt.c (struct noce_if_info): Add comments to the fields.
3882 Remove the b_unconditional field.
3883 (noce_try_sign_mask): Do not look at b_unconditional.
3884 (noce_process_if_block): Do not use merge_if_blocks. Update
3885 the CFG here. Do not set b_unconditional.
3886 (cond_move_process_if_block): Likewise.
3887 (find_cond_trap): Likewise.
3888 (check_cond_move_block): Require simple jump insns at the end
3891 2007-01-11 Jan Hubicka <jh@suse.cz>
3893 PR tree-optimization/1046
3894 * tree-tailcall.c (suitable_for_tail_call_opt_p): Use TREE_ADDRESSABLE
3895 when alias info is not ready.
3896 (pass_tail_recursion): Do not require aliasing.
3897 * tree-ssa-copyrename.c (pass_rename_ssa_cop): Likewise.
3898 * tree-ssa-ccp.c (pass_ccp, pass_fold_builtins): Likewise.
3899 * tree-ssa-copy.c (pass_copy_prop): Likewise.
3900 * tree-ssa-forwprop.c (pass_forwprop): Likewise.
3901 * tree-ssa-dce.c (pass_dce, pass_dce_loop, pass_cd_dce): Likewise.
3902 * passes.c (init_optimization_passes): Execute rename_ssa_copies,
3903 ccp, forwprop, copy_prop, merge_phi, copy_prop, dce and tail recursion
3905 * tree-ssa-operands.c (add_virtual_operand, get_indirect_ref_operand):
3906 When aliasing is not build, mark statement as volatile.
3908 2007-01-11 Tom Tromey <tromey@redhat.com>
3910 PR preprocessor/15185, PR preprocessor/20989:
3911 * doc/cppopts.texi <-MT>: Update description of algorithm for
3912 computing default target.
3913 <-M, -MD>: Reword "basename" text.
3915 2007-01-11 Roger Sayle <roger@eyesopen.com>
3917 * builtins.c (expand_builtin_pow, expand_builtin_powi,
3918 fold_builtin_cabs, fold_builtin_sqrt, fold_builtin_trunc,
3919 fold_builtin_floor, fold_builtin_ceil, fold_builtin_round,
3920 fold_builtin_int_int_roundingfn, fold_builtin_bitop,
3921 fold_builtin_bswap, real_constp, fold_builtin_pow,
3922 fold_builtin_powi, fold_builtin_signbit, fold_builtin_copysign,
3923 do_mpfr_arg1, do_mpfr_arg2, do_mpfr_arg3, do_mpfr_sincos): Replace
3924 uses of the macro TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
3925 * convert.c (convert_to_pointer): Likewise.
3926 * expr.c (highest_pow2_factor, expand_expr_real_1): Likewise.
3927 * fold-const.c (force_fit_type, fold_negate_expr, int_const_binop,
3928 const_binop, fold_convert_const_int_from_int,
3929 fold_convert_const_int_from_real,
3930 fold_convert_const_real_from_real, sign_bit_p,
3931 optimize_minmax_comparison, extract_muldiv_1, fold_div_compare,
3932 fold_sign_changed_comparison, fold_unary, fold_comparison,
3933 fold_binary, multiple_of_p, tree_Expr_non_zero_p,
3934 fold_negate_const, fold_abs_const, fold_not_const): Likewise.
3935 * print-tree.c (print_node_brief, print_node): Likewise.
3936 * stor-layout.c (place_field, layout_type): Likewise.
3937 * tree-chrec.c (keep_cast): Likewise.
3938 * tree.c (build_vector, build_real, build_real_from_int_cst,
3939 build_complex): Likewise.
3941 2007-01-11 Roger Sayle <roger@eyesopen.com>
3943 * tree.h (TREE_CONSTANT_OVERFLOW): Obsolete. For the time being,
3944 treat TREE_CONSTANT_OVERFLOW as a synonym of TREE_OVERFLOW.
3946 2007-01-11 Paolo Bonzini <bonzini@gnu.org>
3948 * configure.ac (strict1_warn): Rename to strict_warn.
3949 (WERROR, --enable-werror, symlink hacks, stage1_cflags,
3950 cc_set_by_configure, quoted_cc_set_by_configure,
3951 stage_prefix_set_by_configure, quoted_stage_prefix_set_by_configure,
3952 all_boot_languages, all_stagestuff): Remove.
3953 (target_list): Remove bootstrap targets.
3954 * Makefile.in (quickstrap): Unconditionally make a synonym of all.
3955 (BOOT_LANGUAGES, STAGE1_CFLAGS, STAGE1_CHECKING,
3956 REMAKEFLAGS, FLAGS_TO_PASS, PREPEND_DOTDOT_TO_RELATIVE_PATHS,
3957 SUBDIR_FLAGS_TO_PASS, WERROR_FLAGS, STRICT2_WARN, LANG_STAGESTUFF,
3958 VOL_FILES, POSTSTAGE1_FLAGS_TO_PASS, STAGE2_FLAGS_TO_PASS,
3959 STAGEPROFILE_FLAGS_TO_PASS, STAGEFEEDBACK_FLAGS_TO_PASS, stage1_build,
3960 stage1_copy, stage2_build, stage2_copy, stageprofile_build,
3961 stageprofile_copy, stage3_build, stage3_copy, stagefeedback_build,
3962 stagefeedback_copy, stage4_build, clean_s1, clean_sw, bootstrap,
3963 bootstrap-lean, bootstrap2, bootstrap2-lean, bootstrap3,
3964 bootstrap3-lean, bootstrap4, bootstrap4-lean, unstage1, unstage2,
3965 unstage3, unstage4, unstageprofile, unstagefeedback, restage, restage2,
3966 restage3, restage4, restageprofile, restagefeedback, bubbleestrap,
3967 cleanstrap, unstrap, restrap, *compare, *compare3, *compare4,
3968 *compare-lean, *compare3-lean, *compare4-lean, stage1-start, stage1,
3969 stage2-start, stage2, stage3-start, stage3, stage4-start, stage4,
3970 stageprofile-start, stageprofile, stagefeedback-start, stagefeedback,
3971 risky-stage1, risky-stage2, risky-stage3, risky-stage4): Remove.
3972 (ORDINARY_FLAGS_TO_PASS): Rename to FLAGS_TO_PASS.
3973 (STAGECOPYSTUFF, STAGEMOVESTUFF): Consolidate into MOSTLYCLEANFILES.
3974 (mostlyclean): Adjust.
3975 (clean, distclean): Don't mention bootstrap stuff.
3976 * configure: Regenerate.
3977 * ada/config-lang.in, cp/config-lang.in, forttran/config-lang.in,
3978 java/config-lang.in, objc/config-lang.in, objcp/config-lang.in,
3979 treelang/config-lang.in (stagestuff): Remove.
3980 * doc/sourcebuild.texi (stage1, stage2, stage3, stage4,
3981 stageprofile, stagefeedback, stagestuff): Remove mention.
3983 2007-01-11 Nick Clifton <nickc@redhat.com>
3985 * config/mcore/predicates.md (mcore_general_movesrc_operand):
3987 (mcore_general_movdst_operand): Do not accept CONST_INTs.
3988 (mcore_arith_K_S_operand): Run the test for the S constraint not
3989 the test for the M constraint.
3990 (mcore_addsub_operand): Do not accept integer values that are
3991 larger than 32 bits.
3992 * config/mcore/mcore.md: Remove unused constraints from split.
3993 (andsi3): Use HOST_WIDE_INT instead of int to hold an INTVAL.
3995 (allocate_stack): Likewise.
3996 * config/mcore/mcore.c (mcore_print_operand): Restrict output of P
3997 operands to 32 bits.
3998 (mcore_const_costs): Use HOST_WIDE_INT instead of int to hold an
4000 (mcore_and_cost, mcore_modify_comparison, const_ok_for_mcore,
4001 mcore_const_ok_for_inline, mcore_const_trick_uses_not,
4002 try_constant_tricks, mcore_num_ones, mcore_num_zeros,
4003 mcore_output_bclri, mcore_output_andn, output_inline_const,
4004 mcore_output_move, mcore_output_movedouble): Likewise.
4005 (mcore_output_cmov): Use CONST_OK_FOR_M and CONST_OK_FOR_N.
4006 (output_inline_const): Likewise.
4007 (output_inline_const): Fix format strings used in sprintf
4009 * config/mcore/mcore-protos.h: Update prototypes for changed
4010 functions in mcore.c.
4011 * config/mcore/mcore.h (CONST_OK_FOR_I): Cast values to
4012 HOST_WIDE_INT and not int.
4013 (CONST_OK_FOR_J, CONST_OK_FOR_K, CONST_OK_FOR_L, CONST_OK_FOR_M,
4014 CONST_OK_FOR_N): Likewise.
4015 (LEGITIMATE_CONSTANT_P): Also check CONSTANT_P.
4016 (GO_IF_LEGITIMATE_INDEX): Use HOST_WIDE_INT instead of int to hold
4019 2007-01-10 Jan Hubicka <jh@suse.cz>
4021 * tree-vrp.c (remove_range_assertions): Release defs.
4022 * tree-ssa-loop-ivopts.c (rmeove_statement): Likewise.
4023 * tree-ssa-dom.c (remove_stmt_or_phi): Likewise.
4025 2007-01-10 Paul Brook <paul@codesourcery.com>
4027 * config/arm/arm.c (arm_rtx_costs_1): Handle mutiply-subtract.
4028 * config/arm/arm.md (mulsi3subsi): New insn.
4030 2007-01-10 Zdenek Dvorak <dvorakz@suse.cz>
4032 * tree-ssa-loop-manip.c (tree_unroll_loop): Make it a wrapper over ...
4033 (tree_transform_and_unroll_loop): New.
4034 * tree-flow.h (transform_callback, tree_transform_and_unroll_loop):
4037 2007-01-10 Robert Kennedy <jimbob@google.com>
4039 * fold-const.c (fold_comparison): Fold comparisons like (x *
4040 1000 < 0) to (x < 0).
4042 2007-01-10 Ian Lance Taylor <iant@google.com>
4044 * tree-pretty-print.c (dump_generic_node): Print parentheses when
4045 operands have the same priority.
4047 2007-01-10 Tom Tromey <tromey@redhat.com>
4049 * fold-const.c (fold_truthop): Don't check can_use_bit_fields_p.
4050 (fold_binary): Likewise.
4051 * langhooks.c (lhd_can_use_bit_fields_p): Removed.
4052 * langhooks-def.h (lhd_can_use_bit_fields_p): Removed.
4053 (LANG_HOOKS_CAN_USE_BIT_FIELDS_P): Removed.
4054 (LANG_HOOKS_INITIALIZER): Remove LANG_HOOKS_CAN_USE_BIT_FIELDS_P.
4055 * langhooks.h (struct lang_hooks): Removed field
4056 'can_use_bit_fields_p'.
4058 2007-01-10 Ralf Corsépius <ralf.corsepius@rtems.org>
4060 * config/bfin/t-bfin, config/bfin/t-bfin-elf: Remove GCC_CFLAGS.
4062 2007-01-10 Razya Ladelsky <razya@il.ibm.com>
4064 * function.c (get_last_funcdef_no): New function.
4065 * function.h (get_last_funcdef_no): Declare.
4066 * tree-inline.c (initialize_cfun): Add initialization.
4067 (tree_function_versioning): Cleanup.
4069 2007-01-10 Jan Hubicka <jh@suse.cz>
4071 * tree-inline.c (setup_one_parameter): Do not propagate into abnormal
4074 2007-01-10 Sa Liu <saliu@de.ibm.com>
4075 Ben Elliston <bje@au.ibm.com>
4077 * spu.h (STACK_SAVE_AREA): Use VOIDmode for SAVE_FUNCTION, SImode
4078 for SAVE_NONLOCAL and Pmode for any other save level.
4079 * spu-protos.h (spu_restore_stack_block): Declare.
4080 * spu.md (save_stack_block): Remove.
4081 (restore_stack_block): Call spu_restore_stack_block.
4082 * spu.c (spu_restore_stack_block): New function.
4083 (spu_expand_epilogue): Remove old comment.
4085 2007-01-09 Zdenek Dvorak <dvorakz@suse.cz>
4087 PR tree-optimization/30322
4088 * tree-ssa-loop-ivopts.c (fold_affine_expr, iv_value): Removed.
4089 (cand_value_at): Return the value as aff_tree.
4090 (may_eliminate_iv): Convert the bound from aff_tree to tree.
4091 * tree-affine.c (aff_combination_add_cst, aff_combination_add_product,
4092 aff_combination_mult): New functions.
4093 (aff_combination_add): Use aff_combination_add_cst.
4094 (aff_combination_convert): Allow conversions to a wider type.
4095 (tree_to_aff_combination): Handle BIT_NOT_EXPR.
4096 * tree-affine.h (aff_combination_mult): Declare.
4098 2007-01-09 Carlos O'Donell <carlos@codesourcery.com>
4100 * doc/tm.texi: Update documentation to reflect reality of exec
4101 and start file search behaviours. Update copyright year.
4102 * doc/invoke.texi: Explain how GCC_EXEC_PREFIX is used to find
4103 header file directories.
4105 2007-01-09 Uros Bizjak <ubizjak@gmail.com>
4107 * config/i386/i386.md (*sinxf2): Rename to *sinxf2_i387.
4108 (*cosxf2): Rename to cosxf2_i387.
4109 (*sindf2, *sinsf2): Extend operand 1 to XFmode. Macroize patterns
4110 using X87MODEF12 mode macro. Rename patterns to
4111 *sin_extend<mode>xf2_i387. Use SSE_FLOAT_MODE_P to disable patterns
4113 (*cosdf2, *cossf2): Ditto.
4114 (sincosdf3, sincossf3): Ditto. Rewrite corresponding splitters
4115 to match extended input operands.
4116 (sincos<mode>3): New expander.
4117 (*sinextendsfdf2, *cosextendsfdf2, *sincosextendsfdf3): Remove
4118 insn patterns and corresponding splitters.
4120 2007-01-09 Kaz Kojima <kkojima@gcc.gnu.org>
4122 * config/sh/t-linux (TARGET_LIBGCC2_CFLAGS): Delete.
4123 (SHLIB_MAPFILES, SHLIB_LINK, SHLIB_INSTALL): Likewise.
4125 2007-01-09 Nicolas Pitre <nico@cam.org>
4128 * arm/ieee754-df.S (Lad_s): Also test the low word of X for zero.
4130 2007-01-08 Geoffrey Keating <geoffk@apple.com>
4132 * target.h (struct gcc_target): New field library_rtti_comdat.
4133 * target-def.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): New.
4134 (TARGET_CXX): Add TARGET_CXX_LIBRARY_RTTI_COMDAT.
4135 * doc/tm.texi (C++ ABI): Document TARGET_CXX_LIBRARY_RTTI_COMDAT.
4136 * config/darwin.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
4138 2007-01-08 Geoffrey Keating <geoffk@apple.com>
4140 * doc/invoke.texi (Optimize Options): Correct description of -O0.
4142 2007-01-08 Richard Guenther <rguenther@suse.de>
4144 * tree.h (force_fit_type_double): Export.
4145 (force_fit_type): Remove.
4146 * fold-const.c (force_fit_type_double): New function.
4147 (force_fit_type): Remove.
4148 (int_const_binop): Use it.
4149 (fold_convert_const_int_from_int): Likewise.
4150 (fold_convert_const_int_from_real): Likewise.
4151 (fold_div_compare): Likewise.
4152 (fold_sign_changed_comparison): Likewise.
4153 (fold_unary): Likewise.
4154 (fold_negate_const): Likewise.
4155 (fold_abs_const): Likewise.
4156 (fold_not_const): Likewise.
4157 * c-common.c (shorten_compare): Use force_fit_type_double.
4158 * convert.c (convert_to_pointer): Likewise.
4160 2007-01-08 Richard Guenther <rguenther@suse.de>
4162 * tree.h (build_int_cst_wide_type): Export.
4163 * tree.c (build_int_cst_wide_type): New function.
4164 (build_int_cst_wide): Fix comment.
4165 * builtins.c (fold_builtin_object_size): Use build_int_cst
4166 to build -1 or 0 of the correct type. Use fit_double_type
4167 to check for overflow.
4168 * fold-const.c (optimize_bit_field_compare): Use build_int_cst_type
4170 (decode_field_reference): Likewise.
4171 (all_ones_mask_p): Likewise.
4172 (native_interpret_int): Use build_int_cst_wide_type.
4173 (fold_binary): Use build_int_cst_type to build an all-ones
4175 * stor-layout.c (set_sizetype): Use build_int_cst_wide_type.
4177 2007-01-08 Daniel Jacobowitz <dan@codesourcery.com>
4179 * config/pa/t-pa64 (libgcc_stub.a): Use $(T).
4181 2007-01-09 Ben Elliston <bje@au.ibm.com>
4183 * genautomata.c (STATS_OPTION): New option.
4184 (stats_flag): New flag.
4185 (gen_automata_option): Handle it.
4186 (initiate_automaton_gen): Ditto.
4187 (write_automata): Output statistics only if stats_flag is
4188 set. Likewise, output time statistics only if time_flag is set.
4189 * doc/md.texi (Processor pipeline description): Document new flag.
4191 2007-01-08 Richard Guenther <rguenther@suse.de>
4193 * builtins.c (fold_builtin_int_roundingfn): Use fit_double_type.
4194 * tree.c (build_int_cst_type): Likewise.
4195 (size_in_bytes): Don't call force_fit_type on the result.
4196 (int_fits_type_p): Use fit_double_type.
4197 * fold-const.c (fit_double_type): New function.
4198 (force_fit_type): Use it.
4199 * tree.h (fit_double_type): Export.
4201 2007-01-08 Jan Hubicka <jh@suse.cz>
4203 * tree-vectorizer.c (gate_increase_alignment): Fix return type.
4204 * ipa.c (function_and_variable_visibility): Fix return type.
4206 2007-01-08 Richard Guenther <rguenther@suse.de>
4208 * tree-ssa-ccp.c (maybe_fold_offset_to_array_ref): Use type
4209 of offset to build the index.
4210 * tree-pretty-print.c (dump_generic_node): Don't build negated
4211 const just for printing.
4212 * c-pretty-print.c (pp_c_integer_constant): Likewise.
4213 * builtins.c (fold_builtin_int_roundingfn): Check if result
4214 fits the type by using force_fit_type and comparing the result.
4215 * predict.c (predict_loops): Use compare_tree_int for comparison.
4216 * tree.c (build_int_cst): Fall back to integer_type_node for
4218 (build_int_cst_wide): Assert type is non-null.
4220 2007-01-08 Roberto Costa <roberto.costa@st.com>
4222 * tree-vrp.c (extract_range_from_cond_expr): New.
4223 (extract_range_from_expr): Handle COND_EXPR nodes used as expressions.
4224 * tree-ssa-ccp.c (get_maxval_strlen): Handle COND_EXPR nodes used
4226 (fold_stmt): Bug fix, avoid infinite recursion when folding COND_EXPRs.
4227 * tree-ssa-forwprop.c (simplify_cond, forward_propagate_into_cond,
4228 tree_ssa_forward_propagate_single_use_vars): Handle COND_EXPR nodes
4229 used as expressions.
4230 * tree-object-size.c (cond_expr_object_size): New.
4231 (collect_object_sizes_for): Handle COND_EXPR nodes used as expressions.
4233 2007-01-08 Jan Hubicka <jh@suse.cz>
4235 * tree-ssa-forwprop.c (forward_propagate_into_cond,
4236 tree_ssa_forward_propagate_single_use_va): Release defs of propagated
4239 2007-01-08 Richard Guenther <rguenther@suse.de>
4241 PR tree-optimization/23603
4242 * tree-vrp.c (set_value_range_to_truthvalue): New function.
4243 (extract_range_from_binary): Fall back to truthvalue instead of
4244 varying for TRUTH_*_EXPR.
4245 (extract_range_from_comparison): Fall back to truthvalue instead of
4247 (vrp_visit_phi_node): Don't adjust new range bounds to +INF/-INF
4248 if all visited PHI values were constant.
4250 2007-01-08 Jan Hubicka <jh@suse.cz>
4252 * cgraphunit.c (cgraph_process_new_functions): Reset reachable flag.
4253 (cgraph_analyze_function): break out from ...
4254 (cgraph_finalize_compilation_unit): ... here.
4255 (cgraph_expand_function): Remove forgoten commented out line.
4256 (cgraph_optimize): Analyze functions.
4258 2007-01-08 Jan Hubicka <jh@suse.cz>
4260 * tree-pas.h (TODO_remove_function): New flag.
4261 (TODO_update*): Renumber.
4262 (pass_ipa_increase_alignment,
4263 pass_ipa_function_and_variable_visibility): New passes.
4264 * cgraphunit.c (cgraph_increase_alignment): Move to tree-vectorizer.c
4265 (cgraph_function_and_variable_visibility): Move to ipa.c
4266 (cgraph_optimize): Don't call cgraph_function_and_variable_visibility,
4267 cgraph_increase_alignment.
4268 * ipa-inline.c (cgraph_decide_inlining): Don't push timevar.
4269 (cgraph_decide_inlining_incrementally): Push TV_INTEGRATION before
4270 calling tree-inline.
4271 (cgraph_early_inlining): Do not call cgraph_remove_unreachable_nodes.
4272 (pass_ipa_inline, pass_early_ipa_inlining): Set TODO_remove_functions
4273 * tree-vectorizer.c (increase_alignment): Move here from cgraphunit.c
4274 (gate_increase_alignment): New function.
4275 (pass_ipa_increase_alignment): New pass.
4276 * ipa.c: Inline tree-pass.h and timevar.h
4277 (function_and_variable_visibility): Move here from cgraphunit.c
4278 * tree-optimize.c (pass_early_local_passes): Add TODO_remove_functions.
4279 * passes.c (init_optimization_passes): Add the two new passes.
4280 (execute_todo): Handle cgraph_remove_functions.
4282 2007-01-08 Nick Clifton <nickc@redhat.com>
4284 * config/frv/predicates.md (reg_or_0_operand): Accept
4287 2007-01-08 Ralf Corsépius <ralf.corsepius@rtems.org>
4289 * config/bfin/rtems.h, config/bfin/t-rtems: New.
4290 * config.gcc: Add bfin*-rtems*.
4292 2007-01-08 Mark Shinwell <shinwell@codesourcery.com>
4294 * c.opt: Add -flax-vector-conversions.
4295 * c-typeck.c (convert_for_assignment): Pass flag to
4296 vector_types_convertible_p to allow emission of note.
4297 (digest_init): Likewise.
4298 * c-opts.c: Handle -flax-vector-conversions.
4299 * c-common.c (flag_lax_vector_conversions): New.
4300 (vector_types_convertible_p): Unless -flax-vector conversions
4301 has been passed, disallow conversions between vectors with
4302 differing numbers of subparts and/or element types. If such
4303 a conversion is disallowed, possibly emit a note on the first
4304 occasion only to inform the user of -flax-vector-conversions.
4305 The new last argument specifies this.
4306 * c-common.h (flag_lax_vector_conversions): New.
4307 (vector_types_convertible_p): Add extra argument.
4308 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Use
4309 char_type_node for V*QI type vectors.
4310 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins):
4311 Update to satisfy new typechecking rules.
4312 * config/rs6000/altivec.h (vec_cmple): Use vec_cmpge, for both
4314 * doc/invoke.texi (C Dialect Options): Document
4315 -flax-vector-conversions.
4317 2007-01-08 Mark Shinwell <shinwell@codesourcery.com>
4319 PR tree-optimization/29877
4320 * tree-ssa-ter.c (is_replaceable_p): Deem assignments with
4321 a register variable on the RHS to not be replaceable.
4323 2007-01-08 Chen Liqin <liqin@sunnorth.com.cn>
4324 * config/score/t-score-elf (MULTILIB_OPTIONS): Change.
4325 * config/score/predicates.md (const_uimm5, sr0_operand, const_simm12,
4326 const_simm15, const_pow2, const_npow2): Added.
4327 * config/score/misc.md (insv, extv, extzv, movmemsi,
4328 move_lbu_a/b, mov_lhu_a/b etc): Added and fix some bug.
4329 * config/score/score.c (score_address_cost, score_select_cc_mode):
4331 Change CONST_OK_FOR_LETTER_P/EXTRA_CONSTRAINT define.
4332 Update score_rtx_costs for MACRO TARGET_RTX_COSTS.
4333 Update score_print_operand.
4334 * config/score/score.h (DATA_ALIGNMENT, SELECT_CC_MODE): Added.
4335 Adjust register allocate order and update some macro define.
4336 * config/score/score-mdaux.c (mdx_unaligned_load, mdx_unsigned_store,
4337 mdx_block_move_straight, mdx_block_move_loop_head,
4338 mdx_block_move_loop_body, mdx_block_move_loop_foot, mdx_block_move_loop,
4339 mdx_block_move): Added.
4340 (mdx_movsicc, mdp_select_add_imm, mdp_select, mds_zero_extract_andi,
4341 mdp_limm): Updated and fix some bug and typo.
4342 * config/score/score.md (movqi/hi/si, add/sub/zero/ext): Updated.
4343 (movsf, movdf, doloop_end): Added.
4345 2007-01-08 Kazu Hirata <kazu@codesourcery.com>
4347 * config/arm/arm.c, config/arm/arm.h, config/arm/arm.md,
4348 config/arm/thumb2.md: Fix comment typos.
4349 * doc/extend.texi: Fix a typo.
4351 2007-01-07 Eric Christopher <echristo@apple.com>
4353 * configure.ac: Check for __stack_chk_fail for darwin.
4354 * configure: Regenerate.
4356 2007-01-07 Richard Guenther <rguenther@suse.de>
4358 * tree-vrp.c (extract_range_from_assert): CSE calls to
4359 compare_values where possible.
4360 (extract_range_from_unary_expr): Likewise.
4362 2007-01-07 Anatoly Sokolov <aesok@post.ru>
4364 * config/avr/avr-protos.h (call_insn_operand): Delete prototype.
4365 * config/avr/avr.c (call_insn_operand): Delete function.
4366 * config/avr/avr.md (*pushqi, *pushhi, *pushsi, *pushsf): Use REG_SP
4367 instead of register number. Use predicates.
4368 * config/avr/predicates.md (const0_operand, reg_or_0_operand,
4369 call_insn_operand): Add.
4371 2007-01-06 Jan Hubicka <jh@suse.cz>
4373 * tree-pass.h (pass_build_cgraph_edges): Declare.
4374 * cgraphunit.c (record_refernece): Move to cgraphbuild.c
4375 (visited_nodes): Remove.
4376 (cgraph_create_edges): Move to cgraphbuild.c; rename to
4377 build_cgrpah_edges; make visited_nodes local.
4378 (cgraph_process_new_functions): DO not call initialize_inline_failed.
4379 (record_references_in_initializer): Move to cgraphbuild.c
4380 (initialize_inline_failed, rebuild_cgraph_edges,
4381 pass_rebuild_cgraph_edges): Move to cgraphbuild.c.
4382 (verify_cgraph_node): Make visited_nodes local.
4383 (cgraph_analyze_function): Do not call cgraph_create_edges and
4384 initialize_inline_failed.
4385 (cgraph_expand_function): Do not call cgraph_lower_function;
4386 assert that function is already lowered.
4387 * Makefile.in (cgraphbuild.o): New.
4388 * passes.c (init_optimization_passes): Add pass_build_cgraph_edges
4389 at the end of lowering passes.
4391 2007-01-06 Steven Bosscher <steven@gcc.gnu.org>
4393 * ifcvt.c (cond_move_convert_if_block): New function, code
4394 factored out from...
4395 (cond_move_process_if_block): ...here. Call the new function
4396 on the THEN and ELSE blocks.
4397 (merge_if_block): Do not copy global_live_at_end, merge_blocks
4398 already takes care of this.
4400 2007-01-05 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
4403 * tree.h (TREE_OVERFLOW_P): New.
4404 * c-typeck.c (parser_build_unary_op): Warn only if result
4405 overflowed and operands did not.
4406 (parser_build_binary_op): Likewise.
4407 (convert_for_assignment): Remove redundant overflow_warning.
4408 * c-common.c (overflow_warning): Don't check or set TREE_OVERFLOW.
4410 2007-01-05 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
4412 * c-typeck.c (store_init_value): Split over two lines to follow
4413 the GNU coding style.
4415 2007-01-05 Benjamin Kosnik <bkoz@redhat.com>
4417 * c-cppbuiltin.c (c_cpp_builtins): __GXX_EXPERIMENTAL_CPP0X__ to
4418 __GXX_EXPERIMENTAL_CXX0X__.
4419 * doc/cpp.texi: Same.
4421 2007-01-05 Richard Guenther <rguenther@suse.de>
4424 * tree.c (get_narrower): Do not construct COMPONENT_REFs
4425 with mismatched types. Instead explicitly build a
4426 conversion NOP_EXPR.
4428 2007-01-05 Ian Lance Taylor <iant@google.com>
4430 * c-common.c (decl_with_nonnull_addr_p): New function.
4431 (c_common_truthvalue_conversion): Call it.
4432 * c-typeck.c (build_binary_op): Likewise.
4433 * c-common.h (decl_with_nonnull_addr_p): Declare.
4435 2007-01-05 Jakub Jelinek <jakub@redhat.com>
4438 * libgcc2.c (__divdc3): Compare c and d against 0.0 instead of
4441 2007-01-05 Joel Brobecker <brobecker@adacore.com>
4443 * doc/install.texi (Final install): Document the fact that
4444 the GNAT runtime should not be stripped.
4446 2007-01-04 Jan Hubicka <jh@suse.cz>
4448 * tree-inline.c (fold_marked_statements): Update operand caches
4449 and EH after folding
4451 2007-01-04 Ian Lance Taylor <iant@google.com>
4453 * c-common.c (check_function_nonnull): Whitespace fix.
4455 2007-01-04 Jan Hubicka <jh@suse.cz>
4457 * tree-optimize.c (execute_fixup_cfg): Correct previously mistakely
4458 comitted older version of patch.
4459 (pass_fixup_cfg): Add TODOs to verify flow and statements, dump
4460 function, celanup cfg and collect garbage.
4462 2007-01-04 Mike Stump <mrs@apple.com>
4464 * Makefile.in (mostlyclean): Don't remove libgcc anymore.
4467 2007-01-04 Eric Christopher <echristo@apple.com>
4469 * libgcc2.c (__bswapsi2): Use SItype.
4470 (__bswapdi2): Use DItype.
4471 * libgcc2.h: Update for above.
4473 2007-01-04 Paul Brook <paul@codesourcery.com>
4475 * config/arm/arm.md (arm_mulsi3, thumb_mulsi3, mulsi3_compare0,
4476 mulsi_compare0_scratch, mulsi3addsi, mulsi3addsi_compare0,
4477 mulsi3addsi_compare0_scratch, mulsidi3adddi, mulsidi3,
4478 umulsidi3, umulsidi3adddi, smulsi3_highpart,
4479 umulsi3_highpart): Make conditional on !arm_arch6.
4480 (arm_mulsi3_v6, thumb_mulsi3_v6, mulsi3_compare0_v6,
4481 mulsi_compare0_scratch_v6, mulsi3addsi_v6, mulsi3addsi_compare0_v6,
4482 mulsi3addsi_compare0_scratch_v6, mulsidi3adddi_v6, mulsidi3_v6,
4483 umulsidi3_v6, umulsidi3adddi_v6, smulsi3_highpart_v6,
4484 umulsi3_highpart_v6): New insns.
4486 2007-01-04 Roger Sayle <roger@eyesopen.com>
4488 * fold-const.c (fold_convert): When casting an expression to void,
4489 fold_ignored_result may discover a GIMPLE_MODIFY_STMT which doesn't
4490 have a type. Instead of attempting to build a NOP_EXPR, return
4491 these "special" trees directly.
4493 2007-01-04 Joseph Myers <joseph@codesourcery.com>
4495 * config/rs6000/rs6000.c (rs6000_rtx_costs): Make adjustment for
4496 MULT inside MINUS as either argument. Use rs6000_cost->dmul -
4497 rs6000_cost->fp not 0 as adjustment for outer NEG.
4499 2007-01-04 Jan Hubicka <jh@suse.cz>
4501 * cgraph.c (cgraph_release_function_body): New function.
4502 (cgraph_remove_node): Use it.
4503 * cgraph.h (cgraph_release_function_body): Declare.
4504 * cgraphunit.c (cgraph_expand_function): Use it.
4505 * ipa.c (cgraph_remove_unreahchable_nodes): Use it.
4506 * tree-ssa.c (delete_tree_ssa): Allow to be called before aliasing
4507 is initialized and while compilation of other function is running.
4508 * tree-optimize.c (execute_free_cfg_annotations): Move code to clear
4509 statement CFG annotations from here to ...
4510 * tree-cfg.c (delete_tree_cfg_annotations): ... here.
4512 2007-01-04 Zdenek Dvorak <dvorakz@suse.cz>
4514 * cfgloop.h (enum li_flags): Make the constants powers of two.
4516 2007-01-04 Jan Hubicka <jh@suse.cz>
4518 * tree-inline.c (copy_bb): Insert new statements to statements_to_fold
4520 (fold_marked_statements): New function.
4521 (optimize_inline_calls, tree_function_versioning): Fold new statements.
4522 * tree-inline.h (copy_body_data): Add statements_to_fold.
4524 2007-01-03 Daniel Jacobowitz <dan@codesourcery.com>
4526 * config.gcc: Mention libgcc/config.host.
4527 * Makefile.in: Update comments mentioning libgcc.
4528 (LIBGCC, INSTALL_LIBGCC, GCC_PARTS, mklibgcc): Delete.
4529 (all.cross, start.encap, rest.encap, rest.cross): Update
4530 dependencies for libgcc move.
4531 (libgcc.mk, LIBGCC_DEPS, libgcov.a, libgcc.a, stmp-multilib)
4532 (clean-target, clean-target-libgcc): Delete.
4533 (srcdirify, GCC_EXTRA_PARTS): New macros.
4534 (libgcc-support, libgcc.mvars): New rules.
4535 (distclean): Remove mention of mklibgcc.
4536 (install): Don't reference INSTALL_LIBGCC.
4537 (install-common): Don't reference EXTRA_PARTS.
4538 (install-libgcc, install-multilib): Delete rules.
4539 * mklibgcc.in: Delete file.
4540 * doc/configfiles.texi: Don't mention mklibgcc.
4542 * config/i386/t-darwin (SHLIB_VERPFX): Delete (moved to libgcc).
4543 * config/i386/t-darwin64 (SHLIB_VERPFX): Likewise.
4544 * config/rs6000/t-darwin (SHLIB_VERPFX): Likewise.
4545 * config/rs6000/t-ppccomm (TARGET_LIBGCC2_CFLAGS, SHLIB_MAPFILES)
4546 (mklibgcc, ldblspecs): Likewise.
4548 * config/i386/t-nwld (libgcc.def, libc.def, libpcre.def)
4549 (posixpre.def): Use $(T).
4550 (SHLIB_EXT, SHLIB_NAME, SHLIB_SLIBDIR_QUAL, SHLIB_DEF, SHLIB_MAP)
4551 (SHLIB_SRC, SHLIB_INSTALL): Delete.
4552 (SHLIB_LINK): Make dummy.
4553 * config/t-slibgcc-darwin: Delete contents except for dummy SHLIB_LINK.
4555 * config/frv/t-linux (EXTRA_MULTILIB_PARTS): Clear.
4557 * config/alpha/t-crtfm: Use $(T) in rules for EXTRA_PARTS.
4558 * config/alpha/t-vms, config/alpha/t-vms64, config/fr30/t-fr30,
4559 config/i386/t-rtems-i386, config/ia64/t-ia64, config/rs6000/t-beos,
4560 config/rs6000/t-newas, config/sparc/t-elf: Likewise.
4562 * configure.ac (all_outputs): Remove mklibgcc.
4563 * configure: Regenerated.
4565 2007-01-03 Josh Conner <jconner@apple.com>
4568 * calls.c (compute_argument_addresses): Set stack and stack_slot
4569 for partial args, too.
4570 (store_one_arg): Use locate.size.constant for the size when
4571 generating a save_area.
4573 2007-01-03 Robert Kennedy <jimbob@google.com>
4575 * tree-cfg.c (tree_merge_blocks): Release SSA_NAME phi results
4576 whose definitions are deleted due to basic block merging.
4578 2007-01-03 Paul Brook <paul@codesourcery.com>
4581 * config/arm/arm.c (output_return_instruction): Pop PC in interrupt
4583 (use_return_insn): Return 0 for Thumb interrupt functions.
4584 (print_multi_reg): Add rfe argument for IRQ returns.
4585 (arm_output_epilogue): Pop interrupt return address directly into PC.
4586 (arm_expand_prologue): Only adjust IRQ return address in Arm mode.
4588 2007-01-03 Paul Brook <paul@codesourcery.com>
4590 Merge from sourcerygxx-4_1.
4591 * config/arm/thumb2.md: New file.
4592 * config/arm/elf.h (JUMP_TABLES_IN_TEXT_SECTION): Return True for
4594 * config/arm/coff.h (JUMP_TABLES_IN_TEXT_SECTION): Ditto.
4595 * config/arm/aout.h (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
4596 (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump tables.
4597 * config/arm/aof.h (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump
4599 (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
4600 * config/arm/ieee754-df.S: Use macros for Thumb-2/Unified asm
4602 * config/arm/ieee754-sf.S: Ditto.
4603 * config/arm/arm.c (thumb_base_register_rtx_p): Rename...
4604 (thumb1_base_register_rtx_p): ... to this.
4605 (thumb_index_register_rtx_p): Rename...
4606 (thumb1_index_register_rtx_p): ... to this.
4607 (thumb_output_function_prologue): Rename...
4608 (thumb1_output_function_prologue): ... to this.
4609 (thumb_legitimate_address_p): Rename...
4610 (thumb1_legitimate_address_p): ... to this.
4611 (thumb_rtx_costs): Rename...
4612 (thumb1_rtx_costs): ... to this.
4613 (thumb_compute_save_reg_mask): Rename...
4614 (thumb1_compute_save_reg_mask): ... to this.
4615 (thumb_final_prescan_insn): Rename...
4616 (thumb1_final_prescan_insn): ... to this.
4617 (thumb_expand_epilogue): Rename...
4618 (thumb1_expand_epilogue): ... to this.
4619 (arm_unwind_emit_stm): Rename...
4620 (arm_unwind_emit_sequence): ... to this.
4621 (thumb2_legitimate_index_p, thumb2_legitimate_address_p,
4622 thumb1_compute_save_reg_mask, arm_dwarf_handle_frame_unspec,
4623 thumb2_index_mul_operand, output_move_vfp, arm_shift_nmem,
4624 arm_save_coproc_regs, thumb_set_frame_pointer, arm_print_condition,
4625 thumb2_final_prescan_insn, thumb2_asm_output_opcode, arm_output_shift,
4626 thumb2_output_casesi): New functions.
4627 (TARGET_DWARF_HANDLE_FRAME_UNSPEC): Define.
4628 (FL_THUMB2, FL_NOTM, FL_DIV, FL_FOR_ARCH6T2, FL_FOR_ARCH7,
4629 FL_FOR_ARCH7A, FL_FOR_ARCH7R, FL_FOR_ARCH7M, ARM_LSL_NAME,
4630 THUMB2_WORK_REGS): Define.
4631 (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv, arm_condexec_count,
4632 arm_condexec_mask, arm_condexec_masklen)): New variables.
4633 (all_architectures): Add armv6t2, armv7, armv7a, armv7r and armv7m.
4634 (arm_override_options): Check new CPU capabilities.
4635 Set new architecture flag variables.
4636 (arm_isr_value): Handle v7m interrupt functions.
4637 (user_return_insn): Return 0 for v7m interrupt functions. Handle
4639 (const_ok_for_arm): Handle Thumb-2 constants.
4640 (arm_gen_constant): Ditto. Use movw when available.
4641 (arm_function_ok_for_sibcall): Return false for v7m interrupt
4643 (legitimize_pic_address, arm_call_tls_get_addr): Handle Thumb-2.
4644 (thumb_find_work_register, arm_load_pic_register,
4645 legitimize_tls_address, arm_address_cost, load_multiple_sequence,
4646 emit_ldm_seq, emit_stm_seq, arm_select_cc_mode, get_jump_table_size,
4647 print_multi_reg, output_mov_long_double_fpa_from_arm,
4648 output_mov_long_double_arm_from_fpa, output_mov_double_fpa_from_arm,
4649 output_mov_double_fpa_from_arm, output_move_double,
4650 arm_compute_save_reg_mask, arm_compute_save_reg0_reg12_mask,
4651 output_return_instruction, arm_output_function_prologue,
4652 arm_output_epilogue, arm_get_frame_offsets, arm_regno_class,
4653 arm_output_mi_thunk, thumb_set_return_address): Ditto.
4654 (arm_expand_prologue): Handle Thumb-2. Use arm_save_coproc_regs.
4655 (arm_coproc_mem_operand): Allow POST_INC/PRE_DEC.
4656 (arithmetic_instr, shift_op): Use arm_shift_nmem.
4657 (arm_print_operand): Use arm_print_condition. Handle '(', ')', '.',
4659 (arm_final_prescan_insn): Use extract_constrain_insn_cached.
4660 (thumb_expand_prologue): Use thumb_set_frame_pointer.
4661 (arm_file_start): Output directive for unified syntax.
4662 (arm_unwind_emit_set): Handle stack alignment instruction.
4663 * config/arm/lib1funcs.asm: Remove default for __ARM_ARCH__.
4664 Add v6t2, v7, v7a, v7r and v7m.
4665 (RETLDM): Add Thumb-2 code.
4666 (do_it, shift1, do_push, do_pop, COND, THUMB_SYNTAX): New macros.
4667 * config/arm/arm.h (TARGET_CPU_CPP_BUILTINS): Define __thumb2__.
4668 (TARGET_THUMB1, TARGET_32BIT, TARGET_THUMB2, TARGET_DSP_MULTIPLY,
4669 TARGET_INT_SIMD, TARGET_UNIFIED_ASM, ARM_FT_STACKALIGN, IS_STACKALIGN,
4670 THUMB2_TRAMPOLINE_TEMPLATE, TRAMPOLINE_ADJUST_ADDRESS,
4671 ASM_OUTPUT_OPCODE, THUMB2_GO_IF_LEGITIMATE_ADDRESS,
4672 THUMB2_LEGITIMIZE_ADDRESS, CASE_VECTOR_PC_RELATIVE,
4673 CASE_VECTOR_SHORTEN_MODE, ADDR_VEC_ALIGN, ASM_OUTPUT_CASE_END,
4674 ADJUST_INSN_LENGTH): Define.
4675 (TARGET_REALLY_IWMMXT, TARGET_IWMMXT_ABI, CONDITIONAL_REGISTER_USAGE,
4676 STATIC_CHAIN_REGNUM, HARD_REGNO_NREGS, INDEX_REG_CLASS,
4677 BASE_REG_CLASS, MODE_BASE_REG_CLASS, SMALL_REGISTER_CLASSES,
4678 PREFERRED_RELOAD_CLASS, SECONDARY_OUTPUT_RELOAD_CLASS,
4679 SECONDARY_INPUT_RELOAD_CLASS, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P,
4680 TRAMPOLINE_SIZE, INITIALIZE_TRAMPOLINE, HAVE_PRE_INCREMENT,
4681 HAVE_POST_DECREMENT, HAVE_PRE_DECREMENT, HAVE_PRE_MODIFY_DISP,
4682 HAVE_POST_MODIFY_DISP, HAVE_PRE_MODIFY_REG, HAVE_POST_MODIFY_REG,
4683 REGNO_MODE_OK_FOR_BASE_P, LEGITIMATE_CONSTANT_P,
4684 REG_MODE_OK_FOR_BASE_P, REG_OK_FOR_INDEX_P, GO_IF_LEGITIMATE_ADDRESS,
4685 LEGITIMIZE_ADDRESS, THUMB2_LEGITIMIZE_ADDRESS,
4686 GO_IF_MODE_DEPENDENT_ADDRESS, MEMORY_MOVE_COST, BRANCH_COST,
4687 ASM_APP_OFF, ASM_OUTPUT_CASE_LABEL, ARM_DECLARE_FUNCTION_NAME,
4688 FINAL_PRESCAN_INSN, PRINT_OPERAND_PUNCT_VALID_P,
4689 PRINT_OPERAND_ADDRESS): Adjust for Thumb-2.
4690 (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv): New declarations.
4691 * config/arm/arm-cores.def: Add arm1156t2-s, cortex-a8, cortex-r4 and
4693 * config/arm/arm-tune.md: Regenerate.
4694 * config/arm/arm-protos.h: Update prototypes.
4695 * config/arm/vfp.md: Enable patterns for Thumb-2.
4696 (arm_movsi_vfp): Add movw alternative. Use output_move_vfp.
4697 (arm_movdi_vfp, movsf_vfp, movdf_vfp): Use output_move_vfp.
4698 (thumb2_movsi_vfp, thumb2_movdi_vfp, thumb2_movsf_vfp,
4699 thumb2_movdf_vfp, thumb2_movsfcc_vfp, thumb2_movdfcc_vfp): New.
4700 * config/arm/libunwind.S: Add Thumb-2 code.
4701 * config/arm/constraints.md: Update include Thumb-2.
4702 * config/arm/ieee754-sf.S: Add Thumb-2/Unified asm support.
4703 * config/arm/ieee754-df.S: Ditto.
4704 * config/arm/bpabi.S: Ditto.
4705 * config/arm/t-arm (MD_INCLUDES): Add thumb2.md.
4706 * config/arm/predicates.md (low_register_operand,
4707 low_reg_or_int_operand, thumb_16bit_operator): New.
4708 (thumb_cmp_operand, thumb_cmpneg_operand): Rename...
4709 (thumb1_cmp_operand, thumb1_cmpneg_operand): ... to this.
4710 * config/arm/t-arm-elf: Add armv7 multilib.
4711 * config/arm/arm.md: Update patterns for Thumb-2 and Unified asm.
4713 (UNSPEC_STACK_ALIGN, ce_count): New.
4714 (arm_incscc, arm_decscc, arm_umaxsi3, arm_uminsi3,
4715 arm_zero_extendsidi2, arm_zero_extendqidi2): New
4717 * config/arm/fpa.md: Update patterns for Thumb-2 and Unified asm.
4718 (thumb2_movsf_fpa, thumb2_movdf_fpa, thumb2_movxf_fpa,
4719 thumb2_movsfcc_fpa, thumb2_movdfcc_fpa): New insns.
4720 * config/arm/cirrus.md: Update patterns for Thumb-2 and Unified asm.
4721 (cirrus_thumb2_movdi, cirrus_thumb2_movsi_insn,
4722 thumb2_cirrus_movsf_hard_insn, thumb2_cirrus_movdf_hard_insn): New
4724 * doc/extend.texi: Document ARMv7-M interrupt functions.
4725 * doc/invoke.texi: Document Thumb-2 new cores+architectures.
4727 2007-01-03 Jakub Jelinek <jakub@redhat.com>
4729 * unwind-dw2.c (SIGNAL_FRAME_BIT, EXTENDED_CONTEXT_BIT): Define.
4730 (struct _Unwind_Context): Rename args_size to flags, remove
4731 signal_frame field, add a new args_size field and version field.
4732 (_Unwind_IsSignalFrame, _Unwind_SetSignalFrame,
4733 _Unwind_IsExtendedContext): New inline functions.
4734 (_Unwind_GetGR, _Unwind_SetGR, _Unwind_GetGRPtr, _Unwind_SetGRPtr):
4735 Assume by_value array is only present if _Unwind_IsExtendedContext.
4736 (_Unwind_GetIPInfo, execute_cfa_program, uw_frame_state_for): Use
4737 _Unwind_IsSignalFrame.
4738 (__frame_state_for): Initialize context.flags to EXTENDED_CONTEXT_BIT.
4739 (uw_update_context_1): Use _Unwind_SetSignalFrame.
4740 (uw_init_context_1): Initialize context->flags to
4741 EXTENDED_CONTEXT_BIT.
4742 * config/rs6000/linux-unwind.h (frob_update_context): Use
4743 _Unwind_SetSignalFrame.
4745 2007-01-03 Andrew Pinski <andrew_pinski@playstation.sony.com>
4748 * gimplify.c (gimplify_modify_expr_complex_part): Move below
4749 tree_to_gimple_tuple. Call tree_to_gimple_tuple when we need
4752 2007-01-03 Kazu Hirata <kazu@codesourcery.com>
4754 * config/i386/i386.h (NON_STACK_REG_P, REGNO_OK_FOR_SIREG_P,
4755 REGNO_OK_FOR_DIREG_P, REWRITE_ADDRESS, ASM_OPERAND_LETTER,
4756 RET, AT_SP): Remove.
4757 * config/i386/i386.md (*sse_prologue_save_insn): Use return
4760 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
4761 * cfgbuild.c (find_basic_blocks): Likewise.
4762 * cfgrtl.c (rtl_create_basic_block): Likewise.
4763 * function.c (temp_slots_at_level): Likewise.
4764 * reg-stack.c (stack_regs_mentioned): Likewise.
4765 * regclass.c (allocate_reg_info): Likewise.
4766 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
4767 set_bb_for_stmt, move_block_to_fn): Likewise.
4768 * tree-complex.c (tree_lower_complex): Likewise.
4769 * vec.h (VEC_safe_grow_cleared): New.
4771 * cgraphunit.c, tree-ssa-alias.c: Fix comment typos.
4773 2007-01-03 Zdenek Dvorak <dvorakz@suse.cz>
4775 * loop-unswitch.c (unswitch_loop): Pass probabilities to loopify.
4776 * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Pass probabilities
4778 * cfgloopmanip.c (scale_loop_frequencies): Export.
4779 (loopify): Scale the frequencies by prescribed coefficients.
4780 (set_zero_probability): New function.
4781 (duplicate_loop_to_header_edge): Improve updating of frequencies.
4782 (lv_adjust_loop_entry_edge, loop_version): Set probabilities
4783 and frequencies according to arguments.
4784 * tree-ssa-loop-manip.c (tree_unroll_loop): Set probabilities
4786 * cfg.c (scale_bbs_frequencies_int): Allow scaling the frequencies up.
4787 * modulo-sched.c (sms_schedule): Set probabilities for entering
4788 versioned loop correctly.
4789 * tree-vect-transform.c (vect_transform_loop): Ditto.
4790 * cfgloop.h (loopify, loop_version): Declaration changed.
4791 (scale_loop_frequencies): Declared.
4793 2007-01-02 Jan Hubicka <jh@suse.cz>
4795 * cgraph.c: Include tree-flow.h
4796 (cgraph_add_new-function): Handle IPA_SSA mode; execute
4798 * cgraph.h (enum cgraph_state): Add CGRAPH_STATE_IPA_SSA.
4799 * tree-pass.h (pass_all_early_optimizations): Declare.
4800 * cgraphunit.c (cgraph_process_new_functions): Add IPA_SSA; execute
4802 (cgraph_analyze_function): Do early_local_passes.
4803 * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
4804 Do not add referenced vars.
4805 * tree-optimize.c (gate_all_optimizations): Do not execute when not in
4807 (gate_all_early_local_passes): New gate.
4808 (pass_early_local_passes): Use new gate.
4809 (execute_early_local_optimizations): New functions.
4810 (gate_all_early_optimizations): New gate.
4811 (pass_all_early_optimizations): New pass.
4812 (execute_free_datastructures): Free SSA only when initialized.
4813 (gate_init_datastructures): Init only when optimizing.
4814 (tree_lowering_passes): Do early local passes when called late.
4815 * tree-profile.c (do_tree_profiling): Don't profile functions added
4817 (do_early_tree_profiling, pass_early_tree_profile): Kill.
4818 * tree-cfg.c (update_modified_stmts): Do not update when operands are
4820 * passes.c (init_optimizations_passes): Reorder so we go into SSA
4821 during early_local_passes.
4822 * Makefile.in (cgraph.o): Add dependency on tree-flow.h.
4825 2007-01-02 Carlos O'Donell <carlos@codesourcery.com>
4827 * Makefile.in: Update copyright year.
4829 2007-01-02 Carlos O'Donell <carlos@codesourcery.com>
4831 * Makefile.in: Export GCC_EXEC_PREFIX before calling $(RUNTEST)
4832 in $(lang_checks) and check-consistency targets.
4834 2007-01-02 Jan Hubicka <jh@suse.cz>
4836 * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
4837 Do not add referenced vars.
4838 * tree-cfg.c (update_modified_stmts): Do not update when SSA operands
4840 * passes.c (init_optimization_passes): Put mudflap_2 after
4841 free_datastructures.
4843 2007-01-02 Jan Hubicka <jh@suse.cz>
4845 * tree-optimize (execute_fixup_cfg): Set after_inlining flag.
4846 Set NOTHROW flag on call statements proved to be nothrow.
4847 Update statement of local calls so new pure/const functions are
4848 updated. Update_ssa when in ssa form. Mark PHI nodes of nonlocal
4850 (tree_rest_of_compilation): Register hooks and initialize bitmap
4851 early. Do not set after_inlining flag.
4853 2007-01-02 Steve Ellcey <sje@cup.hp.com>
4855 * sbitmap.c (HOST_BITS_PER_LONG_LONG): Change to
4856 HOST_BITS_PER_LONGLONG
4858 2007-01-02 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
4861 * c-typeck.c (store_init_value): Don't emit pedantic overflow
4862 warning for non-static initializers.
4864 2007-01-02 Steven Bosscher <steven@gcc.gnu.org>
4866 * config/alpha/alpha.md, arm/arm.c, darwin.c, frv/frv.md,
4867 m32r/m32r.c, m32r/m32r.c, mn10300/mn10300.md, pa/pa.c,
4868 rs6000/rs6000.c, s390/s390.md, sh/sh.md, sparc/sparc.c:
4869 Always use set_unique_reg_note to add REG_EQUAL notes.
4871 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
4874 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
4876 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
4877 * cfgbuild.c (find_basic_blocks): Likewise.
4878 * cfgrtl.c (rtl_create_basic_block): Likewise.
4879 * function.c (temp_slots_at_level): Likewise.
4880 * reg-stack.c (stack_regs_mentioned): Likewise.
4881 * regclass.c (allocate_reg_info): Likewise.
4882 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
4883 set_bb_for_stmt, move_block_to_fn): Likewise.
4884 * tree-complex.c (tree_lower_complex): Likewise.
4885 * vec.h (VEC_safe_grow_cleared): New.
4887 2007-01-02 Ian Lance Taylor <iant@google.com>
4889 * c-common.c (c_common_truthvalue_conversion): When warning about
4890 using an assignment as a truth value, set TREE_NO_WARNING.
4892 2007-01-02 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
4895 * c.opt (Wold-style-declaration): New.
4896 * doc/invoke.texi (C-only Warning Options): New.
4897 (Wold-style-declaration): Document it.
4898 (Wextra): Enabled by -Wextra.
4899 * c-opts.c (c_common_post_options): Enabled by -Wextra.
4900 * c-decl.c (declspecs_add_scspec): Replace -Wextra with
4901 -Wold-style-declaration.
4903 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
4905 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
4906 * cfgbuild.c (find_basic_blocks): Likewise.
4907 * cfgrtl.c (rtl_create_basic_block): Likewise.
4908 * function.c (temp_slots_at_level): Likewise.
4909 * reg-stack.c (stack_regs_mentioned): Likewise.
4910 * regclass.c (allocate_reg_info): Likewise.
4911 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
4912 set_bb_for_stmt, move_block_to_fn): Likewise.
4913 * tree-complex.c (tree_lower_complex): Likewise.
4914 * vec.h (VEC_safe_grow_cleared): New.
4916 2007-01-02 Douglas Gregor <doug.gregor@gmail.com>
4918 * c-common.c (c_common_nodes_and_builtins): Since variants of
4919 void_type_node get built before it is given a name, we need to
4920 give those variants the name, too.
4921 (complete_array_type): We need to work with the canonical main
4922 type of the array, from which we will build the qualified version.
4923 * params.def (PARAM_VERIFY_CANONICAL_TYPES): New.
4924 * print-tree.c (print_node): Display canonical type information
4926 * stor-layout.c (layout_type): When we don't know the
4927 alignment of a type for which we're building an array, we end up
4928 guessing wrong, so make the type require structural equality.
4929 * tree.c (make_node_stat): When we build a new type, it is its
4931 (build_type_attribute_qual_variant): When building an attribute
4932 variant, its canonical type is the non-attribute variant. However,
4933 if the attributes are target-dependent and they differ, we need to
4934 use structural equality checks for this type.
4935 (build_qualified_type): A qualified type is not equivalent to its
4936 unqualified variant; set the canonical type appropriately.
4937 (build_distinct_type_copy): When building a distinct type from
4938 another type, the new type is its own canonical type.
4939 (build_variant_type_copy): When building a new type variant, we
4940 assume that it is equivalent to the original type.
4941 (build_pointer_type_for_mode): When building a pointer type, also
4942 build a canonical type pointer.
4943 (build_reference_type_for_mode): When building a reference type,
4944 also build a canonical type reference.
4945 (build_index_type): When we can't hash an index type (e.g.,
4946 because its maximum value is negative), the index type requires
4947 structural equality tests.
4948 (build_array_type): Build the canonical form of an array type.
4949 (build_function_type): Function types require structural equality,
4950 because they contain default arguments, attributes, etc.
4951 (build_method_type_directly): Ditto for method types.
4952 (build_offset_type): Build the canonical offset type.
4953 (build_complex_type): Build the canonical vector type.
4954 (make_vector_type): Build the canonical vector type.
4955 * tree.h (TYPE_CANONICAL): New.
4956 (TYPE_STRUCTURAL_EQUALITY_P): New.
4957 (SET_TYPE_STRUCTURAL_EQUALITY): New.
4958 (struct tree_type): Added "canonical" field.
4959 * params.h (VERIFY_CANONICAL_TYPES): New.
4960 * doc/c-tree.texi (TYPE_CANONICAL): Document.
4961 (TYPE_STRUCTURAL_EQUALITY_P): Document.
4962 (SET_TYPE_STRUCTURAL_EQUALITY): Document.
4963 * doc/invoke.texi (verify-canonical-types): Document --param
4964 parameter for verifying canonical types.
4966 2007-01-02 Joseph Myers <joseph@codesourcery.com>
4968 * config.gcc (powerpc-*-eabispe*, powerpc-*-eabisimaltivec*,
4969 powerpc-*-eabisim*, powerpc-*-eabialtivec*, powerpc-*-eabi*,
4970 powerpc-*-rtems*, powerpc-wrs-vxworks, powerpc-wrs-vxworksae,
4971 powerpcle-*-eabisim*, powerpcle-*-eabi*): Add rs6000/e500.h to
4973 * config/rs6000/e500.h: New.
4974 * config/rs6000/eabi.h (TARGET_SPE_ABI, TARGET_SPE, TARGET_E500,
4975 TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE, TARGET_E500_DOUBLE):
4977 * config/rs6000/linuxspe.h (TARGET_SPE_ABI, TARGET_SPE,
4978 TARGET_E500, TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE,
4979 TARGET_E500_DOUBLE): Remove.
4980 * config/rs6000/vxworks.h (TARGET_SPE_ABI, TARGET_SPE,
4981 TARGET_E500, TARGET_ISEL, TARGET_FPRS): Remove.
4982 * config/rs6000/rs6000.h (CHECK_E500_OPTIONS): Define.
4983 * config/rs6000/rs6000.c (rs6000_override_options): Use
4986 2007-01-02 Joseph Myers <joseph@codesourcery.com>
4988 * config/rs6000/rs6000.c (print_operand): Check (TARGET_SPE ||
4989 TARGET_E500_DOUBLE), not TARGET_E500, for %y.
4990 (rs6000_generate_compare, rs6000_emit_sCOND, output_cbranch,
4991 rs6000_emit_cmove): Don't check TARGET_E500.
4992 * config/rs6000/rs6000.md (bunordered, bordered, sunordered,
4993 sordered): Don't check TARGET_E500.
4995 2007-01-01 Eric Christopher <echristo@apple.com>
4997 * config/mips/mips.c (mips_regno_mode_ok_for_base_p): Use
4998 HARD_REGISTER_NUM_P.
5000 2007-01-01 Roger Sayle <roger@eyesopen.com>
5002 * fold-const.c (fold_binary) <EQ_EXPR>: Fold "(X^C1) eq/ne C2" into
5003 "X eq/ne (C1^C2)". Fold "(X^Z) eq/ne (Y^Z)" as "X eq/ne Y" when Z
5004 has no side-effects. Fold "(X^C1) eq/ne (Y^C2)" as "(X^(C1^C2))
5007 2007-01-01 Mike Stump <mrs@apple.com>
5009 * configure.ac: Remove support for building with Apple's gcc-3.1.
5011 2007-01-02 Joseph Myers <joseph@codesourcery.com>
5014 * caller-save.c (add_stored_regs): Only handle SUBREGs if inner
5015 REG is a hard register. Do not modify REG before calling
5017 * rtlanal.c (subreg_get_info): Don't assert size of XMODE is a
5018 multiple of the size of YMODE for certain lowpart cases.
5020 2007-01-01 Andrew Pinski <pinskia@gmail.com>
5023 * gimplify (voidify_wrapper_expr): Update for
5026 2007-01-01 Andreas Schwab <schwab@suse.de>
5029 * config/ia64/ia64.c (ia64_compute_frame_size): Account space for
5030 save of BR0 in extra_spill_size instead of spill_size.
5031 (ia64_expand_prologue): Save BR0 outside of the gr/br/fr spill
5033 (ia64_expand_epilogue): Restore BR0 from its new location.
5035 2007-01-01 Andrew Pinski <pinskia@gmail.com>
5037 * gimplify.c (gimplify_init_constructor <case VECTOR_TYPE>):
5038 Use a temporary variable if the left hand side is not a gimple
5041 2007-01-01 Andrew Pinski <pinskia@gmail.com>
5043 * gimplify.c (gimplify_return_expr): Make the temporary variable
5044 for the return expression, a gimple register variable.
5046 2007-01-01 Jan Hubicka <jh@suse.cz>
5048 * emit-rtl.c (emit_copy_of_insn_after): Do not call copy_insn_1 for
5051 2007-01-01 Mike Stump <mrs@apple.com>
5053 * configure.ac (HAVE_GAS_LITERAL16): Add autoconf check for
5055 * config/darwin.c (machopic_select_rtx_section): Use
5057 (darwin_mergeable_constant_section): Likewise.
5058 * configure: Regenerate.
5059 * config.in: Regenerate.
5061 2007-01-01 Jan Hubicka <jh@suse.cz>
5062 Andrew Pinski <pinskia@gmail.com>
5064 * cgraphunit.c (cgraph_optimize): Call cgraph_add_new_functions
5065 before starting IPA passes.