1 2007-02-13 Paul Brook <paul@codesourcery.com>
3 * config.gcc: Add arm*-*-uclinux-*eabi.
4 * config/arm/uclinux-elf.h (TARGET_OS_CPP_BUILTINS): Define.
5 (SUBTARGET_EXTRA_LINK_SPEC): Define.
6 (STARTFILE_SPEC, ENDFILE_SPEC): Remove broken -shared handling.
7 (LINK_GCC_C_SEQUENCE_SPEC): Undef.
10 * config/arm/arm.c (arm_override_options): Use r9 as EABI PIC
12 * config/arm/uclinux-eabi.h: New file.
13 * config/arm/linux-eabi.h (WCHAR_TYPE): Remove.
14 * config/arm/linux-gas.h (WCHAR_TYPE): Use unsigned long on AAPCS
17 2007-02-13 Ian Lance Taylor <iant@google.com>
19 * common.opt: Add Wstrict-overflow and Wstrict-overflow=.
20 * flags.h (warn_strict_overflow): Declare.
21 (enum warn_strict_overflow_code): Define.
22 (issue_strict_overflow_warning): New static inline function.
23 * opts.c (warn_strict_overflow): New variable.
24 (common_handle_option): Handle OPT_Wstrict_overflow and
25 OPT_Wstrict_overflow_.
26 * c-opts.c (c_common_handle_option): Set warn_strict_overflow for
28 * fold-const.c: Include intl.h.
29 (fold_deferring_overflow_warnings): New static variable.
30 (fold_deferred_overflow_warning): New static variable.
31 (fold_deferred_overflow_code): New static variable.
32 (fold_defer_overflow_warnings): New function.
33 (fold_undefer_overflow_warnings): New function.
34 (fold_undefer_and_ignore_overflow_warnings): New function.
35 (fold_deferring_overflow_warnings_p): New function.
36 (fold_overflow_warning): New static function.
37 (make_range): Add strict_overflow_p parameter. Change all
39 (extract_muldiv, extract_muldiv_1): Likewise.
40 (fold_unary) [ABS_EXPR]: Check ABS_EXPR before calling
41 tree_expr_nonnegative_p.
42 (fold_negate_expr): Call fold_overflow_warning.
43 (fold_range_test): Likewise.
44 (fold_comparison): Likewise.
45 (fold_binary): Likewise. Call tree_expr_nonnegative_warnv_p
46 instead of tree_expr_nonnegative_p.
47 (tree_expr_nonnegative_warnv_p): Rename from
48 tree_expr_nonnegative_p, add strict_overflow_p parameter.
49 (tree_expr_nonnegative_p): New function.
50 (tree_expr_nonzero_warnv_p): Rename from tree_expr_nonzero_p, add
51 strict_overflow_p parameter.
52 (tree_expr_nonzero_p): New function.
53 * passes.c (verify_interpass_invariants): New static function.
54 (execute_one_pass): Call it.
55 * tree-ssa-loop-niter.c (expand_simple_operations): Ignore fold
57 (number_of_iterations_exit, loop_niter_by_eval): Likewise.
58 (estimate_numbers_of_iterations): Likewise.
59 (scev_probably_wraps_p): Likewise.
60 * tree-ssa-ccp.c: Include "toplev.h".
61 (evaluate_stmt): Defer fold overflow warnings until we know we are
63 (struct fold_stmt_r_data): Add stmt field.
64 (fold_stmt_r): Defer fold overflow warnings until we know we
66 (fold_stmt): Initialize stmt field of fold_stmt_r_data.
67 (fold_stmt_inplace): Likewise.
68 * tree-cfgcleanup.c: Include "toplev.h" rather than "errors.h".
69 (cleanup_control_expr_graph): Defer fold overflow warnings until
70 we know we are going to optimize.
71 * tree-cfg.c (fold_cond_expr_cond): Likewise.
72 * tree-ssa-threadedge.c (simplify_control_stmt_condition):
74 * tree-vrp.c (vrp_expr_computes_nonnegative): Call
75 tree_expr_nonnegative_warnv_p instead of tree_expr_nonnegative_p.
76 * tree-ssa-loop-manip.c (create_iv): Likewise.
77 * c-typeck.c (build_conditional_expr): Likewise.
78 (build_binary_op): Likewise.
79 * tree-vrp.c (vrp_expr_computes_nonzero): Call
80 tree_expr_nonzero_warnv_p instead of tree_expr_nonzero_p.
81 (extract_range_from_unary_expr): Likewise.
82 * simplify-rtx.c (simplify_const_relational_operation): Warn when
83 assuming that signed overflow does not occur.
84 * c-common.c (pointer_int_sum): Ignore fold overflow warnings.
85 * tree.h (tree_expr_nonnegative_warnv_p): Declare.
86 (fold_defer_overflow_warnings): Declare.
87 (fold_undefer_overflow_warnings): Declare.
88 (fold_undefer_and_ignore_overflow_warnings): Declare.
89 (fold_deferring_overflow_warnings_p): Declare.
90 (tree_expr_nonzero_warnv_p): Declare.
91 * doc/invoke.texi (Option Summary): Add -Wstrict-overflow to list
93 (Warning Options): Document -Wstrict-overflow.
94 * Makefile.in (tree-ssa-threadedge.o): Depend on toplev.h.
95 (tree-ssa-ccp.o): Likewise.
96 (tree-cfgcleanup.o): Change errors.h dependency to toplev.h.
97 (fold-const.o): Depend on intl.h.
99 2007-02-13 Ian Lance Taylor <iant@google.com>
102 * lower-subreg.c (resolve_simple_move): Decompose subregs in
105 2007-02-13 Stuart Hastings <stuart@apple.com>
107 * gcc/config/i386/i386.md (fixuns_truncdfhi2): Require SSE2.
109 2007-02-13 Richard Henderson <rth@redhat.com>
111 * config/alpha/alpha.c (alpha_stdarg_optimize_hook): Strip
112 handled_component_p before looking for the indirect_ref.
114 2007-02-13 Richard Henderson <rth@redhat.com>
116 * config/i386/i386.md (bswapsi_1): Rename from bswapsi2,
117 remove flags clobber.
118 (bswapsi2): New expander, emit code for !TARGET_BSWAP.
119 (bswaphi_lowpart): New.
120 (bswapdi2): Rename from bswapdi2_rex, remove flags clobber,
121 remove TARGET_BSWAP test. Delete expander of the same name.
123 * optabs.c (widen_bswap, expand_doubleword_bswap): New.
124 (expand_unop): Use them.
126 2007-02-13 Uros Bizjak <ubizjak@gmail.com>
128 * config/i386/i386.md (cmpdi_ccno_1_rex64, *cmpsi_ccno_1,
129 *cmphi_ccno_1, *cmpqi_ccno_1, *movsi_xor, *movstricthi_xor,
130 *movstrictqi_xor, *movdi_xor_rex64, *ashldi3_1_rex64,
131 *ashldi3_cmp_rex64, *ashldi3_cconly_rex64, ashlsi3, *ashlsi3_1_zext,
132 *ashlsi3_cmp, *ashlsi3_cconly, *ashlsi3_cmp_zext, *ashlhi3_1_lea,
133 *ashlhi3_1, *ashlhi3_cmp, *ashlhi3_cconly, *ashlqi3_1_lea,
134 *ashlqi3_1, *ashlqi3_cmp, *ashlqi3_cconly): Remove equivalent
135 assembler dialect choice from asm templates.
137 2007-02-12 Richard Henderson <rth@redhat.com>
139 * config/i386/i386.md (fixuns_trunc<SSEMODEF>si_1): New insn.
140 (fixuns_trunc<SSEMODEF>si2): Use it.
141 * config/i386/sse.md (vec_setv4sf_0): Export.
142 * config/i386/i386.c (ix86_build_const_vector): Export.
143 (ix86_split_convert_uns_si_sse): Rename from
144 ix86_expand_convert_uns_si_sse and rewrite as a splitter.
145 * config/i386/i386-protos.h: Update.
147 2007-02-13 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
150 * c-typeck.c (c_finish_return): Improve warning message.
152 2007-02-12 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
154 * alias.c (find_symbolic_term): Delete unused function.
156 2007-02-12 Uros Bizjak <ubizjak@gmail.com>
158 * config/i386/i386.md (paritydi2, paritysi2): New expanders.
159 (paritydi2_cmp, paritydi2_cmp): New insn and split patterns.
160 (*parityhi2_cmp, *parityqi2_cmp): New insn patterns.
162 2007-02-12 Eric Botcazou <ebotcazou@adacore.com>
164 * tree.h (DECL_IGNORED_P): Document further effect for FUNCTION_DECL.
165 * cgraphunit.c (cgraph_expand_function): If DECL_IGNORED_P is set on
166 the function, temporarily point the debug interface to the null one.
168 2007-02-12 Eric Botcazou <ebotcazou@adacore.com>
170 * dwarf2out.c (round_up_to_align): New static function.
171 (field_byte_offset): Use it to round the offset.
173 2007-02-12 Richard Henderson <rth@redhat.com>
175 * config/alpha/alpha.md (bswapsi2, bswapdi2): New.
176 (inswl_const): Export.
178 2007-02-12 Richard Henderson <rth@redhat.com>
180 * calls.c (emit_library_call_value_1): If PROMOTE_MODE modifed the
181 result mode of the libcall, convert back to outmode.
183 2007-02-12 Roger Sayle <roger@eyesopen.com>
185 * config/i386/i386.md (*bswapdi2_rex): Renamed from bswapdi2.
186 (bswapdi2): New define_expand to implement 32-bit implementation.
188 2007-02-12 Nick Clifton <nickc@redhat.com>
190 * doc/invoke.texi (Overall Options): Document --help=.
191 * gcc.c (target_help_flag): Rename to print_subprocess_flag.
192 (cc1_options): Pass --help= on to cc1.
193 (display_help): Add description of --help=.
194 (process_command): Add code to handle --help=. Allow translated
195 --help and --target-help switches to be passed on to compiler
197 (main): Remove unused if statement.
198 * opts.c (columns): Remove.
199 (LEFT_COLUMN): Define.
200 (wrap_help): Add columns argument.
201 (print_filtered_help): Change parameters to be an include bitmask,
202 an exclude bitmask, an any bitmask and the column width. Move the
203 code to display the params list here. Add code to display the
204 status of options rather than their descriptions if the quiet flag
206 (print_specific_help): Change parameters to be an include bitmask,
207 an exclude bitmask and an any bitmask. Move code to look up the
208 column width here. Decide upon the title for an options listing.
209 (common_handle_options): Add code to handle --help=. Adapt code
210 for --help and --target-help to use the revised form of the
211 print_specific_help function.
212 (print_help): Delete.
213 (print_param_help): Delete.
214 (print_switch): Delete.
215 * opts.h (cl_lang_count): Add prototype.
216 (CL_PARAMS, CL_WARNING, CL_OPTIMIZATION, CL_MIN_OPTION_CLASS,
217 CL_MAX_OPTION_CLASS): New defines.
218 * optc-gen.awk: Add construction of cl_lang_count.
219 * c.opt: Add Warning attribute to warning options and Optimization
220 attribute to optimization options.
221 * common.opt: Likewise.
223 Add -fhelp and -ftarget-help as aliases for the transformed --help
224 and --target-help options.
225 * opt-functions.awk: Add code to handle Warning and Optimization
228 2007-02-12 Richard Henderson <rth@redhat.com>
230 * config/alpha/constraints.md: New file.
231 * config/alpha/alpha.c: Include tm-constrs.h.
232 (alpha_const_ok_for_letter_p, alpha_const_double_ok_for_letter_p,
233 alpha_extra_constraint): Remove.
234 (alpha_emit_conditional_branch): Use satisfies_constraint_*.
235 * config/alpha/alpha-protos.h: Update.
236 * config/alpha/alpha.h (REG_CLASS_FROM_LETTER): Remove.
237 (CONST_OK_FOR_LETTER_P, CONST_DOUBLE_OK_FOR_LETTER_P): Remove.
238 (EXTRA_CONSTRAINT): Remove.
239 * config/alpha/alpha.md: Include constraints.md.
240 (adddi splitter): Use satisfies_constraint_*.
241 * config/alpha/predicates.md (add_operand): Likewise.
242 (sext_add_operand, addition_operation): Likewise.
244 2007-02-12 Dorit Nuzman <dorit@il.ibm.com>
246 PR tree-optimization/29145
247 * tree-data-ref.c (base_addr_differ_p): Make us more conservative
248 in our handling of restrict qualified pointers.
250 2007-02-12 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
253 * doc/invoke.texi (Wunused-value): Update description.
254 (Wextra): Delete item.
255 * opts.c (set_Wextra): Don't use the value of Wextra to set the
256 value of Wunused-value.
257 * c-typeck.c (c_process_expr_stmt): Don't check extra_warnings.
258 (c_finish_stmt_expr): Don't check extra_warnings.
259 (emit_side_effect_warnings): The caller is responsible to check
262 2007-02-11 Roger Sayle <roger@eyesopen.com>
263 Matt Thomas <matt@3am-software.com>
265 * simplify-rtx.c (simplify_relational_operation_1): Correct typo.
267 2007-02-11 Roger Sayle <roger@eyesopen.com>
269 * simplify-rtx.c (simplify_relational_operation_1): Optimize
270 comparisons of POPCOUNT against zero.
271 (simplify_const_relational_operation): Likewise.
273 2007-02-11 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
275 * doc/invoke.texi (Wextra): Delete outdated paragraph.
277 2007-02-11 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
279 * dwarf2out.c (root_type): Delete unused function.
281 2007-02-11 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
283 * genattrtab.c (contained_in_p): Delete unused function.
284 (write_expr_attr_cache): Likewise.
286 2007-02-11 Jan Hubicka <jh@suse.cz>
288 * ipa-inline.c (cgraph_edge_badness): Add "else" missing in the
291 2007-02-11 Steven Bosscher <steven@gcc.gnu.org>
293 * fwprop.c (try_fwprop_subst): Use set_unique_reg_note
294 to add the REG_EQ* note.
295 * see.c (see_merge_one_use_extension): Likewise.
296 * local-alloc.c (update_equiv_regs): Likewise. Also don't
297 turn REG_EQUAL notes into REG_EQUIV notes if the target
298 register may have more than one set.
299 * function.c (assign_parm_setup_reg): Use set_unique_reg_note.
300 * gcse.c (try_replace_reg): Likewise.
301 * alias.c (init_alias_analysis): Use find_reg_equal_equiv_note.
302 * calls.c (fixup_tail_calls): Likewise. Abort if there is
303 more than one REG_EQUIV note.
304 * reload1.c (gen_reload): Use set_unique_reg_note.
306 2007-02-11 Uros Bizjak <ubizjak@gmail.com>
308 * config/i386/i386.c (TARGET_VECTORIZE_BUILTIN_CONVERSION): Define.
309 (ix86_builtin_conversion): New function.
311 2007-02-06 Mark Mitchell <mark@codesourcery.com>
314 * tree.h (DECL_REPLACEABLE_P): New macro.
315 * except.c (set_nothrow_function_flags): Likewise.
317 2007-02-11 Tehila Meyzels <tehila@il.ibm.com>
318 Ira Rosen <irar@il.ibm.com>
319 Dorit Nuzman <dorit@il.ibm.com>
321 * doc/tm.texi (TARGET_VECTORIZE_BUILTIN_CONVERSION): New target hook.
322 * targhooks.c (default_builtin_vectorized_conversion): New.
323 * targhooks.h (default_builtin_vectorized_function): New declaration.
324 * target.h (struct vectorize): Add builtin_conversion field.
325 * tree-vectorizer.h (type_conversion_vec_info_type): New enum
326 stmt_vec_info_type value.
327 (vectorizable_conversion): New declaration.
328 * tree-vect-analyze.c (vect_analyze_operations): Add
329 vectorizable_conversion call.
330 * target-def.h (TARGET_VECTORIZE_BUILTIN_CONVERSION): New.
331 * tree-vect-transform.c (vectorizable_conversion): New function.
332 (vect_transform_stmt): Add case for type_conversion_vec_info_type.
333 * tree-vect-generic.c (expand_vector_operations_1): Consider correct
335 * config/rs6000/rs6000.c (rs6000_builtin_conversion): New.
336 (TARGET_VECTORIZE_BUILTIN_CONVERSION): Defined.
337 (rs6000_expand_builtin): Add handling a case of ALTIVEC_BUILTIN_VCFUX
338 or ALTIVEC_BUILTIN_VCFSX.
340 2007-02-10 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
343 * pa.md (movdf): For 64-bit target, fail if operand 1 is a non-zero
344 CONST_DOUBLE and operand 0 is a hard register.
345 (movdi): For 64-bit target, remove code to force CONST_DOUBLE to
346 memory. Fail if operand 1 is a non-zero CONST_INT and operand 0
347 is a hard floating-point register.
349 2007-02-10 Richard Henderson <rth@redhat.com>
350 Jakub Jelinek <jakub@redhat.com>
351 Alexandre Oliva <aoliva@redhat.com>
353 * Makefile.in (libgcc-support, libgcc.mvars): Add emutls.c.
354 * builtin-types.def (BT_WORD): Make unsigned.
355 (BT_FN_VOID_PTR_WORD_WORD_PTR): New.
356 * builtins.def (BUILT_IN_EMUTLS_GET_ADDRESS): New.
357 (BUILT_IN_EMUTLS_REGISTER_COMMON): New.
358 * c-decl.c (grokdeclarator): Don't error if !have_tls.
359 * c-parser.c (c_parser_omp_threadprivate): Likewise.
360 * dwarf2out.c (loc_descriptor_from_tree_1): Don't do anything for
362 * expr.c (emutls_var_address): New.
363 (expand_expr_real_1): Expand emulated tls.
364 (expand_expr_addr_expr_1): Likewise.
365 * libgcc-std.ver: Add __emutls_get_address, __emutls_register_common.
366 * output.h (emutls_finish): Declare.
367 * toplev.c (compile_file): Call it.
368 * tree-ssa-address.c (gen_addr_rtx): Check for const-ness of the
369 address before wrapping in CONST.
370 * varasm.c (emutls_htab, emutls_object_type): New.
371 (EMUTLS_VAR_PREFIX, EMUTLS_TMPL_PREFIX): New.
372 (get_emutls_object_name, get_emutls_object_type): New.
373 (get_emutls_init_templ_addr, emutls_decl): New.
374 (emutls_common_1, emutls_finish): New.
375 (assemble_variable): When emulating tls, swap decls; generate
376 constructor for the emutls objects.
377 (do_assemble_alias): When emulating tls, swap decl and target name.
378 (default_encode_section_info): Don't add SYMBOL_FLAG_TLS_SHIFT
380 * varpool.c (decide_is_variable_needed): Look at force_output.
381 Recurse for emulated tls.
382 (cgraph_varpool_remove_unreferenced_decls): Remove checks redundant
383 with decide_is_variable_needed.
384 * emutls.c: New file.
385 * config/sparc/sol2.h (ASM_DECLARE_OBJECT_NAME): Only emit
386 tls_object for real tls.
388 2007-02-10 Kaz Kojima <kkojima@gcc.gnu.org>
390 PR rtl-optimization/29599
391 * reload1.c (eliminate_regs_in_insn): Take the destination
392 mode into account when computing the offset.
394 2007-02-09 Stuart Hastings <stuart@apple.com>
395 Richard Henderson <rth@redhat.com>
397 * gcc/config/i386/i386.h (TARGET_KEEPS_VECTOR_ALIGNED_STACK): New.
398 * gcc/config/i386/darwin.h: (TARGET_KEEPS_VECTOR_ALIGNED_STACK): New.
399 * gcc/config/i386/i386.md (fixuns_trunc<mode>si2, fixuns_truncsfhi2,
400 fixuns_truncdfhi2): New.
401 (fix_truncsfdi_sse): Call ix86_expand_convert_sign_didf_sse.
402 (floatunsdidf2): Call ix86_expand_convert_uns_didf_sse.
403 (floatunssisf2): Add call to ix86_expand_convert_uns_sisf_sse.
404 (floatunssidf2): Allow nonimmediate source.
405 * gcc/config/i386/sse.md (movdi_to_sse): New.
406 (vec_concatv2di): Drop '*'.
407 * gcc/config/i386/i386-protos.h (ix86_expand_convert_uns_si_sse,
408 ix86_expand_convert_uns_didf_sse, ix86_expand_convert_uns_sidf_sse,
409 ix86_expand_convert_uns_sisf_sse, ix86_expand_convert_sign_didf_sse):
411 * gcc/config/i386/i386.c (ix86_expand_convert_uns_si_sse,
412 ix86_expand_convert_uns_didf_sse, ix86_expand_convert_uns_sidf_sse,
413 ix86_expand_convert_uns_sisf_sse, ix86_expand_convert_sign_didf_sse,
414 ix86_build_const_vector, ix86_expand_vector_init_one_nonzero): New.
415 (ix86_build_signbit_mask): Fix decl of v, refactor to call
416 ix86_build_const_vector.
417 (x86_emit_floatuns): Rewrite.
419 2007-02-10 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
421 * genautomata.c (longest_path_length): Delete unused function.
422 (struct state): Delete unused longest_path_length.
423 (UNDEFINED_LONGEST_PATH_LENGTH): Delete unused macro.
424 (get_free_state): Delete unused.
426 2007-02-09 Jan Hubicka <jh@suse.cz>
428 * params.def (PARAM_INLINE_UNIT_GROWTH): Set to 30.
429 * doc/invoke.texi (inline-unit-growth): Update default value.
431 * Makefile.in (passes.o, ipa-inline.o): Add dependencies.
432 * cgraphbuild.c (build_cgraph_edges): Compute frequencies.
433 (rebuild_cgraph_edges): Likewise.
434 * cgraph.c (cgraph_set_call_stmt): Add new argument frequency.
435 (dump_cgraph_node): Dump frequencies.
436 (cgraph_clone_edge): Add frequency scales.
437 (cgraph_clone_node): Add freuqnecy.
438 * cgraph.h (cgraph_edge): Add freuqnecy argument.
439 (CGRAPH_FREQ_BASE, CGRAPH_FREQ_MAX): New constants.
440 (cgraph_create_edge, cgraph_clone_edge, cgraph_clone_node): Update.
441 * tree-pass.h (TODO_rebuild_frequencies): New constant.
442 * cgraphunit.c (verify_cgraph_node): Verify frequencies.
443 (cgraph_copy_node_for_versioning): Update call of cgraph_clone_edge.
444 (save_inline_function_body): Likewise.
445 * ipa-inline.c: inluce rtl.h
446 (cgraph_clone_inlined_nods): Update call of cgraph_clone_node.
447 (cgraph_edge_badness): Use frequencies.
448 (cgraph_decide_recursive_inlining): Update clonning.
449 (cgraph_decide_inlining_of_small_function): Dump frequency.
450 * predict.c (estimate_bb_frequencies): Export.
451 * predict.h (estimate_bb_frequencies): Declare.
452 * tree-inline.c (copy_bb): Watch overflows.
453 (expand_call_inline): Update call of cgraph_create_edge.
454 (optimize_inline_calls): Use TODO flags to update frequnecies.
455 * passes.h: Include predict.h
456 (init_optimization_passes): Move profile ahead.
457 (execute_function_todo): Handle TODO_rebuild_frequencies.
459 2007-02-09 Roger Sayle <roger@eyesopen.com>
461 * config/alpha/alpha.c (emit_insxl): Force the first operand of
462 the insbl or inswl pattern into a register.
464 2007-02-09 Roger Sayle <roger@eyesopen.com>
466 * config/ia64/ia64.md (bswapdi2): New define_insn.
468 2007-02-09 Richard Henderson <rth@redhat.com>
470 * config/i386/constraints.md (Ym): New constraint.
471 * config/i386/i386.md (movsi_1): Change Y2 to Yi constraints.
472 (movdi_1_rex64): Split sse and xmm general register moves from
473 memory move alternatives. Use conditional register constraints.
474 (movsf_1, movdf_integer): Likewise.
475 (zero_extendsidi2_32, zero_extendsidi2_rex64): Likewise.
476 (movdf_integer_rex64): New.
477 (pushsf_rex64): Fix output constraints.
478 * config/i386/sse.md (sse2_loadld): Split rm alternative, use Yi.
479 (sse2_stored): Likewise.
480 (sse2_storeq_rex64): New.
481 * config/i386/i386.c (x86_inter_unit_moves): Enable for not
483 (ix86_secondary_memory_needed): Don't bypass TARGET_INTER_UNIT_MOVES
484 for optimize_size. Remove SF/DFmode hack.
486 2007-02-09 Dwarakanath Rajagopal <dwarak.rajagopal@amd.com>
488 * config/i386/driver-i386.c: Turn on -mtune=native for AMDFAM10.
491 2007-02-09 Nathan Sidwell <nathan@codesourcery.com>
492 Richard Sandiford <richard@codesourcery.com>
494 * config.gcc (m68010-*-netbsdelf*, m68k*-*-netbsdelf*)
495 (m68k*-*-openbsd*, m68k-*-linux*): Set default_cf_cpu.
496 (m68k-*-aout*, m68k-*-coff*, m68k-*-uclinux*, m68k-*-rtems*): Add
497 m68k/t-mlib to tmake_file.
498 (m68020-*-elf*, m68k-*-elf*): Likewise. Add t-m68kbare as well.
499 (m68k*-*-*): Use --with-arch to pick a default for --with-cpu.
500 (m680[012]0-*-*, m68k*-*-*): Add support for --with-arch.
501 Allow it to be cf or m68k. Set m68k_arch_family. If that
502 variable is not empty, add t-$m68k_arch_family to tmake_file.
503 Add t-mlibs to tmake_file.
504 * doc/install.texi: Document --with-arch=m68k and --with-arch=cf.
505 * config/m68k/t-cf: New file.
506 * config/m68k/t-m68k: Likewise.
507 * config/m68k/t-mlibs: Likewise.
508 * config/m68k/t-m68kbare (MULTILIB_OPTIONS, MULTILIB_DIRNAMES)
509 (MULTILIB_MATCHES, MULTILIB_EXCEPTIONS): Delete.
510 (M68K_MLIB_DIRNAMES, M68K_MLIB_OPTIONS): Define.
511 * config/m68k/t-m68kelf (MULTILIB_OPTIONS, MULTILIB_DIRNAMES)
512 (MULTILIB_MATCHES, MULTILIB_EXCEPTIONS, LIBGCC, INSTALL_LIBGCC):
514 * config/m68k/t-openbsd (MULTILIB_OPTIONS, LIBGCC): Delete.
515 (INSTALL_LIBGCC): Delete.
516 (M68K_MLIB_DIRNAMES, M68K_MLIB_OPTIONS): Define.
517 * config/m68k/t-rtems (MULTILIB_OPTIONS, MULTILIB_DIRNAMES)
518 (MULTILIB_MATCHES, MULTILIB_EXCEPTIONS): Delete.
519 (M68K_MLIB_CPU): Define.
520 * config/m68k/t-uclinux (MULTILIB_OPTIONS, MULTILIB_DIRNAMES)
521 (MULTILIB_MATCHES, MULTILIB_EXCEPTIONS): Delete.
522 (M68K_MLIB_CPU, M68K_MLIB_OPTIONS, M68K_MLIB_DIRNAMES): Define.
524 2007-02-09 Zdenek Dvorak <dvorakz@suse.cz>
525 Richard Guenther <rguenther@suse.de>
528 * fold-const.c (fold_comparison): Handle obfuscated comparisons
529 against INT_MIN/INT_MAX.
530 * tree-ssa-loop-ivcanon.c (remove_empty_loop): Print to dump
531 file if a loop is removed.
533 2007-02-09 Joseph Myers <joseph@codesourcery.com>
535 * calls.c (store_one_arg): Pass correct alignment to
536 emit_push_insn for non-BLKmode values.
537 * expr.c (emit_push_insn): If STRICT_ALIGNMENT, copy to an
538 unaligned stack slot via a suitably aligned slot.
540 2007-02-08 DJ Delorie <dj@redhat.com>
542 * config/m32c/m32c.c (m32c_unpend_compare): Add default to silence
544 (legal_subregs): Use unsigned char, make const.
545 (m32c_illegal_subreg_p): Use ARRAY_SIZE. Delete unused variables.
547 2007-02-08 Paul Brook <paul@codesourcery.com>
549 * config/arm/lib1funcs.asm (RETLDM): Pop directly into PC when no
550 special interworking needed.
552 2007-02-08 Harsha Jagasia <harsha.jagasia@amd.com>
554 * config/i386/xmmintrin.h: Make inclusion of emmintrin.h
555 conditional to __SSE2__.
556 (Entries below should have been added to first ChangeLog
557 entry for amdfam10 dated 2007-02-05)
558 * config/i386/emmintrin.h: Generate #error if __SSE2__ is not
560 * config/i386/pmmintrin.h: Generate #error if __SSE3__ is not
562 * config/i386/tmmintrin.h: Generate #error if __SSSE3__ is not
565 2007-02-08 DJ Delorie <dj@redhat.com>
567 * config/m32c/m32c-protos.h (m32c_illegal_subreg_p): New.
568 * config/m32c/m32c.c (legal_subregs): New.
569 (m32c_illegal_subreg_p): New.
570 * config/m32c/predicates.md (m32c_any_operand): Use it to reject
571 unsupported subregs of hard regs.
573 2007-02-08 Jan Hubicka <jh@suse.cz>
575 * tree-cfg.c (bsi_replace): Shortcut when replacing the statement with
576 the same one; always update histograms.
578 2007-02-08 Diego Novillo <dnovillo@redhat.com>
580 * passes.c (init_optimization_passes): Tidy comment.
582 2007-02-08 Roger Sayle <roger@eyesopen.com>
584 * simplify-rtx.c (simplify_unary_operation_1) <POPCOUNT>: We can
585 strip zero_extend, bswap and rotates from POCOUNT's argument.
586 <PARITY>: Likewise, we can strip not, bswap, sign_extend,
587 zero_extend and rotates from PARITY's argument.
588 <BSWAP>: A byte-swap followed by a byte-swap is an identity.
589 (simplify_const_unary_operation) <BSWAP>: Evaluate the byte-swap
590 of an integer constant at compile-time.
592 2007-02-08 Diego Novillo <dnovillo@redhat.com>
595 * tree-flow.h (struct var_ann_d): Remove field 'is_used'.
597 * tree-ssa-alias.c (compute_is_aliased): Remove. Update all
600 * tree-ssa-live.c (remove_unused_locals): Do not remove
601 TREE_ADDRESSABLE variables.
602 * tree-ssa-structalias.c (compute_points_to_sets): Tidy.
603 * tree-ssa-operands.c (add_virtual_operand): Remove argument
604 FOR_CLOBBER. Update all users.
605 If VAR has an associated alias set, add a virtual operand for
606 it if no alias is found to conflict with the memory reference.
608 2007-02-07 Jan Hubicka <jh@suse.cz>
609 Robert Kidd <rkidd@crhc.uiuc.edu>
611 * value-prof.c (visit_hist, free_hist): Return 1 instead of 0.
613 2007-02-07 Ian Lance Taylor <iant@google.com>
615 * lower-subreg.c (simple_move): Reject PARTIAL_INT modes.
617 2007-02-07 Roger Sayle <roger@eyesopen.com>
619 * config/rs6000/rs6000.md (ctz<mode>2, ffs<mode>2, popcount<mode>2,
620 parity<mode>2, smulsi3_highpart, abstf2_internal, allocate_stack,
621 tablejumpdi, movsi_to_cr_one): Remove constraints from
622 define_expand's match_operands.
624 2007-02-07 Roger Sayle <roger@eyesopen.com>
626 * global.c (compute_regsets): Move declatation of "i" inside of
627 #ifdef ELIMINABLE_REGS to avoid unused variable bootstrap failure.
629 2007-02-07 Jakub Jelinek <jakub@redhat.com>
632 * gimplify.c (gimplify_scan_omp_clauses): Remove special casing
633 of INDIRECT_REF <RESULT_DECL>.
635 * config/i386/i386.c (override_options): Set PTA_SSSE3 for core2.
637 2007-02-06 J"orn Rennecke <joern.rennecke@arc.com>
638 Kaz Kojima <kkojima@gcc.gnu.org>
641 * config/sh/sh.c (expand_cbranchdi4): Use scratch register
643 (sh_initialize_trampoline): Add parentheses to avoid a warning.
645 2007-02-06 Zdenek Dvorak <dvorakz@suse.cz>
647 * doc/loop.texi: Document possibility not to perform disambiguation
648 of loops with multiple latches.
649 * cfgloopmanip.c (alp_enum_p): Removed.
650 (add_loop): Handle subloops. Use get_loop_body_with_size.
651 (create_preheader): Do not allow ENTRY_BLOCK_PTR to be preheader.
652 * cfghooks.c (redirect_edge_and_branch_force): Set dominator for
653 the new forwarder block.
654 (make_forwarder_block): Only call new_bb_cbk if it is not NULL.
655 Handle the case latch is NULL.
656 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Avoid cfg modifications
657 when marking loop exits.
658 * ifcvt.c (if_convert): Ditto. Mark loop exits even if cfg cannot
660 * loop-init.c (loop_optimizer_init): Do not modify cfg. Call
661 disambiguate_loops_with_multiple_latches.
662 * tree-cfgcleanup.c (cleanup_tree_cfg_loop): Calculate dominators
663 before fix_loop_structure.
664 * cfgloop.c: Include pointer-set.h and output.h.
665 (canonicalize_loop_headers, HEADER_BLOCK, LATCH_EDGE,
666 update_latch_info, mfb_keep_just, mfb_keep_nonlatch): Removed.
667 (get_loop_latch_edges, find_subloop_latch_edge_by_profile,
668 find_subloop_latch_edge_by_ivs, find_subloop_latch_edge,
669 mfb_redirect_edges_in_set, form_subloop, merge_latch_edges,
670 disambiguate_multiple_latches, get_loop_body_with_size,
671 disambiguate_loops_with_multiple_latches): New functions.
672 (flow_loop_dump): Dump multiple latch edges.
673 (flow_loop_nodes_find): Handle loops with multiple latches.
674 (flow_loops_find): Ditto. Do not call canonicalize_loop_headers.
675 (glb_enum_p): Modified.
676 (get_loop_body): Use get_loop_body_with_size.
677 * cfgloop.h (LOOPS_HAVE_RECORDED_EXITS): New flag.
678 (AVOID_CFG_MODIFICATIONS): New constant.
679 (disambiguate_loops_with_multiple_latches, add_loop,
680 get_loop_body_with_size): Declare.
681 * Makefile.in (cfgloop.o): Add pointer-set.h and output.h.
683 2007-02-06 Seongbae Park <seongbae.park@gmail.com>
686 * global.c (compute_regsets): New function.
687 (global_alloc): Refactored ELIMINABLE_REGSET
688 and NO_GLOBAL_ALLOC_REGS computation out.
689 (rest_of_handle_global_alloc): Call compute_regsets()
690 for non-optimizing case.
692 2007-02-06 Richard Henderson <rth@redhat.com>
694 * config/i386/constraints.md (Y2): Rename from Y.
695 (Yi): New constraint.
696 * config/i386/i386.md (movsi_1, movdi_2, pushdf_nointeger,
697 pushdf_integer, movdf_nointeger, movdf_integer, zero_extendsidi2_32,
698 zero_extendsidi2_rex64, truncxfdf2_mixed): Change Y constraints to Y2.
699 (extendsfdf2_mixed, extendsfdf2_sse, truncdfsf_fast_mixed,
700 truncdfsf_fast_sse, truncdfsf_mixed, fix_truncdfdi_sse,
701 fix_truncdfsi_sse, floatsidf2_mixed, floatsidf2_sse,
702 floatdidf2_mixed, floatdidf2_sse, absnegdf2_mixed,
703 absnegdf2_sse, sse_setccdf, fop_df_comm_mixed, fop_df_comm_sse,
704 fop_df_1_mixed, fop_df_1_sse): Change Y constraints to x.
705 * config/i386/mmx.md (mov<MMXMODEI>_internal_rex64,
706 mov<MMXMODEI>_internal, movv2sf_internal_rex64, movv2sf_internal,
707 vec_extractv2si_1): Change Y constraints to Y2.
708 * config/i386/sse.md (vec_setv4sf_0, vec_concatv2df, vec_dupv4si,
709 vec_dupv2di, sse2_concatv2si, vec_concatv4si_1, vec_concatv2di):
710 Change Y constraints to Y2.
711 (sse2_loadld): Change Y constraints to x.
713 2007-02-06 Roger Sayle <roger@eyesopen.com>
715 * config/rs6000/rs6000.md (popcount<mode>2): Rewrite.
716 (parity<mode>2): New define_expand using rs6000_emit_parity.
717 * config/rs6000/rs6000.c (rs6000_emit_popcount,
718 rs6000_emit_parity): New functions.
719 * config/rs6000/rs6000-protos.h (rs6000_emit_popcount,
720 rs6000_emit_parity): Prototype here.
722 2007-02-06 Ian Lance Taylor <iant@google.com>
724 * lower-subreg.c (simple_move_operand): Reject CONST.
725 (resolve_clobber): Call validate_change rather than directly
726 assigning to XEXP (pat, 0).
728 2006-02-06 Paolo Bonzini <bonzini@gnu.org>
730 * Makefile.in (tree-ssa-loop-ivopts.o): Add pointer-set.h dependency.
731 (tree-ssa-reassoc.o): Add pointer-set.h dependency.
732 (tree-cfg.o): Remove hashtab.h dependency.
734 * tree-ssa-loop-ivopts.c: Include pointer-set.h.
735 (struct ivopts_data): Change niters to pointer_map_t.
736 (struct nfe_cache_elt, nfe_hash, nfe_eq): Delete.
737 (niter_for_exit): Create pointer_map on demand. Change for
739 (tree_ssa_iv_optimize_init): Initialize data->niters to NULL.
740 (free_loop_data): Destroy data->niters if created and reset field.
741 (tree_ssa_iv_optimize_finalize): Don't delete data->niters here.
742 (tree_ssa_iv_optimize_loop): Check for presence of stale data.
744 * tree-ssa-reassoc.c: Include pointer-set.h.
745 (bb_rank): Change to long *.
746 (operand_rank): Change to pointer_map_t.
747 (find_operand_rank): Return long, -1 if not found. Declare as inline.
748 (insert_operand_rank): Accept long.
749 (operand_entry_hash, operand_entry_eq): Remove.
750 (get_rank): Return long. Adjust for changes above.
751 (init_reassoc): Change rank type to long. Adjust creation of bb_rank
753 (fini_reassoc): Delete operand_rank with pointer_map_destroy.
755 * tree-ssa-structalias.c (vi_for_tree): Change to pointer_map.
756 (struct tree_vi, tree_vi_t, tree_vi_hash, tree_vi_eq): Delete.
757 (insert_vi_for_tree): Rewrite for pointer_map API. Assert argument
759 (lookup_vi_for_tree): Rewrite for pointer_map API. Return varinfo_t
760 directly since it cannot be NULL.
761 (get_vi_for_tree): Rewrite for pointer_map API.
762 (find_what_p_points_to): Adjust for change to lookup_vi_for_tree.
763 (init_alias_vars): Create vi_for_tree as pointer_map.
764 (delete_points_to_sets): Delete vi_for_tree using pointer_map_destroy.
766 * tree-cfg.c: Don't include hashtab.h.
767 (edge_to_cases): Declare as pointer_map.
768 (struct edge_to_cases_elt, edge_to_cases_hash, edge_to_cases_eq):
770 (edge_to_cases_cleanup): Rewrite as pointer_map_traverse callback.
771 (start_recording_case_labels): Create edge_to_cases as pointer_map.
772 (end_recoding_case_labels): Cleanup edge_to_cases manually before
774 (record_switch_edge): Delete.
775 (get_cases_for_edge): Adjust for pointer_map API, inline
776 record_switch_edge (rewritten for new API), remove goto.
778 2006-02-06 Paolo Bonzini <bonzini@gnu.org>
780 * Makefile.in (tree-nested.o): Add pointer-set.h dependency.
781 * tree-nested.c: Include pointer-set.h.
782 (var_map_elt, var_map_eq, var_map_hash): Delete.
783 (struct nesting_info): Remove GTY marker. Change the two htab_t's
785 (nesting_info_bitmap_obstack): New.
786 (lookup_field_for_decl): Adjust for pointer_map API.
787 (lookup_tramp_for_decl): Adjust for pointer_map API.
788 (get_nonlocal_debug_decl): Adjust for pointer_map API.
789 (get_local_debug_decl): Adjust for pointer_map API.
790 (convert_nl_goto_reference): Adjust for pointer_map API.
791 (convert_nl_goto_receiver): Adjust for pointer_map API.
792 (create_nesting_tree): Create outside GGC space. Create bitmap on
793 the new obstack. Create field_map and var_map as pointer_maps.
794 (free_nesting_tree): Adjust for changes to create_nesting_tree.
796 (lower_nested_functions): Move root here, no need to NULL it.
797 Initialize and release the obstack.
799 2007-02-06 Paolo Bonzini <bonzini@gnu.org>
801 * tree.c (tree_int_map_hash, tree_int_map_eq, tree_int_map_marked_p):
802 Remove prototypes and make them non-static.
803 (struct tree_int_map): Remove.
804 * tree.h (struct tree_int_map): Move here, turning TO into an
806 (tree_int_map_hash, tree_int_map_eq, tree_int_map_marked_p): Declare.
808 * tree.h (TREE_COMPLEXITY): Remove.
809 (struct tree_exp): Remove complexity field.
810 * tree.c (build1_stat): Don't set it.
812 2007-02-06 Dorit Nuzman <dorit@il.ibm.com>
813 Victor Kaplansky <victork@il.ibm.com>
815 * tree-vectorizer.c (vect_is_simple_use): Support induction.
816 (vect_is_simple_reduction): Support reduction with induction as
818 (vect_is_simple_iv_evolution): Fix formatting.
819 * tree-vect-analyze.c (vect_mark_stmts_to_be_vectorized): Fix
820 formatting. Don't mark induction phis for vectorization.
821 (vect_analyze_scalar_cycles): Analyze all inductions, then reductions.
822 * tree-vect-transform.c (get_initial_def_for_induction): New function.
823 (vect_get_vec_def_for_operand): Support induction.
824 (vect_get_vec_def_for_stmt_copy): Fix formatting and add check for
826 (vectorizable_reduction): Support reduction with induction as one of
828 (vectorizable_type_demotion): Use def-type of stmt argument rather
831 * tree-ssa-loop.c (gate_scev_const_prop): Return the value of
832 flag_tree_scev_cprop.
833 * common.opt (tree-scev-cprop): New flag.
835 * tree-vect-transform.c (vect_create_destination_var): Use 'kind' in
836 call to vect_get_new_vect_var.
838 2007-02-06 Ira Rosen <irar@il.ibm.com>
840 * tree-vect-patterns.c (vect_recog_widen_mult_pattern): Check that
842 (vect_pattern_recog_1): Likewise.
844 2007-02-05 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
846 * fold-const.c (negate_expr_p): Handle CONJ_EXPR.
847 (fold_negate_expr): Likewise.
849 2007-02-05 Alexandre Oliva <aoliva@redhat.com>
852 * dwarf2out.c (modified_type_die): Follow DECL_ORIGINAL_TYPE
853 even if cv-qualification is the same.
855 2007-02-05 Geoffrey Keating <geoffk@apple.com>
857 * config/rs6000/darwin-tramp.asm (__trampoline_setup): Call
858 __enable_execute_stack on completion.
860 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
862 * config/i386/athlon.md (athlon_fldxf_k8, athlon_fld_k8,
863 athlon_fstxf_k8, athlon_fst_k8, athlon_fist, athlon_fmov,
864 athlon_fadd_load, athlon_fadd_load_k8, athlon_fadd, athlon_fmul,
865 athlon_fmul_load, athlon_fmul_load_k8, athlon_fsgn,
866 athlon_fdiv_load, athlon_fdiv_load_k8, athlon_fdiv_k8,
867 athlon_fpspc_load, athlon_fpspc, athlon_fcmov_load,
868 athlon_fcmov_load_k8, athlon_fcmov_k8, athlon_fcomi_load_k8,
869 athlon_fcomi, athlon_fcom_load_k8, athlon_fcom): Added amdfam10.
871 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
873 * config/i386/i386.md (x86_sahf_1, cmpfp_i_mixed, cmpfp_i_sse,
874 cmpfp_i_i387, cmpfp_iu_mixed, cmpfp_iu_sse, cmpfp_iu_387,
875 swapsi, swaphi_1, swapqi_1, swapdi_rex64, fix_truncsfdi_sse,
876 fix_truncdfdi_sse, fix_truncsfsi_sse, fix_truncdfsi_sse,
877 x86_fldcw_1, floatsisf2_mixed, floatsisf2_sse, floatdisf2_mixed,
878 floatdisf2_sse, floatsidf2_mixed, floatsidf2_sse,
879 floatdidf2_mixed, floatdidf2_sse, muldi3_1_rex64, mulsi3_1,
880 mulsi3_1_zext, mulhi3_1, mulqi3_1, umulqihi3_1, mulqihi3_insn,
881 umulditi3_insn, umulsidi3_insn, mulditi3_insn, mulsidi3_insn,
882 umuldi3_highpart_rex64, umulsi3_highpart_insn,
883 umulsi3_highpart_zext, smuldi3_highpart_rex64,
884 smulsi3_highpart_insn, smulsi3_highpart_zext, x86_64_shld,
885 x86_shld_1, x86_64_shrd, sqrtsf2_mixed, sqrtsf2_sse,
886 sqrtsf2_i387, sqrtdf2_mixed, sqrtdf2_sse, sqrtdf2_i387,
887 sqrtextendsfdf2_i387, sqrtxf2, sqrtextendsfxf2_i387,
888 sqrtextenddfxf2_i387): Added amdfam10_decode.
890 * config/i386/athlon.md (athlon_idirect_amdfam10,
891 athlon_ivector_amdfam10, athlon_idirect_load_amdfam10,
892 athlon_ivector_load_amdfam10, athlon_idirect_both_amdfam10,
893 athlon_ivector_both_amdfam10, athlon_idirect_store_amdfam10,
894 athlon_ivector_store_amdfam10): New define_insn_reservation.
895 (athlon_idirect_loadmov, athlon_idirect_movstore): Added
898 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
900 * config/i386/athlon.md (athlon_call_amdfam10,
901 athlon_pop_amdfam10, athlon_lea_amdfam10): New
902 define_insn_reservation.
903 (athlon_branch, athlon_push, athlon_leave_k8, athlon_imul_k8,
904 athlon_imul_k8_DI, athlon_imul_mem_k8, athlon_imul_mem_k8_DI,
905 athlon_idiv, athlon_idiv_mem, athlon_str): Added amdfam10.
907 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
909 * config/i386/athlon.md (athlon_sseld_amdfam10,
910 athlon_mmxld_amdfam10, athlon_ssest_amdfam10,
911 athlon_mmxssest_short_amdfam10): New define_insn_reservation.
913 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
915 * config/i386/athlon.md (athlon_sseins_amdfam10): New
916 define_insn_reservation.
917 * config/i386/i386.md (sseins): Added sseins to define_attr type
918 and define_attr unit.
919 * config/i386/sse.md: Set type attribute to sseins for insertq
922 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
924 * config/i386/athlon.md (sselog_load_amdfam10, sselog_amdfam10,
925 ssecmpvector_load_amdfam10, ssecmpvector_amdfam10,
926 ssecomi_load_amdfam10, ssecomi_amdfam10,
927 sseaddvector_load_amdfam10, sseaddvector_amdfam10): New
928 define_insn_reservation.
929 (ssecmp_load_k8, ssecmp, sseadd_load_k8, seadd): Added amdfam10.
931 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
933 * config/i386/athlon.md (cvtss2sd_load_amdfam10,
934 cvtss2sd_amdfam10, cvtps2pd_load_amdfam10, cvtps2pd_amdfam10,
935 cvtsi2sd_load_amdfam10, cvtsi2ss_load_amdfam10,
936 cvtsi2sd_amdfam10, cvtsi2ss_amdfam10, cvtsd2ss_load_amdfam10,
937 cvtsd2ss_amdfam10, cvtpd2ps_load_amdfam10, cvtpd2ps_amdfam10,
938 cvtsX2si_load_amdfam10, cvtsX2si_amdfam10): New
939 define_insn_reservation.
941 * config/i386/sse.md (cvtsi2ss, cvtsi2ssq, cvtss2si,
942 cvtss2siq, cvttss2si, cvttss2siq, cvtsi2sd, cvtsi2sdq,
943 cvtsd2si, cvtsd2siq, cvttsd2si, cvttsd2siq,
944 cvtpd2dq, cvttpd2dq, cvtsd2ss, cvtss2sd,
945 cvtpd2ps, cvtps2pd): Added amdfam10_decode attribute.
947 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
949 * config/i386/athlon.md (athlon_ssedivvector_amdfam10,
950 athlon_ssedivvector_load_amdfam10, athlon_ssemulvector_amdfam10,
951 athlon_ssemulvector_load_amdfam10): New define_insn_reservation.
952 (athlon_ssediv, athlon_ssediv_load_k8, athlon_ssemul,
953 athlon_ssemul_load_k8): Added amdfam10.
955 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
957 * config/i386/i386.h (TARGET_SSE_UNALIGNED_MOVE_OPTIMAL): New macro.
958 (x86_sse_unaligned_move_optimal): New variable.
960 * config/i386/i386.c (x86_sse_unaligned_move_optimal): Enable for
962 (ix86_expand_vector_move_misalign): Add code to generate movupd/movups
963 for unaligned vector SSE double/single precision loads for AMDFAM10.
965 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
967 * config/i386/i386.h (TARGET_AMDFAM10): New macro.
968 (TARGET_CPU_CPP_BUILTINS): Add code for amdfam10.
969 Define TARGET_CPU_DEFAULT_amdfam10.
970 (TARGET_CPU_DEFAULT_NAMES): Add amdfam10.
971 (processor_type): Add PROCESSOR_AMDFAM10.
973 * config/i386/i386.md: Add amdfam10 as a new cpu attribute to match
974 processor_type in config/i386/i386.h.
975 Enable imul peepholes for TARGET_AMDFAM10.
977 * config.gcc: Add support for --with-cpu option for amdfam10.
979 * config/i386/i386.c (amdfam10_cost): New variable.
980 (m_AMDFAM10): New macro.
981 (m_ATHLON_K8_AMDFAM10): New macro.
982 (x86_use_leave, x86_push_memory, x86_movx, x86_unroll_strlen,
983 x86_cmove, x86_3dnow_a, x86_deep_branch, x86_use_simode_fiop,
984 x86_promote_QImode, x86_integer_DFmode_moves,
985 x86_partial_reg_dependency, x86_memory_mismatch_stall,
986 x86_accumulate_outgoing_args, x86_arch_always_fancy_math_387,
987 x86_sse_partial_reg_dependency, x86_sse_typeless_stores,
988 x86_use_ffreep, x86_use_incdec, x86_four_jump_limit,
989 x86_schedule, x86_use_bt, x86_cmpxchg16b, x86_pad_returns):
990 Enable/disable for amdfam10.
991 (override_options): Add amdfam10_cost to processor_target_table.
992 Set up PROCESSOR_AMDFAM10 for amdfam10 entry in
993 processor_alias_table.
994 (ix86_issue_rate): Add PROCESSOR_AMDFAM10.
995 (ix86_adjust_cost): Add code for amdfam10.
997 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
999 * config/i386/i386.opt: Add new Advanced Bit Manipulation (-mabm)
1000 instruction set feature flag. Add new (-mpopcnt) flag for popcnt
1001 instruction. Add new SSE4A (-msse4a) instruction set feature flag.
1002 * config/i386/i386.h: Add builtin definition for SSE4A.
1003 * config/i386/i386.md: Add support for ABM instructions
1005 * config/i386/sse.md: Add support for SSE4A instructions
1006 (movntss, movntsd, extrq, insertq).
1007 * config/i386/i386.c: Add support for ABM and SSE4A builtins.
1008 Add -march=amdfam10 flag.
1009 * config/i386/ammintrin.h: Add support for SSE4A intrinsics.
1010 * doc/invoke.texi: Add documentation on flags for sse4a, abm, popcnt
1012 * doc/extend.texi: Add documentation for SSE4A builtins.
1014 2007-02-05 Bob Wilson <bob.wilson@acm.org>
1016 * config/xtensa/xtensa.c (constantpool_mem_p): Skip over SUBREGs.
1018 2007-02-05 Richard Guenther <rguenther@suse.de>
1020 * tree-vectorizer.h (vectorizable_function): Add argument type
1021 argument, change return type.
1022 * tree-vect-patterns.c (vect_recog_pow_pattern): Adjust caller.
1023 * tree-vect-transform.c (vectorizable_function): Handle extra
1024 argument, return vectorized function decl.
1025 (build_vectorized_function_call): Remove.
1026 (vectorizable_call): Handle calls with result and argument types
1027 differing. Handle loop vectorization factor correctly.
1028 * targhooks.c (default_builtin_vectorized_function): Adjust for
1030 * targhooks.h (default_builtin_vectorized_function): Likewise.
1031 * target.h (builtin_vectorized_function): Add argument type
1033 * config/i386/i386.c (ix86_builtin_vectorized_function): Handle
1034 extra argument, allow vectorizing of lrintf.
1035 * doc/tm.texi (TARGET_VECTORIZE_BUILTIN_VECTORIZED_FUNCTION): Adjust
1036 documentation of target hook.
1038 2007-02-05 Hans-Peter Nilsson <hp@axis.com>
1041 * config/cris/cris.md ("*andsi_movu", "*andsi_clear", "*andhi_movu")
1042 ("*andhi_clear", andu (casesi+45)): For size-changed operand where
1043 memory is allowed, require !side_effects_p, not just !MEM_VOLATILE_P.
1045 2007-02-05 Roger Sayle <roger@eyesopen.com>
1047 * fold-const.c (fold_unary) <REAL_PART>: Test for availability of
1048 BUILT_IN_COS before simplifying REAL_PART(CEXPI)) to COS.
1049 <IMAG_PART>: Likewise, check for availability of BUILT_IN_SIN.
1050 * builtins.c (fold_builtin_sincos): Check for TARGET_C99_FUNCTIONS
1051 before canonicalizing sincos to cexpi.
1052 (fold_builtin_cexp): Likewise, for canonicalizing cexp to cexpi.
1054 2007-02-05 Roger Sayle <roger@eyesopen.com>
1056 * config/alpha/alpha.c (alpha_add_builtins): New Helper function.
1057 Set TREE_READONLY and TREE_NOTHROW directly, not via attributes.
1058 (alpha_init_builtins): Use alpha_add_builtins to process tables.
1060 2007-02-05 Roger Sayle <roger@eyesopen.com>
1062 * mips-tfile.c (initialize_init_file): Correct endianness test.
1064 2007-02-05 Kazu Hirata <kazu@codesourcery.com>
1066 * config/m68k/m68k.md (pushdi-1, pushdi, movsi+1): Don't use
1069 2007-02-05 Richard Sandiford <richard@codesourcery.com>
1071 * dwarf2out.c (dwarf2out_frame_debug_expr): Record the register
1072 saves in a PARALLEL before the register assignments.
1074 2007-02-05 Richard Sandiford <richard@codesourcery.com>
1076 * doc/tm.texi (DWARF_ALT_FRAME_RETURN_COLUMN): Do not require
1077 DWARF_FRAME_RETURN_COLUMN to be a general register.
1078 * dwarf2out.c (init_return_column_size): New function, split from...
1079 (expand_builtin_init_dwarf_reg_sizes): ...here. Allow both
1080 DWARF_FRAME_RETURN_COLUMN and DWARF_ALT_FRAME_RETURN_COLUMN
1081 to be nongeneral registers.
1082 * config/m68k/m68k.h (DWARF_FRAME_REGNUM): Only map FP and
1084 (DWARF_FRAME_REGISTERS, DWARF_FRAME_RETURN_COLUMN): Define.
1085 (DWARF_ALT_FRAME_RETURN_COLUMN): Define.
1087 2007-02-04 Zdenek Dvorak <dvorakz@suse.cz>
1089 * cfgcleanup.c (try_optimize_cfg): Avoid removing ENTRY_BLOCK_PTR.
1091 2007-02-04 Zdenek Dvorak <dvorakz@suse.cz>
1093 * cfgloopmanip.c (loop_delete_branch_edge): Removed.
1094 (remove_path): Use can_remove_branch_p and remove_branch instead
1095 of loop_delete_branch_edge.
1096 * tree-ssa-loop-manip.c (scale_dominated_blocks_in_loop): New function.
1097 (tree_transform_and_unroll_loop): Remove dead branches immediately.
1098 Update profile using scale_dominated_blocks_in_loop.
1099 * cfghooks.c (can_remove_branch_p, remove_branch): New functions.
1100 * cfghooks.h (struct cfg_hooks): Add can_remove_branch_p.
1101 (can_remove_branch_p, remove_branch): Declare.
1102 * tree-cfg.c (tree_can_remove_branch_p): New function.
1103 (tree_cfg_hooks): Add tree_can_remove_branch_p.
1104 * cfgrtl.c (rtl_can_remove_branch_p): New function.
1105 (rtl_cfg_hooks, cfg_layout_rtl_cfg_hook): Add rtl_can_remove_branch_p.
1107 2007-02-05 Jan Hubicka <jh@suse.cz>
1110 * ipa-inline.c (cgraph_clone_inlined_nodes): When there are unanalyzed
1111 nodes in cgraph, don't remove offline copy of the function.
1113 2007-02-04 Jan Hubicka <jh@suse.cz>
1115 * tree-sra.c (sra_walk_expr): Add linebreaks. BITFIELD_REFs into
1116 vectors might cause maybe_lookup_element_for_expr to be called
1117 on non-sra-candidate.
1119 2007-02-04 Kazu Hirata <kazu@codesourcery.com>
1121 * config/bfin/bfin-modes.def, config/bfin/bfin.c,
1122 config/bfin/bfin.md, config/bfin/predicates.md: Follow
1123 spelling conventions.
1125 2007-02-04 Richard Guenther <rguenther@suse.de>
1128 * fold-const.c (try_move_mult_to_index): Make sure to not
1129 overflow one dimension of a multi-dimensional array access.
1131 2007-02-04 Jan Hubicka <jh@suse.cz>
1133 * passes.c (init_optimization_passes): Reindent.
1135 2007-02-04 Jan Hubicka <jh@suse.cz>
1136 Eric Botcazou <ebotcazou@adacore.com>
1138 * tree-optimize.c (has_abnormal_outgoing_edge_p): Move to...
1139 (execute_fixup_cfg): Break out the abnormal goto code.
1140 * tree-inline.c (has_abnormal_outgoing_edge_p): ...here.
1141 (make_nonlocal_label_edges): Move here from execute_fixup_cfg.
1142 (optimize_inline_calls): Call make_nonlocal_label_edges.
1144 2007-02-04 Jan Hubicka <jh@suse.cz>
1146 * tree-ssa-copyrename.c (copy_rename_partition_coalesce): Return
1147 true when something was changed.
1148 (rename_ssa_copies): When something was changed, do
1149 TODO_remove_unused_locals.
1150 * tree-ssa-forwprop.c (tree_ssa_forward_propagate_single_use_value):
1151 add TODO_remove_unused_locals when instruction was removed.
1153 2007-02-04 Jan Hubicka <jh@suse.cz>
1155 * ipa-inline.c (try_inline): Improve debug output; work on already
1157 (cgraph_decide_inlining_incrementally): Indent; improve debug output;
1158 call try_inline for already inlined edges too when flattening;
1159 inline also functions that make callee growth but overall unit size
1162 2007-02-04 Kazu Hirata <kazu@codesourcery.com>
1164 * config/m32c/bitops.md, config/m32c/jump.md,
1165 config/m32c/m32c.c, config/m32c/m32c.h, config/m32r/m32r.c,
1166 config/m32r/m32r.h, config/m32r/m32r.md,
1167 config/m32r/predicates.md, config/m68hc11/larith.asm,
1168 config/m68hc11/m68hc11.c, config/m68hc11/m68hc11.h,
1169 config/m68k/m68k.h, config/mcore/mcore.md, config/mips/4k.md,
1170 config/mips/mips-protos.h, config/mips/mips.c,
1171 config/mips/mips.h, config/mips/mips.md, config/mips/mips16.S,
1172 config/mn10300/mn10300.h, config/mn10300/predicates.md,
1173 config/mt/mt.c, config/mt/mt.h, config/mt/mt.md: Follow
1174 spelling conventions.
1176 * config/v850/v850.c, config/v850/v850.h, config/v850/v850.md:
1177 Follow spelling conventions.
1179 2007-02-03 Douglas Gregor <doug.gregor@gmail.com>
1181 * c-opts.c (c_common_post_options): If C++0x mode is enabled, don't
1182 warn about C++0x compatibility.
1184 2007-02-04 Kazu Hirata <kazu@codesourcery.com>
1186 * config/h8300/h8300.c, config/h8300/h8300.h,
1187 config/h8300/h8300.md: Follow spelling conventions.
1189 2007-02-03 Uros Bizjak <ubizjak@gmail.com>
1192 * combine.c (try_combine): Do not substitute source operand
1193 with constants wider than 2 * HOST_BITS_PER_WIDE_INT.
1195 2007-02-03 Jan Hubicka <jh@suse.cz>
1197 PR gcov-profile/30650
1198 * value-prof.c (stringop_block_profile): Fix handling of size counter;
1199 do not divide by zero for never executed counters.
1200 (tree_find_values_to_profile): Fix counters.
1201 * gcov-ui.h (GCOV_COUNTER_AVERAGE, GCOV_COUNTER_IOR): Fix comments.
1203 2007-02-03 Ian Lance Taylor <iant@google.com>
1205 * lower-subreg.c (simple_move_operand): New static function,
1206 broken out of simple_move. Reject LABEL_REF, SYMBOL_REF, and HIGH
1208 (simple_move): Call simple_move_operand.
1209 (find_decomposable_subregs): Add special handling of MEMs.
1210 (can_decompose_p): Rename from cannot_decompose_p. Reverse
1211 meaning of return value. If we see a hard register, test whether
1212 it can store a word_mode value. Change all callers.
1214 2007-02-03 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
1216 * pa.md (addvdi3, addvsi3, subvdi3, subvsi3, negvdi2, negvsi2): New
1217 ftrapv insns and expanders.
1218 (subdi3): Change define_expand operand 1 to arith11_operand, and
1219 operand 2 to reg_or_0_operand. Change constraints of 64-bit insn
1220 pattern to handle reg_or_0 operands. Revise 32-bit insn pattern to
1221 handle 11-bit constants and reg_or_0 operands in operands 1 and 2,
1225 * varasm.c (notice_global_symbol): Treat global objects as weak when
1228 2007-02-03 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1230 * emit-rtl.c (dconstpi): Delete.
1232 (init_emit_once): Delete dconstpi and init dconstsqrt2.
1233 * real.h (dconstpi): Delete.
1235 * builtins.c (fold_builtin_cabs): Use dconstsqrt2.
1236 (fold_builtin_hypot): Likewise.
1238 2007-02-03 Tom Tromey <tromey@redhat.com>
1241 * gcc.c (cpp_unique_options): Any of -ggdb3, -gstabs3,
1242 -gcoff3, -gxcoff3, -gvms3 implies -dD.
1244 2007-02-03 Kazu Hirata <kazu@codesourcery.com>
1246 * c-decl.c, config/avr/avr.c, config/avr/avr.h,
1247 config/m68k/m68k.c, config/m68k/netbsd-elf.h,
1248 config/mn10300/mn10300.c, config/pdp11/pdp11.h,
1249 config/rs6000/cell.md, config/rs6000/darwin.h,
1250 config/sh/sh.md, config/sh/sh4-300.md, config/spu/spu.c,
1251 config/spu/spu.md, cselib.c, expr.c, haifa-sched.c, hwint.h,
1252 jump.c, reload.c, sched-deps.c, sched-int.h, tree-inline.c,
1253 tree-profile.c, tree-ssa-live.h, tree-vrp.c: Fix comment
1254 typos. Follow spelling conventions.
1255 * doc/invoke.texi: Follow spelling conventions.
1257 2007-02-03 Roger Sayle <roger@eyesopen.com>
1259 * simplify-rtx.c (simplify_relational_operation_1): Implement some
1260 canonicalization transformations that attempt to simplify integer
1261 constant comparisons to become comparisons against zero.
1263 2007-02-03 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1265 * builtins.c (fold_builtin_cabs): Fold cabs(x+xi) into
1267 * fold-const.c (fold_binary): Fix comment typos. Fold complex
1268 (x,0)-(0,y) into (x,-y). Likewise (0,y)-(x,0) into (-x,y).
1270 2007-02-02 Mike Stump <mrs@apple.com>
1272 * config/darwin9.h (DARWIN_LINKER_GENERATES_ISLANDS): Add.
1273 * config/rs6000/rs6000.c (DARWIN_GENERATE_ISLANDS): Add.
1274 (output_call): Use DARWIN_GENERATE_ISLANDS to decide when to
1275 generate a branch island.
1277 2007-02-02 Bob Wilson <bob.wilson@acm.org>
1279 * config/xtensa/xtensa.c (smalloffset_mem_p): Use BASE_REG_P.
1280 (xtensa_legitimate_address_p): New.
1281 (xtensa_legitimize_address): New.
1282 (xtensa_output_addr_const_extra): New.
1283 * config/xtensa/xtensa.h (REG_OK_STRICT_FLAG): Define.
1285 (REG_OK_FOR_BASE_P): Use BASE_REG_P.
1286 (GO_IF_LEGITIMATE_ADDRESS): Move code to xtensa_legitimate_address_p.
1287 (LEGITIMIZE_ADDRESS): Move code to xtensa_legitimize_address.
1288 (OUTPUT_ADDR_CONST_EXTRA): Move code to xtensa_output_addr_const_extra.
1289 * config/xtensa/xtensa-protos.h (xtensa_legitimate_address_p): New.
1290 (xtensa_legitimize_address): New.
1291 (xtensa_output_addr_const_extra): New.
1293 2007-02-02 Steve Ellcey <sje@cup.hp.com>
1295 * config/ia64/ia64.c (ia64_print_operand): Fix compare strings.
1297 2007-02-02 Ian Lance Taylor <iant@google.com>
1299 * expmed.c (expand_divmod): Add comment.
1301 2007-02-02 Kazu Hirata <kazu@codesourcery.com>
1303 * emit-rtl.c (renumber_insns): Remove.
1304 * flags.h: Remove the extern for flag_renumber_insns.
1305 * rtl.h: Remove the prototype for renumber_insns.
1306 * toplev.c (flag_renumber_insns): Remove.
1308 2007-02-02 Hui-May Chang <hm.chang@apple.com>
1310 Revert for x86 darwin:
1311 2005-06-19 Uros Bizjak <uros@kss-loka.si>
1313 * config/i386/i386.c (ix86_function_arg_regno_p): Put back the
1314 code before the following patch under TARGET_MACHO.
1315 (ix86_function_value_regno_p): Likewise.
1317 2007-02-02 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1319 * fold-const.c (negate_expr_p, fold_negate_expr): Handle
1322 2007-02-02 Maxim Kuvyrkov <mkuvyrkov@ispras.ru>
1325 * config/ia64/ia64.c (ia64_speculate_insn): Restrict to memory loads to
1326 general or fp registers. Add comments.
1327 * config/ia64/ia64.md (reg_pred_prefix): Add comment.
1329 2007-02-02 Paolo Bonzini <bonzini@gnu.org>
1331 * pointer-set.c (insert_aux): Only return insertion slot.
1332 (pointer_set_insert): Adjust.
1333 (pointer_set_traverse, struct pointer_map_t, pointer_map_create,
1334 pointer_map_destroy, pointer_map_insert, pointer_map_contains,
1335 pointer_map_traverse): New.
1336 * pointer-set.h (pointer_set_traverse, struct pointer_map_t,
1337 pointer_map_create, pointer_map_destroy, pointer_map_insert,
1338 pointer_map_contains, pointer_map_traverse): Declare.
1340 2007-02-02 Jakub Jelinek <jakub@redhat.com>
1343 * builtins.c (fold_builtin_sprintf): Do not attempt to optimize
1344 sprintf (str, "%s"). Do not optimize sprintf (str, "nopercent", p++).
1346 2007-02-02 Maxim Kuvyrkov <mkuvyrkov@ispras.ru>
1348 * sched-int.h (ds_to_dk, dk_to_ds): Declare functions.
1350 (struct _dep): New type.
1351 (dep_t): New typedef.
1352 (DEP_PRO, DEP_CON, DEP_KIND): New access macros.
1353 (DEP_STATUS): New access macro. The macro with the same name was
1354 renamed to DEP_LINK_STATUS.
1355 (dep_init): Declare function
1357 (struct _dep_link): New type.
1358 (dep_link_t): New typedef.
1359 (DEP_LINK_NODE, DEP_LINK_NEXT, DEP_LINK_PREV_NEXTP): New access macros.
1360 (DEP_LINK_DEP, DEP_LINK_PRO, DEP_LINK_CON, DEP_LINK_KIND): New macros.
1361 (DEP_LINK_STATUS): New macro.
1362 (debug_dep_links): New debug function.
1364 (struct _deps_list): New type.
1365 (deps_list_t): New typedef.
1366 (DEPS_LIST_FIRST): New access macro.
1367 (FOR_EACH_DEP_LINK): New cycle macro.
1368 (create_deps_list, free_deps_list, delete_deps_list): Declare
1370 (deps_list_empty_p, debug_deps_list, add_back_dep_to_deps_list): Ditto.
1371 (find_link_by_pro_in_deps_list, find_link_by_con_in_deps_list): Ditto.
1372 (copy_deps_list_change_con): Ditto.
1374 (move_dep_link): Declare function.
1376 (struct _dep_node): New type.
1377 (dep_node_t): New typedef.
1378 (DEP_NODE_BACK, DEP_NODE_DEP, DEP_NODE_FORW): New access macros.
1380 (struct haifa_insn_data.back_deps): New field to hold backward
1381 dependencies of the insn.
1382 (struct haifa_insn_data.depend): Rename to forw_deps. Change its type
1384 (struct haifa_insn_data.resolved_deps): Rename to resolved_back_deps.
1385 Change its type to deps_list_t.
1386 (INSN_BACK_DEPS): New access macro to use instead of LOG_LINKS.
1387 (INSN_DEPEND): Rename to INSN_FORW_DEPS.
1388 (RESOLVED_DEPS): Rename to INSN_RESOLVED_BACK_DEPS.
1390 (INSN_COST): Move to haifa-sched.c. Use insn_cost () instead.
1392 (DEP_STATUS): Rename to DEP_LINK_STATUS. Fix typo in the comment.
1394 (add_forw_dep, delete_back_forw_dep, insn_cost): Update declaration and
1396 (dep_cost): Declare.
1398 * sched-deps.c (CHECK): New macro to (en/dis)able sanity checks.
1399 (ds_to_dk, dk_to_ds): New functions.
1401 (init_dep_1): New static function.
1402 (init_dep): New function.
1403 (copy_dep): New static function.
1405 (dep_link_consistent_p, attach_dep_link, add_to_deps_list): New static
1407 (detach_dep_link): New static function.
1408 (move_dep_link): New function.
1410 (dep_links_consistent_p, dump_dep_links): New static functions.
1411 (debug_dep_links): New debugging function.
1413 (deps_obstack, dl_obstack, dn_obstack): New static variables.
1415 (alloc_deps_list, init_deps_list): New static functions.
1416 (create_deps_list): New function.
1417 (clear_deps_list): New static function.
1418 (free_deps_list, delete_deps_list, deps_list_empty_p): New functions.
1419 (deps_list_consistent_p, dump_deps_list): New static functions.
1420 (debug_deps_list): New function.
1421 (add_back_dep_to_deps_list, find_link_by_pro_in_deps_list): New
1423 (find_link_by_con_in_deps_list, copy_deps_list_change_con): Ditto.
1425 (maybe_add_or_update_back_dep_1, add_or_update_back_dep_1): Update to
1426 use new scheduler dependencies lists.
1427 (add_back_dep, delete_all_dependences, fixup_sched_groups): Ditto.
1428 (sched_analyze): Ditto. Initialize dependencies lists.
1429 (add_forw_dep, compute_forward_dependences): Update to use new
1430 scheduler dependencies lists.
1432 (init_dependency_caches): Init deps_obstack.
1433 (free_dependency_caches): Free deps_obstack.
1435 (adjust_add_sorted_back_dep, adjust_back_add_forw_dep): Update to use
1436 new scheduler dependencies lists.
1437 (delete_forw_dep, add_or_update_back_forw_dep): Ditto.
1438 (add_back_forw_dep, delete_back_forw_dep): Ditto.
1440 * sched-rgn.c (set_spec_fed, find_conditional_protection, is_pfree):
1441 Update to use new scheduler dependencies lists.
1442 (is_conditionally_protected, is_prisky, add_branch_dependences): Ditto.
1443 (debug_dependencies): Ditto.
1444 (schedule_region): Update comments.
1446 * sched-ebb.c (earliest_block_with_similiar_load): Update to use new
1447 scheduler dependencies lists.
1448 (schedule_ebb): Update comments.
1450 * rtl.def (DEPS_LIST): Remove.
1452 * lists.c (unused_deps_list): Remove.
1453 (free_list): Update assertions.
1455 (alloc_DEPS_LIST, free_DEPS_LIST_list, free_DEPS_LIST_node): Remove.
1456 (remove_free_DEPS_LIST_elem, copy_DEPS_LIST_list): Ditto.
1458 * rtl.h (free_DEPS_LIST_list, alloc_DEPS_LIST): Remove declarations.
1459 (remove_free_DEPS_LIST_elem, copy_DEPS_LIST_list): Ditto.
1461 * haifa-sched.c (comments): Update.
1462 (insn_cost1): Remove. Inline the code into insn_cost ().
1463 (insn_cost): Update to use new scheduler dependencies lists. Move
1464 processing of the dependency cost to dep_cost ().
1465 (dep_cost): New function. Use it instead of insn_cost () when
1466 evaluating cost of the dependency. Use compatible interface to
1467 interact with the target.
1468 (priority): Update to use new scheduler dependencies lists.
1469 (rank_for_schedule): Ditto. Optimize heuristic that prefers the insn
1470 with greater number of insns that depend on the insn.
1471 (schedule_insn): Update to use new scheduler dependencies lists. Add
1472 code to free backward dependencies lists. Inline and optimize code
1473 from resolve_dep () - see PR28071.
1474 (ok_for_early_queue_removal): Update to use new scheduler dependencies
1475 lists. Update call to targetm.sched.is_costly_dependence hook.
1477 (fix_inter_tick, try_ready, fix_tick_ready): Update to use new
1478 scheduler dependencies lists.
1480 (resolve_dep): Remove. Move the logic to schedule_insn ().
1481 (init_h_i_d): Initialize dependencies lists.
1483 (process_insn_depend_be_in_spec): Rename to
1484 process_insn_forw_deps_be_in_spec. Update to use new scheduler
1486 (add_to_speculative_block, create_check_block_twin, fix_recovery_deps):
1487 Update to use new scheduler dependencies lists.
1488 (clear_priorities, calc_priorities, add_jump_dependencies): Ditto.
1490 * ddg.c (create_ddg_dependence, create_ddg_dep_no_link): Update to use
1491 new scheduler dependencies lists.
1492 (build_intra_loop_deps): Ditto.
1494 * target.h (struct _dep): Declare to use in
1495 gcc_target.sched.is_costly_dependence.
1496 (struct gcc_target.sched.adjust_cost): Fix typo.
1497 (struct gcc_target.sched.is_costly_dependence): Change signature to use
1498 single dep_t parameter instead of an equivalent triad.
1499 (struct gcc_target.sched.adjust_cost_2): Remove.
1501 * target-def.h (TARGET_SCHED_ADJUST_COST_2): Remove.
1503 * reg-notes.def (DEP_TRUE, DEP_OUTPUT, DEP_ANTI): Update comments.
1505 * doc/tm.texi (TARGET_SCHED_IS_COSTLY_DEPENDENCE): Update
1507 (TARGET_SCHED_ADJUST_COST_2): Remove documentation.
1509 * doc/rtl.texi (LOG_LINKS): Remove part about instruction scheduler.
1510 (REG_DEP_TRUE): Document.
1512 * config/ia64/ia64.c (ia64_adjust_cost_2): Rename to ia64_adjust_cost.
1513 Change signature to correspond to the targetm.sched.adjust_cost hook.
1514 Update use in TARGET_SCHED_ADJUST_COST_2.
1515 (TARGET_SCHED_ADJUST_COST_2): Rename to TARGET_SCHED_ADJUST_COST.
1516 (ia64_dependencies_evaluation_hook, ia64_dfa_new_cycle): Update to use
1517 new scheduler dependencies lists.
1518 (ia64_gen_check): Ditto.
1520 * config/mips/mips.c (vr4130_swap_insns_p): Update to use new scheduler
1523 * config/rs6000/rs6000.c (rs6000_is_costly_dependence): Change
1524 signature to correspond to the targetm.sched.is_costly_dependence hook.
1525 (is_costly_group): Update to use new scheduler dependencies lists.
1527 * config/spu/spu.c (spu_sched_adjust_cost): Use insn_cost () function
1528 instead of INSN_COST () macro.
1530 2007-02-01 Ian Lance Taylor <iant@google.com>
1532 * lower-subreg.c (resolve_clobber): Handle a subreg of a concatn.
1534 2007-02-01 Guy Martin <gmsoft@gentoo.org>
1536 * pa.md (tp_load): Correct mfctl instruction syntax.
1538 2007-02-01 Geoffrey Keating <geoffk@apple.com>
1540 * config/rs6000/rs6000.c (rs6000_stack_info): Correct
1541 altivec_padding_size calculation on AIX. Improve comment, add
1542 assert to verify that it's right.
1544 * config/rs6000/darwin.h (MD_UNWIND_SUPPORT): Don't define for
1547 2007-2-01 Seongbae Park <seongbae.park@gmail.com>
1550 * global.c (global_alloc): Add mising initialization of
1553 2007-02-01 Roger Sayle <roger@eyesopen.com>
1555 * alias.c (init_alias_analysis): Correct whitespace.
1556 * bb-reorder.c (fix_edges_for_rarely_executed_code,
1557 partition_hot_cold_basic_blocks): Likewise.
1558 * builtins.c (expand_builtin_printf, expand_builtin_fprintf,
1559 expand_builtin_sprintf, fold_builtin_carg, fold_builtin_sprintf,
1560 maybe_emit_sprintf_chk_warning, fold_builtin_sprintf_chk,
1561 fold_builtin_snprintf_chk, fold_builtin_printf,
1562 fold_builtin_fprintf, do_mpfr_ckconv, do_mpfr_arg1, do_mpfr_arg2,
1563 do_mpfr_arg3, do_mpfr_sincos): Likewise.
1564 * cfgcleanup.c (cleanup_cfg): Likewise.
1565 * cfgexpand.c (tree_expand_cfg): Likewise.
1566 * fold-const.c (fold_binary) <RDIV_EXPR>: Likewise.
1567 * function.c (get_next_funcdef_no): Likewise.
1568 * gengtype.c (main): Likewise.
1569 * genmodes.c (main): Likewise.
1570 * gcse.c (bypass_conditional_jumps, print_ldst_list): Likewise.
1571 * haifa-sched.c (schedule_block, extend_h_i_d): Likewise.
1572 * ifcvt.c (noce_emit_move_insn): Likewise.
1573 * modulo-sched.c (generate_prolog_epilog, sms_schedule_by_order):
1575 * stor-layout.c (get_best_mode): Likewise.
1576 * tree-ssa-loop-niter.c (get_val_for): Likewise.
1577 * tree-ssa-structalias.c (get_varinfo, get_varinfo_fc,
1578 scc_visit, do_ds_constraint, do_complex_constraint, label_visit,
1579 perform_var_substitution, solve_graph): Likewise.
1580 * tree-vrp.c (vrp_finalize): Likewise.
1582 2007-02-01 Ian Lance Taylor <iant@google.com>
1584 * lower-subreg.c (simplify_gen_subreg_concatn): If we ask for the
1585 high part of a paradoxical subreg, return a constant zero.
1587 2007-02-01 Zdenek Dvorak <dvorakz@suse.cz>
1589 * toplev.c (lang_dependent_init): Call init_set_costs.
1590 * loop-init.c (loop_optimizer_init): Do not call init_set_costs.
1592 2007-02-01 Richard Guenther <rguenther@suse.de>
1595 * fold-const.c (fold_negate_expr): Allow negating a
1596 constant if overflow does not change.
1598 2007-02-01 Ralf Wildenhues <Ralf.Wildenhues@gmx.de>
1600 * doc/c-tree.texi (Expression trees): Improve markup.
1601 * doc/tm.texi (Register Classes, Addressing Modes)
1602 (Floating Point): Fix spacing after abbreviations. Fix some
1605 2007-02-01 Ben Elliston <bje@au.ibm.com>
1607 * doc/invoke.texi: Replace "bugfix" with "bug fix" throughout.
1608 * doc/contrib.texi: Likewise.
1609 * doc/install.texi: Likewise.
1611 2007-01-31 Richard Henderson <rth@redhat.com>
1612 Ian Lance Taylor <iant@google.com>
1614 * lower-subreg.c: New file.
1615 * rtl.def (CONCATN): Define.
1616 * passes.c (init_optimization_passes): Add pass_lower_subreg and
1618 * emit-rtl.c (update_reg_offset): New static function, broken out
1619 of gen_rtx_REG_offset.
1620 (gen_rtx_REG_offset): Call update_reg_offset.
1621 (gen_reg_rtx_offset): New function.
1622 * regclass.c: Revert patch of 2006-03-05, restoring
1624 (clear_reg_info_regno): New function.
1625 * dwarf2out.c (concatn_loc_descriptor): New static function.
1626 (loc_descriptor): Handle CONCATN.
1627 * common.opt (fsplit_wide_types): New option.
1628 * opts.c (decode_options): Set flag_split_wide_types when
1630 * timevar.def (TV_LOWER_SUBREG): Define.
1631 * rtl.h (gen_reg_rtx_offset): Declare.
1632 (reg_scan_update): Declare.
1633 * regs.h (clear_reg_info_regno): Declare.
1634 * tree-pass.h (pass_lower_subreg): Declare.
1635 (pass_lower_subreg2): Declare.
1636 * doc/invoke.texi (Option Summary): List -fno-split-wide-types.
1637 (Optimize Options): Add -fsplit-wide-types to -O1 list. Document
1639 * doc/rtl.texi (Regs and Memory): Document concat and concatn.
1640 * Makefile.in (OBJS-common): Add lower-subreg.o.
1641 (lower-subreg.o): New target.
1643 2007-01-31 Kazu Hirata <kazu@codesourcery.com>
1645 * config/sh/sh.h (HAVE_SECONDARY_RELOADS): Remove.
1647 2007-01-31 Anatoly Sokolov <aesok@post.ru>
1650 * config/avr/avr.c (DWARF2_ADDR_SIZE): Define.
1652 2007-01-31 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1655 * builtins.c (fold_builtin_sqrt): Use MPFR for constant args.
1657 2007-01-31 Zdenek Dvorak <dvorakz@suse.cz>
1659 * cfgloop.h: Include vec-prim.h.
1660 (enum li_flags): Remove LI_ONLY_OLD.
1661 (loop_iterator): Changed.
1662 (fel_next, fel_init): Iterate over loop tree.
1663 (FOR_EACH_LOOP_BREAK): New macro.
1664 * loop-unswitch.c (unswitch_loops): Do not pass LI_ONLY_OLD to
1666 * tree-ssa-loop-unswitch.c (tree_ssa_unswitch_loops): Ditto.
1667 * modulo-sched.c (sms_schedule): Ditto.
1668 * tree-vectorizer.c (vectorize_loops): Ditto.
1669 * doc/loop.texi: Update information on loop numbering and behavior of
1670 FOR_EACH_LOOP wrto new loops.
1671 * tree-scalar-evolution.c (compute_overall_effect_of_inner_loop,
1672 add_to_evolution_1): Test nestedness of loops instead of comparing
1674 * tree-chrec.c (chrec_fold_plus_poly_poly,
1675 chrec_fold_multiply_poly_poly, chrec_evaluate,
1676 hide_evolution_in_other_loops_than_loop, chrec_component_in_loop_num,
1677 reset_evolution_in_loop): Ditto.
1678 * Makefile.in (CFGLOOP_H): Add vecprim.h dependency.
1680 2007-01-31 Dirk Mueller <dmueller@suse.de>
1682 * c-common.c (warn_about_parentheses): Separate warning about
1683 un-parenthized sequence of comparison operators from the one
1684 which is supposed to warn about x <= y <= z.
1686 2007-01-31 Uros Bizjak <ubizjak@gmail.com>
1688 * optabs.h (enum optab_index): Add new OTI_isinf.
1689 (isinf_optab): Define corresponding macro.
1690 * optabs.c (init_optabs): Initialize isinf_optab.
1691 * genopinit.c (optabs): Implement isinf_optab using isinf?f2
1693 * builtins.c (mathfn_built_in): Handle BUILT_IN_ISINF{,F,L}.
1694 (expand_builtin_interclass_mathfn): Expand BUILT_IN_ISINF{,F,L}
1696 (expand_builtin): Expand BUILT_IN_ISINF{,F,L} using
1697 expand_builtin_interclass_mathfn.
1698 * reg_stack.c (subst_stack_regs_pat): Handle UNSPEC_FXAM.
1699 * config/i386/i386.md (UNSPEC_FXAM): New constant.
1700 (fxam<mode>2_i387): New insn pattern.
1701 (isinf<mode>2) New expander to implement isinf, isinff and isinfl
1702 built-in functions as x87 inline asm.
1704 2007-01-31 Kazu Hirata <kazu@codesourcery.com>
1706 * gcc/config/arm/unwind-arm.h (_sleb128_t, _uleb128_t): New.
1708 2007-01-30 Eric Christopher <echristo@apple.com>
1710 * config.gcc: Add geode.
1712 2007-01-31 Kazu Hirata <kazu@codesourcery.com>
1714 * cgraphunit.c, config/arm/arm.c, config/m68k/m68k.c,
1715 ipa-inline.c, tree-profile.c, tree-ssa-live.c,
1716 tree-ssa-math-opts.c, tree-ssanames.c, tree-vect-analyze.c,
1717 value-prof.c: Fix comment typos.
1719 2007-01-30 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
1722 * doc/invoke.texi (Wpointer-arith): Document warning.
1724 2007-01-30 Janis Johnson <janis187@us.ibm.com>
1726 * doc/extend.texi (Decimal Floating Types): Remove decfloat.h from
1727 the list of discrepancies from the draft TR.
1729 2007-01-30 Dirk Mueller <dmueller@suse.de>
1732 * doc/invoke.texi (-Wreturn-type): Update description to
1735 2007-01-30 Richard Sandiford <richard@codesourcery.com>
1737 * cfgrtl.c (try_redirect_by_replacing_jump): Check only_sets_cc0_p.
1739 2007-01-30 Uros Bizjak <ubizjak@gmail.com>
1741 * builtins.c (expand_builtin_int_interclass_roundingfn): New function
1742 to handle optabs that operate on floating point input argument and
1743 output to integer output.
1744 (expand_builtin_mathfn) [BUILT_IN_ILOGB]: Move from here ...
1745 (expand_builtin_interclass_mathfn) [BUILT_IN_ILOGB]: ... to here.
1746 (expand_builtin): Expand BUILT_IN_ILOGB{,F,L} using
1747 expand_builtin_interclass_mathfn ().
1748 * config/i386/i386.md (fxtractxf3_i387): Rename from *fxtractxf3_i387.
1750 (ilogbxf2, ilogb<mode>2): New expanders to implement ilogb, ilogbf and
1751 ilogbl built-in functions as x87 intrinsics.
1753 2007-01-30 Richard Guenther <rguenther@suse.de>
1756 * dwarf2out.c (reference_to_unused): Query varpool if the
1757 variable was output.
1759 2007-01-30 Richard Guenther <rguenther@suse.de>
1762 * passes.c (execute_one_pass): Reset in_gimple_form to not
1763 confuse non-unit-at-a-time mode.
1765 2007-01-29 Roger Sayle <roger@eyesopen.com>
1766 Richard Guenther <rguenther@suse.de>
1768 * fold-const.c (round_up): Make HIGH an unsigned HOST_WIDE_INT to
1769 avoid undefined behaviour on overflow. Use force_fit_type_double
1770 to construct the constant with the specified TREE_OVERFLOW.
1772 2007-01-29 Janis Johnson <janis187@us.ibm.com>
1774 * config/dfp-bit.c: Add parameterized support for fp exceptions.
1775 * config/dfp-bit.h: Ditto.
1777 2007-01-29 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
1779 * c-decl.c (pop_scope): Replace warnings with call to
1780 warn_for_unused_label.
1781 * c-common.h (warn_for_unused_label): Declare.
1782 * c-common.c (warn_for_unused_label): Define.
1784 2007-01-29 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
1786 * tree-optimize.c (update_inlined_to_pointers): Delete unused
1789 2007-01-29 Janis Johnson <janis187@us.ibm.com>
1791 * Makefile.in (USER_H): Remove decfloat.h.
1792 * ginclude/decfloat.h: Delete, moving contents to ...
1793 * ginclude/float.h: Add support for decimal floating point,
1794 guarded by __STDC_WANT_DEC_FP__.
1796 2007-01-29 Mike Stump <mrs@apple.com>
1798 * doc/gccint.texi (Top): Rename Loop Representation to Loop
1799 Analysis and Representation to resolve case insensitive conflict.
1800 * doc/loop.texi (Loop Analysis and Representation): Likewise.
1802 2007-01-28 Daniel Berlin <dberlin@dberlin.org>
1804 * tree.h (struct tree_memory_tag): Add aliases member.
1805 (MTAG_ALIASES): New macro.
1806 * tree-ssa-alias.c (alias_bitmap_obstack): New variable.
1807 (add_may_alias): Remove pointer-set. Update for may_aliases being
1809 (mark_aliases_call_clobbered): Update for may_aliases being a
1811 (compute_tag_properties): Ditto.
1812 (create_partition_for): Ditto.
1813 (compute_memory_partitions): Ditto.
1814 (dump_may_aliases_for): Ditto.
1815 (is_aliased_with): Ditto.
1816 (add_may_alias_for_new_tag): Ditto.
1817 (rewrite_alias_set_for): Rewrite for may_aliases being a bitmap.
1818 (compute_is_aliased): New function.
1819 (compute_may_aliases): Call compute_is_aliased).
1820 (init_alias_info): Initialize alias_bitmap_obstack.
1821 (union_alias_set_into): New function.
1822 (compute_flow_sensitive_aliasing): Use union_aliases_into.
1823 (have_common_aliases_p): Rewrite to take two bitmaps and use
1825 (compute_flow_insensitive_aliasing): Stop using pointer-sets.
1827 (finalize_ref_all_pointers): Update for add_may_alias changes.
1828 (new_type_alias): Ditto.
1829 * tree-flow-inline.h (may_aliases): Return a bitmap.
1830 * tree-dfa.c (dump_variable): Check for MTAG_P'ness.
1831 * tree-ssa.c (verify_flow_insensitive_alias_info): Update for
1832 may_aliases being a bitmap.
1833 * tree-flow.h (struct var_ann_d): Remove may_aliases member.
1834 may_aliases now returns a bitmap.
1835 * tree-ssa-structalias.c (merge_smts_into): Update for may_aliases
1837 * tree-ssa-operands.c (add_virtual_operand): Update for
1838 may_aliases being a bitmap.
1840 2007-01-29 Daniel Berlin <dberlin@dberlin.org>
1842 PR tree-optimization/30630
1843 * tree-ssa-structalias.c (do_complex_constraint): Mark correct
1844 variable as changed.
1846 2007-01-29 Simon Martin <simartin@users.sourceforge.net>
1849 * gimplify.c (gimplify_target_expr): Make sure that the TARGET_EXPR is
1850 expanded only once even if an error occurs.
1852 2007-01-29 Ben Elliston <bje@au.ibm.com>
1854 * gcov-io.h (__gcov_indirect_call_profiler): Declare.
1855 (__gcov_average_profiler): Likewise.
1856 (__gcov_ior_profiler): Likewise.
1857 (__gcov_merge_ior): Likewise.
1859 2007-01-28 Jan Hubicka <jh@suse.cz>
1861 * builtins.c (expand_builtin_memset): Fix typo in my last patch.
1862 * value-prof.c (stringop_block_profile): Likewise.
1864 2007-01-28 Jan Hubicka <jh@suse.cz>
1866 * expr.c (emit_block_move_via_movmem, emit_block_move_via_libcall): Add
1867 variant handling histograms; add wrapper.
1868 (clear_storage_via_libcall): Export.
1869 (emit_block_move_hints): Break out from ...; add histograms.
1870 (emit_block_move): ... this one.
1871 (clear_storage_hints): Break out from ...; add histograms.
1872 (clear_storage): ... this one.
1873 (set_storage_via_memset): Handle histogram.
1874 * expr.h (emit_block_move_via_libcall, emit_block_move_hints): Declare.
1875 (clear_storage_hints, clear_storage_via_libcall): Declare.
1876 (set_storage_via_setmem): Update prototype.
1877 * doc/md.texi (movmem, setmem): Document new arguments.
1879 * value-prof.c (dump_histogram_value, tree_find_values_to_profile): Add
1881 (stringop_block_profile): New global function.
1882 (tree_stringops_values_to_profile): Profile block size and alignment.
1883 * value-prof.h (enum hist_type): add HIST_TYPE_AVERAGE and
1885 (struct profile_hooks): Add gen_average_profiler and gen_ior_profiler.
1886 (stringop_block_profile): Declare.
1887 * builtins.c: Include value-prof.h.
1888 (expand_builtin_memcpy, expand_builtin_memset): Pass block profile.
1889 * gcov-ui.h (GCOV_COUNTER_NAMES): Add new counter.
1890 (GCOV_COUNTER_AVERAGE, GCOV_COUNTER_IOR): New constants.
1891 (GCOV_COUNTERS, GCOV_LAST_VALUE_COUNTER): Update.
1892 * profile.c (instrument_values): Add new counters.
1893 * cfgexpand.c (expand_gimple_basic_block): Propagate histograms to
1895 * tree-profile.c (tree_average_profiler_fn, tree_ior_profiler_fn): New.
1896 (tree_init_edge_profiler): Build new profilers.
1897 (tree_gen_average_profiler, tree_gen_ior_profiler): New.
1898 (pass_tree_profile): Add dump.
1899 (tree_profile_hooks): Update.
1900 * Makefile.in (LIBGCOV): Add new constants.
1901 * libgcov.c (__gcov_merge_ior, __gcov_average_profiler,
1902 __gcov_ior_profiler): New.
1903 * i386.md (movmem/setmem expanders): Add new optional arguments.
1905 2007-01-28 David Edelsohn <edelsohn@gnu.org>
1907 * doc/md.texi (Standard Pattern Names): Document blockage pattern.
1909 2007-01-28 Zdenek Dvorak <dvorakz@suse.cz>
1911 * tree-ssa-loop-unswitch.c: Include tree-inline.h.
1912 (tree_unswitch_single_loop): Pass eni_size_weights to
1913 tree_num_loop_insns.
1914 * tree-ssa-loop-manip.c: Include tree-inline.h.
1915 (can_unroll_loop_p): Pass eni_size_weights to
1916 tree_num_loop_insns.
1917 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p):
1918 Pass eni_size_weights to estimate_num_insns.
1919 * tree.h (init_inline_once): Export.
1920 * toplev.c (backend_init): Call init_inline_once.
1921 * cgraphunit.c (cgraph_process_new_functions,
1922 cgraph_analyze_function): Pass eni_inlining_weights to
1924 * ipa-inline.c (compute_inline_parameters): Ditto.
1925 * tree-ssa-loop-ivcanon.c (tree_num_loop_insns): Pass weights
1926 to estimate_num_insns.
1927 (try_unroll_loop_completely): Pass eni_size_weights to
1928 tree_num_loop_insns.
1929 * tree-eh.c (decide_copy_try_finally): Pass eni_size_weights
1930 ot estimate_num_insns.
1931 * tree-ssa-loop-prefetch.c: Include tree-inline.h.
1932 (loop_prefetch_arrays): Pass eni_time_weights to tree_num_loop_insns.
1933 * tree-inline.c (eni_inlining_weights, eni_size_weights,
1934 eni_time_weights): New variables.
1935 (init_inline_once): Initialize them.
1936 (struct eni_data): Mew.
1937 (estimate_num_insns_1, estimate_num_insns): Use weights.
1938 * tree-inline.h (struct eni_weights_d): New.
1939 (eni_inlining_weights, eni_size_weights, eni_time_weights): Declare.
1940 (estimate_num_insns): Declaration changed.
1941 * cfgloop.h (tree_num_loop_insns): Declaration changed.
1942 * Makefile.in (tree-ssa-loop-unswitch.o, tree-ssa-loop-prefetch.o,
1943 tree-ssa-loop-manip.o): Add TREE_INLINE_H dependency.
1945 2007-01-28 Zdenek Dvorak <dvorakz@suse.cz>
1947 * tree-data-ref.c (conflict_fn): Assert that the number of affine
1948 relations in the conflict function is valid.
1950 2007-01-27 Ian Lance Taylor <iant@google.com>
1952 * common.opt: Add fstrict-overflow.
1953 * opts.c (decode_options): Set flag_strict_overflow if -O2.
1954 * flags.h (TYPE_OVERFLOW_WRAPS): Define.
1955 (TYPE_OVERFLOW_UNDEFINED): Define.
1956 (TYPE_OVERFLOW_TRAPS): Define. This replaces TYPE_TRAP_SIGNED.
1958 * tree.h (TYPE_TRAP_SIGNED): Don't define.
1959 * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_UNDEFINED.
1960 (fold_negate_expr): Likewise.
1961 (make_range): Likewise.
1962 (extract_muldiv_1): Likewise.
1963 (maybe_canonicalize_comparison): Likewise.
1964 (fold_comparison): Likewise.
1965 (fold_binary): Likewise.
1966 (tree_expr_nonnegative_p): Likewise.
1967 (tree_expr_nonzero_p): Likewise.
1968 * tree-vrp.c (compare_values): Likewise.
1969 (extract_range_from_binary_expr): Likewise.
1970 (extract_range_from_unary_expr): Likewise.
1971 * tree-ssa-loop-niter.c (infer_loop_bounds_from_signedness):
1973 (nowrap_type_p): Likewise.
1974 * tree-scalar-evolution.c (simple_iv): Likewise.
1975 * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_WRAPS.
1976 (build_range_check): Likewise.
1977 (extract_muldiv_1): Likewise.
1978 (fold_comparison): Likewise.
1979 * tree-vrp.c (vrp_int_const_binop): Likewise.
1980 (extract_range_from_unary_expr): Likewise.
1981 * convert.c (convert_to_integer): Likewise.
1982 * fold-const.c (fold_negate_expr): Use TYPE_OVERFLOW_TRAPS.
1983 (fold_comparison): Likewise.
1984 (fold_binary): Likewise.
1985 * optabs.c (optab_for_tree_code): Likewise.
1986 * tree-vectorizer.c (vect_is_simple_reduction): Likewise.
1987 * simplify-rtx.c (simplify_const_relational_operation): Check
1988 flag_strict_overflow and flag_trapv.
1989 (simplify_const_relational_operation): Likewise.
1990 * doc/invoke.texi (Option Summary): Mention -fstrict-overflow.
1991 (Optimize Options): Add -fstrict-overflow to -O2 list. Document
1994 2007-01-27 Roger Sayle <roger@eyesopen.com>
1996 * tree.c (tree_fold_gcd): Delete.
1997 * tree.h (tree_fold_gcd): Remove prototype.
1998 * tree-data-ref.c (tree_fold_divides_p): Don't use tree_fold_gcd to
1999 test whether one constant integer is a multiple of another. Instead
2000 call int_const_binop with TRUNC_MOD_EXPR and test for a zero result.
2001 * fold-const.c (multiple_of_p): We've determined both TOP and
2002 BOTTOM are integer constants so we can call int_const_binop directly
2003 instead of the more generic const_binop.
2005 2007-01-27 Roger Sayle <roger@eyesopen.com>
2007 * fold-const.c (size_binop): In the fast-paths for X+0, 0+X, X-0 and
2008 1*X check that the constant hasn't overflowed, to preserve the
2010 (round_up): Provide an efficient implementation when rouding-up an
2011 INTEGER_CST to a power-of-two.
2013 2007-01-28 Ralf Wildenhues <Ralf.Wildenhues@gmx.de>
2015 * doc/sourcebuild.texi: Add comma for clarity.
2016 * doc/extend.texi: Fix some typos.
2017 * doc/passes.texi: Likewise.
2018 * doc/cppinternals.texi: Likewise.
2019 * doc/c-tree.texi: Likewise.
2020 * doc/tree-ssa.texi: Likewise.
2021 * doc/install.texi: Likewise.
2023 2007-01-27 Jan Hubicka <jh@suse.cz>
2025 * tree-sra.c (sra_walk_function): Don't rely on aliases being build.
2026 (pass_sra): Do not require alias information.
2027 * passes.c (init_optimization_passes): Add SRA
2029 2007-01-27 Steven Bosscher <steven@gcc.gnu.org>
2031 * tracer.c (rest_of_handle_tracer): We already cleaned
2032 up the CFG in tracer() so don't do it here again.
2033 * cfgcleanup.c (rest_of_handle_jump2): Don't repeat
2034 cleanup_cfg here, either. And don't call renumber_insns.
2036 * cfgrtl.c (rtl_verify_flow_info_1): Don't verify that BB_END
2037 and BB_HEAD are in the insn stream here. Instead make sure
2038 that BB_INSN is valid on all insns. Also, do check here that
2039 there are no pending branch predictions...
2040 (rtl_verify_flow_info): ...instead of doing it here. Checks
2041 for BB_END and BB_HEAD moved from rtl_verify_flow_info_1 to
2044 2007-01-26 Roger Sayle <roger@eyesopen.com>
2046 * config/i386/i386.c (ix86_swap_binary_operands_p): New helper
2047 function to simplify/factorize operand order canonicalization.
2048 (ix86_fixup_binary_operands): Reorganize using the above function.
2049 (ix86_binary_operator_ok): Likewise.
2051 2007-01-27 Jakub Jelinek <jakub@redhat.com>
2053 * genattrtab.c (struct attr_value_list, insn_code_values): Move to
2054 file scope from optimize_attrs.
2055 (simplify_test_exp): If insn_code_values is not NULL, use it to speed
2057 (optimize_attrs): Clear insn_code_values after freeing it.
2059 2007-01-26 Zdenek Dvorak <dvorakz@suse.cz>
2061 * tree-ssa-address.c (create_mem_ref): Remove ", bsi" from
2062 a parts.base assignment.
2064 2007-01-26 Zdenek Dvorak <dvorakz@suse.cz>
2066 * tree-data-ref.c (dump_subscript): Use dump_conflict_function.
2067 (compute_subscript_distance, initialize_data_dependence_relation,
2068 finalize_ddr_dependent, analyze_ziv_subscript,
2069 analyze_siv_subscript_cst_affine,
2070 compute_overlap_steps_for_affine_univar,
2071 compute_overlap_steps_for_affine_1_2, analyze_subscript_affine_affine,
2072 analyze_siv_subscript, analyze_miv_subscript,
2073 analyze_overlapping_iterations, subscript_dependence_tester_1,
2074 compute_self_dependence, free_dependence_relation): Work
2075 with affine_fn instead of chrecs.
2076 (dump_affine_function, dump_conflict_function, affine_function_equal_p,
2077 common_affine_function, affine_function_base,
2078 affine_function_constant_p, affine_fn_op, affine_fn_plus,
2079 affine_fn_minus, affine_fn_free, conflict_fn_not_known,
2080 conflict_fn_no_dependence, free_conflict_function, free_subscripts,
2081 conflict_fn, affine_fn_cst, affine_fn_univar): New functions.
2082 (all_chrecs_equal_p): Removed.
2083 * tree-data-ref.h (affine_fn, conflict_function): New types.
2084 (struct subscript): Change type of conflicting_iterations_in_a
2085 and conflicting_iterations_in_b.
2087 2007-01-26 Steve Ellcey <sje@cup.hp.com>
2090 * config/pa/pa.h (TARGET_HPUX_11): New.
2091 * config/pa/pa-hpux11.h (TARGET_HPUX_11): New.
2092 * config/pa/pa.c (pa_init_builtins): Use TARGET_HPUX_11.
2094 2007-01-26 Daniel Berlin <dberlin@dberlin.org>
2095 Richard Guenther <rguenther@suse.de>
2097 * tree-ssa-structalias.c (solve_graph): Handle case
2098 we merged the variable to another.
2100 2007-01-25 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
2102 * builtins.c (fold_builtin_1): Treat ccos and ccosh as 'even'
2105 * fold-const.c (negate_mathfn_p): Treat casin, casinh, catan,
2106 catanh, cproj, csin, csinh, ctan and ctanh as 'odd' functions.
2108 2007-01-25 DJ Delorie <dj@redhat.com>
2110 * config/m32c/m32c.c (m32c_cannot_change_mode_class): We don't
2111 allow changes to modes which don't fit in those registers.
2113 * reload1.c (choose_reload_regs): Check for invalid subregs before
2114 computing their locations, not after.
2116 2007-01-25 Geoffrey Keating <geoffk@apple.com>
2119 * config/rs6000/rs6000.c (first_altivec_reg_to_save): On Darwin,
2120 save Altivec registers in an eh_return function.
2121 (compute_vrsave_mask): Likewise.
2122 (rs6000_stack_info): Correct AIX/Darwin stack alignment computation
2123 for saving Altivec registers.
2124 (rs6000_emit_prologue): Don't allocate stack twice in
2125 eh_return function. Correct expected value of altivec_save_offset
2126 when using save_world. Describe save of R0 to stack when using
2127 save_world. Describe stack pointer adjustment when using
2128 save_world. Remove duplicated eh_return parameter register saving.
2129 Update sp_offset variable after save_world.
2130 * config/rs6000/t-darwin (LIB2FUNCS_STATIC_EXTRA): Remove
2132 (LIB2FUNCS_EXTRA): Add darwin-world.asm.
2133 * config/rs6000/darwin.h (SUBTARGET_OVERRIDE_OPTIONS): -m64
2136 2007-01-25 Steve Ellcey <sje@cup.hp.com>
2138 * config.gcc (ia64*-*-hpux*): Make posix threads the default.
2140 2007-01-25 Steve Ellcey <sje@cup.hp.com>
2143 * config/pa/pa.c (pa_init_builtins): Set asm names for finite routines.
2144 * config/ia64/ia64.c (ia64_init_builtins): Ditto.
2146 2007-01-25 Richard Guenther <rguenther@suse.de>
2148 * doc/invoke.texi (-Wcoverage-mismatch): Document.
2149 * common.opt (-Wcoverage-mismatch): New warning option.
2150 * coverage.c (get_coverage_counts): Ignore coverage mismatch
2151 if -Wcoverage-mismatch is given.
2153 2007-01-25 Razya Ladelsky <razya@il.ibm.com>
2155 * ipa-cp.c (ipcp_insert_stage, ipcp_driver): Support for SSA.
2156 (ipcp_driver): Change to static definition.
2157 Add dumping of the ifunctions.
2158 (constant_val_insert): Remove unused parameter. Support for SSA.
2159 (ipcp_propagate_const): Support for SSA.
2160 (ipcp_profile_bb_print): Print only analyzed nodes.
2161 (ipcp_replace_map_create): Remove support for Fortran constant
2163 * ipa-prop.c (ipa_method_modify_stmt,
2164 ipa_callsite_compute_param): Support for SSA.
2165 * ipa-prop.h (ipcp_driver): Remove declaration.
2166 (IS_VALID_TREE_MAP_INDEX): Add define.
2168 2007-01-24 Geoffrey Keating <geoffk@apple.com>
2170 * unwind-dw2.c (execute_stack_op): Handle DW_OP_swap.
2172 2007-01-24 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
2175 * builtins.c (fold_builtin_cabs): Use MPFR to evaluate a
2176 constant argument to cabs and do it without checking for
2177 -funsafe-math-optimizations.
2179 2007-01-24 Douglas Gregor <dgregor@osl.iu.edu>
2181 * c-common.h (RID_FIRST_CXX0X): New.
2182 (RID_LAST_CXX0X): New.
2183 * c-opts.c (c_common_handle_option): -Wc++0x-compat is triggered
2185 * c.opt (Wc++0x-compat): New.
2186 * doc/invoke.texi (-Wc++0x-compat): Document.
2188 2007-01-24 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
2190 * builtins.c (fold_builtin_carg): New.
2191 (fold_builtin_1): Use it.
2193 2007-01-24 Jan Hubicka <jh@suse.cz>
2195 * ipa-inline.c (cgraph_decide_inlining): Initialize initial_insns.
2197 * ipa-inline.c (initial_insns, max_insns): Delete.
2198 (compute_max_insns): New function.
2199 (cgraph_decide_inlining_of_small_function): Use it; take minimal amount
2200 of insns as base for code growth.
2201 (cgraph_decide_inlining): Make initial_insns local; do not compute
2203 * params.def (PARAM_INLINE_UNIT_GROWTH): Set to 60.
2204 * doc/invoke.texi (inline-unit-growth): Update docs.
2206 2007-01-24 Jakub Jelinek <jakub@redhat.com>
2208 * config/i386/i386.h (x86_cmpxchg16b): Remove const.
2209 (TARGET_CMPXCHG16B): Define to x86_cmpxchg16b.
2210 * config/i386/i386.c (x86_cmpxchg16b): Remove const.
2211 (override_options): Add PTA_CX16 flag. Set x86_cmpxchg16b
2212 for CPUs that have PTA_CX16 set.
2215 * gimplify.c (omp_check_private): New function.
2216 (gimplify_scan_omp_clauses): Use it for
2217 firstprivate/lastprivate/reduction.
2220 * gimplify.c (omp_add_variable): Don't call omp_notice_variable
2221 on TYPE_SIZE_UNIT for GOVD_LOCAL VLAs.
2224 * omp-low.c (lower_omp_for_lastprivate): Add dlist argument.
2225 If lower_lastprivate_clauses emits some statements, append them
2226 to dlist rather than body_p and to body_p append an initializer.
2227 (lower_omp_for): Adjust caller.
2229 2007-01-24 Steve Ellcey <sje@cup.hp.com>
2231 * target.h (globalize_decl_name): New.
2232 * target-def.h (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
2233 * output.h (default_globalize_decl_name): New.
2234 * varasm.c (asm_output_bss): Use globalize_decl_name instead of
2236 (globalize_decl): Ditto.
2237 (default_globalize_decl_name): New.
2238 * config/ia64/ia64.c (ia64_globalize_decl_name): New.
2239 (ia64_handle_version_id_attribute): New.
2240 (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
2241 (ia64_asm_output_external): Use globalize_decl_name instead
2243 * doc/extend.texi (version_id): New pragma.
2244 * doc/tm.texi (ARGET_ASM_GLOBALIZE_DECL_NAME): New target hook.
2246 2007-01-24 Andreas Krebbel <krebbel1@de.ibm.com>
2248 * unwind-dw2-fde.c (get_cie_encoding): Replaced _Unwind_Word with
2249 _uleb128_t and _Unwind_SWord with _sleb128_t.
2250 * unwind-dw2.c (extract_cie_info, execute_stack_op, execute_cfa_program,
2251 uw_frame_state_for, uw_update_context_1): Likewise.
2252 * unwind-c.c (parse_lsda_header, PERSONALITY_FUNCTION): Likewise.
2253 * unwind-pe.h (read_uleb128, read_sleb128,
2254 read_encoded_value_with_base): Likewise.
2255 * unwind-generic.h: Define _sleb128_t and _uleb128_t types.
2257 2007-01-24 Richard Guenther <rguenther@suse.de>
2259 * builtins.c (expand_builtin_cexpi): Get the fndecl
2260 for cexp in the correct way.
2262 2007-01-24 Jan Hubicka <jh@suse.cz>
2264 * tree-ssa-dce.c (eliminate_unnecesary_stmts): Remove dead LHS of calls.
2266 2007-01-24 Andreas Krebbel <krebbel1@de.ibm.com>
2268 * c-cppbuiltin.c (builtin_define_type_sizeof): New function.
2269 (c_cpp_builtins): New builtin macros: __SIZEOF_INT__, __SIZEOF_LONG__,
2270 __SIZEOF_LONG_LONG__, __SIZEOF_SHORT__, __SIZEOF_POINTER__,
2271 __SIZEOF_FLOAT__, __SIZEOF_DOUBLE__, __SIZEOF_LONG_DOUBLE__,
2272 __SIZEOF_SIZE_T__, __SIZEOF_WCHAR_T__, __SIZEOF_WINT_T__ and
2273 __SIZEOF_PTRDIFF_T__.
2274 * doc/cpp.texi: Documentation for the new builtin macros added.
2276 2007-01-24 Uros Bizjak <ubizjak@gmail.com>
2278 * config/i386/i386.md (tanxf2, tan<mode>2, atan<mode>2, log<mode>2,
2279 log10<mode>2, log2<mode>2, expxf2, exp10xf2, exp2xf2): Use op2
2280 instead of operands[2] to avoid access past the end of array.
2282 2007-01-24 Richard Sandiford <richard@codesourcery.com>
2284 * reload1.c (emit_reload_insns): Pass the reload register
2285 for a non-spill output reload through forget_old_reloads_1.
2287 2007-01-23 Joseph Myers <joseph@codesourcery.com>
2289 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Define
2290 _SOFT_DOUBLE if doubles use software floating-point.
2291 * config/rs6000/libgcc-ppc-glibc.ver: Export additional long
2292 double functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
2293 * config/rs6000/darwin-ldouble.c: Also compile functions for
2294 hard-float without FPRs. Use fmsub function for all __NO_FPRS__
2295 cases. Compile extra functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
2296 * config/rs6000/linuxspe.h (SUBSUBTARGET_OVERRIDE_OPTIONS): Remove
2297 commented-out long double override.
2298 (CPP_LONGDOUBLE_DEFAULT_SPEC): Likewise.
2299 * config/rs6000/eabispe.h: Likewise.
2300 * config/rs6000/rs6000.c (rs6000_override_options): Don't override
2301 long double for non-SPE.
2302 (rs6000_handle_option): Likewise.
2303 (invalid_e500_subreg): Disallow more subregs involding DImode,
2304 DFmode, TImode or TFmode.
2305 (rs6000_legitimate_offset_address_p): Check TFmode offsets for
2307 (legitimate_lo_sum_address_p): Also check for TFmode for E500
2309 (rs6000_legitimize_address): Also handle TFmode for E500 double.
2310 (rs6000_legitimize_reload_address): Also handle TFmode for E500
2312 (rs6000_legitimate_address): Also check for TFmode for E500
2314 (rs6000_emit_move): Use DFmode subregs of TFmode for E500 double.
2315 (spe_build_register_parallel): Handle TFmode and TCmode.
2316 (rs6000_spe_function_arg): Handle TFmode and TCmode for E500
2318 (function_arg): Handle TFmode and TCmode for E500 double.
2319 (rs6000_init_libfuncs): Initialize extra libfuncs for soft double
2321 (print_operand): Handle TFmode and TImode for %y.
2322 (rs6000_generate_compare): Handle TFmode comparisons for E500
2324 (spe_func_has_64bit_regs_p): Check for TFmode for E500 double.
2325 (rs6000_function_value): Handle TFmode and TCmode for E500 double.
2326 (rs6000_libcall_value): Handle TFmode and TCmode for E500 double.
2327 * config/rs6000/rs6000.h (CANNOT_CHANGE_MODE_CLASS): Check for
2328 TFmode for E500 double.
2329 * config/rs6000/rs6000.md (FP): Allow TF for E500 double.
2330 (floatsidf2): Enable for E500 double.
2331 (movtf_softfloat): Use rs6000_nonimmediate_operand.
2332 (extenddftf2): Change to extenddftf2_fprs.
2333 (extenddftf2): Call gen_spe_extenddftf2 or gen_extenddftf2_fprs
2334 depending on TARGET_E500_DOUBLE.
2335 (extendsftf2): Enable for E500 double.
2336 (trunctfdf2): Enable for E500 double.
2337 (trunctfsf2): Change to trunctfsf2_fprs.
2338 (trunctfsf2): Call gen_spe_trunctfsf2 or gen_trunctfsf2_fprs
2339 depending on TARGET_E500_DOUBLE.
2340 (floatsitf2): Enable for E500 double.
2341 (fix_trunctfsi2): Change to fix_trunctfsi2_fprs.
2342 (fix_trunctfsi2): Call gen_spe_fix_trunctfsi2 or
2343 gen_fix_trunctfsi2_fprs depending on TARGET_E500_DOUBLE.
2344 (negtf2): Change to negtf2_internal.
2345 (negtf2): New expander.
2346 (abstf2): Enable for E500 double. Call gen_spe_abstf2_tst,
2347 gen_spe_abstf2_cmp or gen_abstf2_internal depending on
2348 TARGET_E500_DOUBLE and flag_unsafe_math_optimizations.
2349 (movdi_internal32): Use rs6000_nonimmediate_operand.
2350 (unnamed splitter): Likewise.
2351 * config/rs6000/spe.md (CMPTFEQ_GPR, TSTTFEQ_GPR, CMPTFGT_GPR,
2352 TSTTFGT_GPR, CMPTFLT_GPR, TSTTFLT_GPR): New unspecs.
2353 (SPE64TF, DITI): New mode macros.
2354 (frob_df_di): Change to frob_<SPE64:mode>_<DITI:mode>; allow more
2357 (frob_<mode>_di_2): New.
2358 (frob_tf_di_8_2): New.
2359 (frob_di_df): Change to frob_di_<mode>; allow more modes.
2361 (frob_di_df_2): Change to frob_<DITI:mode>_<SPE64:mode>_2; allow
2363 (frob_ti_<mode>_8_2): New.
2364 (frob_ti_tf_2): New.
2365 (mov_si<mode>_e500_subreg0, mov_si<mode>_e500_subreg0_2,
2366 mov_si<mode>_e500_subreg4, mov_si<mode>_e500_subreg4_2): Allow
2368 (mov_sitf_e500_subreg8, mov_sitf_e500_subreg8_2,
2369 mov_sitf_e500_subreg12, mov_sitf_e500_subreg12_2): New.
2370 (spe_trunctfdf2_internal1, spe_trunctfsf2, spe_extenddftf2,
2371 spe_fix_trunctfsi2, spe_fix_trunctfsi2_internal,
2372 spe_negtf2_internal, spe_abstf2_cmp, spe_abstf2_tst): New.
2373 (cmptfeq_gpr, tsttfeq_gpr, cmptfgt_gpr, tsttfgt_gpr, cmptflt_gpr,
2376 2007-01-23 Ian Lance Taylor <iant@google.com>
2378 * Makefile.in (OBJS-common): Reformat, alphabetize, but put
2380 (OBJS-archive): Reformat, alphabetize.
2381 (OBJS): Change out_object_file to OBJS-md.
2383 2007-01-23 Uros Bizjak <ubizjak@gmail.com>
2385 * config/i386/i386.md: Use REG_P, MEM_P, CONST_INT_P, LABEL_P,
2386 JUMP_P and CALL_P predicates where applicable.
2387 * config/i386/i386.c: Ditto.
2388 * config/i386/i386.md: Ditto.
2389 * config/i386/mmx.md: Ditto.
2390 * config/i386/predicates.md: Ditto.
2392 2007-01-22 Andreas Schwab <schwab@suse.de>
2394 * config/m68k/m68k.h: Fix comment.
2396 2007-01-22 Jan Hubicka <jh@suse.cz>
2398 * passes.c (init_optimization_passes): Do not rerun
2399 pass_early_warn_uninitialized.
2401 2007-01-22 Richard Guenther <rguenther@suse.de>
2403 PR tree-optimization/30038
2404 * tree-ssa-math-opts.c (maybe_record_sincos): New static helper
2406 (execute_cse_sincos_1): Likewise.
2407 (execute_cse_sincos): Likewise.
2408 (gate_cse_sincos): Likewise.
2409 (pass_cse_sincos): New pass CSEing sin() and cos() calls using
2410 the cexpi() canonicalization of sincos().
2411 * tree-pass.h (pass_cse_sincos): Declare.
2412 * passes.c (init_optimization_passes): New pass pas_cse_sincos.
2414 2007-01-21 Eric Botcazou <ebotcazou@libertysurf.fr>
2416 PR rtl-optimization/29329
2417 * combine.c (replaced_rhs_insn): Rename to i2mod.
2418 (replaced_rhs_value): Rename to i2mod_new_rhs.
2419 (i2mod_old_rhs): New global variable.
2420 (combine_instructions): Adjust for above change. Save a copy of
2421 the old RHS into i2mod_old_rhs when the contents of a REG_EQUAL
2422 note are substituted in the second instruction.
2423 (distribute_notes) <REG_DEAD>: Adjust for above change. Do not
2424 ditch the note if it pertains to the second eliminated register
2425 and this register is mentioned in i2mod_old_rhs.
2428 2006-09-12 Eric Botcazou <ebotcazou@libertysurf.fr>
2430 * combine.c (distribute_notes) <REG_DEAD>: Do not consider SETs past
2431 the insn to which the note was originally attached.
2433 2007-01-21 Jan Hubicka <jh@suse.cz>
2435 * ipa-inline.c (inlining_mode): Comment, move up.
2436 (cgraph_decide_inlining_incrementally): Do not perform inlining
2437 itself; fix handling of flattening of self recursive functions.
2438 (cgraph_find_cycles): Remove.
2439 (cgraph_flatten_node): Remove.
2440 (cgraph_decide_inlining): Use incremental inliner to handle flattening.
2441 (try_inline): New function.
2442 (cgraph_early_inlining): Update call of
2443 cgraph_decide_inlining_incrementally. Apply inlining here.
2444 (apply_inline): Update call of cgraph_decide_inlining_incrementally.
2446 2007-01-21 Dirk Mueller <dmueller@suse.de>
2449 * tree-vrp.c (check_array_bounds): do not warn
2450 about ADDR_EXPR's of ARRAY_REF's which are immediately
2451 used in binary expressions.
2453 2007-01-21 Ira Rosen <irar@il.ibm.com>
2455 * tree-vectorizer.h (struct _stmt_vec_info): Add new field
2456 read_write_dep and macros for its access.
2457 * tree-vectorizer.c (new_stmt_vec_info): Initialize the new field.
2458 * tree-vect-analyze.c (vect_analyze_data_ref_dependence): Remove
2459 argument, call vect_check_interleaving for every independent pair of
2460 data-refs. Mark loads that access the same memory location as a store
2462 (vect_check_dependences): Remove.
2463 (vect_analyze_data_ref_dependences): Remove vect_check_dependences
2464 call, fix the call to vect_analyze_data_ref_dependence.
2465 (vect_analyze_data_ref_access): For statements that access the same
2466 data-ref, check that they are not stores; for loads, check that there
2467 is no store that access the same location.
2469 2007-01-20 Roger Sayle <roger@eyesopen.com>
2470 Joseph Myers <joseph@codesourcery.com>
2472 * doc/invoke.texi (-fdump-rtl-combine): Fix under/overfull hbox.
2473 (-fdump-rtl-stack): Likewise.
2474 (-fno-signed-zeros): Use @minus{} for a minus sign. Correct typo.
2475 (-mcheck-zero-division, -mcpu): Fix under/overfull hbox.
2476 (-mpt-fixed): Use @minus{} for minus sign.
2477 (Using Precompiled Headers): Fix under/overfull hbox.
2479 2007-01-20 Jan Hubicka <jh@suse.cz>
2481 * tree-flow.h (struct stmt_ann_d): Move references_memory to proper
2482 place within annotation.
2484 2007-01-20 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
2486 * pa.c (output_move_double): Change array size of xoperands to 4.
2488 2007-01-20 Jan Hubicka <jh@suse.cz>
2490 * tree-tailcall.c (adjust_return_value): Do not use RESULT_DECL
2493 2007-01-19 Ian Lance Taylor <iant@google.com>
2495 * expmed.c (expand_divmod) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]: Cast
2496 constant to unsigned HOST_WIDE_INT before negating.
2498 2007-01-19 Ian Lance Taylor <iant@google.com>
2500 * tree-ssa-operands.h (struct vuse_vec_d): Change num_vuse field
2502 (VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Use unsigned comparison.
2503 (VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
2504 (SET_VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Likewise.
2505 (SET_VUSE_ELEMENT_VAR) [ENABLE_CHECKING]: Likewise.
2506 (SET_VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
2507 (realloc_vdef, realloc_vuse): Change second parameter to
2509 (ssa_operand_iterator_d): Change vuse_index and mayuse_index
2511 * tree-ssa-operands.c (realloc_vop): Change num_elem parameter to
2512 unsigned. Change x and lim locals to unsigned.
2513 (realloc_vdef, realloc_vuse): Change num_elem parameter to
2515 (finalize_ssa_vuse_ops): Change old_i local to unsigned.
2516 (copy_virtual_operands): Change i and n locals to unsigned.
2518 2007-01-19 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
2521 * builtins.c (fold_builtin_1): Handle builtin fdim.
2523 2007-01-20 Jan Hubicka <jh@suse.cz>
2525 * tree-ssa.c (init_tree_ssa): Do not call init_alias_heapvars.
2526 * tree-ssa-structalias.c (compute_points_to_sets): Do call
2527 init_alias_heapvars.
2528 (init_alias_heapvars): Initialize only when not already initialized.
2529 (delete_alias_heapvars): Set heapvar_for_stmt to NULL.
2531 2007-01-19 Roger Sayle <roger@eyesopen.com>
2533 * common.opt (fsigned-zeros): New command line option.
2534 * flags.h (HONOR_SIGNED_ZEROS): Control via flag_signed_zeros instead
2535 of flag_unsafe_math_optimizations.
2536 * opts.c (set_fast_math_flags): The -ffast-math command line option
2537 implies -fno-signed-zeros.
2538 (fast_math_flags_set_p): Likewise.
2540 * doc/invoke.texi: Document new -fno-signed-zeros option, and update
2541 the documentation of -ffast-math appropriately. Wrap long lines.
2543 2007-01-19 Steve Ellcey <sje@cup.hp.com>
2545 * system.h (ASM_MAKE_LABEL_LINKONCE): Poison.
2546 * varasm.c (globalize_decl): Remove ASM_MAKE_LABEL_LINKONCE ifdef.
2548 2007-01-19 Tomas Bily <tbily@suse.cz>
2550 * cgraphunit.c (cgraph_finalize_function): Updating of pid
2552 (tree_init_ic_make_global_vars): New function
2553 (tree_init_edge_profiler): call of tree_init_ic_make_global_vars
2554 (tree_gen_ic_profiler): New function
2555 (tree_gen_ic_func_profiler): New function
2556 (tree_profiling): Added calling of tree_gen_ic_func_profiler
2557 (tree_profile_hooks): Added hook for indirec/virtual calls
2558 * value-prof.c (tree_find_values_to_profile): New case for
2560 (tree_values_to_profile): Call for determining indirect/virtual
2562 (tree_indirect_call_to_profile): New function
2563 (tree_ic_transform): New function
2564 (tree_ic): New function
2565 (find_func_by_pid): New function
2566 (init_pid_map): New function
2567 (tree_value_profile_transformations): Added check for
2568 indirect/virtual call transformation
2569 * value-prof.h (enum hist_type): New counter type for
2570 indirect/virtual calls
2571 (profile_hooks): Added new hook for profiling indirect/virtual
2573 * profile.c (instrument_values): New case for indirect/virtual
2575 * gcov-io.h (GCOV_LAST_VALUE_COUNTER): Changed to 6
2576 (GCOV_COUNTER_V_INDIR): New counter type
2577 (GCOV_COUNTER_NAMES): New name of counter "indirect" added
2578 (GCOV_MERGE_FUNCTIONS): New merge function for indirect/virtual
2580 * cgraph.c: Definition of cgraph_max_pid
2581 (cgraph_create_node): Default init of pid attribute
2582 * cgraph.h: Declaration of cgraph_max_pid
2583 (struct cgraph_node): Added pid attribute
2584 * libgcov.c (__gcov_indirect_call_profiler): New function
2585 (__gcov_one_value_profiler_body): New function
2586 (__gcov_one_value_profiler): Body was moved to
2587 __gcov_one_value_profiler_body and calls it
2589 2007-01-19 Basile Starynkevitch <basile@starynkevitch.net>
2591 * doc/gty.texi (Options): Document the mark_hook option to GTY.
2592 * gengtype.c (write_types_data, write_func_for_structure,
2593 write_types, ggc_wtd, pch_wtd): Add skip_hooks to
2594 write_types_data, ggc_wtd, pch_wtd for processing mark_hook.
2595 (walk_type, write_func_for_structure): Generate the mark_hook if
2598 2007-01-19 Jan Hubicka <jh@suse.cz>
2600 * ipa-inline.c (cgraph_decide_inlining_incrementally): Instead of
2601 'early' argument take inlining mode argument specifying whether to
2602 inline for size/speeed or all functions; add support for flattening;
2604 (cgraph_early_inlining): Update call of decide_inlining_incrementally.
2606 2007-01-19 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
2609 * toplev.c (warn_deprecated_use): Use %qD instead of %qs to print
2610 the name of the declared identifier.
2612 2007-01-19 Dirk Mueller <dmueller@suse.de>
2614 * config/i386.h (CONDITIONAL_REGISTER_USAGE): Store
2615 result of PIC_OFFSET_TABLE_REGNUM in temporary variable to avoid
2616 duplicate evaluation.
2618 2007-01-19 Uros Bizjak <ubizjak@gmail.com>
2620 * config/i386/i386.md (acos<mode>2): Rename from acossf2 and acosdf2.
2621 Macroize expander using X87MODEF12 mode macro. Extend operand 1
2622 to XFMode, use acosxf2 and truncate result to requested mode.
2623 Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
2624 (asin<mode>2): Similarly, with asin expanders.
2625 (*fscalexf4_i387): Rename from *fscalexf4.
2626 (expNcorexf3): New expander.
2627 (expxf2, exp10xf2, exp2xf2): Use expNcorexf3 expander.
2628 (exp<mode>2): Rename from expsf2 and expdf2. Macroize expander using
2629 X87MODEF12 mode macro. Extend operand 1 to XFMode, use expxf2 and
2630 truncate result to requested mode. Use SSE_FLOAT_MODE_P to disable
2631 patterns for SSE math.
2632 (exp10<mode>2): Similarly, with exp10 expanders.
2633 (exp2<mode>2): Similarly, with exp2 expanders.
2634 (expm1<mode>2): Similarly, with expm1 expanders.
2635 (ldexp<mode>3): Similarly, with ldexp expanders.
2636 (log<mode>2, log10<mode>2, log2<mode>2, log1p<mode>2, logb<mode>2):
2637 Use gen_truncxf<mode>2_i387_noop to truncate result.
2639 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2641 * config/m68k/m68k.h (PREFERRED_STACK_BOUNDARY): Define to 32
2642 for ColdFire targets.
2644 2007-01-19 Nathan Sidwell <nathan@codesourcery.com>
2645 Richard Sandiford <richard@codesourcery.com>
2647 * config/m68k/m68k.h (M68K_STATIC_CHAIN_REG_NAME): New macro.
2648 (INITIALIZE_TRAMPOLINE): Use STATIC_CHAIN_REGNUM.
2649 (__transfer_from_trampoline): Use M68K_STATIC_CHAIN_REG_NAME.
2650 * config/m68k/m68kelf.h (STATIC_CHAIN_REGNUM): Override.
2651 (M68K_STATIC_CHAIN_REG_NAME): Likewise.
2652 * config/m68k/netbsd-elf.h (M68K_STATIC_CHAIN_REG_NAME): Likewise.
2654 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2656 * config/m68k/m68k.md (adddi_dilshr32): Rename to...
2657 (*adddi_dilshr32): ...this. Fix formatting. Remove commented-out
2658 non-canonical pattern. Restrict to !TARGET_COLDFIRE.
2659 (*adddi_dilshr32_cf): New pattern.
2660 (adddi3, subdi3): Remove first alternatives.
2662 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2664 * config/m68k/m68k.c (notice_update_cc): If an SFmode move is
2665 implemented using move.l, do not use its cc result for floating-point
2668 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2670 * config/m68k/m68k.h (EXTRA_CONSTRAINT): Stop the 'T' constraint
2671 from accepting 's' constraints if flag_pic.
2673 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2675 * config/m68k/m68k.md (bordered, bunordered, buneq, bunge, bungt)
2676 (bunle, bunlt, bltgt, bordered_rev, bunordered_rev, buneq_rev)
2677 (bunge_rev, bungt_rev, bunle_rev, bunlt_rev, bltgt_rev): Change
2678 condition from TARGET_68881 to TARGET_HARD_FLOAT.
2680 2007-01-19 Sandra Loosemore <sandra@codesourcery.com>
2682 * longlong.h (count_leading_zeros, COUNT_LEADING_ZEROS_0): Add
2683 ColdFire alternatives.
2684 * config/m68k/m68k.h (CLZ_DEFINED_VALUE_AT_ZERO): New macro.
2685 * config/m68k/m68k.md (clzsi2): Define for ColdFire
2686 architectures that support ff1 instruction.
2688 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2689 Julian Brown <julian@codesourcery.com>
2691 * config/m68k/m68k.h (CONST_OK_FOR_LETTER_P): Add an 'R' case.
2692 * config/m68k/m68k.md (*movsi_cfv4): Fold into...
2693 (*movsi_cf): ...here. Remove unnecessary 'R' from 'Rg'.
2696 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2698 * config/m68k/m68k-protos.h (valid_mov3q_const): Take a HOST_WIDE_INT
2700 (output_move_const_into_data_reg, output_move_simode_const): Delete.
2701 * config/m68k/m68k.c (const_method, const_int_cost): Take a
2702 HOST_WIDE_INT instead of an rtx.
2703 (m68k_rtx_costs): Update call accordingly.
2704 (output_move_const_into_data_reg): Likewise. Fix formatting.
2705 (valid_mov3q_const): Take a HOST_WIDE_INT instead of an rtx.
2707 (output_move_simode_const): Update calls after above changes.
2708 Rework to use automatic variables and predicates like MEM_P.
2709 * config/m68k/m68k.md (pushexthisi_const): Update call to
2712 2007-01-19 Dirk Mueller <dmueller@suse.de>
2714 * tree-ssa-alias.c (perform_var_substitution): Fix typo
2717 2007-01-19 Richard Guenther <rguenther@suse.de>
2719 * builtins.c (expand_builtin_cexpi): Fall back to expanding
2720 via cexp in case sincos is not available.
2722 2007-01-19 Richard Guenther <rguenther@suse.de>
2724 * doc/tm.texi (TARGET_HAS_SINCOS): Document new target macro.
2725 * defaults.h (TARGET_HAS_SINCOS): Default to off.
2726 * config/linux.h (TARGET_HAS_SINCOS): Set to on if we have glibc.
2727 * config/alpha/linux.h (TARGET_HAS_SINCOS): Likewise.
2728 * config/sparc/linux.h (TARGET_HAS_SINCOS): Likewise.
2729 * config/sparc/linux64.h (TARGET_HAS_SINCOS): Likewise.
2730 * config/rs6000/linux.h (TARGET_HAS_SINCOS): Likewise.
2731 * config/rs6000/linux64.h (TARGET_HAS_SINCOS): Likewise.
2733 2007-01-19 Uros Bizjak <ubizjak@gmail.com>
2735 * config/i386/i386.md (*fpatanxf3_i387, fpatan_extend<mode>xf3_i387):
2737 (atan2sf3_1, atan2df3_1, atan2xf3_1): Remove insn patterns.
2738 (atan2xf3): Directly generate RTL pattern.
2739 (atan2<mode>3): Rename from atan2sf3 and atan2df3 and macroize insn
2740 patterns using X87MODEF12 mode macro. Use fpatan_extend<mode>xf3_i387
2741 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
2742 disable patterns for SSE math.
2743 (atan<mode>2): Rename from atansf2 and atandf2 and macroize insn
2744 patterns using X87MODEF12 mode macro. Use fpatan_extend<mode>xf3_i387
2745 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
2746 disable patterns for SSE math.
2748 2007-01-19 Alexandre Oliva <aoliva@redhat.com>
2750 * libgcc-std.ver: Fix typo in %inherit for GCC_4.3.0.
2752 2007-01-18 Roger Sayle <roger@eyesopen.com>
2754 * fold-const.c (fold_unary) <VIEW_CONVERT_EXPR>: Optimize away a
2755 VIEW_CONVERT_EXPR to the same type as it's operand.
2757 2007-01-18 David Edelsohn <edelsohn@gnu.org>
2759 * config/rs6000/darwin-ldouble.c: Only build _SOFT_FLOAT if
2760 configured for long double 128.
2762 2007-01-18 Mike Stump <mrs@apple.com>
2764 * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Fix build
2767 2007-01-18 Michael Meissner <michael.meissner@amd.com>
2769 * i386.c (ix86_compute_frame_layout): Make fprintf's in #if 0 code
2772 2007-01-18 Jan Hubicka <jh@suse.cz>
2774 * tree-ssa-operands.c (vop_free_bucket_size): Never return value
2775 greater than NUM_VOP_FREE_BUCKETS.
2777 2007-01-18 Daniel Berlin <dberlin@dberlin.org>
2779 * tree-ssa-structalias.c: Update comments.
2780 (ptabitmap_obstack): Removed.
2782 (oldpta_obstack): Ditto.
2783 (stats): Add a few members.
2784 (struct variable_info): Remove node, complex, address_taken, and
2785 indirect_target members. Add oldsolution member.
2786 (new_var_info): Do not initialize removed members.
2787 (constraint_expr_type): Remove INCLUDES.
2788 (constraint_graph): Add size, implicit_preds, rep,
2789 indirect_cycles, eq_rep, label, direct_nodes, and complex members.
2790 (FIRST_REF_NODE): New macro.
2791 (LAST_REF_NODE): Ditto.
2792 (FIRST_ADDR_NODE): Ditto.
2793 (find): New function.
2795 (dump_constraint): Do not handle INCLUDES.
2796 (insert_into_complex): Do not insert duplicate constraints.
2797 (condense_varmap_nodes): Renamed and rewritten into ...
2798 (merge_node_constraints): This. Also fix bug in handling of
2799 offseted copy constraints.
2800 (clear_edges_for_node): No longer need to deal with preds at all,
2801 or removing associated preds/succs.
2802 (merge_graph_nodes): Deal with indirect_cycles.
2803 Don't deal with predecessors.
2804 (add_implicit_graph_edge): New function.
2805 (add_pred_graph_edge): Ditto.
2806 (add_graph_edge): Don't deal with predecessors.
2807 (build_constraint_graph): Removed.
2808 (build_pred_graph): New function.
2809 (build_succ_graph): Ditto.
2810 (struct scc_info): Removed in_component. Added roots, dfs, and
2811 node_mapping. Remove visited_index, unification_queue.
2812 (scc_visit): Deal with union-find we do now.
2813 Deal with cycles with REF nodes.
2814 (collapse_nodes): Renamed and rewritten to ...
2815 (unify_nodes): This.
2816 (process_unification_queue): Removed.
2817 (topo_visit): Cleanup
2818 (do_da_constraint): Use find.
2819 (do_sd_constraint): Ditto.
2820 (do_ds_constraint): Ditto.
2821 (do_complex_constraint): Ditto.
2822 (init_scc_info): Update for removed and added members.
2823 (find_and_collapse_graph_cycles): Renamed and rewritten into ...
2824 (find_indirect_cycles): This.
2825 (equivalence_class): New variable.
2826 (label_visit): New function.
2827 (perform_variable_substitution): Rewritten.
2828 (free_var_substitution_info): New function.
2829 (find_equivalent_node): Ditto.
2830 (move_complex_constraints): Ditto.
2831 (eliminate_indirect_cycles): Ditto.
2832 (solve_graph): Only propagate changed bits.
2833 Use indirect cycle elimination.
2835 (tree_id_t): Rename to tree_vi_t, delete id member, add vi member.
2836 (tree_id_eq): Renamed to ...
2837 (tree_vi_eq): This. Update for member change
2838 (insert_id_for_tree): Renamed and rewritten to ...
2839 (insert_vi_for_tree): This.
2840 (lookup_id_for_tree): Renamed and rewritten to ...
2841 (lookup_vi_for_tree): This.
2842 (get_id_for_tree): Renamed and rewritten to ...
2843 (get_vi_for_tree): Ditto.
2844 (get_constraint_exp_from_ssa_var): Update to use get_vi_for_tree.
2845 (process_constraint): Don't handle INCLUDES.
2846 Remove special ADDRESSOF case.
2847 (find_func_aliases): Rewrite to use vi functions instead of id
2849 (create_function_info_for): Ditto.
2850 (create_variable_info_for): Ditto.
2851 (intra_create_variable_infos): Ditto.
2852 (merge_smts_into): Ditto.
2853 (find_what_p_points_to): Ditto.
2854 (init_base_vars): Ditto.
2855 (init_alias_vars): Ditto.
2856 (remove_preds_and_fake_succs): New function.
2857 (dump_sa_points_to_info): Dump new stats.
2858 (dump_solution_for_var): Use find.
2859 (set_used_smts): Fix formatting.
2860 (compute_points_to_sets): Updated for new functions.
2861 (ipa_pta_execute): Ditto.
2863 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
2864 Richard Sandiford <richard@codesourcery.com>
2866 * doc/tm.texi (TARGET_FUNCTION_VALUE): Expand documentation of
2868 * calls.c (expand_call): If the return value is a PARALLEL,
2869 extract its first member.
2870 * config/m68k/linux.h (FUNCTION_EXTRA_EPILOGUE): Remove.
2871 * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
2872 use FUNCTION_EXTRA_EPILOGUE.
2873 (m68k_function_value): Return a PARALLEL if the return value
2874 is of a pointer type.
2875 * config/m68k/netbsd-elf.h (current_function_returns_pointer)
2876 (FUNCTION_EXTRA_EPILOGUE): Remove.
2877 * config/m68k/m68k.md (D0_REG): New constant.
2879 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
2881 * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
2882 output a NOP for empty epilogues.
2884 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2886 * config/m68k/m68k.c (m68k_use_return_insn): Update comments
2887 before function. Extend register save check to include all
2888 registers, not just integer ones.
2890 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
2892 * config/m68k/m68k-protos.h (use_return_insn): Rename to...
2893 (m68k_use_return_insn): ...this.
2894 * config/m68k/m68k.h (USE_RETURN_INSN): Delete.
2895 * config/m68k/m68k.c (use_return_insn): Rename to...
2896 (m68k_use_return_insn): ...this.
2897 * config/m68k/m68k.md (return): Use m68k_use_return_insn instead
2900 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
2902 * config/m68k/fpgnulib.c (__truncdfsf2): Implement round to
2903 nearest even, fix denormal rounding overflow.
2905 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2907 * config/m68k/m68k.md (movsf_cf_hard): Use fsmove instead of
2908 f%$smove and f%$move.
2909 (movdf_cf_hard): Use fdmove for cases 0 and 3 and fmove for case 1.
2910 (extendsfdf2_cf): Use fdmove instead of f%&move.
2911 (truncdfsf2_cf): Use fsmove instead of f%$smove.
2912 (add<mode>3_cf, sub<mode>3_cf): Use <FP:prec> instead of <FP:round>.
2914 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2916 * config/m68k/m68k.md (movdf_cf_hard): Use output_move_double for
2919 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2921 * real.h (real_format): Add a canonical_nan_lsbs_set field.
2922 (coldfire_single_format): Declare.
2923 (coldfire_double_format): Likewise.
2924 * real.c (encode_ieee_single): Use canonical_nan_lsbs_set instead
2925 of qnan_msb_set to determine the lower bits of a canonical
2927 (encode_ieee_double): Likewise.
2928 (encode_ieee_quad): Likewise.
2929 (ieee_single_format): Initialize canonical_nan_lsbs_set.
2930 (mips_single_format): Likewise.
2931 (ieee_double_format): Likewise.
2932 (mips_double_format): Likewise.
2933 (ieee_extended_motorola_format): Likewise.
2934 (ieee_extended_intel_96_format): Likewise.
2935 (ieee_extended_intel_128_format): Likewise.
2936 (ieee_extended_intel_96_round_53_format): Likewise.
2937 (ibm_extended_format): Likewise.
2938 (mips_extended_format): Likewise.
2939 (ieee_quad_format): Likewise.
2940 (mips_quad_format): Likewise.
2941 (vax_f_format): Likewise.
2942 (vax_d_format): Likewise.
2943 (vax_g_format): Likewise.
2944 (i370_single_format): Likewise.
2945 (i370_double_format): Likewise.
2946 (decimal_single_format): Likewise.
2947 (decimal_double_format): Likewise.
2948 (decimal_quad_format): Likewise.
2949 (c4x_single_format): Likewise.
2950 (c4x_extended_format): Likewise.
2951 (real_internal_format): Likewise.
2952 (coldfire_single_format): New real_format.
2953 (coldfire_double_format): Likewise.
2954 * config/pdp11/pdp11.c (pdp11_f_format): Initialize
2955 canonical_nan_lsbs_set.
2956 (pdp11_d_format): Likewise.
2957 * config/m68k/m68k.c (override_options): Override REAL_FORMAT_MODE
2958 if TARGET_COLDFIRE_CPU.
2960 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2962 * config/m68k/m68k-protos.h (m68k_output_pic_call): Delete.
2963 (output_call, m68k_legitimize_call_address): Declare.
2964 * config/m68k/m68k.h (EXTRA_CONSTRAINT): Remove unnecessary
2965 parenthesees. Add support for a 'W' constraint.
2966 (LEGITIMATE_PIC_OPERAND_P): Remove SYMBOL_REF_FLAG handling.
2967 (PRINT_OPERAND_PUNCT_VALID_P): Remove comment about 'o'.
2968 (m68k_symbolic_call, m68k_symbolic_jump): Declare.
2969 * config/m68k/m68k.c (m68k_symbolic_call, m68k_symbolic_jump): New
2971 (override_options): Initialize them. Do not set flag_no_function_cse
2972 for TARGET_ID_SHARED_LIBRARY.
2973 (m68k_output_pic_call): Delete.
2974 (m68k_legitimize_call_address): New function.
2975 (print_operand): Remove the %o prefix. Handle the %p prefix.
2976 (output_call): New function.
2977 (m68k_output_mi_thunk): Use m68k_symbolic_jump. Always load the
2978 target address from the GOT if symbolic jumps are not allowed.
2979 * config/m68k/m68k.md (call, general_operand): Do not set
2980 SYMBOL_REF_FLAG. Use m68k_legitimize_call_address instead.
2981 Merge separate flag_pic and !flag_pic define_insns into...
2982 (*call, *call_value): ...these new patterns. Match the address
2983 rather than the containing MEM and require it to be a call_operand.
2984 Use output_call to generate the asm template.
2985 * config/m68k/predicates.md (const_call_operand): New predicate.
2986 (call_operand): Likewise.
2988 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
2990 * config/m68k/m68k.h (REGISTER_MOVE_COST): Simplify definition.
2991 (STACK_GROWS_DOWNWARD): Define to 1.
2992 (FUNCTION_VALUE, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P): Equivocate
2993 comments, emphasizing that these values are only defaults.
2994 * config/m68k/linux.h (LINK_SPEC): Fix formatting in #undef.
2995 * config/m68k/m68k.c (const_method): Remove trailing whitespace.
2997 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2999 * config/m68k/m68k.md (cmpsi): Remove outdated flag_pic handling.
3001 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
3002 Richard Sandiford <richard@codesourcery.com>
3004 * config/m68k/m68k.h (DATA_REGNO_P, ADDRESS_REGNO_P, INT_REGNO_P)
3005 (FP_REGNO_P): New macros.
3006 (REGNO_OK_FOR_INDEX_P, REGNO_OK_FOR_BASE_P, REGNO_OK_FOR_DATA_P)
3007 (REGNO_OK_FOR_FP_P, REG_OK_FOR_INDEX_P, REG_OK_FOR_BASE_P): Use them.
3008 (EH_RETURN_STACKADJ_RTX): Use A0_REG.
3009 * config/m68k/m68k.c (m68k_regno_mode_ok): Use the new REGNO macros.
3011 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
3013 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
3014 (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-rtems*): Add t-floatlib
3016 * config/m68k/t-floatlib: New file, extracting common code from...
3017 * config/m68k/t-m68kbare, config/m68k/t-m68kelf,
3018 * config/m68k/t-uclinux: Here.
3019 * config/m68k/fpgnulib.c: Do not compile extendeed precision
3020 routines on ColdFire targets.
3022 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
3024 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
3025 (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*)
3026 (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-linux*)
3027 (m68k-*-rtems*): Use tm_file rather than m68k/m68k.h and
3028 explicitly set MOTOROLA to 1.
3029 * config/m68k/m68k.h (MOTOROLA): Simplify definition accordingly.
3031 2007-01-18 Richard Sandiford <richard@codesourcery.com>
3032 Nathan Sidwell <nathan@codesourcery.com>
3034 * config/m68k/m68k.h (PCC_STATIC_STRUCT_RETURN): Delete.
3035 (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Add MOTOROLA cases.
3036 * config/m68k/coff.h (REGISTER_PREFIX_MD): Delete.
3037 * config/m68k/m68020-elf.h (LIB_SPEC): Delete.
3038 * config/m68k/m68k-none.h (CC1_SPEC, CPP_SUBTARGET_SPEC): Delete.
3039 * config/m68k/m68kelf.h (IMMEDIATE_PREFIX, REGISTER_PREFIX_MD)
3040 (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Delete.
3041 (NO_DOLLAR_IN_LABEL, PCC_STATIC_STRUCT_RETURN): Don't undefine.
3042 (BSS_ASM_OP, ASM_OUTPUT_SKIP, ASM_OUTPUT_ASCII): Delete.
3043 * config/m68k/m68kemb.h (PCC_STATIC_STRUCT_RETURN): Don't undefine.
3044 * config/m68k/linux.h (SIZE_TYPE, PTRDIFF_TYPE, WCHAR_TYPE)
3045 (WCHAR_TYPE_SIZE, TARGET_OBJFMT_CPP_BUILTINS): Delete.
3046 (TARGET_OS_CPP_BUILTINS): Don't define mc68000 and mc68020 here.
3047 (DBX_REGISTER_NUMBER): Delete.
3048 * config/m68k/netbsd-elf.h (IMMEDIATE_PREFIX): Delete.
3049 (PCC_STATIC_STRUCT_RETURN): Don't undefine.
3050 * config/m68k/openbsd.h (PCC_STATIC_STRUCT_RETURN): Define.
3052 2007-01-18 Richard Sandiford <richard@codesourcery.com>
3054 * config.gcc (m68k-*-uclinux*): Add flat.h to $tm_file.
3055 * config/flat.h: New file.
3056 * crtstuff.c (USE_PT_GNU_EH_FRAME): Don't define if
3058 * config/m68k/m68k.h (ASM_PREFERRED_EH_DATA_FORMAT): Do not use
3059 indirect references for -msep-data or -mid-shared-library.
3060 Do not use PC-relative code addresses either.
3061 * config/m68k/m68k.c (override_options): Restrict -fPIC error
3063 * config/m68k/uclinux.h (STARTFILE_SPEC): Define. Use Scrt1.o
3064 for shared libraries and crt1.o for executables. Use crti.o and
3066 (ENDFILE_SPEC): Use crtend.o and crtn.o.
3067 (LIB_SPEC): Suppress -Rlibc.gdb if -static-libc is given.
3068 Do not add -elf2flt or -shared-lib-id options here.
3069 (LINK_SPEC): Define. Pass -elf2flt if no -elf2flt option is given.
3070 Pass -shared-lib-id if -mid-shared-library, taking the library
3071 identifier from -mshared-library-id if given, otherwise
3073 (EH_FRAME_IN_DATA_SECTION): Do not undefine.
3074 (INIT_SECTION_ASM_OP, FINI_SECTION_ASM_OP): Likewise.
3075 (TARGET_OS_CPP_BUILTINS): Define __GXX_MERGED_TYPEINFO_NAMES=0
3076 and __GXX_TYPEINFO_EQUALITY_INLINE=0 if -mid-shared-library.
3077 (DRIVER_SELF_SPECS): Map unadorned PIC options to -msep-data.
3078 * config/m68k/t-uclinux (EXTRA_MULTILIB_PARTS): Add crtbegin.o
3080 * config/m68k/lb1sf68.asm (PICCALL): Use an lea and pc-relative
3081 jump sequence for ISA A and ISA A+.
3082 (PICJUMP): Likewise.
3084 2007-01-18 Richard Sandiford <richard@codesourcery.com>
3086 * config.gcc (m68k-*-uclinux*): Base the port on the common
3087 and m68k GNU/Linux files rather than on the generic ELF ones.
3088 * config/m68k/uclinux.h (TARGET_VERSION): Override.
3089 (TARGET_OS_CPP_BUILTINS): Use LINUX_TARGET_OS_CPP_BUILTINS.
3091 2007-01-18 Julian Brown <julian@codesourcery.com>
3092 Richard Sandiford <richard@codesourcery.com>
3094 * config/m68k/m68k.h (LONG_DOUBLE_TYPE_SIZE): Make 64-bit on ColdFire.
3095 (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
3096 * config/m68k/netbsd-elf.h (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Undefine
3098 * config/m68k/uclinux-oldabi.h (LONG_DOUBLE_TYPE_SIZE): Redefine to
3100 (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
3102 2007-01-18 Richard Sandiford <richard@codesourcery.com>
3104 * doc/install.texi: Document m68k-uclinuxoldabi.
3105 * config.gcc (m68k-*-uclinuxoldabi): New configuration.
3106 * config/m68k/uclinux-oldabi.h: New file, copied from
3107 config/m68k/uclinux.h.
3109 2007-01-18 Kaz Kojima <kkojima@rr.iij4u.or.jp>
3111 * config/m32r/m32r-protos.h (m32r_expand_epilogue): Declare it.
3112 * config/m32r/m32r.c (m32r_setup_incoming_varargs): Use gen_frame_mem.
3113 (m32r_compute_frame_size): Use unsigned for regno.
3114 (m32r_reload_lr): Use gen_frame_mem.
3116 (m32r_output_function_epilogue): Don't output the function epilogue
3118 (m32r_expand_epilogue): New.
3119 (direct_return): Return false if the function has the interrupt
3121 (m32r_hard_regno_rename_ok): Remove code for the textual epilogue.
3122 * config/m32r/m32r.md (epilogue): New expander.
3123 (return_lr, return_rte): New insns.
3124 (return): Make it expander.
3125 (return_normal): New expander.
3127 2007-01-18 Josh Conner <jconner@apple.com>
3130 * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Add
3131 support for UNLE, UNLT, UNGE, and UNGT.
3133 2007-01-18 Jan Hubicka <jh@suse.cz>
3135 * tree-vrp.c (finalize_jump_threads): Do not call cleanup_cfg by hand.
3136 * tree-tailcall (add_virtual_phis): Likewise.
3137 (optimize_tail_call): Return TODOs.
3138 (execute_tail_calls): Return TODOs.
3139 * tree-ssa-ccp (execute_fold_all_builtins): Do cleanup_cfg via TODO.
3140 * tree-cfgcleanup.c (cleanup_tree_cfg_loop): Return if something
3142 * tree-ssa-forwprop.c (tree_ssa_forward_propagate_single_use_value):
3143 Cleanup cfg using TODO.
3144 * tree-flow.h (cleanup_tree_cfg_loop): Update prototype.
3145 * passes.c (execute_function_todo): When cleanup did something, remove
3147 * tree-cfg.c (pass_build_cfg): Add cleanup_cfg TODO.
3148 (make_edges): Don't cleanup_cfg.
3150 2007-01-18 Uros Bizjak <ubizjak@gmail.com>
3152 * reg-stack.c (subst_stack_regs_pat) [UNSPEC_SINCOS_COS,
3153 UNSPEC_XTRACT_FRACT]: Use generic code for instructions that
3154 operate on the top of stack.
3155 [UNSPEC_SINCOS_SIN, UNSPEC_XTRACT_EXP, UNSPEC_TAN]: Rewrite
3156 register handling of instructions that output to the second
3158 [UNSPEC_TAN_ONE, UNSPEC_TAN_TAN]: Remove.
3159 (move_for_stack_reg): Special-case check for dead destination
3160 stack slot for constant load of 1.0 inside UNSPEC_TAN.
3162 * config/i386/i386.md (UNSPEC_TAN): New constant.
3163 (UNSPEC_TAN_ONE, UNSPEC_TAN_TAN): Remove.
3164 (fptanxf4_i387, fptan_extend<mode>xf4_i387): New patterns
3165 to correctly model move of constant 1.0 to top stack slot.
3166 (*tandf3_1, *tansf3_1, *tanxf3_1): Remove insn patterns.
3167 (unnamed peephole2 pattern): Remove corresponding peephole2
3168 pattern that optimizes tan insn and loading of constant 1.0.
3169 (tanxf2): Use fptanxf4_i387.
3170 (tan<mode>2): Rename from tansf2 and tandf2 and macroize insn
3171 patterns using X87MODEF12 mode macro. Use fptan_extend<mode>xf4_i387
3172 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
3173 disable patterns for SSE math.
3174 (sincos<mode>3): Use truncxf<mode>2_i387_noop for truncation.
3175 (fyl2x_extend<mode>xf3_i387): Use X87MODEF12 for operand 1.
3177 2007-01-18 Dirk Mueller <dmueller@suse.de>
3178 Richard Guenther <rguenther@suse.de>
3181 * doc/invoke.texi (Warray-bounds): Document -Warray-bounds.
3182 * common.opt (Warray-bounds): Add new warning option.
3183 * c-opts.c (c_common_handle_option): Define -Warray-bounds
3185 * Makefile.in: make tree-vrp.o depend on toplev.h
3186 * tree-vrp.c (vrp_finalize): Call check_array_refs if -Warray-bounds
3188 (check_array_refs, check_array_bounds, check_array_ref): New.
3190 2007-01-18 Jan Hubicka <jh@suse.cz>
3192 * tree-ssa-ccp.c (ccp_finalize): Return if something changed.
3193 (execute_ssa_ccp): Return flags conditionally.
3194 * tree-ssa-propagate.c (substitue_and_fold): Return if something was
3196 * tree-ssa-propagate.h (substitute_and_fold): Update prototype.
3198 2007-01-18 Steven Bosscher <steven@gcc.gnu.org>
3200 * cfgcleanup.c (cleanup_cfg): Detect cfglayout mode and set
3201 the CLEANUP_CFGLAYOUT flag when in cfglayout mode.
3203 * Makefile.c (GTFILES): Add cfglayout.h.
3204 * gengtype.c (open_base_files): Likewise.
3205 * cfglayout.c (cfg_layout_function_footer,
3206 cfg_layout_function_header) Reindent to make gengtype happy.
3207 * cfglayout.h (cfg_layout_function_footer,
3208 cfg_layout_function_header): Add GTY(()) marker.
3210 * ifcvt.c (noce_try_sign_mask): Make sure INSN_B is non-null.
3212 2007-01-18 Ben Elliston <bje@au.ibm.com>
3214 * genautomata.c (write_automata): Include xstrerror output in the
3215 error message if writing the DFA description file fails.
3217 2007-01-17 H.J. Lu <hongjiu.lu@intel.com>
3219 * config/mips/mips-protos.h (mips_output_external): Make it
3221 * config/mips/iris.h (TARGET_ASM_EXTERNAL_LIBCALL): Removed.
3222 * config/mips/mips.c (irix_output_external_libcall): Likewise.
3223 (extern_list): Likewise.
3224 (extern_head): Likewise.
3225 (TARGET_ASM_FILE_END): Likewise.
3226 (mips_file_end): Likewise.
3227 (mips_output_external): Rewritten.
3229 2007-01-18 Ben Elliston <bje@au.ibm.com>
3231 * genpreds.c (write_insn_preds_c): Only write out the function
3232 body for regclass_for_constraint if we have register constraints.
3234 2007-01-17 Tom Tromey <tromey@redhat.com>
3236 * doc/sourcebuild.texi (libgcj Tests): Use sourceware.org.
3237 * doc/install.texi (Testing): Use sourceware.org.
3238 (Binaries): Likewise.
3239 (Specific): Likewise.
3240 * doc/contrib.texi (Contributors): Use sourceware.org.
3242 2007-01-17 Anatoly Sokolov <aesok@post.ru>
3244 * config/avr/avr.h (AVR_HAVE_LPMX): New macro.
3245 (AVR_ENHANCED): Rename to ...
3246 (AVR_HAVE_MUL): ... new.
3247 (avr_enhanced_p): Rename to ...
3248 (avr_have_mul_p): ... new.
3249 (TARGET_CPU_CPP_BUILTINS): Use 'avr_have_mul_p' instead of
3250 'avr_enhanced_p' for "__AVR_ENHANCED__". Define "__AVR_HAVE_MUL__".
3251 * config/avr/avr.c (avr_enhanced_p): Rename to ...
3252 (avr_have_mul_p): ... new.
3253 (base_arch_s): Rename 'enhanced' to 'have_mul'.
3254 (avr_override_options): Use 'avr_have_mul_p' and 'have_mul' instead of
3255 'avr_enhanced_p' and 'enhanced'.
3256 (ashlhi3_out, ashrhi3_out, lshrhi3_out, avr_rtx_costs): Use
3257 AVR_HAVE_MUL instead of AVR_ENHANCED.
3258 * avr.md (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
3259 (mulqi3, *mulqi3_enh, *mulqi3_call, mulqihi3, umulqihi3, mulhi3,
3260 *mulhi3_enh, *mulhi3_call, mulsi3, *mulsi3_call): Use AVR_HAVE_MUL
3261 instead of AVR_ENHANCED.
3262 (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
3263 * libgcc.S: Use __AVR_HAVE_MUL__ instead of __AVR_ENHANCED__.
3264 (__tablejump__): Use __AVR_HAVE_LPMX__ instead of __AVR_ENHANCED__.
3266 2007-01-17 Ian Lance Taylor <iant@google.com>
3268 * vec.h (VEC_reserve_exact): Define.
3269 (vec_gc_p_reserve_exact): Declare.
3270 (vec_gc_o_reserve_exact): Declare.
3271 (vec_heap_p_reserve_exact): Declare.
3272 (vec_heap_o_reserve_exact): Declare.
3273 (VEC_OP (T,A,reserve_exact)): New static inline function, three
3275 (VEC_OP (T,A,reserve)) [all versions]: Remove handling of
3277 (VEC_OP (T,A,alloc)) [all versions]: Call ...reserve_exact.
3278 (VEC_OP (T,A,copy)) [all versions]: Likewise.
3279 (VEC_OP (T,a,safe_grow)) [all versions]: Likewise.
3280 * vec.c (calculate_allocation): Add exact parameter. Change all
3282 (vec_gc_o_reserve_1): New static function, from vec_gc_o_reserve.
3283 (vec_gc_p_reserve, vec_gc_o_reserve): Call vec_gc_o_reserve_1.
3284 (vec_gc_p_reserve_exact, vec_gc_o_reserve_exact): New functions.
3285 (vec_heap_o_reserve_1): New static function, from vec_heap_o_reserve.
3286 (vec_heap_p_reserve, vec_heap_o_reserve): Call vec_heap_o_reserve_1.
3287 (vec_heap_p_reserve_exact): New function.
3288 (vec_heap_o_reserve_exact): New function.
3290 2007-01-17 Jan Hubicka <jh@suse.cz>
3292 * ipa-type-escape.c (look_for_casts): Revamp using handled_component_p.
3294 2007-01-17 Eric Christopher <echristo@apple.com>
3296 * config.gcc: Support core2 processor.
3298 2007-01-16 Jan Hubicka <jh@suse.cz>
3300 * tree-ssanames.c (release_dead_ssa_names): Instead of ggc_freeing
3301 the names, just unlink the chain so we don't crash on dangling pointers
3304 2007-01-16 Jan Hubicka <jh@suse.cz>
3306 * cgraph.h (cgraph_decide_inlining_incrementally): Kill.
3307 * tree-pass.h: Reorder to make IPA passes appear toegher.
3308 (pass_early_inline, pass_inline_parameters, pass_apply_inline): Declare.
3309 * cgraphunit.c (cgraph_finalize_function): Do not compute inling
3310 parameters, do not call early inliner.
3311 * ipa-inline.c: Update comments. Include tree-flow.h
3312 (cgraph_decide_inlining): Do not compute inlining parameters.
3313 (cgraph_decide_inlining_incrementally): Return TODOs; assume to
3314 be called with function context set up.
3315 (pass_ipa_inline): Remove unreachable functions before pass.
3316 (cgraph_early_inlining): Simplify assuming to be called from the
3318 (pass_early_inline): New pass.
3319 (cgraph_gate_ipa_early_inlining): New gate.
3320 (pass_ipa_early_inline): Turn into simple wrapper.
3321 (compute_inline_parameters): New function.
3322 (gate_inline_passes): New gate.
3323 (pass_inline_parameters): New pass.
3324 (apply_inline): Move here from tree-optimize.c
3325 (pass_apply_inline): New pass.
3326 * ipa.c (cgraph_remove_unreachable_nodes): Verify cgraph after
3328 * tree-inline.c (optimize_inline_calls): Return TODOs rather than
3330 (tree_function_versioning): Do not allocate dummy struct function.
3331 * tree-inline.h (optimize_inline_calls): Update prototype.
3332 * tree-optimize.c (execute_fixup_cfg): Export.
3333 (pass_fixup_cfg): Remove
3334 (tree_rest_of_compilation): Do not apply inlines.
3335 * tree-flow.h (execute_fixup_cfg): Declare.
3336 * Makefile.in (gt-passes.c): New.
3337 * passes.c: Include gt-passes.h
3338 (init_optimization_passes): New passes.
3339 (nnodes, order): New static vars.
3340 (do_per_function_toporder): New function.
3341 (execute_one_pass): Dump current pass here.
3342 (execute_ipa_pass_list): Don't dump current pass here.
3344 2007-01-16 Janis Johnson <janis187@us.ibm.com>
3346 * config/dfp-bit.c (dfp_compare_op): Return separate value for NaN.
3347 (DFP_NE, DFP_LE, DFP_GE): Return false for NaN.
3349 2007-01-16 David Edelsohn <edelsohn@gnu.org>
3351 * config/rs6000/darwin-ldouble.c: Build file for SOFT_FLOAT.
3352 (strong_alias): Define.
3353 (__gcc_qmul): Provide non-FMA for soft-float.
3359 (__gcc_qunord): New.
3369 * config/rs6000/rs6000.c (rs6000_init_libfuncs): Initialize
3370 soft-float functions.
3371 * config/rs6000/libgcc-ppc-glibc.ver: Version soft-float symbols.
3372 * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Do not warn
3373 about long double soft float.
3375 2007-01-16 Dorit Nuzman <dorit@il.ibm.com>
3376 Tehila Meyzels <tehila@il.ibm.com>
3378 * tree-vectorizer.h (is_pattern_stmt_p): New.
3379 * tree-vect-analyze.c (vect_determine_vectorization_factor): Fix
3380 formatting (tabs instead of spaces). Cleanup and clarify setting
3381 of STMT_VINFO_VECTYPE. Call is_pattern_stmt_p.
3382 * tree-vect-transform.c (vect_get_vec_def_for_operand): Fix typo.
3383 (vectorizable_type_demotion): Check that types are integral.
3384 (vectorizable_type_promotion): Likewise.
3385 (vectorizable_store): Fix typo. Eliminate new-line at end of
3388 2007-01-16 Jan Hubicka <jh@suse.cz>
3390 * tree-ssanames.c (release_dead_ssa_names): Remove invalidated
3393 2007-01-15 Eric Christopher <echristo@apple.com>
3395 * ifcvt.c: Include vec.h, vecprim.h.
3396 (check_cond_move_block): New argument regs.
3397 Reorganize. Add registers used to regs.
3398 (cond_move_process_if_block): Use regs set above as
3401 2007-01-15 Eric Christopher <echristo@apple.com>
3403 * config/darwin.h: Update copyright.
3404 (TARGET_OPTION_TRANSLATE_TABLE): Add umbrella.
3405 (LINK_COMMAND_SPEC): Add -u.
3406 (LINK_SPEC): Fix umbrella for above.
3408 2007-01-15 Joseph S. Myers <joseph@codesourcery.com>
3410 * config/soft-fp/op-common.h, config/soft-fp/op-4.h: Update from
3413 2007-01-15 Tom Tromey <tromey@redhat.com>
3415 * doc/sourcebuild.texi (libgcj Tests): Don't mention jacks.
3416 * doc/install.texi (Testing): Don't mention jacks.
3417 (Configuration): Document --enable-java-maintainer-mode. Move
3418 --with-java-home to libgcj-specific section. Document
3420 (Prerequisites): Mention --enable-java-maintainer-mode, ecj1.
3422 2007-01-15 Jan Hubicka <jh@suse.cz>
3424 * tree-ssa-dce.c (DCE_TODOs): New.
3425 (propagate_necessity): Return if something changed.
3426 (eliminate_unnecessary_stmts): Likewise.
3427 (perform_tree_ssa_dce): Return TODO flags when needed.
3428 (pass_dce, pass_dce_loop, pass_cd_dce): Remove TODO flags.
3430 2007-01-15 Uros Bizjak <ubizjak@gmail.com>
3432 * config/i386/i386.md (fyl2xxf3_i387): Rename from fyl2x_xf3.
3433 (fyl2x_extend<mode>xf3_i387): New insn pattern.
3434 (log<mode>2): Rename from logsf2 and logdf2 and macroize insn
3435 patterns using X87MODEF12 mode macro. Extend operand 1
3436 to XFmode. Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
3437 (log10<mode>2): Ditto.
3438 (log2<mode>2): Ditto.
3439 (log1p<mode>2): Ditto.
3440 (logb<mode>2): Ditto.
3441 (fyl2xp1xf3_i387): Rename from fyl2xp1_xf3.
3442 (fyl2xp1_extend<mode>xf3_i387): New insn pattern.
3443 (*fxtractxf3_i387): Rename from *fxtractxf3.
3444 (fxtract_extend<mode>xf3_i387): New insn pattern.
3445 (ilogbsi2): Use match_dup 3, not match_operand:XF 3.
3446 * config/i386/i386.c (ix86_emit_i387_log1p): Use gen_fyl2xp1xf3_i387()
3447 and gen_fyl2xxf3_i387().
3449 2007-01-14 Zdenek Dvorak <dvorakz@suse.cz>
3451 * loop-unswitch.c (unswitch_loop): Do not call fix_loop_placement.
3452 * cfgloopmanip.c (fix_loop_placement): Made static. Use
3453 get_loop_exit_edges. Changed return type to bool.
3454 * cfgloop.h (fix_loop_placement): Declaration removed.
3456 2007-01-14 Dorit Nuzman <dorit@il.ibm.com>
3458 * param.h (MIN_VECT_LOOP_BOUND): New.
3459 * params.def (MIN_VECT_LOOP_BOUND): New.
3460 * tree-vectorizer.c (slpeel_tree_peel_loop_to_edge): Takes another
3461 argument - minimum threshold for number of iterations.
3462 * tree-vectorizer.h (slpeel_tree_peel_loop_to_edge): Add another
3463 argument to declaration.
3464 * tree-vect-analyze.c (vect_analyze_operations): Check value of
3465 MIN_VECT_LOOP_BOUND.
3466 * tree-vect-transform.c (vect_do_peeling_for_loop_bound): Call
3467 slpeel_tree_peel_loop_to_edge with additional argument.
3468 (vect_do_peeling_for_alignment): Likewise.
3469 * doc/invoke.texi (min-vect-loop-bound): Document new param option.
3471 2007-01-14 Uros Bizjak <ubizjak@gmail.com>
3474 * config/i386/i386.c (print_operand) ['z']: Output 'b' for
3477 2007-01-14 Jan Hubicka <jh@suse.cz>
3479 * tree-dfa.c (remove_referenced_var): New function.
3480 * tree-ssa-live.c (remove_unused_locals): Walk referenced vars and
3481 prune referenced vars list too.
3482 * tree-flow.h (remove_referenced_var): Declare.
3484 2007-01-14 Jan Hubicka <jh@suse.cz>
3486 * tree-eh.c (add_stmt_to_eh_region_fn): Do not add call_exprs
3488 (remove_stmt_from_eh_region_fn): Do not remove call_exprs.
3489 (verify_eh_throw_stmt_node, verify_eh_throw_table_statements): Kill.
3490 * except.h (verify_eh_throw_table_statements): Kill prototype.
3491 * cfgexpand.c (expand_gimple_basic_block): Propagate Eh regions
3492 into call exrepssions.
3493 * tree-optimize.c (execute_free_cfg_annotatiosn): Do not call
3495 * tree-cfg.c: Include pointer-set.h.
3496 (verify_node_sharing): Work on pointer set.
3497 (verify_eh_throw_stmt_node): New.
3498 (verify_stmts): Use pointers sets, verify throw_stmt.
3500 2007-01-13 Zdenek Dvorak <dvorakz@suse.cz>
3502 * ipa-reference.c (analyze_function): Consider also addresses taken
3505 2007-01-12 Roger Sayle <roger@eyesopen.com>
3507 * c-typeck.c (null_pointer_constant_p): Replace use of
3508 TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
3509 (build_c_cast): Likewise.
3511 2007-01-12 Roger Sayle <roger@eyesopen.com>
3513 * tree.h (force_fit_type_double): Remove unused final argument.
3514 * c-common.c (constant_expression_warning): Replace use of
3515 TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
3516 (convert_and_check): Likewise.
3517 (shorten_compare): Update call to force_fit_type_double.
3518 (c_common_truthvalue_conversion) <INTEGER_CST>: Use integer_zerop.
3519 * convert.c (convert_to_pointer): Update call to
3520 force_fit_type_double.
3521 * fold-const.c (force_fit_type_double): Remove overflowed_const
3523 (int_const_binop, fold_convert_const_int_from_int,
3524 fold_convert_const_int_from_real, fold_div_compare,
3525 fold_sign_changed_comparison, fold_unary, fold_negate_const,
3526 fold_abs_const, fold_not_const): Remove the final argument from
3527 calls to force_fit_type_double.
3529 2007-01-12 Andrew Pinski <andrew_pinski@playstation.sony.com>
3531 * configure.ac: Set insn to "nop" for spu-*-* also.
3532 * configure: Regenerate.
3534 2007-01-12 Olga Golovanevsky <olga@il.ibm.com>
3536 * builtins.def : Add BUILT_IN_FREE.
3538 2007-01-12 Jan Hubicka <jh@suse.cz>
3540 PR tree-optimization/30443
3541 * tree-inline.c (tree_function_versioning): Do not optimize when
3542 cloning for inlining.
3544 2007-01-12 Zdenek Dvorak <dvorakz@suse.cz>
3546 * doc/loop.texi: Document recording of loop exits.
3547 * cfgloopmanip.c (loopify, duplicate_loop): Use alloc_loop.
3548 (update_single_exits_after_duplication,
3549 update_single_exit_for_duplicated_loop,
3550 update_single_exit_for_duplicated_loops): Removed.
3551 (duplicate_loop_to_header_edge): Do not call
3552 update_single_exits_after_duplication and
3553 update_single_exit_for_duplicated_loops.
3554 (loop_version): Do not update single_exit information.
3555 (fix_loop_structure): Use record_loop_exits instead of
3556 mark_single_exit_loops.
3557 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update
3558 the lists of loop exits.
3559 * cfghooks.c (redirect_edge_and_branch, redirect_edge_and_branch_force,
3560 split_edge, merge_blocks): Update the lists of loop exits.
3561 * modulo-sched.c (sms_schedule): Pass LOOPS_HAVE_RECORDED_EXITS to
3562 loop_optimizer_init.
3563 * loop-init.c (loop_optimizer_init): Call record_loop_exits instead
3564 of mark_single_exit_loops.
3565 (loop_optimizer_finalize): Call release_recorded_exits.
3566 * tree-ssa-loop.c (tree_loop_optimizer_init): Pass
3567 LOOPS_HAVE_RECORDED_EXITS to loop_optimizer_init.
3568 * tree-vectorizer.c (slpeel_tree_duplicate_loop_to_edge_cfg): Do not
3569 update single exit information.
3570 * lambda-code.c (perfect_nestify): Ditto.
3571 * cfgloop.c (flow_loop_free): Destroy the list of exits of the loop.
3572 (mark_single_exit_loops): Removed.
3573 (alloc_loop, loop_exit_hash, loop_exit_eq, loop_exit_free,
3574 get_exit_descriptions, rescan_loop_exit, record_loop_exits,
3575 dump_recorded_exit, dump_recorded_exits, release_recorded_exits): New
3577 (get_loop_exit_edges, single_exit): Use recorded exit lists.
3578 (add_bb_to_loop, remove_bb_from_loops): Update the lists of loop exits.
3579 (verify_loop_structure): Verify consistency of the exit lists.
3580 (flow_loops_find): Use alloc_loop. Initialize exits hash.
3581 (set_single_exit): Removed.
3582 * cfgloop.h (struct loop_exit): New function.
3583 (struct loop): single_exit_ field replaced by exits field.
3584 (LOOPS_HAVE_MARKED_SINGLE_EXITS): Replaced by LOOPS_HAVE_RECORDED_EXITS.
3585 (struct loops): Added exits hash.
3586 (mark_single_exit_loops, set_single_exit): Declaration removed.
3587 (release_recorded_exits, record_loop_exits, rescan_loop_exit): Declare.
3589 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3591 * doc/invoke.texi: Avoid use of @headitem.
3593 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3595 * cse.c (cse_insn): Move HAVE_CC0 code after declarations.
3597 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3599 * doc/install.texi: Fix m68k-*-* anchor and add m68k-*-* to the
3602 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
3603 Richard Sandiford <richard@codesourcery.com>
3605 * doc/invoke.texi: Document -mno-bitfield, -mno-rtd and -mno-short.
3606 * config/m68k/m68k.opt: Resort options.
3607 (mbitfield, mrtd, mshort): Remove RejectNegative properties.
3609 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
3610 Richard Sandiford <richard@codesourcery.com>
3612 * doc/invoke.texi: Document the macros that are defined by
3613 m68k's -mtune and -mhard-float options.
3614 * config/m68k/m68k-protos.h (m68k_cpp_cpu_ident) Declare.
3615 (m68k_cpp_cpu_family): Likewise.
3616 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Add a full set
3617 of __ucfv*__ macros. Define __mcffpu__ if generating code for
3618 ColdFire FPUs. Define __mcf_cpu_* and __mcf_family_* macros.
3619 * config/m68k/m68k.c (m68k_cpp_cpu_ident): New function.
3620 (m68k_cpp_cpu_family): Likewise.
3622 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3624 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Treat all mc68*
3625 macros besides mc68000 as tuning macros. Use a switch statement
3626 to set them and mcpu32.
3628 2007-01-12 Julian Brown <julian@codesourcery.com>
3630 * config/m68k/m68k.h: Use TARGET_68040 instead of TARGET_68040_ONLY.
3631 (TARGET_68040_ONLY): Rename to...
3632 (TARGET_68040): ...this.
3633 * config/m68k/m68k.c: Use TARGET_68040 instead of TARGET_68040_ONLY.
3634 * config/m68k/m68k.md: Likewise.
3636 2007-01-12 Julian Brown <julian@codesourcery.com>
3637 Nathan Sidwell <nathan@codesourcery.com>
3638 Richard Sandiford <richard@codesourcery.com>
3640 * config.gcc (m680[012]0-*-*, m68k*-*-*): Set m68k_cpu_ident to
3641 the -mcpu= argument associated with the --with-cpu setting.
3642 Define M68K_DEFAULT_TUNE to the default -mtune= option,
3643 if different from the one implied by the -mcpu setting.
3644 Accept --with-cpu=FOO if FOO is listed in m68k-devices.def,
3645 using mcpu=FOO as the default CPU option. Set target_cpu_default2.
3646 * doc/invoke.texi: Mention ColdFire in the introduction to the
3647 m68k options. Document the new -march, -mcpu, -mtune, -mdiv,
3648 -mno-div and -mhard-float options. Make -m68881 a synonym for
3649 -mhard-float. Document the previously-undocumented -m5206e,
3650 -m528x, -m5307 and -m5407 options. Tweak the existing option
3651 documentation for consistency.
3652 * doc/install.texi: Mention new --with-cpu arguments.
3653 * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Only use the
3654 default CPU if neither -mcpu nor -march are specified.
3655 (ASM_CPU_SPEC): Pass down -mcpu and -march options.
3656 (TARGET_CPU_CPP_BUILTINS): Set __mcfisa*__ macros from
3657 TARGET_ISA*. Set the legacy __mcf*__ cpu macros in the same way,
3658 using m68k_tune to decide between families that implement the
3659 same ISA. Use m68k_tune to set __mcfv4e__.
3660 (FL_BITFIELD, FL_68881, FL_COLDFIRE, FL_CF_HWDIV, FL_CF_MAC)
3661 (FL_CF_EMAC, FL_CF_EMAC_B, FL_CF_USP, FL_CF_FPU, FL_ISA_68000)
3662 (FL_ISA_68010, FL_ISA_68020, FL_ISA_68040, FL_ISA_A, FL_ISA_B)
3663 (FL_ISA_C, FL_ISA_MMU): New macros.
3664 (MASK_COLDFIRE): Delete.
3665 (TARGET_68010, TARGET_68020, TARGET_68040_ONLY, TARGET_COLDFIRE)
3666 (TARGET_ISAB): Redefine in terms of m68k_cpu_flags.
3667 (TARGET_68881, TARGET_COLDFIRE_FPU): Redefine in terms of m68k_fpu.
3668 (TARGET_HARD_FLOAT): Do not define here.
3669 (TARGET_ISAAPLUS, TARGET_ISAC): New macros.
3670 (TUNE_68000): New macro.
3671 (TUNE_68000_10): Redefine in terms of TUNE_68000 and TUNE_68010.
3672 (TUNE_68010, TUNE_68030, TUNE_68040, TUNE_68060, TUNE_CPU32)
3673 (TUNE_CFV2): Redefine in terms of m68k_tune.
3674 (uarch_type, target_device, fpu_type): New enums.
3675 (m68k_cpu, m68k_tune, m68k_fpu, m68k_cpu_flags): Declare.
3676 * config/m68k/m68k.c (TARGET_DEFAULT): Remove MASK_68881.
3677 (FL_FOR_isa_00, FL_FOR_isa_10, FL_FOR_isa_20, FL_FOR_isa_40)
3678 (FL_FOR_isa_cpu32, FL_FOR_isa_a, FL_FOR_isa_aplus, FL_FOR_isa_b)
3679 (FL_FOR_isa_c): New macros.
3680 (m68k_isa): New enum.
3681 (m68k_target_selection): New structure.
3682 (all_devices, all_isas, all_microarchs): New tables.
3683 (m68k_cpu_entry, m68k_arch_entry, m68k_tune_entry, m68k_cpu)
3684 (m68k_tune, m68k_fpu, m68k_cpu_flags): New variables.
3685 (MASK_ALL_CPU_BITS): Delete.
3686 (m68k_find_selection): New function.
3687 (m68k_handle_option): Handle -mcpu=, -march= and -mtune=.
3688 Map the legacy target options to a combination of the new ones.
3689 (override_options): Set m68k_cpu, m68k_tune, m68k_fpu and
3690 m68k_cpu_flags. Handle M68K_DEFAULT_TUNE. Use m68k_cpu_flags
3691 to derive default MASK_BITFIELD, MASK_CF_HWDIV and MASK_HARD_FLOAT
3693 * config/m68k/m68k.opt (m5200, m5206e, m528x, m5307, m5407, mcfv4e)
3694 (m68010, m68020, m68020-40, m68020-60, m68030, m68040): Remove Mask
3696 (m68881, msoft-float): Change mask from 68881 to HARD_FLOAT.
3697 (march=, mcpu=, mdiv, mhard-float, mtune=): New options.
3698 * config/m68k/m68k-devices.def: New file.
3700 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3701 Nathan Sidwell <nathan@codesourcery.com>
3703 * config/m68k/m68k.h (ASM_CPU_SPEC, ASM_SPEC, EXTRA_SPECS)
3704 (SUBTARGET_EXTRA_SPECS): New macros.
3705 * config/m68k/linux.h (ASM_SPEC): Remove CPU flags;
3706 use %(asm_cpu_spec) instead.
3707 * config/m68k/m68k-none.h (ASM_SPEC): Likewise.
3708 * config/m68k/openbsd.h (ASM_SPEC): Likewise.
3709 * config/m68k/netbsd-elf.h (ASM_SPEC): Likewise.
3710 (EXTRA_SPECS): Rename to...
3711 (SUBTARGET_EXTRA_SPECS): ...this.
3713 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
3714 Richard Sandiford <richard@codesourcery.com>
3715 Julian Brown <julian@codesourcery.com>
3717 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
3718 (m68k-*-uclinux*, m68k-*-linux*, m68k-*-rtems*): Set default_m68k_cpu
3719 to the configuration's default CPU.
3720 (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*): Likewise.
3721 Remove default masks.
3722 (m680[012]0-*-*): Set the default with_cpu to the first part of
3724 (m68k*-*-*): Set the default with_cpu to m$default_m68k_cpu.
3725 (m68k*-*-linux): Extend the --with-cpu handling to...
3726 (m680[012]0-*-*, m68k*-*-*): ...these configurations. Allow m68000
3727 and m68010. Don't set target_cpu_default2.
3728 * doc/install.texi: Document --with-cpu for m68k.
3729 * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Define.
3730 * config/m68k/m68k-none.h (TARGET_CPU_DEFAULT, M68K_CPU_m68k)
3731 (M68K_CPU_m68000, M68K_CPU_m68010, M68K_CPU_m68020, M68K_CPU_m68030)
3732 (M68K_CPU_m68040, M68K_CPU_m68302, M68K_CPU_m68332, TARGET_DEFAULT)
3733 (ASM_CPU_DEFAULT_SPEC, CC1_CPU_DEFAULT_SPEC): Delete.
3734 (ASM_SPEC): Remove use of %(asm_cpu_default).
3735 (EXTRA_SPECS, SUBTARGET_EXTRA_SPECS, MULTILIB_DEFAULTS): Delete.
3736 * config/m68k/linux.h (TARGET_DEFAULT): Delete.
3737 (CPP_SPEC): Merge definitions. Do not handle __HAVE_68881__ here.
3738 * config/m68k/netbsd-elf.h (TARGET_OS_CPP_BUILTINS): Define
3739 __HAVE_FPU__ if TARGET_HARD_FLOAT.
3740 (TARGET_DEFAULT): Delete.
3741 (EXTRA_SPECS): Delete cpp_cpu_default_spec, cpp_cpu_spec,
3742 cpp_fpu_spec, asm_default_spec and netbsd_cpp_spec.
3743 (CPP_CPU_SPEC): Delete.
3744 (TARGET_VERSION): Merge definitions, using TARGET_68010 to pick
3745 the appropriate string.
3746 (CPP_CPU_DEFAULT_SPEC, ASM_DEFAULT_SPEC, CPP_FPU_SPEC): Delete.
3747 (CPP_SPEC): Define to NETBSD_CPP_SPEC.
3748 (ASM_SPEC): Don't use %(asm_default_spec).
3749 * config/m68k/m68k.c (TARGET_DEFAULT_TARGET_FLAGS): Remove
3750 TARGET_DEFAULT and add MASK_68881.
3751 * config/m68k/m68k.md: Remove mention of TARGET_DEFAULT from comments.
3753 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3755 * config.gcc (m68010-*-netbsdelf*): Add MASK_68010.
3756 (m68k*-*-netbsdelf*, m68k*-*-openbsd*, m68k*-linux*): Add
3757 MASK_68010 alongside MASK_68020.
3758 * doc/invoke.texi: Document -m68010.
3759 * config/m68k/m68k.opt (m68010): New.
3760 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Define mc68010
3762 (TUNE_68010): New macro.
3763 * config/m68k/m68k-none.h (M68K_CPU_m68k, M68K_CPU_m68010)
3764 (M68K_CPU_m68020, M68K_CPU_m68030, M68K_CPU_m68040)
3765 (M68K_CPU_m68332): Add MASK_68010.
3766 * config/m68k/linux.h (TARGET_DEFAULT): Add MASK_68010 to
3767 fallback definition.
3768 * config/m68k/netbsd-elf.h (CPP_CPU_SPEC): Remove now-redundant
3770 * config/m68k/m68k.c (MASK_ALL_CPU_BITS): Add MASK_68010.
3771 (m68k_handle_option): Handle OPT_m68010. Add MASK_68010
3772 to all entries that use MASK_68020.
3773 (output_move_simode_const, output_move_himode, output_move_qimode)
3774 (output_move_stricthi, output_move_strictqi): Use TARGET_68010
3775 instead of TARGET_68020 to select clr behavior. Remove comment
3776 about there being no TARGET_68010.
3777 * config/m68k/m68k.md: Likewise throughout.
3779 2007-01-12 Julian Brown <julian@codesourcery.com>
3781 * config/m68k/m68k.h (TARGET_ISAB): New macro.
3782 * config/m68k/m68k.c: Use TARGET_ISAB rather than TARGET_CFV4.
3783 * config/m68k/m68k.md: Likewise.
3785 2007-01-12 Julian Brown <julian@codesourcery.com>
3787 * config/m68k/m68k.h (LEGITIMATE_INDEX_P, LEGITIMIZE_ADDRESS): Use
3788 TARGET_COLDFIRE_FPU instead of TARGET_CFV4E.
3790 2007-01-12 Julian Brown <julian@codesourcery.com>
3792 * config/m68k/m68k.h (TUNE_68040_60): New macro.
3793 * config/m68k/m68k.c (standard_68881_constant_p): Use it.
3794 * config/m68k/m68k.md: Likewise.
3796 2007-01-12 Julian Brown <julian@codesourcery.com>
3797 Richard Sandiford <richard@codesourcery.com>
3799 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Use TUNE_68030
3800 instead of TARGET_68030, TUNE_68040 instead of TARGET_68040,
3801 TUNE_68060 instead of TARGET_68060 and TUNE_CPU32 instead of
3803 (TARGET_CPU32): Rename to...
3804 (TUNE_CPU32): ...this.
3805 (TUNE_68000_10, TUNE_68030, TUNE_68040, TUNE_68060)
3806 (TUNE_CFV2): New macros.
3807 * config/m68k/netbsd-elf.h (LONG_DOUBLE_TYPE_SIZE): Simplify;
3808 remove conditions that are implied by TARGET_68020.
3809 * config/m68k/m68k.c (m68k_output_function_prologue): Use TUNE_68040
3810 instead of TARGET_68040 and TUNE_CPU32 instead of TARGET_CPU32.
3811 (m68k_output_function_epilogue): Likewise.
3812 (m68k_rtx_costs): Likewise. Use TUNE_68060 instead of TARGET_68060
3813 and TUNE_CFV2 instead of TARGET_5200. Use TUNE_68000_10 instead of
3814 "!TARGET_68020 && !TARGET_COLDFIRE" to choose between 68000 and
3815 non-68000 timings. Refactor multiplication and division costs.
3816 (output_addsi3): Use TUNE_68040 instead of TARGET_68040 and
3817 TUNE_CPU32 instead of TARGET_CPU32.
3818 (standard_68881_constant_p): Use TUNE_68040 instead of TARGET_68040
3819 and TUNE_68060 instead of TARGET_68060.
3820 * config/m68k/m68k.md: Use TUNE_68040 instead of TARGET_68040,
3821 TUNE_68060 instead of TARGET_68060, and TUNE_CPU32 instead of
3823 (movsi_const0): Use TUNE_68000_10 rather than "!TARGET_68020
3824 && !TARGET_COLDFIRE" to choose between moveq and clr.
3825 Likewise in the unnamed movsf pattern.
3826 (ashlsi_17_24, lshrsi_17_24): Guard with TUNE_68000_10 rather than
3827 "!TARGET_68020 && !TARGET_COLDFIRE". Likewise the unnamed
3830 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3832 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Increase amount
3833 of tabbing before backslashes.
3835 2007-01-11 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
3837 * pa-linux.h (ASM_OUTPUT_INTERNAL_LABEL): Undefine.
3838 * pa.h (ASM_OUTPUT_LABEL): Output colon when using GAS.
3839 (ASM_OUTPUT_INTERNAL_LABEL): Define.
3841 2007-01-11 Zdenek Dvorak <dvorakz@suse.cz>
3843 * tree-ssa-loop-ivopts.c (extract_cond_operands): Split from
3844 find_interesting_uses_cond.
3845 (find_interesting_uses_cond): Use extract_cond_operands.
3846 (rewrite_use_compare): Use extract_cond_operands and
3847 force_gimple_operand_bsi. Do not call update_stmt.
3848 (determine_use_iv_cost_condition): Use extract_cond_operands.
3849 Return cheaper of using original bound and changing the exit bound.
3851 2007-01-11 Zdenek Dvorak <dvorakz@suse.cz>
3853 PR tree-optimization/29516
3854 * tree-ssa-address.c (tree_mem_ref_addr, add_to_parts,
3855 most_expensive_mult_to_index, addr_to_parts,
3856 create_mem_ref, maybe_fold_tmr): Make the type of
3857 fields of TARGET_MEM_REF sizetype.
3858 (move_fixed_address_to_symbol, move_pointer_to_base):
3860 * tree.def (TARGET_MEM_REF): Add comment on types of
3863 2007-01-11 Joseph Myers <joseph@codesourcery.com>
3865 * c-common.c (vector_types_convertible_p): Treat opaque types as
3866 always convertible if they have the same size, but not otherwise.
3868 2007-01-11 Steven Bosscher <steven@gcc.gnu.org>
3870 * ifcvt.c (struct noce_if_info): Add comments to the fields.
3871 Remove the b_unconditional field.
3872 (noce_try_sign_mask): Do not look at b_unconditional.
3873 (noce_process_if_block): Do not use merge_if_blocks. Update
3874 the CFG here. Do not set b_unconditional.
3875 (cond_move_process_if_block): Likewise.
3876 (find_cond_trap): Likewise.
3877 (check_cond_move_block): Require simple jump insns at the end
3880 2007-01-11 Jan Hubicka <jh@suse.cz>
3882 PR tree-optimization/1046
3883 * tree-tailcall.c (suitable_for_tail_call_opt_p): Use TREE_ADDRESSABLE
3884 when alias info is not ready.
3885 (pass_tail_recursion): Do not require aliasing.
3886 * tree-ssa-copyrename.c (pass_rename_ssa_cop): Likewise.
3887 * tree-ssa-ccp.c (pass_ccp, pass_fold_builtins): Likewise.
3888 * tree-ssa-copy.c (pass_copy_prop): Likewise.
3889 * tree-ssa-forwprop.c (pass_forwprop): Likewise.
3890 * tree-ssa-dce.c (pass_dce, pass_dce_loop, pass_cd_dce): Likewise.
3891 * passes.c (init_optimization_passes): Execute rename_ssa_copies,
3892 ccp, forwprop, copy_prop, merge_phi, copy_prop, dce and tail recursion
3894 * tree-ssa-operands.c (add_virtual_operand, get_indirect_ref_operand):
3895 When aliasing is not build, mark statement as volatile.
3897 2007-01-11 Tom Tromey <tromey@redhat.com>
3899 PR preprocessor/15185, PR preprocessor/20989:
3900 * doc/cppopts.texi <-MT>: Update description of algorithm for
3901 computing default target.
3902 <-M, -MD>: Reword "basename" text.
3904 2007-01-11 Roger Sayle <roger@eyesopen.com>
3906 * builtins.c (expand_builtin_pow, expand_builtin_powi,
3907 fold_builtin_cabs, fold_builtin_sqrt, fold_builtin_trunc,
3908 fold_builtin_floor, fold_builtin_ceil, fold_builtin_round,
3909 fold_builtin_int_int_roundingfn, fold_builtin_bitop,
3910 fold_builtin_bswap, real_constp, fold_builtin_pow,
3911 fold_builtin_powi, fold_builtin_signbit, fold_builtin_copysign,
3912 do_mpfr_arg1, do_mpfr_arg2, do_mpfr_arg3, do_mpfr_sincos): Replace
3913 uses of the macro TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
3914 * convert.c (convert_to_pointer): Likewise.
3915 * expr.c (highest_pow2_factor, expand_expr_real_1): Likewise.
3916 * fold-const.c (force_fit_type, fold_negate_expr, int_const_binop,
3917 const_binop, fold_convert_const_int_from_int,
3918 fold_convert_const_int_from_real,
3919 fold_convert_const_real_from_real, sign_bit_p,
3920 optimize_minmax_comparison, extract_muldiv_1, fold_div_compare,
3921 fold_sign_changed_comparison, fold_unary, fold_comparison,
3922 fold_binary, multiple_of_p, tree_Expr_non_zero_p,
3923 fold_negate_const, fold_abs_const, fold_not_const): Likewise.
3924 * print-tree.c (print_node_brief, print_node): Likewise.
3925 * stor-layout.c (place_field, layout_type): Likewise.
3926 * tree-chrec.c (keep_cast): Likewise.
3927 * tree.c (build_vector, build_real, build_real_from_int_cst,
3928 build_complex): Likewise.
3930 2007-01-11 Roger Sayle <roger@eyesopen.com>
3932 * tree.h (TREE_CONSTANT_OVERFLOW): Obsolete. For the time being,
3933 treat TREE_CONSTANT_OVERFLOW as a synonym of TREE_OVERFLOW.
3935 2007-01-11 Paolo Bonzini <bonzini@gnu.org>
3937 * configure.ac (strict1_warn): Rename to strict_warn.
3938 (WERROR, --enable-werror, symlink hacks, stage1_cflags,
3939 cc_set_by_configure, quoted_cc_set_by_configure,
3940 stage_prefix_set_by_configure, quoted_stage_prefix_set_by_configure,
3941 all_boot_languages, all_stagestuff): Remove.
3942 (target_list): Remove bootstrap targets.
3943 * Makefile.in (quickstrap): Unconditionally make a synonym of all.
3944 (BOOT_LANGUAGES, STAGE1_CFLAGS, STAGE1_CHECKING,
3945 REMAKEFLAGS, FLAGS_TO_PASS, PREPEND_DOTDOT_TO_RELATIVE_PATHS,
3946 SUBDIR_FLAGS_TO_PASS, WERROR_FLAGS, STRICT2_WARN, LANG_STAGESTUFF,
3947 VOL_FILES, POSTSTAGE1_FLAGS_TO_PASS, STAGE2_FLAGS_TO_PASS,
3948 STAGEPROFILE_FLAGS_TO_PASS, STAGEFEEDBACK_FLAGS_TO_PASS, stage1_build,
3949 stage1_copy, stage2_build, stage2_copy, stageprofile_build,
3950 stageprofile_copy, stage3_build, stage3_copy, stagefeedback_build,
3951 stagefeedback_copy, stage4_build, clean_s1, clean_sw, bootstrap,
3952 bootstrap-lean, bootstrap2, bootstrap2-lean, bootstrap3,
3953 bootstrap3-lean, bootstrap4, bootstrap4-lean, unstage1, unstage2,
3954 unstage3, unstage4, unstageprofile, unstagefeedback, restage, restage2,
3955 restage3, restage4, restageprofile, restagefeedback, bubbleestrap,
3956 cleanstrap, unstrap, restrap, *compare, *compare3, *compare4,
3957 *compare-lean, *compare3-lean, *compare4-lean, stage1-start, stage1,
3958 stage2-start, stage2, stage3-start, stage3, stage4-start, stage4,
3959 stageprofile-start, stageprofile, stagefeedback-start, stagefeedback,
3960 risky-stage1, risky-stage2, risky-stage3, risky-stage4): Remove.
3961 (ORDINARY_FLAGS_TO_PASS): Rename to FLAGS_TO_PASS.
3962 (STAGECOPYSTUFF, STAGEMOVESTUFF): Consolidate into MOSTLYCLEANFILES.
3963 (mostlyclean): Adjust.
3964 (clean, distclean): Don't mention bootstrap stuff.
3965 * configure: Regenerate.
3966 * ada/config-lang.in, cp/config-lang.in, forttran/config-lang.in,
3967 java/config-lang.in, objc/config-lang.in, objcp/config-lang.in,
3968 treelang/config-lang.in (stagestuff): Remove.
3969 * doc/sourcebuild.texi (stage1, stage2, stage3, stage4,
3970 stageprofile, stagefeedback, stagestuff): Remove mention.
3972 2007-01-11 Nick Clifton <nickc@redhat.com>
3974 * config/mcore/predicates.md (mcore_general_movesrc_operand):
3976 (mcore_general_movdst_operand): Do not accept CONST_INTs.
3977 (mcore_arith_K_S_operand): Run the test for the S constraint not
3978 the test for the M constraint.
3979 (mcore_addsub_operand): Do not accept integer values that are
3980 larger than 32 bits.
3981 * config/mcore/mcore.md: Remove unused constraints from split.
3982 (andsi3): Use HOST_WIDE_INT instead of int to hold an INTVAL.
3984 (allocate_stack): Likewise.
3985 * config/mcore/mcore.c (mcore_print_operand): Restrict output of P
3986 operands to 32 bits.
3987 (mcore_const_costs): Use HOST_WIDE_INT instead of int to hold an
3989 (mcore_and_cost, mcore_modify_comparison, const_ok_for_mcore,
3990 mcore_const_ok_for_inline, mcore_const_trick_uses_not,
3991 try_constant_tricks, mcore_num_ones, mcore_num_zeros,
3992 mcore_output_bclri, mcore_output_andn, output_inline_const,
3993 mcore_output_move, mcore_output_movedouble): Likewise.
3994 (mcore_output_cmov): Use CONST_OK_FOR_M and CONST_OK_FOR_N.
3995 (output_inline_const): Likewise.
3996 (output_inline_const): Fix format strings used in sprintf
3998 * config/mcore/mcore-protos.h: Update prototypes for changed
3999 functions in mcore.c.
4000 * config/mcore/mcore.h (CONST_OK_FOR_I): Cast values to
4001 HOST_WIDE_INT and not int.
4002 (CONST_OK_FOR_J, CONST_OK_FOR_K, CONST_OK_FOR_L, CONST_OK_FOR_M,
4003 CONST_OK_FOR_N): Likewise.
4004 (LEGITIMATE_CONSTANT_P): Also check CONSTANT_P.
4005 (GO_IF_LEGITIMATE_INDEX): Use HOST_WIDE_INT instead of int to hold
4008 2007-01-10 Jan Hubicka <jh@suse.cz>
4010 * tree-vrp.c (remove_range_assertions): Release defs.
4011 * tree-ssa-loop-ivopts.c (rmeove_statement): Likewise.
4012 * tree-ssa-dom.c (remove_stmt_or_phi): Likewise.
4014 2007-01-10 Paul Brook <paul@codesourcery.com>
4016 * config/arm/arm.c (arm_rtx_costs_1): Handle mutiply-subtract.
4017 * config/arm/arm.md (mulsi3subsi): New insn.
4019 2007-01-10 Zdenek Dvorak <dvorakz@suse.cz>
4021 * tree-ssa-loop-manip.c (tree_unroll_loop): Make it a wrapper over ...
4022 (tree_transform_and_unroll_loop): New.
4023 * tree-flow.h (transform_callback, tree_transform_and_unroll_loop):
4026 2007-01-10 Robert Kennedy <jimbob@google.com>
4028 * fold-const.c (fold_comparison): Fold comparisons like (x *
4029 1000 < 0) to (x < 0).
4031 2007-01-10 Ian Lance Taylor <iant@google.com>
4033 * tree-pretty-print.c (dump_generic_node): Print parentheses when
4034 operands have the same priority.
4036 2007-01-10 Tom Tromey <tromey@redhat.com>
4038 * fold-const.c (fold_truthop): Don't check can_use_bit_fields_p.
4039 (fold_binary): Likewise.
4040 * langhooks.c (lhd_can_use_bit_fields_p): Removed.
4041 * langhooks-def.h (lhd_can_use_bit_fields_p): Removed.
4042 (LANG_HOOKS_CAN_USE_BIT_FIELDS_P): Removed.
4043 (LANG_HOOKS_INITIALIZER): Remove LANG_HOOKS_CAN_USE_BIT_FIELDS_P.
4044 * langhooks.h (struct lang_hooks): Removed field
4045 'can_use_bit_fields_p'.
4047 2007-01-10 Ralf Corsépius <ralf.corsepius@rtems.org>
4049 * config/bfin/t-bfin, config/bfin/t-bfin-elf: Remove GCC_CFLAGS.
4051 2007-01-10 Razya Ladelsky <razya@il.ibm.com>
4053 * function.c (get_last_funcdef_no): New function.
4054 * function.h (get_last_funcdef_no): Declare.
4055 * tree-inline.c (initialize_cfun): Add initialization.
4056 (tree_function_versioning): Cleanup.
4058 2007-01-10 Jan Hubicka <jh@suse.cz>
4060 * tree-inline.c (setup_one_parameter): Do not propagate into abnormal
4063 2007-01-10 Sa Liu <saliu@de.ibm.com>
4064 Ben Elliston <bje@au.ibm.com>
4066 * spu.h (STACK_SAVE_AREA): Use VOIDmode for SAVE_FUNCTION, SImode
4067 for SAVE_NONLOCAL and Pmode for any other save level.
4068 * spu-protos.h (spu_restore_stack_block): Declare.
4069 * spu.md (save_stack_block): Remove.
4070 (restore_stack_block): Call spu_restore_stack_block.
4071 * spu.c (spu_restore_stack_block): New function.
4072 (spu_expand_epilogue): Remove old comment.
4074 2007-01-09 Zdenek Dvorak <dvorakz@suse.cz>
4076 PR tree-optimization/30322
4077 * tree-ssa-loop-ivopts.c (fold_affine_expr, iv_value): Removed.
4078 (cand_value_at): Return the value as aff_tree.
4079 (may_eliminate_iv): Convert the bound from aff_tree to tree.
4080 * tree-affine.c (aff_combination_add_cst, aff_combination_add_product,
4081 aff_combination_mult): New functions.
4082 (aff_combination_add): Use aff_combination_add_cst.
4083 (aff_combination_convert): Allow conversions to a wider type.
4084 (tree_to_aff_combination): Handle BIT_NOT_EXPR.
4085 * tree-affine.h (aff_combination_mult): Declare.
4087 2007-01-09 Carlos O'Donell <carlos@codesourcery.com>
4089 * doc/tm.texi: Update documentation to reflect reality of exec
4090 and start file search behaviours. Update copyright year.
4091 * doc/invoke.texi: Explain how GCC_EXEC_PREFIX is used to find
4092 header file directories.
4094 2007-01-09 Uros Bizjak <ubizjak@gmail.com>
4096 * config/i386/i386.md (*sinxf2): Rename to *sinxf2_i387.
4097 (*cosxf2): Rename to cosxf2_i387.
4098 (*sindf2, *sinsf2): Extend operand 1 to XFmode. Macroize patterns
4099 using X87MODEF12 mode macro. Rename patterns to
4100 *sin_extend<mode>xf2_i387. Use SSE_FLOAT_MODE_P to disable patterns
4102 (*cosdf2, *cossf2): Ditto.
4103 (sincosdf3, sincossf3): Ditto. Rewrite corresponding splitters
4104 to match extended input operands.
4105 (sincos<mode>3): New expander.
4106 (*sinextendsfdf2, *cosextendsfdf2, *sincosextendsfdf3): Remove
4107 insn patterns and corresponding splitters.
4109 2007-01-09 Kaz Kojima <kkojima@gcc.gnu.org>
4111 * config/sh/t-linux (TARGET_LIBGCC2_CFLAGS): Delete.
4112 (SHLIB_MAPFILES, SHLIB_LINK, SHLIB_INSTALL): Likewise.
4114 2007-01-09 Nicolas Pitre <nico@cam.org>
4117 * arm/ieee754-df.S (Lad_s): Also test the low word of X for zero.
4119 2007-01-08 Geoffrey Keating <geoffk@apple.com>
4121 * target.h (struct gcc_target): New field library_rtti_comdat.
4122 * target-def.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): New.
4123 (TARGET_CXX): Add TARGET_CXX_LIBRARY_RTTI_COMDAT.
4124 * doc/tm.texi (C++ ABI): Document TARGET_CXX_LIBRARY_RTTI_COMDAT.
4125 * config/darwin.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
4127 2007-01-08 Geoffrey Keating <geoffk@apple.com>
4129 * doc/invoke.texi (Optimize Options): Correct description of -O0.
4131 2007-01-08 Richard Guenther <rguenther@suse.de>
4133 * tree.h (force_fit_type_double): Export.
4134 (force_fit_type): Remove.
4135 * fold-const.c (force_fit_type_double): New function.
4136 (force_fit_type): Remove.
4137 (int_const_binop): Use it.
4138 (fold_convert_const_int_from_int): Likewise.
4139 (fold_convert_const_int_from_real): Likewise.
4140 (fold_div_compare): Likewise.
4141 (fold_sign_changed_comparison): Likewise.
4142 (fold_unary): Likewise.
4143 (fold_negate_const): Likewise.
4144 (fold_abs_const): Likewise.
4145 (fold_not_const): Likewise.
4146 * c-common.c (shorten_compare): Use force_fit_type_double.
4147 * convert.c (convert_to_pointer): Likewise.
4149 2007-01-08 Richard Guenther <rguenther@suse.de>
4151 * tree.h (build_int_cst_wide_type): Export.
4152 * tree.c (build_int_cst_wide_type): New function.
4153 (build_int_cst_wide): Fix comment.
4154 * builtins.c (fold_builtin_object_size): Use build_int_cst
4155 to build -1 or 0 of the correct type. Use fit_double_type
4156 to check for overflow.
4157 * fold-const.c (optimize_bit_field_compare): Use build_int_cst_type
4159 (decode_field_reference): Likewise.
4160 (all_ones_mask_p): Likewise.
4161 (native_interpret_int): Use build_int_cst_wide_type.
4162 (fold_binary): Use build_int_cst_type to build an all-ones
4164 * stor-layout.c (set_sizetype): Use build_int_cst_wide_type.
4166 2007-01-08 Daniel Jacobowitz <dan@codesourcery.com>
4168 * config/pa/t-pa64 (libgcc_stub.a): Use $(T).
4170 2007-01-09 Ben Elliston <bje@au.ibm.com>
4172 * genautomata.c (STATS_OPTION): New option.
4173 (stats_flag): New flag.
4174 (gen_automata_option): Handle it.
4175 (initiate_automaton_gen): Ditto.
4176 (write_automata): Output statistics only if stats_flag is
4177 set. Likewise, output time statistics only if time_flag is set.
4178 * doc/md.texi (Processor pipeline description): Document new flag.
4180 2007-01-08 Richard Guenther <rguenther@suse.de>
4182 * builtins.c (fold_builtin_int_roundingfn): Use fit_double_type.
4183 * tree.c (build_int_cst_type): Likewise.
4184 (size_in_bytes): Don't call force_fit_type on the result.
4185 (int_fits_type_p): Use fit_double_type.
4186 * fold-const.c (fit_double_type): New function.
4187 (force_fit_type): Use it.
4188 * tree.h (fit_double_type): Export.
4190 2007-01-08 Jan Hubicka <jh@suse.cz>
4192 * tree-vectorizer.c (gate_increase_alignment): Fix return type.
4193 * ipa.c (function_and_variable_visibility): Fix return type.
4195 2007-01-08 Richard Guenther <rguenther@suse.de>
4197 * tree-ssa-ccp.c (maybe_fold_offset_to_array_ref): Use type
4198 of offset to build the index.
4199 * tree-pretty-print.c (dump_generic_node): Don't build negated
4200 const just for printing.
4201 * c-pretty-print.c (pp_c_integer_constant): Likewise.
4202 * builtins.c (fold_builtin_int_roundingfn): Check if result
4203 fits the type by using force_fit_type and comparing the result.
4204 * predict.c (predict_loops): Use compare_tree_int for comparison.
4205 * tree.c (build_int_cst): Fall back to integer_type_node for
4207 (build_int_cst_wide): Assert type is non-null.
4209 2007-01-08 Roberto Costa <roberto.costa@st.com>
4211 * tree-vrp.c (extract_range_from_cond_expr): New.
4212 (extract_range_from_expr): Handle COND_EXPR nodes used as expressions.
4213 * tree-ssa-ccp.c (get_maxval_strlen): Handle COND_EXPR nodes used
4215 (fold_stmt): Bug fix, avoid infinite recursion when folding COND_EXPRs.
4216 * tree-ssa-forwprop.c (simplify_cond, forward_propagate_into_cond,
4217 tree_ssa_forward_propagate_single_use_vars): Handle COND_EXPR nodes
4218 used as expressions.
4219 * tree-object-size.c (cond_expr_object_size): New.
4220 (collect_object_sizes_for): Handle COND_EXPR nodes used as expressions.
4222 2007-01-08 Jan Hubicka <jh@suse.cz>
4224 * tree-ssa-forwprop.c (forward_propagate_into_cond,
4225 tree_ssa_forward_propagate_single_use_va): Release defs of propagated
4228 2007-01-08 Richard Guenther <rguenther@suse.de>
4230 PR tree-optimization/23603
4231 * tree-vrp.c (set_value_range_to_truthvalue): New function.
4232 (extract_range_from_binary): Fall back to truthvalue instead of
4233 varying for TRUTH_*_EXPR.
4234 (extract_range_from_comparison): Fall back to truthvalue instead of
4236 (vrp_visit_phi_node): Don't adjust new range bounds to +INF/-INF
4237 if all visited PHI values were constant.
4239 2007-01-08 Jan Hubicka <jh@suse.cz>
4241 * cgraphunit.c (cgraph_process_new_functions): Reset reachable flag.
4242 (cgraph_analyze_function): break out from ...
4243 (cgraph_finalize_compilation_unit): ... here.
4244 (cgraph_expand_function): Remove forgoten commented out line.
4245 (cgraph_optimize): Analyze functions.
4247 2007-01-08 Jan Hubicka <jh@suse.cz>
4249 * tree-pas.h (TODO_remove_function): New flag.
4250 (TODO_update*): Renumber.
4251 (pass_ipa_increase_alignment,
4252 pass_ipa_function_and_variable_visibility): New passes.
4253 * cgraphunit.c (cgraph_increase_alignment): Move to tree-vectorizer.c
4254 (cgraph_function_and_variable_visibility): Move to ipa.c
4255 (cgraph_optimize): Don't call cgraph_function_and_variable_visibility,
4256 cgraph_increase_alignment.
4257 * ipa-inline.c (cgraph_decide_inlining): Don't push timevar.
4258 (cgraph_decide_inlining_incrementally): Push TV_INTEGRATION before
4259 calling tree-inline.
4260 (cgraph_early_inlining): Do not call cgraph_remove_unreachable_nodes.
4261 (pass_ipa_inline, pass_early_ipa_inlining): Set TODO_remove_functions
4262 * tree-vectorizer.c (increase_alignment): Move here from cgraphunit.c
4263 (gate_increase_alignment): New function.
4264 (pass_ipa_increase_alignment): New pass.
4265 * ipa.c: Inline tree-pass.h and timevar.h
4266 (function_and_variable_visibility): Move here from cgraphunit.c
4267 * tree-optimize.c (pass_early_local_passes): Add TODO_remove_functions.
4268 * passes.c (init_optimization_passes): Add the two new passes.
4269 (execute_todo): Handle cgraph_remove_functions.
4271 2007-01-08 Nick Clifton <nickc@redhat.com>
4273 * config/frv/predicates.md (reg_or_0_operand): Accept
4276 2007-01-08 Ralf Corsépius <ralf.corsepius@rtems.org>
4278 * config/bfin/rtems.h, config/bfin/t-rtems: New.
4279 * config.gcc: Add bfin*-rtems*.
4281 2007-01-08 Mark Shinwell <shinwell@codesourcery.com>
4283 * c.opt: Add -flax-vector-conversions.
4284 * c-typeck.c (convert_for_assignment): Pass flag to
4285 vector_types_convertible_p to allow emission of note.
4286 (digest_init): Likewise.
4287 * c-opts.c: Handle -flax-vector-conversions.
4288 * c-common.c (flag_lax_vector_conversions): New.
4289 (vector_types_convertible_p): Unless -flax-vector conversions
4290 has been passed, disallow conversions between vectors with
4291 differing numbers of subparts and/or element types. If such
4292 a conversion is disallowed, possibly emit a note on the first
4293 occasion only to inform the user of -flax-vector-conversions.
4294 The new last argument specifies this.
4295 * c-common.h (flag_lax_vector_conversions): New.
4296 (vector_types_convertible_p): Add extra argument.
4297 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Use
4298 char_type_node for V*QI type vectors.
4299 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins):
4300 Update to satisfy new typechecking rules.
4301 * config/rs6000/altivec.h (vec_cmple): Use vec_cmpge, for both
4303 * doc/invoke.texi (C Dialect Options): Document
4304 -flax-vector-conversions.
4306 2007-01-08 Mark Shinwell <shinwell@codesourcery.com>
4308 PR tree-optimization/29877
4309 * tree-ssa-ter.c (is_replaceable_p): Deem assignments with
4310 a register variable on the RHS to not be replaceable.
4312 2007-01-08 Chen Liqin <liqin@sunnorth.com.cn>
4313 * config/score/t-score-elf (MULTILIB_OPTIONS): Change.
4314 * config/score/predicates.md (const_uimm5, sr0_operand, const_simm12,
4315 const_simm15, const_pow2, const_npow2): Added.
4316 * config/score/misc.md (insv, extv, extzv, movmemsi,
4317 move_lbu_a/b, mov_lhu_a/b etc): Added and fix some bug.
4318 * config/score/score.c (score_address_cost, score_select_cc_mode):
4320 Change CONST_OK_FOR_LETTER_P/EXTRA_CONSTRAINT define.
4321 Update score_rtx_costs for MACRO TARGET_RTX_COSTS.
4322 Update score_print_operand.
4323 * config/score/score.h (DATA_ALIGNMENT, SELECT_CC_MODE): Added.
4324 Adjust register allocate order and update some macro define.
4325 * config/score/score-mdaux.c (mdx_unaligned_load, mdx_unsigned_store,
4326 mdx_block_move_straight, mdx_block_move_loop_head,
4327 mdx_block_move_loop_body, mdx_block_move_loop_foot, mdx_block_move_loop,
4328 mdx_block_move): Added.
4329 (mdx_movsicc, mdp_select_add_imm, mdp_select, mds_zero_extract_andi,
4330 mdp_limm): Updated and fix some bug and typo.
4331 * config/score/score.md (movqi/hi/si, add/sub/zero/ext): Updated.
4332 (movsf, movdf, doloop_end): Added.
4334 2007-01-08 Kazu Hirata <kazu@codesourcery.com>
4336 * config/arm/arm.c, config/arm/arm.h, config/arm/arm.md,
4337 config/arm/thumb2.md: Fix comment typos.
4338 * doc/extend.texi: Fix a typo.
4340 2007-01-07 Eric Christopher <echristo@apple.com>
4342 * configure.ac: Check for __stack_chk_fail for darwin.
4343 * configure: Regenerate.
4345 2007-01-07 Richard Guenther <rguenther@suse.de>
4347 * tree-vrp.c (extract_range_from_assert): CSE calls to
4348 compare_values where possible.
4349 (extract_range_from_unary_expr): Likewise.
4351 2007-01-07 Anatoly Sokolov <aesok@post.ru>
4353 * config/avr/avr-protos.h (call_insn_operand): Delete prototype.
4354 * config/avr/avr.c (call_insn_operand): Delete function.
4355 * config/avr/avr.md (*pushqi, *pushhi, *pushsi, *pushsf): Use REG_SP
4356 instead of register number. Use predicates.
4357 * config/avr/predicates.md (const0_operand, reg_or_0_operand,
4358 call_insn_operand): Add.
4360 2007-01-06 Jan Hubicka <jh@suse.cz>
4362 * tree-pass.h (pass_build_cgraph_edges): Declare.
4363 * cgraphunit.c (record_refernece): Move to cgraphbuild.c
4364 (visited_nodes): Remove.
4365 (cgraph_create_edges): Move to cgraphbuild.c; rename to
4366 build_cgrpah_edges; make visited_nodes local.
4367 (cgraph_process_new_functions): DO not call initialize_inline_failed.
4368 (record_references_in_initializer): Move to cgraphbuild.c
4369 (initialize_inline_failed, rebuild_cgraph_edges,
4370 pass_rebuild_cgraph_edges): Move to cgraphbuild.c.
4371 (verify_cgraph_node): Make visited_nodes local.
4372 (cgraph_analyze_function): Do not call cgraph_create_edges and
4373 initialize_inline_failed.
4374 (cgraph_expand_function): Do not call cgraph_lower_function;
4375 assert that function is already lowered.
4376 * Makefile.in (cgraphbuild.o): New.
4377 * passes.c (init_optimization_passes): Add pass_build_cgraph_edges
4378 at the end of lowering passes.
4380 2007-01-06 Steven Bosscher <steven@gcc.gnu.org>
4382 * ifcvt.c (cond_move_convert_if_block): New function, code
4383 factored out from...
4384 (cond_move_process_if_block): ...here. Call the new function
4385 on the THEN and ELSE blocks.
4386 (merge_if_block): Do not copy global_live_at_end, merge_blocks
4387 already takes care of this.
4389 2007-01-05 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
4392 * tree.h (TREE_OVERFLOW_P): New.
4393 * c-typeck.c (parser_build_unary_op): Warn only if result
4394 overflowed and operands did not.
4395 (parser_build_binary_op): Likewise.
4396 (convert_for_assignment): Remove redundant overflow_warning.
4397 * c-common.c (overflow_warning): Don't check or set TREE_OVERFLOW.
4399 2007-01-05 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
4401 * c-typeck.c (store_init_value): Split over two lines to follow
4402 the GNU coding style.
4404 2007-01-05 Benjamin Kosnik <bkoz@redhat.com>
4406 * c-cppbuiltin.c (c_cpp_builtins): __GXX_EXPERIMENTAL_CPP0X__ to
4407 __GXX_EXPERIMENTAL_CXX0X__.
4408 * doc/cpp.texi: Same.
4410 2007-01-05 Richard Guenther <rguenther@suse.de>
4413 * tree.c (get_narrower): Do not construct COMPONENT_REFs
4414 with mismatched types. Instead explicitly build a
4415 conversion NOP_EXPR.
4417 2007-01-05 Ian Lance Taylor <iant@google.com>
4419 * c-common.c (decl_with_nonnull_addr_p): New function.
4420 (c_common_truthvalue_conversion): Call it.
4421 * c-typeck.c (build_binary_op): Likewise.
4422 * c-common.h (decl_with_nonnull_addr_p): Declare.
4424 2007-01-05 Jakub Jelinek <jakub@redhat.com>
4427 * libgcc2.c (__divdc3): Compare c and d against 0.0 instead of
4430 2007-01-05 Joel Brobecker <brobecker@adacore.com>
4432 * doc/install.texi (Final install): Document the fact that
4433 the GNAT runtime should not be stripped.
4435 2007-01-04 Jan Hubicka <jh@suse.cz>
4437 * tree-inline.c (fold_marked_statements): Update operand caches
4438 and EH after folding
4440 2007-01-04 Ian Lance Taylor <iant@google.com>
4442 * c-common.c (check_function_nonnull): Whitespace fix.
4444 2007-01-04 Jan Hubicka <jh@suse.cz>
4446 * tree-optimize.c (execute_fixup_cfg): Correct previously mistakely
4447 comitted older version of patch.
4448 (pass_fixup_cfg): Add TODOs to verify flow and statements, dump
4449 function, celanup cfg and collect garbage.
4451 2007-01-04 Mike Stump <mrs@apple.com>
4453 * Makefile.in (mostlyclean): Don't remove libgcc anymore.
4456 2007-01-04 Eric Christopher <echristo@apple.com>
4458 * libgcc2.c (__bswapsi2): Use SItype.
4459 (__bswapdi2): Use DItype.
4460 * libgcc2.h: Update for above.
4462 2007-01-04 Paul Brook <paul@codesourcery.com>
4464 * config/arm/arm.md (arm_mulsi3, thumb_mulsi3, mulsi3_compare0,
4465 mulsi_compare0_scratch, mulsi3addsi, mulsi3addsi_compare0,
4466 mulsi3addsi_compare0_scratch, mulsidi3adddi, mulsidi3,
4467 umulsidi3, umulsidi3adddi, smulsi3_highpart,
4468 umulsi3_highpart): Make conditional on !arm_arch6.
4469 (arm_mulsi3_v6, thumb_mulsi3_v6, mulsi3_compare0_v6,
4470 mulsi_compare0_scratch_v6, mulsi3addsi_v6, mulsi3addsi_compare0_v6,
4471 mulsi3addsi_compare0_scratch_v6, mulsidi3adddi_v6, mulsidi3_v6,
4472 umulsidi3_v6, umulsidi3adddi_v6, smulsi3_highpart_v6,
4473 umulsi3_highpart_v6): New insns.
4475 2007-01-04 Roger Sayle <roger@eyesopen.com>
4477 * fold-const.c (fold_convert): When casting an expression to void,
4478 fold_ignored_result may discover a GIMPLE_MODIFY_STMT which doesn't
4479 have a type. Instead of attempting to build a NOP_EXPR, return
4480 these "special" trees directly.
4482 2007-01-04 Joseph Myers <joseph@codesourcery.com>
4484 * config/rs6000/rs6000.c (rs6000_rtx_costs): Make adjustment for
4485 MULT inside MINUS as either argument. Use rs6000_cost->dmul -
4486 rs6000_cost->fp not 0 as adjustment for outer NEG.
4488 2007-01-04 Jan Hubicka <jh@suse.cz>
4490 * cgraph.c (cgraph_release_function_body): New function.
4491 (cgraph_remove_node): Use it.
4492 * cgraph.h (cgraph_release_function_body): Declare.
4493 * cgraphunit.c (cgraph_expand_function): Use it.
4494 * ipa.c (cgraph_remove_unreahchable_nodes): Use it.
4495 * tree-ssa.c (delete_tree_ssa): Allow to be called before aliasing
4496 is initialized and while compilation of other function is running.
4497 * tree-optimize.c (execute_free_cfg_annotations): Move code to clear
4498 statement CFG annotations from here to ...
4499 * tree-cfg.c (delete_tree_cfg_annotations): ... here.
4501 2007-01-04 Zdenek Dvorak <dvorakz@suse.cz>
4503 * cfgloop.h (enum li_flags): Make the constants powers of two.
4505 2007-01-04 Jan Hubicka <jh@suse.cz>
4507 * tree-inline.c (copy_bb): Insert new statements to statements_to_fold
4509 (fold_marked_statements): New function.
4510 (optimize_inline_calls, tree_function_versioning): Fold new statements.
4511 * tree-inline.h (copy_body_data): Add statements_to_fold.
4513 2007-01-03 Daniel Jacobowitz <dan@codesourcery.com>
4515 * config.gcc: Mention libgcc/config.host.
4516 * Makefile.in: Update comments mentioning libgcc.
4517 (LIBGCC, INSTALL_LIBGCC, GCC_PARTS, mklibgcc): Delete.
4518 (all.cross, start.encap, rest.encap, rest.cross): Update
4519 dependencies for libgcc move.
4520 (libgcc.mk, LIBGCC_DEPS, libgcov.a, libgcc.a, stmp-multilib)
4521 (clean-target, clean-target-libgcc): Delete.
4522 (srcdirify, GCC_EXTRA_PARTS): New macros.
4523 (libgcc-support, libgcc.mvars): New rules.
4524 (distclean): Remove mention of mklibgcc.
4525 (install): Don't reference INSTALL_LIBGCC.
4526 (install-common): Don't reference EXTRA_PARTS.
4527 (install-libgcc, install-multilib): Delete rules.
4528 * mklibgcc.in: Delete file.
4529 * doc/configfiles.texi: Don't mention mklibgcc.
4531 * config/i386/t-darwin (SHLIB_VERPFX): Delete (moved to libgcc).
4532 * config/i386/t-darwin64 (SHLIB_VERPFX): Likewise.
4533 * config/rs6000/t-darwin (SHLIB_VERPFX): Likewise.
4534 * config/rs6000/t-ppccomm (TARGET_LIBGCC2_CFLAGS, SHLIB_MAPFILES)
4535 (mklibgcc, ldblspecs): Likewise.
4537 * config/i386/t-nwld (libgcc.def, libc.def, libpcre.def)
4538 (posixpre.def): Use $(T).
4539 (SHLIB_EXT, SHLIB_NAME, SHLIB_SLIBDIR_QUAL, SHLIB_DEF, SHLIB_MAP)
4540 (SHLIB_SRC, SHLIB_INSTALL): Delete.
4541 (SHLIB_LINK): Make dummy.
4542 * config/t-slibgcc-darwin: Delete contents except for dummy SHLIB_LINK.
4544 * config/frv/t-linux (EXTRA_MULTILIB_PARTS): Clear.
4546 * config/alpha/t-crtfm: Use $(T) in rules for EXTRA_PARTS.
4547 * config/alpha/t-vms, config/alpha/t-vms64, config/fr30/t-fr30,
4548 config/i386/t-rtems-i386, config/ia64/t-ia64, config/rs6000/t-beos,
4549 config/rs6000/t-newas, config/sparc/t-elf: Likewise.
4551 * configure.ac (all_outputs): Remove mklibgcc.
4552 * configure: Regenerated.
4554 2007-01-03 Josh Conner <jconner@apple.com>
4557 * calls.c (compute_argument_addresses): Set stack and stack_slot
4558 for partial args, too.
4559 (store_one_arg): Use locate.size.constant for the size when
4560 generating a save_area.
4562 2007-01-03 Robert Kennedy <jimbob@google.com>
4564 * tree-cfg.c (tree_merge_blocks): Release SSA_NAME phi results
4565 whose definitions are deleted due to basic block merging.
4567 2007-01-03 Paul Brook <paul@codesourcery.com>
4570 * config/arm/arm.c (output_return_instruction): Pop PC in interrupt
4572 (use_return_insn): Return 0 for Thumb interrupt functions.
4573 (print_multi_reg): Add rfe argument for IRQ returns.
4574 (arm_output_epilogue): Pop interrupt return address directly into PC.
4575 (arm_expand_prologue): Only adjust IRQ return address in Arm mode.
4577 2007-01-03 Paul Brook <paul@codesourcery.com>
4579 Merge from sourcerygxx-4_1.
4580 * config/arm/thumb2.md: New file.
4581 * config/arm/elf.h (JUMP_TABLES_IN_TEXT_SECTION): Return True for
4583 * config/arm/coff.h (JUMP_TABLES_IN_TEXT_SECTION): Ditto.
4584 * config/arm/aout.h (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
4585 (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump tables.
4586 * config/arm/aof.h (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump
4588 (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
4589 * config/arm/ieee754-df.S: Use macros for Thumb-2/Unified asm
4591 * config/arm/ieee754-sf.S: Ditto.
4592 * config/arm/arm.c (thumb_base_register_rtx_p): Rename...
4593 (thumb1_base_register_rtx_p): ... to this.
4594 (thumb_index_register_rtx_p): Rename...
4595 (thumb1_index_register_rtx_p): ... to this.
4596 (thumb_output_function_prologue): Rename...
4597 (thumb1_output_function_prologue): ... to this.
4598 (thumb_legitimate_address_p): Rename...
4599 (thumb1_legitimate_address_p): ... to this.
4600 (thumb_rtx_costs): Rename...
4601 (thumb1_rtx_costs): ... to this.
4602 (thumb_compute_save_reg_mask): Rename...
4603 (thumb1_compute_save_reg_mask): ... to this.
4604 (thumb_final_prescan_insn): Rename...
4605 (thumb1_final_prescan_insn): ... to this.
4606 (thumb_expand_epilogue): Rename...
4607 (thumb1_expand_epilogue): ... to this.
4608 (arm_unwind_emit_stm): Rename...
4609 (arm_unwind_emit_sequence): ... to this.
4610 (thumb2_legitimate_index_p, thumb2_legitimate_address_p,
4611 thumb1_compute_save_reg_mask, arm_dwarf_handle_frame_unspec,
4612 thumb2_index_mul_operand, output_move_vfp, arm_shift_nmem,
4613 arm_save_coproc_regs, thumb_set_frame_pointer, arm_print_condition,
4614 thumb2_final_prescan_insn, thumb2_asm_output_opcode, arm_output_shift,
4615 thumb2_output_casesi): New functions.
4616 (TARGET_DWARF_HANDLE_FRAME_UNSPEC): Define.
4617 (FL_THUMB2, FL_NOTM, FL_DIV, FL_FOR_ARCH6T2, FL_FOR_ARCH7,
4618 FL_FOR_ARCH7A, FL_FOR_ARCH7R, FL_FOR_ARCH7M, ARM_LSL_NAME,
4619 THUMB2_WORK_REGS): Define.
4620 (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv, arm_condexec_count,
4621 arm_condexec_mask, arm_condexec_masklen)): New variables.
4622 (all_architectures): Add armv6t2, armv7, armv7a, armv7r and armv7m.
4623 (arm_override_options): Check new CPU capabilities.
4624 Set new architecture flag variables.
4625 (arm_isr_value): Handle v7m interrupt functions.
4626 (user_return_insn): Return 0 for v7m interrupt functions. Handle
4628 (const_ok_for_arm): Handle Thumb-2 constants.
4629 (arm_gen_constant): Ditto. Use movw when available.
4630 (arm_function_ok_for_sibcall): Return false for v7m interrupt
4632 (legitimize_pic_address, arm_call_tls_get_addr): Handle Thumb-2.
4633 (thumb_find_work_register, arm_load_pic_register,
4634 legitimize_tls_address, arm_address_cost, load_multiple_sequence,
4635 emit_ldm_seq, emit_stm_seq, arm_select_cc_mode, get_jump_table_size,
4636 print_multi_reg, output_mov_long_double_fpa_from_arm,
4637 output_mov_long_double_arm_from_fpa, output_mov_double_fpa_from_arm,
4638 output_mov_double_fpa_from_arm, output_move_double,
4639 arm_compute_save_reg_mask, arm_compute_save_reg0_reg12_mask,
4640 output_return_instruction, arm_output_function_prologue,
4641 arm_output_epilogue, arm_get_frame_offsets, arm_regno_class,
4642 arm_output_mi_thunk, thumb_set_return_address): Ditto.
4643 (arm_expand_prologue): Handle Thumb-2. Use arm_save_coproc_regs.
4644 (arm_coproc_mem_operand): Allow POST_INC/PRE_DEC.
4645 (arithmetic_instr, shift_op): Use arm_shift_nmem.
4646 (arm_print_operand): Use arm_print_condition. Handle '(', ')', '.',
4648 (arm_final_prescan_insn): Use extract_constrain_insn_cached.
4649 (thumb_expand_prologue): Use thumb_set_frame_pointer.
4650 (arm_file_start): Output directive for unified syntax.
4651 (arm_unwind_emit_set): Handle stack alignment instruction.
4652 * config/arm/lib1funcs.asm: Remove default for __ARM_ARCH__.
4653 Add v6t2, v7, v7a, v7r and v7m.
4654 (RETLDM): Add Thumb-2 code.
4655 (do_it, shift1, do_push, do_pop, COND, THUMB_SYNTAX): New macros.
4656 * config/arm/arm.h (TARGET_CPU_CPP_BUILTINS): Define __thumb2__.
4657 (TARGET_THUMB1, TARGET_32BIT, TARGET_THUMB2, TARGET_DSP_MULTIPLY,
4658 TARGET_INT_SIMD, TARGET_UNIFIED_ASM, ARM_FT_STACKALIGN, IS_STACKALIGN,
4659 THUMB2_TRAMPOLINE_TEMPLATE, TRAMPOLINE_ADJUST_ADDRESS,
4660 ASM_OUTPUT_OPCODE, THUMB2_GO_IF_LEGITIMATE_ADDRESS,
4661 THUMB2_LEGITIMIZE_ADDRESS, CASE_VECTOR_PC_RELATIVE,
4662 CASE_VECTOR_SHORTEN_MODE, ADDR_VEC_ALIGN, ASM_OUTPUT_CASE_END,
4663 ADJUST_INSN_LENGTH): Define.
4664 (TARGET_REALLY_IWMMXT, TARGET_IWMMXT_ABI, CONDITIONAL_REGISTER_USAGE,
4665 STATIC_CHAIN_REGNUM, HARD_REGNO_NREGS, INDEX_REG_CLASS,
4666 BASE_REG_CLASS, MODE_BASE_REG_CLASS, SMALL_REGISTER_CLASSES,
4667 PREFERRED_RELOAD_CLASS, SECONDARY_OUTPUT_RELOAD_CLASS,
4668 SECONDARY_INPUT_RELOAD_CLASS, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P,
4669 TRAMPOLINE_SIZE, INITIALIZE_TRAMPOLINE, HAVE_PRE_INCREMENT,
4670 HAVE_POST_DECREMENT, HAVE_PRE_DECREMENT, HAVE_PRE_MODIFY_DISP,
4671 HAVE_POST_MODIFY_DISP, HAVE_PRE_MODIFY_REG, HAVE_POST_MODIFY_REG,
4672 REGNO_MODE_OK_FOR_BASE_P, LEGITIMATE_CONSTANT_P,
4673 REG_MODE_OK_FOR_BASE_P, REG_OK_FOR_INDEX_P, GO_IF_LEGITIMATE_ADDRESS,
4674 LEGITIMIZE_ADDRESS, THUMB2_LEGITIMIZE_ADDRESS,
4675 GO_IF_MODE_DEPENDENT_ADDRESS, MEMORY_MOVE_COST, BRANCH_COST,
4676 ASM_APP_OFF, ASM_OUTPUT_CASE_LABEL, ARM_DECLARE_FUNCTION_NAME,
4677 FINAL_PRESCAN_INSN, PRINT_OPERAND_PUNCT_VALID_P,
4678 PRINT_OPERAND_ADDRESS): Adjust for Thumb-2.
4679 (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv): New declarations.
4680 * config/arm/arm-cores.def: Add arm1156t2-s, cortex-a8, cortex-r4 and
4682 * config/arm/arm-tune.md: Regenerate.
4683 * config/arm/arm-protos.h: Update prototypes.
4684 * config/arm/vfp.md: Enable patterns for Thumb-2.
4685 (arm_movsi_vfp): Add movw alternative. Use output_move_vfp.
4686 (arm_movdi_vfp, movsf_vfp, movdf_vfp): Use output_move_vfp.
4687 (thumb2_movsi_vfp, thumb2_movdi_vfp, thumb2_movsf_vfp,
4688 thumb2_movdf_vfp, thumb2_movsfcc_vfp, thumb2_movdfcc_vfp): New.
4689 * config/arm/libunwind.S: Add Thumb-2 code.
4690 * config/arm/constraints.md: Update include Thumb-2.
4691 * config/arm/ieee754-sf.S: Add Thumb-2/Unified asm support.
4692 * config/arm/ieee754-df.S: Ditto.
4693 * config/arm/bpabi.S: Ditto.
4694 * config/arm/t-arm (MD_INCLUDES): Add thumb2.md.
4695 * config/arm/predicates.md (low_register_operand,
4696 low_reg_or_int_operand, thumb_16bit_operator): New.
4697 (thumb_cmp_operand, thumb_cmpneg_operand): Rename...
4698 (thumb1_cmp_operand, thumb1_cmpneg_operand): ... to this.
4699 * config/arm/t-arm-elf: Add armv7 multilib.
4700 * config/arm/arm.md: Update patterns for Thumb-2 and Unified asm.
4702 (UNSPEC_STACK_ALIGN, ce_count): New.
4703 (arm_incscc, arm_decscc, arm_umaxsi3, arm_uminsi3,
4704 arm_zero_extendsidi2, arm_zero_extendqidi2): New
4706 * config/arm/fpa.md: Update patterns for Thumb-2 and Unified asm.
4707 (thumb2_movsf_fpa, thumb2_movdf_fpa, thumb2_movxf_fpa,
4708 thumb2_movsfcc_fpa, thumb2_movdfcc_fpa): New insns.
4709 * config/arm/cirrus.md: Update patterns for Thumb-2 and Unified asm.
4710 (cirrus_thumb2_movdi, cirrus_thumb2_movsi_insn,
4711 thumb2_cirrus_movsf_hard_insn, thumb2_cirrus_movdf_hard_insn): New
4713 * doc/extend.texi: Document ARMv7-M interrupt functions.
4714 * doc/invoke.texi: Document Thumb-2 new cores+architectures.
4716 2007-01-03 Jakub Jelinek <jakub@redhat.com>
4718 * unwind-dw2.c (SIGNAL_FRAME_BIT, EXTENDED_CONTEXT_BIT): Define.
4719 (struct _Unwind_Context): Rename args_size to flags, remove
4720 signal_frame field, add a new args_size field and version field.
4721 (_Unwind_IsSignalFrame, _Unwind_SetSignalFrame,
4722 _Unwind_IsExtendedContext): New inline functions.
4723 (_Unwind_GetGR, _Unwind_SetGR, _Unwind_GetGRPtr, _Unwind_SetGRPtr):
4724 Assume by_value array is only present if _Unwind_IsExtendedContext.
4725 (_Unwind_GetIPInfo, execute_cfa_program, uw_frame_state_for): Use
4726 _Unwind_IsSignalFrame.
4727 (__frame_state_for): Initialize context.flags to EXTENDED_CONTEXT_BIT.
4728 (uw_update_context_1): Use _Unwind_SetSignalFrame.
4729 (uw_init_context_1): Initialize context->flags to
4730 EXTENDED_CONTEXT_BIT.
4731 * config/rs6000/linux-unwind.h (frob_update_context): Use
4732 _Unwind_SetSignalFrame.
4734 2007-01-03 Andrew Pinski <andrew_pinski@playstation.sony.com>
4737 * gimplify.c (gimplify_modify_expr_complex_part): Move below
4738 tree_to_gimple_tuple. Call tree_to_gimple_tuple when we need
4741 2007-01-03 Kazu Hirata <kazu@codesourcery.com>
4743 * config/i386/i386.h (NON_STACK_REG_P, REGNO_OK_FOR_SIREG_P,
4744 REGNO_OK_FOR_DIREG_P, REWRITE_ADDRESS, ASM_OPERAND_LETTER,
4745 RET, AT_SP): Remove.
4746 * config/i386/i386.md (*sse_prologue_save_insn): Use return
4749 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
4750 * cfgbuild.c (find_basic_blocks): Likewise.
4751 * cfgrtl.c (rtl_create_basic_block): Likewise.
4752 * function.c (temp_slots_at_level): Likewise.
4753 * reg-stack.c (stack_regs_mentioned): Likewise.
4754 * regclass.c (allocate_reg_info): Likewise.
4755 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
4756 set_bb_for_stmt, move_block_to_fn): Likewise.
4757 * tree-complex.c (tree_lower_complex): Likewise.
4758 * vec.h (VEC_safe_grow_cleared): New.
4760 * cgraphunit.c, tree-ssa-alias.c: Fix comment typos.
4762 2007-01-03 Zdenek Dvorak <dvorakz@suse.cz>
4764 * loop-unswitch.c (unswitch_loop): Pass probabilities to loopify.
4765 * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Pass probabilities
4767 * cfgloopmanip.c (scale_loop_frequencies): Export.
4768 (loopify): Scale the frequencies by prescribed coefficients.
4769 (set_zero_probability): New function.
4770 (duplicate_loop_to_header_edge): Improve updating of frequencies.
4771 (lv_adjust_loop_entry_edge, loop_version): Set probabilities
4772 and frequencies according to arguments.
4773 * tree-ssa-loop-manip.c (tree_unroll_loop): Set probabilities
4775 * cfg.c (scale_bbs_frequencies_int): Allow scaling the frequencies up.
4776 * modulo-sched.c (sms_schedule): Set probabilities for entering
4777 versioned loop correctly.
4778 * tree-vect-transform.c (vect_transform_loop): Ditto.
4779 * cfgloop.h (loopify, loop_version): Declaration changed.
4780 (scale_loop_frequencies): Declared.
4782 2007-01-02 Jan Hubicka <jh@suse.cz>
4784 * cgraph.c: Include tree-flow.h
4785 (cgraph_add_new-function): Handle IPA_SSA mode; execute
4787 * cgraph.h (enum cgraph_state): Add CGRAPH_STATE_IPA_SSA.
4788 * tree-pass.h (pass_all_early_optimizations): Declare.
4789 * cgraphunit.c (cgraph_process_new_functions): Add IPA_SSA; execute
4791 (cgraph_analyze_function): Do early_local_passes.
4792 * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
4793 Do not add referenced vars.
4794 * tree-optimize.c (gate_all_optimizations): Do not execute when not in
4796 (gate_all_early_local_passes): New gate.
4797 (pass_early_local_passes): Use new gate.
4798 (execute_early_local_optimizations): New functions.
4799 (gate_all_early_optimizations): New gate.
4800 (pass_all_early_optimizations): New pass.
4801 (execute_free_datastructures): Free SSA only when initialized.
4802 (gate_init_datastructures): Init only when optimizing.
4803 (tree_lowering_passes): Do early local passes when called late.
4804 * tree-profile.c (do_tree_profiling): Don't profile functions added
4806 (do_early_tree_profiling, pass_early_tree_profile): Kill.
4807 * tree-cfg.c (update_modified_stmts): Do not update when operands are
4809 * passes.c (init_optimizations_passes): Reorder so we go into SSA
4810 during early_local_passes.
4811 * Makefile.in (cgraph.o): Add dependency on tree-flow.h.
4814 2007-01-02 Carlos O'Donell <carlos@codesourcery.com>
4816 * Makefile.in: Update copyright year.
4818 2007-01-02 Carlos O'Donell <carlos@codesourcery.com>
4820 * Makefile.in: Export GCC_EXEC_PREFIX before calling $(RUNTEST)
4821 in $(lang_checks) and check-consistency targets.
4823 2007-01-02 Jan Hubicka <jh@suse.cz>
4825 * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
4826 Do not add referenced vars.
4827 * tree-cfg.c (update_modified_stmts): Do not update when SSA operands
4829 * passes.c (init_optimization_passes): Put mudflap_2 after
4830 free_datastructures.
4832 2007-01-02 Jan Hubicka <jh@suse.cz>
4834 * tree-optimize (execute_fixup_cfg): Set after_inlining flag.
4835 Set NOTHROW flag on call statements proved to be nothrow.
4836 Update statement of local calls so new pure/const functions are
4837 updated. Update_ssa when in ssa form. Mark PHI nodes of nonlocal
4839 (tree_rest_of_compilation): Register hooks and initialize bitmap
4840 early. Do not set after_inlining flag.
4842 2007-01-02 Steve Ellcey <sje@cup.hp.com>
4844 * sbitmap.c (HOST_BITS_PER_LONG_LONG): Change to
4845 HOST_BITS_PER_LONGLONG
4847 2007-01-02 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
4850 * c-typeck.c (store_init_value): Don't emit pedantic overflow
4851 warning for non-static initializers.
4853 2007-01-02 Steven Bosscher <steven@gcc.gnu.org>
4855 * config/alpha/alpha.md, arm/arm.c, darwin.c, frv/frv.md,
4856 m32r/m32r.c, m32r/m32r.c, mn10300/mn10300.md, pa/pa.c,
4857 rs6000/rs6000.c, s390/s390.md, sh/sh.md, sparc/sparc.c:
4858 Always use set_unique_reg_note to add REG_EQUAL notes.
4860 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
4863 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
4865 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
4866 * cfgbuild.c (find_basic_blocks): Likewise.
4867 * cfgrtl.c (rtl_create_basic_block): Likewise.
4868 * function.c (temp_slots_at_level): Likewise.
4869 * reg-stack.c (stack_regs_mentioned): Likewise.
4870 * regclass.c (allocate_reg_info): Likewise.
4871 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
4872 set_bb_for_stmt, move_block_to_fn): Likewise.
4873 * tree-complex.c (tree_lower_complex): Likewise.
4874 * vec.h (VEC_safe_grow_cleared): New.
4876 2007-01-02 Ian Lance Taylor <iant@google.com>
4878 * c-common.c (c_common_truthvalue_conversion): When warning about
4879 using an assignment as a truth value, set TREE_NO_WARNING.
4881 2007-01-02 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
4884 * c.opt (Wold-style-declaration): New.
4885 * doc/invoke.texi (C-only Warning Options): New.
4886 (Wold-style-declaration): Document it.
4887 (Wextra): Enabled by -Wextra.
4888 * c-opts.c (c_common_post_options): Enabled by -Wextra.
4889 * c-decl.c (declspecs_add_scspec): Replace -Wextra with
4890 -Wold-style-declaration.
4892 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
4894 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
4895 * cfgbuild.c (find_basic_blocks): Likewise.
4896 * cfgrtl.c (rtl_create_basic_block): Likewise.
4897 * function.c (temp_slots_at_level): Likewise.
4898 * reg-stack.c (stack_regs_mentioned): Likewise.
4899 * regclass.c (allocate_reg_info): Likewise.
4900 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
4901 set_bb_for_stmt, move_block_to_fn): Likewise.
4902 * tree-complex.c (tree_lower_complex): Likewise.
4903 * vec.h (VEC_safe_grow_cleared): New.
4905 2007-01-02 Douglas Gregor <doug.gregor@gmail.com>
4907 * c-common.c (c_common_nodes_and_builtins): Since variants of
4908 void_type_node get built before it is given a name, we need to
4909 give those variants the name, too.
4910 (complete_array_type): We need to work with the canonical main
4911 type of the array, from which we will build the qualified version.
4912 * params.def (PARAM_VERIFY_CANONICAL_TYPES): New.
4913 * print-tree.c (print_node): Display canonical type information
4915 * stor-layout.c (layout_type): When we don't know the
4916 alignment of a type for which we're building an array, we end up
4917 guessing wrong, so make the type require structural equality.
4918 * tree.c (make_node_stat): When we build a new type, it is its
4920 (build_type_attribute_qual_variant): When building an attribute
4921 variant, its canonical type is the non-attribute variant. However,
4922 if the attributes are target-dependent and they differ, we need to
4923 use structural equality checks for this type.
4924 (build_qualified_type): A qualified type is not equivalent to its
4925 unqualified variant; set the canonical type appropriately.
4926 (build_distinct_type_copy): When building a distinct type from
4927 another type, the new type is its own canonical type.
4928 (build_variant_type_copy): When building a new type variant, we
4929 assume that it is equivalent to the original type.
4930 (build_pointer_type_for_mode): When building a pointer type, also
4931 build a canonical type pointer.
4932 (build_reference_type_for_mode): When building a reference type,
4933 also build a canonical type reference.
4934 (build_index_type): When we can't hash an index type (e.g.,
4935 because its maximum value is negative), the index type requires
4936 structural equality tests.
4937 (build_array_type): Build the canonical form of an array type.
4938 (build_function_type): Function types require structural equality,
4939 because they contain default arguments, attributes, etc.
4940 (build_method_type_directly): Ditto for method types.
4941 (build_offset_type): Build the canonical offset type.
4942 (build_complex_type): Build the canonical vector type.
4943 (make_vector_type): Build the canonical vector type.
4944 * tree.h (TYPE_CANONICAL): New.
4945 (TYPE_STRUCTURAL_EQUALITY_P): New.
4946 (SET_TYPE_STRUCTURAL_EQUALITY): New.
4947 (struct tree_type): Added "canonical" field.
4948 * params.h (VERIFY_CANONICAL_TYPES): New.
4949 * doc/c-tree.texi (TYPE_CANONICAL): Document.
4950 (TYPE_STRUCTURAL_EQUALITY_P): Document.
4951 (SET_TYPE_STRUCTURAL_EQUALITY): Document.
4952 * doc/invoke.texi (verify-canonical-types): Document --param
4953 parameter for verifying canonical types.
4955 2007-01-02 Joseph Myers <joseph@codesourcery.com>
4957 * config.gcc (powerpc-*-eabispe*, powerpc-*-eabisimaltivec*,
4958 powerpc-*-eabisim*, powerpc-*-eabialtivec*, powerpc-*-eabi*,
4959 powerpc-*-rtems*, powerpc-wrs-vxworks, powerpc-wrs-vxworksae,
4960 powerpcle-*-eabisim*, powerpcle-*-eabi*): Add rs6000/e500.h to
4962 * config/rs6000/e500.h: New.
4963 * config/rs6000/eabi.h (TARGET_SPE_ABI, TARGET_SPE, TARGET_E500,
4964 TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE, TARGET_E500_DOUBLE):
4966 * config/rs6000/linuxspe.h (TARGET_SPE_ABI, TARGET_SPE,
4967 TARGET_E500, TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE,
4968 TARGET_E500_DOUBLE): Remove.
4969 * config/rs6000/vxworks.h (TARGET_SPE_ABI, TARGET_SPE,
4970 TARGET_E500, TARGET_ISEL, TARGET_FPRS): Remove.
4971 * config/rs6000/rs6000.h (CHECK_E500_OPTIONS): Define.
4972 * config/rs6000/rs6000.c (rs6000_override_options): Use
4975 2007-01-02 Joseph Myers <joseph@codesourcery.com>
4977 * config/rs6000/rs6000.c (print_operand): Check (TARGET_SPE ||
4978 TARGET_E500_DOUBLE), not TARGET_E500, for %y.
4979 (rs6000_generate_compare, rs6000_emit_sCOND, output_cbranch,
4980 rs6000_emit_cmove): Don't check TARGET_E500.
4981 * config/rs6000/rs6000.md (bunordered, bordered, sunordered,
4982 sordered): Don't check TARGET_E500.
4984 2007-01-01 Eric Christopher <echristo@apple.com>
4986 * config/mips/mips.c (mips_regno_mode_ok_for_base_p): Use
4987 HARD_REGISTER_NUM_P.
4989 2007-01-01 Roger Sayle <roger@eyesopen.com>
4991 * fold-const.c (fold_binary) <EQ_EXPR>: Fold "(X^C1) eq/ne C2" into
4992 "X eq/ne (C1^C2)". Fold "(X^Z) eq/ne (Y^Z)" as "X eq/ne Y" when Z
4993 has no side-effects. Fold "(X^C1) eq/ne (Y^C2)" as "(X^(C1^C2))
4996 2007-01-01 Mike Stump <mrs@apple.com>
4998 * configure.ac: Remove support for building with Apple's gcc-3.1.
5000 2007-01-02 Joseph Myers <joseph@codesourcery.com>
5003 * caller-save.c (add_stored_regs): Only handle SUBREGs if inner
5004 REG is a hard register. Do not modify REG before calling
5006 * rtlanal.c (subreg_get_info): Don't assert size of XMODE is a
5007 multiple of the size of YMODE for certain lowpart cases.
5009 2007-01-01 Andrew Pinski <pinskia@gmail.com>
5012 * gimplify (voidify_wrapper_expr): Update for
5015 2007-01-01 Andreas Schwab <schwab@suse.de>
5018 * config/ia64/ia64.c (ia64_compute_frame_size): Account space for
5019 save of BR0 in extra_spill_size instead of spill_size.
5020 (ia64_expand_prologue): Save BR0 outside of the gr/br/fr spill
5022 (ia64_expand_epilogue): Restore BR0 from its new location.
5024 2007-01-01 Andrew Pinski <pinskia@gmail.com>
5026 * gimplify.c (gimplify_init_constructor <case VECTOR_TYPE>):
5027 Use a temporary variable if the left hand side is not a gimple
5030 2007-01-01 Andrew Pinski <pinskia@gmail.com>
5032 * gimplify.c (gimplify_return_expr): Make the temporary variable
5033 for the return expression, a gimple register variable.
5035 2007-01-01 Jan Hubicka <jh@suse.cz>
5037 * emit-rtl.c (emit_copy_of_insn_after): Do not call copy_insn_1 for
5040 2007-01-01 Mike Stump <mrs@apple.com>
5042 * configure.ac (HAVE_GAS_LITERAL16): Add autoconf check for
5044 * config/darwin.c (machopic_select_rtx_section): Use
5046 (darwin_mergeable_constant_section): Likewise.
5047 * configure: Regenerate.
5048 * config.in: Regenerate.
5050 2007-01-01 Jan Hubicka <jh@suse.cz>
5051 Andrew Pinski <pinskia@gmail.com>
5053 * cgraphunit.c (cgraph_optimize): Call cgraph_add_new_functions
5054 before starting IPA passes.