| blob | 46bfb56d7eacae5355cb0f4eccb6e9b417500b0f |
1 2007-02-01 Ian Lance Taylor <iant@google.com>
3 * lower-subreg.c (resolve_clobber): Handle a subreg of a concatn.
5 2007-02-01 Guy Martin <gmsoft@gentoo.org>
7 * pa.md (tp_load): Correct mfctl instruction syntax.
9 2007-02-01 Geoffrey Keating <geoffk@apple.com>
11 * config/rs6000/rs6000.c (rs6000_stack_info): Correct
12 altivec_padding_size calculation on AIX. Improve comment, add
13 assert to verify that it's right.
15 * config/rs6000/darwin.h (MD_UNWIND_SUPPORT): Don't define for
16 64-bit.
18 2007-2-01 Seongbae Park <seongbae.park@gmail.com>
20 PR inline-asm/28686
21 * global.c (global_alloc): Add mising initialization of
22 ELIMINABLE_REGSET.
24 2007-02-01 Roger Sayle <roger@eyesopen.com>
26 * alias.c (init_alias_analysis): Correct whitespace.
27 * bb-reorder.c (fix_edges_for_rarely_executed_code,
28 partition_hot_cold_basic_blocks): Likewise.
29 * builtins.c (expand_builtin_printf, expand_builtin_fprintf,
30 expand_builtin_sprintf, fold_builtin_carg, fold_builtin_sprintf,
31 maybe_emit_sprintf_chk_warning, fold_builtin_sprintf_chk,
32 fold_builtin_snprintf_chk, fold_builtin_printf,
33 fold_builtin_fprintf, do_mpfr_ckconv, do_mpfr_arg1, do_mpfr_arg2,
34 do_mpfr_arg3, do_mpfr_sincos): Likewise.
35 * cfgcleanup.c (cleanup_cfg): Likewise.
36 * cfgexpand.c (tree_expand_cfg): Likewise.
37 * fold-const.c (fold_binary) <RDIV_EXPR>: Likewise.
38 * function.c (get_next_funcdef_no): Likewise.
39 * gengtype.c (main): Likewise.
40 * genmodes.c (main): Likewise.
41 * gcse.c (bypass_conditional_jumps, print_ldst_list): Likewise.
42 * haifa-sched.c (schedule_block, extend_h_i_d): Likewise.
43 * ifcvt.c (noce_emit_move_insn): Likewise.
44 * modulo-sched.c (generate_prolog_epilog, sms_schedule_by_order):
45 Likewise.
46 * stor-layout.c (get_best_mode): Likewise.
47 * tree-ssa-loop-niter.c (get_val_for): Likewise.
48 * tree-ssa-structalias.c (get_varinfo, get_varinfo_fc,
49 scc_visit, do_ds_constraint, do_complex_constraint, label_visit,
50 perform_var_substitution, solve_graph): Likewise.
51 * tree-vrp.c (vrp_finalize): Likewise.
53 2007-02-01 Ian Lance Taylor <iant@google.com>
55 * lower-subreg.c (simplify_gen_subreg_concatn): If we ask for the
56 high part of a paradoxical subreg, return a constant zero.
58 2007-02-01 Zdenek Dvorak <dvorakz@suse.cz>
60 * toplev.c (lang_dependent_init): Call init_set_costs.
61 * loop-init.c (loop_optimizer_init): Do not call init_set_costs.
63 2007-02-01 Richard Guenther <rguenther@suse.de>
65 PR middle-end/30656
66 * fold-const.c (fold_negate_expr): Allow negating a
67 constant if overflow does not change.
69 2007-02-01 Ralf Wildenhues <Ralf.Wildenhues@gmx.de>
71 * doc/c-tree.texi (Expression trees): Improve markup.
72 * doc/tm.texi (Register Classes, Addressing Modes)
73 (Floating Point): Fix spacing after abbreviations. Fix some
74 typos.
76 2007-02-01 Ben Elliston <bje@au.ibm.com>
78 * doc/invoke.texi: Replace "bugfix" with "bug fix" throughout.
79 * doc/contrib.texi: Likewise.
80 * doc/install.texi: Likewise.
82 2007-01-31 Richard Henderson <rth@redhat.com>
83 Ian Lance Taylor <iant@google.com>
85 * lower-subreg.c: New file.
86 * rtl.def (CONCATN): Define.
87 * passes.c (init_optimization_passes): Add pass_lower_subreg and
88 pass_lower_subreg2.
89 * emit-rtl.c (update_reg_offset): New static function, broken out
90 of gen_rtx_REG_offset.
91 (gen_rtx_REG_offset): Call update_reg_offset.
92 (gen_reg_rtx_offset): New function.
93 * regclass.c: Revert patch of 2006-03-05, restoring
94 reg_scan_update.
95 (clear_reg_info_regno): New function.
96 * dwarf2out.c (concatn_loc_descriptor): New static function.
97 (loc_descriptor): Handle CONCATN.
98 * common.opt (fsplit_wide_types): New option.
99 * opts.c (decode_options): Set flag_split_wide_types when
100 optimizing.
101 * timevar.def (TV_LOWER_SUBREG): Define.
102 * rtl.h (gen_reg_rtx_offset): Declare.
103 (reg_scan_update): Declare.
104 * regs.h (clear_reg_info_regno): Declare.
105 * tree-pass.h (pass_lower_subreg): Declare.
106 (pass_lower_subreg2): Declare.
107 * doc/invoke.texi (Option Summary): List -fno-split-wide-types.
108 (Optimize Options): Add -fsplit-wide-types to -O1 list. Document
109 -fsplit-wide-types.
110 * doc/rtl.texi (Regs and Memory): Document concat and concatn.
111 * Makefile.in (OBJS-common): Add lower-subreg.o.
112 (lower-subreg.o): New target.
114 2007-01-31 Kazu Hirata <kazu@codesourcery.com>
116 * config/sh/sh.h (HAVE_SECONDARY_RELOADS): Remove.
118 2007-01-31 Anatoly Sokolov <aesok@post.ru>
120 PR target/19087
121 * config/avr/avr.c (DWARF2_ADDR_SIZE): Define.
123 2007-01-31 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
125 PR middle-end/29335
126 * builtins.c (fold_builtin_sqrt): Use MPFR for constant args.
128 2007-01-31 Zdenek Dvorak <dvorakz@suse.cz>
130 * cfgloop.h: Include vec-prim.h.
131 (enum li_flags): Remove LI_ONLY_OLD.
132 (loop_iterator): Changed.
133 (fel_next, fel_init): Iterate over loop tree.
134 (FOR_EACH_LOOP_BREAK): New macro.
135 * loop-unswitch.c (unswitch_loops): Do not pass LI_ONLY_OLD to
136 FOR_EACH_LOOP.
137 * tree-ssa-loop-unswitch.c (tree_ssa_unswitch_loops): Ditto.
138 * modulo-sched.c (sms_schedule): Ditto.
139 * tree-vectorizer.c (vectorize_loops): Ditto.
140 * doc/loop.texi: Update information on loop numbering and behavior of
141 FOR_EACH_LOOP wrto new loops.
142 * tree-scalar-evolution.c (compute_overall_effect_of_inner_loop,
143 add_to_evolution_1): Test nestedness of loops instead of comparing
144 their numbers.
145 * tree-chrec.c (chrec_fold_plus_poly_poly,
146 chrec_fold_multiply_poly_poly, chrec_evaluate,
147 hide_evolution_in_other_loops_than_loop, chrec_component_in_loop_num,
148 reset_evolution_in_loop): Ditto.
149 * Makefile.in (CFGLOOP_H): Add vecprim.h dependency.
151 2007-01-31 Dirk Mueller <dmueller@suse.de>
153 * c-common.c (warn_about_parentheses): Separate warning about
154 un-parenthized sequence of comparison operators from the one
155 which is supposed to warn about x <= y <= z.
157 2007-01-31 Uros Bizjak <ubizjak@gmail.com>
159 * optabs.h (enum optab_index): Add new OTI_isinf.
160 (isinf_optab): Define corresponding macro.
161 * optabs.c (init_optabs): Initialize isinf_optab.
162 * genopinit.c (optabs): Implement isinf_optab using isinf?f2
163 patterns.
164 * builtins.c (mathfn_built_in): Handle BUILT_IN_ISINF{,F,L}.
165 (expand_builtin_interclass_mathfn): Expand BUILT_IN_ISINF{,F,L}
166 using isinf_optab.
167 (expand_builtin): Expand BUILT_IN_ISINF{,F,L} using
168 expand_builtin_interclass_mathfn.
169 * reg_stack.c (subst_stack_regs_pat): Handle UNSPEC_FXAM.
170 * config/i386/i386.md (UNSPEC_FXAM): New constant.
171 (fxam<mode>2_i387): New insn pattern.
172 (isinf<mode>2) New expander to implement isinf, isinff and isinfl
173 built-in functions as x87 inline asm.
175 2007-01-31 Kazu Hirata <kazu@codesourcery.com>
177 * gcc/config/arm/unwind-arm.h (_sleb128_t, _uleb128_t): New.
179 2007-01-30 Eric Christopher <echristo@apple.com>
181 * config.gcc: Add geode.
183 2007-01-31 Kazu Hirata <kazu@codesourcery.com>
185 * cgraphunit.c, config/arm/arm.c, config/m68k/m68k.c,
186 ipa-inline.c, tree-profile.c, tree-ssa-live.c,
187 tree-ssa-math-opts.c, tree-ssanames.c, tree-vect-analyze.c,
188 value-prof.c: Fix comment typos.
190 2007-01-30 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
192 PR c++/24745
193 * doc/invoke.texi (Wpointer-arith): Document warning.
195 2007-01-30 Janis Johnson
197 * doc/extend.texi (Decimal Floating Types): Remove decfloat.h from
198 the list of discrepancies from the draft TR.
200 2007-01-30 Dirk Mueller <dmueller@suse.de>
202 PR c++/30601
203 * doc/invoke.texi (-Wreturn-type): Update description to
204 match new behavior.
206 2007-01-30 Richard Sandiford <richard@codesourcery.com>
208 * cfgrtl.c (try_redirect_by_replacing_jump): Check only_sets_cc0_p.
210 2007-01-30 Uros Bizjak <ubizjak@gmail.com>
212 * builtins.c (expand_builtin_int_interclass_roundingfn): New function
213 to handle optabs that operate on floating point input argument and
214 output to integer output.
215 (expand_builtin_mathfn) [BUILT_IN_ILOGB]: Move from here ...
216 (expand_builtin_interclass_mathfn) [BUILT_IN_ILOGB]: ... to here.
217 (expand_builtin): Expand BUILT_IN_ILOGB{,F,L} using
218 expand_builtin_interclass_mathfn ().
219 * config/i386/i386.md (fxtractxf3_i387): Rename from *fxtractxf3_i387.
220 (ilogbsi2): Remove.
221 (ilogbxf2, ilogb<mode>2): New expanders to implement ilogb, ilogbf and
222 ilogbl built-in functions as x87 intrinsics.
224 2007-01-30 Richard Guenther <rguenther@suse.de>
226 PR middle-end/27657
227 * dwarf2out.c (reference_to_unused): Query varpool if the
228 variable was output.
230 2007-01-30 Richard Guenther <rguenther@suse.de>
232 PR middle-end/30313
233 * passes.c (execute_one_pass): Reset in_gimple_form to not
234 confuse non-unit-at-a-time mode.
236 2007-01-29 Roger Sayle <roger@eyesopen.com>
237 Richard Guenther <rguenther@suse.de>
239 * fold-const.c (round_up): Make HIGH an unsigned HOST_WIDE_INT to
240 avoid undefined behaviour on overflow. Use force_fit_type_double
241 to construct the constant with the specified TREE_OVERFLOW.
243 2007-01-29 Janis Johnson <janis187@us.ibm.com>
245 * config/dfp-bit.c: Add parameterized support for fp exceptions.
246 * config/dfp-bit.h: Ditto.
248 2007-01-29 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
250 * c-decl.c (pop_scope): Replace warnings with call to
251 warn_for_unused_label.
252 * c-common.h (warn_for_unused_label): Declare.
253 * c-common.c (warn_for_unused_label): Define.
255 2007-01-29 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
257 * tree-optimize.c (update_inlined_to_pointers): Delete unused
258 function.
260 2007-01-29 Janis Johnson <janis187@us.ibm.com>
262 * Makefile.in (USER_H): Remove decfloat.h.
263 * ginclude/decfloat.h: Delete, moving contents to ...
264 * ginclude/float.h: Add support for decimal floating point,
265 guarded by __STDC_WANT_DEC_FP__.
267 2007-01-29 Mike Stump <mrs@apple.com>
269 * doc/gccint.texi (Top): Rename Loop Representation to Loop
270 Analysis and Representation to resolve case insensitive conflict.
271 * doc/loop.texi (Loop Analysis and Representation): Likewise.
273 2007-01-28 Daniel Berlin <dberlin@dberlin.org>
275 * tree.h (struct tree_memory_tag): Add aliases member.
276 (MTAG_ALIASES): New macro.
277 * tree-ssa-alias.c (alias_bitmap_obstack): New variable.
278 (add_may_alias): Remove pointer-set. Update for may_aliases being
279 a bitmap.
280 (mark_aliases_call_clobbered): Update for may_aliases being a
281 bitmap.
282 (compute_tag_properties): Ditto.
283 (create_partition_for): Ditto.
284 (compute_memory_partitions): Ditto.
285 (dump_may_aliases_for): Ditto.
286 (is_aliased_with): Ditto.
287 (add_may_alias_for_new_tag): Ditto.
288 (rewrite_alias_set_for): Rewrite for may_aliases being a bitmap.
289 (compute_is_aliased): New function.
290 (compute_may_aliases): Call compute_is_aliased).
291 (init_alias_info): Initialize alias_bitmap_obstack.
292 (union_alias_set_into): New function.
293 (compute_flow_sensitive_aliasing): Use union_aliases_into.
294 (have_common_aliases_p): Rewrite to take two bitmaps and use
295 intersection.
296 (compute_flow_insensitive_aliasing): Stop using pointer-sets.
297 Update for bitmaps.
298 (finalize_ref_all_pointers): Update for add_may_alias changes.
299 (new_type_alias): Ditto.
300 * tree-flow-inline.h (may_aliases): Return a bitmap.
301 * tree-dfa.c (dump_variable): Check for MTAG_P'ness.
302 * tree-ssa.c (verify_flow_insensitive_alias_info): Update for
303 may_aliases being a bitmap.
304 * tree-flow.h (struct var_ann_d): Remove may_aliases member.
305 may_aliases now returns a bitmap.
306 * tree-ssa-structalias.c (merge_smts_into): Update for may_aliases
307 being a bitmap.
308 * tree-ssa-operands.c (add_virtual_operand): Update for
309 may_aliases being a bitmap.
311 2007-01-29 Daniel Berlin <dberlin@dberlin.org>
313 PR tree-optimization/30630
314 * tree-ssa-structalias.c (do_complex_constraint): Mark correct
315 variable as changed.
317 2007-01-29 Simon Martin <simartin@users.sourceforge.net>
319 PR c++/28266
320 * gimplify.c (gimplify_target_expr): Make sure that the TARGET_EXPR is
321 expanded only once even if an error occurs.
323 2007-01-29 Ben Elliston <bje@au.ibm.com>
325 * gcov-io.h (__gcov_indirect_call_profiler): Declare.
326 (__gcov_average_profiler): Likewise.
327 (__gcov_ior_profiler): Likewise.
328 (__gcov_merge_ior): Likewise.
330 2007-01-28 Jan Hubicka <jh@suse.cz>
332 * builtins.c (expand_builtin_memset): Fix typo in my last patch.
333 * value-prof.c (stringop_block_profile): Likewise.
335 2007-01-28 Jan Hubicka <jh@suse.cz>
337 * expr.c (emit_block_move_via_movmem, emit_block_move_via_libcall): Add
338 variant handling histograms; add wrapper.
339 (clear_storage_via_libcall): Export.
340 (emit_block_move_hints): Break out from ...; add histograms.
341 (emit_block_move): ... this one.
342 (clear_storage_hints): Break out from ...; add histograms.
343 (clear_storage): ... this one.
344 (set_storage_via_memset): Handle histogram.
345 * expr.h (emit_block_move_via_libcall, emit_block_move_hints): Declare.
346 (clear_storage_hints, clear_storage_via_libcall): Declare.
347 (set_storage_via_setmem): Update prototype.
348 * doc/md.texi (movmem, setmem): Document new arguments.
350 * value-prof.c (dump_histogram_value, tree_find_values_to_profile): Add
351 new histograms.
352 (stringop_block_profile): New global function.
353 (tree_stringops_values_to_profile): Profile block size and alignment.
354 * value-prof.h (enum hist_type): add HIST_TYPE_AVERAGE and
355 HIST_TYPE_IOR.
356 (struct profile_hooks): Add gen_average_profiler and gen_ior_profiler.
357 (stringop_block_profile): Declare.
358 * builtins.c: Include value-prof.h.
359 (expand_builtin_memcpy, expand_builtin_memset): Pass block profile.
360 * gcov-ui.h (GCOV_COUNTER_NAMES): Add new counter.
361 (GCOV_COUNTER_AVERAGE, GCOV_COUNTER_IOR): New constants.
362 (GCOV_COUNTERS, GCOV_LAST_VALUE_COUNTER): Update.
363 * profile.c (instrument_values): Add new counters.
364 * cfgexpand.c (expand_gimple_basic_block): Propagate histograms to
365 calls.
366 * tree-profile.c (tree_average_profiler_fn, tree_ior_profiler_fn): New.
367 (tree_init_edge_profiler): Build new profilers.
368 (tree_gen_average_profiler, tree_gen_ior_profiler): New.
369 (pass_tree_profile): Add dump.
370 (tree_profile_hooks): Update.
371 * Makefile.in (LIBGCOV): Add new constants.
372 * libgcov.c (__gcov_merge_ior, __gcov_average_profiler,
373 __gcov_ior_profiler): New.
374 * i386.md (movmem/setmem expanders): Add new optional arguments.
376 2007-01-28 David Edelsohn <edelsohn@gnu.org>
378 * doc/md.texi (Standard Pattern Names): Document blockage pattern.
380 2007-01-28 Zdenek Dvorak <dvorakz@suse.cz>
382 * tree-ssa-loop-unswitch.c: Include tree-inline.h.
383 (tree_unswitch_single_loop): Pass eni_size_weights to
384 tree_num_loop_insns.
385 * tree-ssa-loop-manip.c: Include tree-inline.h.
386 (can_unroll_loop_p): Pass eni_size_weights to
387 tree_num_loop_insns.
388 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p):
389 Pass eni_size_weights to estimate_num_insns.
390 * tree.h (init_inline_once): Export.
391 * toplev.c (backend_init): Call init_inline_once.
392 * cgraphunit.c (cgraph_process_new_functions,
393 cgraph_analyze_function): Pass eni_inlining_weights to
394 estimate_num_insns.
395 * ipa-inline.c (compute_inline_parameters): Ditto.
396 * tree-ssa-loop-ivcanon.c (tree_num_loop_insns): Pass weights
397 to estimate_num_insns.
398 (try_unroll_loop_completely): Pass eni_size_weights to
399 tree_num_loop_insns.
400 * tree-eh.c (decide_copy_try_finally): Pass eni_size_weights
401 ot estimate_num_insns.
402 * tree-ssa-loop-prefetch.c: Include tree-inline.h.
403 (loop_prefetch_arrays): Pass eni_time_weights to tree_num_loop_insns.
404 * tree-inline.c (eni_inlining_weights, eni_size_weights,
405 eni_time_weights): New variables.
406 (init_inline_once): Initialize them.
407 (struct eni_data): Mew.
408 (estimate_num_insns_1, estimate_num_insns): Use weights.
409 * tree-inline.h (struct eni_weights_d): New.
410 (eni_inlining_weights, eni_size_weights, eni_time_weights): Declare.
411 (estimate_num_insns): Declaration changed.
412 * cfgloop.h (tree_num_loop_insns): Declaration changed.
413 * Makefile.in (tree-ssa-loop-unswitch.o, tree-ssa-loop-prefetch.o,
414 tree-ssa-loop-manip.o): Add TREE_INLINE_H dependency.
416 2007-01-28 Zdenek Dvorak <dvorakz@suse.cz>
418 * tree-data-ref.c (conflict_fn): Assert that the number of affine
419 relations in the conflict function is valid.
421 2007-01-27 Ian Lance Taylor <iant@google.com>
423 * common.opt: Add fstrict-overflow.
424 * opts.c (decode_options): Set flag_strict_overflow if -O2.
425 * flags.h (TYPE_OVERFLOW_WRAPS): Define.
426 (TYPE_OVERFLOW_UNDEFINED): Define.
427 (TYPE_OVERFLOW_TRAPS): Define. This replaces TYPE_TRAP_SIGNED.
428 Replace all uses.
429 * tree.h (TYPE_TRAP_SIGNED): Don't define.
430 * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_UNDEFINED.
431 (fold_negate_expr): Likewise.
432 (make_range): Likewise.
433 (extract_muldiv_1): Likewise.
434 (maybe_canonicalize_comparison): Likewise.
435 (fold_comparison): Likewise.
436 (fold_binary): Likewise.
437 (tree_expr_nonnegative_p): Likewise.
438 (tree_expr_nonzero_p): Likewise.
439 * tree-vrp.c (compare_values): Likewise.
440 (extract_range_from_binary_expr): Likewise.
441 (extract_range_from_unary_expr): Likewise.
442 * tree-ssa-loop-niter.c (infer_loop_bounds_from_signedness):
443 Likewise.
444 (nowrap_type_p): Likewise.
445 * tree-scalar-evolution.c (simple_iv): Likewise.
446 * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_WRAPS.
447 (build_range_check): Likewise.
448 (extract_muldiv_1): Likewise.
449 (fold_comparison): Likewise.
450 * tree-vrp.c (vrp_int_const_binop): Likewise.
451 (extract_range_from_unary_expr): Likewise.
452 * convert.c (convert_to_integer): Likewise.
453 * fold-const.c (fold_negate_expr): Use TYPE_OVERFLOW_TRAPS.
454 (fold_comparison): Likewise.
455 (fold_binary): Likewise.
456 * optabs.c (optab_for_tree_code): Likewise.
457 * tree-vectorizer.c (vect_is_simple_reduction): Likewise.
458 * simplify-rtx.c (simplify_const_relational_operation): Check
459 flag_strict_overflow and flag_trapv.
460 (simplify_const_relational_operation): Likewise.
461 * doc/invoke.texi (Option Summary): Mention -fstrict-overflow.
462 (Optimize Options): Add -fstrict-overflow to -O2 list. Document
463 -fstrict-overflow.
465 2007-01-27 Roger Sayle <roger@eyesopen.com>
467 * tree.c (tree_fold_gcd): Delete.
468 * tree.h (tree_fold_gcd): Remove prototype.
469 * tree-data-ref.c (tree_fold_divides_p): Don't use tree_fold_gcd to
470 test whether one constant integer is a multiple of another. Instead
471 call int_const_binop with TRUNC_MOD_EXPR and test for a zero result.
472 * fold-const.c (multiple_of_p): We've determined both TOP and
473 BOTTOM are integer constants so we can call int_const_binop directly
474 instead of the more generic const_binop.
476 2007-01-27 Roger Sayle <roger@eyesopen.com>
478 * fold-const.c (size_binop): In the fast-paths for X+0, 0+X, X-0 and
479 1*X check that the constant hasn't overflowed, to preserve the
480 TREE_OVERFLOW bit.
481 (round_up): Provide an efficient implementation when rouding-up an
482 INTEGER_CST to a power-of-two.
484 2007-01-28 Ralf Wildenhues <Ralf.Wildenhues@gmx.de>
486 * doc/sourcebuild.texi: Add comma for clarity.
487 * doc/extend.texi: Fix some typos.
488 * doc/passes.texi: Likewise.
489 * doc/cppinternals.texi: Likewise.
490 * doc/c-tree.texi: Likewise.
491 * doc/tree-ssa.texi: Likewise.
492 * doc/install.texi: Likewise.
494 2007-01-27 Jan Hubicka <jh@suse.cz>
496 * tree-sra.c (sra_walk_function): Don't rely on aliases being build.
497 (pass_sra): Do not require alias information.
498 * passes.c (init_optimization_passes): Add SRA
500 2007-01-27 Steven Bosscher <steven@gcc.gnu.org>
502 * tracer.c (rest_of_handle_tracer): We already cleaned
503 up the CFG in tracer() so don't do it here again.
504 * cfgcleanup.c (rest_of_handle_jump2): Don't repeat
505 cleanup_cfg here, either. And don't call renumber_insns.
507 * cfgrtl.c (rtl_verify_flow_info_1): Don't verify that BB_END
508 and BB_HEAD are in the insn stream here. Instead make sure
509 that BB_INSN is valid on all insns. Also, do check here that
510 there are no pending branch predictions...
511 (rtl_verify_flow_info): ...instead of doing it here. Checks
512 for BB_END and BB_HEAD moved from rtl_verify_flow_info_1 to
513 here.
515 2007-01-26 Roger Sayle <roger@eyesopen.com>
517 * config/i386/i386.c (ix86_swap_binary_operands_p): New helper
518 function to simplify/factorize operand order canonicalization.
519 (ix86_fixup_binary_operands): Reorganize using the above function.
520 (ix86_binary_operator_ok): Likewise.
522 2007-01-27 Jakub Jelinek <jakub@redhat.com>
524 * genattrtab.c (struct attr_value_list, insn_code_values): Move to
525 file scope from optimize_attrs.
526 (simplify_test_exp): If insn_code_values is not NULL, use it to speed
527 up search.
528 (optimize_attrs): Clear insn_code_values after freeing it.
530 2007-01-26 Zdenek Dvorak <dvorakz@suse.cz>
532 * tree-ssa-address.c (create_mem_ref): Remove ", bsi" from
533 a parts.base assignment.
535 2007-01-26 Zdenek Dvorak <dvorakz@suse.cz>
537 * tree-data-ref.c (dump_subscript): Use dump_conflict_function.
538 (compute_subscript_distance, initialize_data_dependence_relation,
539 finalize_ddr_dependent, analyze_ziv_subscript,
540 analyze_siv_subscript_cst_affine,
541 compute_overlap_steps_for_affine_univar,
542 compute_overlap_steps_for_affine_1_2, analyze_subscript_affine_affine,
543 analyze_siv_subscript, analyze_miv_subscript,
544 analyze_overlapping_iterations, subscript_dependence_tester_1,
545 compute_self_dependence, free_dependence_relation): Work
546 with affine_fn instead of chrecs.
547 (dump_affine_function, dump_conflict_function, affine_function_equal_p,
548 common_affine_function, affine_function_base,
549 affine_function_constant_p, affine_fn_op, affine_fn_plus,
550 affine_fn_minus, affine_fn_free, conflict_fn_not_known,
551 conflict_fn_no_dependence, free_conflict_function, free_subscripts,
552 conflict_fn, affine_fn_cst, affine_fn_univar): New functions.
553 (all_chrecs_equal_p): Removed.
554 * tree-data-ref.h (affine_fn, conflict_function): New types.
555 (struct subscript): Change type of conflicting_iterations_in_a
556 and conflicting_iterations_in_b.
558 2007-01-26 Steve Ellcey <sje@cup.hp.com>
560 PR other/30182
561 * config/pa/pa.h (TARGET_HPUX_11): New.
562 * config/pa/pa-hpux11.h (TARGET_HPUX_11): New.
563 * config/pa/pa.c (pa_init_builtins): Use TARGET_HPUX_11.
565 2007-01-26 Daniel Berlin <dberlin@dberlin.org>
566 Richard Guenther <rguenther@suse.de>
568 * tree-ssa-structalias.c (solve_graph): Handle case
569 we merged the variable to another.
571 2007-01-25 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
573 * builtins.c (fold_builtin_1): Treat ccos and ccosh as 'even'
574 functions.
576 * fold-const.c (negate_mathfn_p): Treat casin, casinh, catan,
577 catanh, cproj, csin, csinh, ctan and ctanh as 'odd' functions.
579 2007-01-25 DJ Delorie <dj@redhat.com>
581 * config/m32c/m32c.c (m32c_cannot_change_mode_class): We don't
582 allow changes to modes which don't fit in those registers.
584 * reload1.c (choose_reload_regs): Check for invalid subregs before
585 computing their locations, not after.
587 2007-01-25 Geoffrey Keating <geoffk@apple.com>
589 PR 25127
590 * config/rs6000/rs6000.c (first_altivec_reg_to_save): On Darwin,
591 save Altivec registers in an eh_return function.
592 (compute_vrsave_mask): Likewise.
593 (rs6000_stack_info): Correct AIX/Darwin stack alignment computation
594 for saving Altivec registers.
595 (rs6000_emit_prologue): Don't allocate stack twice in
596 eh_return function. Correct expected value of altivec_save_offset
597 when using save_world. Describe save of R0 to stack when using
598 save_world. Describe stack pointer adjustment when using
599 save_world. Remove duplicated eh_return parameter register saving.
600 Update sp_offset variable after save_world.
601 * config/rs6000/t-darwin (LIB2FUNCS_STATIC_EXTRA): Remove
602 darwin-world.asm.
603 (LIB2FUNCS_EXTRA): Add darwin-world.asm.
604 * config/rs6000/darwin.h (SUBTARGET_OVERRIDE_OPTIONS): -m64
605 implies Altivec.
607 2007-01-25 Steve Ellcey <sje@cup.hp.com>
609 * config.gcc (ia64*-*-hpux*): Make posix threads the default.
611 2007-01-25 Steve Ellcey <sje@cup.hp.com>
613 PR other/30182
614 * config/pa/pa.c (pa_init_builtins): Set asm names for finite routines.
615 * config/ia64/ia64.c (ia64_init_builtins): Ditto.
617 2007-01-25 Richard Guenther <rguenther@suse.de>
619 * doc/invoke.texi (-Wcoverage-mismatch): Document.
620 * common.opt (-Wcoverage-mismatch): New warning option.
621 * coverage.c (get_coverage_counts): Ignore coverage mismatch
622 if -Wcoverage-mismatch is given.
624 2007-01-25 Razya Ladelsky <razya@il.ibm.com>
626 * ipa-cp.c (ipcp_insert_stage, ipcp_driver): Support for SSA.
627 (ipcp_driver): Change to static definition.
628 Add dumping of the ifunctions.
629 (constant_val_insert): Remove unused parameter. Support for SSA.
630 (ipcp_propagate_const): Support for SSA.
631 (ipcp_profile_bb_print): Print only analyzed nodes.
632 (ipcp_replace_map_create): Remove support for Fortran constant
633 for now.
634 * ipa-prop.c (ipa_method_modify_stmt,
635 ipa_callsite_compute_param): Support for SSA.
636 * ipa-prop.h (ipcp_driver): Remove declaration.
637 (IS_VALID_TREE_MAP_INDEX): Add define.
639 2007-01-24 Geoffrey Keating <geoffk@apple.com>
641 * unwind-dw2.c (execute_stack_op): Handle DW_OP_swap.
643 2007-01-24 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
645 PR middle-end/30447
646 * builtins.c (fold_builtin_cabs): Use MPFR to evaluate a
647 constant argument to cabs and do it without checking for
648 -funsafe-math-optimizations.
650 2007-01-24 Douglas Gregor <dgregor@osl.iu.edu>
652 * c-common.h (RID_FIRST_CXX0X): New.
653 (RID_LAST_CXX0X): New.
654 * c-opts.c (c_common_handle_option): -Wc++0x-compat is triggered
655 by -Wall.
656 * c.opt (Wc++0x-compat): New.
657 * doc/invoke.texi (-Wc++0x-compat): Document.
659 2007-01-24 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
661 * builtins.c (fold_builtin_carg): New.
662 (fold_builtin_1): Use it.
664 2007-01-24 Jan Hubicka <jh@suse.cz>
666 * ipa-inline.c (cgraph_decide_inlining): Initialize initial_insns.
668 * ipa-inline.c (initial_insns, max_insns): Delete.
669 (compute_max_insns): New function.
670 (cgraph_decide_inlining_of_small_function): Use it; take minimal amount
671 of insns as base for code growth.
672 (cgraph_decide_inlining): Make initial_insns local; do not compute
673 max_insns.
674 * params.def (PARAM_INLINE_UNIT_GROWTH): Set to 60.
675 * doc/invoke.texi (inline-unit-growth): Update docs.
677 2007-01-24 Jakub Jelinek <jakub@redhat.com>
679 * config/i386/i386.h (x86_cmpxchg16b): Remove const.
680 (TARGET_CMPXCHG16B): Define to x86_cmpxchg16b.
681 * config/i386/i386.c (x86_cmpxchg16b): Remove const.
682 (override_options): Add PTA_CX16 flag. Set x86_cmpxchg16b
683 for CPUs that have PTA_CX16 set.
685 PR middle-end/27416
686 * gimplify.c (omp_check_private): New function.
687 (gimplify_scan_omp_clauses): Use it for
688 firstprivate/lastprivate/reduction.
690 PR middle-end/30494
691 * gimplify.c (omp_add_variable): Don't call omp_notice_variable
692 on TYPE_SIZE_UNIT for GOVD_LOCAL VLAs.
694 PR middle-end/30421
695 * omp-low.c (lower_omp_for_lastprivate): Add dlist argument.
696 If lower_lastprivate_clauses emits some statements, append them
697 to dlist rather than body_p and to body_p append an initializer.
698 (lower_omp_for): Adjust caller.
700 2007-01-24 Steve Ellcey <sje@cup.hp.com>
702 * target.h (globalize_decl_name): New.
703 * target-def.h (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
704 * output.h (default_globalize_decl_name): New.
705 * varasm.c (asm_output_bss): Use globalize_decl_name instead of
706 globalize_label.
707 (globalize_decl): Ditto.
708 (default_globalize_decl_name): New.
709 * config/ia64/ia64.c (ia64_globalize_decl_name): New.
710 (ia64_handle_version_id_attribute): New.
711 (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
712 (ia64_asm_output_external): Use globalize_decl_name instead
713 of globalize_label.
714 * doc/extend.texi (version_id): New pragma.
715 * doc/tm.texi (ARGET_ASM_GLOBALIZE_DECL_NAME): New target hook.
717 2007-01-24 Andreas Krebbel <krebbel1@de.ibm.com>
719 * unwind-dw2-fde.c (get_cie_encoding): Replaced _Unwind_Word with
720 _uleb128_t and _Unwind_SWord with _sleb128_t.
721 * unwind-dw2.c (extract_cie_info, execute_stack_op, execute_cfa_program,
722 uw_frame_state_for, uw_update_context_1): Likewise.
723 * unwind-c.c (parse_lsda_header, PERSONALITY_FUNCTION): Likewise.
724 * unwind-pe.h (read_uleb128, read_sleb128,
725 read_encoded_value_with_base): Likewise.
726 * unwind-generic.h: Define _sleb128_t and _uleb128_t types.
728 2007-01-24 Richard Guenther <rguenther@suse.de>
730 * builtins.c (expand_builtin_cexpi): Get the fndecl
731 for cexp in the correct way.
733 2007-01-24 Jan Hubicka <jh@suse.cz>
735 * tree-ssa-dce.c (eliminate_unnecesary_stmts): Remove dead LHS of calls.
737 2007-01-24 Andreas Krebbel <krebbel1@de.ibm.com>
739 * c-cppbuiltin.c (builtin_define_type_sizeof): New function.
740 (c_cpp_builtins): New builtin macros: __SIZEOF_INT__, __SIZEOF_LONG__,
741 __SIZEOF_LONG_LONG__, __SIZEOF_SHORT__, __SIZEOF_POINTER__,
742 __SIZEOF_FLOAT__, __SIZEOF_DOUBLE__, __SIZEOF_LONG_DOUBLE__,
743 __SIZEOF_SIZE_T__, __SIZEOF_WCHAR_T__, __SIZEOF_WINT_T__ and
744 __SIZEOF_PTRDIFF_T__.
745 * doc/cpp.texi: Documentation for the new builtin macros added.
747 2007-01-24 Uros Bizjak <ubizjak@gmail.com>
749 * config/i386/i386.md (tanxf2, tan<mode>2, atan<mode>2, log<mode>2,
750 log10<mode>2, log2<mode>2, expxf2, exp10xf2, exp2xf2): Use op2
751 instead of operands[2] to avoid access past the end of array.
753 2007-01-24 Richard Sandiford <richard@codesourcery.com>
755 * reload1.c (emit_reload_insns): Pass the reload register
756 for a non-spill output reload through forget_old_reloads_1.
758 2007-01-23 Joseph Myers <joseph@codesourcery.com>
760 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Define
761 _SOFT_DOUBLE if doubles use software floating-point.
762 * config/rs6000/libgcc-ppc-glibc.ver: Export additional long
763 double functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
764 * config/rs6000/darwin-ldouble.c: Also compile functions for
765 hard-float without FPRs. Use fmsub function for all __NO_FPRS__
766 cases. Compile extra functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
767 * config/rs6000/linuxspe.h (SUBSUBTARGET_OVERRIDE_OPTIONS): Remove
768 commented-out long double override.
769 (CPP_LONGDOUBLE_DEFAULT_SPEC): Likewise.
770 * config/rs6000/eabispe.h: Likewise.
771 * config/rs6000/rs6000.c (rs6000_override_options): Don't override
772 long double for non-SPE.
773 (rs6000_handle_option): Likewise.
774 (invalid_e500_subreg): Disallow more subregs involding DImode,
775 DFmode, TImode or TFmode.
776 (rs6000_legitimate_offset_address_p): Check TFmode offsets for
777 E500 double.
778 (legitimate_lo_sum_address_p): Also check for TFmode for E500
779 double.
780 (rs6000_legitimize_address): Also handle TFmode for E500 double.
781 (rs6000_legitimize_reload_address): Also handle TFmode for E500
782 double.
783 (rs6000_legitimate_address): Also check for TFmode for E500
784 double.
785 (rs6000_emit_move): Use DFmode subregs of TFmode for E500 double.
786 (spe_build_register_parallel): Handle TFmode and TCmode.
787 (rs6000_spe_function_arg): Handle TFmode and TCmode for E500
788 double.
789 (function_arg): Handle TFmode and TCmode for E500 double.
790 (rs6000_init_libfuncs): Initialize extra libfuncs for soft double
791 in general.
792 (print_operand): Handle TFmode and TImode for %y.
793 (rs6000_generate_compare): Handle TFmode comparisons for E500
794 double.
795 (spe_func_has_64bit_regs_p): Check for TFmode for E500 double.
796 (rs6000_function_value): Handle TFmode and TCmode for E500 double.
797 (rs6000_libcall_value): Handle TFmode and TCmode for E500 double.
798 * config/rs6000/rs6000.h (CANNOT_CHANGE_MODE_CLASS): Check for
799 TFmode for E500 double.
800 * config/rs6000/rs6000.md (FP): Allow TF for E500 double.
801 (floatsidf2): Enable for E500 double.
802 (movtf_softfloat): Use rs6000_nonimmediate_operand.
803 (extenddftf2): Change to extenddftf2_fprs.
804 (extenddftf2): Call gen_spe_extenddftf2 or gen_extenddftf2_fprs
805 depending on TARGET_E500_DOUBLE.
806 (extendsftf2): Enable for E500 double.
807 (trunctfdf2): Enable for E500 double.
808 (trunctfsf2): Change to trunctfsf2_fprs.
809 (trunctfsf2): Call gen_spe_trunctfsf2 or gen_trunctfsf2_fprs
810 depending on TARGET_E500_DOUBLE.
811 (floatsitf2): Enable for E500 double.
812 (fix_trunctfsi2): Change to fix_trunctfsi2_fprs.
813 (fix_trunctfsi2): Call gen_spe_fix_trunctfsi2 or
814 gen_fix_trunctfsi2_fprs depending on TARGET_E500_DOUBLE.
815 (negtf2): Change to negtf2_internal.
816 (negtf2): New expander.
817 (abstf2): Enable for E500 double. Call gen_spe_abstf2_tst,
818 gen_spe_abstf2_cmp or gen_abstf2_internal depending on
819 TARGET_E500_DOUBLE and flag_unsafe_math_optimizations.
820 (movdi_internal32): Use rs6000_nonimmediate_operand.
821 (unnamed splitter): Likewise.
822 * config/rs6000/spe.md (CMPTFEQ_GPR, TSTTFEQ_GPR, CMPTFGT_GPR,
823 TSTTFGT_GPR, CMPTFLT_GPR, TSTTFLT_GPR): New unspecs.
824 (SPE64TF, DITI): New mode macros.
825 (frob_df_di): Change to frob_<SPE64:mode>_<DITI:mode>; allow more
826 modes.
827 (frob_tf_ti): New.
828 (frob_<mode>_di_2): New.
829 (frob_tf_di_8_2): New.
830 (frob_di_df): Change to frob_di_<mode>; allow more modes.
831 (frob_ti_tf): New.
832 (frob_di_df_2): Change to frob_<DITI:mode>_<SPE64:mode>_2; allow
833 more modes.
834 (frob_ti_<mode>_8_2): New.
835 (frob_ti_tf_2): New.
836 (mov_si<mode>_e500_subreg0, mov_si<mode>_e500_subreg0_2,
837 mov_si<mode>_e500_subreg4, mov_si<mode>_e500_subreg4_2): Allow
838 TFmode.
839 (mov_sitf_e500_subreg8, mov_sitf_e500_subreg8_2,
840 mov_sitf_e500_subreg12, mov_sitf_e500_subreg12_2): New.
841 (spe_trunctfdf2_internal1, spe_trunctfsf2, spe_extenddftf2,
842 spe_fix_trunctfsi2, spe_fix_trunctfsi2_internal,
843 spe_negtf2_internal, spe_abstf2_cmp, spe_abstf2_tst): New.
844 (cmptfeq_gpr, tsttfeq_gpr, cmptfgt_gpr, tsttfgt_gpr, cmptflt_gpr,
845 tsttflt_gp): New.
847 2007-01-23 Ian Lance Taylor <iant@google.com>
849 * Makefile.in (OBJS-common): Reformat, alphabetize, but put
850 insn-*.o first.
851 (OBJS-archive): Reformat, alphabetize.
852 (OBJS): Change out_object_file to OBJS-md.
854 2007-01-23 Uros Bizjak <ubizjak@gmail.com>
856 * config/i386/i386.md: Use REG_P, MEM_P, CONST_INT_P, LABEL_P,
857 JUMP_P and CALL_P predicates where applicable.
858 * config/i386/i386.c: Ditto.
859 * config/i386/i386.md: Ditto.
860 * config/i386/mmx.md: Ditto.
861 * config/i386/predicates.md: Ditto.
863 2007-01-22 Andreas Schwab <schwab@suse.de>
865 * config/m68k/m68k.h: Fix comment.
867 2007-01-22 Jan Hubicka <jh@suse.cz>
869 * passes.c (init_optimization_passes): Do not rerun
870 pass_early_warn_uninitialized.
872 2007-01-22 Richard Guenther <rguenther@suse.de>
874 PR tree-optimization/30038
875 * tree-ssa-math-opts.c (maybe_record_sincos): New static helper
876 function.
877 (execute_cse_sincos_1): Likewise.
878 (execute_cse_sincos): Likewise.
879 (gate_cse_sincos): Likewise.
880 (pass_cse_sincos): New pass CSEing sin() and cos() calls using
881 the cexpi() canonicalization of sincos().
882 * tree-pass.h (pass_cse_sincos): Declare.
883 * passes.c (init_optimization_passes): New pass pas_cse_sincos.
885 2007-01-21 Eric Botcazou <ebotcazou@libertysurf.fr>
887 PR rtl-optimization/29329
888 * combine.c (replaced_rhs_insn): Rename to i2mod.
889 (replaced_rhs_value): Rename to i2mod_new_rhs.
890 (i2mod_old_rhs): New global variable.
891 (combine_instructions): Adjust for above change. Save a copy of
892 the old RHS into i2mod_old_rhs when the contents of a REG_EQUAL
893 note are substituted in the second instruction.
894 (distribute_notes) <REG_DEAD>: Adjust for above change. Do not
895 ditch the note if it pertains to the second eliminated register
896 and this register is mentioned in i2mod_old_rhs.
898 Revert:
899 2006-09-12 Eric Botcazou <ebotcazou@libertysurf.fr>
901 * combine.c (distribute_notes) <REG_DEAD>: Do not consider SETs past
902 the insn to which the note was originally attached.
904 2007-01-21 Jan Hubicka <jh@suse.cz>
906 * ipa-inline.c (inlining_mode): Comment, move up.
907 (cgraph_decide_inlining_incrementally): Do not perform inlining
908 itself; fix handling of flattening of self recursive functions.
909 (cgraph_find_cycles): Remove.
910 (cgraph_flatten_node): Remove.
911 (cgraph_decide_inlining): Use incremental inliner to handle flattening.
912 (try_inline): New function.
913 (cgraph_early_inlining): Update call of
914 cgraph_decide_inlining_incrementally. Apply inlining here.
915 (apply_inline): Update call of cgraph_decide_inlining_incrementally.
917 2007-01-21 Dirk Mueller <dmueller@suse.de>
919 PR bootstrap/30511
920 * tree-vrp.c (check_array_bounds): do not warn
921 about ADDR_EXPR's of ARRAY_REF's which are immediately
922 used in binary expressions.
924 2007-01-21 Ira Rosen <irar@il.ibm.com>
926 * tree-vectorizer.h (struct _stmt_vec_info): Add new field
927 read_write_dep and macros for its access.
928 * tree-vectorizer.c (new_stmt_vec_info): Initialize the new field.
929 * tree-vect-analyze.c (vect_analyze_data_ref_dependence): Remove
930 argument, call vect_check_interleaving for every independent pair of
931 data-refs. Mark loads that access the same memory location as a store
932 in the loop.
933 (vect_check_dependences): Remove.
934 (vect_analyze_data_ref_dependences): Remove vect_check_dependences
935 call, fix the call to vect_analyze_data_ref_dependence.
936 (vect_analyze_data_ref_access): For statements that access the same
937 data-ref, check that they are not stores; for loads, check that there
938 is no store that access the same location.
940 2007-01-20 Roger Sayle <roger@eyesopen.com>
941 Joseph Myers <joseph@codesourcery.com>
943 * doc/invoke.texi (-fdump-rtl-combine): Fix under/overfull hbox.
944 (-fdump-rtl-stack): Likewise.
945 (-fno-signed-zeros): Use @minus{} for a minus sign. Correct typo.
946 (-mcheck-zero-division, -mcpu): Fix under/overfull hbox.
947 (-mpt-fixed): Use @minus{} for minus sign.
948 (Using Precompiled Headers): Fix under/overfull hbox.
950 2007-01-20 Jan Hubicka <jh@suse.cz>
952 * tree-flow.h (struct stmt_ann_d): Move references_memory to proper
953 place within annotation.
955 2007-01-20 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
957 * pa.c (output_move_double): Change array size of xoperands to 4.
959 2007-01-20 Jan Hubicka <jh@suse.cz>
961 * tree-tailcall.c (adjust_return_value): Do not use RESULT_DECL
962 as temporary.
964 2007-01-19 Ian Lance Taylor <iant@google.com>
966 * expmed.c (expand_divmod) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]: Cast
967 constant to unsigned HOST_WIDE_INT before negating.
969 2007-01-19 Ian Lance Taylor <iant@google.com>
971 * tree-ssa-operands.h (struct vuse_vec_d): Change num_vuse field
972 to unsigned.
973 (VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Use unsigned comparison.
974 (VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
975 (SET_VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Likewise.
976 (SET_VUSE_ELEMENT_VAR) [ENABLE_CHECKING]: Likewise.
977 (SET_VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
978 (realloc_vdef, realloc_vuse): Change second parameter to
979 unsigned.
980 (ssa_operand_iterator_d): Change vuse_index and mayuse_index
981 fields to unsigned.
982 * tree-ssa-operands.c (realloc_vop): Change num_elem parameter to
983 unsigned. Change x and lim locals to unsigned.
984 (realloc_vdef, realloc_vuse): Change num_elem parameter to
985 unsigned.
986 (finalize_ssa_vuse_ops): Change old_i local to unsigned.
987 (copy_virtual_operands): Change i and n locals to unsigned.
989 2007-01-19 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
991 PR middle-end/29335
992 * builtins.c (fold_builtin_1): Handle builtin fdim.
994 2007-01-20 Jan Hubicka <jh@suse.cz>
996 * tree-ssa.c (init_tree_ssa): Do not call init_alias_heapvars.
997 * tree-ssa-structalias.c (compute_points_to_sets): Do call
998 init_alias_heapvars.
999 (init_alias_heapvars): Initialize only when not already initialized.
1000 (delete_alias_heapvars): Set heapvar_for_stmt to NULL.
1002 2007-01-19 Roger Sayle <roger@eyesopen.com>
1004 * common.opt (fsigned-zeros): New command line option.
1005 * flags.h (HONOR_SIGNED_ZEROS): Control via flag_signed_zeros instead
1006 of flag_unsafe_math_optimizations.
1007 * opts.c (set_fast_math_flags): The -ffast-math command line option
1008 implies -fno-signed-zeros.
1009 (fast_math_flags_set_p): Likewise.
1011 * doc/invoke.texi: Document new -fno-signed-zeros option, and update
1012 the documentation of -ffast-math appropriately. Wrap long lines.
1014 2007-01-19 Steve Ellcey <sje@cup.hp.com>
1016 * system.h (ASM_MAKE_LABEL_LINKONCE): Poison.
1017 * varasm.c (globalize_decl): Remove ASM_MAKE_LABEL_LINKONCE ifdef.
1019 2007-01-19 Tomas Bily <tbily@suse.cz>
1021 * cgraphunit.c (cgraph_finalize_function): Updating of pid
1022 * tree-profile.c:
1023 (tree_init_ic_make_global_vars): New function
1024 (tree_init_edge_profiler): call of tree_init_ic_make_global_vars
1025 (tree_gen_ic_profiler): New function
1026 (tree_gen_ic_func_profiler): New function
1027 (tree_profiling): Added calling of tree_gen_ic_func_profiler
1028 (tree_profile_hooks): Added hook for indirec/virtual calls
1029 * value-prof.c (tree_find_values_to_profile): New case for
1030 indirect calls
1031 (tree_values_to_profile): Call for determining indirect/virtual
1032 counters
1033 (tree_indirect_call_to_profile): New function
1034 (tree_ic_transform): New function
1035 (tree_ic): New function
1036 (find_func_by_pid): New function
1037 (init_pid_map): New function
1038 (tree_value_profile_transformations): Added check for
1039 indirect/virtual call transformation
1040 * value-prof.h (enum hist_type): New counter type for
1041 indirect/virtual calls
1042 (profile_hooks): Added new hook for profiling indirect/virtual
1043 calls
1044 * profile.c (instrument_values): New case for indirect/virtual
1045 call added
1046 * gcov-io.h (GCOV_LAST_VALUE_COUNTER): Changed to 6
1047 (GCOV_COUNTER_V_INDIR): New counter type
1048 (GCOV_COUNTER_NAMES): New name of counter "indirect" added
1049 (GCOV_MERGE_FUNCTIONS): New merge function for indirect/virtual
1050 call added
1051 * cgraph.c: Definition of cgraph_max_pid
1052 (cgraph_create_node): Default init of pid attribute
1053 * cgraph.h: Declaration of cgraph_max_pid
1054 (struct cgraph_node): Added pid attribute
1055 * libgcov.c (__gcov_indirect_call_profiler): New function
1056 (__gcov_one_value_profiler_body): New function
1057 (__gcov_one_value_profiler): Body was moved to
1058 __gcov_one_value_profiler_body and calls it
1060 2007-01-19 Basile Starynkevitch <basile@starynkevitch.net>
1062 * doc/gty.texi (Options): Document the mark_hook option to GTY.
1063 * gengtype.c (write_types_data, write_func_for_structure,
1064 write_types, ggc_wtd, pch_wtd): Add skip_hooks to
1065 write_types_data, ggc_wtd, pch_wtd for processing mark_hook.
1066 (walk_type, write_func_for_structure): Generate the mark_hook if
1067 needed.
1069 2007-01-19 Jan Hubicka <jh@suse.cz>
1071 * ipa-inline.c (cgraph_decide_inlining_incrementally): Instead of
1072 'early' argument take inlining mode argument specifying whether to
1073 inline for size/speeed or all functions; add support for flattening;
1074 improve dumpting.
1075 (cgraph_early_inlining): Update call of decide_inlining_incrementally.
1077 2007-01-19 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
1079 PR c++/17947
1080 * toplev.c (warn_deprecated_use): Use %qD instead of %qs to print
1081 the name of the declared identifier.
1083 2007-01-19 Dirk Mueller <dmueller@suse.de>
1085 * config/i386.h (CONDITIONAL_REGISTER_USAGE): Store
1086 result of PIC_OFFSET_TABLE_REGNUM in temporary variable to avoid
1087 duplicate evaluation.
1089 2007-01-19 Uros Bizjak <ubizjak@gmail.com>
1091 * config/i386/i386.md (acos<mode>2): Rename from acossf2 and acosdf2.
1092 Macroize expander using X87MODEF12 mode macro. Extend operand 1
1093 to XFMode, use acosxf2 and truncate result to requested mode.
1094 Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
1095 (asin<mode>2): Similarly, with asin expanders.
1096 (*fscalexf4_i387): Rename from *fscalexf4.
1097 (expNcorexf3): New expander.
1098 (expxf2, exp10xf2, exp2xf2): Use expNcorexf3 expander.
1099 (exp<mode>2): Rename from expsf2 and expdf2. Macroize expander using
1100 X87MODEF12 mode macro. Extend operand 1 to XFMode, use expxf2 and
1101 truncate result to requested mode. Use SSE_FLOAT_MODE_P to disable
1102 patterns for SSE math.
1103 (exp10<mode>2): Similarly, with exp10 expanders.
1104 (exp2<mode>2): Similarly, with exp2 expanders.
1105 (expm1<mode>2): Similarly, with expm1 expanders.
1106 (ldexp<mode>3): Similarly, with ldexp expanders.
1107 (log<mode>2, log10<mode>2, log2<mode>2, log1p<mode>2, logb<mode>2):
1108 Use gen_truncxf<mode>2_i387_noop to truncate result.
1110 2007-01-19 Richard Sandiford <richard@codesourcery.com>
1112 * config/m68k/m68k.h (PREFERRED_STACK_BOUNDARY): Define to 32
1113 for ColdFire targets.
1115 2007-01-19 Nathan Sidwell <nathan@codesourcery.com>
1116 Richard Sandiford <richard@codesourcery.com>
1118 * config/m68k/m68k.h (M68K_STATIC_CHAIN_REG_NAME): New macro.
1119 (INITIALIZE_TRAMPOLINE): Use STATIC_CHAIN_REGNUM.
1120 (__transfer_from_trampoline): Use M68K_STATIC_CHAIN_REG_NAME.
1121 * config/m68k/m68kelf.h (STATIC_CHAIN_REGNUM): Override.
1122 (M68K_STATIC_CHAIN_REG_NAME): Likewise.
1123 * config/m68k/netbsd-elf.h (M68K_STATIC_CHAIN_REG_NAME): Likewise.
1125 2007-01-19 Richard Sandiford <richard@codesourcery.com>
1127 * config/m68k/m68k.md (adddi_dilshr32): Rename to...
1128 (*adddi_dilshr32): ...this. Fix formatting. Remove commented-out
1129 non-canonical pattern. Restrict to !TARGET_COLDFIRE.
1130 (*adddi_dilshr32_cf): New pattern.
1131 (adddi3, subdi3): Remove first alternatives.
1133 2007-01-19 Richard Sandiford <richard@codesourcery.com>
1135 * config/m68k/m68k.c (notice_update_cc): If an SFmode move is
1136 implemented using move.l, do not use its cc result for floating-point
1137 comparisons.
1139 2007-01-19 Richard Sandiford <richard@codesourcery.com>
1141 * config/m68k/m68k.h (EXTRA_CONSTRAINT): Stop the 'T' constraint
1142 from accepting 's' constraints if flag_pic.
1144 2007-01-19 Richard Sandiford <richard@codesourcery.com>
1146 * config/m68k/m68k.md (bordered, bunordered, buneq, bunge, bungt)
1147 (bunle, bunlt, bltgt, bordered_rev, bunordered_rev, buneq_rev)
1148 (bunge_rev, bungt_rev, bunle_rev, bunlt_rev, bltgt_rev): Change
1149 condition from TARGET_68881 to TARGET_HARD_FLOAT.
1151 2007-01-19 Sandra Loosemore <sandra@codesourcery.com>
1153 * longlong.h (count_leading_zeros, COUNT_LEADING_ZEROS_0): Add
1154 ColdFire alternatives.
1155 * config/m68k/m68k.h (CLZ_DEFINED_VALUE_AT_ZERO): New macro.
1156 * config/m68k/m68k.md (clzsi2): Define for ColdFire
1157 architectures that support ff1 instruction.
1159 2007-01-19 Richard Sandiford <richard@codesourcery.com>
1160 Julian Brown <julian@codesourcery.com>
1162 * config/m68k/m68k.h (CONST_OK_FOR_LETTER_P): Add an 'R' case.
1163 * config/m68k/m68k.md (*movsi_cfv4): Fold into...
1164 (*movsi_cf): ...here. Remove unnecessary 'R' from 'Rg'.
1165 Add commentary.
1167 2007-01-19 Richard Sandiford <richard@codesourcery.com>
1169 * config/m68k/m68k-protos.h (valid_mov3q_const): Take a HOST_WIDE_INT
1170 and return a bool.
1171 (output_move_const_into_data_reg, output_move_simode_const): Delete.
1172 * config/m68k/m68k.c (const_method, const_int_cost): Take a
1173 HOST_WIDE_INT instead of an rtx.
1174 (m68k_rtx_costs): Update call accordingly.
1175 (output_move_const_into_data_reg): Likewise. Fix formatting.
1176 (valid_mov3q_const): Take a HOST_WIDE_INT instead of an rtx.
1177 Return a bool.
1178 (output_move_simode_const): Update calls after above changes.
1179 Rework to use automatic variables and predicates like MEM_P.
1180 * config/m68k/m68k.md (pushexthisi_const): Update call to
1181 valid_mov3q_const.
1183 2007-01-19 Dirk Mueller <dmueller@suse.de>
1185 * tree-ssa-alias.c (perform_var_substitution): Fix typo
1186 in dump_flags test.
1188 2007-01-19 Richard Guenther <rguenther@suse.de>
1190 * builtins.c (expand_builtin_cexpi): Fall back to expanding
1191 via cexp in case sincos is not available.
1193 2007-01-19 Richard Guenther <rguenther@suse.de>
1195 * doc/tm.texi (TARGET_HAS_SINCOS): Document new target macro.
1196 * defaults.h (TARGET_HAS_SINCOS): Default to off.
1197 * config/linux.h (TARGET_HAS_SINCOS): Set to on if we have glibc.
1198 * config/alpha/linux.h (TARGET_HAS_SINCOS): Likewise.
1199 * config/sparc/linux.h (TARGET_HAS_SINCOS): Likewise.
1200 * config/sparc/linux64.h (TARGET_HAS_SINCOS): Likewise.
1201 * config/rs6000/linux.h (TARGET_HAS_SINCOS): Likewise.
1202 * config/rs6000/linux64.h (TARGET_HAS_SINCOS): Likewise.
1204 2007-01-19 Uros Bizjak <ubizjak@gmail.com>
1206 * config/i386/i386.md (*fpatanxf3_i387, fpatan_extend<mode>xf3_i387):
1207 New insn patterns.
1208 (atan2sf3_1, atan2df3_1, atan2xf3_1): Remove insn patterns.
1209 (atan2xf3): Directly generate RTL pattern.
1210 (atan2<mode>3): Rename from atan2sf3 and atan2df3 and macroize insn
1211 patterns using X87MODEF12 mode macro. Use fpatan_extend<mode>xf3_i387
1212 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
1213 disable patterns for SSE math.
1214 (atan<mode>2): Rename from atansf2 and atandf2 and macroize insn
1215 patterns using X87MODEF12 mode macro. Use fpatan_extend<mode>xf3_i387
1216 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
1217 disable patterns for SSE math.
1219 2007-01-19 Alexandre Oliva <aoliva@redhat.com>
1221 * libgcc-std.ver: Fix typo in %inherit for GCC_4.3.0.
1223 2007-01-18 Roger Sayle <roger@eyesopen.com>
1225 * fold-const.c (fold_unary) <VIEW_CONVERT_EXPR>: Optimize away a
1226 VIEW_CONVERT_EXPR to the same type as it's operand.
1228 2007-01-18 David Edelsohn <edelsohn@gnu.org>
1230 * config/rs6000/darwin-ldouble.c: Only build _SOFT_FLOAT if
1231 configured for long double 128.
1233 2007-01-18 Mike Stump <mrs@apple.com>
1235 * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Fix build
1236 error.
1238 2007-01-18 Michael Meissner <michael.meissner@amd.com>
1240 * i386.c (ix86_compute_frame_layout): Make fprintf's in #if 0 code
1241 type correct.
1243 2007-01-18 Jan Hubicka <jh@suse.cz>
1245 * tree-ssa-operands.c (vop_free_bucket_size): Never return value
1246 greater than NUM_VOP_FREE_BUCKETS.
1248 2007-01-18 Daniel Berlin <dberlin@dberlin.org>
1250 * tree-ssa-structalias.c: Update comments.
1251 (ptabitmap_obstack): Removed.
1252 (pta_obstack): New.
1253 (oldpta_obstack): Ditto.
1254 (stats): Add a few members.
1255 (struct variable_info): Remove node, complex, address_taken, and
1256 indirect_target members. Add oldsolution member.
1257 (new_var_info): Do not initialize removed members.
1258 (constraint_expr_type): Remove INCLUDES.
1259 (constraint_graph): Add size, implicit_preds, rep,
1260 indirect_cycles, eq_rep, label, direct_nodes, and complex members.
1261 (FIRST_REF_NODE): New macro.
1262 (LAST_REF_NODE): Ditto.
1263 (FIRST_ADDR_NODE): Ditto.
1264 (find): New function.
1265 (unite): Ditto.
1266 (dump_constraint): Do not handle INCLUDES.
1267 (insert_into_complex): Do not insert duplicate constraints.
1268 (condense_varmap_nodes): Renamed and rewritten into ...
1269 (merge_node_constraints): This. Also fix bug in handling of
1270 offseted copy constraints.
1271 (clear_edges_for_node): No longer need to deal with preds at all,
1272 or removing associated preds/succs.
1273 (merge_graph_nodes): Deal with indirect_cycles.
1274 Don't deal with predecessors.
1275 (add_implicit_graph_edge): New function.
1276 (add_pred_graph_edge): Ditto.
1277 (add_graph_edge): Don't deal with predecessors.
1278 (build_constraint_graph): Removed.
1279 (build_pred_graph): New function.
1280 (build_succ_graph): Ditto.
1281 (struct scc_info): Removed in_component. Added roots, dfs, and
1282 node_mapping. Remove visited_index, unification_queue.
1283 (scc_visit): Deal with union-find we do now.
1284 Deal with cycles with REF nodes.
1285 (collapse_nodes): Renamed and rewritten to ...
1286 (unify_nodes): This.
1287 (process_unification_queue): Removed.
1288 (topo_visit): Cleanup
1289 (do_da_constraint): Use find.
1290 (do_sd_constraint): Ditto.
1291 (do_ds_constraint): Ditto.
1292 (do_complex_constraint): Ditto.
1293 (init_scc_info): Update for removed and added members.
1294 (find_and_collapse_graph_cycles): Renamed and rewritten into ...
1295 (find_indirect_cycles): This.
1296 (equivalence_class): New variable.
1297 (label_visit): New function.
1298 (perform_variable_substitution): Rewritten.
1299 (free_var_substitution_info): New function.
1300 (find_equivalent_node): Ditto.
1301 (move_complex_constraints): Ditto.
1302 (eliminate_indirect_cycles): Ditto.
1303 (solve_graph): Only propagate changed bits.
1304 Use indirect cycle elimination.
1305 Use find.
1306 (tree_id_t): Rename to tree_vi_t, delete id member, add vi member.
1307 (tree_id_eq): Renamed to ...
1308 (tree_vi_eq): This. Update for member change
1309 (insert_id_for_tree): Renamed and rewritten to ...
1310 (insert_vi_for_tree): This.
1311 (lookup_id_for_tree): Renamed and rewritten to ...
1312 (lookup_vi_for_tree): This.
1313 (get_id_for_tree): Renamed and rewritten to ...
1314 (get_vi_for_tree): Ditto.
1315 (get_constraint_exp_from_ssa_var): Update to use get_vi_for_tree.
1316 (process_constraint): Don't handle INCLUDES.
1317 Remove special ADDRESSOF case.
1318 (find_func_aliases): Rewrite to use vi functions instead of id
1319 ones.
1320 (create_function_info_for): Ditto.
1321 (create_variable_info_for): Ditto.
1322 (intra_create_variable_infos): Ditto.
1323 (merge_smts_into): Ditto.
1324 (find_what_p_points_to): Ditto.
1325 (init_base_vars): Ditto.
1326 (init_alias_vars): Ditto.
1327 (remove_preds_and_fake_succs): New function.
1328 (dump_sa_points_to_info): Dump new stats.
1329 (dump_solution_for_var): Use find.
1330 (set_used_smts): Fix formatting.
1331 (compute_points_to_sets): Updated for new functions.
1332 (ipa_pta_execute): Ditto.
1334 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
1335 Richard Sandiford <richard@codesourcery.com>
1337 * doc/tm.texi (TARGET_FUNCTION_VALUE): Expand documentation of
1338 parallels.
1339 * calls.c (expand_call): If the return value is a PARALLEL,
1340 extract its first member.
1341 * config/m68k/linux.h (FUNCTION_EXTRA_EPILOGUE): Remove.
1342 * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
1343 use FUNCTION_EXTRA_EPILOGUE.
1344 (m68k_function_value): Return a PARALLEL if the return value
1345 is of a pointer type.
1346 * config/m68k/netbsd-elf.h (current_function_returns_pointer)
1347 (FUNCTION_EXTRA_EPILOGUE): Remove.
1348 * config/m68k/m68k.md (D0_REG): New constant.
1350 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
1352 * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
1353 output a NOP for empty epilogues.
1355 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1357 * config/m68k/m68k.c (m68k_use_return_insn): Update comments
1358 before function. Extend register save check to include all
1359 registers, not just integer ones.
1361 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
1363 * config/m68k/m68k-protos.h (use_return_insn): Rename to...
1364 (m68k_use_return_insn): ...this.
1365 * config/m68k/m68k.h (USE_RETURN_INSN): Delete.
1366 * config/m68k/m68k.c (use_return_insn): Rename to...
1367 (m68k_use_return_insn): ...this.
1368 * config/m68k/m68k.md (return): Use m68k_use_return_insn instead
1369 of USE_RETURN_INSN.
1371 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
1373 * config/m68k/fpgnulib.c (__truncdfsf2): Implement round to
1374 nearest even, fix denormal rounding overflow.
1376 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1378 * config/m68k/m68k.md (movsf_cf_hard): Use fsmove instead of
1379 f%$smove and f%$move.
1380 (movdf_cf_hard): Use fdmove for cases 0 and 3 and fmove for case 1.
1381 (extendsfdf2_cf): Use fdmove instead of f%&move.
1382 (truncdfsf2_cf): Use fsmove instead of f%$smove.
1383 (add<mode>3_cf, sub<mode>3_cf): Use <FP:prec> instead of <FP:round>.
1385 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1387 * config/m68k/m68k.md (movdf_cf_hard): Use output_move_double for
1388 GPR<-GPR moves.
1390 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1392 * real.h (real_format): Add a canonical_nan_lsbs_set field.
1393 (coldfire_single_format): Declare.
1394 (coldfire_double_format): Likewise.
1395 * real.c (encode_ieee_single): Use canonical_nan_lsbs_set instead
1396 of qnan_msb_set to determine the lower bits of a canonical
1397 NaN significand.
1398 (encode_ieee_double): Likewise.
1399 (encode_ieee_quad): Likewise.
1400 (ieee_single_format): Initialize canonical_nan_lsbs_set.
1401 (mips_single_format): Likewise.
1402 (ieee_double_format): Likewise.
1403 (mips_double_format): Likewise.
1404 (ieee_extended_motorola_format): Likewise.
1405 (ieee_extended_intel_96_format): Likewise.
1406 (ieee_extended_intel_128_format): Likewise.
1407 (ieee_extended_intel_96_round_53_format): Likewise.
1408 (ibm_extended_format): Likewise.
1409 (mips_extended_format): Likewise.
1410 (ieee_quad_format): Likewise.
1411 (mips_quad_format): Likewise.
1412 (vax_f_format): Likewise.
1413 (vax_d_format): Likewise.
1414 (vax_g_format): Likewise.
1415 (i370_single_format): Likewise.
1416 (i370_double_format): Likewise.
1417 (decimal_single_format): Likewise.
1418 (decimal_double_format): Likewise.
1419 (decimal_quad_format): Likewise.
1420 (c4x_single_format): Likewise.
1421 (c4x_extended_format): Likewise.
1422 (real_internal_format): Likewise.
1423 (coldfire_single_format): New real_format.
1424 (coldfire_double_format): Likewise.
1425 * config/pdp11/pdp11.c (pdp11_f_format): Initialize
1426 canonical_nan_lsbs_set.
1427 (pdp11_d_format): Likewise.
1428 * config/m68k/m68k.c (override_options): Override REAL_FORMAT_MODE
1429 if TARGET_COLDFIRE_CPU.
1431 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1433 * config/m68k/m68k-protos.h (m68k_output_pic_call): Delete.
1434 (output_call, m68k_legitimize_call_address): Declare.
1435 * config/m68k/m68k.h (EXTRA_CONSTRAINT): Remove unnecessary
1436 parenthesees. Add support for a 'W' constraint.
1437 (LEGITIMATE_PIC_OPERAND_P): Remove SYMBOL_REF_FLAG handling.
1438 (PRINT_OPERAND_PUNCT_VALID_P): Remove comment about 'o'.
1439 (m68k_symbolic_call, m68k_symbolic_jump): Declare.
1440 * config/m68k/m68k.c (m68k_symbolic_call, m68k_symbolic_jump): New
1441 variables.
1442 (override_options): Initialize them. Do not set flag_no_function_cse
1443 for TARGET_ID_SHARED_LIBRARY.
1444 (m68k_output_pic_call): Delete.
1445 (m68k_legitimize_call_address): New function.
1446 (print_operand): Remove the %o prefix. Handle the %p prefix.
1447 (output_call): New function.
1448 (m68k_output_mi_thunk): Use m68k_symbolic_jump. Always load the
1449 target address from the GOT if symbolic jumps are not allowed.
1450 * config/m68k/m68k.md (call, general_operand): Do not set
1451 SYMBOL_REF_FLAG. Use m68k_legitimize_call_address instead.
1452 Merge separate flag_pic and !flag_pic define_insns into...
1453 (*call, *call_value): ...these new patterns. Match the address
1454 rather than the containing MEM and require it to be a call_operand.
1455 Use output_call to generate the asm template.
1456 * config/m68k/predicates.md (const_call_operand): New predicate.
1457 (call_operand): Likewise.
1459 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
1461 * config/m68k/m68k.h (REGISTER_MOVE_COST): Simplify definition.
1462 (STACK_GROWS_DOWNWARD): Define to 1.
1463 (FUNCTION_VALUE, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P): Equivocate
1464 comments, emphasizing that these values are only defaults.
1465 * config/m68k/linux.h (LINK_SPEC): Fix formatting in #undef.
1466 * config/m68k/m68k.c (const_method): Remove trailing whitespace.
1468 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1470 * config/m68k/m68k.md (cmpsi): Remove outdated flag_pic handling.
1472 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
1473 Richard Sandiford <richard@codesourcery.com>
1475 * config/m68k/m68k.h (DATA_REGNO_P, ADDRESS_REGNO_P, INT_REGNO_P)
1476 (FP_REGNO_P): New macros.
1477 (REGNO_OK_FOR_INDEX_P, REGNO_OK_FOR_BASE_P, REGNO_OK_FOR_DATA_P)
1478 (REGNO_OK_FOR_FP_P, REG_OK_FOR_INDEX_P, REG_OK_FOR_BASE_P): Use them.
1479 (EH_RETURN_STACKADJ_RTX): Use A0_REG.
1480 * config/m68k/m68k.c (m68k_regno_mode_ok): Use the new REGNO macros.
1482 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
1484 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
1485 (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-rtems*): Add t-floatlib
1486 to $tmake_file.
1487 * config/m68k/t-floatlib: New file, extracting common code from...
1488 * config/m68k/t-m68kbare, config/m68k/t-m68kelf,
1489 * config/m68k/t-uclinux: Here.
1490 * config/m68k/fpgnulib.c: Do not compile extendeed precision
1491 routines on ColdFire targets.
1493 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
1495 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
1496 (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*)
1497 (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-linux*)
1498 (m68k-*-rtems*): Use tm_file rather than m68k/m68k.h and
1499 explicitly set MOTOROLA to 1.
1500 * config/m68k/m68k.h (MOTOROLA): Simplify definition accordingly.
1502 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1503 Nathan Sidwell <nathan@codesourcery.com>
1505 * config/m68k/m68k.h (PCC_STATIC_STRUCT_RETURN): Delete.
1506 (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Add MOTOROLA cases.
1507 * config/m68k/coff.h (REGISTER_PREFIX_MD): Delete.
1508 * config/m68k/m68020-elf.h (LIB_SPEC): Delete.
1509 * config/m68k/m68k-none.h (CC1_SPEC, CPP_SUBTARGET_SPEC): Delete.
1510 * config/m68k/m68kelf.h (IMMEDIATE_PREFIX, REGISTER_PREFIX_MD)
1511 (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Delete.
1512 (NO_DOLLAR_IN_LABEL, PCC_STATIC_STRUCT_RETURN): Don't undefine.
1513 (BSS_ASM_OP, ASM_OUTPUT_SKIP, ASM_OUTPUT_ASCII): Delete.
1514 * config/m68k/m68kemb.h (PCC_STATIC_STRUCT_RETURN): Don't undefine.
1515 * config/m68k/linux.h (SIZE_TYPE, PTRDIFF_TYPE, WCHAR_TYPE)
1516 (WCHAR_TYPE_SIZE, TARGET_OBJFMT_CPP_BUILTINS): Delete.
1517 (TARGET_OS_CPP_BUILTINS): Don't define mc68000 and mc68020 here.
1518 (DBX_REGISTER_NUMBER): Delete.
1519 * config/m68k/netbsd-elf.h (IMMEDIATE_PREFIX): Delete.
1520 (PCC_STATIC_STRUCT_RETURN): Don't undefine.
1521 * config/m68k/openbsd.h (PCC_STATIC_STRUCT_RETURN): Define.
1523 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1525 * config.gcc (m68k-*-uclinux*): Add flat.h to $tm_file.
1526 * config/flat.h: New file.
1527 * crtstuff.c (USE_PT_GNU_EH_FRAME): Don't define if
1528 OBJECT_FORMAT_FLAT.
1529 * config/m68k/m68k.h (ASM_PREFERRED_EH_DATA_FORMAT): Do not use
1530 indirect references for -msep-data or -mid-shared-library.
1531 Do not use PC-relative code addresses either.
1532 * config/m68k/m68k.c (override_options): Restrict -fPIC error
1533 to -mpcrel.
1534 * config/m68k/uclinux.h (STARTFILE_SPEC): Define. Use Scrt1.o
1535 for shared libraries and crt1.o for executables. Use crti.o and
1536 crtbegin.o.
1537 (ENDFILE_SPEC): Use crtend.o and crtn.o.
1538 (LIB_SPEC): Suppress -Rlibc.gdb if -static-libc is given.
1539 Do not add -elf2flt or -shared-lib-id options here.
1540 (LINK_SPEC): Define. Pass -elf2flt if no -elf2flt option is given.
1541 Pass -shared-lib-id if -mid-shared-library, taking the library
1542 identifier from -mshared-library-id if given, otherwise
1543 defaulting to 0.
1544 (EH_FRAME_IN_DATA_SECTION): Do not undefine.
1545 (INIT_SECTION_ASM_OP, FINI_SECTION_ASM_OP): Likewise.
1546 (TARGET_OS_CPP_BUILTINS): Define __GXX_MERGED_TYPEINFO_NAMES=0
1547 and __GXX_TYPEINFO_EQUALITY_INLINE=0 if -mid-shared-library.
1548 (DRIVER_SELF_SPECS): Map unadorned PIC options to -msep-data.
1549 * config/m68k/t-uclinux (EXTRA_MULTILIB_PARTS): Add crtbegin.o
1550 and crtend.o.
1551 * config/m68k/lb1sf68.asm (PICCALL): Use an lea and pc-relative
1552 jump sequence for ISA A and ISA A+.
1553 (PICJUMP): Likewise.
1555 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1557 * config.gcc (m68k-*-uclinux*): Base the port on the common
1558 and m68k GNU/Linux files rather than on the generic ELF ones.
1559 * config/m68k/uclinux.h (TARGET_VERSION): Override.
1560 (TARGET_OS_CPP_BUILTINS): Use LINUX_TARGET_OS_CPP_BUILTINS.
1562 2007-01-18 Julian Brown <julian@codesourcery.com>
1563 Richard Sandiford <richard@codesourcery.com>
1565 * config/m68k/m68k.h (LONG_DOUBLE_TYPE_SIZE): Make 64-bit on ColdFire.
1566 (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
1567 * config/m68k/netbsd-elf.h (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Undefine
1568 before redefining.
1569 * config/m68k/uclinux-oldabi.h (LONG_DOUBLE_TYPE_SIZE): Redefine to
1570 80 unconditionally.
1571 (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
1573 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1575 * doc/install.texi: Document m68k-uclinuxoldabi.
1576 * config.gcc (m68k-*-uclinuxoldabi): New configuration.
1577 * config/m68k/uclinux-oldabi.h: New file, copied from
1578 config/m68k/uclinux.h.
1580 2007-01-18 Kaz Kojima <kkojima@rr.iij4u.or.jp>
1582 * config/m32r/m32r-protos.h (m32r_expand_epilogue): Declare it.
1583 * config/m32r/m32r.c (m32r_setup_incoming_varargs): Use gen_frame_mem.
1584 (m32r_compute_frame_size): Use unsigned for regno.
1585 (m32r_reload_lr): Use gen_frame_mem.
1586 (pop): New.
1587 (m32r_output_function_epilogue): Don't output the function epilogue
1588 textually here.
1589 (m32r_expand_epilogue): New.
1590 (direct_return): Return false if the function has the interrupt
1591 attribute.
1592 (m32r_hard_regno_rename_ok): Remove code for the textual epilogue.
1593 * config/m32r/m32r.md (epilogue): New expander.
1594 (return_lr, return_rte): New insns.
1595 (return): Make it expander.
1596 (return_normal): New expander.
1598 2007-01-18 Josh Conner <jconner@apple.com>
1600 PR target/30485
1601 * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Add
1602 support for UNLE, UNLT, UNGE, and UNGT.
1604 2007-01-18 Jan Hubicka <jh@suse.cz>
1606 * tree-vrp.c (finalize_jump_threads): Do not call cleanup_cfg by hand.
1607 * tree-tailcall (add_virtual_phis): Likewise.
1608 (optimize_tail_call): Return TODOs.
1609 (execute_tail_calls): Return TODOs.
1610 * tree-ssa-ccp (execute_fold_all_builtins): Do cleanup_cfg via TODO.
1611 * tree-cfgcleanup.c (cleanup_tree_cfg_loop): Return if something
1612 changed.
1613 * tree-ssa-forwprop.c (tree_ssa_forward_propagate_single_use_value):
1614 Cleanup cfg using TODO.
1615 * tree-flow.h (cleanup_tree_cfg_loop): Update prototype.
1616 * passes.c (execute_function_todo): When cleanup did something, remove
1617 unused locals.
1618 * tree-cfg.c (pass_build_cfg): Add cleanup_cfg TODO.
1619 (make_edges): Don't cleanup_cfg.
1621 2007-01-18 Uros Bizjak <ubizjak@gmail.com>
1623 * reg-stack.c (subst_stack_regs_pat) [UNSPEC_SINCOS_COS,
1624 UNSPEC_XTRACT_FRACT]: Use generic code for instructions that
1625 operate on the top of stack.
1626 [UNSPEC_SINCOS_SIN, UNSPEC_XTRACT_EXP, UNSPEC_TAN]: Rewrite
1627 register handling of instructions that output to the second
1628 stack slot.
1629 [UNSPEC_TAN_ONE, UNSPEC_TAN_TAN]: Remove.
1630 (move_for_stack_reg): Special-case check for dead destination
1631 stack slot for constant load of 1.0 inside UNSPEC_TAN.
1633 * config/i386/i386.md (UNSPEC_TAN): New constant.
1634 (UNSPEC_TAN_ONE, UNSPEC_TAN_TAN): Remove.
1635 (fptanxf4_i387, fptan_extend<mode>xf4_i387): New patterns
1636 to correctly model move of constant 1.0 to top stack slot.
1637 (*tandf3_1, *tansf3_1, *tanxf3_1): Remove insn patterns.
1638 (unnamed peephole2 pattern): Remove corresponding peephole2
1639 pattern that optimizes tan insn and loading of constant 1.0.
1640 (tanxf2): Use fptanxf4_i387.
1641 (tan<mode>2): Rename from tansf2 and tandf2 and macroize insn
1642 patterns using X87MODEF12 mode macro. Use fptan_extend<mode>xf4_i387
1643 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
1644 disable patterns for SSE math.
1645 (sincos<mode>3): Use truncxf<mode>2_i387_noop for truncation.
1646 (fyl2x_extend<mode>xf3_i387): Use X87MODEF12 for operand 1.
1648 2007-01-18 Dirk Mueller <dmueller@suse.de>
1649 Richard Guenther <rguenther@suse.de>
1651 PR diagnostic/8268
1652 * doc/invoke.texi (Warray-bounds): Document -Warray-bounds.
1653 * common.opt (Warray-bounds): Add new warning option.
1654 * c-opts.c (c_common_handle_option): Define -Warray-bounds
1655 if -Wall is given.
1656 * Makefile.in: make tree-vrp.o depend on toplev.h
1657 * tree-vrp.c (vrp_finalize): Call check_array_refs if -Warray-bounds
1658 is enabled.
1659 (check_array_refs, check_array_bounds, check_array_ref): New.
1661 2007-01-18 Jan Hubicka <jh@suse.cz>
1663 * tree-ssa-ccp.c (ccp_finalize): Return if something changed.
1664 (execute_ssa_ccp): Return flags conditionally.
1665 * tree-ssa-propagate.c (substitue_and_fold): Return if something was
1666 changed.
1667 * tree-ssa-propagate.h (substitute_and_fold): Update prototype.
1669 2007-01-18 Steven Bosscher <steven@gcc.gnu.org>
1671 * cfgcleanup.c (cleanup_cfg): Detect cfglayout mode and set
1672 the CLEANUP_CFGLAYOUT flag when in cfglayout mode.
1674 * Makefile.c (GTFILES): Add cfglayout.h.
1675 * gengtype.c (open_base_files): Likewise.
1676 * cfglayout.c (cfg_layout_function_footer,
1677 cfg_layout_function_header) Reindent to make gengtype happy.
1678 * cfglayout.h (cfg_layout_function_footer,
1679 cfg_layout_function_header): Add GTY(()) marker.
1681 * ifcvt.c (noce_try_sign_mask): Make sure INSN_B is non-null.
1683 2007-01-18 Ben Elliston <bje@au.ibm.com>
1685 * genautomata.c (write_automata): Include xstrerror output in the
1686 error message if writing the DFA description file fails.
1688 2007-01-17 H.J. Lu <hongjiu.lu@intel.com>
1690 * config/mips/mips-protos.h (mips_output_external): Make it
1691 return void.
1692 * config/mips/iris.h (TARGET_ASM_EXTERNAL_LIBCALL): Removed.
1693 * config/mips/mips.c (irix_output_external_libcall): Likewise.
1694 (extern_list): Likewise.
1695 (extern_head): Likewise.
1696 (TARGET_ASM_FILE_END): Likewise.
1697 (mips_file_end): Likewise.
1698 (mips_output_external): Rewritten.
1700 2007-01-18 Ben Elliston <bje@au.ibm.com>
1702 * genpreds.c (write_insn_preds_c): Only write out the function
1703 body for regclass_for_constraint if we have register constraints.
1705 2007-01-17 Tom Tromey <tromey@redhat.com>
1707 * doc/sourcebuild.texi (libgcj Tests): Use sourceware.org.
1708 * doc/install.texi (Testing): Use sourceware.org.
1709 (Binaries): Likewise.
1710 (Specific): Likewise.
1711 * doc/contrib.texi (Contributors): Use sourceware.org.
1713 2007-01-17 Anatoly Sokolov <aesok@post.ru>
1715 * config/avr/avr.h (AVR_HAVE_LPMX): New macro.
1716 (AVR_ENHANCED): Rename to ...
1717 (AVR_HAVE_MUL): ... new.
1718 (avr_enhanced_p): Rename to ...
1719 (avr_have_mul_p): ... new.
1720 (TARGET_CPU_CPP_BUILTINS): Use 'avr_have_mul_p' instead of
1721 'avr_enhanced_p' for "__AVR_ENHANCED__". Define "__AVR_HAVE_MUL__".
1722 * config/avr/avr.c (avr_enhanced_p): Rename to ...
1723 (avr_have_mul_p): ... new.
1724 (base_arch_s): Rename 'enhanced' to 'have_mul'.
1725 (avr_override_options): Use 'avr_have_mul_p' and 'have_mul' instead of
1726 'avr_enhanced_p' and 'enhanced'.
1727 (ashlhi3_out, ashrhi3_out, lshrhi3_out, avr_rtx_costs): Use
1728 AVR_HAVE_MUL instead of AVR_ENHANCED.
1729 * avr.md (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
1730 (mulqi3, *mulqi3_enh, *mulqi3_call, mulqihi3, umulqihi3, mulhi3,
1731 *mulhi3_enh, *mulhi3_call, mulsi3, *mulsi3_call): Use AVR_HAVE_MUL
1732 instead of AVR_ENHANCED.
1733 (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
1734 * libgcc.S: Use __AVR_HAVE_MUL__ instead of __AVR_ENHANCED__.
1735 (__tablejump__): Use __AVR_HAVE_LPMX__ instead of __AVR_ENHANCED__.
1737 2007-01-17 Ian Lance Taylor <iant@google.com>
1739 * vec.h (VEC_reserve_exact): Define.
1740 (vec_gc_p_reserve_exact): Declare.
1741 (vec_gc_o_reserve_exact): Declare.
1742 (vec_heap_p_reserve_exact): Declare.
1743 (vec_heap_o_reserve_exact): Declare.
1744 (VEC_OP (T,A,reserve_exact)): New static inline function, three
1745 versions.
1746 (VEC_OP (T,A,reserve)) [all versions]: Remove handling of
1747 negative parameter.
1748 (VEC_OP (T,A,alloc)) [all versions]: Call ...reserve_exact.
1749 (VEC_OP (T,A,copy)) [all versions]: Likewise.
1750 (VEC_OP (T,a,safe_grow)) [all versions]: Likewise.
1751 * vec.c (calculate_allocation): Add exact parameter. Change all
1752 callers.
1753 (vec_gc_o_reserve_1): New static function, from vec_gc_o_reserve.
1754 (vec_gc_p_reserve, vec_gc_o_reserve): Call vec_gc_o_reserve_1.
1755 (vec_gc_p_reserve_exact, vec_gc_o_reserve_exact): New functions.
1756 (vec_heap_o_reserve_1): New static function, from vec_heap_o_reserve.
1757 (vec_heap_p_reserve, vec_heap_o_reserve): Call vec_heap_o_reserve_1.
1758 (vec_heap_p_reserve_exact): New function.
1759 (vec_heap_o_reserve_exact): New function.
1761 2007-01-17 Jan Hubicka <jh@suse.cz>
1763 * ipa-type-escape.c (look_for_casts): Revamp using handled_component_p.
1765 2007-01-17 Eric Christopher <echristo@apple.com>
1767 * config.gcc: Support core2 processor.
1769 2007-01-16 Jan Hubicka <jh@suse.cz>
1771 * tree-ssanames.c (release_dead_ssa_names): Instead of ggc_freeing
1772 the names, just unlink the chain so we don't crash on dangling pointers
1773 to dead SSA names.
1775 2007-01-16 Jan Hubicka <jh@suse.cz>
1777 * cgraph.h (cgraph_decide_inlining_incrementally): Kill.
1778 * tree-pass.h: Reorder to make IPA passes appear toegher.
1779 (pass_early_inline, pass_inline_parameters, pass_apply_inline): Declare.
1780 * cgraphunit.c (cgraph_finalize_function): Do not compute inling
1781 parameters, do not call early inliner.
1782 * ipa-inline.c: Update comments. Include tree-flow.h
1783 (cgraph_decide_inlining): Do not compute inlining parameters.
1784 (cgraph_decide_inlining_incrementally): Return TODOs; assume to
1785 be called with function context set up.
1786 (pass_ipa_inline): Remove unreachable functions before pass.
1787 (cgraph_early_inlining): Simplify assuming to be called from the
1788 PM as local pass.
1789 (pass_early_inline): New pass.
1790 (cgraph_gate_ipa_early_inlining): New gate.
1791 (pass_ipa_early_inline): Turn into simple wrapper.
1792 (compute_inline_parameters): New function.
1793 (gate_inline_passes): New gate.
1794 (pass_inline_parameters): New pass.
1795 (apply_inline): Move here from tree-optimize.c
1796 (pass_apply_inline): New pass.
1797 * ipa.c (cgraph_remove_unreachable_nodes): Verify cgraph after
1798 transforming.
1799 * tree-inline.c (optimize_inline_calls): Return TODOs rather than
1800 doing them by hand.
1801 (tree_function_versioning): Do not allocate dummy struct function.
1802 * tree-inline.h (optimize_inline_calls): Update prototype.
1803 * tree-optimize.c (execute_fixup_cfg): Export.
1804 (pass_fixup_cfg): Remove
1805 (tree_rest_of_compilation): Do not apply inlines.
1806 * tree-flow.h (execute_fixup_cfg): Declare.
1807 * Makefile.in (gt-passes.c): New.
1808 * passes.c: Include gt-passes.h
1809 (init_optimization_passes): New passes.
1810 (nnodes, order): New static vars.
1811 (do_per_function_toporder): New function.
1812 (execute_one_pass): Dump current pass here.
1813 (execute_ipa_pass_list): Don't dump current pass here.
1815 2007-01-16 Janis Johnson <janis187@us.ibm.com>
1817 * config/dfp-bit.c (dfp_compare_op): Return separate value for NaN.
1818 (DFP_NE, DFP_LE, DFP_GE): Return false for NaN.
1820 2007-01-16 David Edelsohn <edelsohn@gnu.org>
1822 * config/rs6000/darwin-ldouble.c: Build file for SOFT_FLOAT.
1823 (strong_alias): Define.
1824 (__gcc_qmul): Provide non-FMA for soft-float.
1825 (__gcc_qdiv): Same.
1826 (__gcc_qneg): New.
1827 (__gcc_qeq): New.
1828 (__gcc_qle): New.
1829 (__gcc_qge): New.
1830 (__gcc_qunord): New.
1831 (__gcc_stoq): New.
1832 (__gcc_dtoq): New.
1833 (__gcc_qtos): New.
1834 (__gcc_qtod): New.
1835 (__gcc_qtoi): New.
1836 (__gcc_qtou): New.
1837 (__gcc_itoq): New.
1838 (__gcc_utoq): New.
1839 (fmsub): New.
1840 * config/rs6000/rs6000.c (rs6000_init_libfuncs): Initialize
1841 soft-float functions.
1842 * config/rs6000/libgcc-ppc-glibc.ver: Version soft-float symbols.
1843 * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Do not warn
1844 about long double soft float.
1846 2007-01-16 Dorit Nuzman <dorit@il.ibm.com>
1847 Tehila Meyzels <tehila@il.ibm.com>
1849 * tree-vectorizer.h (is_pattern_stmt_p): New.
1850 * tree-vect-analyze.c (vect_determine_vectorization_factor): Fix
1851 formatting (tabs instead of spaces). Cleanup and clarify setting
1852 of STMT_VINFO_VECTYPE. Call is_pattern_stmt_p.
1853 * tree-vect-transform.c (vect_get_vec_def_for_operand): Fix typo.
1854 (vectorizable_type_demotion): Check that types are integral.
1855 (vectorizable_type_promotion): Likewise.
1856 (vectorizable_store): Fix typo. Eliminate new-line at end of
1857 comments.
1859 2007-01-16 Jan Hubicka <jh@suse.cz>
1861 * tree-ssanames.c (release_dead_ssa_names): Remove invalidated
1862 cgraph edges too.
1864 2007-01-15 Eric Christopher <echristo@apple.com>
1866 * ifcvt.c: Include vec.h, vecprim.h.
1867 (check_cond_move_block): New argument regs.
1868 Reorganize. Add registers used to regs.
1869 (cond_move_process_if_block): Use regs set above as
1870 loop bounds.
1872 2007-01-15 Eric Christopher <echristo@apple.com>
1874 * config/darwin.h: Update copyright.
1875 (TARGET_OPTION_TRANSLATE_TABLE): Add umbrella.
1876 (LINK_COMMAND_SPEC): Add -u.
1877 (LINK_SPEC): Fix umbrella for above.
1879 2007-01-15 Joseph S. Myers <joseph@codesourcery.com>
1881 * config/soft-fp/op-common.h, config/soft-fp/op-4.h: Update from
1882 glibc CVS.
1884 2007-01-15 Tom Tromey <tromey@redhat.com>
1886 * doc/sourcebuild.texi (libgcj Tests): Don't mention jacks.
1887 * doc/install.texi (Testing): Don't mention jacks.
1888 (Configuration): Document --enable-java-maintainer-mode. Move
1889 --with-java-home to libgcj-specific section. Document
1890 --with-ecj-jar.
1891 (Prerequisites): Mention --enable-java-maintainer-mode, ecj1.
1893 2007-01-15 Jan Hubicka <jh@suse.cz>
1895 * tree-ssa-dce.c (DCE_TODOs): New.
1896 (propagate_necessity): Return if something changed.
1897 (eliminate_unnecessary_stmts): Likewise.
1898 (perform_tree_ssa_dce): Return TODO flags when needed.
1899 (pass_dce, pass_dce_loop, pass_cd_dce): Remove TODO flags.
1901 2007-01-15 Uros Bizjak <ubizjak@gmail.com>
1903 * config/i386/i386.md (fyl2xxf3_i387): Rename from fyl2x_xf3.
1904 (fyl2x_extend<mode>xf3_i387): New insn pattern.
1905 (log<mode>2): Rename from logsf2 and logdf2 and macroize insn
1906 patterns using X87MODEF12 mode macro. Extend operand 1
1907 to XFmode. Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
1908 (log10<mode>2): Ditto.
1909 (log2<mode>2): Ditto.
1910 (log1p<mode>2): Ditto.
1911 (logb<mode>2): Ditto.
1912 (fyl2xp1xf3_i387): Rename from fyl2xp1_xf3.
1913 (fyl2xp1_extend<mode>xf3_i387): New insn pattern.
1914 (*fxtractxf3_i387): Rename from *fxtractxf3.
1915 (fxtract_extend<mode>xf3_i387): New insn pattern.
1916 (ilogbsi2): Use match_dup 3, not match_operand:XF 3.
1917 * config/i386/i386.c (ix86_emit_i387_log1p): Use gen_fyl2xp1xf3_i387()
1918 and gen_fyl2xxf3_i387().
1920 2007-01-14 Zdenek Dvorak <dvorakz@suse.cz>
1922 * loop-unswitch.c (unswitch_loop): Do not call fix_loop_placement.
1923 * cfgloopmanip.c (fix_loop_placement): Made static. Use
1924 get_loop_exit_edges. Changed return type to bool.
1925 * cfgloop.h (fix_loop_placement): Declaration removed.
1927 2007-01-14 Dorit Nuzman <dorit@il.ibm.com>
1929 * param.h (MIN_VECT_LOOP_BOUND): New.
1930 * params.def (MIN_VECT_LOOP_BOUND): New.
1931 * tree-vectorizer.c (slpeel_tree_peel_loop_to_edge): Takes another
1932 argument - minimum threshold for number of iterations.
1933 * tree-vectorizer.h (slpeel_tree_peel_loop_to_edge): Add another
1934 argument to declaration.
1935 * tree-vect-analyze.c (vect_analyze_operations): Check value of
1936 MIN_VECT_LOOP_BOUND.
1937 * tree-vect-transform.c (vect_do_peeling_for_loop_bound): Call
1938 slpeel_tree_peel_loop_to_edge with additional argument.
1939 (vect_do_peeling_for_alignment): Likewise.
1940 * doc/invoke.texi (min-vect-loop-bound): Document new param option.
1942 2007-01-14 Uros Bizjak <ubizjak@gmail.com>
1944 PR target/30413
1945 * config/i386/i386.c (print_operand) ['z']: Output 'b' for
1946 operands of size 1.
1948 2007-01-14 Jan Hubicka <jh@suse.cz>
1950 * tree-dfa.c (remove_referenced_var): New function.
1951 * tree-ssa-live.c (remove_unused_locals): Walk referenced vars and
1952 prune referenced vars list too.
1953 * tree-flow.h (remove_referenced_var): Declare.
1955 2007-01-14 Jan Hubicka <jh@suse.cz>
1957 * tree-eh.c (add_stmt_to_eh_region_fn): Do not add call_exprs
1958 separately.
1959 (remove_stmt_from_eh_region_fn): Do not remove call_exprs.
1960 (verify_eh_throw_stmt_node, verify_eh_throw_table_statements): Kill.
1961 * except.h (verify_eh_throw_table_statements): Kill prototype.
1962 * cfgexpand.c (expand_gimple_basic_block): Propagate Eh regions
1963 into call exrepssions.
1964 * tree-optimize.c (execute_free_cfg_annotatiosn): Do not call
1965 eh trhow verifier.
1966 * tree-cfg.c: Include pointer-set.h.
1967 (verify_node_sharing): Work on pointer set.
1968 (verify_eh_throw_stmt_node): New.
1969 (verify_stmts): Use pointers sets, verify throw_stmt.
1971 2007-01-13 Zdenek Dvorak <dvorakz@suse.cz>
1973 * ipa-reference.c (analyze_function): Consider also addresses taken
1974 in phi nodes.
1976 2007-01-12 Roger Sayle <roger@eyesopen.com>
1978 * c-typeck.c (null_pointer_constant_p): Replace use of
1979 TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
1980 (build_c_cast): Likewise.
1982 2007-01-12 Roger Sayle <roger@eyesopen.com>
1984 * tree.h (force_fit_type_double): Remove unused final argument.
1985 * c-common.c (constant_expression_warning): Replace use of
1986 TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
1987 (convert_and_check): Likewise.
1988 (shorten_compare): Update call to force_fit_type_double.
1989 (c_common_truthvalue_conversion) <INTEGER_CST>: Use integer_zerop.
1990 * convert.c (convert_to_pointer): Update call to
1991 force_fit_type_double.
1992 * fold-const.c (force_fit_type_double): Remove overflowed_const
1993 argument.
1994 (int_const_binop, fold_convert_const_int_from_int,
1995 fold_convert_const_int_from_real, fold_div_compare,
1996 fold_sign_changed_comparison, fold_unary, fold_negate_const,
1997 fold_abs_const, fold_not_const): Remove the final argument from
1998 calls to force_fit_type_double.
2000 2007-01-12 Andrew Pinski <andrew_pinski@playstation.sony.com>
2002 * configure.ac: Set insn to "nop" for spu-*-* also.
2003 * configure: Regenerate.
2005 2007-01-12 Olga Golovanevsky <olga@il.ibm.com>
2007 * builtins.def : Add BUILT_IN_FREE.
2009 2007-01-12 Jan Hubicka <jh@suse.cz>
2011 PR tree-optimization/30443
2012 * tree-inline.c (tree_function_versioning): Do not optimize when
2013 cloning for inlining.
2015 2007-01-12 Zdenek Dvorak <dvorakz@suse.cz>
2017 * doc/loop.texi: Document recording of loop exits.
2018 * cfgloopmanip.c (loopify, duplicate_loop): Use alloc_loop.
2019 (update_single_exits_after_duplication,
2020 update_single_exit_for_duplicated_loop,
2021 update_single_exit_for_duplicated_loops): Removed.
2022 (duplicate_loop_to_header_edge): Do not call
2023 update_single_exits_after_duplication and
2024 update_single_exit_for_duplicated_loops.
2025 (loop_version): Do not update single_exit information.
2026 (fix_loop_structure): Use record_loop_exits instead of
2027 mark_single_exit_loops.
2028 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update
2029 the lists of loop exits.
2030 * cfghooks.c (redirect_edge_and_branch, redirect_edge_and_branch_force,
2031 split_edge, merge_blocks): Update the lists of loop exits.
2032 * modulo-sched.c (sms_schedule): Pass LOOPS_HAVE_RECORDED_EXITS to
2033 loop_optimizer_init.
2034 * loop-init.c (loop_optimizer_init): Call record_loop_exits instead
2035 of mark_single_exit_loops.
2036 (loop_optimizer_finalize): Call release_recorded_exits.
2037 * tree-ssa-loop.c (tree_loop_optimizer_init): Pass
2038 LOOPS_HAVE_RECORDED_EXITS to loop_optimizer_init.
2039 * tree-vectorizer.c (slpeel_tree_duplicate_loop_to_edge_cfg): Do not
2040 update single exit information.
2041 * lambda-code.c (perfect_nestify): Ditto.
2042 * cfgloop.c (flow_loop_free): Destroy the list of exits of the loop.
2043 (mark_single_exit_loops): Removed.
2044 (alloc_loop, loop_exit_hash, loop_exit_eq, loop_exit_free,
2045 get_exit_descriptions, rescan_loop_exit, record_loop_exits,
2046 dump_recorded_exit, dump_recorded_exits, release_recorded_exits): New
2047 functions.
2048 (get_loop_exit_edges, single_exit): Use recorded exit lists.
2049 (add_bb_to_loop, remove_bb_from_loops): Update the lists of loop exits.
2050 (verify_loop_structure): Verify consistency of the exit lists.
2051 (flow_loops_find): Use alloc_loop. Initialize exits hash.
2052 (set_single_exit): Removed.
2053 * cfgloop.h (struct loop_exit): New function.
2054 (struct loop): single_exit_ field replaced by exits field.
2055 (LOOPS_HAVE_MARKED_SINGLE_EXITS): Replaced by LOOPS_HAVE_RECORDED_EXITS.
2056 (struct loops): Added exits hash.
2057 (mark_single_exit_loops, set_single_exit): Declaration removed.
2058 (release_recorded_exits, record_loop_exits, rescan_loop_exit): Declare.
2060 2007-01-12 Richard Sandiford <richard@codesourcery.com>
2062 * doc/invoke.texi: Avoid use of @headitem.
2064 2007-01-12 Richard Sandiford <richard@codesourcery.com>
2066 * cse.c (cse_insn): Move HAVE_CC0 code after declarations.
2068 2007-01-12 Richard Sandiford <richard@codesourcery.com>
2070 * doc/install.texi: Fix m68k-*-* anchor and add m68k-*-* to the
2071 list of targets.
2073 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
2074 Richard Sandiford <richard@codesourcery.com>
2076 * doc/invoke.texi: Document -mno-bitfield, -mno-rtd and -mno-short.
2077 * config/m68k/m68k.opt: Resort options.
2078 (mbitfield, mrtd, mshort): Remove RejectNegative properties.
2080 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
2081 Richard Sandiford <richard@codesourcery.com>
2083 * doc/invoke.texi: Document the macros that are defined by
2084 m68k's -mtune and -mhard-float options.
2085 * config/m68k/m68k-protos.h (m68k_cpp_cpu_ident) Declare.
2086 (m68k_cpp_cpu_family): Likewise.
2087 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Add a full set
2088 of __ucfv*__ macros. Define __mcffpu__ if generating code for
2089 ColdFire FPUs. Define __mcf_cpu_* and __mcf_family_* macros.
2090 * config/m68k/m68k.c (m68k_cpp_cpu_ident): New function.
2091 (m68k_cpp_cpu_family): Likewise.
2093 2007-01-12 Richard Sandiford <richard@codesourcery.com>
2095 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Treat all mc68*
2096 macros besides mc68000 as tuning macros. Use a switch statement
2097 to set them and mcpu32.
2099 2007-01-12 Julian Brown <julian@codesourcery.com>
2101 * config/m68k/m68k.h: Use TARGET_68040 instead of TARGET_68040_ONLY.
2102 (TARGET_68040_ONLY): Rename to...
2103 (TARGET_68040): ...this.
2104 * config/m68k/m68k.c: Use TARGET_68040 instead of TARGET_68040_ONLY.
2105 * config/m68k/m68k.md: Likewise.
2107 2007-01-12 Julian Brown <julian@codesourcery.com>
2108 Nathan Sidwell <nathan@codesourcery.com>
2109 Richard Sandiford <richard@codesourcery.com>
2111 * config.gcc (m680[012]0-*-*, m68k*-*-*): Set m68k_cpu_ident to
2112 the -mcpu= argument associated with the --with-cpu setting.
2113 Define M68K_DEFAULT_TUNE to the default -mtune= option,
2114 if different from the one implied by the -mcpu setting.
2115 Accept --with-cpu=FOO if FOO is listed in m68k-devices.def,
2116 using mcpu=FOO as the default CPU option. Set target_cpu_default2.
2117 * doc/invoke.texi: Mention ColdFire in the introduction to the
2118 m68k options. Document the new -march, -mcpu, -mtune, -mdiv,
2119 -mno-div and -mhard-float options. Make -m68881 a synonym for
2120 -mhard-float. Document the previously-undocumented -m5206e,
2121 -m528x, -m5307 and -m5407 options. Tweak the existing option
2122 documentation for consistency.
2123 * doc/install.texi: Mention new --with-cpu arguments.
2124 * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Only use the
2125 default CPU if neither -mcpu nor -march are specified.
2126 (ASM_CPU_SPEC): Pass down -mcpu and -march options.
2127 (TARGET_CPU_CPP_BUILTINS): Set __mcfisa*__ macros from
2128 TARGET_ISA*. Set the legacy __mcf*__ cpu macros in the same way,
2129 using m68k_tune to decide between families that implement the
2130 same ISA. Use m68k_tune to set __mcfv4e__.
2131 (FL_BITFIELD, FL_68881, FL_COLDFIRE, FL_CF_HWDIV, FL_CF_MAC)
2132 (FL_CF_EMAC, FL_CF_EMAC_B, FL_CF_USP, FL_CF_FPU, FL_ISA_68000)
2133 (FL_ISA_68010, FL_ISA_68020, FL_ISA_68040, FL_ISA_A, FL_ISA_B)
2134 (FL_ISA_C, FL_ISA_MMU): New macros.
2135 (MASK_COLDFIRE): Delete.
2136 (TARGET_68010, TARGET_68020, TARGET_68040_ONLY, TARGET_COLDFIRE)
2137 (TARGET_ISAB): Redefine in terms of m68k_cpu_flags.
2138 (TARGET_68881, TARGET_COLDFIRE_FPU): Redefine in terms of m68k_fpu.
2139 (TARGET_HARD_FLOAT): Do not define here.
2140 (TARGET_ISAAPLUS, TARGET_ISAC): New macros.
2141 (TUNE_68000): New macro.
2142 (TUNE_68000_10): Redefine in terms of TUNE_68000 and TUNE_68010.
2143 (TUNE_68010, TUNE_68030, TUNE_68040, TUNE_68060, TUNE_CPU32)
2144 (TUNE_CFV2): Redefine in terms of m68k_tune.
2145 (uarch_type, target_device, fpu_type): New enums.
2146 (m68k_cpu, m68k_tune, m68k_fpu, m68k_cpu_flags): Declare.
2147 * config/m68k/m68k.c (TARGET_DEFAULT): Remove MASK_68881.
2148 (FL_FOR_isa_00, FL_FOR_isa_10, FL_FOR_isa_20, FL_FOR_isa_40)
2149 (FL_FOR_isa_cpu32, FL_FOR_isa_a, FL_FOR_isa_aplus, FL_FOR_isa_b)
2150 (FL_FOR_isa_c): New macros.
2151 (m68k_isa): New enum.
2152 (m68k_target_selection): New structure.
2153 (all_devices, all_isas, all_microarchs): New tables.
2154 (m68k_cpu_entry, m68k_arch_entry, m68k_tune_entry, m68k_cpu)
2155 (m68k_tune, m68k_fpu, m68k_cpu_flags): New variables.
2156 (MASK_ALL_CPU_BITS): Delete.
2157 (m68k_find_selection): New function.
2158 (m68k_handle_option): Handle -mcpu=, -march= and -mtune=.
2159 Map the legacy target options to a combination of the new ones.
2160 (override_options): Set m68k_cpu, m68k_tune, m68k_fpu and
2161 m68k_cpu_flags. Handle M68K_DEFAULT_TUNE. Use m68k_cpu_flags
2162 to derive default MASK_BITFIELD, MASK_CF_HWDIV and MASK_HARD_FLOAT
2163 settings.
2164 * config/m68k/m68k.opt (m5200, m5206e, m528x, m5307, m5407, mcfv4e)
2165 (m68010, m68020, m68020-40, m68020-60, m68030, m68040): Remove Mask
2166 properties.
2167 (m68881, msoft-float): Change mask from 68881 to HARD_FLOAT.
2168 (march=, mcpu=, mdiv, mhard-float, mtune=): New options.
2169 * config/m68k/m68k-devices.def: New file.
2171 2007-01-12 Richard Sandiford <richard@codesourcery.com>
2172 Nathan Sidwell <nathan@codesourcery.com>
2174 * config/m68k/m68k.h (ASM_CPU_SPEC, ASM_SPEC, EXTRA_SPECS)
2175 (SUBTARGET_EXTRA_SPECS): New macros.
2176 * config/m68k/linux.h (ASM_SPEC): Remove CPU flags;
2177 use %(asm_cpu_spec) instead.
2178 * config/m68k/m68k-none.h (ASM_SPEC): Likewise.
2179 * config/m68k/openbsd.h (ASM_SPEC): Likewise.
2180 * config/m68k/netbsd-elf.h (ASM_SPEC): Likewise.
2181 (EXTRA_SPECS): Rename to...
2182 (SUBTARGET_EXTRA_SPECS): ...this.
2184 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
2185 Richard Sandiford <richard@codesourcery.com>
2186 Julian Brown <julian@codesourcery.com>
2188 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
2189 (m68k-*-uclinux*, m68k-*-linux*, m68k-*-rtems*): Set default_m68k_cpu
2190 to the configuration's default CPU.
2191 (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*): Likewise.
2192 Remove default masks.
2193 (m680[012]0-*-*): Set the default with_cpu to the first part of
2194 the target name.
2195 (m68k*-*-*): Set the default with_cpu to m$default_m68k_cpu.
2196 (m68k*-*-linux): Extend the --with-cpu handling to...
2197 (m680[012]0-*-*, m68k*-*-*): ...these configurations. Allow m68000
2198 and m68010. Don't set target_cpu_default2.
2199 * doc/install.texi: Document --with-cpu for m68k.
2200 * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Define.
2201 * config/m68k/m68k-none.h (TARGET_CPU_DEFAULT, M68K_CPU_m68k)
2202 (M68K_CPU_m68000, M68K_CPU_m68010, M68K_CPU_m68020, M68K_CPU_m68030)
2203 (M68K_CPU_m68040, M68K_CPU_m68302, M68K_CPU_m68332, TARGET_DEFAULT)
2204 (ASM_CPU_DEFAULT_SPEC, CC1_CPU_DEFAULT_SPEC): Delete.
2205 (ASM_SPEC): Remove use of %(asm_cpu_default).
2206 (EXTRA_SPECS, SUBTARGET_EXTRA_SPECS, MULTILIB_DEFAULTS): Delete.
2207 * config/m68k/linux.h (TARGET_DEFAULT): Delete.
2208 (CPP_SPEC): Merge definitions. Do not handle __HAVE_68881__ here.
2209 * config/m68k/netbsd-elf.h (TARGET_OS_CPP_BUILTINS): Define
2210 __HAVE_FPU__ if TARGET_HARD_FLOAT.
2211 (TARGET_DEFAULT): Delete.
2212 (EXTRA_SPECS): Delete cpp_cpu_default_spec, cpp_cpu_spec,
2213 cpp_fpu_spec, asm_default_spec and netbsd_cpp_spec.
2214 (CPP_CPU_SPEC): Delete.
2215 (TARGET_VERSION): Merge definitions, using TARGET_68010 to pick
2216 the appropriate string.
2217 (CPP_CPU_DEFAULT_SPEC, ASM_DEFAULT_SPEC, CPP_FPU_SPEC): Delete.
2218 (CPP_SPEC): Define to NETBSD_CPP_SPEC.
2219 (ASM_SPEC): Don't use %(asm_default_spec).
2220 * config/m68k/m68k.c (TARGET_DEFAULT_TARGET_FLAGS): Remove
2221 TARGET_DEFAULT and add MASK_68881.
2222 * config/m68k/m68k.md: Remove mention of TARGET_DEFAULT from comments.
2224 2007-01-12 Richard Sandiford <richard@codesourcery.com>
2226 * config.gcc (m68010-*-netbsdelf*): Add MASK_68010.
2227 (m68k*-*-netbsdelf*, m68k*-*-openbsd*, m68k*-linux*): Add
2228 MASK_68010 alongside MASK_68020.
2229 * doc/invoke.texi: Document -m68010.
2230 * config/m68k/m68k.opt (m68010): New.
2231 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Define mc68010
2232 if TUNE_68010.
2233 (TUNE_68010): New macro.
2234 * config/m68k/m68k-none.h (M68K_CPU_m68k, M68K_CPU_m68010)
2235 (M68K_CPU_m68020, M68K_CPU_m68030, M68K_CPU_m68040)
2236 (M68K_CPU_m68332): Add MASK_68010.
2237 * config/m68k/linux.h (TARGET_DEFAULT): Add MASK_68010 to
2238 fallback definition.
2239 * config/m68k/netbsd-elf.h (CPP_CPU_SPEC): Remove now-redundant
2240 defines.
2241 * config/m68k/m68k.c (MASK_ALL_CPU_BITS): Add MASK_68010.
2242 (m68k_handle_option): Handle OPT_m68010. Add MASK_68010
2243 to all entries that use MASK_68020.
2244 (output_move_simode_const, output_move_himode, output_move_qimode)
2245 (output_move_stricthi, output_move_strictqi): Use TARGET_68010
2246 instead of TARGET_68020 to select clr behavior. Remove comment
2247 about there being no TARGET_68010.
2248 * config/m68k/m68k.md: Likewise throughout.
2250 2007-01-12 Julian Brown <julian@codesourcery.com>
2252 * config/m68k/m68k.h (TARGET_ISAB): New macro.
2253 * config/m68k/m68k.c: Use TARGET_ISAB rather than TARGET_CFV4.
2254 * config/m68k/m68k.md: Likewise.
2256 2007-01-12 Julian Brown <julian@codesourcery.com>
2258 * config/m68k/m68k.h (LEGITIMATE_INDEX_P, LEGITIMIZE_ADDRESS): Use
2259 TARGET_COLDFIRE_FPU instead of TARGET_CFV4E.
2261 2007-01-12 Julian Brown <julian@codesourcery.com>
2263 * config/m68k/m68k.h (TUNE_68040_60): New macro.
2264 * config/m68k/m68k.c (standard_68881_constant_p): Use it.
2265 * config/m68k/m68k.md: Likewise.
2267 2007-01-12 Julian Brown <julian@codesourcery.com>
2268 Richard Sandiford <richard@codesourcery.com>
2270 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Use TUNE_68030
2271 instead of TARGET_68030, TUNE_68040 instead of TARGET_68040,
2272 TUNE_68060 instead of TARGET_68060 and TUNE_CPU32 instead of
2273 TARGET_CPU32.
2274 (TARGET_CPU32): Rename to...
2275 (TUNE_CPU32): ...this.
2276 (TUNE_68000_10, TUNE_68030, TUNE_68040, TUNE_68060)
2277 (TUNE_CFV2): New macros.
2278 * config/m68k/netbsd-elf.h (LONG_DOUBLE_TYPE_SIZE): Simplify;
2279 remove conditions that are implied by TARGET_68020.
2280 * config/m68k/m68k.c (m68k_output_function_prologue): Use TUNE_68040
2281 instead of TARGET_68040 and TUNE_CPU32 instead of TARGET_CPU32.
2282 (m68k_output_function_epilogue): Likewise.
2283 (m68k_rtx_costs): Likewise. Use TUNE_68060 instead of TARGET_68060
2284 and TUNE_CFV2 instead of TARGET_5200. Use TUNE_68000_10 instead of
2285 "!TARGET_68020 && !TARGET_COLDFIRE" to choose between 68000 and
2286 non-68000 timings. Refactor multiplication and division costs.
2287 (output_addsi3): Use TUNE_68040 instead of TARGET_68040 and
2288 TUNE_CPU32 instead of TARGET_CPU32.
2289 (standard_68881_constant_p): Use TUNE_68040 instead of TARGET_68040
2290 and TUNE_68060 instead of TARGET_68060.
2291 * config/m68k/m68k.md: Use TUNE_68040 instead of TARGET_68040,
2292 TUNE_68060 instead of TARGET_68060, and TUNE_CPU32 instead of
2293 TARGET_CPU32.
2294 (movsi_const0): Use TUNE_68000_10 rather than "!TARGET_68020
2295 && !TARGET_COLDFIRE" to choose between moveq and clr.
2296 Likewise in the unnamed movsf pattern.
2297 (ashlsi_17_24, lshrsi_17_24): Guard with TUNE_68000_10 rather than
2298 "!TARGET_68020 && !TARGET_COLDFIRE". Likewise the unnamed
2299 ashiftrt pattern.
2301 2007-01-12 Richard Sandiford <richard@codesourcery.com>
2303 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Increase amount
2304 of tabbing before backslashes.
2306 2007-01-11 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
2308 * pa-linux.h (ASM_OUTPUT_INTERNAL_LABEL): Undefine.
2309 * pa.h (ASM_OUTPUT_LABEL): Output colon when using GAS.
2310 (ASM_OUTPUT_INTERNAL_LABEL): Define.
2312 2007-01-11 Zdenek Dvorak <dvorakz@suse.cz>
2314 * tree-ssa-loop-ivopts.c (extract_cond_operands): Split from
2315 find_interesting_uses_cond.
2316 (find_interesting_uses_cond): Use extract_cond_operands.
2317 (rewrite_use_compare): Use extract_cond_operands and
2318 force_gimple_operand_bsi. Do not call update_stmt.
2319 (determine_use_iv_cost_condition): Use extract_cond_operands.
2320 Return cheaper of using original bound and changing the exit bound.
2322 2007-01-11 Zdenek Dvorak <dvorakz@suse.cz>
2324 PR tree-optimization/29516
2325 * tree-ssa-address.c (tree_mem_ref_addr, add_to_parts,
2326 most_expensive_mult_to_index, addr_to_parts,
2327 create_mem_ref, maybe_fold_tmr): Make the type of
2328 fields of TARGET_MEM_REF sizetype.
2329 (move_fixed_address_to_symbol, move_pointer_to_base):
2330 New functions.
2331 * tree.def (TARGET_MEM_REF): Add comment on types of
2332 the operands.
2334 2007-01-11 Joseph Myers <joseph@codesourcery.com>
2336 * c-common.c (vector_types_convertible_p): Treat opaque types as
2337 always convertible if they have the same size, but not otherwise.
2339 2007-01-11 Steven Bosscher <steven@gcc.gnu.org>
2341 * ifcvt.c (struct noce_if_info): Add comments to the fields.
2342 Remove the b_unconditional field.
2343 (noce_try_sign_mask): Do not look at b_unconditional.
2344 (noce_process_if_block): Do not use merge_if_blocks. Update
2345 the CFG here. Do not set b_unconditional.
2346 (cond_move_process_if_block): Likewise.
2347 (find_cond_trap): Likewise.
2348 (check_cond_move_block): Require simple jump insns at the end
2349 of the basic block.
2351 2007-01-11 Jan Hubicka <jh@suse.cz>
2353 PR tree-optimization/1046
2354 * tree-tailcall.c (suitable_for_tail_call_opt_p): Use TREE_ADDRESSABLE
2355 when alias info is not ready.
2356 (pass_tail_recursion): Do not require aliasing.
2357 * tree-ssa-copyrename.c (pass_rename_ssa_cop): Likewise.
2358 * tree-ssa-ccp.c (pass_ccp, pass_fold_builtins): Likewise.
2359 * tree-ssa-copy.c (pass_copy_prop): Likewise.
2360 * tree-ssa-forwprop.c (pass_forwprop): Likewise.
2361 * tree-ssa-dce.c (pass_dce, pass_dce_loop, pass_cd_dce): Likewise.
2362 * passes.c (init_optimization_passes): Execute rename_ssa_copies,
2363 ccp, forwprop, copy_prop, merge_phi, copy_prop, dce and tail recursion
2364 before inlining.
2365 * tree-ssa-operands.c (add_virtual_operand, get_indirect_ref_operand):
2366 When aliasing is not build, mark statement as volatile.
2368 2007-01-11 Tom Tromey <tromey@redhat.com>
2370 PR preprocessor/15185, PR preprocessor/20989:
2371 * doc/cppopts.texi <-MT>: Update description of algorithm for
2372 computing default target.
2373 <-M, -MD>: Reword "basename" text.
2375 2007-01-11 Roger Sayle <roger@eyesopen.com>
2377 * builtins.c (expand_builtin_pow, expand_builtin_powi,
2378 fold_builtin_cabs, fold_builtin_sqrt, fold_builtin_trunc,
2379 fold_builtin_floor, fold_builtin_ceil, fold_builtin_round,
2380 fold_builtin_int_int_roundingfn, fold_builtin_bitop,
2381 fold_builtin_bswap, real_constp, fold_builtin_pow,
2382 fold_builtin_powi, fold_builtin_signbit, fold_builtin_copysign,
2383 do_mpfr_arg1, do_mpfr_arg2, do_mpfr_arg3, do_mpfr_sincos): Replace
2384 uses of the macro TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
2385 * convert.c (convert_to_pointer): Likewise.
2386 * expr.c (highest_pow2_factor, expand_expr_real_1): Likewise.
2387 * fold-const.c (force_fit_type, fold_negate_expr, int_const_binop,
2388 const_binop, fold_convert_const_int_from_int,
2389 fold_convert_const_int_from_real,
2390 fold_convert_const_real_from_real, sign_bit_p,
2391 optimize_minmax_comparison, extract_muldiv_1, fold_div_compare,
2392 fold_sign_changed_comparison, fold_unary, fold_comparison,
2393 fold_binary, multiple_of_p, tree_Expr_non_zero_p,
2394 fold_negate_const, fold_abs_const, fold_not_const): Likewise.
2395 * print-tree.c (print_node_brief, print_node): Likewise.
2396 * stor-layout.c (place_field, layout_type): Likewise.
2397 * tree-chrec.c (keep_cast): Likewise.
2398 * tree.c (build_vector, build_real, build_real_from_int_cst,
2399 build_complex): Likewise.
2401 2007-01-11 Roger Sayle <roger@eyesopen.com>
2403 * tree.h (TREE_CONSTANT_OVERFLOW): Obsolete. For the time being,
2404 treat TREE_CONSTANT_OVERFLOW as a synonym of TREE_OVERFLOW.
2406 2007-01-11 Paolo Bonzini <bonzini@gnu.org>
2408 * configure.ac (strict1_warn): Rename to strict_warn.
2409 (WERROR, --enable-werror, symlink hacks, stage1_cflags,
2410 cc_set_by_configure, quoted_cc_set_by_configure,
2411 stage_prefix_set_by_configure, quoted_stage_prefix_set_by_configure,
2412 all_boot_languages, all_stagestuff): Remove.
2413 (target_list): Remove bootstrap targets.
2414 * Makefile.in (quickstrap): Unconditionally make a synonym of all.
2415 (BOOT_LANGUAGES, STAGE1_CFLAGS, STAGE1_CHECKING,
2416 REMAKEFLAGS, FLAGS_TO_PASS, PREPEND_DOTDOT_TO_RELATIVE_PATHS,
2417 SUBDIR_FLAGS_TO_PASS, WERROR_FLAGS, STRICT2_WARN, LANG_STAGESTUFF,
2418 VOL_FILES, POSTSTAGE1_FLAGS_TO_PASS, STAGE2_FLAGS_TO_PASS,
2419 STAGEPROFILE_FLAGS_TO_PASS, STAGEFEEDBACK_FLAGS_TO_PASS, stage1_build,
2420 stage1_copy, stage2_build, stage2_copy, stageprofile_build,
2421 stageprofile_copy, stage3_build, stage3_copy, stagefeedback_build,
2422 stagefeedback_copy, stage4_build, clean_s1, clean_sw, bootstrap,
2423 bootstrap-lean, bootstrap2, bootstrap2-lean, bootstrap3,
2424 bootstrap3-lean, bootstrap4, bootstrap4-lean, unstage1, unstage2,
2425 unstage3, unstage4, unstageprofile, unstagefeedback, restage, restage2,
2426 restage3, restage4, restageprofile, restagefeedback, bubbleestrap,
2427 cleanstrap, unstrap, restrap, *compare, *compare3, *compare4,
2428 *compare-lean, *compare3-lean, *compare4-lean, stage1-start, stage1,
2429 stage2-start, stage2, stage3-start, stage3, stage4-start, stage4,
2430 stageprofile-start, stageprofile, stagefeedback-start, stagefeedback,
2431 risky-stage1, risky-stage2, risky-stage3, risky-stage4): Remove.
2432 (ORDINARY_FLAGS_TO_PASS): Rename to FLAGS_TO_PASS.
2433 (STAGECOPYSTUFF, STAGEMOVESTUFF): Consolidate into MOSTLYCLEANFILES.
2434 (mostlyclean): Adjust.
2435 (clean, distclean): Don't mention bootstrap stuff.
2436 * configure: Regenerate.
2437 * ada/config-lang.in, cp/config-lang.in, forttran/config-lang.in,
2438 java/config-lang.in, objc/config-lang.in, objcp/config-lang.in,
2439 treelang/config-lang.in (stagestuff): Remove.
2440 * doc/sourcebuild.texi (stage1, stage2, stage3, stage4,
2441 stageprofile, stagefeedback, stagestuff): Remove mention.
2443 2007-01-11 Nick Clifton <nickc@redhat.com>
2445 * config/mcore/predicates.md (mcore_general_movesrc_operand):
2446 Accept CONSTs.
2447 (mcore_general_movdst_operand): Do not accept CONST_INTs.
2448 (mcore_arith_K_S_operand): Run the test for the S constraint not
2449 the test for the M constraint.
2450 (mcore_addsub_operand): Do not accept integer values that are
2451 larger than 32 bits.
2452 * config/mcore/mcore.md: Remove unused constraints from split.
2453 (andsi3): Use HOST_WIDE_INT instead of int to hold an INTVAL.
2454 (addsi3): Likewise.
2455 (allocate_stack): Likewise.
2456 * config/mcore/mcore.c (mcore_print_operand): Restrict output of P
2457 operands to 32 bits.
2458 (mcore_const_costs): Use HOST_WIDE_INT instead of int to hold an
2459 INTVAL.
2460 (mcore_and_cost, mcore_modify_comparison, const_ok_for_mcore,
2461 mcore_const_ok_for_inline, mcore_const_trick_uses_not,
2462 try_constant_tricks, mcore_num_ones, mcore_num_zeros,
2463 mcore_output_bclri, mcore_output_andn, output_inline_const,
2464 mcore_output_move, mcore_output_movedouble): Likewise.
2465 (mcore_output_cmov): Use CONST_OK_FOR_M and CONST_OK_FOR_N.
2466 (output_inline_const): Likewise.
2467 (output_inline_const): Fix format strings used in sprintf
2468 statements.
2469 * config/mcore/mcore-protos.h: Update prototypes for changed
2470 functions in mcore.c.
2471 * config/mcore/mcore.h (CONST_OK_FOR_I): Cast values to
2472 HOST_WIDE_INT and not int.
2473 (CONST_OK_FOR_J, CONST_OK_FOR_K, CONST_OK_FOR_L, CONST_OK_FOR_M,
2474 CONST_OK_FOR_N): Likewise.
2475 (LEGITIMATE_CONSTANT_P): Also check CONSTANT_P.
2476 (GO_IF_LEGITIMATE_INDEX): Use HOST_WIDE_INT instead of int to hold
2477 an INTVAL.
2479 2007-01-10 Jan Hubicka <jh@suse.cz>
2481 * tree-vrp.c (remove_range_assertions): Release defs.
2482 * tree-ssa-loop-ivopts.c (rmeove_statement): Likewise.
2483 * tree-ssa-dom.c (remove_stmt_or_phi): Likewise.
2485 2007-01-10 Paul Brook <paul@codesourcery.com>
2487 * config/arm/arm.c (arm_rtx_costs_1): Handle mutiply-subtract.
2488 * config/arm/arm.md (mulsi3subsi): New insn.
2490 2007-01-10 Zdenek Dvorak <dvorakz@suse.cz>
2492 * tree-ssa-loop-manip.c (tree_unroll_loop): Make it a wrapper over ...
2493 (tree_transform_and_unroll_loop): New.
2494 * tree-flow.h (transform_callback, tree_transform_and_unroll_loop):
2495 Declare.
2497 2007-01-10 Robert Kennedy <jimbob@google.com>
2499 * fold-const.c (fold_comparison): Fold comparisons like (x *
2500 1000 < 0) to (x < 0).
2502 2007-01-10 Ian Lance Taylor <iant@google.com>
2504 * tree-pretty-print.c (dump_generic_node): Print parentheses when
2505 operands have the same priority.
2507 2007-01-10 Tom Tromey <tromey@redhat.com>
2509 * fold-const.c (fold_truthop): Don't check can_use_bit_fields_p.
2510 (fold_binary): Likewise.
2511 * langhooks.c (lhd_can_use_bit_fields_p): Removed.
2512 * langhooks-def.h (lhd_can_use_bit_fields_p): Removed.
2513 (LANG_HOOKS_CAN_USE_BIT_FIELDS_P): Removed.
2514 (LANG_HOOKS_INITIALIZER): Remove LANG_HOOKS_CAN_USE_BIT_FIELDS_P.
2515 * langhooks.h (struct lang_hooks): Removed field
2516 'can_use_bit_fields_p'.
2518 2007-01-10 Ralf Corsépius <ralf.corsepius@rtems.org>
2520 * config/bfin/t-bfin, config/bfin/t-bfin-elf: Remove GCC_CFLAGS.
2522 2007-01-10 Razya Ladelsky <razya@il.ibm.com>
2524 * function.c (get_last_funcdef_no): New function.
2525 * function.h (get_last_funcdef_no): Declare.
2526 * tree-inline.c (initialize_cfun): Add initialization.
2527 (tree_function_versioning): Cleanup.
2529 2007-01-10 Jan Hubicka <jh@suse.cz>
2531 * tree-inline.c (setup_one_parameter): Do not propagate into abnormal
2532 PHIs.
2534 2007-01-10 Sa Liu <saliu@de.ibm.com>
2535 Ben Elliston <bje@au.ibm.com>
2537 * spu.h (STACK_SAVE_AREA): Use VOIDmode for SAVE_FUNCTION, SImode
2538 for SAVE_NONLOCAL and Pmode for any other save level.
2539 * spu-protos.h (spu_restore_stack_block): Declare.
2540 * spu.md (save_stack_block): Remove.
2541 (restore_stack_block): Call spu_restore_stack_block.
2542 * spu.c (spu_restore_stack_block): New function.
2543 (spu_expand_epilogue): Remove old comment.
2545 2007-01-09 Zdenek Dvorak <dvorakz@suse.cz>
2547 PR tree-optimization/30322
2548 * tree-ssa-loop-ivopts.c (fold_affine_expr, iv_value): Removed.
2549 (cand_value_at): Return the value as aff_tree.
2550 (may_eliminate_iv): Convert the bound from aff_tree to tree.
2551 * tree-affine.c (aff_combination_add_cst, aff_combination_add_product,
2552 aff_combination_mult): New functions.
2553 (aff_combination_add): Use aff_combination_add_cst.
2554 (aff_combination_convert): Allow conversions to a wider type.
2555 (tree_to_aff_combination): Handle BIT_NOT_EXPR.
2556 * tree-affine.h (aff_combination_mult): Declare.
2558 2007-01-09 Carlos O'Donell <carlos@codesourcery.com>
2560 * doc/tm.texi: Update documentation to reflect reality of exec
2561 and start file search behaviours. Update copyright year.
2562 * doc/invoke.texi: Explain how GCC_EXEC_PREFIX is used to find
2563 header file directories.
2565 2007-01-09 Uros Bizjak <ubizjak@gmail.com>
2567 * config/i386/i386.md (*sinxf2): Rename to *sinxf2_i387.
2568 (*cosxf2): Rename to cosxf2_i387.
2569 (*sindf2, *sinsf2): Extend operand 1 to XFmode. Macroize patterns
2570 using X87MODEF12 mode macro. Rename patterns to
2571 *sin_extend<mode>xf2_i387. Use SSE_FLOAT_MODE_P to disable patterns
2572 for SSE math.
2573 (*cosdf2, *cossf2): Ditto.
2574 (sincosdf3, sincossf3): Ditto. Rewrite corresponding splitters
2575 to match extended input operands.
2576 (sincos<mode>3): New expander.
2577 (*sinextendsfdf2, *cosextendsfdf2, *sincosextendsfdf3): Remove
2578 insn patterns and corresponding splitters.
2580 2007-01-09 Kaz Kojima <kkojima@gcc.gnu.org>
2582 * config/sh/t-linux (TARGET_LIBGCC2_CFLAGS): Delete.
2583 (SHLIB_MAPFILES, SHLIB_LINK, SHLIB_INSTALL): Likewise.
2585 2007-01-09 Nicolas Pitre <nico@cam.org>
2587 PR target/30173
2588 * arm/ieee754-df.S (Lad_s): Also test the low word of X for zero.
2590 2007-01-08 Geoffrey Keating <geoffk@apple.com>
2592 * target.h (struct gcc_target): New field library_rtti_comdat.
2593 * target-def.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): New.
2594 (TARGET_CXX): Add TARGET_CXX_LIBRARY_RTTI_COMDAT.
2595 * doc/tm.texi (C++ ABI): Document TARGET_CXX_LIBRARY_RTTI_COMDAT.
2596 * config/darwin.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
2598 2007-01-08 Geoffrey Keating <geoffk@apple.com>
2600 * doc/invoke.texi (Optimize Options): Correct description of -O0.
2602 2007-01-08 Richard Guenther <rguenther@suse.de>
2604 * tree.h (force_fit_type_double): Export.
2605 (force_fit_type): Remove.
2606 * fold-const.c (force_fit_type_double): New function.
2607 (force_fit_type): Remove.
2608 (int_const_binop): Use it.
2609 (fold_convert_const_int_from_int): Likewise.
2610 (fold_convert_const_int_from_real): Likewise.
2611 (fold_div_compare): Likewise.
2612 (fold_sign_changed_comparison): Likewise.
2613 (fold_unary): Likewise.
2614 (fold_negate_const): Likewise.
2615 (fold_abs_const): Likewise.
2616 (fold_not_const): Likewise.
2617 * c-common.c (shorten_compare): Use force_fit_type_double.
2618 * convert.c (convert_to_pointer): Likewise.
2620 2007-01-08 Richard Guenther <rguenther@suse.de>
2622 * tree.h (build_int_cst_wide_type): Export.
2623 * tree.c (build_int_cst_wide_type): New function.
2624 (build_int_cst_wide): Fix comment.
2625 * builtins.c (fold_builtin_object_size): Use build_int_cst
2626 to build -1 or 0 of the correct type. Use fit_double_type
2627 to check for overflow.
2628 * fold-const.c (optimize_bit_field_compare): Use build_int_cst_type
2629 to build the mask.
2630 (decode_field_reference): Likewise.
2631 (all_ones_mask_p): Likewise.
2632 (native_interpret_int): Use build_int_cst_wide_type.
2633 (fold_binary): Use build_int_cst_type to build an all-ones
2634 value.
2635 * stor-layout.c (set_sizetype): Use build_int_cst_wide_type.
2637 2007-01-08 Daniel Jacobowitz <dan@codesourcery.com>
2639 * config/pa/t-pa64 (libgcc_stub.a): Use $(T).
2641 2007-01-09 Ben Elliston <bje@au.ibm.com>
2643 * genautomata.c (STATS_OPTION): New option.
2644 (stats_flag): New flag.
2645 (gen_automata_option): Handle it.
2646 (initiate_automaton_gen): Ditto.
2647 (write_automata): Output statistics only if stats_flag is
2648 set. Likewise, output time statistics only if time_flag is set.
2649 * doc/md.texi (Processor pipeline description): Document new flag.
2651 2007-01-08 Richard Guenther <rguenther@suse.de>
2653 * builtins.c (fold_builtin_int_roundingfn): Use fit_double_type.
2654 * tree.c (build_int_cst_type): Likewise.
2655 (size_in_bytes): Don't call force_fit_type on the result.
2656 (int_fits_type_p): Use fit_double_type.
2657 * fold-const.c (fit_double_type): New function.
2658 (force_fit_type): Use it.
2659 * tree.h (fit_double_type): Export.
2661 2007-01-08 Jan Hubicka <jh@suse.cz>
2663 * tree-vectorizer.c (gate_increase_alignment): Fix return type.
2664 * ipa.c (function_and_variable_visibility): Fix return type.
2666 2007-01-08 Richard Guenther <rguenther@suse.de>
2668 * tree-ssa-ccp.c (maybe_fold_offset_to_array_ref): Use type
2669 of offset to build the index.
2670 * tree-pretty-print.c (dump_generic_node): Don't build negated
2671 const just for printing.
2672 * c-pretty-print.c (pp_c_integer_constant): Likewise.
2673 * builtins.c (fold_builtin_int_roundingfn): Check if result
2674 fits the type by using force_fit_type and comparing the result.
2675 * predict.c (predict_loops): Use compare_tree_int for comparison.
2676 * tree.c (build_int_cst): Fall back to integer_type_node for
2677 NULL_TREE type.
2678 (build_int_cst_wide): Assert type is non-null.
2680 2007-01-08 Roberto Costa <roberto.costa@st.com>
2682 * tree-vrp.c (extract_range_from_cond_expr): New.
2683 (extract_range_from_expr): Handle COND_EXPR nodes used as expressions.
2684 * tree-ssa-ccp.c (get_maxval_strlen): Handle COND_EXPR nodes used
2685 as expressions.
2686 (fold_stmt): Bug fix, avoid infinite recursion when folding COND_EXPRs.
2687 * tree-ssa-forwprop.c (simplify_cond, forward_propagate_into_cond,
2688 tree_ssa_forward_propagate_single_use_vars): Handle COND_EXPR nodes
2689 used as expressions.
2690 * tree-object-size.c (cond_expr_object_size): New.
2691 (collect_object_sizes_for): Handle COND_EXPR nodes used as expressions.
2693 2007-01-08 Jan Hubicka <jh@suse.cz>
2695 * tree-ssa-forwprop.c (forward_propagate_into_cond,
2696 tree_ssa_forward_propagate_single_use_va): Release defs of propagated
2697 statement.
2699 2007-01-08 Richard Guenther <rguenther@suse.de>
2701 PR tree-optimization/23603
2702 * tree-vrp.c (set_value_range_to_truthvalue): New function.
2703 (extract_range_from_binary): Fall back to truthvalue instead of
2704 varying for TRUTH_*_EXPR.
2705 (extract_range_from_comparison): Fall back to truthvalue instead of
2706 varying.
2707 (vrp_visit_phi_node): Don't adjust new range bounds to +INF/-INF
2708 if all visited PHI values were constant.
2710 2007-01-08 Jan Hubicka <jh@suse.cz>
2712 * cgraphunit.c (cgraph_process_new_functions): Reset reachable flag.
2713 (cgraph_analyze_function): break out from ...
2714 (cgraph_finalize_compilation_unit): ... here.
2715 (cgraph_expand_function): Remove forgoten commented out line.
2716 (cgraph_optimize): Analyze functions.
2718 2007-01-08 Jan Hubicka <jh@suse.cz>
2720 * tree-pas.h (TODO_remove_function): New flag.
2721 (TODO_update*): Renumber.
2722 (pass_ipa_increase_alignment,
2723 pass_ipa_function_and_variable_visibility): New passes.
2724 * cgraphunit.c (cgraph_increase_alignment): Move to tree-vectorizer.c
2725 (cgraph_function_and_variable_visibility): Move to ipa.c
2726 (cgraph_optimize): Don't call cgraph_function_and_variable_visibility,
2727 cgraph_increase_alignment.
2728 * ipa-inline.c (cgraph_decide_inlining): Don't push timevar.
2729 (cgraph_decide_inlining_incrementally): Push TV_INTEGRATION before
2730 calling tree-inline.
2731 (cgraph_early_inlining): Do not call cgraph_remove_unreachable_nodes.
2732 (pass_ipa_inline, pass_early_ipa_inlining): Set TODO_remove_functions
2733 * tree-vectorizer.c (increase_alignment): Move here from cgraphunit.c
2734 (gate_increase_alignment): New function.
2735 (pass_ipa_increase_alignment): New pass.
2736 * ipa.c: Inline tree-pass.h and timevar.h
2737 (function_and_variable_visibility): Move here from cgraphunit.c
2738 * tree-optimize.c (pass_early_local_passes): Add TODO_remove_functions.
2739 * passes.c (init_optimization_passes): Add the two new passes.
2740 (execute_todo): Handle cgraph_remove_functions.
2742 2007-01-08 Nick Clifton <nickc@redhat.com>
2744 * config/frv/predicates.md (reg_or_0_operand): Accept
2745 CONST_DOUBLEs.
2747 2007-01-08 Ralf Corsépius <ralf.corsepius@rtems.org>
2749 * config/bfin/rtems.h, config/bfin/t-rtems: New.
2750 * config.gcc: Add bfin*-rtems*.
2752 2007-01-08 Mark Shinwell <shinwell@codesourcery.com>
2754 * c.opt: Add -flax-vector-conversions.
2755 * c-typeck.c (convert_for_assignment): Pass flag to
2756 vector_types_convertible_p to allow emission of note.
2757 (digest_init): Likewise.
2758 * c-opts.c: Handle -flax-vector-conversions.
2759 * c-common.c (flag_lax_vector_conversions): New.
2760 (vector_types_convertible_p): Unless -flax-vector conversions
2761 has been passed, disallow conversions between vectors with
2762 differing numbers of subparts and/or element types. If such
2763 a conversion is disallowed, possibly emit a note on the first
2764 occasion only to inform the user of -flax-vector-conversions.
2765 The new last argument specifies this.
2766 * c-common.h (flag_lax_vector_conversions): New.
2767 (vector_types_convertible_p): Add extra argument.
2768 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Use
2769 char_type_node for V*QI type vectors.
2770 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins):
2771 Update to satisfy new typechecking rules.
2772 * config/rs6000/altivec.h (vec_cmple): Use vec_cmpge, for both
2773 C and C++ variants.
2774 * doc/invoke.texi (C Dialect Options): Document
2775 -flax-vector-conversions.
2777 2007-01-08 Mark Shinwell <shinwell@codesourcery.com>
2779 PR tree-optimization/29877
2780 * tree-ssa-ter.c (is_replaceable_p): Deem assignments with
2781 a register variable on the RHS to not be replaceable.
2783 2007-01-08 Chen Liqin <liqin@sunnorth.com.cn>
2784 * config/score/t-score-elf (MULTILIB_OPTIONS): Change.
2785 * config/score/predicates.md (const_uimm5, sr0_operand, const_simm12,
2786 const_simm15, const_pow2, const_npow2): Added.
2787 * config/score/misc.md (insv, extv, extzv, movmemsi,
2788 move_lbu_a/b, mov_lhu_a/b etc): Added and fix some bug.
2789 * config/score/score.c (score_address_cost, score_select_cc_mode):
2790 Added.
2791 Change CONST_OK_FOR_LETTER_P/EXTRA_CONSTRAINT define.
2792 Update score_rtx_costs for MACRO TARGET_RTX_COSTS.
2793 Update score_print_operand.
2794 * config/score/score.h (DATA_ALIGNMENT, SELECT_CC_MODE): Added.
2795 Adjust register allocate order and update some macro define.
2796 * config/score/score-mdaux.c (mdx_unaligned_load, mdx_unsigned_store,
2797 mdx_block_move_straight, mdx_block_move_loop_head,
2798 mdx_block_move_loop_body, mdx_block_move_loop_foot, mdx_block_move_loop,
2799 mdx_block_move): Added.
2800 (mdx_movsicc, mdp_select_add_imm, mdp_select, mds_zero_extract_andi,
2801 mdp_limm): Updated and fix some bug and typo.
2802 * config/score/score.md (movqi/hi/si, add/sub/zero/ext): Updated.
2803 (movsf, movdf, doloop_end): Added.
2805 2007-01-08 Kazu Hirata <kazu@codesourcery.com>
2807 * config/arm/arm.c, config/arm/arm.h, config/arm/arm.md,
2808 config/arm/thumb2.md: Fix comment typos.
2809 * doc/extend.texi: Fix a typo.
2811 2007-01-07 Eric Christopher <echristo@apple.com>
2813 * configure.ac: Check for __stack_chk_fail for darwin.
2814 * configure: Regenerate.
2816 2007-01-07 Richard Guenther <rguenther@suse.de>
2818 * tree-vrp.c (extract_range_from_assert): CSE calls to
2819 compare_values where possible.
2820 (extract_range_from_unary_expr): Likewise.
2822 2007-01-07 Anatoly Sokolov <aesok@post.ru>
2824 * config/avr/avr-protos.h (call_insn_operand): Delete prototype.
2825 * config/avr/avr.c (call_insn_operand): Delete function.
2826 * config/avr/avr.md (*pushqi, *pushhi, *pushsi, *pushsf): Use REG_SP
2827 instead of register number. Use predicates.
2828 * config/avr/predicates.md (const0_operand, reg_or_0_operand,
2829 call_insn_operand): Add.
2831 2007-01-06 Jan Hubicka <jh@suse.cz>
2833 * tree-pass.h (pass_build_cgraph_edges): Declare.
2834 * cgraphunit.c (record_refernece): Move to cgraphbuild.c
2835 (visited_nodes): Remove.
2836 (cgraph_create_edges): Move to cgraphbuild.c; rename to
2837 build_cgrpah_edges; make visited_nodes local.
2838 (cgraph_process_new_functions): DO not call initialize_inline_failed.
2839 (record_references_in_initializer): Move to cgraphbuild.c
2840 (initialize_inline_failed, rebuild_cgraph_edges,
2841 pass_rebuild_cgraph_edges): Move to cgraphbuild.c.
2842 (verify_cgraph_node): Make visited_nodes local.
2843 (cgraph_analyze_function): Do not call cgraph_create_edges and
2844 initialize_inline_failed.
2845 (cgraph_expand_function): Do not call cgraph_lower_function;
2846 assert that function is already lowered.
2847 * Makefile.in (cgraphbuild.o): New.
2848 * passes.c (init_optimization_passes): Add pass_build_cgraph_edges
2849 at the end of lowering passes.
2851 2007-01-06 Steven Bosscher <steven@gcc.gnu.org>
2853 * ifcvt.c (cond_move_convert_if_block): New function, code
2854 factored out from...
2855 (cond_move_process_if_block): ...here. Call the new function
2856 on the THEN and ELSE blocks.
2857 (merge_if_block): Do not copy global_live_at_end, merge_blocks
2858 already takes care of this.
2860 2007-01-05 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
2862 PR c/19978
2863 * tree.h (TREE_OVERFLOW_P): New.
2864 * c-typeck.c (parser_build_unary_op): Warn only if result
2865 overflowed and operands did not.
2866 (parser_build_binary_op): Likewise.
2867 (convert_for_assignment): Remove redundant overflow_warning.
2868 * c-common.c (overflow_warning): Don't check or set TREE_OVERFLOW.
2870 2007-01-05 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
2872 * c-typeck.c (store_init_value): Split over two lines to follow
2873 the GNU coding style.
2875 2007-01-05 Benjamin Kosnik <bkoz@redhat.com>
2877 * c-cppbuiltin.c (c_cpp_builtins): __GXX_EXPERIMENTAL_CPP0X__ to
2878 __GXX_EXPERIMENTAL_CXX0X__.
2879 * doc/cpp.texi: Same.
2881 2007-01-05 Richard Guenther <rguenther@suse.de>
2883 PR middle-end/27826
2884 * tree.c (get_narrower): Do not construct COMPONENT_REFs
2885 with mismatched types. Instead explicitly build a
2886 conversion NOP_EXPR.
2888 2007-01-05 Ian Lance Taylor <iant@google.com>
2890 * c-common.c (decl_with_nonnull_addr_p): New function.
2891 (c_common_truthvalue_conversion): Call it.
2892 * c-typeck.c (build_binary_op): Likewise.
2893 * c-common.h (decl_with_nonnull_addr_p): Declare.
2895 2007-01-05 Jakub Jelinek <jakub@redhat.com>
2897 PR c/30360
2898 * libgcc2.c (__divdc3): Compare c and d against 0.0 instead of
2899 denom against 0.0.
2901 2007-01-05 Joel Brobecker <brobecker@adacore.com>
2903 * doc/install.texi (Final install): Document the fact that
2904 the GNAT runtime should not be stripped.
2906 2007-01-04 Jan Hubicka <jh@suse.cz>
2908 * tree-inline.c (fold_marked_statements): Update operand caches
2909 and EH after folding
2911 2007-01-04 Ian Lance Taylor <iant@google.com>
2913 * c-common.c (check_function_nonnull): Whitespace fix.
2915 2007-01-04 Jan Hubicka <jh@suse.cz>
2917 * tree-optimize.c (execute_fixup_cfg): Correct previously mistakely
2918 comitted older version of patch.
2919 (pass_fixup_cfg): Add TODOs to verify flow and statements, dump
2920 function, celanup cfg and collect garbage.
2922 2007-01-04 Mike Stump <mrs@apple.com>
2924 * Makefile.in (mostlyclean): Don't remove libgcc anymore.
2925 (clean): Likewise.
2927 2007-01-04 Eric Christopher <echristo@apple.com>
2929 * libgcc2.c (__bswapsi2): Use SItype.
2930 (__bswapdi2): Use DItype.
2931 * libgcc2.h: Update for above.
2933 2007-01-04 Paul Brook <paul@codesourcery.com>
2935 * config/arm/arm.md (arm_mulsi3, thumb_mulsi3, mulsi3_compare0,
2936 mulsi_compare0_scratch, mulsi3addsi, mulsi3addsi_compare0,
2937 mulsi3addsi_compare0_scratch, mulsidi3adddi, mulsidi3,
2938 umulsidi3, umulsidi3adddi, smulsi3_highpart,
2939 umulsi3_highpart): Make conditional on !arm_arch6.
2940 (arm_mulsi3_v6, thumb_mulsi3_v6, mulsi3_compare0_v6,
2941 mulsi_compare0_scratch_v6, mulsi3addsi_v6, mulsi3addsi_compare0_v6,
2942 mulsi3addsi_compare0_scratch_v6, mulsidi3adddi_v6, mulsidi3_v6,
2943 umulsidi3_v6, umulsidi3adddi_v6, smulsi3_highpart_v6,
2944 umulsi3_highpart_v6): New insns.
2946 2007-01-04 Roger Sayle <roger@eyesopen.com>
2948 * fold-const.c (fold_convert): When casting an expression to void,
2949 fold_ignored_result may discover a GIMPLE_MODIFY_STMT which doesn't
2950 have a type. Instead of attempting to build a NOP_EXPR, return
2951 these "special" trees directly.
2953 2007-01-04 Joseph Myers <joseph@codesourcery.com>
2955 * config/rs6000/rs6000.c (rs6000_rtx_costs): Make adjustment for
2956 MULT inside MINUS as either argument. Use rs6000_cost->dmul -
2957 rs6000_cost->fp not 0 as adjustment for outer NEG.
2959 2007-01-04 Jan Hubicka <jh@suse.cz>
2961 * cgraph.c (cgraph_release_function_body): New function.
2962 (cgraph_remove_node): Use it.
2963 * cgraph.h (cgraph_release_function_body): Declare.
2964 * cgraphunit.c (cgraph_expand_function): Use it.
2965 * ipa.c (cgraph_remove_unreahchable_nodes): Use it.
2966 * tree-ssa.c (delete_tree_ssa): Allow to be called before aliasing
2967 is initialized and while compilation of other function is running.
2968 * tree-optimize.c (execute_free_cfg_annotations): Move code to clear
2969 statement CFG annotations from here to ...
2970 * tree-cfg.c (delete_tree_cfg_annotations): ... here.
2972 2007-01-04 Zdenek Dvorak <dvorakz@suse.cz>
2974 * cfgloop.h (enum li_flags): Make the constants powers of two.
2976 2007-01-04 Jan Hubicka <jh@suse.cz>
2978 * tree-inline.c (copy_bb): Insert new statements to statements_to_fold
2979 set.
2980 (fold_marked_statements): New function.
2981 (optimize_inline_calls, tree_function_versioning): Fold new statements.
2982 * tree-inline.h (copy_body_data): Add statements_to_fold.
2984 2007-01-03 Daniel Jacobowitz <dan@codesourcery.com>
2986 * config.gcc: Mention libgcc/config.host.
2987 * Makefile.in: Update comments mentioning libgcc.
2988 (LIBGCC, INSTALL_LIBGCC, GCC_PARTS, mklibgcc): Delete.
2989 (all.cross, start.encap, rest.encap, rest.cross): Update
2990 dependencies for libgcc move.
2991 (libgcc.mk, LIBGCC_DEPS, libgcov.a, libgcc.a, stmp-multilib)
2992 (clean-target, clean-target-libgcc): Delete.
2993 (srcdirify, GCC_EXTRA_PARTS): New macros.
2994 (libgcc-support, libgcc.mvars): New rules.
2995 (distclean): Remove mention of mklibgcc.
2996 (install): Don't reference INSTALL_LIBGCC.
2997 (install-common): Don't reference EXTRA_PARTS.
2998 (install-libgcc, install-multilib): Delete rules.
2999 * mklibgcc.in: Delete file.
3000 * doc/configfiles.texi: Don't mention mklibgcc.
3002 * config/i386/t-darwin (SHLIB_VERPFX): Delete (moved to libgcc).
3003 * config/i386/t-darwin64 (SHLIB_VERPFX): Likewise.
3004 * config/rs6000/t-darwin (SHLIB_VERPFX): Likewise.
3005 * config/rs6000/t-ppccomm (TARGET_LIBGCC2_CFLAGS, SHLIB_MAPFILES)
3006 (mklibgcc, ldblspecs): Likewise.
3008 * config/i386/t-nwld (libgcc.def, libc.def, libpcre.def)
3009 (posixpre.def): Use $(T).
3010 (SHLIB_EXT, SHLIB_NAME, SHLIB_SLIBDIR_QUAL, SHLIB_DEF, SHLIB_MAP)
3011 (SHLIB_SRC, SHLIB_INSTALL): Delete.
3012 (SHLIB_LINK): Make dummy.
3013 * config/t-slibgcc-darwin: Delete contents except for dummy SHLIB_LINK.
3015 * config/frv/t-linux (EXTRA_MULTILIB_PARTS): Clear.
3017 * config/alpha/t-crtfm: Use $(T) in rules for EXTRA_PARTS.
3018 * config/alpha/t-vms, config/alpha/t-vms64, config/fr30/t-fr30,
3019 config/i386/t-rtems-i386, config/ia64/t-ia64, config/rs6000/t-beos,
3020 config/rs6000/t-newas, config/sparc/t-elf: Likewise.
3022 * configure.ac (all_outputs): Remove mklibgcc.
3023 * configure: Regenerated.
3025 2007-01-03 Josh Conner <jconner@apple.com>
3027 PR middle-end/29683
3028 * calls.c (compute_argument_addresses): Set stack and stack_slot
3029 for partial args, too.
3030 (store_one_arg): Use locate.size.constant for the size when
3031 generating a save_area.
3033 2007-01-03 Robert Kennedy <jimbob@google.com>
3035 * tree-cfg.c (tree_merge_blocks): Release SSA_NAME phi results
3036 whose definitions are deleted due to basic block merging.
3038 2007-01-03 Paul Brook <paul@codesourcery.com>
3040 PR target/16634
3041 * config/arm/arm.c (output_return_instruction): Pop PC in interrupt
3042 functions.
3043 (use_return_insn): Return 0 for Thumb interrupt functions.
3044 (print_multi_reg): Add rfe argument for IRQ returns.
3045 (arm_output_epilogue): Pop interrupt return address directly into PC.
3046 (arm_expand_prologue): Only adjust IRQ return address in Arm mode.
3048 2007-01-03 Paul Brook <paul@codesourcery.com>
3050 Merge from sourcerygxx-4_1.
3051 * config/arm/thumb2.md: New file.
3052 * config/arm/elf.h (JUMP_TABLES_IN_TEXT_SECTION): Return True for
3053 Thumb-2.
3054 * config/arm/coff.h (JUMP_TABLES_IN_TEXT_SECTION): Ditto.
3055 * config/arm/aout.h (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
3056 (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump tables.
3057 * config/arm/aof.h (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump
3058 tables.
3059 (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
3060 * config/arm/ieee754-df.S: Use macros for Thumb-2/Unified asm
3061 comptibility.
3062 * config/arm/ieee754-sf.S: Ditto.
3063 * config/arm/arm.c (thumb_base_register_rtx_p): Rename...
3064 (thumb1_base_register_rtx_p): ... to this.
3065 (thumb_index_register_rtx_p): Rename...
3066 (thumb1_index_register_rtx_p): ... to this.
3067 (thumb_output_function_prologue): Rename...
3068 (thumb1_output_function_prologue): ... to this.
3069 (thumb_legitimate_address_p): Rename...
3070 (thumb1_legitimate_address_p): ... to this.
3071 (thumb_rtx_costs): Rename...
3072 (thumb1_rtx_costs): ... to this.
3073 (thumb_compute_save_reg_mask): Rename...
3074 (thumb1_compute_save_reg_mask): ... to this.
3075 (thumb_final_prescan_insn): Rename...
3076 (thumb1_final_prescan_insn): ... to this.
3077 (thumb_expand_epilogue): Rename...
3078 (thumb1_expand_epilogue): ... to this.
3079 (arm_unwind_emit_stm): Rename...
3080 (arm_unwind_emit_sequence): ... to this.
3081 (thumb2_legitimate_index_p, thumb2_legitimate_address_p,
3082 thumb1_compute_save_reg_mask, arm_dwarf_handle_frame_unspec,
3083 thumb2_index_mul_operand, output_move_vfp, arm_shift_nmem,
3084 arm_save_coproc_regs, thumb_set_frame_pointer, arm_print_condition,
3085 thumb2_final_prescan_insn, thumb2_asm_output_opcode, arm_output_shift,
3086 thumb2_output_casesi): New functions.
3087 (TARGET_DWARF_HANDLE_FRAME_UNSPEC): Define.
3088 (FL_THUMB2, FL_NOTM, FL_DIV, FL_FOR_ARCH6T2, FL_FOR_ARCH7,
3089 FL_FOR_ARCH7A, FL_FOR_ARCH7R, FL_FOR_ARCH7M, ARM_LSL_NAME,
3090 THUMB2_WORK_REGS): Define.
3091 (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv, arm_condexec_count,
3092 arm_condexec_mask, arm_condexec_masklen)): New variables.
3093 (all_architectures): Add armv6t2, armv7, armv7a, armv7r and armv7m.
3094 (arm_override_options): Check new CPU capabilities.
3095 Set new architecture flag variables.
3096 (arm_isr_value): Handle v7m interrupt functions.
3097 (user_return_insn): Return 0 for v7m interrupt functions. Handle
3098 Thumb-2.
3099 (const_ok_for_arm): Handle Thumb-2 constants.
3100 (arm_gen_constant): Ditto. Use movw when available.
3101 (arm_function_ok_for_sibcall): Return false for v7m interrupt
3102 functions.
3103 (legitimize_pic_address, arm_call_tls_get_addr): Handle Thumb-2.
3104 (thumb_find_work_register, arm_load_pic_register,
3105 legitimize_tls_address, arm_address_cost, load_multiple_sequence,
3106 emit_ldm_seq, emit_stm_seq, arm_select_cc_mode, get_jump_table_size,
3107 print_multi_reg, output_mov_long_double_fpa_from_arm,
3108 output_mov_long_double_arm_from_fpa, output_mov_double_fpa_from_arm,
3109 output_mov_double_fpa_from_arm, output_move_double,
3110 arm_compute_save_reg_mask, arm_compute_save_reg0_reg12_mask,
3111 output_return_instruction, arm_output_function_prologue,
3112 arm_output_epilogue, arm_get_frame_offsets, arm_regno_class,
3113 arm_output_mi_thunk, thumb_set_return_address): Ditto.
3114 (arm_expand_prologue): Handle Thumb-2. Use arm_save_coproc_regs.
3115 (arm_coproc_mem_operand): Allow POST_INC/PRE_DEC.
3116 (arithmetic_instr, shift_op): Use arm_shift_nmem.
3117 (arm_print_operand): Use arm_print_condition. Handle '(', ')', '.',
3118 '!' and 'L'.
3119 (arm_final_prescan_insn): Use extract_constrain_insn_cached.
3120 (thumb_expand_prologue): Use thumb_set_frame_pointer.
3121 (arm_file_start): Output directive for unified syntax.
3122 (arm_unwind_emit_set): Handle stack alignment instruction.
3123 * config/arm/lib1funcs.asm: Remove default for __ARM_ARCH__.
3124 Add v6t2, v7, v7a, v7r and v7m.
3125 (RETLDM): Add Thumb-2 code.
3126 (do_it, shift1, do_push, do_pop, COND, THUMB_SYNTAX): New macros.
3127 * config/arm/arm.h (TARGET_CPU_CPP_BUILTINS): Define __thumb2__.
3128 (TARGET_THUMB1, TARGET_32BIT, TARGET_THUMB2, TARGET_DSP_MULTIPLY,
3129 TARGET_INT_SIMD, TARGET_UNIFIED_ASM, ARM_FT_STACKALIGN, IS_STACKALIGN,
3130 THUMB2_TRAMPOLINE_TEMPLATE, TRAMPOLINE_ADJUST_ADDRESS,
3131 ASM_OUTPUT_OPCODE, THUMB2_GO_IF_LEGITIMATE_ADDRESS,
3132 THUMB2_LEGITIMIZE_ADDRESS, CASE_VECTOR_PC_RELATIVE,
3133 CASE_VECTOR_SHORTEN_MODE, ADDR_VEC_ALIGN, ASM_OUTPUT_CASE_END,
3134 ADJUST_INSN_LENGTH): Define.
3135 (TARGET_REALLY_IWMMXT, TARGET_IWMMXT_ABI, CONDITIONAL_REGISTER_USAGE,
3136 STATIC_CHAIN_REGNUM, HARD_REGNO_NREGS, INDEX_REG_CLASS,
3137 BASE_REG_CLASS, MODE_BASE_REG_CLASS, SMALL_REGISTER_CLASSES,
3138 PREFERRED_RELOAD_CLASS, SECONDARY_OUTPUT_RELOAD_CLASS,
3139 SECONDARY_INPUT_RELOAD_CLASS, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P,
3140 TRAMPOLINE_SIZE, INITIALIZE_TRAMPOLINE, HAVE_PRE_INCREMENT,
3141 HAVE_POST_DECREMENT, HAVE_PRE_DECREMENT, HAVE_PRE_MODIFY_DISP,
3142 HAVE_POST_MODIFY_DISP, HAVE_PRE_MODIFY_REG, HAVE_POST_MODIFY_REG,
3143 REGNO_MODE_OK_FOR_BASE_P, LEGITIMATE_CONSTANT_P,
3144 REG_MODE_OK_FOR_BASE_P, REG_OK_FOR_INDEX_P, GO_IF_LEGITIMATE_ADDRESS,
3145 LEGITIMIZE_ADDRESS, THUMB2_LEGITIMIZE_ADDRESS,
3146 GO_IF_MODE_DEPENDENT_ADDRESS, MEMORY_MOVE_COST, BRANCH_COST,
3147 ASM_APP_OFF, ASM_OUTPUT_CASE_LABEL, ARM_DECLARE_FUNCTION_NAME,
3148 FINAL_PRESCAN_INSN, PRINT_OPERAND_PUNCT_VALID_P,
3149 PRINT_OPERAND_ADDRESS): Adjust for Thumb-2.
3150 (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv): New declarations.
3151 * config/arm/arm-cores.def: Add arm1156t2-s, cortex-a8, cortex-r4 and
3152 cortex-m3.
3153 * config/arm/arm-tune.md: Regenerate.
3154 * config/arm/arm-protos.h: Update prototypes.
3155 * config/arm/vfp.md: Enable patterns for Thumb-2.
3156 (arm_movsi_vfp): Add movw alternative. Use output_move_vfp.
3157 (arm_movdi_vfp, movsf_vfp, movdf_vfp): Use output_move_vfp.
3158 (thumb2_movsi_vfp, thumb2_movdi_vfp, thumb2_movsf_vfp,
3159 thumb2_movdf_vfp, thumb2_movsfcc_vfp, thumb2_movdfcc_vfp): New.
3160 * config/arm/libunwind.S: Add Thumb-2 code.
3161 * config/arm/constraints.md: Update include Thumb-2.
3162 * config/arm/ieee754-sf.S: Add Thumb-2/Unified asm support.
3163 * config/arm/ieee754-df.S: Ditto.
3164 * config/arm/bpabi.S: Ditto.
3165 * config/arm/t-arm (MD_INCLUDES): Add thumb2.md.
3166 * config/arm/predicates.md (low_register_operand,
3167 low_reg_or_int_operand, thumb_16bit_operator): New.
3168 (thumb_cmp_operand, thumb_cmpneg_operand): Rename...
3169 (thumb1_cmp_operand, thumb1_cmpneg_operand): ... to this.
3170 * config/arm/t-arm-elf: Add armv7 multilib.
3171 * config/arm/arm.md: Update patterns for Thumb-2 and Unified asm.
3172 Include thumb2.md.
3173 (UNSPEC_STACK_ALIGN, ce_count): New.
3174 (arm_incscc, arm_decscc, arm_umaxsi3, arm_uminsi3,
3175 arm_zero_extendsidi2, arm_zero_extendqidi2): New
3176 insns/expanders.
3177 * config/arm/fpa.md: Update patterns for Thumb-2 and Unified asm.
3178 (thumb2_movsf_fpa, thumb2_movdf_fpa, thumb2_movxf_fpa,
3179 thumb2_movsfcc_fpa, thumb2_movdfcc_fpa): New insns.
3180 * config/arm/cirrus.md: Update patterns for Thumb-2 and Unified asm.
3181 (cirrus_thumb2_movdi, cirrus_thumb2_movsi_insn,
3182 thumb2_cirrus_movsf_hard_insn, thumb2_cirrus_movdf_hard_insn): New
3183 insns.
3184 * doc/extend.texi: Document ARMv7-M interrupt functions.
3185 * doc/invoke.texi: Document Thumb-2 new cores+architectures.
3187 2007-01-03 Jakub Jelinek <jakub@redhat.com>
3189 * unwind-dw2.c (SIGNAL_FRAME_BIT, EXTENDED_CONTEXT_BIT): Define.
3190 (struct _Unwind_Context): Rename args_size to flags, remove
3191 signal_frame field, add a new args_size field and version field.
3192 (_Unwind_IsSignalFrame, _Unwind_SetSignalFrame,
3193 _Unwind_IsExtendedContext): New inline functions.
3194 (_Unwind_GetGR, _Unwind_SetGR, _Unwind_GetGRPtr, _Unwind_SetGRPtr):
3195 Assume by_value array is only present if _Unwind_IsExtendedContext.
3196 (_Unwind_GetIPInfo, execute_cfa_program, uw_frame_state_for): Use
3197 _Unwind_IsSignalFrame.
3198 (__frame_state_for): Initialize context.flags to EXTENDED_CONTEXT_BIT.
3199 (uw_update_context_1): Use _Unwind_SetSignalFrame.
3200 (uw_init_context_1): Initialize context->flags to
3201 EXTENDED_CONTEXT_BIT.
3202 * config/rs6000/linux-unwind.h (frob_update_context): Use
3203 _Unwind_SetSignalFrame.
3205 2007-01-03 Andrew Pinski <andrew_pinski@playstation.sony.com>
3207 PR middle-end/30353
3208 * gimplify.c (gimplify_modify_expr_complex_part): Move below
3209 tree_to_gimple_tuple. Call tree_to_gimple_tuple when we need
3210 the value.
3212 2007-01-03 Kazu Hirata <kazu@codesourcery.com>
3214 * config/i386/i386.h (NON_STACK_REG_P, REGNO_OK_FOR_SIREG_P,
3215 REGNO_OK_FOR_DIREG_P, REWRITE_ADDRESS, ASM_OPERAND_LETTER,
3216 RET, AT_SP): Remove.
3217 * config/i386/i386.md (*sse_prologue_save_insn): Use return
3218 instead of RET.
3220 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
3221 * cfgbuild.c (find_basic_blocks): Likewise.
3222 * cfgrtl.c (rtl_create_basic_block): Likewise.
3223 * function.c (temp_slots_at_level): Likewise.
3224 * reg-stack.c (stack_regs_mentioned): Likewise.
3225 * regclass.c (allocate_reg_info): Likewise.
3226 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
3227 set_bb_for_stmt, move_block_to_fn): Likewise.
3228 * tree-complex.c (tree_lower_complex): Likewise.
3229 * vec.h (VEC_safe_grow_cleared): New.
3231 * cgraphunit.c, tree-ssa-alias.c: Fix comment typos.
3233 2007-01-03 Zdenek Dvorak <dvorakz@suse.cz>
3235 * loop-unswitch.c (unswitch_loop): Pass probabilities to loopify.
3236 * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Pass probabilities
3237 to loop_version.
3238 * cfgloopmanip.c (scale_loop_frequencies): Export.
3239 (loopify): Scale the frequencies by prescribed coefficients.
3240 (set_zero_probability): New function.
3241 (duplicate_loop_to_header_edge): Improve updating of frequencies.
3242 (lv_adjust_loop_entry_edge, loop_version): Set probabilities
3243 and frequencies according to arguments.
3244 * tree-ssa-loop-manip.c (tree_unroll_loop): Set probabilities
3245 correctly.
3246 * cfg.c (scale_bbs_frequencies_int): Allow scaling the frequencies up.
3247 * modulo-sched.c (sms_schedule): Set probabilities for entering
3248 versioned loop correctly.
3249 * tree-vect-transform.c (vect_transform_loop): Ditto.
3250 * cfgloop.h (loopify, loop_version): Declaration changed.
3251 (scale_loop_frequencies): Declared.
3253 2007-01-02 Jan Hubicka <jh@suse.cz>
3255 * cgraph.c: Include tree-flow.h
3256 (cgraph_add_new-function): Handle IPA_SSA mode; execute
3257 early_local_passes.
3258 * cgraph.h (enum cgraph_state): Add CGRAPH_STATE_IPA_SSA.
3259 * tree-pass.h (pass_all_early_optimizations): Declare.
3260 * cgraphunit.c (cgraph_process_new_functions): Add IPA_SSA; execute
3261 early_local_passes.
3262 (cgraph_analyze_function): Do early_local_passes.
3263 * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
3264 Do not add referenced vars.
3265 * tree-optimize.c (gate_all_optimizations): Do not execute when not in
3266 SSA form.
3267 (gate_all_early_local_passes): New gate.
3268 (pass_early_local_passes): Use new gate.
3269 (execute_early_local_optimizations): New functions.
3270 (gate_all_early_optimizations): New gate.
3271 (pass_all_early_optimizations): New pass.
3272 (execute_free_datastructures): Free SSA only when initialized.
3273 (gate_init_datastructures): Init only when optimizing.
3274 (tree_lowering_passes): Do early local passes when called late.
3275 * tree-profile.c (do_tree_profiling): Don't profile functions added
3276 late.
3277 (do_early_tree_profiling, pass_early_tree_profile): Kill.
3278 * tree-cfg.c (update_modified_stmts): Do not update when operands are
3279 not active.
3280 * passes.c (init_optimizations_passes): Reorder so we go into SSA
3281 during early_local_passes.
3282 * Makefile.in (cgraph.o): Add dependency on tree-flow.h.
3285 2007-01-02 Carlos O'Donell <carlos@codesourcery.com>
3287 * Makefile.in: Update copyright year.
3289 2007-01-02 Carlos O'Donell <carlos@codesourcery.com>
3291 * Makefile.in: Export GCC_EXEC_PREFIX before calling $(RUNTEST)
3292 in $(lang_checks) and check-consistency targets.
3294 2007-01-02 Jan Hubicka <jh@suse.cz>
3296 * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
3297 Do not add referenced vars.
3298 * tree-cfg.c (update_modified_stmts): Do not update when SSA operands
3299 are not active.
3300 * passes.c (init_optimization_passes): Put mudflap_2 after
3301 free_datastructures.
3303 2007-01-02 Jan Hubicka <jh@suse.cz>
3305 * tree-optimize (execute_fixup_cfg): Set after_inlining flag.
3306 Set NOTHROW flag on call statements proved to be nothrow.
3307 Update statement of local calls so new pure/const functions are
3308 updated. Update_ssa when in ssa form. Mark PHI nodes of nonlocal
3309 goto receivers.
3310 (tree_rest_of_compilation): Register hooks and initialize bitmap
3311 early. Do not set after_inlining flag.
3313 2007-01-02 Steve Ellcey <sje@cup.hp.com>
3315 * sbitmap.c (HOST_BITS_PER_LONG_LONG): Change to
3316 HOST_BITS_PER_LONGLONG
3318 2007-01-02 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
3320 PR c/19977
3321 * c-typeck.c (store_init_value): Don't emit pedantic overflow
3322 warning for non-static initializers.
3324 2007-01-02 Steven Bosscher <steven@gcc.gnu.org>
3326 * config/alpha/alpha.md, arm/arm.c, darwin.c, frv/frv.md,
3327 m32r/m32r.c, m32r/m32r.c, mn10300/mn10300.md, pa/pa.c,
3328 rs6000/rs6000.c, s390/s390.md, sh/sh.md, sparc/sparc.c:
3329 Always use set_unique_reg_note to add REG_EQUAL notes.
3331 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
3333 Revert:
3334 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
3336 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
3337 * cfgbuild.c (find_basic_blocks): Likewise.
3338 * cfgrtl.c (rtl_create_basic_block): Likewise.
3339 * function.c (temp_slots_at_level): Likewise.
3340 * reg-stack.c (stack_regs_mentioned): Likewise.
3341 * regclass.c (allocate_reg_info): Likewise.
3342 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
3343 set_bb_for_stmt, move_block_to_fn): Likewise.
3344 * tree-complex.c (tree_lower_complex): Likewise.
3345 * vec.h (VEC_safe_grow_cleared): New.
3347 2007-01-02 Ian Lance Taylor <iant@google.com>
3349 * c-common.c (c_common_truthvalue_conversion): When warning about
3350 using an assignment as a truth value, set TREE_NO_WARNING.
3352 2007-01-02 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
3354 PR middle-end/7651
3355 * c.opt (Wold-style-declaration): New.
3356 * doc/invoke.texi (C-only Warning Options): New.
3357 (Wold-style-declaration): Document it.
3358 (Wextra): Enabled by -Wextra.
3359 * c-opts.c (c_common_post_options): Enabled by -Wextra.
3360 * c-decl.c (declspecs_add_scspec): Replace -Wextra with
3361 -Wold-style-declaration.
3363 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
3365 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
3366 * cfgbuild.c (find_basic_blocks): Likewise.
3367 * cfgrtl.c (rtl_create_basic_block): Likewise.
3368 * function.c (temp_slots_at_level): Likewise.
3369 * reg-stack.c (stack_regs_mentioned): Likewise.
3370 * regclass.c (allocate_reg_info): Likewise.
3371 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
3372 set_bb_for_stmt, move_block_to_fn): Likewise.
3373 * tree-complex.c (tree_lower_complex): Likewise.
3374 * vec.h (VEC_safe_grow_cleared): New.
3376 2007-01-02 Douglas Gregor <doug.gregor@gmail.com>
3378 * c-common.c (c_common_nodes_and_builtins): Since variants of
3379 void_type_node get built before it is given a name, we need to
3380 give those variants the name, too.
3381 (complete_array_type): We need to work with the canonical main
3382 type of the array, from which we will build the qualified version.
3383 * params.def (PARAM_VERIFY_CANONICAL_TYPES): New.
3384 * print-tree.c (print_node): Display canonical type information
3385 for each type.
3386 * stor-layout.c (layout_type): When we don't know the
3387 alignment of a type for which we're building an array, we end up
3388 guessing wrong, so make the type require structural equality.
3389 * tree.c (make_node_stat): When we build a new type, it is its
3390 own canonical type.
3391 (build_type_attribute_qual_variant): When building an attribute
3392 variant, its canonical type is the non-attribute variant. However,
3393 if the attributes are target-dependent and they differ, we need to
3394 use structural equality checks for this type.
3395 (build_qualified_type): A qualified type is not equivalent to its
3396 unqualified variant; set the canonical type appropriately.
3397 (build_distinct_type_copy): When building a distinct type from
3398 another type, the new type is its own canonical type.
3399 (build_variant_type_copy): When building a new type variant, we
3400 assume that it is equivalent to the original type.
3401 (build_pointer_type_for_mode): When building a pointer type, also
3402 build a canonical type pointer.
3403 (build_reference_type_for_mode): When building a reference type,
3404 also build a canonical type reference.
3405 (build_index_type): When we can't hash an index type (e.g.,
3406 because its maximum value is negative), the index type requires
3407 structural equality tests.
3408 (build_array_type): Build the canonical form of an array type.
3409 (build_function_type): Function types require structural equality,
3410 because they contain default arguments, attributes, etc.
3411 (build_method_type_directly): Ditto for method types.
3412 (build_offset_type): Build the canonical offset type.
3413 (build_complex_type): Build the canonical vector type.
3414 (make_vector_type): Build the canonical vector type.
3415 * tree.h (TYPE_CANONICAL): New.
3416 (TYPE_STRUCTURAL_EQUALITY_P): New.
3417 (SET_TYPE_STRUCTURAL_EQUALITY): New.
3418 (struct tree_type): Added "canonical" field.
3419 * params.h (VERIFY_CANONICAL_TYPES): New.
3420 * doc/c-tree.texi (TYPE_CANONICAL): Document.
3421 (TYPE_STRUCTURAL_EQUALITY_P): Document.
3422 (SET_TYPE_STRUCTURAL_EQUALITY): Document.
3423 * doc/invoke.texi (verify-canonical-types): Document --param
3424 parameter for verifying canonical types.
3426 2007-01-02 Joseph Myers <joseph@codesourcery.com>
3428 * config.gcc (powerpc-*-eabispe*, powerpc-*-eabisimaltivec*,
3429 powerpc-*-eabisim*, powerpc-*-eabialtivec*, powerpc-*-eabi*,
3430 powerpc-*-rtems*, powerpc-wrs-vxworks, powerpc-wrs-vxworksae,
3431 powerpcle-*-eabisim*, powerpcle-*-eabi*): Add rs6000/e500.h to
3432 tm_file.
3433 * config/rs6000/e500.h: New.
3434 * config/rs6000/eabi.h (TARGET_SPE_ABI, TARGET_SPE, TARGET_E500,
3435 TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE, TARGET_E500_DOUBLE):
3436 Remove.
3437 * config/rs6000/linuxspe.h (TARGET_SPE_ABI, TARGET_SPE,
3438 TARGET_E500, TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE,
3439 TARGET_E500_DOUBLE): Remove.
3440 * config/rs6000/vxworks.h (TARGET_SPE_ABI, TARGET_SPE,
3441 TARGET_E500, TARGET_ISEL, TARGET_FPRS): Remove.
3442 * config/rs6000/rs6000.h (CHECK_E500_OPTIONS): Define.
3443 * config/rs6000/rs6000.c (rs6000_override_options): Use
3444 CHECK_E500_OPTIONS.
3446 2007-01-02 Joseph Myers <joseph@codesourcery.com>
3448 * config/rs6000/rs6000.c (print_operand): Check (TARGET_SPE ||
3449 TARGET_E500_DOUBLE), not TARGET_E500, for %y.
3450 (rs6000_generate_compare, rs6000_emit_sCOND, output_cbranch,
3451 rs6000_emit_cmove): Don't check TARGET_E500.
3452 * config/rs6000/rs6000.md (bunordered, bordered, sunordered,
3453 sordered): Don't check TARGET_E500.
3455 2007-01-01 Eric Christopher <echristo@apple.com>
3457 * config/mips/mips.c (mips_regno_mode_ok_for_base_p): Use
3458 HARD_REGISTER_NUM_P.
3460 2007-01-01 Roger Sayle <roger@eyesopen.com>
3462 * fold-const.c (fold_binary) <EQ_EXPR>: Fold "(X^C1) eq/ne C2" into
3463 "X eq/ne (C1^C2)". Fold "(X^Z) eq/ne (Y^Z)" as "X eq/ne Y" when Z
3464 has no side-effects. Fold "(X^C1) eq/ne (Y^C2)" as "(X^(C1^C2))
3465 eq/ne Y".
3467 2007-01-01 Mike Stump <mrs@apple.com>
3469 * configure.ac: Remove support for building with Apple's gcc-3.1.
3471 2007-01-02 Joseph Myers <joseph@codesourcery.com>
3473 PR middle-end/30311
3474 * caller-save.c (add_stored_regs): Only handle SUBREGs if inner
3475 REG is a hard register. Do not modify REG before calling
3476 subreg_nregs.
3477 * rtlanal.c (subreg_get_info): Don't assert size of XMODE is a
3478 multiple of the size of YMODE for certain lowpart cases.
3480 2007-01-01 Andrew Pinski <pinskia@gmail.com>
3482 PR middle-end/30253
3483 * gimplify (voidify_wrapper_expr): Update for
3484 GIMPLE_MODIFY_STMT.
3486 2007-01-01 Andreas Schwab <schwab@suse.de>
3488 PR target/29166
3489 * config/ia64/ia64.c (ia64_compute_frame_size): Account space for
3490 save of BR0 in extra_spill_size instead of spill_size.
3491 (ia64_expand_prologue): Save BR0 outside of the gr/br/fr spill
3492 area.
3493 (ia64_expand_epilogue): Restore BR0 from its new location.
3495 2007-01-01 Andrew Pinski <pinskia@gmail.com>
3497 * gimplify.c (gimplify_init_constructor <case VECTOR_TYPE>):
3498 Use a temporary variable if the left hand side is not a gimple
3499 register.
3501 2007-01-01 Andrew Pinski <pinskia@gmail.com>
3503 * gimplify.c (gimplify_return_expr): Make the temporary variable
3504 for the return expression, a gimple register variable.
3506 2007-01-01 Jan Hubicka <jh@suse.cz>
3508 * emit-rtl.c (emit_copy_of_insn_after): Do not call copy_insn_1 for
3509 INSN_LIST.
3511 2007-01-01 Mike Stump <mrs@apple.com>
3513 * configure.ac (HAVE_GAS_LITERAL16): Add autoconf check for
3514 .literal16.
3515 * config/darwin.c (machopic_select_rtx_section): Use
3516 HAVE_GAS_LITERAL16.
3517 (darwin_mergeable_constant_section): Likewise.
3518 * configure: Regenerate.
3519 * config.in: Regenerate.
3521 2007-01-01 Jan Hubicka <jh@suse.cz>
3522 Andrew Pinski <pinskia@gmail.com>
3524 * cgraphunit.c (cgraph_optimize): Call cgraph_add_new_functions
3525 before starting IPA passes.