1 2007-01-31 Richard Henderson <rth@redhat.com>
2 Ian Lance Taylor <iant@google.com>
4 * lower-subreg.c: New file.
5 * rtl.def (CONCATN): Define.
6 * passes.c (init_optimization_passes): Add pass_lower_subreg and
8 * emit-rtl.c (update_reg_offset): New static function, broken out
10 (gen_rtx_REG_offset): Call update_reg_offset.
11 (gen_reg_rtx_offset): New function.
12 * regclass.c: Revert patch of 2006-03-05, restoring
14 (clear_reg_info_regno): New function.
15 * dwarf2out.c (concatn_loc_descriptor): New static function.
16 (loc_descriptor): Handle CONCATN.
17 * common.opt (fsplit_wide_types): New option.
18 * opts.c (decode_options): Set flag_split_wide_types when
20 * timevar.def (TV_LOWER_SUBREG): Define.
21 * rtl.h (gen_reg_rtx_offset): Declare.
22 (reg_scan_update): Declare.
23 * regs.h (clear_reg_info_regno): Declare.
24 * tree-pass.h (pass_lower_subreg): Declare.
25 (pass_lower_subreg2): Declare.
26 * doc/invoke.texi (Option Summary): List -fno-split-wide-types.
27 (Optimize Options): Add -fsplit-wide-types to -O1 list. Document
29 * doc/rtl.texi (Regs and Memory): Document concat and concatn.
30 * Makefile.in (OBJS-common): Add lower-subreg.o.
31 (lower-subreg.o): New target.
33 2007-01-31 Kazu Hirata <kazu@codesourcery.com>
35 * config/sh/sh.h (HAVE_SECONDARY_RELOADS): Remove.
37 2007-01-31 Anatoly Sokolov <aesok@post.ru>
40 * config/avr/avr.c (DWARF2_ADDR_SIZE): Define.
42 2007-01-31 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
45 * builtins.c (fold_builtin_sqrt): Use MPFR for constant args.
47 2007-01-31 Zdenek Dvorak <dvorakz@suse.cz>
49 * cfgloop.h: Include vec-prim.h.
50 (enum li_flags): Remove LI_ONLY_OLD.
51 (loop_iterator): Changed.
52 (fel_next, fel_init): Iterate over loop tree.
53 (FOR_EACH_LOOP_BREAK): New macro.
54 * loop-unswitch.c (unswitch_loops): Do not pass LI_ONLY_OLD to
56 * tree-ssa-loop-unswitch.c (tree_ssa_unswitch_loops): Ditto.
57 * modulo-sched.c (sms_schedule): Ditto.
58 * tree-vectorizer.c (vectorize_loops): Ditto.
59 * doc/loop.texi: Update information on loop numbering and behavior of
60 FOR_EACH_LOOP wrto new loops.
61 * tree-scalar-evolution.c (compute_overall_effect_of_inner_loop,
62 add_to_evolution_1): Test nestedness of loops instead of comparing
64 * tree-chrec.c (chrec_fold_plus_poly_poly,
65 chrec_fold_multiply_poly_poly, chrec_evaluate,
66 hide_evolution_in_other_loops_than_loop, chrec_component_in_loop_num,
67 reset_evolution_in_loop): Ditto.
68 * Makefile.in (CFGLOOP_H): Add vecprim.h dependency.
70 2007-01-31 Dirk Mueller <dmueller@suse.de>
72 * c-common.c (warn_about_parentheses): Separate warning about
73 un-parenthized sequence of comparison operators from the one
74 which is supposed to warn about x <= y <= z.
76 2007-01-31 Uros Bizjak <ubizjak@gmail.com>
78 * optabs.h (enum optab_index): Add new OTI_isinf.
79 (isinf_optab): Define corresponding macro.
80 * optabs.c (init_optabs): Initialize isinf_optab.
81 * genopinit.c (optabs): Implement isinf_optab using isinf?f2
83 * builtins.c (mathfn_built_in): Handle BUILT_IN_ISINF{,F,L}.
84 (expand_builtin_interclass_mathfn): Expand BUILT_IN_ISINF{,F,L}
86 (expand_builtin): Expand BUILT_IN_ISINF{,F,L} using
87 expand_builtin_interclass_mathfn.
88 * reg_stack.c (subst_stack_regs_pat): Handle UNSPEC_FXAM.
89 * config/i386/i386.md (UNSPEC_FXAM): New constant.
90 (fxam<mode>2_i387): New insn pattern.
91 (isinf<mode>2) New expander to implement isinf, isinff and isinfl
92 built-in functions as x87 inline asm.
94 2007-01-31 Kazu Hirata <kazu@codesourcery.com>
96 * gcc/config/arm/unwind-arm.h (_sleb128_t, _uleb128_t): New.
98 2007-01-30 Eric Christopher <echristo@apple.com>
100 * config.gcc: Add geode.
102 2007-01-31 Kazu Hirata <kazu@codesourcery.com>
104 * cgraphunit.c, config/arm/arm.c, config/m68k/m68k.c,
105 ipa-inline.c, tree-profile.c, tree-ssa-live.c,
106 tree-ssa-math-opts.c, tree-ssanames.c, tree-vect-analyze.c,
107 value-prof.c: Fix comment typos.
109 2007-01-30 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
112 * doc/invoke.texi (Wpointer-arith): Document warning.
114 2007-01-30 Janis Johnson
116 * doc/extend.texi (Decimal Floating Types): Remove decfloat.h from
117 the list of discrepancies from the draft TR.
119 2007-01-30 Dirk Mueller <dmueller@suse.de>
122 * doc/invoke.texi (-Wreturn-type): Update description to
125 2007-01-30 Richard Sandiford <richard@codesourcery.com>
127 * cfgrtl.c (try_redirect_by_replacing_jump): Check only_sets_cc0_p.
129 2007-01-30 Uros Bizjak <ubizjak@gmail.com>
131 * builtins.c (expand_builtin_int_interclass_roundingfn): New function
132 to handle optabs that operate on floating point input argument and
133 output to integer output.
134 (expand_builtin_mathfn) [BUILT_IN_ILOGB]: Move from here ...
135 (expand_builtin_interclass_mathfn) [BUILT_IN_ILOGB]: ... to here.
136 (expand_builtin): Expand BUILT_IN_ILOGB{,F,L} using
137 expand_builtin_interclass_mathfn ().
138 * config/i386/i386.md (fxtractxf3_i387): Rename from *fxtractxf3_i387.
140 (ilogbxf2, ilogb<mode>2): New expanders to implement ilogb, ilogbf and
141 ilogbl built-in functions as x87 intrinsics.
143 2007-01-30 Richard Guenther <rguenther@suse.de>
146 * dwarf2out.c (reference_to_unused): Query varpool if the
149 2007-01-30 Richard Guenther <rguenther@suse.de>
152 * passes.c (execute_one_pass): Reset in_gimple_form to not
153 confuse non-unit-at-a-time mode.
155 2007-01-29 Roger Sayle <roger@eyesopen.com>
156 Richard Guenther <rguenther@suse.de>
158 * fold-const.c (round_up): Make HIGH an unsigned HOST_WIDE_INT to
159 avoid undefined behaviour on overflow. Use force_fit_type_double
160 to construct the constant with the specified TREE_OVERFLOW.
162 2007-01-29 Janis Johnson <janis187@us.ibm.com>
164 * config/dfp-bit.c: Add parameterized support for fp exceptions.
165 * config/dfp-bit.h: Ditto.
167 2007-01-29 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
169 * c-decl.c (pop_scope): Replace warnings with call to
170 warn_for_unused_label.
171 * c-common.h (warn_for_unused_label): Declare.
172 * c-common.c (warn_for_unused_label): Define.
174 2007-01-29 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
176 * tree-optimize.c (update_inlined_to_pointers): Delete unused
179 2007-01-29 Janis Johnson <janis187@us.ibm.com>
181 * Makefile.in (USER_H): Remove decfloat.h.
182 * ginclude/decfloat.h: Delete, moving contents to ...
183 * ginclude/float.h: Add support for decimal floating point,
184 guarded by __STDC_WANT_DEC_FP__.
186 2007-01-29 Mike Stump <mrs@apple.com>
188 * doc/gccint.texi (Top): Rename Loop Representation to Loop
189 Analysis and Representation to resolve case insensitive conflict.
190 * doc/loop.texi (Loop Analysis and Representation): Likewise.
192 2007-01-28 Daniel Berlin <dberlin@dberlin.org>
194 * tree.h (struct tree_memory_tag): Add aliases member.
195 (MTAG_ALIASES): New macro.
196 * tree-ssa-alias.c (alias_bitmap_obstack): New variable.
197 (add_may_alias): Remove pointer-set. Update for may_aliases being
199 (mark_aliases_call_clobbered): Update for may_aliases being a
201 (compute_tag_properties): Ditto.
202 (create_partition_for): Ditto.
203 (compute_memory_partitions): Ditto.
204 (dump_may_aliases_for): Ditto.
205 (is_aliased_with): Ditto.
206 (add_may_alias_for_new_tag): Ditto.
207 (rewrite_alias_set_for): Rewrite for may_aliases being a bitmap.
208 (compute_is_aliased): New function.
209 (compute_may_aliases): Call compute_is_aliased).
210 (init_alias_info): Initialize alias_bitmap_obstack.
211 (union_alias_set_into): New function.
212 (compute_flow_sensitive_aliasing): Use union_aliases_into.
213 (have_common_aliases_p): Rewrite to take two bitmaps and use
215 (compute_flow_insensitive_aliasing): Stop using pointer-sets.
217 (finalize_ref_all_pointers): Update for add_may_alias changes.
218 (new_type_alias): Ditto.
219 * tree-flow-inline.h (may_aliases): Return a bitmap.
220 * tree-dfa.c (dump_variable): Check for MTAG_P'ness.
221 * tree-ssa.c (verify_flow_insensitive_alias_info): Update for
222 may_aliases being a bitmap.
223 * tree-flow.h (struct var_ann_d): Remove may_aliases member.
224 may_aliases now returns a bitmap.
225 * tree-ssa-structalias.c (merge_smts_into): Update for may_aliases
227 * tree-ssa-operands.c (add_virtual_operand): Update for
228 may_aliases being a bitmap.
230 2007-01-29 Daniel Berlin <dberlin@dberlin.org>
232 PR tree-optimization/30630
233 * tree-ssa-structalias.c (do_complex_constraint): Mark correct
236 2007-01-29 Simon Martin <simartin@users.sourceforge.net>
239 * gimplify.c (gimplify_target_expr): Make sure that the TARGET_EXPR is
240 expanded only once even if an error occurs.
242 2007-01-29 Ben Elliston <bje@au.ibm.com>
244 * gcov-io.h (__gcov_indirect_call_profiler): Declare.
245 (__gcov_average_profiler): Likewise.
246 (__gcov_ior_profiler): Likewise.
247 (__gcov_merge_ior): Likewise.
249 2007-01-28 Jan Hubicka <jh@suse.cz>
251 * builtins.c (expand_builtin_memset): Fix typo in my last patch.
252 * value-prof.c (stringop_block_profile): Likewise.
254 2007-01-28 Jan Hubicka <jh@suse.cz>
256 * expr.c (emit_block_move_via_movmem, emit_block_move_via_libcall): Add
257 variant handling histograms; add wrapper.
258 (clear_storage_via_libcall): Export.
259 (emit_block_move_hints): Break out from ...; add histograms.
260 (emit_block_move): ... this one.
261 (clear_storage_hints): Break out from ...; add histograms.
262 (clear_storage): ... this one.
263 (set_storage_via_memset): Handle histogram.
264 * expr.h (emit_block_move_via_libcall, emit_block_move_hints): Declare.
265 (clear_storage_hints, clear_storage_via_libcall): Declare.
266 (set_storage_via_setmem): Update prototype.
267 * doc/md.texi (movmem, setmem): Document new arguments.
269 * value-prof.c (dump_histogram_value, tree_find_values_to_profile): Add
271 (stringop_block_profile): New global function.
272 (tree_stringops_values_to_profile): Profile block size and alignment.
273 * value-prof.h (enum hist_type): add HIST_TYPE_AVERAGE and
275 (struct profile_hooks): Add gen_average_profiler and gen_ior_profiler.
276 (stringop_block_profile): Declare.
277 * builtins.c: Include value-prof.h.
278 (expand_builtin_memcpy, expand_builtin_memset): Pass block profile.
279 * gcov-ui.h (GCOV_COUNTER_NAMES): Add new counter.
280 (GCOV_COUNTER_AVERAGE, GCOV_COUNTER_IOR): New constants.
281 (GCOV_COUNTERS, GCOV_LAST_VALUE_COUNTER): Update.
282 * profile.c (instrument_values): Add new counters.
283 * cfgexpand.c (expand_gimple_basic_block): Propagate histograms to
285 * tree-profile.c (tree_average_profiler_fn, tree_ior_profiler_fn): New.
286 (tree_init_edge_profiler): Build new profilers.
287 (tree_gen_average_profiler, tree_gen_ior_profiler): New.
288 (pass_tree_profile): Add dump.
289 (tree_profile_hooks): Update.
290 * Makefile.in (LIBGCOV): Add new constants.
291 * libgcov.c (__gcov_merge_ior, __gcov_average_profiler,
292 __gcov_ior_profiler): New.
293 * i386.md (movmem/setmem expanders): Add new optional arguments.
295 2007-01-28 David Edelsohn <edelsohn@gnu.org>
297 * doc/md.texi (Standard Pattern Names): Document blockage pattern.
299 2007-01-28 Zdenek Dvorak <dvorakz@suse.cz>
301 * tree-ssa-loop-unswitch.c: Include tree-inline.h.
302 (tree_unswitch_single_loop): Pass eni_size_weights to
304 * tree-ssa-loop-manip.c: Include tree-inline.h.
305 (can_unroll_loop_p): Pass eni_size_weights to
307 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p):
308 Pass eni_size_weights to estimate_num_insns.
309 * tree.h (init_inline_once): Export.
310 * toplev.c (backend_init): Call init_inline_once.
311 * cgraphunit.c (cgraph_process_new_functions,
312 cgraph_analyze_function): Pass eni_inlining_weights to
314 * ipa-inline.c (compute_inline_parameters): Ditto.
315 * tree-ssa-loop-ivcanon.c (tree_num_loop_insns): Pass weights
316 to estimate_num_insns.
317 (try_unroll_loop_completely): Pass eni_size_weights to
319 * tree-eh.c (decide_copy_try_finally): Pass eni_size_weights
320 ot estimate_num_insns.
321 * tree-ssa-loop-prefetch.c: Include tree-inline.h.
322 (loop_prefetch_arrays): Pass eni_time_weights to tree_num_loop_insns.
323 * tree-inline.c (eni_inlining_weights, eni_size_weights,
324 eni_time_weights): New variables.
325 (init_inline_once): Initialize them.
326 (struct eni_data): Mew.
327 (estimate_num_insns_1, estimate_num_insns): Use weights.
328 * tree-inline.h (struct eni_weights_d): New.
329 (eni_inlining_weights, eni_size_weights, eni_time_weights): Declare.
330 (estimate_num_insns): Declaration changed.
331 * cfgloop.h (tree_num_loop_insns): Declaration changed.
332 * Makefile.in (tree-ssa-loop-unswitch.o, tree-ssa-loop-prefetch.o,
333 tree-ssa-loop-manip.o): Add TREE_INLINE_H dependency.
335 2007-01-28 Zdenek Dvorak <dvorakz@suse.cz>
337 * tree-data-ref.c (conflict_fn): Assert that the number of affine
338 relations in the conflict function is valid.
340 2007-01-27 Ian Lance Taylor <iant@google.com>
342 * common.opt: Add fstrict-overflow.
343 * opts.c (decode_options): Set flag_strict_overflow if -O2.
344 * flags.h (TYPE_OVERFLOW_WRAPS): Define.
345 (TYPE_OVERFLOW_UNDEFINED): Define.
346 (TYPE_OVERFLOW_TRAPS): Define. This replaces TYPE_TRAP_SIGNED.
348 * tree.h (TYPE_TRAP_SIGNED): Don't define.
349 * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_UNDEFINED.
350 (fold_negate_expr): Likewise.
351 (make_range): Likewise.
352 (extract_muldiv_1): Likewise.
353 (maybe_canonicalize_comparison): Likewise.
354 (fold_comparison): Likewise.
355 (fold_binary): Likewise.
356 (tree_expr_nonnegative_p): Likewise.
357 (tree_expr_nonzero_p): Likewise.
358 * tree-vrp.c (compare_values): Likewise.
359 (extract_range_from_binary_expr): Likewise.
360 (extract_range_from_unary_expr): Likewise.
361 * tree-ssa-loop-niter.c (infer_loop_bounds_from_signedness):
363 (nowrap_type_p): Likewise.
364 * tree-scalar-evolution.c (simple_iv): Likewise.
365 * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_WRAPS.
366 (build_range_check): Likewise.
367 (extract_muldiv_1): Likewise.
368 (fold_comparison): Likewise.
369 * tree-vrp.c (vrp_int_const_binop): Likewise.
370 (extract_range_from_unary_expr): Likewise.
371 * convert.c (convert_to_integer): Likewise.
372 * fold-const.c (fold_negate_expr): Use TYPE_OVERFLOW_TRAPS.
373 (fold_comparison): Likewise.
374 (fold_binary): Likewise.
375 * optabs.c (optab_for_tree_code): Likewise.
376 * tree-vectorizer.c (vect_is_simple_reduction): Likewise.
377 * simplify-rtx.c (simplify_const_relational_operation): Check
378 flag_strict_overflow and flag_trapv.
379 (simplify_const_relational_operation): Likewise.
380 * doc/invoke.texi (Option Summary): Mention -fstrict-overflow.
381 (Optimize Options): Add -fstrict-overflow to -O2 list. Document
384 2007-01-27 Roger Sayle <roger@eyesopen.com>
386 * tree.c (tree_fold_gcd): Delete.
387 * tree.h (tree_fold_gcd): Remove prototype.
388 * tree-data-ref.c (tree_fold_divides_p): Don't use tree_fold_gcd to
389 test whether one constant integer is a multiple of another. Instead
390 call int_const_binop with TRUNC_MOD_EXPR and test for a zero result.
391 * fold-const.c (multiple_of_p): We've determined both TOP and
392 BOTTOM are integer constants so we can call int_const_binop directly
393 instead of the more generic const_binop.
395 2007-01-27 Roger Sayle <roger@eyesopen.com>
397 * fold-const.c (size_binop): In the fast-paths for X+0, 0+X, X-0 and
398 1*X check that the constant hasn't overflowed, to preserve the
400 (round_up): Provide an efficient implementation when rouding-up an
401 INTEGER_CST to a power-of-two.
403 2007-01-28 Ralf Wildenhues <Ralf.Wildenhues@gmx.de>
405 * doc/sourcebuild.texi: Add comma for clarity.
406 * doc/extend.texi: Fix some typos.
407 * doc/passes.texi: Likewise.
408 * doc/cppinternals.texi: Likewise.
409 * doc/c-tree.texi: Likewise.
410 * doc/tree-ssa.texi: Likewise.
411 * doc/install.texi: Likewise.
413 2007-01-27 Jan Hubicka <jh@suse.cz>
415 * tree-sra.c (sra_walk_function): Don't rely on aliases being build.
416 (pass_sra): Do not require alias information.
417 * passes.c (init_optimization_passes): Add SRA
419 2007-01-27 Steven Bosscher <steven@gcc.gnu.org>
421 * tracer.c (rest_of_handle_tracer): We already cleaned
422 up the CFG in tracer() so don't do it here again.
423 * cfgcleanup.c (rest_of_handle_jump2): Don't repeat
424 cleanup_cfg here, either. And don't call renumber_insns.
426 * cfgrtl.c (rtl_verify_flow_info_1): Don't verify that BB_END
427 and BB_HEAD are in the insn stream here. Instead make sure
428 that BB_INSN is valid on all insns. Also, do check here that
429 there are no pending branch predictions...
430 (rtl_verify_flow_info): ...instead of doing it here. Checks
431 for BB_END and BB_HEAD moved from rtl_verify_flow_info_1 to
434 2007-01-26 Roger Sayle <roger@eyesopen.com>
436 * config/i386/i386.c (ix86_swap_binary_operands_p): New helper
437 function to simplify/factorize operand order canonicalization.
438 (ix86_fixup_binary_operands): Reorganize using the above function.
439 (ix86_binary_operator_ok): Likewise.
441 2007-01-27 Jakub Jelinek <jakub@redhat.com>
443 * genattrtab.c (struct attr_value_list, insn_code_values): Move to
444 file scope from optimize_attrs.
445 (simplify_test_exp): If insn_code_values is not NULL, use it to speed
447 (optimize_attrs): Clear insn_code_values after freeing it.
449 2007-01-26 Zdenek Dvorak <dvorakz@suse.cz>
451 * tree-ssa-address.c (create_mem_ref): Remove ", bsi" from
452 a parts.base assignment.
454 2007-01-26 Zdenek Dvorak <dvorakz@suse.cz>
456 * tree-data-ref.c (dump_subscript): Use dump_conflict_function.
457 (compute_subscript_distance, initialize_data_dependence_relation,
458 finalize_ddr_dependent, analyze_ziv_subscript,
459 analyze_siv_subscript_cst_affine,
460 compute_overlap_steps_for_affine_univar,
461 compute_overlap_steps_for_affine_1_2, analyze_subscript_affine_affine,
462 analyze_siv_subscript, analyze_miv_subscript,
463 analyze_overlapping_iterations, subscript_dependence_tester_1,
464 compute_self_dependence, free_dependence_relation): Work
465 with affine_fn instead of chrecs.
466 (dump_affine_function, dump_conflict_function, affine_function_equal_p,
467 common_affine_function, affine_function_base,
468 affine_function_constant_p, affine_fn_op, affine_fn_plus,
469 affine_fn_minus, affine_fn_free, conflict_fn_not_known,
470 conflict_fn_no_dependence, free_conflict_function, free_subscripts,
471 conflict_fn, affine_fn_cst, affine_fn_univar): New functions.
472 (all_chrecs_equal_p): Removed.
473 * tree-data-ref.h (affine_fn, conflict_function): New types.
474 (struct subscript): Change type of conflicting_iterations_in_a
475 and conflicting_iterations_in_b.
477 2007-01-26 Steve Ellcey <sje@cup.hp.com>
480 * config/pa/pa.h (TARGET_HPUX_11): New.
481 * config/pa/pa-hpux11.h (TARGET_HPUX_11): New.
482 * config/pa/pa.c (pa_init_builtins): Use TARGET_HPUX_11.
484 2007-01-26 Daniel Berlin <dberlin@dberlin.org>
485 Richard Guenther <rguenther@suse.de>
487 * tree-ssa-structalias.c (solve_graph): Handle case
488 we merged the variable to another.
490 2007-01-25 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
492 * builtins.c (fold_builtin_1): Treat ccos and ccosh as 'even'
495 * fold-const.c (negate_mathfn_p): Treat casin, casinh, catan,
496 catanh, cproj, csin, csinh, ctan and ctanh as 'odd' functions.
498 2007-01-25 DJ Delorie <dj@redhat.com>
500 * config/m32c/m32c.c (m32c_cannot_change_mode_class): We don't
501 allow changes to modes which don't fit in those registers.
503 * reload1.c (choose_reload_regs): Check for invalid subregs before
504 computing their locations, not after.
506 2007-01-25 Geoffrey Keating <geoffk@apple.com>
509 * config/rs6000/rs6000.c (first_altivec_reg_to_save): On Darwin,
510 save Altivec registers in an eh_return function.
511 (compute_vrsave_mask): Likewise.
512 (rs6000_stack_info): Correct AIX/Darwin stack alignment computation
513 for saving Altivec registers.
514 (rs6000_emit_prologue): Don't allocate stack twice in
515 eh_return function. Correct expected value of altivec_save_offset
516 when using save_world. Describe save of R0 to stack when using
517 save_world. Describe stack pointer adjustment when using
518 save_world. Remove duplicated eh_return parameter register saving.
519 Update sp_offset variable after save_world.
520 * config/rs6000/t-darwin (LIB2FUNCS_STATIC_EXTRA): Remove
522 (LIB2FUNCS_EXTRA): Add darwin-world.asm.
523 * config/rs6000/darwin.h (SUBTARGET_OVERRIDE_OPTIONS): -m64
526 2007-01-25 Steve Ellcey <sje@cup.hp.com>
528 * config.gcc (ia64*-*-hpux*): Make posix threads the default.
530 2007-01-25 Steve Ellcey <sje@cup.hp.com>
533 * config/pa/pa.c (pa_init_builtins): Set asm names for finite routines.
534 * config/ia64/ia64.c (ia64_init_builtins): Ditto.
536 2007-01-25 Richard Guenther <rguenther@suse.de>
538 * doc/invoke.texi (-Wcoverage-mismatch): Document.
539 * common.opt (-Wcoverage-mismatch): New warning option.
540 * coverage.c (get_coverage_counts): Ignore coverage mismatch
541 if -Wcoverage-mismatch is given.
543 2007-01-25 Razya Ladelsky <razya@il.ibm.com>
545 * ipa-cp.c (ipcp_insert_stage, ipcp_driver): Support for SSA.
546 (ipcp_driver): Change to static definition.
547 Add dumping of the ifunctions.
548 (constant_val_insert): Remove unused parameter. Support for SSA.
549 (ipcp_propagate_const): Support for SSA.
550 (ipcp_profile_bb_print): Print only analyzed nodes.
551 (ipcp_replace_map_create): Remove support for Fortran constant
553 * ipa-prop.c (ipa_method_modify_stmt,
554 ipa_callsite_compute_param): Support for SSA.
555 * ipa-prop.h (ipcp_driver): Remove declaration.
556 (IS_VALID_TREE_MAP_INDEX): Add define.
558 2007-01-24 Geoffrey Keating <geoffk@apple.com>
560 * unwind-dw2.c (execute_stack_op): Handle DW_OP_swap.
562 2007-01-24 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
565 * builtins.c (fold_builtin_cabs): Use MPFR to evaluate a
566 constant argument to cabs and do it without checking for
567 -funsafe-math-optimizations.
569 2007-01-24 Douglas Gregor <dgregor@osl.iu.edu>
571 * c-common.h (RID_FIRST_CXX0X): New.
572 (RID_LAST_CXX0X): New.
573 * c-opts.c (c_common_handle_option): -Wc++0x-compat is triggered
575 * c.opt (Wc++0x-compat): New.
576 * doc/invoke.texi (-Wc++0x-compat): Document.
578 2007-01-24 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
580 * builtins.c (fold_builtin_carg): New.
581 (fold_builtin_1): Use it.
583 2007-01-24 Jan Hubicka <jh@suse.cz>
585 * ipa-inline.c (cgraph_decide_inlining): Initialize initial_insns.
587 * ipa-inline.c (initial_insns, max_insns): Delete.
588 (compute_max_insns): New function.
589 (cgraph_decide_inlining_of_small_function): Use it; take minimal amount
590 of insns as base for code growth.
591 (cgraph_decide_inlining): Make initial_insns local; do not compute
593 * params.def (PARAM_INLINE_UNIT_GROWTH): Set to 60.
594 * doc/invoke.texi (inline-unit-growth): Update docs.
596 2007-01-24 Jakub Jelinek <jakub@redhat.com>
598 * config/i386/i386.h (x86_cmpxchg16b): Remove const.
599 (TARGET_CMPXCHG16B): Define to x86_cmpxchg16b.
600 * config/i386/i386.c (x86_cmpxchg16b): Remove const.
601 (override_options): Add PTA_CX16 flag. Set x86_cmpxchg16b
602 for CPUs that have PTA_CX16 set.
605 * gimplify.c (omp_check_private): New function.
606 (gimplify_scan_omp_clauses): Use it for
607 firstprivate/lastprivate/reduction.
610 * gimplify.c (omp_add_variable): Don't call omp_notice_variable
611 on TYPE_SIZE_UNIT for GOVD_LOCAL VLAs.
614 * omp-low.c (lower_omp_for_lastprivate): Add dlist argument.
615 If lower_lastprivate_clauses emits some statements, append them
616 to dlist rather than body_p and to body_p append an initializer.
617 (lower_omp_for): Adjust caller.
619 2007-01-24 Steve Ellcey <sje@cup.hp.com>
621 * target.h (globalize_decl_name): New.
622 * target-def.h (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
623 * output.h (default_globalize_decl_name): New.
624 * varasm.c (asm_output_bss): Use globalize_decl_name instead of
626 (globalize_decl): Ditto.
627 (default_globalize_decl_name): New.
628 * config/ia64/ia64.c (ia64_globalize_decl_name): New.
629 (ia64_handle_version_id_attribute): New.
630 (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
631 (ia64_asm_output_external): Use globalize_decl_name instead
633 * doc/extend.texi (version_id): New pragma.
634 * doc/tm.texi (ARGET_ASM_GLOBALIZE_DECL_NAME): New target hook.
636 2007-01-24 Andreas Krebbel <krebbel1@de.ibm.com>
638 * unwind-dw2-fde.c (get_cie_encoding): Replaced _Unwind_Word with
639 _uleb128_t and _Unwind_SWord with _sleb128_t.
640 * unwind-dw2.c (extract_cie_info, execute_stack_op, execute_cfa_program,
641 uw_frame_state_for, uw_update_context_1): Likewise.
642 * unwind-c.c (parse_lsda_header, PERSONALITY_FUNCTION): Likewise.
643 * unwind-pe.h (read_uleb128, read_sleb128,
644 read_encoded_value_with_base): Likewise.
645 * unwind-generic.h: Define _sleb128_t and _uleb128_t types.
647 2007-01-24 Richard Guenther <rguenther@suse.de>
649 * builtins.c (expand_builtin_cexpi): Get the fndecl
650 for cexp in the correct way.
652 2007-01-24 Jan Hubicka <jh@suse.cz>
654 * tree-ssa-dce.c (eliminate_unnecesary_stmts): Remove dead LHS of calls.
656 2007-01-24 Andreas Krebbel <krebbel1@de.ibm.com>
658 * c-cppbuiltin.c (builtin_define_type_sizeof): New function.
659 (c_cpp_builtins): New builtin macros: __SIZEOF_INT__, __SIZEOF_LONG__,
660 __SIZEOF_LONG_LONG__, __SIZEOF_SHORT__, __SIZEOF_POINTER__,
661 __SIZEOF_FLOAT__, __SIZEOF_DOUBLE__, __SIZEOF_LONG_DOUBLE__,
662 __SIZEOF_SIZE_T__, __SIZEOF_WCHAR_T__, __SIZEOF_WINT_T__ and
663 __SIZEOF_PTRDIFF_T__.
664 * doc/cpp.texi: Documentation for the new builtin macros added.
666 2007-01-24 Uros Bizjak <ubizjak@gmail.com>
668 * config/i386/i386.md (tanxf2, tan<mode>2, atan<mode>2, log<mode>2,
669 log10<mode>2, log2<mode>2, expxf2, exp10xf2, exp2xf2): Use op2
670 instead of operands[2] to avoid access past the end of array.
672 2007-01-24 Richard Sandiford <richard@codesourcery.com>
674 * reload1.c (emit_reload_insns): Pass the reload register
675 for a non-spill output reload through forget_old_reloads_1.
677 2007-01-23 Joseph Myers <joseph@codesourcery.com>
679 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Define
680 _SOFT_DOUBLE if doubles use software floating-point.
681 * config/rs6000/libgcc-ppc-glibc.ver: Export additional long
682 double functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
683 * config/rs6000/darwin-ldouble.c: Also compile functions for
684 hard-float without FPRs. Use fmsub function for all __NO_FPRS__
685 cases. Compile extra functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
686 * config/rs6000/linuxspe.h (SUBSUBTARGET_OVERRIDE_OPTIONS): Remove
687 commented-out long double override.
688 (CPP_LONGDOUBLE_DEFAULT_SPEC): Likewise.
689 * config/rs6000/eabispe.h: Likewise.
690 * config/rs6000/rs6000.c (rs6000_override_options): Don't override
691 long double for non-SPE.
692 (rs6000_handle_option): Likewise.
693 (invalid_e500_subreg): Disallow more subregs involding DImode,
694 DFmode, TImode or TFmode.
695 (rs6000_legitimate_offset_address_p): Check TFmode offsets for
697 (legitimate_lo_sum_address_p): Also check for TFmode for E500
699 (rs6000_legitimize_address): Also handle TFmode for E500 double.
700 (rs6000_legitimize_reload_address): Also handle TFmode for E500
702 (rs6000_legitimate_address): Also check for TFmode for E500
704 (rs6000_emit_move): Use DFmode subregs of TFmode for E500 double.
705 (spe_build_register_parallel): Handle TFmode and TCmode.
706 (rs6000_spe_function_arg): Handle TFmode and TCmode for E500
708 (function_arg): Handle TFmode and TCmode for E500 double.
709 (rs6000_init_libfuncs): Initialize extra libfuncs for soft double
711 (print_operand): Handle TFmode and TImode for %y.
712 (rs6000_generate_compare): Handle TFmode comparisons for E500
714 (spe_func_has_64bit_regs_p): Check for TFmode for E500 double.
715 (rs6000_function_value): Handle TFmode and TCmode for E500 double.
716 (rs6000_libcall_value): Handle TFmode and TCmode for E500 double.
717 * config/rs6000/rs6000.h (CANNOT_CHANGE_MODE_CLASS): Check for
718 TFmode for E500 double.
719 * config/rs6000/rs6000.md (FP): Allow TF for E500 double.
720 (floatsidf2): Enable for E500 double.
721 (movtf_softfloat): Use rs6000_nonimmediate_operand.
722 (extenddftf2): Change to extenddftf2_fprs.
723 (extenddftf2): Call gen_spe_extenddftf2 or gen_extenddftf2_fprs
724 depending on TARGET_E500_DOUBLE.
725 (extendsftf2): Enable for E500 double.
726 (trunctfdf2): Enable for E500 double.
727 (trunctfsf2): Change to trunctfsf2_fprs.
728 (trunctfsf2): Call gen_spe_trunctfsf2 or gen_trunctfsf2_fprs
729 depending on TARGET_E500_DOUBLE.
730 (floatsitf2): Enable for E500 double.
731 (fix_trunctfsi2): Change to fix_trunctfsi2_fprs.
732 (fix_trunctfsi2): Call gen_spe_fix_trunctfsi2 or
733 gen_fix_trunctfsi2_fprs depending on TARGET_E500_DOUBLE.
734 (negtf2): Change to negtf2_internal.
735 (negtf2): New expander.
736 (abstf2): Enable for E500 double. Call gen_spe_abstf2_tst,
737 gen_spe_abstf2_cmp or gen_abstf2_internal depending on
738 TARGET_E500_DOUBLE and flag_unsafe_math_optimizations.
739 (movdi_internal32): Use rs6000_nonimmediate_operand.
740 (unnamed splitter): Likewise.
741 * config/rs6000/spe.md (CMPTFEQ_GPR, TSTTFEQ_GPR, CMPTFGT_GPR,
742 TSTTFGT_GPR, CMPTFLT_GPR, TSTTFLT_GPR): New unspecs.
743 (SPE64TF, DITI): New mode macros.
744 (frob_df_di): Change to frob_<SPE64:mode>_<DITI:mode>; allow more
747 (frob_<mode>_di_2): New.
748 (frob_tf_di_8_2): New.
749 (frob_di_df): Change to frob_di_<mode>; allow more modes.
751 (frob_di_df_2): Change to frob_<DITI:mode>_<SPE64:mode>_2; allow
753 (frob_ti_<mode>_8_2): New.
755 (mov_si<mode>_e500_subreg0, mov_si<mode>_e500_subreg0_2,
756 mov_si<mode>_e500_subreg4, mov_si<mode>_e500_subreg4_2): Allow
758 (mov_sitf_e500_subreg8, mov_sitf_e500_subreg8_2,
759 mov_sitf_e500_subreg12, mov_sitf_e500_subreg12_2): New.
760 (spe_trunctfdf2_internal1, spe_trunctfsf2, spe_extenddftf2,
761 spe_fix_trunctfsi2, spe_fix_trunctfsi2_internal,
762 spe_negtf2_internal, spe_abstf2_cmp, spe_abstf2_tst): New.
763 (cmptfeq_gpr, tsttfeq_gpr, cmptfgt_gpr, tsttfgt_gpr, cmptflt_gpr,
766 2007-01-23 Ian Lance Taylor <iant@google.com>
768 * Makefile.in (OBJS-common): Reformat, alphabetize, but put
770 (OBJS-archive): Reformat, alphabetize.
771 (OBJS): Change out_object_file to OBJS-md.
773 2007-01-23 Uros Bizjak <ubizjak@gmail.com>
775 * config/i386/i386.md: Use REG_P, MEM_P, CONST_INT_P, LABEL_P,
776 JUMP_P and CALL_P predicates where applicable.
777 * config/i386/i386.c: Ditto.
778 * config/i386/i386.md: Ditto.
779 * config/i386/mmx.md: Ditto.
780 * config/i386/predicates.md: Ditto.
782 2007-01-22 Andreas Schwab <schwab@suse.de>
784 * config/m68k/m68k.h: Fix comment.
786 2007-01-22 Jan Hubicka <jh@suse.cz>
788 * passes.c (init_optimization_passes): Do not rerun
789 pass_early_warn_uninitialized.
791 2007-01-22 Richard Guenther <rguenther@suse.de>
793 PR tree-optimization/30038
794 * tree-ssa-math-opts.c (maybe_record_sincos): New static helper
796 (execute_cse_sincos_1): Likewise.
797 (execute_cse_sincos): Likewise.
798 (gate_cse_sincos): Likewise.
799 (pass_cse_sincos): New pass CSEing sin() and cos() calls using
800 the cexpi() canonicalization of sincos().
801 * tree-pass.h (pass_cse_sincos): Declare.
802 * passes.c (init_optimization_passes): New pass pas_cse_sincos.
804 2007-01-21 Eric Botcazou <ebotcazou@libertysurf.fr>
806 PR rtl-optimization/29329
807 * combine.c (replaced_rhs_insn): Rename to i2mod.
808 (replaced_rhs_value): Rename to i2mod_new_rhs.
809 (i2mod_old_rhs): New global variable.
810 (combine_instructions): Adjust for above change. Save a copy of
811 the old RHS into i2mod_old_rhs when the contents of a REG_EQUAL
812 note are substituted in the second instruction.
813 (distribute_notes) <REG_DEAD>: Adjust for above change. Do not
814 ditch the note if it pertains to the second eliminated register
815 and this register is mentioned in i2mod_old_rhs.
818 2006-09-12 Eric Botcazou <ebotcazou@libertysurf.fr>
820 * combine.c (distribute_notes) <REG_DEAD>: Do not consider SETs past
821 the insn to which the note was originally attached.
823 2007-01-21 Jan Hubicka <jh@suse.cz>
825 * ipa-inline.c (inlining_mode): Comment, move up.
826 (cgraph_decide_inlining_incrementally): Do not perform inlining
827 itself; fix handling of flattening of self recursive functions.
828 (cgraph_find_cycles): Remove.
829 (cgraph_flatten_node): Remove.
830 (cgraph_decide_inlining): Use incremental inliner to handle flattening.
831 (try_inline): New function.
832 (cgraph_early_inlining): Update call of
833 cgraph_decide_inlining_incrementally. Apply inlining here.
834 (apply_inline): Update call of cgraph_decide_inlining_incrementally.
836 2007-01-21 Dirk Mueller <dmueller@suse.de>
839 * tree-vrp.c (check_array_bounds): do not warn
840 about ADDR_EXPR's of ARRAY_REF's which are immediately
841 used in binary expressions.
843 2007-01-21 Ira Rosen <irar@il.ibm.com>
845 * tree-vectorizer.h (struct _stmt_vec_info): Add new field
846 read_write_dep and macros for its access.
847 * tree-vectorizer.c (new_stmt_vec_info): Initialize the new field.
848 * tree-vect-analyze.c (vect_analyze_data_ref_dependence): Remove
849 argument, call vect_check_interleaving for every independent pair of
850 data-refs. Mark loads that access the same memory location as a store
852 (vect_check_dependences): Remove.
853 (vect_analyze_data_ref_dependences): Remove vect_check_dependences
854 call, fix the call to vect_analyze_data_ref_dependence.
855 (vect_analyze_data_ref_access): For statements that access the same
856 data-ref, check that they are not stores; for loads, check that there
857 is no store that access the same location.
859 2007-01-20 Roger Sayle <roger@eyesopen.com>
860 Joseph Myers <joseph@codesourcery.com>
862 * doc/invoke.texi (-fdump-rtl-combine): Fix under/overfull hbox.
863 (-fdump-rtl-stack): Likewise.
864 (-fno-signed-zeros): Use @minus{} for a minus sign. Correct typo.
865 (-mcheck-zero-division, -mcpu): Fix under/overfull hbox.
866 (-mpt-fixed): Use @minus{} for minus sign.
867 (Using Precompiled Headers): Fix under/overfull hbox.
869 2007-01-20 Jan Hubicka <jh@suse.cz>
871 * tree-flow.h (struct stmt_ann_d): Move references_memory to proper
872 place within annotation.
874 2007-01-20 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
876 * pa.c (output_move_double): Change array size of xoperands to 4.
878 2007-01-20 Jan Hubicka <jh@suse.cz>
880 * tree-tailcall.c (adjust_return_value): Do not use RESULT_DECL
883 2007-01-19 Ian Lance Taylor <iant@google.com>
885 * expmed.c (expand_divmod) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]: Cast
886 constant to unsigned HOST_WIDE_INT before negating.
888 2007-01-19 Ian Lance Taylor <iant@google.com>
890 * tree-ssa-operands.h (struct vuse_vec_d): Change num_vuse field
892 (VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Use unsigned comparison.
893 (VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
894 (SET_VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Likewise.
895 (SET_VUSE_ELEMENT_VAR) [ENABLE_CHECKING]: Likewise.
896 (SET_VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
897 (realloc_vdef, realloc_vuse): Change second parameter to
899 (ssa_operand_iterator_d): Change vuse_index and mayuse_index
901 * tree-ssa-operands.c (realloc_vop): Change num_elem parameter to
902 unsigned. Change x and lim locals to unsigned.
903 (realloc_vdef, realloc_vuse): Change num_elem parameter to
905 (finalize_ssa_vuse_ops): Change old_i local to unsigned.
906 (copy_virtual_operands): Change i and n locals to unsigned.
908 2007-01-19 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
911 * builtins.c (fold_builtin_1): Handle builtin fdim.
913 2007-01-20 Jan Hubicka <jh@suse.cz>
915 * tree-ssa.c (init_tree_ssa): Do not call init_alias_heapvars.
916 * tree-ssa-structalias.c (compute_points_to_sets): Do call
918 (init_alias_heapvars): Initialize only when not already initialized.
919 (delete_alias_heapvars): Set heapvar_for_stmt to NULL.
921 2007-01-19 Roger Sayle <roger@eyesopen.com>
923 * common.opt (fsigned-zeros): New command line option.
924 * flags.h (HONOR_SIGNED_ZEROS): Control via flag_signed_zeros instead
925 of flag_unsafe_math_optimizations.
926 * opts.c (set_fast_math_flags): The -ffast-math command line option
927 implies -fno-signed-zeros.
928 (fast_math_flags_set_p): Likewise.
930 * doc/invoke.texi: Document new -fno-signed-zeros option, and update
931 the documentation of -ffast-math appropriately. Wrap long lines.
933 2007-01-19 Steve Ellcey <sje@cup.hp.com>
935 * system.h (ASM_MAKE_LABEL_LINKONCE): Poison.
936 * varasm.c (globalize_decl): Remove ASM_MAKE_LABEL_LINKONCE ifdef.
938 2007-01-19 Tomas Bily <tbily@suse.cz>
940 * cgraphunit.c (cgraph_finalize_function): Updating of pid
942 (tree_init_ic_make_global_vars): New function
943 (tree_init_edge_profiler): call of tree_init_ic_make_global_vars
944 (tree_gen_ic_profiler): New function
945 (tree_gen_ic_func_profiler): New function
946 (tree_profiling): Added calling of tree_gen_ic_func_profiler
947 (tree_profile_hooks): Added hook for indirec/virtual calls
948 * value-prof.c (tree_find_values_to_profile): New case for
950 (tree_values_to_profile): Call for determining indirect/virtual
952 (tree_indirect_call_to_profile): New function
953 (tree_ic_transform): New function
954 (tree_ic): New function
955 (find_func_by_pid): New function
956 (init_pid_map): New function
957 (tree_value_profile_transformations): Added check for
958 indirect/virtual call transformation
959 * value-prof.h (enum hist_type): New counter type for
960 indirect/virtual calls
961 (profile_hooks): Added new hook for profiling indirect/virtual
963 * profile.c (instrument_values): New case for indirect/virtual
965 * gcov-io.h (GCOV_LAST_VALUE_COUNTER): Changed to 6
966 (GCOV_COUNTER_V_INDIR): New counter type
967 (GCOV_COUNTER_NAMES): New name of counter "indirect" added
968 (GCOV_MERGE_FUNCTIONS): New merge function for indirect/virtual
970 * cgraph.c: Definition of cgraph_max_pid
971 (cgraph_create_node): Default init of pid attribute
972 * cgraph.h: Declaration of cgraph_max_pid
973 (struct cgraph_node): Added pid attribute
974 * libgcov.c (__gcov_indirect_call_profiler): New function
975 (__gcov_one_value_profiler_body): New function
976 (__gcov_one_value_profiler): Body was moved to
977 __gcov_one_value_profiler_body and calls it
979 2007-01-19 Basile Starynkevitch <basile@starynkevitch.net>
981 * doc/gty.texi (Options): Document the mark_hook option to GTY.
982 * gengtype.c (write_types_data, write_func_for_structure,
983 write_types, ggc_wtd, pch_wtd): Add skip_hooks to
984 write_types_data, ggc_wtd, pch_wtd for processing mark_hook.
985 (walk_type, write_func_for_structure): Generate the mark_hook if
988 2007-01-19 Jan Hubicka <jh@suse.cz>
990 * ipa-inline.c (cgraph_decide_inlining_incrementally): Instead of
991 'early' argument take inlining mode argument specifying whether to
992 inline for size/speeed or all functions; add support for flattening;
994 (cgraph_early_inlining): Update call of decide_inlining_incrementally.
996 2007-01-19 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
999 * toplev.c (warn_deprecated_use): Use %qD instead of %qs to print
1000 the name of the declared identifier.
1002 2007-01-19 Dirk Mueller <dmueller@suse.de>
1004 * config/i386.h (CONDITIONAL_REGISTER_USAGE): Store
1005 result of PIC_OFFSET_TABLE_REGNUM in temporary variable to avoid
1006 duplicate evaluation.
1008 2007-01-19 Uros Bizjak <ubizjak@gmail.com>
1010 * config/i386/i386.md (acos<mode>2): Rename from acossf2 and acosdf2.
1011 Macroize expander using X87MODEF12 mode macro. Extend operand 1
1012 to XFMode, use acosxf2 and truncate result to requested mode.
1013 Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
1014 (asin<mode>2): Similarly, with asin expanders.
1015 (*fscalexf4_i387): Rename from *fscalexf4.
1016 (expNcorexf3): New expander.
1017 (expxf2, exp10xf2, exp2xf2): Use expNcorexf3 expander.
1018 (exp<mode>2): Rename from expsf2 and expdf2. Macroize expander using
1019 X87MODEF12 mode macro. Extend operand 1 to XFMode, use expxf2 and
1020 truncate result to requested mode. Use SSE_FLOAT_MODE_P to disable
1021 patterns for SSE math.
1022 (exp10<mode>2): Similarly, with exp10 expanders.
1023 (exp2<mode>2): Similarly, with exp2 expanders.
1024 (expm1<mode>2): Similarly, with expm1 expanders.
1025 (ldexp<mode>3): Similarly, with ldexp expanders.
1026 (log<mode>2, log10<mode>2, log2<mode>2, log1p<mode>2, logb<mode>2):
1027 Use gen_truncxf<mode>2_i387_noop to truncate result.
1029 2007-01-19 Richard Sandiford <richard@codesourcery.com>
1031 * config/m68k/m68k.h (PREFERRED_STACK_BOUNDARY): Define to 32
1032 for ColdFire targets.
1034 2007-01-19 Nathan Sidwell <nathan@codesourcery.com>
1035 Richard Sandiford <richard@codesourcery.com>
1037 * config/m68k/m68k.h (M68K_STATIC_CHAIN_REG_NAME): New macro.
1038 (INITIALIZE_TRAMPOLINE): Use STATIC_CHAIN_REGNUM.
1039 (__transfer_from_trampoline): Use M68K_STATIC_CHAIN_REG_NAME.
1040 * config/m68k/m68kelf.h (STATIC_CHAIN_REGNUM): Override.
1041 (M68K_STATIC_CHAIN_REG_NAME): Likewise.
1042 * config/m68k/netbsd-elf.h (M68K_STATIC_CHAIN_REG_NAME): Likewise.
1044 2007-01-19 Richard Sandiford <richard@codesourcery.com>
1046 * config/m68k/m68k.md (adddi_dilshr32): Rename to...
1047 (*adddi_dilshr32): ...this. Fix formatting. Remove commented-out
1048 non-canonical pattern. Restrict to !TARGET_COLDFIRE.
1049 (*adddi_dilshr32_cf): New pattern.
1050 (adddi3, subdi3): Remove first alternatives.
1052 2007-01-19 Richard Sandiford <richard@codesourcery.com>
1054 * config/m68k/m68k.c (notice_update_cc): If an SFmode move is
1055 implemented using move.l, do not use its cc result for floating-point
1058 2007-01-19 Richard Sandiford <richard@codesourcery.com>
1060 * config/m68k/m68k.h (EXTRA_CONSTRAINT): Stop the 'T' constraint
1061 from accepting 's' constraints if flag_pic.
1063 2007-01-19 Richard Sandiford <richard@codesourcery.com>
1065 * config/m68k/m68k.md (bordered, bunordered, buneq, bunge, bungt)
1066 (bunle, bunlt, bltgt, bordered_rev, bunordered_rev, buneq_rev)
1067 (bunge_rev, bungt_rev, bunle_rev, bunlt_rev, bltgt_rev): Change
1068 condition from TARGET_68881 to TARGET_HARD_FLOAT.
1070 2007-01-19 Sandra Loosemore <sandra@codesourcery.com>
1072 * longlong.h (count_leading_zeros, COUNT_LEADING_ZEROS_0): Add
1073 ColdFire alternatives.
1074 * config/m68k/m68k.h (CLZ_DEFINED_VALUE_AT_ZERO): New macro.
1075 * config/m68k/m68k.md (clzsi2): Define for ColdFire
1076 architectures that support ff1 instruction.
1078 2007-01-19 Richard Sandiford <richard@codesourcery.com>
1079 Julian Brown <julian@codesourcery.com>
1081 * config/m68k/m68k.h (CONST_OK_FOR_LETTER_P): Add an 'R' case.
1082 * config/m68k/m68k.md (*movsi_cfv4): Fold into...
1083 (*movsi_cf): ...here. Remove unnecessary 'R' from 'Rg'.
1086 2007-01-19 Richard Sandiford <richard@codesourcery.com>
1088 * config/m68k/m68k-protos.h (valid_mov3q_const): Take a HOST_WIDE_INT
1090 (output_move_const_into_data_reg, output_move_simode_const): Delete.
1091 * config/m68k/m68k.c (const_method, const_int_cost): Take a
1092 HOST_WIDE_INT instead of an rtx.
1093 (m68k_rtx_costs): Update call accordingly.
1094 (output_move_const_into_data_reg): Likewise. Fix formatting.
1095 (valid_mov3q_const): Take a HOST_WIDE_INT instead of an rtx.
1097 (output_move_simode_const): Update calls after above changes.
1098 Rework to use automatic variables and predicates like MEM_P.
1099 * config/m68k/m68k.md (pushexthisi_const): Update call to
1102 2007-01-19 Dirk Mueller <dmueller@suse.de>
1104 * tree-ssa-alias.c (perform_var_substitution): Fix typo
1107 2007-01-19 Richard Guenther <rguenther@suse.de>
1109 * builtins.c (expand_builtin_cexpi): Fall back to expanding
1110 via cexp in case sincos is not available.
1112 2007-01-19 Richard Guenther <rguenther@suse.de>
1114 * doc/tm.texi (TARGET_HAS_SINCOS): Document new target macro.
1115 * defaults.h (TARGET_HAS_SINCOS): Default to off.
1116 * config/linux.h (TARGET_HAS_SINCOS): Set to on if we have glibc.
1117 * config/alpha/linux.h (TARGET_HAS_SINCOS): Likewise.
1118 * config/sparc/linux.h (TARGET_HAS_SINCOS): Likewise.
1119 * config/sparc/linux64.h (TARGET_HAS_SINCOS): Likewise.
1120 * config/rs6000/linux.h (TARGET_HAS_SINCOS): Likewise.
1121 * config/rs6000/linux64.h (TARGET_HAS_SINCOS): Likewise.
1123 2007-01-19 Uros Bizjak <ubizjak@gmail.com>
1125 * config/i386/i386.md (*fpatanxf3_i387, fpatan_extend<mode>xf3_i387):
1127 (atan2sf3_1, atan2df3_1, atan2xf3_1): Remove insn patterns.
1128 (atan2xf3): Directly generate RTL pattern.
1129 (atan2<mode>3): Rename from atan2sf3 and atan2df3 and macroize insn
1130 patterns using X87MODEF12 mode macro. Use fpatan_extend<mode>xf3_i387
1131 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
1132 disable patterns for SSE math.
1133 (atan<mode>2): Rename from atansf2 and atandf2 and macroize insn
1134 patterns using X87MODEF12 mode macro. Use fpatan_extend<mode>xf3_i387
1135 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
1136 disable patterns for SSE math.
1138 2007-01-19 Alexandre Oliva <aoliva@redhat.com>
1140 * libgcc-std.ver: Fix typo in %inherit for GCC_4.3.0.
1142 2007-01-18 Roger Sayle <roger@eyesopen.com>
1144 * fold-const.c (fold_unary) <VIEW_CONVERT_EXPR>: Optimize away a
1145 VIEW_CONVERT_EXPR to the same type as it's operand.
1147 2007-01-18 David Edelsohn <edelsohn@gnu.org>
1149 * config/rs6000/darwin-ldouble.c: Only build _SOFT_FLOAT if
1150 configured for long double 128.
1152 2007-01-18 Mike Stump <mrs@apple.com>
1154 * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Fix build
1157 2007-01-18 Michael Meissner <michael.meissner@amd.com>
1159 * i386.c (ix86_compute_frame_layout): Make fprintf's in #if 0 code
1162 2007-01-18 Jan Hubicka <jh@suse.cz>
1164 * tree-ssa-operands.c (vop_free_bucket_size): Never return value
1165 greater than NUM_VOP_FREE_BUCKETS.
1167 2007-01-18 Daniel Berlin <dberlin@dberlin.org>
1169 * tree-ssa-structalias.c: Update comments.
1170 (ptabitmap_obstack): Removed.
1172 (oldpta_obstack): Ditto.
1173 (stats): Add a few members.
1174 (struct variable_info): Remove node, complex, address_taken, and
1175 indirect_target members. Add oldsolution member.
1176 (new_var_info): Do not initialize removed members.
1177 (constraint_expr_type): Remove INCLUDES.
1178 (constraint_graph): Add size, implicit_preds, rep,
1179 indirect_cycles, eq_rep, label, direct_nodes, and complex members.
1180 (FIRST_REF_NODE): New macro.
1181 (LAST_REF_NODE): Ditto.
1182 (FIRST_ADDR_NODE): Ditto.
1183 (find): New function.
1185 (dump_constraint): Do not handle INCLUDES.
1186 (insert_into_complex): Do not insert duplicate constraints.
1187 (condense_varmap_nodes): Renamed and rewritten into ...
1188 (merge_node_constraints): This. Also fix bug in handling of
1189 offseted copy constraints.
1190 (clear_edges_for_node): No longer need to deal with preds at all,
1191 or removing associated preds/succs.
1192 (merge_graph_nodes): Deal with indirect_cycles.
1193 Don't deal with predecessors.
1194 (add_implicit_graph_edge): New function.
1195 (add_pred_graph_edge): Ditto.
1196 (add_graph_edge): Don't deal with predecessors.
1197 (build_constraint_graph): Removed.
1198 (build_pred_graph): New function.
1199 (build_succ_graph): Ditto.
1200 (struct scc_info): Removed in_component. Added roots, dfs, and
1201 node_mapping. Remove visited_index, unification_queue.
1202 (scc_visit): Deal with union-find we do now.
1203 Deal with cycles with REF nodes.
1204 (collapse_nodes): Renamed and rewritten to ...
1205 (unify_nodes): This.
1206 (process_unification_queue): Removed.
1207 (topo_visit): Cleanup
1208 (do_da_constraint): Use find.
1209 (do_sd_constraint): Ditto.
1210 (do_ds_constraint): Ditto.
1211 (do_complex_constraint): Ditto.
1212 (init_scc_info): Update for removed and added members.
1213 (find_and_collapse_graph_cycles): Renamed and rewritten into ...
1214 (find_indirect_cycles): This.
1215 (equivalence_class): New variable.
1216 (label_visit): New function.
1217 (perform_variable_substitution): Rewritten.
1218 (free_var_substitution_info): New function.
1219 (find_equivalent_node): Ditto.
1220 (move_complex_constraints): Ditto.
1221 (eliminate_indirect_cycles): Ditto.
1222 (solve_graph): Only propagate changed bits.
1223 Use indirect cycle elimination.
1225 (tree_id_t): Rename to tree_vi_t, delete id member, add vi member.
1226 (tree_id_eq): Renamed to ...
1227 (tree_vi_eq): This. Update for member change
1228 (insert_id_for_tree): Renamed and rewritten to ...
1229 (insert_vi_for_tree): This.
1230 (lookup_id_for_tree): Renamed and rewritten to ...
1231 (lookup_vi_for_tree): This.
1232 (get_id_for_tree): Renamed and rewritten to ...
1233 (get_vi_for_tree): Ditto.
1234 (get_constraint_exp_from_ssa_var): Update to use get_vi_for_tree.
1235 (process_constraint): Don't handle INCLUDES.
1236 Remove special ADDRESSOF case.
1237 (find_func_aliases): Rewrite to use vi functions instead of id
1239 (create_function_info_for): Ditto.
1240 (create_variable_info_for): Ditto.
1241 (intra_create_variable_infos): Ditto.
1242 (merge_smts_into): Ditto.
1243 (find_what_p_points_to): Ditto.
1244 (init_base_vars): Ditto.
1245 (init_alias_vars): Ditto.
1246 (remove_preds_and_fake_succs): New function.
1247 (dump_sa_points_to_info): Dump new stats.
1248 (dump_solution_for_var): Use find.
1249 (set_used_smts): Fix formatting.
1250 (compute_points_to_sets): Updated for new functions.
1251 (ipa_pta_execute): Ditto.
1253 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
1254 Richard Sandiford <richard@codesourcery.com>
1256 * doc/tm.texi (TARGET_FUNCTION_VALUE): Expand documentation of
1258 * calls.c (expand_call): If the return value is a PARALLEL,
1259 extract its first member.
1260 * config/m68k/linux.h (FUNCTION_EXTRA_EPILOGUE): Remove.
1261 * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
1262 use FUNCTION_EXTRA_EPILOGUE.
1263 (m68k_function_value): Return a PARALLEL if the return value
1264 is of a pointer type.
1265 * config/m68k/netbsd-elf.h (current_function_returns_pointer)
1266 (FUNCTION_EXTRA_EPILOGUE): Remove.
1267 * config/m68k/m68k.md (D0_REG): New constant.
1269 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
1271 * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
1272 output a NOP for empty epilogues.
1274 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1276 * config/m68k/m68k.c (m68k_use_return_insn): Update comments
1277 before function. Extend register save check to include all
1278 registers, not just integer ones.
1280 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
1282 * config/m68k/m68k-protos.h (use_return_insn): Rename to...
1283 (m68k_use_return_insn): ...this.
1284 * config/m68k/m68k.h (USE_RETURN_INSN): Delete.
1285 * config/m68k/m68k.c (use_return_insn): Rename to...
1286 (m68k_use_return_insn): ...this.
1287 * config/m68k/m68k.md (return): Use m68k_use_return_insn instead
1290 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
1292 * config/m68k/fpgnulib.c (__truncdfsf2): Implement round to
1293 nearest even, fix denormal rounding overflow.
1295 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1297 * config/m68k/m68k.md (movsf_cf_hard): Use fsmove instead of
1298 f%$smove and f%$move.
1299 (movdf_cf_hard): Use fdmove for cases 0 and 3 and fmove for case 1.
1300 (extendsfdf2_cf): Use fdmove instead of f%&move.
1301 (truncdfsf2_cf): Use fsmove instead of f%$smove.
1302 (add<mode>3_cf, sub<mode>3_cf): Use <FP:prec> instead of <FP:round>.
1304 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1306 * config/m68k/m68k.md (movdf_cf_hard): Use output_move_double for
1309 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1311 * real.h (real_format): Add a canonical_nan_lsbs_set field.
1312 (coldfire_single_format): Declare.
1313 (coldfire_double_format): Likewise.
1314 * real.c (encode_ieee_single): Use canonical_nan_lsbs_set instead
1315 of qnan_msb_set to determine the lower bits of a canonical
1317 (encode_ieee_double): Likewise.
1318 (encode_ieee_quad): Likewise.
1319 (ieee_single_format): Initialize canonical_nan_lsbs_set.
1320 (mips_single_format): Likewise.
1321 (ieee_double_format): Likewise.
1322 (mips_double_format): Likewise.
1323 (ieee_extended_motorola_format): Likewise.
1324 (ieee_extended_intel_96_format): Likewise.
1325 (ieee_extended_intel_128_format): Likewise.
1326 (ieee_extended_intel_96_round_53_format): Likewise.
1327 (ibm_extended_format): Likewise.
1328 (mips_extended_format): Likewise.
1329 (ieee_quad_format): Likewise.
1330 (mips_quad_format): Likewise.
1331 (vax_f_format): Likewise.
1332 (vax_d_format): Likewise.
1333 (vax_g_format): Likewise.
1334 (i370_single_format): Likewise.
1335 (i370_double_format): Likewise.
1336 (decimal_single_format): Likewise.
1337 (decimal_double_format): Likewise.
1338 (decimal_quad_format): Likewise.
1339 (c4x_single_format): Likewise.
1340 (c4x_extended_format): Likewise.
1341 (real_internal_format): Likewise.
1342 (coldfire_single_format): New real_format.
1343 (coldfire_double_format): Likewise.
1344 * config/pdp11/pdp11.c (pdp11_f_format): Initialize
1345 canonical_nan_lsbs_set.
1346 (pdp11_d_format): Likewise.
1347 * config/m68k/m68k.c (override_options): Override REAL_FORMAT_MODE
1348 if TARGET_COLDFIRE_CPU.
1350 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1352 * config/m68k/m68k-protos.h (m68k_output_pic_call): Delete.
1353 (output_call, m68k_legitimize_call_address): Declare.
1354 * config/m68k/m68k.h (EXTRA_CONSTRAINT): Remove unnecessary
1355 parenthesees. Add support for a 'W' constraint.
1356 (LEGITIMATE_PIC_OPERAND_P): Remove SYMBOL_REF_FLAG handling.
1357 (PRINT_OPERAND_PUNCT_VALID_P): Remove comment about 'o'.
1358 (m68k_symbolic_call, m68k_symbolic_jump): Declare.
1359 * config/m68k/m68k.c (m68k_symbolic_call, m68k_symbolic_jump): New
1361 (override_options): Initialize them. Do not set flag_no_function_cse
1362 for TARGET_ID_SHARED_LIBRARY.
1363 (m68k_output_pic_call): Delete.
1364 (m68k_legitimize_call_address): New function.
1365 (print_operand): Remove the %o prefix. Handle the %p prefix.
1366 (output_call): New function.
1367 (m68k_output_mi_thunk): Use m68k_symbolic_jump. Always load the
1368 target address from the GOT if symbolic jumps are not allowed.
1369 * config/m68k/m68k.md (call, general_operand): Do not set
1370 SYMBOL_REF_FLAG. Use m68k_legitimize_call_address instead.
1371 Merge separate flag_pic and !flag_pic define_insns into...
1372 (*call, *call_value): ...these new patterns. Match the address
1373 rather than the containing MEM and require it to be a call_operand.
1374 Use output_call to generate the asm template.
1375 * config/m68k/predicates.md (const_call_operand): New predicate.
1376 (call_operand): Likewise.
1378 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
1380 * config/m68k/m68k.h (REGISTER_MOVE_COST): Simplify definition.
1381 (STACK_GROWS_DOWNWARD): Define to 1.
1382 (FUNCTION_VALUE, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P): Equivocate
1383 comments, emphasizing that these values are only defaults.
1384 * config/m68k/linux.h (LINK_SPEC): Fix formatting in #undef.
1385 * config/m68k/m68k.c (const_method): Remove trailing whitespace.
1387 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1389 * config/m68k/m68k.md (cmpsi): Remove outdated flag_pic handling.
1391 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
1392 Richard Sandiford <richard@codesourcery.com>
1394 * config/m68k/m68k.h (DATA_REGNO_P, ADDRESS_REGNO_P, INT_REGNO_P)
1395 (FP_REGNO_P): New macros.
1396 (REGNO_OK_FOR_INDEX_P, REGNO_OK_FOR_BASE_P, REGNO_OK_FOR_DATA_P)
1397 (REGNO_OK_FOR_FP_P, REG_OK_FOR_INDEX_P, REG_OK_FOR_BASE_P): Use them.
1398 (EH_RETURN_STACKADJ_RTX): Use A0_REG.
1399 * config/m68k/m68k.c (m68k_regno_mode_ok): Use the new REGNO macros.
1401 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
1403 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
1404 (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-rtems*): Add t-floatlib
1406 * config/m68k/t-floatlib: New file, extracting common code from...
1407 * config/m68k/t-m68kbare, config/m68k/t-m68kelf,
1408 * config/m68k/t-uclinux: Here.
1409 * config/m68k/fpgnulib.c: Do not compile extendeed precision
1410 routines on ColdFire targets.
1412 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
1414 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
1415 (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*)
1416 (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-linux*)
1417 (m68k-*-rtems*): Use tm_file rather than m68k/m68k.h and
1418 explicitly set MOTOROLA to 1.
1419 * config/m68k/m68k.h (MOTOROLA): Simplify definition accordingly.
1421 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1422 Nathan Sidwell <nathan@codesourcery.com>
1424 * config/m68k/m68k.h (PCC_STATIC_STRUCT_RETURN): Delete.
1425 (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Add MOTOROLA cases.
1426 * config/m68k/coff.h (REGISTER_PREFIX_MD): Delete.
1427 * config/m68k/m68020-elf.h (LIB_SPEC): Delete.
1428 * config/m68k/m68k-none.h (CC1_SPEC, CPP_SUBTARGET_SPEC): Delete.
1429 * config/m68k/m68kelf.h (IMMEDIATE_PREFIX, REGISTER_PREFIX_MD)
1430 (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Delete.
1431 (NO_DOLLAR_IN_LABEL, PCC_STATIC_STRUCT_RETURN): Don't undefine.
1432 (BSS_ASM_OP, ASM_OUTPUT_SKIP, ASM_OUTPUT_ASCII): Delete.
1433 * config/m68k/m68kemb.h (PCC_STATIC_STRUCT_RETURN): Don't undefine.
1434 * config/m68k/linux.h (SIZE_TYPE, PTRDIFF_TYPE, WCHAR_TYPE)
1435 (WCHAR_TYPE_SIZE, TARGET_OBJFMT_CPP_BUILTINS): Delete.
1436 (TARGET_OS_CPP_BUILTINS): Don't define mc68000 and mc68020 here.
1437 (DBX_REGISTER_NUMBER): Delete.
1438 * config/m68k/netbsd-elf.h (IMMEDIATE_PREFIX): Delete.
1439 (PCC_STATIC_STRUCT_RETURN): Don't undefine.
1440 * config/m68k/openbsd.h (PCC_STATIC_STRUCT_RETURN): Define.
1442 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1444 * config.gcc (m68k-*-uclinux*): Add flat.h to $tm_file.
1445 * config/flat.h: New file.
1446 * crtstuff.c (USE_PT_GNU_EH_FRAME): Don't define if
1448 * config/m68k/m68k.h (ASM_PREFERRED_EH_DATA_FORMAT): Do not use
1449 indirect references for -msep-data or -mid-shared-library.
1450 Do not use PC-relative code addresses either.
1451 * config/m68k/m68k.c (override_options): Restrict -fPIC error
1453 * config/m68k/uclinux.h (STARTFILE_SPEC): Define. Use Scrt1.o
1454 for shared libraries and crt1.o for executables. Use crti.o and
1456 (ENDFILE_SPEC): Use crtend.o and crtn.o.
1457 (LIB_SPEC): Suppress -Rlibc.gdb if -static-libc is given.
1458 Do not add -elf2flt or -shared-lib-id options here.
1459 (LINK_SPEC): Define. Pass -elf2flt if no -elf2flt option is given.
1460 Pass -shared-lib-id if -mid-shared-library, taking the library
1461 identifier from -mshared-library-id if given, otherwise
1463 (EH_FRAME_IN_DATA_SECTION): Do not undefine.
1464 (INIT_SECTION_ASM_OP, FINI_SECTION_ASM_OP): Likewise.
1465 (TARGET_OS_CPP_BUILTINS): Define __GXX_MERGED_TYPEINFO_NAMES=0
1466 and __GXX_TYPEINFO_EQUALITY_INLINE=0 if -mid-shared-library.
1467 (DRIVER_SELF_SPECS): Map unadorned PIC options to -msep-data.
1468 * config/m68k/t-uclinux (EXTRA_MULTILIB_PARTS): Add crtbegin.o
1470 * config/m68k/lb1sf68.asm (PICCALL): Use an lea and pc-relative
1471 jump sequence for ISA A and ISA A+.
1472 (PICJUMP): Likewise.
1474 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1476 * config.gcc (m68k-*-uclinux*): Base the port on the common
1477 and m68k GNU/Linux files rather than on the generic ELF ones.
1478 * config/m68k/uclinux.h (TARGET_VERSION): Override.
1479 (TARGET_OS_CPP_BUILTINS): Use LINUX_TARGET_OS_CPP_BUILTINS.
1481 2007-01-18 Julian Brown <julian@codesourcery.com>
1482 Richard Sandiford <richard@codesourcery.com>
1484 * config/m68k/m68k.h (LONG_DOUBLE_TYPE_SIZE): Make 64-bit on ColdFire.
1485 (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
1486 * config/m68k/netbsd-elf.h (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Undefine
1488 * config/m68k/uclinux-oldabi.h (LONG_DOUBLE_TYPE_SIZE): Redefine to
1490 (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
1492 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1494 * doc/install.texi: Document m68k-uclinuxoldabi.
1495 * config.gcc (m68k-*-uclinuxoldabi): New configuration.
1496 * config/m68k/uclinux-oldabi.h: New file, copied from
1497 config/m68k/uclinux.h.
1499 2007-01-18 Kaz Kojima <kkojima@rr.iij4u.or.jp>
1501 * config/m32r/m32r-protos.h (m32r_expand_epilogue): Declare it.
1502 * config/m32r/m32r.c (m32r_setup_incoming_varargs): Use gen_frame_mem.
1503 (m32r_compute_frame_size): Use unsigned for regno.
1504 (m32r_reload_lr): Use gen_frame_mem.
1506 (m32r_output_function_epilogue): Don't output the function epilogue
1508 (m32r_expand_epilogue): New.
1509 (direct_return): Return false if the function has the interrupt
1511 (m32r_hard_regno_rename_ok): Remove code for the textual epilogue.
1512 * config/m32r/m32r.md (epilogue): New expander.
1513 (return_lr, return_rte): New insns.
1514 (return): Make it expander.
1515 (return_normal): New expander.
1517 2007-01-18 Josh Conner <jconner@apple.com>
1520 * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Add
1521 support for UNLE, UNLT, UNGE, and UNGT.
1523 2007-01-18 Jan Hubicka <jh@suse.cz>
1525 * tree-vrp.c (finalize_jump_threads): Do not call cleanup_cfg by hand.
1526 * tree-tailcall (add_virtual_phis): Likewise.
1527 (optimize_tail_call): Return TODOs.
1528 (execute_tail_calls): Return TODOs.
1529 * tree-ssa-ccp (execute_fold_all_builtins): Do cleanup_cfg via TODO.
1530 * tree-cfgcleanup.c (cleanup_tree_cfg_loop): Return if something
1532 * tree-ssa-forwprop.c (tree_ssa_forward_propagate_single_use_value):
1533 Cleanup cfg using TODO.
1534 * tree-flow.h (cleanup_tree_cfg_loop): Update prototype.
1535 * passes.c (execute_function_todo): When cleanup did something, remove
1537 * tree-cfg.c (pass_build_cfg): Add cleanup_cfg TODO.
1538 (make_edges): Don't cleanup_cfg.
1540 2007-01-18 Uros Bizjak <ubizjak@gmail.com>
1542 * reg-stack.c (subst_stack_regs_pat) [UNSPEC_SINCOS_COS,
1543 UNSPEC_XTRACT_FRACT]: Use generic code for instructions that
1544 operate on the top of stack.
1545 [UNSPEC_SINCOS_SIN, UNSPEC_XTRACT_EXP, UNSPEC_TAN]: Rewrite
1546 register handling of instructions that output to the second
1548 [UNSPEC_TAN_ONE, UNSPEC_TAN_TAN]: Remove.
1549 (move_for_stack_reg): Special-case check for dead destination
1550 stack slot for constant load of 1.0 inside UNSPEC_TAN.
1552 * config/i386/i386.md (UNSPEC_TAN): New constant.
1553 (UNSPEC_TAN_ONE, UNSPEC_TAN_TAN): Remove.
1554 (fptanxf4_i387, fptan_extend<mode>xf4_i387): New patterns
1555 to correctly model move of constant 1.0 to top stack slot.
1556 (*tandf3_1, *tansf3_1, *tanxf3_1): Remove insn patterns.
1557 (unnamed peephole2 pattern): Remove corresponding peephole2
1558 pattern that optimizes tan insn and loading of constant 1.0.
1559 (tanxf2): Use fptanxf4_i387.
1560 (tan<mode>2): Rename from tansf2 and tandf2 and macroize insn
1561 patterns using X87MODEF12 mode macro. Use fptan_extend<mode>xf4_i387
1562 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
1563 disable patterns for SSE math.
1564 (sincos<mode>3): Use truncxf<mode>2_i387_noop for truncation.
1565 (fyl2x_extend<mode>xf3_i387): Use X87MODEF12 for operand 1.
1567 2007-01-18 Dirk Mueller <dmueller@suse.de>
1568 Richard Guenther <rguenther@suse.de>
1571 * doc/invoke.texi (Warray-bounds): Document -Warray-bounds.
1572 * common.opt (Warray-bounds): Add new warning option.
1573 * c-opts.c (c_common_handle_option): Define -Warray-bounds
1575 * Makefile.in: make tree-vrp.o depend on toplev.h
1576 * tree-vrp.c (vrp_finalize): Call check_array_refs if -Warray-bounds
1578 (check_array_refs, check_array_bounds, check_array_ref): New.
1580 2007-01-18 Jan Hubicka <jh@suse.cz>
1582 * tree-ssa-ccp.c (ccp_finalize): Return if something changed.
1583 (execute_ssa_ccp): Return flags conditionally.
1584 * tree-ssa-propagate.c (substitue_and_fold): Return if something was
1586 * tree-ssa-propagate.h (substitute_and_fold): Update prototype.
1588 2007-01-18 Steven Bosscher <steven@gcc.gnu.org>
1590 * cfgcleanup.c (cleanup_cfg): Detect cfglayout mode and set
1591 the CLEANUP_CFGLAYOUT flag when in cfglayout mode.
1593 * Makefile.c (GTFILES): Add cfglayout.h.
1594 * gengtype.c (open_base_files): Likewise.
1595 * cfglayout.c (cfg_layout_function_footer,
1596 cfg_layout_function_header) Reindent to make gengtype happy.
1597 * cfglayout.h (cfg_layout_function_footer,
1598 cfg_layout_function_header): Add GTY(()) marker.
1600 * ifcvt.c (noce_try_sign_mask): Make sure INSN_B is non-null.
1602 2007-01-18 Ben Elliston <bje@au.ibm.com>
1604 * genautomata.c (write_automata): Include xstrerror output in the
1605 error message if writing the DFA description file fails.
1607 2007-01-17 H.J. Lu <hongjiu.lu@intel.com>
1609 * config/mips/mips-protos.h (mips_output_external): Make it
1611 * config/mips/iris.h (TARGET_ASM_EXTERNAL_LIBCALL): Removed.
1612 * config/mips/mips.c (irix_output_external_libcall): Likewise.
1613 (extern_list): Likewise.
1614 (extern_head): Likewise.
1615 (TARGET_ASM_FILE_END): Likewise.
1616 (mips_file_end): Likewise.
1617 (mips_output_external): Rewritten.
1619 2007-01-18 Ben Elliston <bje@au.ibm.com>
1621 * genpreds.c (write_insn_preds_c): Only write out the function
1622 body for regclass_for_constraint if we have register constraints.
1624 2007-01-17 Tom Tromey <tromey@redhat.com>
1626 * doc/sourcebuild.texi (libgcj Tests): Use sourceware.org.
1627 * doc/install.texi (Testing): Use sourceware.org.
1628 (Binaries): Likewise.
1629 (Specific): Likewise.
1630 * doc/contrib.texi (Contributors): Use sourceware.org.
1632 2007-01-17 Anatoly Sokolov <aesok@post.ru>
1634 * config/avr/avr.h (AVR_HAVE_LPMX): New macro.
1635 (AVR_ENHANCED): Rename to ...
1636 (AVR_HAVE_MUL): ... new.
1637 (avr_enhanced_p): Rename to ...
1638 (avr_have_mul_p): ... new.
1639 (TARGET_CPU_CPP_BUILTINS): Use 'avr_have_mul_p' instead of
1640 'avr_enhanced_p' for "__AVR_ENHANCED__". Define "__AVR_HAVE_MUL__".
1641 * config/avr/avr.c (avr_enhanced_p): Rename to ...
1642 (avr_have_mul_p): ... new.
1643 (base_arch_s): Rename 'enhanced' to 'have_mul'.
1644 (avr_override_options): Use 'avr_have_mul_p' and 'have_mul' instead of
1645 'avr_enhanced_p' and 'enhanced'.
1646 (ashlhi3_out, ashrhi3_out, lshrhi3_out, avr_rtx_costs): Use
1647 AVR_HAVE_MUL instead of AVR_ENHANCED.
1648 * avr.md (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
1649 (mulqi3, *mulqi3_enh, *mulqi3_call, mulqihi3, umulqihi3, mulhi3,
1650 *mulhi3_enh, *mulhi3_call, mulsi3, *mulsi3_call): Use AVR_HAVE_MUL
1651 instead of AVR_ENHANCED.
1652 (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
1653 * libgcc.S: Use __AVR_HAVE_MUL__ instead of __AVR_ENHANCED__.
1654 (__tablejump__): Use __AVR_HAVE_LPMX__ instead of __AVR_ENHANCED__.
1656 2007-01-17 Ian Lance Taylor <iant@google.com>
1658 * vec.h (VEC_reserve_exact): Define.
1659 (vec_gc_p_reserve_exact): Declare.
1660 (vec_gc_o_reserve_exact): Declare.
1661 (vec_heap_p_reserve_exact): Declare.
1662 (vec_heap_o_reserve_exact): Declare.
1663 (VEC_OP (T,A,reserve_exact)): New static inline function, three
1665 (VEC_OP (T,A,reserve)) [all versions]: Remove handling of
1667 (VEC_OP (T,A,alloc)) [all versions]: Call ...reserve_exact.
1668 (VEC_OP (T,A,copy)) [all versions]: Likewise.
1669 (VEC_OP (T,a,safe_grow)) [all versions]: Likewise.
1670 * vec.c (calculate_allocation): Add exact parameter. Change all
1672 (vec_gc_o_reserve_1): New static function, from vec_gc_o_reserve.
1673 (vec_gc_p_reserve, vec_gc_o_reserve): Call vec_gc_o_reserve_1.
1674 (vec_gc_p_reserve_exact, vec_gc_o_reserve_exact): New functions.
1675 (vec_heap_o_reserve_1): New static function, from vec_heap_o_reserve.
1676 (vec_heap_p_reserve, vec_heap_o_reserve): Call vec_heap_o_reserve_1.
1677 (vec_heap_p_reserve_exact): New function.
1678 (vec_heap_o_reserve_exact): New function.
1680 2007-01-17 Jan Hubicka <jh@suse.cz>
1682 * ipa-type-escape.c (look_for_casts): Revamp using handled_component_p.
1684 2007-01-17 Eric Christopher <echristo@apple.com>
1686 * config.gcc: Support core2 processor.
1688 2007-01-16 Jan Hubicka <jh@suse.cz>
1690 * tree-ssanames.c (release_dead_ssa_names): Instead of ggc_freeing
1691 the names, just unlink the chain so we don't crash on dangling pointers
1694 2007-01-16 Jan Hubicka <jh@suse.cz>
1696 * cgraph.h (cgraph_decide_inlining_incrementally): Kill.
1697 * tree-pass.h: Reorder to make IPA passes appear toegher.
1698 (pass_early_inline, pass_inline_parameters, pass_apply_inline): Declare.
1699 * cgraphunit.c (cgraph_finalize_function): Do not compute inling
1700 parameters, do not call early inliner.
1701 * ipa-inline.c: Update comments. Include tree-flow.h
1702 (cgraph_decide_inlining): Do not compute inlining parameters.
1703 (cgraph_decide_inlining_incrementally): Return TODOs; assume to
1704 be called with function context set up.
1705 (pass_ipa_inline): Remove unreachable functions before pass.
1706 (cgraph_early_inlining): Simplify assuming to be called from the
1708 (pass_early_inline): New pass.
1709 (cgraph_gate_ipa_early_inlining): New gate.
1710 (pass_ipa_early_inline): Turn into simple wrapper.
1711 (compute_inline_parameters): New function.
1712 (gate_inline_passes): New gate.
1713 (pass_inline_parameters): New pass.
1714 (apply_inline): Move here from tree-optimize.c
1715 (pass_apply_inline): New pass.
1716 * ipa.c (cgraph_remove_unreachable_nodes): Verify cgraph after
1718 * tree-inline.c (optimize_inline_calls): Return TODOs rather than
1720 (tree_function_versioning): Do not allocate dummy struct function.
1721 * tree-inline.h (optimize_inline_calls): Update prototype.
1722 * tree-optimize.c (execute_fixup_cfg): Export.
1723 (pass_fixup_cfg): Remove
1724 (tree_rest_of_compilation): Do not apply inlines.
1725 * tree-flow.h (execute_fixup_cfg): Declare.
1726 * Makefile.in (gt-passes.c): New.
1727 * passes.c: Include gt-passes.h
1728 (init_optimization_passes): New passes.
1729 (nnodes, order): New static vars.
1730 (do_per_function_toporder): New function.
1731 (execute_one_pass): Dump current pass here.
1732 (execute_ipa_pass_list): Don't dump current pass here.
1734 2007-01-16 Janis Johnson <janis187@us.ibm.com>
1736 * config/dfp-bit.c (dfp_compare_op): Return separate value for NaN.
1737 (DFP_NE, DFP_LE, DFP_GE): Return false for NaN.
1739 2007-01-16 David Edelsohn <edelsohn@gnu.org>
1741 * config/rs6000/darwin-ldouble.c: Build file for SOFT_FLOAT.
1742 (strong_alias): Define.
1743 (__gcc_qmul): Provide non-FMA for soft-float.
1749 (__gcc_qunord): New.
1759 * config/rs6000/rs6000.c (rs6000_init_libfuncs): Initialize
1760 soft-float functions.
1761 * config/rs6000/libgcc-ppc-glibc.ver: Version soft-float symbols.
1762 * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Do not warn
1763 about long double soft float.
1765 2007-01-16 Dorit Nuzman <dorit@il.ibm.com>
1766 Tehila Meyzels <tehila@il.ibm.com>
1768 * tree-vectorizer.h (is_pattern_stmt_p): New.
1769 * tree-vect-analyze.c (vect_determine_vectorization_factor): Fix
1770 formatting (tabs instead of spaces). Cleanup and clarify setting
1771 of STMT_VINFO_VECTYPE. Call is_pattern_stmt_p.
1772 * tree-vect-transform.c (vect_get_vec_def_for_operand): Fix typo.
1773 (vectorizable_type_demotion): Check that types are integral.
1774 (vectorizable_type_promotion): Likewise.
1775 (vectorizable_store): Fix typo. Eliminate new-line at end of
1778 2007-01-16 Jan Hubicka <jh@suse.cz>
1780 * tree-ssanames.c (release_dead_ssa_names): Remove invalidated
1783 2007-01-15 Eric Christopher <echristo@apple.com>
1785 * ifcvt.c: Include vec.h, vecprim.h.
1786 (check_cond_move_block): New argument regs.
1787 Reorganize. Add registers used to regs.
1788 (cond_move_process_if_block): Use regs set above as
1791 2007-01-15 Eric Christopher <echristo@apple.com>
1793 * config/darwin.h: Update copyright.
1794 (TARGET_OPTION_TRANSLATE_TABLE): Add umbrella.
1795 (LINK_COMMAND_SPEC): Add -u.
1796 (LINK_SPEC): Fix umbrella for above.
1798 2007-01-15 Joseph S. Myers <joseph@codesourcery.com>
1800 * config/soft-fp/op-common.h, config/soft-fp/op-4.h: Update from
1803 2007-01-15 Tom Tromey <tromey@redhat.com>
1805 * doc/sourcebuild.texi (libgcj Tests): Don't mention jacks.
1806 * doc/install.texi (Testing): Don't mention jacks.
1807 (Configuration): Document --enable-java-maintainer-mode. Move
1808 --with-java-home to libgcj-specific section. Document
1810 (Prerequisites): Mention --enable-java-maintainer-mode, ecj1.
1812 2007-01-15 Jan Hubicka <jh@suse.cz>
1814 * tree-ssa-dce.c (DCE_TODOs): New.
1815 (propagate_necessity): Return if something changed.
1816 (eliminate_unnecessary_stmts): Likewise.
1817 (perform_tree_ssa_dce): Return TODO flags when needed.
1818 (pass_dce, pass_dce_loop, pass_cd_dce): Remove TODO flags.
1820 2007-01-15 Uros Bizjak <ubizjak@gmail.com>
1822 * config/i386/i386.md (fyl2xxf3_i387): Rename from fyl2x_xf3.
1823 (fyl2x_extend<mode>xf3_i387): New insn pattern.
1824 (log<mode>2): Rename from logsf2 and logdf2 and macroize insn
1825 patterns using X87MODEF12 mode macro. Extend operand 1
1826 to XFmode. Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
1827 (log10<mode>2): Ditto.
1828 (log2<mode>2): Ditto.
1829 (log1p<mode>2): Ditto.
1830 (logb<mode>2): Ditto.
1831 (fyl2xp1xf3_i387): Rename from fyl2xp1_xf3.
1832 (fyl2xp1_extend<mode>xf3_i387): New insn pattern.
1833 (*fxtractxf3_i387): Rename from *fxtractxf3.
1834 (fxtract_extend<mode>xf3_i387): New insn pattern.
1835 (ilogbsi2): Use match_dup 3, not match_operand:XF 3.
1836 * config/i386/i386.c (ix86_emit_i387_log1p): Use gen_fyl2xp1xf3_i387()
1837 and gen_fyl2xxf3_i387().
1839 2007-01-14 Zdenek Dvorak <dvorakz@suse.cz>
1841 * loop-unswitch.c (unswitch_loop): Do not call fix_loop_placement.
1842 * cfgloopmanip.c (fix_loop_placement): Made static. Use
1843 get_loop_exit_edges. Changed return type to bool.
1844 * cfgloop.h (fix_loop_placement): Declaration removed.
1846 2007-01-14 Dorit Nuzman <dorit@il.ibm.com>
1848 * param.h (MIN_VECT_LOOP_BOUND): New.
1849 * params.def (MIN_VECT_LOOP_BOUND): New.
1850 * tree-vectorizer.c (slpeel_tree_peel_loop_to_edge): Takes another
1851 argument - minimum threshold for number of iterations.
1852 * tree-vectorizer.h (slpeel_tree_peel_loop_to_edge): Add another
1853 argument to declaration.
1854 * tree-vect-analyze.c (vect_analyze_operations): Check value of
1855 MIN_VECT_LOOP_BOUND.
1856 * tree-vect-transform.c (vect_do_peeling_for_loop_bound): Call
1857 slpeel_tree_peel_loop_to_edge with additional argument.
1858 (vect_do_peeling_for_alignment): Likewise.
1859 * doc/invoke.texi (min-vect-loop-bound): Document new param option.
1861 2007-01-14 Uros Bizjak <ubizjak@gmail.com>
1864 * config/i386/i386.c (print_operand) ['z']: Output 'b' for
1867 2007-01-14 Jan Hubicka <jh@suse.cz>
1869 * tree-dfa.c (remove_referenced_var): New function.
1870 * tree-ssa-live.c (remove_unused_locals): Walk referenced vars and
1871 prune referenced vars list too.
1872 * tree-flow.h (remove_referenced_var): Declare.
1874 2007-01-14 Jan Hubicka <jh@suse.cz>
1876 * tree-eh.c (add_stmt_to_eh_region_fn): Do not add call_exprs
1878 (remove_stmt_from_eh_region_fn): Do not remove call_exprs.
1879 (verify_eh_throw_stmt_node, verify_eh_throw_table_statements): Kill.
1880 * except.h (verify_eh_throw_table_statements): Kill prototype.
1881 * cfgexpand.c (expand_gimple_basic_block): Propagate Eh regions
1882 into call exrepssions.
1883 * tree-optimize.c (execute_free_cfg_annotatiosn): Do not call
1885 * tree-cfg.c: Include pointer-set.h.
1886 (verify_node_sharing): Work on pointer set.
1887 (verify_eh_throw_stmt_node): New.
1888 (verify_stmts): Use pointers sets, verify throw_stmt.
1890 2007-01-13 Zdenek Dvorak <dvorakz@suse.cz>
1892 * ipa-reference.c (analyze_function): Consider also addresses taken
1895 2007-01-12 Roger Sayle <roger@eyesopen.com>
1897 * c-typeck.c (null_pointer_constant_p): Replace use of
1898 TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
1899 (build_c_cast): Likewise.
1901 2007-01-12 Roger Sayle <roger@eyesopen.com>
1903 * tree.h (force_fit_type_double): Remove unused final argument.
1904 * c-common.c (constant_expression_warning): Replace use of
1905 TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
1906 (convert_and_check): Likewise.
1907 (shorten_compare): Update call to force_fit_type_double.
1908 (c_common_truthvalue_conversion) <INTEGER_CST>: Use integer_zerop.
1909 * convert.c (convert_to_pointer): Update call to
1910 force_fit_type_double.
1911 * fold-const.c (force_fit_type_double): Remove overflowed_const
1913 (int_const_binop, fold_convert_const_int_from_int,
1914 fold_convert_const_int_from_real, fold_div_compare,
1915 fold_sign_changed_comparison, fold_unary, fold_negate_const,
1916 fold_abs_const, fold_not_const): Remove the final argument from
1917 calls to force_fit_type_double.
1919 2007-01-12 Andrew Pinski <andrew_pinski@playstation.sony.com>
1921 * configure.ac: Set insn to "nop" for spu-*-* also.
1922 * configure: Regenerate.
1924 2007-01-12 Olga Golovanevsky <olga@il.ibm.com>
1926 * builtins.def : Add BUILT_IN_FREE.
1928 2007-01-12 Jan Hubicka <jh@suse.cz>
1930 PR tree-optimization/30443
1931 * tree-inline.c (tree_function_versioning): Do not optimize when
1932 cloning for inlining.
1934 2007-01-12 Zdenek Dvorak <dvorakz@suse.cz>
1936 * doc/loop.texi: Document recording of loop exits.
1937 * cfgloopmanip.c (loopify, duplicate_loop): Use alloc_loop.
1938 (update_single_exits_after_duplication,
1939 update_single_exit_for_duplicated_loop,
1940 update_single_exit_for_duplicated_loops): Removed.
1941 (duplicate_loop_to_header_edge): Do not call
1942 update_single_exits_after_duplication and
1943 update_single_exit_for_duplicated_loops.
1944 (loop_version): Do not update single_exit information.
1945 (fix_loop_structure): Use record_loop_exits instead of
1946 mark_single_exit_loops.
1947 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update
1948 the lists of loop exits.
1949 * cfghooks.c (redirect_edge_and_branch, redirect_edge_and_branch_force,
1950 split_edge, merge_blocks): Update the lists of loop exits.
1951 * modulo-sched.c (sms_schedule): Pass LOOPS_HAVE_RECORDED_EXITS to
1952 loop_optimizer_init.
1953 * loop-init.c (loop_optimizer_init): Call record_loop_exits instead
1954 of mark_single_exit_loops.
1955 (loop_optimizer_finalize): Call release_recorded_exits.
1956 * tree-ssa-loop.c (tree_loop_optimizer_init): Pass
1957 LOOPS_HAVE_RECORDED_EXITS to loop_optimizer_init.
1958 * tree-vectorizer.c (slpeel_tree_duplicate_loop_to_edge_cfg): Do not
1959 update single exit information.
1960 * lambda-code.c (perfect_nestify): Ditto.
1961 * cfgloop.c (flow_loop_free): Destroy the list of exits of the loop.
1962 (mark_single_exit_loops): Removed.
1963 (alloc_loop, loop_exit_hash, loop_exit_eq, loop_exit_free,
1964 get_exit_descriptions, rescan_loop_exit, record_loop_exits,
1965 dump_recorded_exit, dump_recorded_exits, release_recorded_exits): New
1967 (get_loop_exit_edges, single_exit): Use recorded exit lists.
1968 (add_bb_to_loop, remove_bb_from_loops): Update the lists of loop exits.
1969 (verify_loop_structure): Verify consistency of the exit lists.
1970 (flow_loops_find): Use alloc_loop. Initialize exits hash.
1971 (set_single_exit): Removed.
1972 * cfgloop.h (struct loop_exit): New function.
1973 (struct loop): single_exit_ field replaced by exits field.
1974 (LOOPS_HAVE_MARKED_SINGLE_EXITS): Replaced by LOOPS_HAVE_RECORDED_EXITS.
1975 (struct loops): Added exits hash.
1976 (mark_single_exit_loops, set_single_exit): Declaration removed.
1977 (release_recorded_exits, record_loop_exits, rescan_loop_exit): Declare.
1979 2007-01-12 Richard Sandiford <richard@codesourcery.com>
1981 * doc/invoke.texi: Avoid use of @headitem.
1983 2007-01-12 Richard Sandiford <richard@codesourcery.com>
1985 * cse.c (cse_insn): Move HAVE_CC0 code after declarations.
1987 2007-01-12 Richard Sandiford <richard@codesourcery.com>
1989 * doc/install.texi: Fix m68k-*-* anchor and add m68k-*-* to the
1992 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
1993 Richard Sandiford <richard@codesourcery.com>
1995 * doc/invoke.texi: Document -mno-bitfield, -mno-rtd and -mno-short.
1996 * config/m68k/m68k.opt: Resort options.
1997 (mbitfield, mrtd, mshort): Remove RejectNegative properties.
1999 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
2000 Richard Sandiford <richard@codesourcery.com>
2002 * doc/invoke.texi: Document the macros that are defined by
2003 m68k's -mtune and -mhard-float options.
2004 * config/m68k/m68k-protos.h (m68k_cpp_cpu_ident) Declare.
2005 (m68k_cpp_cpu_family): Likewise.
2006 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Add a full set
2007 of __ucfv*__ macros. Define __mcffpu__ if generating code for
2008 ColdFire FPUs. Define __mcf_cpu_* and __mcf_family_* macros.
2009 * config/m68k/m68k.c (m68k_cpp_cpu_ident): New function.
2010 (m68k_cpp_cpu_family): Likewise.
2012 2007-01-12 Richard Sandiford <richard@codesourcery.com>
2014 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Treat all mc68*
2015 macros besides mc68000 as tuning macros. Use a switch statement
2016 to set them and mcpu32.
2018 2007-01-12 Julian Brown <julian@codesourcery.com>
2020 * config/m68k/m68k.h: Use TARGET_68040 instead of TARGET_68040_ONLY.
2021 (TARGET_68040_ONLY): Rename to...
2022 (TARGET_68040): ...this.
2023 * config/m68k/m68k.c: Use TARGET_68040 instead of TARGET_68040_ONLY.
2024 * config/m68k/m68k.md: Likewise.
2026 2007-01-12 Julian Brown <julian@codesourcery.com>
2027 Nathan Sidwell <nathan@codesourcery.com>
2028 Richard Sandiford <richard@codesourcery.com>
2030 * config.gcc (m680[012]0-*-*, m68k*-*-*): Set m68k_cpu_ident to
2031 the -mcpu= argument associated with the --with-cpu setting.
2032 Define M68K_DEFAULT_TUNE to the default -mtune= option,
2033 if different from the one implied by the -mcpu setting.
2034 Accept --with-cpu=FOO if FOO is listed in m68k-devices.def,
2035 using mcpu=FOO as the default CPU option. Set target_cpu_default2.
2036 * doc/invoke.texi: Mention ColdFire in the introduction to the
2037 m68k options. Document the new -march, -mcpu, -mtune, -mdiv,
2038 -mno-div and -mhard-float options. Make -m68881 a synonym for
2039 -mhard-float. Document the previously-undocumented -m5206e,
2040 -m528x, -m5307 and -m5407 options. Tweak the existing option
2041 documentation for consistency.
2042 * doc/install.texi: Mention new --with-cpu arguments.
2043 * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Only use the
2044 default CPU if neither -mcpu nor -march are specified.
2045 (ASM_CPU_SPEC): Pass down -mcpu and -march options.
2046 (TARGET_CPU_CPP_BUILTINS): Set __mcfisa*__ macros from
2047 TARGET_ISA*. Set the legacy __mcf*__ cpu macros in the same way,
2048 using m68k_tune to decide between families that implement the
2049 same ISA. Use m68k_tune to set __mcfv4e__.
2050 (FL_BITFIELD, FL_68881, FL_COLDFIRE, FL_CF_HWDIV, FL_CF_MAC)
2051 (FL_CF_EMAC, FL_CF_EMAC_B, FL_CF_USP, FL_CF_FPU, FL_ISA_68000)
2052 (FL_ISA_68010, FL_ISA_68020, FL_ISA_68040, FL_ISA_A, FL_ISA_B)
2053 (FL_ISA_C, FL_ISA_MMU): New macros.
2054 (MASK_COLDFIRE): Delete.
2055 (TARGET_68010, TARGET_68020, TARGET_68040_ONLY, TARGET_COLDFIRE)
2056 (TARGET_ISAB): Redefine in terms of m68k_cpu_flags.
2057 (TARGET_68881, TARGET_COLDFIRE_FPU): Redefine in terms of m68k_fpu.
2058 (TARGET_HARD_FLOAT): Do not define here.
2059 (TARGET_ISAAPLUS, TARGET_ISAC): New macros.
2060 (TUNE_68000): New macro.
2061 (TUNE_68000_10): Redefine in terms of TUNE_68000 and TUNE_68010.
2062 (TUNE_68010, TUNE_68030, TUNE_68040, TUNE_68060, TUNE_CPU32)
2063 (TUNE_CFV2): Redefine in terms of m68k_tune.
2064 (uarch_type, target_device, fpu_type): New enums.
2065 (m68k_cpu, m68k_tune, m68k_fpu, m68k_cpu_flags): Declare.
2066 * config/m68k/m68k.c (TARGET_DEFAULT): Remove MASK_68881.
2067 (FL_FOR_isa_00, FL_FOR_isa_10, FL_FOR_isa_20, FL_FOR_isa_40)
2068 (FL_FOR_isa_cpu32, FL_FOR_isa_a, FL_FOR_isa_aplus, FL_FOR_isa_b)
2069 (FL_FOR_isa_c): New macros.
2070 (m68k_isa): New enum.
2071 (m68k_target_selection): New structure.
2072 (all_devices, all_isas, all_microarchs): New tables.
2073 (m68k_cpu_entry, m68k_arch_entry, m68k_tune_entry, m68k_cpu)
2074 (m68k_tune, m68k_fpu, m68k_cpu_flags): New variables.
2075 (MASK_ALL_CPU_BITS): Delete.
2076 (m68k_find_selection): New function.
2077 (m68k_handle_option): Handle -mcpu=, -march= and -mtune=.
2078 Map the legacy target options to a combination of the new ones.
2079 (override_options): Set m68k_cpu, m68k_tune, m68k_fpu and
2080 m68k_cpu_flags. Handle M68K_DEFAULT_TUNE. Use m68k_cpu_flags
2081 to derive default MASK_BITFIELD, MASK_CF_HWDIV and MASK_HARD_FLOAT
2083 * config/m68k/m68k.opt (m5200, m5206e, m528x, m5307, m5407, mcfv4e)
2084 (m68010, m68020, m68020-40, m68020-60, m68030, m68040): Remove Mask
2086 (m68881, msoft-float): Change mask from 68881 to HARD_FLOAT.
2087 (march=, mcpu=, mdiv, mhard-float, mtune=): New options.
2088 * config/m68k/m68k-devices.def: New file.
2090 2007-01-12 Richard Sandiford <richard@codesourcery.com>
2091 Nathan Sidwell <nathan@codesourcery.com>
2093 * config/m68k/m68k.h (ASM_CPU_SPEC, ASM_SPEC, EXTRA_SPECS)
2094 (SUBTARGET_EXTRA_SPECS): New macros.
2095 * config/m68k/linux.h (ASM_SPEC): Remove CPU flags;
2096 use %(asm_cpu_spec) instead.
2097 * config/m68k/m68k-none.h (ASM_SPEC): Likewise.
2098 * config/m68k/openbsd.h (ASM_SPEC): Likewise.
2099 * config/m68k/netbsd-elf.h (ASM_SPEC): Likewise.
2100 (EXTRA_SPECS): Rename to...
2101 (SUBTARGET_EXTRA_SPECS): ...this.
2103 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
2104 Richard Sandiford <richard@codesourcery.com>
2105 Julian Brown <julian@codesourcery.com>
2107 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
2108 (m68k-*-uclinux*, m68k-*-linux*, m68k-*-rtems*): Set default_m68k_cpu
2109 to the configuration's default CPU.
2110 (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*): Likewise.
2111 Remove default masks.
2112 (m680[012]0-*-*): Set the default with_cpu to the first part of
2114 (m68k*-*-*): Set the default with_cpu to m$default_m68k_cpu.
2115 (m68k*-*-linux): Extend the --with-cpu handling to...
2116 (m680[012]0-*-*, m68k*-*-*): ...these configurations. Allow m68000
2117 and m68010. Don't set target_cpu_default2.
2118 * doc/install.texi: Document --with-cpu for m68k.
2119 * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Define.
2120 * config/m68k/m68k-none.h (TARGET_CPU_DEFAULT, M68K_CPU_m68k)
2121 (M68K_CPU_m68000, M68K_CPU_m68010, M68K_CPU_m68020, M68K_CPU_m68030)
2122 (M68K_CPU_m68040, M68K_CPU_m68302, M68K_CPU_m68332, TARGET_DEFAULT)
2123 (ASM_CPU_DEFAULT_SPEC, CC1_CPU_DEFAULT_SPEC): Delete.
2124 (ASM_SPEC): Remove use of %(asm_cpu_default).
2125 (EXTRA_SPECS, SUBTARGET_EXTRA_SPECS, MULTILIB_DEFAULTS): Delete.
2126 * config/m68k/linux.h (TARGET_DEFAULT): Delete.
2127 (CPP_SPEC): Merge definitions. Do not handle __HAVE_68881__ here.
2128 * config/m68k/netbsd-elf.h (TARGET_OS_CPP_BUILTINS): Define
2129 __HAVE_FPU__ if TARGET_HARD_FLOAT.
2130 (TARGET_DEFAULT): Delete.
2131 (EXTRA_SPECS): Delete cpp_cpu_default_spec, cpp_cpu_spec,
2132 cpp_fpu_spec, asm_default_spec and netbsd_cpp_spec.
2133 (CPP_CPU_SPEC): Delete.
2134 (TARGET_VERSION): Merge definitions, using TARGET_68010 to pick
2135 the appropriate string.
2136 (CPP_CPU_DEFAULT_SPEC, ASM_DEFAULT_SPEC, CPP_FPU_SPEC): Delete.
2137 (CPP_SPEC): Define to NETBSD_CPP_SPEC.
2138 (ASM_SPEC): Don't use %(asm_default_spec).
2139 * config/m68k/m68k.c (TARGET_DEFAULT_TARGET_FLAGS): Remove
2140 TARGET_DEFAULT and add MASK_68881.
2141 * config/m68k/m68k.md: Remove mention of TARGET_DEFAULT from comments.
2143 2007-01-12 Richard Sandiford <richard@codesourcery.com>
2145 * config.gcc (m68010-*-netbsdelf*): Add MASK_68010.
2146 (m68k*-*-netbsdelf*, m68k*-*-openbsd*, m68k*-linux*): Add
2147 MASK_68010 alongside MASK_68020.
2148 * doc/invoke.texi: Document -m68010.
2149 * config/m68k/m68k.opt (m68010): New.
2150 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Define mc68010
2152 (TUNE_68010): New macro.
2153 * config/m68k/m68k-none.h (M68K_CPU_m68k, M68K_CPU_m68010)
2154 (M68K_CPU_m68020, M68K_CPU_m68030, M68K_CPU_m68040)
2155 (M68K_CPU_m68332): Add MASK_68010.
2156 * config/m68k/linux.h (TARGET_DEFAULT): Add MASK_68010 to
2157 fallback definition.
2158 * config/m68k/netbsd-elf.h (CPP_CPU_SPEC): Remove now-redundant
2160 * config/m68k/m68k.c (MASK_ALL_CPU_BITS): Add MASK_68010.
2161 (m68k_handle_option): Handle OPT_m68010. Add MASK_68010
2162 to all entries that use MASK_68020.
2163 (output_move_simode_const, output_move_himode, output_move_qimode)
2164 (output_move_stricthi, output_move_strictqi): Use TARGET_68010
2165 instead of TARGET_68020 to select clr behavior. Remove comment
2166 about there being no TARGET_68010.
2167 * config/m68k/m68k.md: Likewise throughout.
2169 2007-01-12 Julian Brown <julian@codesourcery.com>
2171 * config/m68k/m68k.h (TARGET_ISAB): New macro.
2172 * config/m68k/m68k.c: Use TARGET_ISAB rather than TARGET_CFV4.
2173 * config/m68k/m68k.md: Likewise.
2175 2007-01-12 Julian Brown <julian@codesourcery.com>
2177 * config/m68k/m68k.h (LEGITIMATE_INDEX_P, LEGITIMIZE_ADDRESS): Use
2178 TARGET_COLDFIRE_FPU instead of TARGET_CFV4E.
2180 2007-01-12 Julian Brown <julian@codesourcery.com>
2182 * config/m68k/m68k.h (TUNE_68040_60): New macro.
2183 * config/m68k/m68k.c (standard_68881_constant_p): Use it.
2184 * config/m68k/m68k.md: Likewise.
2186 2007-01-12 Julian Brown <julian@codesourcery.com>
2187 Richard Sandiford <richard@codesourcery.com>
2189 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Use TUNE_68030
2190 instead of TARGET_68030, TUNE_68040 instead of TARGET_68040,
2191 TUNE_68060 instead of TARGET_68060 and TUNE_CPU32 instead of
2193 (TARGET_CPU32): Rename to...
2194 (TUNE_CPU32): ...this.
2195 (TUNE_68000_10, TUNE_68030, TUNE_68040, TUNE_68060)
2196 (TUNE_CFV2): New macros.
2197 * config/m68k/netbsd-elf.h (LONG_DOUBLE_TYPE_SIZE): Simplify;
2198 remove conditions that are implied by TARGET_68020.
2199 * config/m68k/m68k.c (m68k_output_function_prologue): Use TUNE_68040
2200 instead of TARGET_68040 and TUNE_CPU32 instead of TARGET_CPU32.
2201 (m68k_output_function_epilogue): Likewise.
2202 (m68k_rtx_costs): Likewise. Use TUNE_68060 instead of TARGET_68060
2203 and TUNE_CFV2 instead of TARGET_5200. Use TUNE_68000_10 instead of
2204 "!TARGET_68020 && !TARGET_COLDFIRE" to choose between 68000 and
2205 non-68000 timings. Refactor multiplication and division costs.
2206 (output_addsi3): Use TUNE_68040 instead of TARGET_68040 and
2207 TUNE_CPU32 instead of TARGET_CPU32.
2208 (standard_68881_constant_p): Use TUNE_68040 instead of TARGET_68040
2209 and TUNE_68060 instead of TARGET_68060.
2210 * config/m68k/m68k.md: Use TUNE_68040 instead of TARGET_68040,
2211 TUNE_68060 instead of TARGET_68060, and TUNE_CPU32 instead of
2213 (movsi_const0): Use TUNE_68000_10 rather than "!TARGET_68020
2214 && !TARGET_COLDFIRE" to choose between moveq and clr.
2215 Likewise in the unnamed movsf pattern.
2216 (ashlsi_17_24, lshrsi_17_24): Guard with TUNE_68000_10 rather than
2217 "!TARGET_68020 && !TARGET_COLDFIRE". Likewise the unnamed
2220 2007-01-12 Richard Sandiford <richard@codesourcery.com>
2222 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Increase amount
2223 of tabbing before backslashes.
2225 2007-01-11 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
2227 * pa-linux.h (ASM_OUTPUT_INTERNAL_LABEL): Undefine.
2228 * pa.h (ASM_OUTPUT_LABEL): Output colon when using GAS.
2229 (ASM_OUTPUT_INTERNAL_LABEL): Define.
2231 2007-01-11 Zdenek Dvorak <dvorakz@suse.cz>
2233 * tree-ssa-loop-ivopts.c (extract_cond_operands): Split from
2234 find_interesting_uses_cond.
2235 (find_interesting_uses_cond): Use extract_cond_operands.
2236 (rewrite_use_compare): Use extract_cond_operands and
2237 force_gimple_operand_bsi. Do not call update_stmt.
2238 (determine_use_iv_cost_condition): Use extract_cond_operands.
2239 Return cheaper of using original bound and changing the exit bound.
2241 2007-01-11 Zdenek Dvorak <dvorakz@suse.cz>
2243 PR tree-optimization/29516
2244 * tree-ssa-address.c (tree_mem_ref_addr, add_to_parts,
2245 most_expensive_mult_to_index, addr_to_parts,
2246 create_mem_ref, maybe_fold_tmr): Make the type of
2247 fields of TARGET_MEM_REF sizetype.
2248 (move_fixed_address_to_symbol, move_pointer_to_base):
2250 * tree.def (TARGET_MEM_REF): Add comment on types of
2253 2007-01-11 Joseph Myers <joseph@codesourcery.com>
2255 * c-common.c (vector_types_convertible_p): Treat opaque types as
2256 always convertible if they have the same size, but not otherwise.
2258 2007-01-11 Steven Bosscher <steven@gcc.gnu.org>
2260 * ifcvt.c (struct noce_if_info): Add comments to the fields.
2261 Remove the b_unconditional field.
2262 (noce_try_sign_mask): Do not look at b_unconditional.
2263 (noce_process_if_block): Do not use merge_if_blocks. Update
2264 the CFG here. Do not set b_unconditional.
2265 (cond_move_process_if_block): Likewise.
2266 (find_cond_trap): Likewise.
2267 (check_cond_move_block): Require simple jump insns at the end
2270 2007-01-11 Jan Hubicka <jh@suse.cz>
2272 PR tree-optimization/1046
2273 * tree-tailcall.c (suitable_for_tail_call_opt_p): Use TREE_ADDRESSABLE
2274 when alias info is not ready.
2275 (pass_tail_recursion): Do not require aliasing.
2276 * tree-ssa-copyrename.c (pass_rename_ssa_cop): Likewise.
2277 * tree-ssa-ccp.c (pass_ccp, pass_fold_builtins): Likewise.
2278 * tree-ssa-copy.c (pass_copy_prop): Likewise.
2279 * tree-ssa-forwprop.c (pass_forwprop): Likewise.
2280 * tree-ssa-dce.c (pass_dce, pass_dce_loop, pass_cd_dce): Likewise.
2281 * passes.c (init_optimization_passes): Execute rename_ssa_copies,
2282 ccp, forwprop, copy_prop, merge_phi, copy_prop, dce and tail recursion
2284 * tree-ssa-operands.c (add_virtual_operand, get_indirect_ref_operand):
2285 When aliasing is not build, mark statement as volatile.
2287 2007-01-11 Tom Tromey <tromey@redhat.com>
2289 PR preprocessor/15185, PR preprocessor/20989:
2290 * doc/cppopts.texi <-MT>: Update description of algorithm for
2291 computing default target.
2292 <-M, -MD>: Reword "basename" text.
2294 2007-01-11 Roger Sayle <roger@eyesopen.com>
2296 * builtins.c (expand_builtin_pow, expand_builtin_powi,
2297 fold_builtin_cabs, fold_builtin_sqrt, fold_builtin_trunc,
2298 fold_builtin_floor, fold_builtin_ceil, fold_builtin_round,
2299 fold_builtin_int_int_roundingfn, fold_builtin_bitop,
2300 fold_builtin_bswap, real_constp, fold_builtin_pow,
2301 fold_builtin_powi, fold_builtin_signbit, fold_builtin_copysign,
2302 do_mpfr_arg1, do_mpfr_arg2, do_mpfr_arg3, do_mpfr_sincos): Replace
2303 uses of the macro TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
2304 * convert.c (convert_to_pointer): Likewise.
2305 * expr.c (highest_pow2_factor, expand_expr_real_1): Likewise.
2306 * fold-const.c (force_fit_type, fold_negate_expr, int_const_binop,
2307 const_binop, fold_convert_const_int_from_int,
2308 fold_convert_const_int_from_real,
2309 fold_convert_const_real_from_real, sign_bit_p,
2310 optimize_minmax_comparison, extract_muldiv_1, fold_div_compare,
2311 fold_sign_changed_comparison, fold_unary, fold_comparison,
2312 fold_binary, multiple_of_p, tree_Expr_non_zero_p,
2313 fold_negate_const, fold_abs_const, fold_not_const): Likewise.
2314 * print-tree.c (print_node_brief, print_node): Likewise.
2315 * stor-layout.c (place_field, layout_type): Likewise.
2316 * tree-chrec.c (keep_cast): Likewise.
2317 * tree.c (build_vector, build_real, build_real_from_int_cst,
2318 build_complex): Likewise.
2320 2007-01-11 Roger Sayle <roger@eyesopen.com>
2322 * tree.h (TREE_CONSTANT_OVERFLOW): Obsolete. For the time being,
2323 treat TREE_CONSTANT_OVERFLOW as a synonym of TREE_OVERFLOW.
2325 2007-01-11 Paolo Bonzini <bonzini@gnu.org>
2327 * configure.ac (strict1_warn): Rename to strict_warn.
2328 (WERROR, --enable-werror, symlink hacks, stage1_cflags,
2329 cc_set_by_configure, quoted_cc_set_by_configure,
2330 stage_prefix_set_by_configure, quoted_stage_prefix_set_by_configure,
2331 all_boot_languages, all_stagestuff): Remove.
2332 (target_list): Remove bootstrap targets.
2333 * Makefile.in (quickstrap): Unconditionally make a synonym of all.
2334 (BOOT_LANGUAGES, STAGE1_CFLAGS, STAGE1_CHECKING,
2335 REMAKEFLAGS, FLAGS_TO_PASS, PREPEND_DOTDOT_TO_RELATIVE_PATHS,
2336 SUBDIR_FLAGS_TO_PASS, WERROR_FLAGS, STRICT2_WARN, LANG_STAGESTUFF,
2337 VOL_FILES, POSTSTAGE1_FLAGS_TO_PASS, STAGE2_FLAGS_TO_PASS,
2338 STAGEPROFILE_FLAGS_TO_PASS, STAGEFEEDBACK_FLAGS_TO_PASS, stage1_build,
2339 stage1_copy, stage2_build, stage2_copy, stageprofile_build,
2340 stageprofile_copy, stage3_build, stage3_copy, stagefeedback_build,
2341 stagefeedback_copy, stage4_build, clean_s1, clean_sw, bootstrap,
2342 bootstrap-lean, bootstrap2, bootstrap2-lean, bootstrap3,
2343 bootstrap3-lean, bootstrap4, bootstrap4-lean, unstage1, unstage2,
2344 unstage3, unstage4, unstageprofile, unstagefeedback, restage, restage2,
2345 restage3, restage4, restageprofile, restagefeedback, bubbleestrap,
2346 cleanstrap, unstrap, restrap, *compare, *compare3, *compare4,
2347 *compare-lean, *compare3-lean, *compare4-lean, stage1-start, stage1,
2348 stage2-start, stage2, stage3-start, stage3, stage4-start, stage4,
2349 stageprofile-start, stageprofile, stagefeedback-start, stagefeedback,
2350 risky-stage1, risky-stage2, risky-stage3, risky-stage4): Remove.
2351 (ORDINARY_FLAGS_TO_PASS): Rename to FLAGS_TO_PASS.
2352 (STAGECOPYSTUFF, STAGEMOVESTUFF): Consolidate into MOSTLYCLEANFILES.
2353 (mostlyclean): Adjust.
2354 (clean, distclean): Don't mention bootstrap stuff.
2355 * configure: Regenerate.
2356 * ada/config-lang.in, cp/config-lang.in, forttran/config-lang.in,
2357 java/config-lang.in, objc/config-lang.in, objcp/config-lang.in,
2358 treelang/config-lang.in (stagestuff): Remove.
2359 * doc/sourcebuild.texi (stage1, stage2, stage3, stage4,
2360 stageprofile, stagefeedback, stagestuff): Remove mention.
2362 2007-01-11 Nick Clifton <nickc@redhat.com>
2364 * config/mcore/predicates.md (mcore_general_movesrc_operand):
2366 (mcore_general_movdst_operand): Do not accept CONST_INTs.
2367 (mcore_arith_K_S_operand): Run the test for the S constraint not
2368 the test for the M constraint.
2369 (mcore_addsub_operand): Do not accept integer values that are
2370 larger than 32 bits.
2371 * config/mcore/mcore.md: Remove unused constraints from split.
2372 (andsi3): Use HOST_WIDE_INT instead of int to hold an INTVAL.
2374 (allocate_stack): Likewise.
2375 * config/mcore/mcore.c (mcore_print_operand): Restrict output of P
2376 operands to 32 bits.
2377 (mcore_const_costs): Use HOST_WIDE_INT instead of int to hold an
2379 (mcore_and_cost, mcore_modify_comparison, const_ok_for_mcore,
2380 mcore_const_ok_for_inline, mcore_const_trick_uses_not,
2381 try_constant_tricks, mcore_num_ones, mcore_num_zeros,
2382 mcore_output_bclri, mcore_output_andn, output_inline_const,
2383 mcore_output_move, mcore_output_movedouble): Likewise.
2384 (mcore_output_cmov): Use CONST_OK_FOR_M and CONST_OK_FOR_N.
2385 (output_inline_const): Likewise.
2386 (output_inline_const): Fix format strings used in sprintf
2388 * config/mcore/mcore-protos.h: Update prototypes for changed
2389 functions in mcore.c.
2390 * config/mcore/mcore.h (CONST_OK_FOR_I): Cast values to
2391 HOST_WIDE_INT and not int.
2392 (CONST_OK_FOR_J, CONST_OK_FOR_K, CONST_OK_FOR_L, CONST_OK_FOR_M,
2393 CONST_OK_FOR_N): Likewise.
2394 (LEGITIMATE_CONSTANT_P): Also check CONSTANT_P.
2395 (GO_IF_LEGITIMATE_INDEX): Use HOST_WIDE_INT instead of int to hold
2398 2007-01-10 Jan Hubicka <jh@suse.cz>
2400 * tree-vrp.c (remove_range_assertions): Release defs.
2401 * tree-ssa-loop-ivopts.c (rmeove_statement): Likewise.
2402 * tree-ssa-dom.c (remove_stmt_or_phi): Likewise.
2404 2007-01-10 Paul Brook <paul@codesourcery.com>
2406 * config/arm/arm.c (arm_rtx_costs_1): Handle mutiply-subtract.
2407 * config/arm/arm.md (mulsi3subsi): New insn.
2409 2007-01-10 Zdenek Dvorak <dvorakz@suse.cz>
2411 * tree-ssa-loop-manip.c (tree_unroll_loop): Make it a wrapper over ...
2412 (tree_transform_and_unroll_loop): New.
2413 * tree-flow.h (transform_callback, tree_transform_and_unroll_loop):
2416 2007-01-10 Robert Kennedy <jimbob@google.com>
2418 * fold-const.c (fold_comparison): Fold comparisons like (x *
2419 1000 < 0) to (x < 0).
2421 2007-01-10 Ian Lance Taylor <iant@google.com>
2423 * tree-pretty-print.c (dump_generic_node): Print parentheses when
2424 operands have the same priority.
2426 2007-01-10 Tom Tromey <tromey@redhat.com>
2428 * fold-const.c (fold_truthop): Don't check can_use_bit_fields_p.
2429 (fold_binary): Likewise.
2430 * langhooks.c (lhd_can_use_bit_fields_p): Removed.
2431 * langhooks-def.h (lhd_can_use_bit_fields_p): Removed.
2432 (LANG_HOOKS_CAN_USE_BIT_FIELDS_P): Removed.
2433 (LANG_HOOKS_INITIALIZER): Remove LANG_HOOKS_CAN_USE_BIT_FIELDS_P.
2434 * langhooks.h (struct lang_hooks): Removed field
2435 'can_use_bit_fields_p'.
2437 2007-01-10 Ralf Corsépius <ralf.corsepius@rtems.org>
2439 * config/bfin/t-bfin, config/bfin/t-bfin-elf: Remove GCC_CFLAGS.
2441 2007-01-10 Razya Ladelsky <razya@il.ibm.com>
2443 * function.c (get_last_funcdef_no): New function.
2444 * function.h (get_last_funcdef_no): Declare.
2445 * tree-inline.c (initialize_cfun): Add initialization.
2446 (tree_function_versioning): Cleanup.
2448 2007-01-10 Jan Hubicka <jh@suse.cz>
2450 * tree-inline.c (setup_one_parameter): Do not propagate into abnormal
2453 2007-01-10 Sa Liu <saliu@de.ibm.com>
2454 Ben Elliston <bje@au.ibm.com>
2456 * spu.h (STACK_SAVE_AREA): Use VOIDmode for SAVE_FUNCTION, SImode
2457 for SAVE_NONLOCAL and Pmode for any other save level.
2458 * spu-protos.h (spu_restore_stack_block): Declare.
2459 * spu.md (save_stack_block): Remove.
2460 (restore_stack_block): Call spu_restore_stack_block.
2461 * spu.c (spu_restore_stack_block): New function.
2462 (spu_expand_epilogue): Remove old comment.
2464 2007-01-09 Zdenek Dvorak <dvorakz@suse.cz>
2466 PR tree-optimization/30322
2467 * tree-ssa-loop-ivopts.c (fold_affine_expr, iv_value): Removed.
2468 (cand_value_at): Return the value as aff_tree.
2469 (may_eliminate_iv): Convert the bound from aff_tree to tree.
2470 * tree-affine.c (aff_combination_add_cst, aff_combination_add_product,
2471 aff_combination_mult): New functions.
2472 (aff_combination_add): Use aff_combination_add_cst.
2473 (aff_combination_convert): Allow conversions to a wider type.
2474 (tree_to_aff_combination): Handle BIT_NOT_EXPR.
2475 * tree-affine.h (aff_combination_mult): Declare.
2477 2007-01-09 Carlos O'Donell <carlos@codesourcery.com>
2479 * doc/tm.texi: Update documentation to reflect reality of exec
2480 and start file search behaviours. Update copyright year.
2481 * doc/invoke.texi: Explain how GCC_EXEC_PREFIX is used to find
2482 header file directories.
2484 2007-01-09 Uros Bizjak <ubizjak@gmail.com>
2486 * config/i386/i386.md (*sinxf2): Rename to *sinxf2_i387.
2487 (*cosxf2): Rename to cosxf2_i387.
2488 (*sindf2, *sinsf2): Extend operand 1 to XFmode. Macroize patterns
2489 using X87MODEF12 mode macro. Rename patterns to
2490 *sin_extend<mode>xf2_i387. Use SSE_FLOAT_MODE_P to disable patterns
2492 (*cosdf2, *cossf2): Ditto.
2493 (sincosdf3, sincossf3): Ditto. Rewrite corresponding splitters
2494 to match extended input operands.
2495 (sincos<mode>3): New expander.
2496 (*sinextendsfdf2, *cosextendsfdf2, *sincosextendsfdf3): Remove
2497 insn patterns and corresponding splitters.
2499 2007-01-09 Kaz Kojima <kkojima@gcc.gnu.org>
2501 * config/sh/t-linux (TARGET_LIBGCC2_CFLAGS): Delete.
2502 (SHLIB_MAPFILES, SHLIB_LINK, SHLIB_INSTALL): Likewise.
2504 2007-01-09 Nicolas Pitre <nico@cam.org>
2507 * arm/ieee754-df.S (Lad_s): Also test the low word of X for zero.
2509 2007-01-08 Geoffrey Keating <geoffk@apple.com>
2511 * target.h (struct gcc_target): New field library_rtti_comdat.
2512 * target-def.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): New.
2513 (TARGET_CXX): Add TARGET_CXX_LIBRARY_RTTI_COMDAT.
2514 * doc/tm.texi (C++ ABI): Document TARGET_CXX_LIBRARY_RTTI_COMDAT.
2515 * config/darwin.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
2517 2007-01-08 Geoffrey Keating <geoffk@apple.com>
2519 * doc/invoke.texi (Optimize Options): Correct description of -O0.
2521 2007-01-08 Richard Guenther <rguenther@suse.de>
2523 * tree.h (force_fit_type_double): Export.
2524 (force_fit_type): Remove.
2525 * fold-const.c (force_fit_type_double): New function.
2526 (force_fit_type): Remove.
2527 (int_const_binop): Use it.
2528 (fold_convert_const_int_from_int): Likewise.
2529 (fold_convert_const_int_from_real): Likewise.
2530 (fold_div_compare): Likewise.
2531 (fold_sign_changed_comparison): Likewise.
2532 (fold_unary): Likewise.
2533 (fold_negate_const): Likewise.
2534 (fold_abs_const): Likewise.
2535 (fold_not_const): Likewise.
2536 * c-common.c (shorten_compare): Use force_fit_type_double.
2537 * convert.c (convert_to_pointer): Likewise.
2539 2007-01-08 Richard Guenther <rguenther@suse.de>
2541 * tree.h (build_int_cst_wide_type): Export.
2542 * tree.c (build_int_cst_wide_type): New function.
2543 (build_int_cst_wide): Fix comment.
2544 * builtins.c (fold_builtin_object_size): Use build_int_cst
2545 to build -1 or 0 of the correct type. Use fit_double_type
2546 to check for overflow.
2547 * fold-const.c (optimize_bit_field_compare): Use build_int_cst_type
2549 (decode_field_reference): Likewise.
2550 (all_ones_mask_p): Likewise.
2551 (native_interpret_int): Use build_int_cst_wide_type.
2552 (fold_binary): Use build_int_cst_type to build an all-ones
2554 * stor-layout.c (set_sizetype): Use build_int_cst_wide_type.
2556 2007-01-08 Daniel Jacobowitz <dan@codesourcery.com>
2558 * config/pa/t-pa64 (libgcc_stub.a): Use $(T).
2560 2007-01-09 Ben Elliston <bje@au.ibm.com>
2562 * genautomata.c (STATS_OPTION): New option.
2563 (stats_flag): New flag.
2564 (gen_automata_option): Handle it.
2565 (initiate_automaton_gen): Ditto.
2566 (write_automata): Output statistics only if stats_flag is
2567 set. Likewise, output time statistics only if time_flag is set.
2568 * doc/md.texi (Processor pipeline description): Document new flag.
2570 2007-01-08 Richard Guenther <rguenther@suse.de>
2572 * builtins.c (fold_builtin_int_roundingfn): Use fit_double_type.
2573 * tree.c (build_int_cst_type): Likewise.
2574 (size_in_bytes): Don't call force_fit_type on the result.
2575 (int_fits_type_p): Use fit_double_type.
2576 * fold-const.c (fit_double_type): New function.
2577 (force_fit_type): Use it.
2578 * tree.h (fit_double_type): Export.
2580 2007-01-08 Jan Hubicka <jh@suse.cz>
2582 * tree-vectorizer.c (gate_increase_alignment): Fix return type.
2583 * ipa.c (function_and_variable_visibility): Fix return type.
2585 2007-01-08 Richard Guenther <rguenther@suse.de>
2587 * tree-ssa-ccp.c (maybe_fold_offset_to_array_ref): Use type
2588 of offset to build the index.
2589 * tree-pretty-print.c (dump_generic_node): Don't build negated
2590 const just for printing.
2591 * c-pretty-print.c (pp_c_integer_constant): Likewise.
2592 * builtins.c (fold_builtin_int_roundingfn): Check if result
2593 fits the type by using force_fit_type and comparing the result.
2594 * predict.c (predict_loops): Use compare_tree_int for comparison.
2595 * tree.c (build_int_cst): Fall back to integer_type_node for
2597 (build_int_cst_wide): Assert type is non-null.
2599 2007-01-08 Roberto Costa <roberto.costa@st.com>
2601 * tree-vrp.c (extract_range_from_cond_expr): New.
2602 (extract_range_from_expr): Handle COND_EXPR nodes used as expressions.
2603 * tree-ssa-ccp.c (get_maxval_strlen): Handle COND_EXPR nodes used
2605 (fold_stmt): Bug fix, avoid infinite recursion when folding COND_EXPRs.
2606 * tree-ssa-forwprop.c (simplify_cond, forward_propagate_into_cond,
2607 tree_ssa_forward_propagate_single_use_vars): Handle COND_EXPR nodes
2608 used as expressions.
2609 * tree-object-size.c (cond_expr_object_size): New.
2610 (collect_object_sizes_for): Handle COND_EXPR nodes used as expressions.
2612 2007-01-08 Jan Hubicka <jh@suse.cz>
2614 * tree-ssa-forwprop.c (forward_propagate_into_cond,
2615 tree_ssa_forward_propagate_single_use_va): Release defs of propagated
2618 2007-01-08 Richard Guenther <rguenther@suse.de>
2620 PR tree-optimization/23603
2621 * tree-vrp.c (set_value_range_to_truthvalue): New function.
2622 (extract_range_from_binary): Fall back to truthvalue instead of
2623 varying for TRUTH_*_EXPR.
2624 (extract_range_from_comparison): Fall back to truthvalue instead of
2626 (vrp_visit_phi_node): Don't adjust new range bounds to +INF/-INF
2627 if all visited PHI values were constant.
2629 2007-01-08 Jan Hubicka <jh@suse.cz>
2631 * cgraphunit.c (cgraph_process_new_functions): Reset reachable flag.
2632 (cgraph_analyze_function): break out from ...
2633 (cgraph_finalize_compilation_unit): ... here.
2634 (cgraph_expand_function): Remove forgoten commented out line.
2635 (cgraph_optimize): Analyze functions.
2637 2007-01-08 Jan Hubicka <jh@suse.cz>
2639 * tree-pas.h (TODO_remove_function): New flag.
2640 (TODO_update*): Renumber.
2641 (pass_ipa_increase_alignment,
2642 pass_ipa_function_and_variable_visibility): New passes.
2643 * cgraphunit.c (cgraph_increase_alignment): Move to tree-vectorizer.c
2644 (cgraph_function_and_variable_visibility): Move to ipa.c
2645 (cgraph_optimize): Don't call cgraph_function_and_variable_visibility,
2646 cgraph_increase_alignment.
2647 * ipa-inline.c (cgraph_decide_inlining): Don't push timevar.
2648 (cgraph_decide_inlining_incrementally): Push TV_INTEGRATION before
2649 calling tree-inline.
2650 (cgraph_early_inlining): Do not call cgraph_remove_unreachable_nodes.
2651 (pass_ipa_inline, pass_early_ipa_inlining): Set TODO_remove_functions
2652 * tree-vectorizer.c (increase_alignment): Move here from cgraphunit.c
2653 (gate_increase_alignment): New function.
2654 (pass_ipa_increase_alignment): New pass.
2655 * ipa.c: Inline tree-pass.h and timevar.h
2656 (function_and_variable_visibility): Move here from cgraphunit.c
2657 * tree-optimize.c (pass_early_local_passes): Add TODO_remove_functions.
2658 * passes.c (init_optimization_passes): Add the two new passes.
2659 (execute_todo): Handle cgraph_remove_functions.
2661 2007-01-08 Nick Clifton <nickc@redhat.com>
2663 * config/frv/predicates.md (reg_or_0_operand): Accept
2666 2007-01-08 Ralf Corsépius <ralf.corsepius@rtems.org>
2668 * config/bfin/rtems.h, config/bfin/t-rtems: New.
2669 * config.gcc: Add bfin*-rtems*.
2671 2007-01-08 Mark Shinwell <shinwell@codesourcery.com>
2673 * c.opt: Add -flax-vector-conversions.
2674 * c-typeck.c (convert_for_assignment): Pass flag to
2675 vector_types_convertible_p to allow emission of note.
2676 (digest_init): Likewise.
2677 * c-opts.c: Handle -flax-vector-conversions.
2678 * c-common.c (flag_lax_vector_conversions): New.
2679 (vector_types_convertible_p): Unless -flax-vector conversions
2680 has been passed, disallow conversions between vectors with
2681 differing numbers of subparts and/or element types. If such
2682 a conversion is disallowed, possibly emit a note on the first
2683 occasion only to inform the user of -flax-vector-conversions.
2684 The new last argument specifies this.
2685 * c-common.h (flag_lax_vector_conversions): New.
2686 (vector_types_convertible_p): Add extra argument.
2687 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Use
2688 char_type_node for V*QI type vectors.
2689 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins):
2690 Update to satisfy new typechecking rules.
2691 * config/rs6000/altivec.h (vec_cmple): Use vec_cmpge, for both
2693 * doc/invoke.texi (C Dialect Options): Document
2694 -flax-vector-conversions.
2696 2007-01-08 Mark Shinwell <shinwell@codesourcery.com>
2698 PR tree-optimization/29877
2699 * tree-ssa-ter.c (is_replaceable_p): Deem assignments with
2700 a register variable on the RHS to not be replaceable.
2702 2007-01-08 Chen Liqin <liqin@sunnorth.com.cn>
2703 * config/score/t-score-elf (MULTILIB_OPTIONS): Change.
2704 * config/score/predicates.md (const_uimm5, sr0_operand, const_simm12,
2705 const_simm15, const_pow2, const_npow2): Added.
2706 * config/score/misc.md (insv, extv, extzv, movmemsi,
2707 move_lbu_a/b, mov_lhu_a/b etc): Added and fix some bug.
2708 * config/score/score.c (score_address_cost, score_select_cc_mode):
2710 Change CONST_OK_FOR_LETTER_P/EXTRA_CONSTRAINT define.
2711 Update score_rtx_costs for MACRO TARGET_RTX_COSTS.
2712 Update score_print_operand.
2713 * config/score/score.h (DATA_ALIGNMENT, SELECT_CC_MODE): Added.
2714 Adjust register allocate order and update some macro define.
2715 * config/score/score-mdaux.c (mdx_unaligned_load, mdx_unsigned_store,
2716 mdx_block_move_straight, mdx_block_move_loop_head,
2717 mdx_block_move_loop_body, mdx_block_move_loop_foot, mdx_block_move_loop,
2718 mdx_block_move): Added.
2719 (mdx_movsicc, mdp_select_add_imm, mdp_select, mds_zero_extract_andi,
2720 mdp_limm): Updated and fix some bug and typo.
2721 * config/score/score.md (movqi/hi/si, add/sub/zero/ext): Updated.
2722 (movsf, movdf, doloop_end): Added.
2724 2007-01-08 Kazu Hirata <kazu@codesourcery.com>
2726 * config/arm/arm.c, config/arm/arm.h, config/arm/arm.md,
2727 config/arm/thumb2.md: Fix comment typos.
2728 * doc/extend.texi: Fix a typo.
2730 2007-01-07 Eric Christopher <echristo@apple.com>
2732 * configure.ac: Check for __stack_chk_fail for darwin.
2733 * configure: Regenerate.
2735 2007-01-07 Richard Guenther <rguenther@suse.de>
2737 * tree-vrp.c (extract_range_from_assert): CSE calls to
2738 compare_values where possible.
2739 (extract_range_from_unary_expr): Likewise.
2741 2007-01-07 Anatoly Sokolov <aesok@post.ru>
2743 * config/avr/avr-protos.h (call_insn_operand): Delete prototype.
2744 * config/avr/avr.c (call_insn_operand): Delete function.
2745 * config/avr/avr.md (*pushqi, *pushhi, *pushsi, *pushsf): Use REG_SP
2746 instead of register number. Use predicates.
2747 * config/avr/predicates.md (const0_operand, reg_or_0_operand,
2748 call_insn_operand): Add.
2750 2007-01-06 Jan Hubicka <jh@suse.cz>
2752 * tree-pass.h (pass_build_cgraph_edges): Declare.
2753 * cgraphunit.c (record_refernece): Move to cgraphbuild.c
2754 (visited_nodes): Remove.
2755 (cgraph_create_edges): Move to cgraphbuild.c; rename to
2756 build_cgrpah_edges; make visited_nodes local.
2757 (cgraph_process_new_functions): DO not call initialize_inline_failed.
2758 (record_references_in_initializer): Move to cgraphbuild.c
2759 (initialize_inline_failed, rebuild_cgraph_edges,
2760 pass_rebuild_cgraph_edges): Move to cgraphbuild.c.
2761 (verify_cgraph_node): Make visited_nodes local.
2762 (cgraph_analyze_function): Do not call cgraph_create_edges and
2763 initialize_inline_failed.
2764 (cgraph_expand_function): Do not call cgraph_lower_function;
2765 assert that function is already lowered.
2766 * Makefile.in (cgraphbuild.o): New.
2767 * passes.c (init_optimization_passes): Add pass_build_cgraph_edges
2768 at the end of lowering passes.
2770 2007-01-06 Steven Bosscher <steven@gcc.gnu.org>
2772 * ifcvt.c (cond_move_convert_if_block): New function, code
2773 factored out from...
2774 (cond_move_process_if_block): ...here. Call the new function
2775 on the THEN and ELSE blocks.
2776 (merge_if_block): Do not copy global_live_at_end, merge_blocks
2777 already takes care of this.
2779 2007-01-05 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
2782 * tree.h (TREE_OVERFLOW_P): New.
2783 * c-typeck.c (parser_build_unary_op): Warn only if result
2784 overflowed and operands did not.
2785 (parser_build_binary_op): Likewise.
2786 (convert_for_assignment): Remove redundant overflow_warning.
2787 * c-common.c (overflow_warning): Don't check or set TREE_OVERFLOW.
2789 2007-01-05 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
2791 * c-typeck.c (store_init_value): Split over two lines to follow
2792 the GNU coding style.
2794 2007-01-05 Benjamin Kosnik <bkoz@redhat.com>
2796 * c-cppbuiltin.c (c_cpp_builtins): __GXX_EXPERIMENTAL_CPP0X__ to
2797 __GXX_EXPERIMENTAL_CXX0X__.
2798 * doc/cpp.texi: Same.
2800 2007-01-05 Richard Guenther <rguenther@suse.de>
2803 * tree.c (get_narrower): Do not construct COMPONENT_REFs
2804 with mismatched types. Instead explicitly build a
2805 conversion NOP_EXPR.
2807 2007-01-05 Ian Lance Taylor <iant@google.com>
2809 * c-common.c (decl_with_nonnull_addr_p): New function.
2810 (c_common_truthvalue_conversion): Call it.
2811 * c-typeck.c (build_binary_op): Likewise.
2812 * c-common.h (decl_with_nonnull_addr_p): Declare.
2814 2007-01-05 Jakub Jelinek <jakub@redhat.com>
2817 * libgcc2.c (__divdc3): Compare c and d against 0.0 instead of
2820 2007-01-05 Joel Brobecker <brobecker@adacore.com>
2822 * doc/install.texi (Final install): Document the fact that
2823 the GNAT runtime should not be stripped.
2825 2007-01-04 Jan Hubicka <jh@suse.cz>
2827 * tree-inline.c (fold_marked_statements): Update operand caches
2828 and EH after folding
2830 2007-01-04 Ian Lance Taylor <iant@google.com>
2832 * c-common.c (check_function_nonnull): Whitespace fix.
2834 2007-01-04 Jan Hubicka <jh@suse.cz>
2836 * tree-optimize.c (execute_fixup_cfg): Correct previously mistakely
2837 comitted older version of patch.
2838 (pass_fixup_cfg): Add TODOs to verify flow and statements, dump
2839 function, celanup cfg and collect garbage.
2841 2007-01-04 Mike Stump <mrs@apple.com>
2843 * Makefile.in (mostlyclean): Don't remove libgcc anymore.
2846 2007-01-04 Eric Christopher <echristo@apple.com>
2848 * libgcc2.c (__bswapsi2): Use SItype.
2849 (__bswapdi2): Use DItype.
2850 * libgcc2.h: Update for above.
2852 2007-01-04 Paul Brook <paul@codesourcery.com>
2854 * config/arm/arm.md (arm_mulsi3, thumb_mulsi3, mulsi3_compare0,
2855 mulsi_compare0_scratch, mulsi3addsi, mulsi3addsi_compare0,
2856 mulsi3addsi_compare0_scratch, mulsidi3adddi, mulsidi3,
2857 umulsidi3, umulsidi3adddi, smulsi3_highpart,
2858 umulsi3_highpart): Make conditional on !arm_arch6.
2859 (arm_mulsi3_v6, thumb_mulsi3_v6, mulsi3_compare0_v6,
2860 mulsi_compare0_scratch_v6, mulsi3addsi_v6, mulsi3addsi_compare0_v6,
2861 mulsi3addsi_compare0_scratch_v6, mulsidi3adddi_v6, mulsidi3_v6,
2862 umulsidi3_v6, umulsidi3adddi_v6, smulsi3_highpart_v6,
2863 umulsi3_highpart_v6): New insns.
2865 2007-01-04 Roger Sayle <roger@eyesopen.com>
2867 * fold-const.c (fold_convert): When casting an expression to void,
2868 fold_ignored_result may discover a GIMPLE_MODIFY_STMT which doesn't
2869 have a type. Instead of attempting to build a NOP_EXPR, return
2870 these "special" trees directly.
2872 2007-01-04 Joseph Myers <joseph@codesourcery.com>
2874 * config/rs6000/rs6000.c (rs6000_rtx_costs): Make adjustment for
2875 MULT inside MINUS as either argument. Use rs6000_cost->dmul -
2876 rs6000_cost->fp not 0 as adjustment for outer NEG.
2878 2007-01-04 Jan Hubicka <jh@suse.cz>
2880 * cgraph.c (cgraph_release_function_body): New function.
2881 (cgraph_remove_node): Use it.
2882 * cgraph.h (cgraph_release_function_body): Declare.
2883 * cgraphunit.c (cgraph_expand_function): Use it.
2884 * ipa.c (cgraph_remove_unreahchable_nodes): Use it.
2885 * tree-ssa.c (delete_tree_ssa): Allow to be called before aliasing
2886 is initialized and while compilation of other function is running.
2887 * tree-optimize.c (execute_free_cfg_annotations): Move code to clear
2888 statement CFG annotations from here to ...
2889 * tree-cfg.c (delete_tree_cfg_annotations): ... here.
2891 2007-01-04 Zdenek Dvorak <dvorakz@suse.cz>
2893 * cfgloop.h (enum li_flags): Make the constants powers of two.
2895 2007-01-04 Jan Hubicka <jh@suse.cz>
2897 * tree-inline.c (copy_bb): Insert new statements to statements_to_fold
2899 (fold_marked_statements): New function.
2900 (optimize_inline_calls, tree_function_versioning): Fold new statements.
2901 * tree-inline.h (copy_body_data): Add statements_to_fold.
2903 2007-01-03 Daniel Jacobowitz <dan@codesourcery.com>
2905 * config.gcc: Mention libgcc/config.host.
2906 * Makefile.in: Update comments mentioning libgcc.
2907 (LIBGCC, INSTALL_LIBGCC, GCC_PARTS, mklibgcc): Delete.
2908 (all.cross, start.encap, rest.encap, rest.cross): Update
2909 dependencies for libgcc move.
2910 (libgcc.mk, LIBGCC_DEPS, libgcov.a, libgcc.a, stmp-multilib)
2911 (clean-target, clean-target-libgcc): Delete.
2912 (srcdirify, GCC_EXTRA_PARTS): New macros.
2913 (libgcc-support, libgcc.mvars): New rules.
2914 (distclean): Remove mention of mklibgcc.
2915 (install): Don't reference INSTALL_LIBGCC.
2916 (install-common): Don't reference EXTRA_PARTS.
2917 (install-libgcc, install-multilib): Delete rules.
2918 * mklibgcc.in: Delete file.
2919 * doc/configfiles.texi: Don't mention mklibgcc.
2921 * config/i386/t-darwin (SHLIB_VERPFX): Delete (moved to libgcc).
2922 * config/i386/t-darwin64 (SHLIB_VERPFX): Likewise.
2923 * config/rs6000/t-darwin (SHLIB_VERPFX): Likewise.
2924 * config/rs6000/t-ppccomm (TARGET_LIBGCC2_CFLAGS, SHLIB_MAPFILES)
2925 (mklibgcc, ldblspecs): Likewise.
2927 * config/i386/t-nwld (libgcc.def, libc.def, libpcre.def)
2928 (posixpre.def): Use $(T).
2929 (SHLIB_EXT, SHLIB_NAME, SHLIB_SLIBDIR_QUAL, SHLIB_DEF, SHLIB_MAP)
2930 (SHLIB_SRC, SHLIB_INSTALL): Delete.
2931 (SHLIB_LINK): Make dummy.
2932 * config/t-slibgcc-darwin: Delete contents except for dummy SHLIB_LINK.
2934 * config/frv/t-linux (EXTRA_MULTILIB_PARTS): Clear.
2936 * config/alpha/t-crtfm: Use $(T) in rules for EXTRA_PARTS.
2937 * config/alpha/t-vms, config/alpha/t-vms64, config/fr30/t-fr30,
2938 config/i386/t-rtems-i386, config/ia64/t-ia64, config/rs6000/t-beos,
2939 config/rs6000/t-newas, config/sparc/t-elf: Likewise.
2941 * configure.ac (all_outputs): Remove mklibgcc.
2942 * configure: Regenerated.
2944 2007-01-03 Josh Conner <jconner@apple.com>
2947 * calls.c (compute_argument_addresses): Set stack and stack_slot
2948 for partial args, too.
2949 (store_one_arg): Use locate.size.constant for the size when
2950 generating a save_area.
2952 2007-01-03 Robert Kennedy <jimbob@google.com>
2954 * tree-cfg.c (tree_merge_blocks): Release SSA_NAME phi results
2955 whose definitions are deleted due to basic block merging.
2957 2007-01-03 Paul Brook <paul@codesourcery.com>
2960 * config/arm/arm.c (output_return_instruction): Pop PC in interrupt
2962 (use_return_insn): Return 0 for Thumb interrupt functions.
2963 (print_multi_reg): Add rfe argument for IRQ returns.
2964 (arm_output_epilogue): Pop interrupt return address directly into PC.
2965 (arm_expand_prologue): Only adjust IRQ return address in Arm mode.
2967 2007-01-03 Paul Brook <paul@codesourcery.com>
2969 Merge from sourcerygxx-4_1.
2970 * config/arm/thumb2.md: New file.
2971 * config/arm/elf.h (JUMP_TABLES_IN_TEXT_SECTION): Return True for
2973 * config/arm/coff.h (JUMP_TABLES_IN_TEXT_SECTION): Ditto.
2974 * config/arm/aout.h (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
2975 (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump tables.
2976 * config/arm/aof.h (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump
2978 (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
2979 * config/arm/ieee754-df.S: Use macros for Thumb-2/Unified asm
2981 * config/arm/ieee754-sf.S: Ditto.
2982 * config/arm/arm.c (thumb_base_register_rtx_p): Rename...
2983 (thumb1_base_register_rtx_p): ... to this.
2984 (thumb_index_register_rtx_p): Rename...
2985 (thumb1_index_register_rtx_p): ... to this.
2986 (thumb_output_function_prologue): Rename...
2987 (thumb1_output_function_prologue): ... to this.
2988 (thumb_legitimate_address_p): Rename...
2989 (thumb1_legitimate_address_p): ... to this.
2990 (thumb_rtx_costs): Rename...
2991 (thumb1_rtx_costs): ... to this.
2992 (thumb_compute_save_reg_mask): Rename...
2993 (thumb1_compute_save_reg_mask): ... to this.
2994 (thumb_final_prescan_insn): Rename...
2995 (thumb1_final_prescan_insn): ... to this.
2996 (thumb_expand_epilogue): Rename...
2997 (thumb1_expand_epilogue): ... to this.
2998 (arm_unwind_emit_stm): Rename...
2999 (arm_unwind_emit_sequence): ... to this.
3000 (thumb2_legitimate_index_p, thumb2_legitimate_address_p,
3001 thumb1_compute_save_reg_mask, arm_dwarf_handle_frame_unspec,
3002 thumb2_index_mul_operand, output_move_vfp, arm_shift_nmem,
3003 arm_save_coproc_regs, thumb_set_frame_pointer, arm_print_condition,
3004 thumb2_final_prescan_insn, thumb2_asm_output_opcode, arm_output_shift,
3005 thumb2_output_casesi): New functions.
3006 (TARGET_DWARF_HANDLE_FRAME_UNSPEC): Define.
3007 (FL_THUMB2, FL_NOTM, FL_DIV, FL_FOR_ARCH6T2, FL_FOR_ARCH7,
3008 FL_FOR_ARCH7A, FL_FOR_ARCH7R, FL_FOR_ARCH7M, ARM_LSL_NAME,
3009 THUMB2_WORK_REGS): Define.
3010 (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv, arm_condexec_count,
3011 arm_condexec_mask, arm_condexec_masklen)): New variables.
3012 (all_architectures): Add armv6t2, armv7, armv7a, armv7r and armv7m.
3013 (arm_override_options): Check new CPU capabilities.
3014 Set new architecture flag variables.
3015 (arm_isr_value): Handle v7m interrupt functions.
3016 (user_return_insn): Return 0 for v7m interrupt functions. Handle
3018 (const_ok_for_arm): Handle Thumb-2 constants.
3019 (arm_gen_constant): Ditto. Use movw when available.
3020 (arm_function_ok_for_sibcall): Return false for v7m interrupt
3022 (legitimize_pic_address, arm_call_tls_get_addr): Handle Thumb-2.
3023 (thumb_find_work_register, arm_load_pic_register,
3024 legitimize_tls_address, arm_address_cost, load_multiple_sequence,
3025 emit_ldm_seq, emit_stm_seq, arm_select_cc_mode, get_jump_table_size,
3026 print_multi_reg, output_mov_long_double_fpa_from_arm,
3027 output_mov_long_double_arm_from_fpa, output_mov_double_fpa_from_arm,
3028 output_mov_double_fpa_from_arm, output_move_double,
3029 arm_compute_save_reg_mask, arm_compute_save_reg0_reg12_mask,
3030 output_return_instruction, arm_output_function_prologue,
3031 arm_output_epilogue, arm_get_frame_offsets, arm_regno_class,
3032 arm_output_mi_thunk, thumb_set_return_address): Ditto.
3033 (arm_expand_prologue): Handle Thumb-2. Use arm_save_coproc_regs.
3034 (arm_coproc_mem_operand): Allow POST_INC/PRE_DEC.
3035 (arithmetic_instr, shift_op): Use arm_shift_nmem.
3036 (arm_print_operand): Use arm_print_condition. Handle '(', ')', '.',
3038 (arm_final_prescan_insn): Use extract_constrain_insn_cached.
3039 (thumb_expand_prologue): Use thumb_set_frame_pointer.
3040 (arm_file_start): Output directive for unified syntax.
3041 (arm_unwind_emit_set): Handle stack alignment instruction.
3042 * config/arm/lib1funcs.asm: Remove default for __ARM_ARCH__.
3043 Add v6t2, v7, v7a, v7r and v7m.
3044 (RETLDM): Add Thumb-2 code.
3045 (do_it, shift1, do_push, do_pop, COND, THUMB_SYNTAX): New macros.
3046 * config/arm/arm.h (TARGET_CPU_CPP_BUILTINS): Define __thumb2__.
3047 (TARGET_THUMB1, TARGET_32BIT, TARGET_THUMB2, TARGET_DSP_MULTIPLY,
3048 TARGET_INT_SIMD, TARGET_UNIFIED_ASM, ARM_FT_STACKALIGN, IS_STACKALIGN,
3049 THUMB2_TRAMPOLINE_TEMPLATE, TRAMPOLINE_ADJUST_ADDRESS,
3050 ASM_OUTPUT_OPCODE, THUMB2_GO_IF_LEGITIMATE_ADDRESS,
3051 THUMB2_LEGITIMIZE_ADDRESS, CASE_VECTOR_PC_RELATIVE,
3052 CASE_VECTOR_SHORTEN_MODE, ADDR_VEC_ALIGN, ASM_OUTPUT_CASE_END,
3053 ADJUST_INSN_LENGTH): Define.
3054 (TARGET_REALLY_IWMMXT, TARGET_IWMMXT_ABI, CONDITIONAL_REGISTER_USAGE,
3055 STATIC_CHAIN_REGNUM, HARD_REGNO_NREGS, INDEX_REG_CLASS,
3056 BASE_REG_CLASS, MODE_BASE_REG_CLASS, SMALL_REGISTER_CLASSES,
3057 PREFERRED_RELOAD_CLASS, SECONDARY_OUTPUT_RELOAD_CLASS,
3058 SECONDARY_INPUT_RELOAD_CLASS, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P,
3059 TRAMPOLINE_SIZE, INITIALIZE_TRAMPOLINE, HAVE_PRE_INCREMENT,
3060 HAVE_POST_DECREMENT, HAVE_PRE_DECREMENT, HAVE_PRE_MODIFY_DISP,
3061 HAVE_POST_MODIFY_DISP, HAVE_PRE_MODIFY_REG, HAVE_POST_MODIFY_REG,
3062 REGNO_MODE_OK_FOR_BASE_P, LEGITIMATE_CONSTANT_P,
3063 REG_MODE_OK_FOR_BASE_P, REG_OK_FOR_INDEX_P, GO_IF_LEGITIMATE_ADDRESS,
3064 LEGITIMIZE_ADDRESS, THUMB2_LEGITIMIZE_ADDRESS,
3065 GO_IF_MODE_DEPENDENT_ADDRESS, MEMORY_MOVE_COST, BRANCH_COST,
3066 ASM_APP_OFF, ASM_OUTPUT_CASE_LABEL, ARM_DECLARE_FUNCTION_NAME,
3067 FINAL_PRESCAN_INSN, PRINT_OPERAND_PUNCT_VALID_P,
3068 PRINT_OPERAND_ADDRESS): Adjust for Thumb-2.
3069 (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv): New declarations.
3070 * config/arm/arm-cores.def: Add arm1156t2-s, cortex-a8, cortex-r4 and
3072 * config/arm/arm-tune.md: Regenerate.
3073 * config/arm/arm-protos.h: Update prototypes.
3074 * config/arm/vfp.md: Enable patterns for Thumb-2.
3075 (arm_movsi_vfp): Add movw alternative. Use output_move_vfp.
3076 (arm_movdi_vfp, movsf_vfp, movdf_vfp): Use output_move_vfp.
3077 (thumb2_movsi_vfp, thumb2_movdi_vfp, thumb2_movsf_vfp,
3078 thumb2_movdf_vfp, thumb2_movsfcc_vfp, thumb2_movdfcc_vfp): New.
3079 * config/arm/libunwind.S: Add Thumb-2 code.
3080 * config/arm/constraints.md: Update include Thumb-2.
3081 * config/arm/ieee754-sf.S: Add Thumb-2/Unified asm support.
3082 * config/arm/ieee754-df.S: Ditto.
3083 * config/arm/bpabi.S: Ditto.
3084 * config/arm/t-arm (MD_INCLUDES): Add thumb2.md.
3085 * config/arm/predicates.md (low_register_operand,
3086 low_reg_or_int_operand, thumb_16bit_operator): New.
3087 (thumb_cmp_operand, thumb_cmpneg_operand): Rename...
3088 (thumb1_cmp_operand, thumb1_cmpneg_operand): ... to this.
3089 * config/arm/t-arm-elf: Add armv7 multilib.
3090 * config/arm/arm.md: Update patterns for Thumb-2 and Unified asm.
3092 (UNSPEC_STACK_ALIGN, ce_count): New.
3093 (arm_incscc, arm_decscc, arm_umaxsi3, arm_uminsi3,
3094 arm_zero_extendsidi2, arm_zero_extendqidi2): New
3096 * config/arm/fpa.md: Update patterns for Thumb-2 and Unified asm.
3097 (thumb2_movsf_fpa, thumb2_movdf_fpa, thumb2_movxf_fpa,
3098 thumb2_movsfcc_fpa, thumb2_movdfcc_fpa): New insns.
3099 * config/arm/cirrus.md: Update patterns for Thumb-2 and Unified asm.
3100 (cirrus_thumb2_movdi, cirrus_thumb2_movsi_insn,
3101 thumb2_cirrus_movsf_hard_insn, thumb2_cirrus_movdf_hard_insn): New
3103 * doc/extend.texi: Document ARMv7-M interrupt functions.
3104 * doc/invoke.texi: Document Thumb-2 new cores+architectures.
3106 2007-01-03 Jakub Jelinek <jakub@redhat.com>
3108 * unwind-dw2.c (SIGNAL_FRAME_BIT, EXTENDED_CONTEXT_BIT): Define.
3109 (struct _Unwind_Context): Rename args_size to flags, remove
3110 signal_frame field, add a new args_size field and version field.
3111 (_Unwind_IsSignalFrame, _Unwind_SetSignalFrame,
3112 _Unwind_IsExtendedContext): New inline functions.
3113 (_Unwind_GetGR, _Unwind_SetGR, _Unwind_GetGRPtr, _Unwind_SetGRPtr):
3114 Assume by_value array is only present if _Unwind_IsExtendedContext.
3115 (_Unwind_GetIPInfo, execute_cfa_program, uw_frame_state_for): Use
3116 _Unwind_IsSignalFrame.
3117 (__frame_state_for): Initialize context.flags to EXTENDED_CONTEXT_BIT.
3118 (uw_update_context_1): Use _Unwind_SetSignalFrame.
3119 (uw_init_context_1): Initialize context->flags to
3120 EXTENDED_CONTEXT_BIT.
3121 * config/rs6000/linux-unwind.h (frob_update_context): Use
3122 _Unwind_SetSignalFrame.
3124 2007-01-03 Andrew Pinski <andrew_pinski@playstation.sony.com>
3127 * gimplify.c (gimplify_modify_expr_complex_part): Move below
3128 tree_to_gimple_tuple. Call tree_to_gimple_tuple when we need
3131 2007-01-03 Kazu Hirata <kazu@codesourcery.com>
3133 * config/i386/i386.h (NON_STACK_REG_P, REGNO_OK_FOR_SIREG_P,
3134 REGNO_OK_FOR_DIREG_P, REWRITE_ADDRESS, ASM_OPERAND_LETTER,
3135 RET, AT_SP): Remove.
3136 * config/i386/i386.md (*sse_prologue_save_insn): Use return
3139 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
3140 * cfgbuild.c (find_basic_blocks): Likewise.
3141 * cfgrtl.c (rtl_create_basic_block): Likewise.
3142 * function.c (temp_slots_at_level): Likewise.
3143 * reg-stack.c (stack_regs_mentioned): Likewise.
3144 * regclass.c (allocate_reg_info): Likewise.
3145 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
3146 set_bb_for_stmt, move_block_to_fn): Likewise.
3147 * tree-complex.c (tree_lower_complex): Likewise.
3148 * vec.h (VEC_safe_grow_cleared): New.
3150 * cgraphunit.c, tree-ssa-alias.c: Fix comment typos.
3152 2007-01-03 Zdenek Dvorak <dvorakz@suse.cz>
3154 * loop-unswitch.c (unswitch_loop): Pass probabilities to loopify.
3155 * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Pass probabilities
3157 * cfgloopmanip.c (scale_loop_frequencies): Export.
3158 (loopify): Scale the frequencies by prescribed coefficients.
3159 (set_zero_probability): New function.
3160 (duplicate_loop_to_header_edge): Improve updating of frequencies.
3161 (lv_adjust_loop_entry_edge, loop_version): Set probabilities
3162 and frequencies according to arguments.
3163 * tree-ssa-loop-manip.c (tree_unroll_loop): Set probabilities
3165 * cfg.c (scale_bbs_frequencies_int): Allow scaling the frequencies up.
3166 * modulo-sched.c (sms_schedule): Set probabilities for entering
3167 versioned loop correctly.
3168 * tree-vect-transform.c (vect_transform_loop): Ditto.
3169 * cfgloop.h (loopify, loop_version): Declaration changed.
3170 (scale_loop_frequencies): Declared.
3172 2007-01-02 Jan Hubicka <jh@suse.cz>
3174 * cgraph.c: Include tree-flow.h
3175 (cgraph_add_new-function): Handle IPA_SSA mode; execute
3177 * cgraph.h (enum cgraph_state): Add CGRAPH_STATE_IPA_SSA.
3178 * tree-pass.h (pass_all_early_optimizations): Declare.
3179 * cgraphunit.c (cgraph_process_new_functions): Add IPA_SSA; execute
3181 (cgraph_analyze_function): Do early_local_passes.
3182 * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
3183 Do not add referenced vars.
3184 * tree-optimize.c (gate_all_optimizations): Do not execute when not in
3186 (gate_all_early_local_passes): New gate.
3187 (pass_early_local_passes): Use new gate.
3188 (execute_early_local_optimizations): New functions.
3189 (gate_all_early_optimizations): New gate.
3190 (pass_all_early_optimizations): New pass.
3191 (execute_free_datastructures): Free SSA only when initialized.
3192 (gate_init_datastructures): Init only when optimizing.
3193 (tree_lowering_passes): Do early local passes when called late.
3194 * tree-profile.c (do_tree_profiling): Don't profile functions added
3196 (do_early_tree_profiling, pass_early_tree_profile): Kill.
3197 * tree-cfg.c (update_modified_stmts): Do not update when operands are
3199 * passes.c (init_optimizations_passes): Reorder so we go into SSA
3200 during early_local_passes.
3201 * Makefile.in (cgraph.o): Add dependency on tree-flow.h.
3204 2007-01-02 Carlos O'Donell <carlos@codesourcery.com>
3206 * Makefile.in: Update copyright year.
3208 2007-01-02 Carlos O'Donell <carlos@codesourcery.com>
3210 * Makefile.in: Export GCC_EXEC_PREFIX before calling $(RUNTEST)
3211 in $(lang_checks) and check-consistency targets.
3213 2007-01-02 Jan Hubicka <jh@suse.cz>
3215 * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
3216 Do not add referenced vars.
3217 * tree-cfg.c (update_modified_stmts): Do not update when SSA operands
3219 * passes.c (init_optimization_passes): Put mudflap_2 after
3220 free_datastructures.
3222 2007-01-02 Jan Hubicka <jh@suse.cz>
3224 * tree-optimize (execute_fixup_cfg): Set after_inlining flag.
3225 Set NOTHROW flag on call statements proved to be nothrow.
3226 Update statement of local calls so new pure/const functions are
3227 updated. Update_ssa when in ssa form. Mark PHI nodes of nonlocal
3229 (tree_rest_of_compilation): Register hooks and initialize bitmap
3230 early. Do not set after_inlining flag.
3232 2007-01-02 Steve Ellcey <sje@cup.hp.com>
3234 * sbitmap.c (HOST_BITS_PER_LONG_LONG): Change to
3235 HOST_BITS_PER_LONGLONG
3237 2007-01-02 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
3240 * c-typeck.c (store_init_value): Don't emit pedantic overflow
3241 warning for non-static initializers.
3243 2007-01-02 Steven Bosscher <steven@gcc.gnu.org>
3245 * config/alpha/alpha.md, arm/arm.c, darwin.c, frv/frv.md,
3246 m32r/m32r.c, m32r/m32r.c, mn10300/mn10300.md, pa/pa.c,
3247 rs6000/rs6000.c, s390/s390.md, sh/sh.md, sparc/sparc.c:
3248 Always use set_unique_reg_note to add REG_EQUAL notes.
3250 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
3253 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
3255 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
3256 * cfgbuild.c (find_basic_blocks): Likewise.
3257 * cfgrtl.c (rtl_create_basic_block): Likewise.
3258 * function.c (temp_slots_at_level): Likewise.
3259 * reg-stack.c (stack_regs_mentioned): Likewise.
3260 * regclass.c (allocate_reg_info): Likewise.
3261 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
3262 set_bb_for_stmt, move_block_to_fn): Likewise.
3263 * tree-complex.c (tree_lower_complex): Likewise.
3264 * vec.h (VEC_safe_grow_cleared): New.
3266 2007-01-02 Ian Lance Taylor <iant@google.com>
3268 * c-common.c (c_common_truthvalue_conversion): When warning about
3269 using an assignment as a truth value, set TREE_NO_WARNING.
3271 2007-01-02 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
3274 * c.opt (Wold-style-declaration): New.
3275 * doc/invoke.texi (C-only Warning Options): New.
3276 (Wold-style-declaration): Document it.
3277 (Wextra): Enabled by -Wextra.
3278 * c-opts.c (c_common_post_options): Enabled by -Wextra.
3279 * c-decl.c (declspecs_add_scspec): Replace -Wextra with
3280 -Wold-style-declaration.
3282 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
3284 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
3285 * cfgbuild.c (find_basic_blocks): Likewise.
3286 * cfgrtl.c (rtl_create_basic_block): Likewise.
3287 * function.c (temp_slots_at_level): Likewise.
3288 * reg-stack.c (stack_regs_mentioned): Likewise.
3289 * regclass.c (allocate_reg_info): Likewise.
3290 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
3291 set_bb_for_stmt, move_block_to_fn): Likewise.
3292 * tree-complex.c (tree_lower_complex): Likewise.
3293 * vec.h (VEC_safe_grow_cleared): New.
3295 2007-01-02 Douglas Gregor <doug.gregor@gmail.com>
3297 * c-common.c (c_common_nodes_and_builtins): Since variants of
3298 void_type_node get built before it is given a name, we need to
3299 give those variants the name, too.
3300 (complete_array_type): We need to work with the canonical main
3301 type of the array, from which we will build the qualified version.
3302 * params.def (PARAM_VERIFY_CANONICAL_TYPES): New.
3303 * print-tree.c (print_node): Display canonical type information
3305 * stor-layout.c (layout_type): When we don't know the
3306 alignment of a type for which we're building an array, we end up
3307 guessing wrong, so make the type require structural equality.
3308 * tree.c (make_node_stat): When we build a new type, it is its
3310 (build_type_attribute_qual_variant): When building an attribute
3311 variant, its canonical type is the non-attribute variant. However,
3312 if the attributes are target-dependent and they differ, we need to
3313 use structural equality checks for this type.
3314 (build_qualified_type): A qualified type is not equivalent to its
3315 unqualified variant; set the canonical type appropriately.
3316 (build_distinct_type_copy): When building a distinct type from
3317 another type, the new type is its own canonical type.
3318 (build_variant_type_copy): When building a new type variant, we
3319 assume that it is equivalent to the original type.
3320 (build_pointer_type_for_mode): When building a pointer type, also
3321 build a canonical type pointer.
3322 (build_reference_type_for_mode): When building a reference type,
3323 also build a canonical type reference.
3324 (build_index_type): When we can't hash an index type (e.g.,
3325 because its maximum value is negative), the index type requires
3326 structural equality tests.
3327 (build_array_type): Build the canonical form of an array type.
3328 (build_function_type): Function types require structural equality,
3329 because they contain default arguments, attributes, etc.
3330 (build_method_type_directly): Ditto for method types.
3331 (build_offset_type): Build the canonical offset type.
3332 (build_complex_type): Build the canonical vector type.
3333 (make_vector_type): Build the canonical vector type.
3334 * tree.h (TYPE_CANONICAL): New.
3335 (TYPE_STRUCTURAL_EQUALITY_P): New.
3336 (SET_TYPE_STRUCTURAL_EQUALITY): New.
3337 (struct tree_type): Added "canonical" field.
3338 * params.h (VERIFY_CANONICAL_TYPES): New.
3339 * doc/c-tree.texi (TYPE_CANONICAL): Document.
3340 (TYPE_STRUCTURAL_EQUALITY_P): Document.
3341 (SET_TYPE_STRUCTURAL_EQUALITY): Document.
3342 * doc/invoke.texi (verify-canonical-types): Document --param
3343 parameter for verifying canonical types.
3345 2007-01-02 Joseph Myers <joseph@codesourcery.com>
3347 * config.gcc (powerpc-*-eabispe*, powerpc-*-eabisimaltivec*,
3348 powerpc-*-eabisim*, powerpc-*-eabialtivec*, powerpc-*-eabi*,
3349 powerpc-*-rtems*, powerpc-wrs-vxworks, powerpc-wrs-vxworksae,
3350 powerpcle-*-eabisim*, powerpcle-*-eabi*): Add rs6000/e500.h to
3352 * config/rs6000/e500.h: New.
3353 * config/rs6000/eabi.h (TARGET_SPE_ABI, TARGET_SPE, TARGET_E500,
3354 TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE, TARGET_E500_DOUBLE):
3356 * config/rs6000/linuxspe.h (TARGET_SPE_ABI, TARGET_SPE,
3357 TARGET_E500, TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE,
3358 TARGET_E500_DOUBLE): Remove.
3359 * config/rs6000/vxworks.h (TARGET_SPE_ABI, TARGET_SPE,
3360 TARGET_E500, TARGET_ISEL, TARGET_FPRS): Remove.
3361 * config/rs6000/rs6000.h (CHECK_E500_OPTIONS): Define.
3362 * config/rs6000/rs6000.c (rs6000_override_options): Use
3365 2007-01-02 Joseph Myers <joseph@codesourcery.com>
3367 * config/rs6000/rs6000.c (print_operand): Check (TARGET_SPE ||
3368 TARGET_E500_DOUBLE), not TARGET_E500, for %y.
3369 (rs6000_generate_compare, rs6000_emit_sCOND, output_cbranch,
3370 rs6000_emit_cmove): Don't check TARGET_E500.
3371 * config/rs6000/rs6000.md (bunordered, bordered, sunordered,
3372 sordered): Don't check TARGET_E500.
3374 2007-01-01 Eric Christopher <echristo@apple.com>
3376 * config/mips/mips.c (mips_regno_mode_ok_for_base_p): Use
3377 HARD_REGISTER_NUM_P.
3379 2007-01-01 Roger Sayle <roger@eyesopen.com>
3381 * fold-const.c (fold_binary) <EQ_EXPR>: Fold "(X^C1) eq/ne C2" into
3382 "X eq/ne (C1^C2)". Fold "(X^Z) eq/ne (Y^Z)" as "X eq/ne Y" when Z
3383 has no side-effects. Fold "(X^C1) eq/ne (Y^C2)" as "(X^(C1^C2))
3386 2007-01-01 Mike Stump <mrs@apple.com>
3388 * configure.ac: Remove support for building with Apple's gcc-3.1.
3390 2007-01-02 Joseph Myers <joseph@codesourcery.com>
3393 * caller-save.c (add_stored_regs): Only handle SUBREGs if inner
3394 REG is a hard register. Do not modify REG before calling
3396 * rtlanal.c (subreg_get_info): Don't assert size of XMODE is a
3397 multiple of the size of YMODE for certain lowpart cases.
3399 2007-01-01 Andrew Pinski <pinskia@gmail.com>
3402 * gimplify (voidify_wrapper_expr): Update for
3405 2007-01-01 Andreas Schwab <schwab@suse.de>
3408 * config/ia64/ia64.c (ia64_compute_frame_size): Account space for
3409 save of BR0 in extra_spill_size instead of spill_size.
3410 (ia64_expand_prologue): Save BR0 outside of the gr/br/fr spill
3412 (ia64_expand_epilogue): Restore BR0 from its new location.
3414 2007-01-01 Andrew Pinski <pinskia@gmail.com>
3416 * gimplify.c (gimplify_init_constructor <case VECTOR_TYPE>):
3417 Use a temporary variable if the left hand side is not a gimple
3420 2007-01-01 Andrew Pinski <pinskia@gmail.com>
3422 * gimplify.c (gimplify_return_expr): Make the temporary variable
3423 for the return expression, a gimple register variable.
3425 2007-01-01 Jan Hubicka <jh@suse.cz>
3427 * emit-rtl.c (emit_copy_of_insn_after): Do not call copy_insn_1 for
3430 2007-01-01 Mike Stump <mrs@apple.com>
3432 * configure.ac (HAVE_GAS_LITERAL16): Add autoconf check for
3434 * config/darwin.c (machopic_select_rtx_section): Use
3436 (darwin_mergeable_constant_section): Likewise.
3437 * configure: Regenerate.
3438 * config.in: Regenerate.
3440 2007-01-01 Jan Hubicka <jh@suse.cz>
3441 Andrew Pinski <pinskia@gmail.com>
3443 * cgraphunit.c (cgraph_optimize): Call cgraph_add_new_functions
3444 before starting IPA passes.