| blob | 7aadf4e6930ae3c301f3b9f464012e392b9b81cd |
1 2007-02-01 Geoffrey Keating <geoffk@apple.com>
3 * config/rs6000/darwin.h (MD_UNWIND_SUPPORT): Don't define for
4 64-bit.
6 2007-2-01 Seongbae Park <seongbae.park@gmail.com>
8 PR inline-asm/28686
9 * global.c (global_alloc): Add mising initialization of
10 ELIMINABLE_REGSET.
12 2007-02-01 Roger Sayle <roger@eyesopen.com>
14 * alias.c (init_alias_analysis): Correct whitespace.
15 * bb-reorder.c (fix_edges_for_rarely_executed_code,
16 partition_hot_cold_basic_blocks): Likewise.
17 * builtins.c (expand_builtin_printf, expand_builtin_fprintf,
18 expand_builtin_sprintf, fold_builtin_carg, fold_builtin_sprintf,
19 maybe_emit_sprintf_chk_warning, fold_builtin_sprintf_chk,
20 fold_builtin_snprintf_chk, fold_builtin_printf,
21 fold_builtin_fprintf, do_mpfr_ckconv, do_mpfr_arg1, do_mpfr_arg2,
22 do_mpfr_arg3, do_mpfr_sincos): Likewise.
23 * cfgcleanup.c (cleanup_cfg): Likewise.
24 * cfgexpand.c (tree_expand_cfg): Likewise.
25 * fold-const.c (fold_binary) <RDIV_EXPR>: Likewise.
26 * function.c (get_next_funcdef_no): Likewise.
27 * gengtype.c (main): Likewise.
28 * genmodes.c (main): Likewise.
29 * gcse.c (bypass_conditional_jumps, print_ldst_list): Likewise.
30 * haifa-sched.c (schedule_block, extend_h_i_d): Likewise.
31 * ifcvt.c (noce_emit_move_insn): Likewise.
32 * modulo-sched.c (generate_prolog_epilog, sms_schedule_by_order):
33 Likewise.
34 * stor-layout.c (get_best_mode): Likewise.
35 * tree-ssa-loop-niter.c (get_val_for): Likewise.
36 * tree-ssa-structalias.c (get_varinfo, get_varinfo_fc,
37 scc_visit, do_ds_constraint, do_complex_constraint, label_visit,
38 perform_var_substitution, solve_graph): Likewise.
39 * tree-vrp.c (vrp_finalize): Likewise.
41 2007-02-01 Ian Lance Taylor <iant@google.com>
43 * lower-subreg.c (simplify_gen_subreg_concatn): If we ask for the
44 high part of a paradoxical subreg, return a constant zero.
46 2007-02-01 Zdenek Dvorak <dvorakz@suse.cz>
48 * toplev.c (lang_dependent_init): Call init_set_costs.
49 * loop-init.c (loop_optimizer_init): Do not call init_set_costs.
51 2007-02-01 Richard Guenther <rguenther@suse.de>
53 PR middle-end/30656
54 * fold-const.c (fold_negate_expr): Allow negating a
55 constant if overflow does not change.
57 2007-02-01 Ralf Wildenhues <Ralf.Wildenhues@gmx.de>
59 * doc/c-tree.texi (Expression trees): Improve markup.
60 * doc/tm.texi (Register Classes, Addressing Modes)
61 (Floating Point): Fix spacing after abbreviations. Fix some
62 typos.
64 2007-02-01 Ben Elliston <bje@au.ibm.com>
66 * doc/invoke.texi: Replace "bugfix" with "bug fix" throughout.
67 * doc/contrib.texi: Likewise.
68 * doc/install.texi: Likewise.
70 2007-01-31 Richard Henderson <rth@redhat.com>
71 Ian Lance Taylor <iant@google.com>
73 * lower-subreg.c: New file.
74 * rtl.def (CONCATN): Define.
75 * passes.c (init_optimization_passes): Add pass_lower_subreg and
76 pass_lower_subreg2.
77 * emit-rtl.c (update_reg_offset): New static function, broken out
78 of gen_rtx_REG_offset.
79 (gen_rtx_REG_offset): Call update_reg_offset.
80 (gen_reg_rtx_offset): New function.
81 * regclass.c: Revert patch of 2006-03-05, restoring
82 reg_scan_update.
83 (clear_reg_info_regno): New function.
84 * dwarf2out.c (concatn_loc_descriptor): New static function.
85 (loc_descriptor): Handle CONCATN.
86 * common.opt (fsplit_wide_types): New option.
87 * opts.c (decode_options): Set flag_split_wide_types when
88 optimizing.
89 * timevar.def (TV_LOWER_SUBREG): Define.
90 * rtl.h (gen_reg_rtx_offset): Declare.
91 (reg_scan_update): Declare.
92 * regs.h (clear_reg_info_regno): Declare.
93 * tree-pass.h (pass_lower_subreg): Declare.
94 (pass_lower_subreg2): Declare.
95 * doc/invoke.texi (Option Summary): List -fno-split-wide-types.
96 (Optimize Options): Add -fsplit-wide-types to -O1 list. Document
97 -fsplit-wide-types.
98 * doc/rtl.texi (Regs and Memory): Document concat and concatn.
99 * Makefile.in (OBJS-common): Add lower-subreg.o.
100 (lower-subreg.o): New target.
102 2007-01-31 Kazu Hirata <kazu@codesourcery.com>
104 * config/sh/sh.h (HAVE_SECONDARY_RELOADS): Remove.
106 2007-01-31 Anatoly Sokolov <aesok@post.ru>
108 PR target/19087
109 * config/avr/avr.c (DWARF2_ADDR_SIZE): Define.
111 2007-01-31 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
113 PR middle-end/29335
114 * builtins.c (fold_builtin_sqrt): Use MPFR for constant args.
116 2007-01-31 Zdenek Dvorak <dvorakz@suse.cz>
118 * cfgloop.h: Include vec-prim.h.
119 (enum li_flags): Remove LI_ONLY_OLD.
120 (loop_iterator): Changed.
121 (fel_next, fel_init): Iterate over loop tree.
122 (FOR_EACH_LOOP_BREAK): New macro.
123 * loop-unswitch.c (unswitch_loops): Do not pass LI_ONLY_OLD to
124 FOR_EACH_LOOP.
125 * tree-ssa-loop-unswitch.c (tree_ssa_unswitch_loops): Ditto.
126 * modulo-sched.c (sms_schedule): Ditto.
127 * tree-vectorizer.c (vectorize_loops): Ditto.
128 * doc/loop.texi: Update information on loop numbering and behavior of
129 FOR_EACH_LOOP wrto new loops.
130 * tree-scalar-evolution.c (compute_overall_effect_of_inner_loop,
131 add_to_evolution_1): Test nestedness of loops instead of comparing
132 their numbers.
133 * tree-chrec.c (chrec_fold_plus_poly_poly,
134 chrec_fold_multiply_poly_poly, chrec_evaluate,
135 hide_evolution_in_other_loops_than_loop, chrec_component_in_loop_num,
136 reset_evolution_in_loop): Ditto.
137 * Makefile.in (CFGLOOP_H): Add vecprim.h dependency.
139 2007-01-31 Dirk Mueller <dmueller@suse.de>
141 * c-common.c (warn_about_parentheses): Separate warning about
142 un-parenthized sequence of comparison operators from the one
143 which is supposed to warn about x <= y <= z.
145 2007-01-31 Uros Bizjak <ubizjak@gmail.com>
147 * optabs.h (enum optab_index): Add new OTI_isinf.
148 (isinf_optab): Define corresponding macro.
149 * optabs.c (init_optabs): Initialize isinf_optab.
150 * genopinit.c (optabs): Implement isinf_optab using isinf?f2
151 patterns.
152 * builtins.c (mathfn_built_in): Handle BUILT_IN_ISINF{,F,L}.
153 (expand_builtin_interclass_mathfn): Expand BUILT_IN_ISINF{,F,L}
154 using isinf_optab.
155 (expand_builtin): Expand BUILT_IN_ISINF{,F,L} using
156 expand_builtin_interclass_mathfn.
157 * reg_stack.c (subst_stack_regs_pat): Handle UNSPEC_FXAM.
158 * config/i386/i386.md (UNSPEC_FXAM): New constant.
159 (fxam<mode>2_i387): New insn pattern.
160 (isinf<mode>2) New expander to implement isinf, isinff and isinfl
161 built-in functions as x87 inline asm.
163 2007-01-31 Kazu Hirata <kazu@codesourcery.com>
165 * gcc/config/arm/unwind-arm.h (_sleb128_t, _uleb128_t): New.
167 2007-01-30 Eric Christopher <echristo@apple.com>
169 * config.gcc: Add geode.
171 2007-01-31 Kazu Hirata <kazu@codesourcery.com>
173 * cgraphunit.c, config/arm/arm.c, config/m68k/m68k.c,
174 ipa-inline.c, tree-profile.c, tree-ssa-live.c,
175 tree-ssa-math-opts.c, tree-ssanames.c, tree-vect-analyze.c,
176 value-prof.c: Fix comment typos.
178 2007-01-30 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
180 PR c++/24745
181 * doc/invoke.texi (Wpointer-arith): Document warning.
183 2007-01-30 Janis Johnson
185 * doc/extend.texi (Decimal Floating Types): Remove decfloat.h from
186 the list of discrepancies from the draft TR.
188 2007-01-30 Dirk Mueller <dmueller@suse.de>
190 PR c++/30601
191 * doc/invoke.texi (-Wreturn-type): Update description to
192 match new behavior.
194 2007-01-30 Richard Sandiford <richard@codesourcery.com>
196 * cfgrtl.c (try_redirect_by_replacing_jump): Check only_sets_cc0_p.
198 2007-01-30 Uros Bizjak <ubizjak@gmail.com>
200 * builtins.c (expand_builtin_int_interclass_roundingfn): New function
201 to handle optabs that operate on floating point input argument and
202 output to integer output.
203 (expand_builtin_mathfn) [BUILT_IN_ILOGB]: Move from here ...
204 (expand_builtin_interclass_mathfn) [BUILT_IN_ILOGB]: ... to here.
205 (expand_builtin): Expand BUILT_IN_ILOGB{,F,L} using
206 expand_builtin_interclass_mathfn ().
207 * config/i386/i386.md (fxtractxf3_i387): Rename from *fxtractxf3_i387.
208 (ilogbsi2): Remove.
209 (ilogbxf2, ilogb<mode>2): New expanders to implement ilogb, ilogbf and
210 ilogbl built-in functions as x87 intrinsics.
212 2007-01-30 Richard Guenther <rguenther@suse.de>
214 PR middle-end/27657
215 * dwarf2out.c (reference_to_unused): Query varpool if the
216 variable was output.
218 2007-01-30 Richard Guenther <rguenther@suse.de>
220 PR middle-end/30313
221 * passes.c (execute_one_pass): Reset in_gimple_form to not
222 confuse non-unit-at-a-time mode.
224 2007-01-29 Roger Sayle <roger@eyesopen.com>
225 Richard Guenther <rguenther@suse.de>
227 * fold-const.c (round_up): Make HIGH an unsigned HOST_WIDE_INT to
228 avoid undefined behaviour on overflow. Use force_fit_type_double
229 to construct the constant with the specified TREE_OVERFLOW.
231 2007-01-29 Janis Johnson <janis187@us.ibm.com>
233 * config/dfp-bit.c: Add parameterized support for fp exceptions.
234 * config/dfp-bit.h: Ditto.
236 2007-01-29 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
238 * c-decl.c (pop_scope): Replace warnings with call to
239 warn_for_unused_label.
240 * c-common.h (warn_for_unused_label): Declare.
241 * c-common.c (warn_for_unused_label): Define.
243 2007-01-29 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
245 * tree-optimize.c (update_inlined_to_pointers): Delete unused
246 function.
248 2007-01-29 Janis Johnson <janis187@us.ibm.com>
250 * Makefile.in (USER_H): Remove decfloat.h.
251 * ginclude/decfloat.h: Delete, moving contents to ...
252 * ginclude/float.h: Add support for decimal floating point,
253 guarded by __STDC_WANT_DEC_FP__.
255 2007-01-29 Mike Stump <mrs@apple.com>
257 * doc/gccint.texi (Top): Rename Loop Representation to Loop
258 Analysis and Representation to resolve case insensitive conflict.
259 * doc/loop.texi (Loop Analysis and Representation): Likewise.
261 2007-01-28 Daniel Berlin <dberlin@dberlin.org>
263 * tree.h (struct tree_memory_tag): Add aliases member.
264 (MTAG_ALIASES): New macro.
265 * tree-ssa-alias.c (alias_bitmap_obstack): New variable.
266 (add_may_alias): Remove pointer-set. Update for may_aliases being
267 a bitmap.
268 (mark_aliases_call_clobbered): Update for may_aliases being a
269 bitmap.
270 (compute_tag_properties): Ditto.
271 (create_partition_for): Ditto.
272 (compute_memory_partitions): Ditto.
273 (dump_may_aliases_for): Ditto.
274 (is_aliased_with): Ditto.
275 (add_may_alias_for_new_tag): Ditto.
276 (rewrite_alias_set_for): Rewrite for may_aliases being a bitmap.
277 (compute_is_aliased): New function.
278 (compute_may_aliases): Call compute_is_aliased).
279 (init_alias_info): Initialize alias_bitmap_obstack.
280 (union_alias_set_into): New function.
281 (compute_flow_sensitive_aliasing): Use union_aliases_into.
282 (have_common_aliases_p): Rewrite to take two bitmaps and use
283 intersection.
284 (compute_flow_insensitive_aliasing): Stop using pointer-sets.
285 Update for bitmaps.
286 (finalize_ref_all_pointers): Update for add_may_alias changes.
287 (new_type_alias): Ditto.
288 * tree-flow-inline.h (may_aliases): Return a bitmap.
289 * tree-dfa.c (dump_variable): Check for MTAG_P'ness.
290 * tree-ssa.c (verify_flow_insensitive_alias_info): Update for
291 may_aliases being a bitmap.
292 * tree-flow.h (struct var_ann_d): Remove may_aliases member.
293 may_aliases now returns a bitmap.
294 * tree-ssa-structalias.c (merge_smts_into): Update for may_aliases
295 being a bitmap.
296 * tree-ssa-operands.c (add_virtual_operand): Update for
297 may_aliases being a bitmap.
299 2007-01-29 Daniel Berlin <dberlin@dberlin.org>
301 PR tree-optimization/30630
302 * tree-ssa-structalias.c (do_complex_constraint): Mark correct
303 variable as changed.
305 2007-01-29 Simon Martin <simartin@users.sourceforge.net>
307 PR c++/28266
308 * gimplify.c (gimplify_target_expr): Make sure that the TARGET_EXPR is
309 expanded only once even if an error occurs.
311 2007-01-29 Ben Elliston <bje@au.ibm.com>
313 * gcov-io.h (__gcov_indirect_call_profiler): Declare.
314 (__gcov_average_profiler): Likewise.
315 (__gcov_ior_profiler): Likewise.
316 (__gcov_merge_ior): Likewise.
318 2007-01-28 Jan Hubicka <jh@suse.cz>
320 * builtins.c (expand_builtin_memset): Fix typo in my last patch.
321 * value-prof.c (stringop_block_profile): Likewise.
323 2007-01-28 Jan Hubicka <jh@suse.cz>
325 * expr.c (emit_block_move_via_movmem, emit_block_move_via_libcall): Add
326 variant handling histograms; add wrapper.
327 (clear_storage_via_libcall): Export.
328 (emit_block_move_hints): Break out from ...; add histograms.
329 (emit_block_move): ... this one.
330 (clear_storage_hints): Break out from ...; add histograms.
331 (clear_storage): ... this one.
332 (set_storage_via_memset): Handle histogram.
333 * expr.h (emit_block_move_via_libcall, emit_block_move_hints): Declare.
334 (clear_storage_hints, clear_storage_via_libcall): Declare.
335 (set_storage_via_setmem): Update prototype.
336 * doc/md.texi (movmem, setmem): Document new arguments.
338 * value-prof.c (dump_histogram_value, tree_find_values_to_profile): Add
339 new histograms.
340 (stringop_block_profile): New global function.
341 (tree_stringops_values_to_profile): Profile block size and alignment.
342 * value-prof.h (enum hist_type): add HIST_TYPE_AVERAGE and
343 HIST_TYPE_IOR.
344 (struct profile_hooks): Add gen_average_profiler and gen_ior_profiler.
345 (stringop_block_profile): Declare.
346 * builtins.c: Include value-prof.h.
347 (expand_builtin_memcpy, expand_builtin_memset): Pass block profile.
348 * gcov-ui.h (GCOV_COUNTER_NAMES): Add new counter.
349 (GCOV_COUNTER_AVERAGE, GCOV_COUNTER_IOR): New constants.
350 (GCOV_COUNTERS, GCOV_LAST_VALUE_COUNTER): Update.
351 * profile.c (instrument_values): Add new counters.
352 * cfgexpand.c (expand_gimple_basic_block): Propagate histograms to
353 calls.
354 * tree-profile.c (tree_average_profiler_fn, tree_ior_profiler_fn): New.
355 (tree_init_edge_profiler): Build new profilers.
356 (tree_gen_average_profiler, tree_gen_ior_profiler): New.
357 (pass_tree_profile): Add dump.
358 (tree_profile_hooks): Update.
359 * Makefile.in (LIBGCOV): Add new constants.
360 * libgcov.c (__gcov_merge_ior, __gcov_average_profiler,
361 __gcov_ior_profiler): New.
362 * i386.md (movmem/setmem expanders): Add new optional arguments.
364 2007-01-28 David Edelsohn <edelsohn@gnu.org>
366 * doc/md.texi (Standard Pattern Names): Document blockage pattern.
368 2007-01-28 Zdenek Dvorak <dvorakz@suse.cz>
370 * tree-ssa-loop-unswitch.c: Include tree-inline.h.
371 (tree_unswitch_single_loop): Pass eni_size_weights to
372 tree_num_loop_insns.
373 * tree-ssa-loop-manip.c: Include tree-inline.h.
374 (can_unroll_loop_p): Pass eni_size_weights to
375 tree_num_loop_insns.
376 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p):
377 Pass eni_size_weights to estimate_num_insns.
378 * tree.h (init_inline_once): Export.
379 * toplev.c (backend_init): Call init_inline_once.
380 * cgraphunit.c (cgraph_process_new_functions,
381 cgraph_analyze_function): Pass eni_inlining_weights to
382 estimate_num_insns.
383 * ipa-inline.c (compute_inline_parameters): Ditto.
384 * tree-ssa-loop-ivcanon.c (tree_num_loop_insns): Pass weights
385 to estimate_num_insns.
386 (try_unroll_loop_completely): Pass eni_size_weights to
387 tree_num_loop_insns.
388 * tree-eh.c (decide_copy_try_finally): Pass eni_size_weights
389 ot estimate_num_insns.
390 * tree-ssa-loop-prefetch.c: Include tree-inline.h.
391 (loop_prefetch_arrays): Pass eni_time_weights to tree_num_loop_insns.
392 * tree-inline.c (eni_inlining_weights, eni_size_weights,
393 eni_time_weights): New variables.
394 (init_inline_once): Initialize them.
395 (struct eni_data): Mew.
396 (estimate_num_insns_1, estimate_num_insns): Use weights.
397 * tree-inline.h (struct eni_weights_d): New.
398 (eni_inlining_weights, eni_size_weights, eni_time_weights): Declare.
399 (estimate_num_insns): Declaration changed.
400 * cfgloop.h (tree_num_loop_insns): Declaration changed.
401 * Makefile.in (tree-ssa-loop-unswitch.o, tree-ssa-loop-prefetch.o,
402 tree-ssa-loop-manip.o): Add TREE_INLINE_H dependency.
404 2007-01-28 Zdenek Dvorak <dvorakz@suse.cz>
406 * tree-data-ref.c (conflict_fn): Assert that the number of affine
407 relations in the conflict function is valid.
409 2007-01-27 Ian Lance Taylor <iant@google.com>
411 * common.opt: Add fstrict-overflow.
412 * opts.c (decode_options): Set flag_strict_overflow if -O2.
413 * flags.h (TYPE_OVERFLOW_WRAPS): Define.
414 (TYPE_OVERFLOW_UNDEFINED): Define.
415 (TYPE_OVERFLOW_TRAPS): Define. This replaces TYPE_TRAP_SIGNED.
416 Replace all uses.
417 * tree.h (TYPE_TRAP_SIGNED): Don't define.
418 * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_UNDEFINED.
419 (fold_negate_expr): Likewise.
420 (make_range): Likewise.
421 (extract_muldiv_1): Likewise.
422 (maybe_canonicalize_comparison): Likewise.
423 (fold_comparison): Likewise.
424 (fold_binary): Likewise.
425 (tree_expr_nonnegative_p): Likewise.
426 (tree_expr_nonzero_p): Likewise.
427 * tree-vrp.c (compare_values): Likewise.
428 (extract_range_from_binary_expr): Likewise.
429 (extract_range_from_unary_expr): Likewise.
430 * tree-ssa-loop-niter.c (infer_loop_bounds_from_signedness):
431 Likewise.
432 (nowrap_type_p): Likewise.
433 * tree-scalar-evolution.c (simple_iv): Likewise.
434 * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_WRAPS.
435 (build_range_check): Likewise.
436 (extract_muldiv_1): Likewise.
437 (fold_comparison): Likewise.
438 * tree-vrp.c (vrp_int_const_binop): Likewise.
439 (extract_range_from_unary_expr): Likewise.
440 * convert.c (convert_to_integer): Likewise.
441 * fold-const.c (fold_negate_expr): Use TYPE_OVERFLOW_TRAPS.
442 (fold_comparison): Likewise.
443 (fold_binary): Likewise.
444 * optabs.c (optab_for_tree_code): Likewise.
445 * tree-vectorizer.c (vect_is_simple_reduction): Likewise.
446 * simplify-rtx.c (simplify_const_relational_operation): Check
447 flag_strict_overflow and flag_trapv.
448 (simplify_const_relational_operation): Likewise.
449 * doc/invoke.texi (Option Summary): Mention -fstrict-overflow.
450 (Optimize Options): Add -fstrict-overflow to -O2 list. Document
451 -fstrict-overflow.
453 2007-01-27 Roger Sayle <roger@eyesopen.com>
455 * tree.c (tree_fold_gcd): Delete.
456 * tree.h (tree_fold_gcd): Remove prototype.
457 * tree-data-ref.c (tree_fold_divides_p): Don't use tree_fold_gcd to
458 test whether one constant integer is a multiple of another. Instead
459 call int_const_binop with TRUNC_MOD_EXPR and test for a zero result.
460 * fold-const.c (multiple_of_p): We've determined both TOP and
461 BOTTOM are integer constants so we can call int_const_binop directly
462 instead of the more generic const_binop.
464 2007-01-27 Roger Sayle <roger@eyesopen.com>
466 * fold-const.c (size_binop): In the fast-paths for X+0, 0+X, X-0 and
467 1*X check that the constant hasn't overflowed, to preserve the
468 TREE_OVERFLOW bit.
469 (round_up): Provide an efficient implementation when rouding-up an
470 INTEGER_CST to a power-of-two.
472 2007-01-28 Ralf Wildenhues <Ralf.Wildenhues@gmx.de>
474 * doc/sourcebuild.texi: Add comma for clarity.
475 * doc/extend.texi: Fix some typos.
476 * doc/passes.texi: Likewise.
477 * doc/cppinternals.texi: Likewise.
478 * doc/c-tree.texi: Likewise.
479 * doc/tree-ssa.texi: Likewise.
480 * doc/install.texi: Likewise.
482 2007-01-27 Jan Hubicka <jh@suse.cz>
484 * tree-sra.c (sra_walk_function): Don't rely on aliases being build.
485 (pass_sra): Do not require alias information.
486 * passes.c (init_optimization_passes): Add SRA
488 2007-01-27 Steven Bosscher <steven@gcc.gnu.org>
490 * tracer.c (rest_of_handle_tracer): We already cleaned
491 up the CFG in tracer() so don't do it here again.
492 * cfgcleanup.c (rest_of_handle_jump2): Don't repeat
493 cleanup_cfg here, either. And don't call renumber_insns.
495 * cfgrtl.c (rtl_verify_flow_info_1): Don't verify that BB_END
496 and BB_HEAD are in the insn stream here. Instead make sure
497 that BB_INSN is valid on all insns. Also, do check here that
498 there are no pending branch predictions...
499 (rtl_verify_flow_info): ...instead of doing it here. Checks
500 for BB_END and BB_HEAD moved from rtl_verify_flow_info_1 to
501 here.
503 2007-01-26 Roger Sayle <roger@eyesopen.com>
505 * config/i386/i386.c (ix86_swap_binary_operands_p): New helper
506 function to simplify/factorize operand order canonicalization.
507 (ix86_fixup_binary_operands): Reorganize using the above function.
508 (ix86_binary_operator_ok): Likewise.
510 2007-01-27 Jakub Jelinek <jakub@redhat.com>
512 * genattrtab.c (struct attr_value_list, insn_code_values): Move to
513 file scope from optimize_attrs.
514 (simplify_test_exp): If insn_code_values is not NULL, use it to speed
515 up search.
516 (optimize_attrs): Clear insn_code_values after freeing it.
518 2007-01-26 Zdenek Dvorak <dvorakz@suse.cz>
520 * tree-ssa-address.c (create_mem_ref): Remove ", bsi" from
521 a parts.base assignment.
523 2007-01-26 Zdenek Dvorak <dvorakz@suse.cz>
525 * tree-data-ref.c (dump_subscript): Use dump_conflict_function.
526 (compute_subscript_distance, initialize_data_dependence_relation,
527 finalize_ddr_dependent, analyze_ziv_subscript,
528 analyze_siv_subscript_cst_affine,
529 compute_overlap_steps_for_affine_univar,
530 compute_overlap_steps_for_affine_1_2, analyze_subscript_affine_affine,
531 analyze_siv_subscript, analyze_miv_subscript,
532 analyze_overlapping_iterations, subscript_dependence_tester_1,
533 compute_self_dependence, free_dependence_relation): Work
534 with affine_fn instead of chrecs.
535 (dump_affine_function, dump_conflict_function, affine_function_equal_p,
536 common_affine_function, affine_function_base,
537 affine_function_constant_p, affine_fn_op, affine_fn_plus,
538 affine_fn_minus, affine_fn_free, conflict_fn_not_known,
539 conflict_fn_no_dependence, free_conflict_function, free_subscripts,
540 conflict_fn, affine_fn_cst, affine_fn_univar): New functions.
541 (all_chrecs_equal_p): Removed.
542 * tree-data-ref.h (affine_fn, conflict_function): New types.
543 (struct subscript): Change type of conflicting_iterations_in_a
544 and conflicting_iterations_in_b.
546 2007-01-26 Steve Ellcey <sje@cup.hp.com>
548 PR other/30182
549 * config/pa/pa.h (TARGET_HPUX_11): New.
550 * config/pa/pa-hpux11.h (TARGET_HPUX_11): New.
551 * config/pa/pa.c (pa_init_builtins): Use TARGET_HPUX_11.
553 2007-01-26 Daniel Berlin <dberlin@dberlin.org>
554 Richard Guenther <rguenther@suse.de>
556 * tree-ssa-structalias.c (solve_graph): Handle case
557 we merged the variable to another.
559 2007-01-25 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
561 * builtins.c (fold_builtin_1): Treat ccos and ccosh as 'even'
562 functions.
564 * fold-const.c (negate_mathfn_p): Treat casin, casinh, catan,
565 catanh, cproj, csin, csinh, ctan and ctanh as 'odd' functions.
567 2007-01-25 DJ Delorie <dj@redhat.com>
569 * config/m32c/m32c.c (m32c_cannot_change_mode_class): We don't
570 allow changes to modes which don't fit in those registers.
572 * reload1.c (choose_reload_regs): Check for invalid subregs before
573 computing their locations, not after.
575 2007-01-25 Geoffrey Keating <geoffk@apple.com>
577 PR 25127
578 * config/rs6000/rs6000.c (first_altivec_reg_to_save): On Darwin,
579 save Altivec registers in an eh_return function.
580 (compute_vrsave_mask): Likewise.
581 (rs6000_stack_info): Correct AIX/Darwin stack alignment computation
582 for saving Altivec registers.
583 (rs6000_emit_prologue): Don't allocate stack twice in
584 eh_return function. Correct expected value of altivec_save_offset
585 when using save_world. Describe save of R0 to stack when using
586 save_world. Describe stack pointer adjustment when using
587 save_world. Remove duplicated eh_return parameter register saving.
588 Update sp_offset variable after save_world.
589 * config/rs6000/t-darwin (LIB2FUNCS_STATIC_EXTRA): Remove
590 darwin-world.asm.
591 (LIB2FUNCS_EXTRA): Add darwin-world.asm.
592 * config/rs6000/darwin.h (SUBTARGET_OVERRIDE_OPTIONS): -m64
593 implies Altivec.
595 2007-01-25 Steve Ellcey <sje@cup.hp.com>
597 * config.gcc (ia64*-*-hpux*): Make posix threads the default.
599 2007-01-25 Steve Ellcey <sje@cup.hp.com>
601 PR other/30182
602 * config/pa/pa.c (pa_init_builtins): Set asm names for finite routines.
603 * config/ia64/ia64.c (ia64_init_builtins): Ditto.
605 2007-01-25 Richard Guenther <rguenther@suse.de>
607 * doc/invoke.texi (-Wcoverage-mismatch): Document.
608 * common.opt (-Wcoverage-mismatch): New warning option.
609 * coverage.c (get_coverage_counts): Ignore coverage mismatch
610 if -Wcoverage-mismatch is given.
612 2007-01-25 Razya Ladelsky <razya@il.ibm.com>
614 * ipa-cp.c (ipcp_insert_stage, ipcp_driver): Support for SSA.
615 (ipcp_driver): Change to static definition.
616 Add dumping of the ifunctions.
617 (constant_val_insert): Remove unused parameter. Support for SSA.
618 (ipcp_propagate_const): Support for SSA.
619 (ipcp_profile_bb_print): Print only analyzed nodes.
620 (ipcp_replace_map_create): Remove support for Fortran constant
621 for now.
622 * ipa-prop.c (ipa_method_modify_stmt,
623 ipa_callsite_compute_param): Support for SSA.
624 * ipa-prop.h (ipcp_driver): Remove declaration.
625 (IS_VALID_TREE_MAP_INDEX): Add define.
627 2007-01-24 Geoffrey Keating <geoffk@apple.com>
629 * unwind-dw2.c (execute_stack_op): Handle DW_OP_swap.
631 2007-01-24 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
633 PR middle-end/30447
634 * builtins.c (fold_builtin_cabs): Use MPFR to evaluate a
635 constant argument to cabs and do it without checking for
636 -funsafe-math-optimizations.
638 2007-01-24 Douglas Gregor <dgregor@osl.iu.edu>
640 * c-common.h (RID_FIRST_CXX0X): New.
641 (RID_LAST_CXX0X): New.
642 * c-opts.c (c_common_handle_option): -Wc++0x-compat is triggered
643 by -Wall.
644 * c.opt (Wc++0x-compat): New.
645 * doc/invoke.texi (-Wc++0x-compat): Document.
647 2007-01-24 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
649 * builtins.c (fold_builtin_carg): New.
650 (fold_builtin_1): Use it.
652 2007-01-24 Jan Hubicka <jh@suse.cz>
654 * ipa-inline.c (cgraph_decide_inlining): Initialize initial_insns.
656 * ipa-inline.c (initial_insns, max_insns): Delete.
657 (compute_max_insns): New function.
658 (cgraph_decide_inlining_of_small_function): Use it; take minimal amount
659 of insns as base for code growth.
660 (cgraph_decide_inlining): Make initial_insns local; do not compute
661 max_insns.
662 * params.def (PARAM_INLINE_UNIT_GROWTH): Set to 60.
663 * doc/invoke.texi (inline-unit-growth): Update docs.
665 2007-01-24 Jakub Jelinek <jakub@redhat.com>
667 * config/i386/i386.h (x86_cmpxchg16b): Remove const.
668 (TARGET_CMPXCHG16B): Define to x86_cmpxchg16b.
669 * config/i386/i386.c (x86_cmpxchg16b): Remove const.
670 (override_options): Add PTA_CX16 flag. Set x86_cmpxchg16b
671 for CPUs that have PTA_CX16 set.
673 PR middle-end/27416
674 * gimplify.c (omp_check_private): New function.
675 (gimplify_scan_omp_clauses): Use it for
676 firstprivate/lastprivate/reduction.
678 PR middle-end/30494
679 * gimplify.c (omp_add_variable): Don't call omp_notice_variable
680 on TYPE_SIZE_UNIT for GOVD_LOCAL VLAs.
682 PR middle-end/30421
683 * omp-low.c (lower_omp_for_lastprivate): Add dlist argument.
684 If lower_lastprivate_clauses emits some statements, append them
685 to dlist rather than body_p and to body_p append an initializer.
686 (lower_omp_for): Adjust caller.
688 2007-01-24 Steve Ellcey <sje@cup.hp.com>
690 * target.h (globalize_decl_name): New.
691 * target-def.h (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
692 * output.h (default_globalize_decl_name): New.
693 * varasm.c (asm_output_bss): Use globalize_decl_name instead of
694 globalize_label.
695 (globalize_decl): Ditto.
696 (default_globalize_decl_name): New.
697 * config/ia64/ia64.c (ia64_globalize_decl_name): New.
698 (ia64_handle_version_id_attribute): New.
699 (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
700 (ia64_asm_output_external): Use globalize_decl_name instead
701 of globalize_label.
702 * doc/extend.texi (version_id): New pragma.
703 * doc/tm.texi (ARGET_ASM_GLOBALIZE_DECL_NAME): New target hook.
705 2007-01-24 Andreas Krebbel <krebbel1@de.ibm.com>
707 * unwind-dw2-fde.c (get_cie_encoding): Replaced _Unwind_Word with
708 _uleb128_t and _Unwind_SWord with _sleb128_t.
709 * unwind-dw2.c (extract_cie_info, execute_stack_op, execute_cfa_program,
710 uw_frame_state_for, uw_update_context_1): Likewise.
711 * unwind-c.c (parse_lsda_header, PERSONALITY_FUNCTION): Likewise.
712 * unwind-pe.h (read_uleb128, read_sleb128,
713 read_encoded_value_with_base): Likewise.
714 * unwind-generic.h: Define _sleb128_t and _uleb128_t types.
716 2007-01-24 Richard Guenther <rguenther@suse.de>
718 * builtins.c (expand_builtin_cexpi): Get the fndecl
719 for cexp in the correct way.
721 2007-01-24 Jan Hubicka <jh@suse.cz>
723 * tree-ssa-dce.c (eliminate_unnecesary_stmts): Remove dead LHS of calls.
725 2007-01-24 Andreas Krebbel <krebbel1@de.ibm.com>
727 * c-cppbuiltin.c (builtin_define_type_sizeof): New function.
728 (c_cpp_builtins): New builtin macros: __SIZEOF_INT__, __SIZEOF_LONG__,
729 __SIZEOF_LONG_LONG__, __SIZEOF_SHORT__, __SIZEOF_POINTER__,
730 __SIZEOF_FLOAT__, __SIZEOF_DOUBLE__, __SIZEOF_LONG_DOUBLE__,
731 __SIZEOF_SIZE_T__, __SIZEOF_WCHAR_T__, __SIZEOF_WINT_T__ and
732 __SIZEOF_PTRDIFF_T__.
733 * doc/cpp.texi: Documentation for the new builtin macros added.
735 2007-01-24 Uros Bizjak <ubizjak@gmail.com>
737 * config/i386/i386.md (tanxf2, tan<mode>2, atan<mode>2, log<mode>2,
738 log10<mode>2, log2<mode>2, expxf2, exp10xf2, exp2xf2): Use op2
739 instead of operands[2] to avoid access past the end of array.
741 2007-01-24 Richard Sandiford <richard@codesourcery.com>
743 * reload1.c (emit_reload_insns): Pass the reload register
744 for a non-spill output reload through forget_old_reloads_1.
746 2007-01-23 Joseph Myers <joseph@codesourcery.com>
748 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Define
749 _SOFT_DOUBLE if doubles use software floating-point.
750 * config/rs6000/libgcc-ppc-glibc.ver: Export additional long
751 double functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
752 * config/rs6000/darwin-ldouble.c: Also compile functions for
753 hard-float without FPRs. Use fmsub function for all __NO_FPRS__
754 cases. Compile extra functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
755 * config/rs6000/linuxspe.h (SUBSUBTARGET_OVERRIDE_OPTIONS): Remove
756 commented-out long double override.
757 (CPP_LONGDOUBLE_DEFAULT_SPEC): Likewise.
758 * config/rs6000/eabispe.h: Likewise.
759 * config/rs6000/rs6000.c (rs6000_override_options): Don't override
760 long double for non-SPE.
761 (rs6000_handle_option): Likewise.
762 (invalid_e500_subreg): Disallow more subregs involding DImode,
763 DFmode, TImode or TFmode.
764 (rs6000_legitimate_offset_address_p): Check TFmode offsets for
765 E500 double.
766 (legitimate_lo_sum_address_p): Also check for TFmode for E500
767 double.
768 (rs6000_legitimize_address): Also handle TFmode for E500 double.
769 (rs6000_legitimize_reload_address): Also handle TFmode for E500
770 double.
771 (rs6000_legitimate_address): Also check for TFmode for E500
772 double.
773 (rs6000_emit_move): Use DFmode subregs of TFmode for E500 double.
774 (spe_build_register_parallel): Handle TFmode and TCmode.
775 (rs6000_spe_function_arg): Handle TFmode and TCmode for E500
776 double.
777 (function_arg): Handle TFmode and TCmode for E500 double.
778 (rs6000_init_libfuncs): Initialize extra libfuncs for soft double
779 in general.
780 (print_operand): Handle TFmode and TImode for %y.
781 (rs6000_generate_compare): Handle TFmode comparisons for E500
782 double.
783 (spe_func_has_64bit_regs_p): Check for TFmode for E500 double.
784 (rs6000_function_value): Handle TFmode and TCmode for E500 double.
785 (rs6000_libcall_value): Handle TFmode and TCmode for E500 double.
786 * config/rs6000/rs6000.h (CANNOT_CHANGE_MODE_CLASS): Check for
787 TFmode for E500 double.
788 * config/rs6000/rs6000.md (FP): Allow TF for E500 double.
789 (floatsidf2): Enable for E500 double.
790 (movtf_softfloat): Use rs6000_nonimmediate_operand.
791 (extenddftf2): Change to extenddftf2_fprs.
792 (extenddftf2): Call gen_spe_extenddftf2 or gen_extenddftf2_fprs
793 depending on TARGET_E500_DOUBLE.
794 (extendsftf2): Enable for E500 double.
795 (trunctfdf2): Enable for E500 double.
796 (trunctfsf2): Change to trunctfsf2_fprs.
797 (trunctfsf2): Call gen_spe_trunctfsf2 or gen_trunctfsf2_fprs
798 depending on TARGET_E500_DOUBLE.
799 (floatsitf2): Enable for E500 double.
800 (fix_trunctfsi2): Change to fix_trunctfsi2_fprs.
801 (fix_trunctfsi2): Call gen_spe_fix_trunctfsi2 or
802 gen_fix_trunctfsi2_fprs depending on TARGET_E500_DOUBLE.
803 (negtf2): Change to negtf2_internal.
804 (negtf2): New expander.
805 (abstf2): Enable for E500 double. Call gen_spe_abstf2_tst,
806 gen_spe_abstf2_cmp or gen_abstf2_internal depending on
807 TARGET_E500_DOUBLE and flag_unsafe_math_optimizations.
808 (movdi_internal32): Use rs6000_nonimmediate_operand.
809 (unnamed splitter): Likewise.
810 * config/rs6000/spe.md (CMPTFEQ_GPR, TSTTFEQ_GPR, CMPTFGT_GPR,
811 TSTTFGT_GPR, CMPTFLT_GPR, TSTTFLT_GPR): New unspecs.
812 (SPE64TF, DITI): New mode macros.
813 (frob_df_di): Change to frob_<SPE64:mode>_<DITI:mode>; allow more
814 modes.
815 (frob_tf_ti): New.
816 (frob_<mode>_di_2): New.
817 (frob_tf_di_8_2): New.
818 (frob_di_df): Change to frob_di_<mode>; allow more modes.
819 (frob_ti_tf): New.
820 (frob_di_df_2): Change to frob_<DITI:mode>_<SPE64:mode>_2; allow
821 more modes.
822 (frob_ti_<mode>_8_2): New.
823 (frob_ti_tf_2): New.
824 (mov_si<mode>_e500_subreg0, mov_si<mode>_e500_subreg0_2,
825 mov_si<mode>_e500_subreg4, mov_si<mode>_e500_subreg4_2): Allow
826 TFmode.
827 (mov_sitf_e500_subreg8, mov_sitf_e500_subreg8_2,
828 mov_sitf_e500_subreg12, mov_sitf_e500_subreg12_2): New.
829 (spe_trunctfdf2_internal1, spe_trunctfsf2, spe_extenddftf2,
830 spe_fix_trunctfsi2, spe_fix_trunctfsi2_internal,
831 spe_negtf2_internal, spe_abstf2_cmp, spe_abstf2_tst): New.
832 (cmptfeq_gpr, tsttfeq_gpr, cmptfgt_gpr, tsttfgt_gpr, cmptflt_gpr,
833 tsttflt_gp): New.
835 2007-01-23 Ian Lance Taylor <iant@google.com>
837 * Makefile.in (OBJS-common): Reformat, alphabetize, but put
838 insn-*.o first.
839 (OBJS-archive): Reformat, alphabetize.
840 (OBJS): Change out_object_file to OBJS-md.
842 2007-01-23 Uros Bizjak <ubizjak@gmail.com>
844 * config/i386/i386.md: Use REG_P, MEM_P, CONST_INT_P, LABEL_P,
845 JUMP_P and CALL_P predicates where applicable.
846 * config/i386/i386.c: Ditto.
847 * config/i386/i386.md: Ditto.
848 * config/i386/mmx.md: Ditto.
849 * config/i386/predicates.md: Ditto.
851 2007-01-22 Andreas Schwab <schwab@suse.de>
853 * config/m68k/m68k.h: Fix comment.
855 2007-01-22 Jan Hubicka <jh@suse.cz>
857 * passes.c (init_optimization_passes): Do not rerun
858 pass_early_warn_uninitialized.
860 2007-01-22 Richard Guenther <rguenther@suse.de>
862 PR tree-optimization/30038
863 * tree-ssa-math-opts.c (maybe_record_sincos): New static helper
864 function.
865 (execute_cse_sincos_1): Likewise.
866 (execute_cse_sincos): Likewise.
867 (gate_cse_sincos): Likewise.
868 (pass_cse_sincos): New pass CSEing sin() and cos() calls using
869 the cexpi() canonicalization of sincos().
870 * tree-pass.h (pass_cse_sincos): Declare.
871 * passes.c (init_optimization_passes): New pass pas_cse_sincos.
873 2007-01-21 Eric Botcazou <ebotcazou@libertysurf.fr>
875 PR rtl-optimization/29329
876 * combine.c (replaced_rhs_insn): Rename to i2mod.
877 (replaced_rhs_value): Rename to i2mod_new_rhs.
878 (i2mod_old_rhs): New global variable.
879 (combine_instructions): Adjust for above change. Save a copy of
880 the old RHS into i2mod_old_rhs when the contents of a REG_EQUAL
881 note are substituted in the second instruction.
882 (distribute_notes) <REG_DEAD>: Adjust for above change. Do not
883 ditch the note if it pertains to the second eliminated register
884 and this register is mentioned in i2mod_old_rhs.
886 Revert:
887 2006-09-12 Eric Botcazou <ebotcazou@libertysurf.fr>
889 * combine.c (distribute_notes) <REG_DEAD>: Do not consider SETs past
890 the insn to which the note was originally attached.
892 2007-01-21 Jan Hubicka <jh@suse.cz>
894 * ipa-inline.c (inlining_mode): Comment, move up.
895 (cgraph_decide_inlining_incrementally): Do not perform inlining
896 itself; fix handling of flattening of self recursive functions.
897 (cgraph_find_cycles): Remove.
898 (cgraph_flatten_node): Remove.
899 (cgraph_decide_inlining): Use incremental inliner to handle flattening.
900 (try_inline): New function.
901 (cgraph_early_inlining): Update call of
902 cgraph_decide_inlining_incrementally. Apply inlining here.
903 (apply_inline): Update call of cgraph_decide_inlining_incrementally.
905 2007-01-21 Dirk Mueller <dmueller@suse.de>
907 PR bootstrap/30511
908 * tree-vrp.c (check_array_bounds): do not warn
909 about ADDR_EXPR's of ARRAY_REF's which are immediately
910 used in binary expressions.
912 2007-01-21 Ira Rosen <irar@il.ibm.com>
914 * tree-vectorizer.h (struct _stmt_vec_info): Add new field
915 read_write_dep and macros for its access.
916 * tree-vectorizer.c (new_stmt_vec_info): Initialize the new field.
917 * tree-vect-analyze.c (vect_analyze_data_ref_dependence): Remove
918 argument, call vect_check_interleaving for every independent pair of
919 data-refs. Mark loads that access the same memory location as a store
920 in the loop.
921 (vect_check_dependences): Remove.
922 (vect_analyze_data_ref_dependences): Remove vect_check_dependences
923 call, fix the call to vect_analyze_data_ref_dependence.
924 (vect_analyze_data_ref_access): For statements that access the same
925 data-ref, check that they are not stores; for loads, check that there
926 is no store that access the same location.
928 2007-01-20 Roger Sayle <roger@eyesopen.com>
929 Joseph Myers <joseph@codesourcery.com>
931 * doc/invoke.texi (-fdump-rtl-combine): Fix under/overfull hbox.
932 (-fdump-rtl-stack): Likewise.
933 (-fno-signed-zeros): Use @minus{} for a minus sign. Correct typo.
934 (-mcheck-zero-division, -mcpu): Fix under/overfull hbox.
935 (-mpt-fixed): Use @minus{} for minus sign.
936 (Using Precompiled Headers): Fix under/overfull hbox.
938 2007-01-20 Jan Hubicka <jh@suse.cz>
940 * tree-flow.h (struct stmt_ann_d): Move references_memory to proper
941 place within annotation.
943 2007-01-20 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
945 * pa.c (output_move_double): Change array size of xoperands to 4.
947 2007-01-20 Jan Hubicka <jh@suse.cz>
949 * tree-tailcall.c (adjust_return_value): Do not use RESULT_DECL
950 as temporary.
952 2007-01-19 Ian Lance Taylor <iant@google.com>
954 * expmed.c (expand_divmod) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]: Cast
955 constant to unsigned HOST_WIDE_INT before negating.
957 2007-01-19 Ian Lance Taylor <iant@google.com>
959 * tree-ssa-operands.h (struct vuse_vec_d): Change num_vuse field
960 to unsigned.
961 (VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Use unsigned comparison.
962 (VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
963 (SET_VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Likewise.
964 (SET_VUSE_ELEMENT_VAR) [ENABLE_CHECKING]: Likewise.
965 (SET_VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
966 (realloc_vdef, realloc_vuse): Change second parameter to
967 unsigned.
968 (ssa_operand_iterator_d): Change vuse_index and mayuse_index
969 fields to unsigned.
970 * tree-ssa-operands.c (realloc_vop): Change num_elem parameter to
971 unsigned. Change x and lim locals to unsigned.
972 (realloc_vdef, realloc_vuse): Change num_elem parameter to
973 unsigned.
974 (finalize_ssa_vuse_ops): Change old_i local to unsigned.
975 (copy_virtual_operands): Change i and n locals to unsigned.
977 2007-01-19 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
979 PR middle-end/29335
980 * builtins.c (fold_builtin_1): Handle builtin fdim.
982 2007-01-20 Jan Hubicka <jh@suse.cz>
984 * tree-ssa.c (init_tree_ssa): Do not call init_alias_heapvars.
985 * tree-ssa-structalias.c (compute_points_to_sets): Do call
986 init_alias_heapvars.
987 (init_alias_heapvars): Initialize only when not already initialized.
988 (delete_alias_heapvars): Set heapvar_for_stmt to NULL.
990 2007-01-19 Roger Sayle <roger@eyesopen.com>
992 * common.opt (fsigned-zeros): New command line option.
993 * flags.h (HONOR_SIGNED_ZEROS): Control via flag_signed_zeros instead
994 of flag_unsafe_math_optimizations.
995 * opts.c (set_fast_math_flags): The -ffast-math command line option
996 implies -fno-signed-zeros.
997 (fast_math_flags_set_p): Likewise.
999 * doc/invoke.texi: Document new -fno-signed-zeros option, and update
1000 the documentation of -ffast-math appropriately. Wrap long lines.
1002 2007-01-19 Steve Ellcey <sje@cup.hp.com>
1004 * system.h (ASM_MAKE_LABEL_LINKONCE): Poison.
1005 * varasm.c (globalize_decl): Remove ASM_MAKE_LABEL_LINKONCE ifdef.
1007 2007-01-19 Tomas Bily <tbily@suse.cz>
1009 * cgraphunit.c (cgraph_finalize_function): Updating of pid
1010 * tree-profile.c:
1011 (tree_init_ic_make_global_vars): New function
1012 (tree_init_edge_profiler): call of tree_init_ic_make_global_vars
1013 (tree_gen_ic_profiler): New function
1014 (tree_gen_ic_func_profiler): New function
1015 (tree_profiling): Added calling of tree_gen_ic_func_profiler
1016 (tree_profile_hooks): Added hook for indirec/virtual calls
1017 * value-prof.c (tree_find_values_to_profile): New case for
1018 indirect calls
1019 (tree_values_to_profile): Call for determining indirect/virtual
1020 counters
1021 (tree_indirect_call_to_profile): New function
1022 (tree_ic_transform): New function
1023 (tree_ic): New function
1024 (find_func_by_pid): New function
1025 (init_pid_map): New function
1026 (tree_value_profile_transformations): Added check for
1027 indirect/virtual call transformation
1028 * value-prof.h (enum hist_type): New counter type for
1029 indirect/virtual calls
1030 (profile_hooks): Added new hook for profiling indirect/virtual
1031 calls
1032 * profile.c (instrument_values): New case for indirect/virtual
1033 call added
1034 * gcov-io.h (GCOV_LAST_VALUE_COUNTER): Changed to 6
1035 (GCOV_COUNTER_V_INDIR): New counter type
1036 (GCOV_COUNTER_NAMES): New name of counter "indirect" added
1037 (GCOV_MERGE_FUNCTIONS): New merge function for indirect/virtual
1038 call added
1039 * cgraph.c: Definition of cgraph_max_pid
1040 (cgraph_create_node): Default init of pid attribute
1041 * cgraph.h: Declaration of cgraph_max_pid
1042 (struct cgraph_node): Added pid attribute
1043 * libgcov.c (__gcov_indirect_call_profiler): New function
1044 (__gcov_one_value_profiler_body): New function
1045 (__gcov_one_value_profiler): Body was moved to
1046 __gcov_one_value_profiler_body and calls it
1048 2007-01-19 Basile Starynkevitch <basile@starynkevitch.net>
1050 * doc/gty.texi (Options): Document the mark_hook option to GTY.
1051 * gengtype.c (write_types_data, write_func_for_structure,
1052 write_types, ggc_wtd, pch_wtd): Add skip_hooks to
1053 write_types_data, ggc_wtd, pch_wtd for processing mark_hook.
1054 (walk_type, write_func_for_structure): Generate the mark_hook if
1055 needed.
1057 2007-01-19 Jan Hubicka <jh@suse.cz>
1059 * ipa-inline.c (cgraph_decide_inlining_incrementally): Instead of
1060 'early' argument take inlining mode argument specifying whether to
1061 inline for size/speeed or all functions; add support for flattening;
1062 improve dumpting.
1063 (cgraph_early_inlining): Update call of decide_inlining_incrementally.
1065 2007-01-19 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
1067 PR c++/17947
1068 * toplev.c (warn_deprecated_use): Use %qD instead of %qs to print
1069 the name of the declared identifier.
1071 2007-01-19 Dirk Mueller <dmueller@suse.de>
1073 * config/i386.h (CONDITIONAL_REGISTER_USAGE): Store
1074 result of PIC_OFFSET_TABLE_REGNUM in temporary variable to avoid
1075 duplicate evaluation.
1077 2007-01-19 Uros Bizjak <ubizjak@gmail.com>
1079 * config/i386/i386.md (acos<mode>2): Rename from acossf2 and acosdf2.
1080 Macroize expander using X87MODEF12 mode macro. Extend operand 1
1081 to XFMode, use acosxf2 and truncate result to requested mode.
1082 Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
1083 (asin<mode>2): Similarly, with asin expanders.
1084 (*fscalexf4_i387): Rename from *fscalexf4.
1085 (expNcorexf3): New expander.
1086 (expxf2, exp10xf2, exp2xf2): Use expNcorexf3 expander.
1087 (exp<mode>2): Rename from expsf2 and expdf2. Macroize expander using
1088 X87MODEF12 mode macro. Extend operand 1 to XFMode, use expxf2 and
1089 truncate result to requested mode. Use SSE_FLOAT_MODE_P to disable
1090 patterns for SSE math.
1091 (exp10<mode>2): Similarly, with exp10 expanders.
1092 (exp2<mode>2): Similarly, with exp2 expanders.
1093 (expm1<mode>2): Similarly, with expm1 expanders.
1094 (ldexp<mode>3): Similarly, with ldexp expanders.
1095 (log<mode>2, log10<mode>2, log2<mode>2, log1p<mode>2, logb<mode>2):
1096 Use gen_truncxf<mode>2_i387_noop to truncate result.
1098 2007-01-19 Richard Sandiford <richard@codesourcery.com>
1100 * config/m68k/m68k.h (PREFERRED_STACK_BOUNDARY): Define to 32
1101 for ColdFire targets.
1103 2007-01-19 Nathan Sidwell <nathan@codesourcery.com>
1104 Richard Sandiford <richard@codesourcery.com>
1106 * config/m68k/m68k.h (M68K_STATIC_CHAIN_REG_NAME): New macro.
1107 (INITIALIZE_TRAMPOLINE): Use STATIC_CHAIN_REGNUM.
1108 (__transfer_from_trampoline): Use M68K_STATIC_CHAIN_REG_NAME.
1109 * config/m68k/m68kelf.h (STATIC_CHAIN_REGNUM): Override.
1110 (M68K_STATIC_CHAIN_REG_NAME): Likewise.
1111 * config/m68k/netbsd-elf.h (M68K_STATIC_CHAIN_REG_NAME): Likewise.
1113 2007-01-19 Richard Sandiford <richard@codesourcery.com>
1115 * config/m68k/m68k.md (adddi_dilshr32): Rename to...
1116 (*adddi_dilshr32): ...this. Fix formatting. Remove commented-out
1117 non-canonical pattern. Restrict to !TARGET_COLDFIRE.
1118 (*adddi_dilshr32_cf): New pattern.
1119 (adddi3, subdi3): Remove first alternatives.
1121 2007-01-19 Richard Sandiford <richard@codesourcery.com>
1123 * config/m68k/m68k.c (notice_update_cc): If an SFmode move is
1124 implemented using move.l, do not use its cc result for floating-point
1125 comparisons.
1127 2007-01-19 Richard Sandiford <richard@codesourcery.com>
1129 * config/m68k/m68k.h (EXTRA_CONSTRAINT): Stop the 'T' constraint
1130 from accepting 's' constraints if flag_pic.
1132 2007-01-19 Richard Sandiford <richard@codesourcery.com>
1134 * config/m68k/m68k.md (bordered, bunordered, buneq, bunge, bungt)
1135 (bunle, bunlt, bltgt, bordered_rev, bunordered_rev, buneq_rev)
1136 (bunge_rev, bungt_rev, bunle_rev, bunlt_rev, bltgt_rev): Change
1137 condition from TARGET_68881 to TARGET_HARD_FLOAT.
1139 2007-01-19 Sandra Loosemore <sandra@codesourcery.com>
1141 * longlong.h (count_leading_zeros, COUNT_LEADING_ZEROS_0): Add
1142 ColdFire alternatives.
1143 * config/m68k/m68k.h (CLZ_DEFINED_VALUE_AT_ZERO): New macro.
1144 * config/m68k/m68k.md (clzsi2): Define for ColdFire
1145 architectures that support ff1 instruction.
1147 2007-01-19 Richard Sandiford <richard@codesourcery.com>
1148 Julian Brown <julian@codesourcery.com>
1150 * config/m68k/m68k.h (CONST_OK_FOR_LETTER_P): Add an 'R' case.
1151 * config/m68k/m68k.md (*movsi_cfv4): Fold into...
1152 (*movsi_cf): ...here. Remove unnecessary 'R' from 'Rg'.
1153 Add commentary.
1155 2007-01-19 Richard Sandiford <richard@codesourcery.com>
1157 * config/m68k/m68k-protos.h (valid_mov3q_const): Take a HOST_WIDE_INT
1158 and return a bool.
1159 (output_move_const_into_data_reg, output_move_simode_const): Delete.
1160 * config/m68k/m68k.c (const_method, const_int_cost): Take a
1161 HOST_WIDE_INT instead of an rtx.
1162 (m68k_rtx_costs): Update call accordingly.
1163 (output_move_const_into_data_reg): Likewise. Fix formatting.
1164 (valid_mov3q_const): Take a HOST_WIDE_INT instead of an rtx.
1165 Return a bool.
1166 (output_move_simode_const): Update calls after above changes.
1167 Rework to use automatic variables and predicates like MEM_P.
1168 * config/m68k/m68k.md (pushexthisi_const): Update call to
1169 valid_mov3q_const.
1171 2007-01-19 Dirk Mueller <dmueller@suse.de>
1173 * tree-ssa-alias.c (perform_var_substitution): Fix typo
1174 in dump_flags test.
1176 2007-01-19 Richard Guenther <rguenther@suse.de>
1178 * builtins.c (expand_builtin_cexpi): Fall back to expanding
1179 via cexp in case sincos is not available.
1181 2007-01-19 Richard Guenther <rguenther@suse.de>
1183 * doc/tm.texi (TARGET_HAS_SINCOS): Document new target macro.
1184 * defaults.h (TARGET_HAS_SINCOS): Default to off.
1185 * config/linux.h (TARGET_HAS_SINCOS): Set to on if we have glibc.
1186 * config/alpha/linux.h (TARGET_HAS_SINCOS): Likewise.
1187 * config/sparc/linux.h (TARGET_HAS_SINCOS): Likewise.
1188 * config/sparc/linux64.h (TARGET_HAS_SINCOS): Likewise.
1189 * config/rs6000/linux.h (TARGET_HAS_SINCOS): Likewise.
1190 * config/rs6000/linux64.h (TARGET_HAS_SINCOS): Likewise.
1192 2007-01-19 Uros Bizjak <ubizjak@gmail.com>
1194 * config/i386/i386.md (*fpatanxf3_i387, fpatan_extend<mode>xf3_i387):
1195 New insn patterns.
1196 (atan2sf3_1, atan2df3_1, atan2xf3_1): Remove insn patterns.
1197 (atan2xf3): Directly generate RTL pattern.
1198 (atan2<mode>3): Rename from atan2sf3 and atan2df3 and macroize insn
1199 patterns using X87MODEF12 mode macro. Use fpatan_extend<mode>xf3_i387
1200 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
1201 disable patterns for SSE math.
1202 (atan<mode>2): Rename from atansf2 and atandf2 and macroize insn
1203 patterns using X87MODEF12 mode macro. Use fpatan_extend<mode>xf3_i387
1204 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
1205 disable patterns for SSE math.
1207 2007-01-19 Alexandre Oliva <aoliva@redhat.com>
1209 * libgcc-std.ver: Fix typo in %inherit for GCC_4.3.0.
1211 2007-01-18 Roger Sayle <roger@eyesopen.com>
1213 * fold-const.c (fold_unary) <VIEW_CONVERT_EXPR>: Optimize away a
1214 VIEW_CONVERT_EXPR to the same type as it's operand.
1216 2007-01-18 David Edelsohn <edelsohn@gnu.org>
1218 * config/rs6000/darwin-ldouble.c: Only build _SOFT_FLOAT if
1219 configured for long double 128.
1221 2007-01-18 Mike Stump <mrs@apple.com>
1223 * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Fix build
1224 error.
1226 2007-01-18 Michael Meissner <michael.meissner@amd.com>
1228 * i386.c (ix86_compute_frame_layout): Make fprintf's in #if 0 code
1229 type correct.
1231 2007-01-18 Jan Hubicka <jh@suse.cz>
1233 * tree-ssa-operands.c (vop_free_bucket_size): Never return value
1234 greater than NUM_VOP_FREE_BUCKETS.
1236 2007-01-18 Daniel Berlin <dberlin@dberlin.org>
1238 * tree-ssa-structalias.c: Update comments.
1239 (ptabitmap_obstack): Removed.
1240 (pta_obstack): New.
1241 (oldpta_obstack): Ditto.
1242 (stats): Add a few members.
1243 (struct variable_info): Remove node, complex, address_taken, and
1244 indirect_target members. Add oldsolution member.
1245 (new_var_info): Do not initialize removed members.
1246 (constraint_expr_type): Remove INCLUDES.
1247 (constraint_graph): Add size, implicit_preds, rep,
1248 indirect_cycles, eq_rep, label, direct_nodes, and complex members.
1249 (FIRST_REF_NODE): New macro.
1250 (LAST_REF_NODE): Ditto.
1251 (FIRST_ADDR_NODE): Ditto.
1252 (find): New function.
1253 (unite): Ditto.
1254 (dump_constraint): Do not handle INCLUDES.
1255 (insert_into_complex): Do not insert duplicate constraints.
1256 (condense_varmap_nodes): Renamed and rewritten into ...
1257 (merge_node_constraints): This. Also fix bug in handling of
1258 offseted copy constraints.
1259 (clear_edges_for_node): No longer need to deal with preds at all,
1260 or removing associated preds/succs.
1261 (merge_graph_nodes): Deal with indirect_cycles.
1262 Don't deal with predecessors.
1263 (add_implicit_graph_edge): New function.
1264 (add_pred_graph_edge): Ditto.
1265 (add_graph_edge): Don't deal with predecessors.
1266 (build_constraint_graph): Removed.
1267 (build_pred_graph): New function.
1268 (build_succ_graph): Ditto.
1269 (struct scc_info): Removed in_component. Added roots, dfs, and
1270 node_mapping. Remove visited_index, unification_queue.
1271 (scc_visit): Deal with union-find we do now.
1272 Deal with cycles with REF nodes.
1273 (collapse_nodes): Renamed and rewritten to ...
1274 (unify_nodes): This.
1275 (process_unification_queue): Removed.
1276 (topo_visit): Cleanup
1277 (do_da_constraint): Use find.
1278 (do_sd_constraint): Ditto.
1279 (do_ds_constraint): Ditto.
1280 (do_complex_constraint): Ditto.
1281 (init_scc_info): Update for removed and added members.
1282 (find_and_collapse_graph_cycles): Renamed and rewritten into ...
1283 (find_indirect_cycles): This.
1284 (equivalence_class): New variable.
1285 (label_visit): New function.
1286 (perform_variable_substitution): Rewritten.
1287 (free_var_substitution_info): New function.
1288 (find_equivalent_node): Ditto.
1289 (move_complex_constraints): Ditto.
1290 (eliminate_indirect_cycles): Ditto.
1291 (solve_graph): Only propagate changed bits.
1292 Use indirect cycle elimination.
1293 Use find.
1294 (tree_id_t): Rename to tree_vi_t, delete id member, add vi member.
1295 (tree_id_eq): Renamed to ...
1296 (tree_vi_eq): This. Update for member change
1297 (insert_id_for_tree): Renamed and rewritten to ...
1298 (insert_vi_for_tree): This.
1299 (lookup_id_for_tree): Renamed and rewritten to ...
1300 (lookup_vi_for_tree): This.
1301 (get_id_for_tree): Renamed and rewritten to ...
1302 (get_vi_for_tree): Ditto.
1303 (get_constraint_exp_from_ssa_var): Update to use get_vi_for_tree.
1304 (process_constraint): Don't handle INCLUDES.
1305 Remove special ADDRESSOF case.
1306 (find_func_aliases): Rewrite to use vi functions instead of id
1307 ones.
1308 (create_function_info_for): Ditto.
1309 (create_variable_info_for): Ditto.
1310 (intra_create_variable_infos): Ditto.
1311 (merge_smts_into): Ditto.
1312 (find_what_p_points_to): Ditto.
1313 (init_base_vars): Ditto.
1314 (init_alias_vars): Ditto.
1315 (remove_preds_and_fake_succs): New function.
1316 (dump_sa_points_to_info): Dump new stats.
1317 (dump_solution_for_var): Use find.
1318 (set_used_smts): Fix formatting.
1319 (compute_points_to_sets): Updated for new functions.
1320 (ipa_pta_execute): Ditto.
1322 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
1323 Richard Sandiford <richard@codesourcery.com>
1325 * doc/tm.texi (TARGET_FUNCTION_VALUE): Expand documentation of
1326 parallels.
1327 * calls.c (expand_call): If the return value is a PARALLEL,
1328 extract its first member.
1329 * config/m68k/linux.h (FUNCTION_EXTRA_EPILOGUE): Remove.
1330 * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
1331 use FUNCTION_EXTRA_EPILOGUE.
1332 (m68k_function_value): Return a PARALLEL if the return value
1333 is of a pointer type.
1334 * config/m68k/netbsd-elf.h (current_function_returns_pointer)
1335 (FUNCTION_EXTRA_EPILOGUE): Remove.
1336 * config/m68k/m68k.md (D0_REG): New constant.
1338 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
1340 * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
1341 output a NOP for empty epilogues.
1343 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1345 * config/m68k/m68k.c (m68k_use_return_insn): Update comments
1346 before function. Extend register save check to include all
1347 registers, not just integer ones.
1349 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
1351 * config/m68k/m68k-protos.h (use_return_insn): Rename to...
1352 (m68k_use_return_insn): ...this.
1353 * config/m68k/m68k.h (USE_RETURN_INSN): Delete.
1354 * config/m68k/m68k.c (use_return_insn): Rename to...
1355 (m68k_use_return_insn): ...this.
1356 * config/m68k/m68k.md (return): Use m68k_use_return_insn instead
1357 of USE_RETURN_INSN.
1359 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
1361 * config/m68k/fpgnulib.c (__truncdfsf2): Implement round to
1362 nearest even, fix denormal rounding overflow.
1364 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1366 * config/m68k/m68k.md (movsf_cf_hard): Use fsmove instead of
1367 f%$smove and f%$move.
1368 (movdf_cf_hard): Use fdmove for cases 0 and 3 and fmove for case 1.
1369 (extendsfdf2_cf): Use fdmove instead of f%&move.
1370 (truncdfsf2_cf): Use fsmove instead of f%$smove.
1371 (add<mode>3_cf, sub<mode>3_cf): Use <FP:prec> instead of <FP:round>.
1373 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1375 * config/m68k/m68k.md (movdf_cf_hard): Use output_move_double for
1376 GPR<-GPR moves.
1378 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1380 * real.h (real_format): Add a canonical_nan_lsbs_set field.
1381 (coldfire_single_format): Declare.
1382 (coldfire_double_format): Likewise.
1383 * real.c (encode_ieee_single): Use canonical_nan_lsbs_set instead
1384 of qnan_msb_set to determine the lower bits of a canonical
1385 NaN significand.
1386 (encode_ieee_double): Likewise.
1387 (encode_ieee_quad): Likewise.
1388 (ieee_single_format): Initialize canonical_nan_lsbs_set.
1389 (mips_single_format): Likewise.
1390 (ieee_double_format): Likewise.
1391 (mips_double_format): Likewise.
1392 (ieee_extended_motorola_format): Likewise.
1393 (ieee_extended_intel_96_format): Likewise.
1394 (ieee_extended_intel_128_format): Likewise.
1395 (ieee_extended_intel_96_round_53_format): Likewise.
1396 (ibm_extended_format): Likewise.
1397 (mips_extended_format): Likewise.
1398 (ieee_quad_format): Likewise.
1399 (mips_quad_format): Likewise.
1400 (vax_f_format): Likewise.
1401 (vax_d_format): Likewise.
1402 (vax_g_format): Likewise.
1403 (i370_single_format): Likewise.
1404 (i370_double_format): Likewise.
1405 (decimal_single_format): Likewise.
1406 (decimal_double_format): Likewise.
1407 (decimal_quad_format): Likewise.
1408 (c4x_single_format): Likewise.
1409 (c4x_extended_format): Likewise.
1410 (real_internal_format): Likewise.
1411 (coldfire_single_format): New real_format.
1412 (coldfire_double_format): Likewise.
1413 * config/pdp11/pdp11.c (pdp11_f_format): Initialize
1414 canonical_nan_lsbs_set.
1415 (pdp11_d_format): Likewise.
1416 * config/m68k/m68k.c (override_options): Override REAL_FORMAT_MODE
1417 if TARGET_COLDFIRE_CPU.
1419 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1421 * config/m68k/m68k-protos.h (m68k_output_pic_call): Delete.
1422 (output_call, m68k_legitimize_call_address): Declare.
1423 * config/m68k/m68k.h (EXTRA_CONSTRAINT): Remove unnecessary
1424 parenthesees. Add support for a 'W' constraint.
1425 (LEGITIMATE_PIC_OPERAND_P): Remove SYMBOL_REF_FLAG handling.
1426 (PRINT_OPERAND_PUNCT_VALID_P): Remove comment about 'o'.
1427 (m68k_symbolic_call, m68k_symbolic_jump): Declare.
1428 * config/m68k/m68k.c (m68k_symbolic_call, m68k_symbolic_jump): New
1429 variables.
1430 (override_options): Initialize them. Do not set flag_no_function_cse
1431 for TARGET_ID_SHARED_LIBRARY.
1432 (m68k_output_pic_call): Delete.
1433 (m68k_legitimize_call_address): New function.
1434 (print_operand): Remove the %o prefix. Handle the %p prefix.
1435 (output_call): New function.
1436 (m68k_output_mi_thunk): Use m68k_symbolic_jump. Always load the
1437 target address from the GOT if symbolic jumps are not allowed.
1438 * config/m68k/m68k.md (call, general_operand): Do not set
1439 SYMBOL_REF_FLAG. Use m68k_legitimize_call_address instead.
1440 Merge separate flag_pic and !flag_pic define_insns into...
1441 (*call, *call_value): ...these new patterns. Match the address
1442 rather than the containing MEM and require it to be a call_operand.
1443 Use output_call to generate the asm template.
1444 * config/m68k/predicates.md (const_call_operand): New predicate.
1445 (call_operand): Likewise.
1447 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
1449 * config/m68k/m68k.h (REGISTER_MOVE_COST): Simplify definition.
1450 (STACK_GROWS_DOWNWARD): Define to 1.
1451 (FUNCTION_VALUE, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P): Equivocate
1452 comments, emphasizing that these values are only defaults.
1453 * config/m68k/linux.h (LINK_SPEC): Fix formatting in #undef.
1454 * config/m68k/m68k.c (const_method): Remove trailing whitespace.
1456 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1458 * config/m68k/m68k.md (cmpsi): Remove outdated flag_pic handling.
1460 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
1461 Richard Sandiford <richard@codesourcery.com>
1463 * config/m68k/m68k.h (DATA_REGNO_P, ADDRESS_REGNO_P, INT_REGNO_P)
1464 (FP_REGNO_P): New macros.
1465 (REGNO_OK_FOR_INDEX_P, REGNO_OK_FOR_BASE_P, REGNO_OK_FOR_DATA_P)
1466 (REGNO_OK_FOR_FP_P, REG_OK_FOR_INDEX_P, REG_OK_FOR_BASE_P): Use them.
1467 (EH_RETURN_STACKADJ_RTX): Use A0_REG.
1468 * config/m68k/m68k.c (m68k_regno_mode_ok): Use the new REGNO macros.
1470 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
1472 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
1473 (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-rtems*): Add t-floatlib
1474 to $tmake_file.
1475 * config/m68k/t-floatlib: New file, extracting common code from...
1476 * config/m68k/t-m68kbare, config/m68k/t-m68kelf,
1477 * config/m68k/t-uclinux: Here.
1478 * config/m68k/fpgnulib.c: Do not compile extendeed precision
1479 routines on ColdFire targets.
1481 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
1483 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
1484 (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*)
1485 (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-linux*)
1486 (m68k-*-rtems*): Use tm_file rather than m68k/m68k.h and
1487 explicitly set MOTOROLA to 1.
1488 * config/m68k/m68k.h (MOTOROLA): Simplify definition accordingly.
1490 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1491 Nathan Sidwell <nathan@codesourcery.com>
1493 * config/m68k/m68k.h (PCC_STATIC_STRUCT_RETURN): Delete.
1494 (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Add MOTOROLA cases.
1495 * config/m68k/coff.h (REGISTER_PREFIX_MD): Delete.
1496 * config/m68k/m68020-elf.h (LIB_SPEC): Delete.
1497 * config/m68k/m68k-none.h (CC1_SPEC, CPP_SUBTARGET_SPEC): Delete.
1498 * config/m68k/m68kelf.h (IMMEDIATE_PREFIX, REGISTER_PREFIX_MD)
1499 (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Delete.
1500 (NO_DOLLAR_IN_LABEL, PCC_STATIC_STRUCT_RETURN): Don't undefine.
1501 (BSS_ASM_OP, ASM_OUTPUT_SKIP, ASM_OUTPUT_ASCII): Delete.
1502 * config/m68k/m68kemb.h (PCC_STATIC_STRUCT_RETURN): Don't undefine.
1503 * config/m68k/linux.h (SIZE_TYPE, PTRDIFF_TYPE, WCHAR_TYPE)
1504 (WCHAR_TYPE_SIZE, TARGET_OBJFMT_CPP_BUILTINS): Delete.
1505 (TARGET_OS_CPP_BUILTINS): Don't define mc68000 and mc68020 here.
1506 (DBX_REGISTER_NUMBER): Delete.
1507 * config/m68k/netbsd-elf.h (IMMEDIATE_PREFIX): Delete.
1508 (PCC_STATIC_STRUCT_RETURN): Don't undefine.
1509 * config/m68k/openbsd.h (PCC_STATIC_STRUCT_RETURN): Define.
1511 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1513 * config.gcc (m68k-*-uclinux*): Add flat.h to $tm_file.
1514 * config/flat.h: New file.
1515 * crtstuff.c (USE_PT_GNU_EH_FRAME): Don't define if
1516 OBJECT_FORMAT_FLAT.
1517 * config/m68k/m68k.h (ASM_PREFERRED_EH_DATA_FORMAT): Do not use
1518 indirect references for -msep-data or -mid-shared-library.
1519 Do not use PC-relative code addresses either.
1520 * config/m68k/m68k.c (override_options): Restrict -fPIC error
1521 to -mpcrel.
1522 * config/m68k/uclinux.h (STARTFILE_SPEC): Define. Use Scrt1.o
1523 for shared libraries and crt1.o for executables. Use crti.o and
1524 crtbegin.o.
1525 (ENDFILE_SPEC): Use crtend.o and crtn.o.
1526 (LIB_SPEC): Suppress -Rlibc.gdb if -static-libc is given.
1527 Do not add -elf2flt or -shared-lib-id options here.
1528 (LINK_SPEC): Define. Pass -elf2flt if no -elf2flt option is given.
1529 Pass -shared-lib-id if -mid-shared-library, taking the library
1530 identifier from -mshared-library-id if given, otherwise
1531 defaulting to 0.
1532 (EH_FRAME_IN_DATA_SECTION): Do not undefine.
1533 (INIT_SECTION_ASM_OP, FINI_SECTION_ASM_OP): Likewise.
1534 (TARGET_OS_CPP_BUILTINS): Define __GXX_MERGED_TYPEINFO_NAMES=0
1535 and __GXX_TYPEINFO_EQUALITY_INLINE=0 if -mid-shared-library.
1536 (DRIVER_SELF_SPECS): Map unadorned PIC options to -msep-data.
1537 * config/m68k/t-uclinux (EXTRA_MULTILIB_PARTS): Add crtbegin.o
1538 and crtend.o.
1539 * config/m68k/lb1sf68.asm (PICCALL): Use an lea and pc-relative
1540 jump sequence for ISA A and ISA A+.
1541 (PICJUMP): Likewise.
1543 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1545 * config.gcc (m68k-*-uclinux*): Base the port on the common
1546 and m68k GNU/Linux files rather than on the generic ELF ones.
1547 * config/m68k/uclinux.h (TARGET_VERSION): Override.
1548 (TARGET_OS_CPP_BUILTINS): Use LINUX_TARGET_OS_CPP_BUILTINS.
1550 2007-01-18 Julian Brown <julian@codesourcery.com>
1551 Richard Sandiford <richard@codesourcery.com>
1553 * config/m68k/m68k.h (LONG_DOUBLE_TYPE_SIZE): Make 64-bit on ColdFire.
1554 (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
1555 * config/m68k/netbsd-elf.h (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Undefine
1556 before redefining.
1557 * config/m68k/uclinux-oldabi.h (LONG_DOUBLE_TYPE_SIZE): Redefine to
1558 80 unconditionally.
1559 (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
1561 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1563 * doc/install.texi: Document m68k-uclinuxoldabi.
1564 * config.gcc (m68k-*-uclinuxoldabi): New configuration.
1565 * config/m68k/uclinux-oldabi.h: New file, copied from
1566 config/m68k/uclinux.h.
1568 2007-01-18 Kaz Kojima <kkojima@rr.iij4u.or.jp>
1570 * config/m32r/m32r-protos.h (m32r_expand_epilogue): Declare it.
1571 * config/m32r/m32r.c (m32r_setup_incoming_varargs): Use gen_frame_mem.
1572 (m32r_compute_frame_size): Use unsigned for regno.
1573 (m32r_reload_lr): Use gen_frame_mem.
1574 (pop): New.
1575 (m32r_output_function_epilogue): Don't output the function epilogue
1576 textually here.
1577 (m32r_expand_epilogue): New.
1578 (direct_return): Return false if the function has the interrupt
1579 attribute.
1580 (m32r_hard_regno_rename_ok): Remove code for the textual epilogue.
1581 * config/m32r/m32r.md (epilogue): New expander.
1582 (return_lr, return_rte): New insns.
1583 (return): Make it expander.
1584 (return_normal): New expander.
1586 2007-01-18 Josh Conner <jconner@apple.com>
1588 PR target/30485
1589 * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Add
1590 support for UNLE, UNLT, UNGE, and UNGT.
1592 2007-01-18 Jan Hubicka <jh@suse.cz>
1594 * tree-vrp.c (finalize_jump_threads): Do not call cleanup_cfg by hand.
1595 * tree-tailcall (add_virtual_phis): Likewise.
1596 (optimize_tail_call): Return TODOs.
1597 (execute_tail_calls): Return TODOs.
1598 * tree-ssa-ccp (execute_fold_all_builtins): Do cleanup_cfg via TODO.
1599 * tree-cfgcleanup.c (cleanup_tree_cfg_loop): Return if something
1600 changed.
1601 * tree-ssa-forwprop.c (tree_ssa_forward_propagate_single_use_value):
1602 Cleanup cfg using TODO.
1603 * tree-flow.h (cleanup_tree_cfg_loop): Update prototype.
1604 * passes.c (execute_function_todo): When cleanup did something, remove
1605 unused locals.
1606 * tree-cfg.c (pass_build_cfg): Add cleanup_cfg TODO.
1607 (make_edges): Don't cleanup_cfg.
1609 2007-01-18 Uros Bizjak <ubizjak@gmail.com>
1611 * reg-stack.c (subst_stack_regs_pat) [UNSPEC_SINCOS_COS,
1612 UNSPEC_XTRACT_FRACT]: Use generic code for instructions that
1613 operate on the top of stack.
1614 [UNSPEC_SINCOS_SIN, UNSPEC_XTRACT_EXP, UNSPEC_TAN]: Rewrite
1615 register handling of instructions that output to the second
1616 stack slot.
1617 [UNSPEC_TAN_ONE, UNSPEC_TAN_TAN]: Remove.
1618 (move_for_stack_reg): Special-case check for dead destination
1619 stack slot for constant load of 1.0 inside UNSPEC_TAN.
1621 * config/i386/i386.md (UNSPEC_TAN): New constant.
1622 (UNSPEC_TAN_ONE, UNSPEC_TAN_TAN): Remove.
1623 (fptanxf4_i387, fptan_extend<mode>xf4_i387): New patterns
1624 to correctly model move of constant 1.0 to top stack slot.
1625 (*tandf3_1, *tansf3_1, *tanxf3_1): Remove insn patterns.
1626 (unnamed peephole2 pattern): Remove corresponding peephole2
1627 pattern that optimizes tan insn and loading of constant 1.0.
1628 (tanxf2): Use fptanxf4_i387.
1629 (tan<mode>2): Rename from tansf2 and tandf2 and macroize insn
1630 patterns using X87MODEF12 mode macro. Use fptan_extend<mode>xf4_i387
1631 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
1632 disable patterns for SSE math.
1633 (sincos<mode>3): Use truncxf<mode>2_i387_noop for truncation.
1634 (fyl2x_extend<mode>xf3_i387): Use X87MODEF12 for operand 1.
1636 2007-01-18 Dirk Mueller <dmueller@suse.de>
1637 Richard Guenther <rguenther@suse.de>
1639 PR diagnostic/8268
1640 * doc/invoke.texi (Warray-bounds): Document -Warray-bounds.
1641 * common.opt (Warray-bounds): Add new warning option.
1642 * c-opts.c (c_common_handle_option): Define -Warray-bounds
1643 if -Wall is given.
1644 * Makefile.in: make tree-vrp.o depend on toplev.h
1645 * tree-vrp.c (vrp_finalize): Call check_array_refs if -Warray-bounds
1646 is enabled.
1647 (check_array_refs, check_array_bounds, check_array_ref): New.
1649 2007-01-18 Jan Hubicka <jh@suse.cz>
1651 * tree-ssa-ccp.c (ccp_finalize): Return if something changed.
1652 (execute_ssa_ccp): Return flags conditionally.
1653 * tree-ssa-propagate.c (substitue_and_fold): Return if something was
1654 changed.
1655 * tree-ssa-propagate.h (substitute_and_fold): Update prototype.
1657 2007-01-18 Steven Bosscher <steven@gcc.gnu.org>
1659 * cfgcleanup.c (cleanup_cfg): Detect cfglayout mode and set
1660 the CLEANUP_CFGLAYOUT flag when in cfglayout mode.
1662 * Makefile.c (GTFILES): Add cfglayout.h.
1663 * gengtype.c (open_base_files): Likewise.
1664 * cfglayout.c (cfg_layout_function_footer,
1665 cfg_layout_function_header) Reindent to make gengtype happy.
1666 * cfglayout.h (cfg_layout_function_footer,
1667 cfg_layout_function_header): Add GTY(()) marker.
1669 * ifcvt.c (noce_try_sign_mask): Make sure INSN_B is non-null.
1671 2007-01-18 Ben Elliston <bje@au.ibm.com>
1673 * genautomata.c (write_automata): Include xstrerror output in the
1674 error message if writing the DFA description file fails.
1676 2007-01-17 H.J. Lu <hongjiu.lu@intel.com>
1678 * config/mips/mips-protos.h (mips_output_external): Make it
1679 return void.
1680 * config/mips/iris.h (TARGET_ASM_EXTERNAL_LIBCALL): Removed.
1681 * config/mips/mips.c (irix_output_external_libcall): Likewise.
1682 (extern_list): Likewise.
1683 (extern_head): Likewise.
1684 (TARGET_ASM_FILE_END): Likewise.
1685 (mips_file_end): Likewise.
1686 (mips_output_external): Rewritten.
1688 2007-01-18 Ben Elliston <bje@au.ibm.com>
1690 * genpreds.c (write_insn_preds_c): Only write out the function
1691 body for regclass_for_constraint if we have register constraints.
1693 2007-01-17 Tom Tromey <tromey@redhat.com>
1695 * doc/sourcebuild.texi (libgcj Tests): Use sourceware.org.
1696 * doc/install.texi (Testing): Use sourceware.org.
1697 (Binaries): Likewise.
1698 (Specific): Likewise.
1699 * doc/contrib.texi (Contributors): Use sourceware.org.
1701 2007-01-17 Anatoly Sokolov <aesok@post.ru>
1703 * config/avr/avr.h (AVR_HAVE_LPMX): New macro.
1704 (AVR_ENHANCED): Rename to ...
1705 (AVR_HAVE_MUL): ... new.
1706 (avr_enhanced_p): Rename to ...
1707 (avr_have_mul_p): ... new.
1708 (TARGET_CPU_CPP_BUILTINS): Use 'avr_have_mul_p' instead of
1709 'avr_enhanced_p' for "__AVR_ENHANCED__". Define "__AVR_HAVE_MUL__".
1710 * config/avr/avr.c (avr_enhanced_p): Rename to ...
1711 (avr_have_mul_p): ... new.
1712 (base_arch_s): Rename 'enhanced' to 'have_mul'.
1713 (avr_override_options): Use 'avr_have_mul_p' and 'have_mul' instead of
1714 'avr_enhanced_p' and 'enhanced'.
1715 (ashlhi3_out, ashrhi3_out, lshrhi3_out, avr_rtx_costs): Use
1716 AVR_HAVE_MUL instead of AVR_ENHANCED.
1717 * avr.md (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
1718 (mulqi3, *mulqi3_enh, *mulqi3_call, mulqihi3, umulqihi3, mulhi3,
1719 *mulhi3_enh, *mulhi3_call, mulsi3, *mulsi3_call): Use AVR_HAVE_MUL
1720 instead of AVR_ENHANCED.
1721 (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
1722 * libgcc.S: Use __AVR_HAVE_MUL__ instead of __AVR_ENHANCED__.
1723 (__tablejump__): Use __AVR_HAVE_LPMX__ instead of __AVR_ENHANCED__.
1725 2007-01-17 Ian Lance Taylor <iant@google.com>
1727 * vec.h (VEC_reserve_exact): Define.
1728 (vec_gc_p_reserve_exact): Declare.
1729 (vec_gc_o_reserve_exact): Declare.
1730 (vec_heap_p_reserve_exact): Declare.
1731 (vec_heap_o_reserve_exact): Declare.
1732 (VEC_OP (T,A,reserve_exact)): New static inline function, three
1733 versions.
1734 (VEC_OP (T,A,reserve)) [all versions]: Remove handling of
1735 negative parameter.
1736 (VEC_OP (T,A,alloc)) [all versions]: Call ...reserve_exact.
1737 (VEC_OP (T,A,copy)) [all versions]: Likewise.
1738 (VEC_OP (T,a,safe_grow)) [all versions]: Likewise.
1739 * vec.c (calculate_allocation): Add exact parameter. Change all
1740 callers.
1741 (vec_gc_o_reserve_1): New static function, from vec_gc_o_reserve.
1742 (vec_gc_p_reserve, vec_gc_o_reserve): Call vec_gc_o_reserve_1.
1743 (vec_gc_p_reserve_exact, vec_gc_o_reserve_exact): New functions.
1744 (vec_heap_o_reserve_1): New static function, from vec_heap_o_reserve.
1745 (vec_heap_p_reserve, vec_heap_o_reserve): Call vec_heap_o_reserve_1.
1746 (vec_heap_p_reserve_exact): New function.
1747 (vec_heap_o_reserve_exact): New function.
1749 2007-01-17 Jan Hubicka <jh@suse.cz>
1751 * ipa-type-escape.c (look_for_casts): Revamp using handled_component_p.
1753 2007-01-17 Eric Christopher <echristo@apple.com>
1755 * config.gcc: Support core2 processor.
1757 2007-01-16 Jan Hubicka <jh@suse.cz>
1759 * tree-ssanames.c (release_dead_ssa_names): Instead of ggc_freeing
1760 the names, just unlink the chain so we don't crash on dangling pointers
1761 to dead SSA names.
1763 2007-01-16 Jan Hubicka <jh@suse.cz>
1765 * cgraph.h (cgraph_decide_inlining_incrementally): Kill.
1766 * tree-pass.h: Reorder to make IPA passes appear toegher.
1767 (pass_early_inline, pass_inline_parameters, pass_apply_inline): Declare.
1768 * cgraphunit.c (cgraph_finalize_function): Do not compute inling
1769 parameters, do not call early inliner.
1770 * ipa-inline.c: Update comments. Include tree-flow.h
1771 (cgraph_decide_inlining): Do not compute inlining parameters.
1772 (cgraph_decide_inlining_incrementally): Return TODOs; assume to
1773 be called with function context set up.
1774 (pass_ipa_inline): Remove unreachable functions before pass.
1775 (cgraph_early_inlining): Simplify assuming to be called from the
1776 PM as local pass.
1777 (pass_early_inline): New pass.
1778 (cgraph_gate_ipa_early_inlining): New gate.
1779 (pass_ipa_early_inline): Turn into simple wrapper.
1780 (compute_inline_parameters): New function.
1781 (gate_inline_passes): New gate.
1782 (pass_inline_parameters): New pass.
1783 (apply_inline): Move here from tree-optimize.c
1784 (pass_apply_inline): New pass.
1785 * ipa.c (cgraph_remove_unreachable_nodes): Verify cgraph after
1786 transforming.
1787 * tree-inline.c (optimize_inline_calls): Return TODOs rather than
1788 doing them by hand.
1789 (tree_function_versioning): Do not allocate dummy struct function.
1790 * tree-inline.h (optimize_inline_calls): Update prototype.
1791 * tree-optimize.c (execute_fixup_cfg): Export.
1792 (pass_fixup_cfg): Remove
1793 (tree_rest_of_compilation): Do not apply inlines.
1794 * tree-flow.h (execute_fixup_cfg): Declare.
1795 * Makefile.in (gt-passes.c): New.
1796 * passes.c: Include gt-passes.h
1797 (init_optimization_passes): New passes.
1798 (nnodes, order): New static vars.
1799 (do_per_function_toporder): New function.
1800 (execute_one_pass): Dump current pass here.
1801 (execute_ipa_pass_list): Don't dump current pass here.
1803 2007-01-16 Janis Johnson <janis187@us.ibm.com>
1805 * config/dfp-bit.c (dfp_compare_op): Return separate value for NaN.
1806 (DFP_NE, DFP_LE, DFP_GE): Return false for NaN.
1808 2007-01-16 David Edelsohn <edelsohn@gnu.org>
1810 * config/rs6000/darwin-ldouble.c: Build file for SOFT_FLOAT.
1811 (strong_alias): Define.
1812 (__gcc_qmul): Provide non-FMA for soft-float.
1813 (__gcc_qdiv): Same.
1814 (__gcc_qneg): New.
1815 (__gcc_qeq): New.
1816 (__gcc_qle): New.
1817 (__gcc_qge): New.
1818 (__gcc_qunord): New.
1819 (__gcc_stoq): New.
1820 (__gcc_dtoq): New.
1821 (__gcc_qtos): New.
1822 (__gcc_qtod): New.
1823 (__gcc_qtoi): New.
1824 (__gcc_qtou): New.
1825 (__gcc_itoq): New.
1826 (__gcc_utoq): New.
1827 (fmsub): New.
1828 * config/rs6000/rs6000.c (rs6000_init_libfuncs): Initialize
1829 soft-float functions.
1830 * config/rs6000/libgcc-ppc-glibc.ver: Version soft-float symbols.
1831 * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Do not warn
1832 about long double soft float.
1834 2007-01-16 Dorit Nuzman <dorit@il.ibm.com>
1835 Tehila Meyzels <tehila@il.ibm.com>
1837 * tree-vectorizer.h (is_pattern_stmt_p): New.
1838 * tree-vect-analyze.c (vect_determine_vectorization_factor): Fix
1839 formatting (tabs instead of spaces). Cleanup and clarify setting
1840 of STMT_VINFO_VECTYPE. Call is_pattern_stmt_p.
1841 * tree-vect-transform.c (vect_get_vec_def_for_operand): Fix typo.
1842 (vectorizable_type_demotion): Check that types are integral.
1843 (vectorizable_type_promotion): Likewise.
1844 (vectorizable_store): Fix typo. Eliminate new-line at end of
1845 comments.
1847 2007-01-16 Jan Hubicka <jh@suse.cz>
1849 * tree-ssanames.c (release_dead_ssa_names): Remove invalidated
1850 cgraph edges too.
1852 2007-01-15 Eric Christopher <echristo@apple.com>
1854 * ifcvt.c: Include vec.h, vecprim.h.
1855 (check_cond_move_block): New argument regs.
1856 Reorganize. Add registers used to regs.
1857 (cond_move_process_if_block): Use regs set above as
1858 loop bounds.
1860 2007-01-15 Eric Christopher <echristo@apple.com>
1862 * config/darwin.h: Update copyright.
1863 (TARGET_OPTION_TRANSLATE_TABLE): Add umbrella.
1864 (LINK_COMMAND_SPEC): Add -u.
1865 (LINK_SPEC): Fix umbrella for above.
1867 2007-01-15 Joseph S. Myers <joseph@codesourcery.com>
1869 * config/soft-fp/op-common.h, config/soft-fp/op-4.h: Update from
1870 glibc CVS.
1872 2007-01-15 Tom Tromey <tromey@redhat.com>
1874 * doc/sourcebuild.texi (libgcj Tests): Don't mention jacks.
1875 * doc/install.texi (Testing): Don't mention jacks.
1876 (Configuration): Document --enable-java-maintainer-mode. Move
1877 --with-java-home to libgcj-specific section. Document
1878 --with-ecj-jar.
1879 (Prerequisites): Mention --enable-java-maintainer-mode, ecj1.
1881 2007-01-15 Jan Hubicka <jh@suse.cz>
1883 * tree-ssa-dce.c (DCE_TODOs): New.
1884 (propagate_necessity): Return if something changed.
1885 (eliminate_unnecessary_stmts): Likewise.
1886 (perform_tree_ssa_dce): Return TODO flags when needed.
1887 (pass_dce, pass_dce_loop, pass_cd_dce): Remove TODO flags.
1889 2007-01-15 Uros Bizjak <ubizjak@gmail.com>
1891 * config/i386/i386.md (fyl2xxf3_i387): Rename from fyl2x_xf3.
1892 (fyl2x_extend<mode>xf3_i387): New insn pattern.
1893 (log<mode>2): Rename from logsf2 and logdf2 and macroize insn
1894 patterns using X87MODEF12 mode macro. Extend operand 1
1895 to XFmode. Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
1896 (log10<mode>2): Ditto.
1897 (log2<mode>2): Ditto.
1898 (log1p<mode>2): Ditto.
1899 (logb<mode>2): Ditto.
1900 (fyl2xp1xf3_i387): Rename from fyl2xp1_xf3.
1901 (fyl2xp1_extend<mode>xf3_i387): New insn pattern.
1902 (*fxtractxf3_i387): Rename from *fxtractxf3.
1903 (fxtract_extend<mode>xf3_i387): New insn pattern.
1904 (ilogbsi2): Use match_dup 3, not match_operand:XF 3.
1905 * config/i386/i386.c (ix86_emit_i387_log1p): Use gen_fyl2xp1xf3_i387()
1906 and gen_fyl2xxf3_i387().
1908 2007-01-14 Zdenek Dvorak <dvorakz@suse.cz>
1910 * loop-unswitch.c (unswitch_loop): Do not call fix_loop_placement.
1911 * cfgloopmanip.c (fix_loop_placement): Made static. Use
1912 get_loop_exit_edges. Changed return type to bool.
1913 * cfgloop.h (fix_loop_placement): Declaration removed.
1915 2007-01-14 Dorit Nuzman <dorit@il.ibm.com>
1917 * param.h (MIN_VECT_LOOP_BOUND): New.
1918 * params.def (MIN_VECT_LOOP_BOUND): New.
1919 * tree-vectorizer.c (slpeel_tree_peel_loop_to_edge): Takes another
1920 argument - minimum threshold for number of iterations.
1921 * tree-vectorizer.h (slpeel_tree_peel_loop_to_edge): Add another
1922 argument to declaration.
1923 * tree-vect-analyze.c (vect_analyze_operations): Check value of
1924 MIN_VECT_LOOP_BOUND.
1925 * tree-vect-transform.c (vect_do_peeling_for_loop_bound): Call
1926 slpeel_tree_peel_loop_to_edge with additional argument.
1927 (vect_do_peeling_for_alignment): Likewise.
1928 * doc/invoke.texi (min-vect-loop-bound): Document new param option.
1930 2007-01-14 Uros Bizjak <ubizjak@gmail.com>
1932 PR target/30413
1933 * config/i386/i386.c (print_operand) ['z']: Output 'b' for
1934 operands of size 1.
1936 2007-01-14 Jan Hubicka <jh@suse.cz>
1938 * tree-dfa.c (remove_referenced_var): New function.
1939 * tree-ssa-live.c (remove_unused_locals): Walk referenced vars and
1940 prune referenced vars list too.
1941 * tree-flow.h (remove_referenced_var): Declare.
1943 2007-01-14 Jan Hubicka <jh@suse.cz>
1945 * tree-eh.c (add_stmt_to_eh_region_fn): Do not add call_exprs
1946 separately.
1947 (remove_stmt_from_eh_region_fn): Do not remove call_exprs.
1948 (verify_eh_throw_stmt_node, verify_eh_throw_table_statements): Kill.
1949 * except.h (verify_eh_throw_table_statements): Kill prototype.
1950 * cfgexpand.c (expand_gimple_basic_block): Propagate Eh regions
1951 into call exrepssions.
1952 * tree-optimize.c (execute_free_cfg_annotatiosn): Do not call
1953 eh trhow verifier.
1954 * tree-cfg.c: Include pointer-set.h.
1955 (verify_node_sharing): Work on pointer set.
1956 (verify_eh_throw_stmt_node): New.
1957 (verify_stmts): Use pointers sets, verify throw_stmt.
1959 2007-01-13 Zdenek Dvorak <dvorakz@suse.cz>
1961 * ipa-reference.c (analyze_function): Consider also addresses taken
1962 in phi nodes.
1964 2007-01-12 Roger Sayle <roger@eyesopen.com>
1966 * c-typeck.c (null_pointer_constant_p): Replace use of
1967 TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
1968 (build_c_cast): Likewise.
1970 2007-01-12 Roger Sayle <roger@eyesopen.com>
1972 * tree.h (force_fit_type_double): Remove unused final argument.
1973 * c-common.c (constant_expression_warning): Replace use of
1974 TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
1975 (convert_and_check): Likewise.
1976 (shorten_compare): Update call to force_fit_type_double.
1977 (c_common_truthvalue_conversion) <INTEGER_CST>: Use integer_zerop.
1978 * convert.c (convert_to_pointer): Update call to
1979 force_fit_type_double.
1980 * fold-const.c (force_fit_type_double): Remove overflowed_const
1981 argument.
1982 (int_const_binop, fold_convert_const_int_from_int,
1983 fold_convert_const_int_from_real, fold_div_compare,
1984 fold_sign_changed_comparison, fold_unary, fold_negate_const,
1985 fold_abs_const, fold_not_const): Remove the final argument from
1986 calls to force_fit_type_double.
1988 2007-01-12 Andrew Pinski <andrew_pinski@playstation.sony.com>
1990 * configure.ac: Set insn to "nop" for spu-*-* also.
1991 * configure: Regenerate.
1993 2007-01-12 Olga Golovanevsky <olga@il.ibm.com>
1995 * builtins.def : Add BUILT_IN_FREE.
1997 2007-01-12 Jan Hubicka <jh@suse.cz>
1999 PR tree-optimization/30443
2000 * tree-inline.c (tree_function_versioning): Do not optimize when
2001 cloning for inlining.
2003 2007-01-12 Zdenek Dvorak <dvorakz@suse.cz>
2005 * doc/loop.texi: Document recording of loop exits.
2006 * cfgloopmanip.c (loopify, duplicate_loop): Use alloc_loop.
2007 (update_single_exits_after_duplication,
2008 update_single_exit_for_duplicated_loop,
2009 update_single_exit_for_duplicated_loops): Removed.
2010 (duplicate_loop_to_header_edge): Do not call
2011 update_single_exits_after_duplication and
2012 update_single_exit_for_duplicated_loops.
2013 (loop_version): Do not update single_exit information.
2014 (fix_loop_structure): Use record_loop_exits instead of
2015 mark_single_exit_loops.
2016 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update
2017 the lists of loop exits.
2018 * cfghooks.c (redirect_edge_and_branch, redirect_edge_and_branch_force,
2019 split_edge, merge_blocks): Update the lists of loop exits.
2020 * modulo-sched.c (sms_schedule): Pass LOOPS_HAVE_RECORDED_EXITS to
2021 loop_optimizer_init.
2022 * loop-init.c (loop_optimizer_init): Call record_loop_exits instead
2023 of mark_single_exit_loops.
2024 (loop_optimizer_finalize): Call release_recorded_exits.
2025 * tree-ssa-loop.c (tree_loop_optimizer_init): Pass
2026 LOOPS_HAVE_RECORDED_EXITS to loop_optimizer_init.
2027 * tree-vectorizer.c (slpeel_tree_duplicate_loop_to_edge_cfg): Do not
2028 update single exit information.
2029 * lambda-code.c (perfect_nestify): Ditto.
2030 * cfgloop.c (flow_loop_free): Destroy the list of exits of the loop.
2031 (mark_single_exit_loops): Removed.
2032 (alloc_loop, loop_exit_hash, loop_exit_eq, loop_exit_free,
2033 get_exit_descriptions, rescan_loop_exit, record_loop_exits,
2034 dump_recorded_exit, dump_recorded_exits, release_recorded_exits): New
2035 functions.
2036 (get_loop_exit_edges, single_exit): Use recorded exit lists.
2037 (add_bb_to_loop, remove_bb_from_loops): Update the lists of loop exits.
2038 (verify_loop_structure): Verify consistency of the exit lists.
2039 (flow_loops_find): Use alloc_loop. Initialize exits hash.
2040 (set_single_exit): Removed.
2041 * cfgloop.h (struct loop_exit): New function.
2042 (struct loop): single_exit_ field replaced by exits field.
2043 (LOOPS_HAVE_MARKED_SINGLE_EXITS): Replaced by LOOPS_HAVE_RECORDED_EXITS.
2044 (struct loops): Added exits hash.
2045 (mark_single_exit_loops, set_single_exit): Declaration removed.
2046 (release_recorded_exits, record_loop_exits, rescan_loop_exit): Declare.
2048 2007-01-12 Richard Sandiford <richard@codesourcery.com>
2050 * doc/invoke.texi: Avoid use of @headitem.
2052 2007-01-12 Richard Sandiford <richard@codesourcery.com>
2054 * cse.c (cse_insn): Move HAVE_CC0 code after declarations.
2056 2007-01-12 Richard Sandiford <richard@codesourcery.com>
2058 * doc/install.texi: Fix m68k-*-* anchor and add m68k-*-* to the
2059 list of targets.
2061 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
2062 Richard Sandiford <richard@codesourcery.com>
2064 * doc/invoke.texi: Document -mno-bitfield, -mno-rtd and -mno-short.
2065 * config/m68k/m68k.opt: Resort options.
2066 (mbitfield, mrtd, mshort): Remove RejectNegative properties.
2068 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
2069 Richard Sandiford <richard@codesourcery.com>
2071 * doc/invoke.texi: Document the macros that are defined by
2072 m68k's -mtune and -mhard-float options.
2073 * config/m68k/m68k-protos.h (m68k_cpp_cpu_ident) Declare.
2074 (m68k_cpp_cpu_family): Likewise.
2075 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Add a full set
2076 of __ucfv*__ macros. Define __mcffpu__ if generating code for
2077 ColdFire FPUs. Define __mcf_cpu_* and __mcf_family_* macros.
2078 * config/m68k/m68k.c (m68k_cpp_cpu_ident): New function.
2079 (m68k_cpp_cpu_family): Likewise.
2081 2007-01-12 Richard Sandiford <richard@codesourcery.com>
2083 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Treat all mc68*
2084 macros besides mc68000 as tuning macros. Use a switch statement
2085 to set them and mcpu32.
2087 2007-01-12 Julian Brown <julian@codesourcery.com>
2089 * config/m68k/m68k.h: Use TARGET_68040 instead of TARGET_68040_ONLY.
2090 (TARGET_68040_ONLY): Rename to...
2091 (TARGET_68040): ...this.
2092 * config/m68k/m68k.c: Use TARGET_68040 instead of TARGET_68040_ONLY.
2093 * config/m68k/m68k.md: Likewise.
2095 2007-01-12 Julian Brown <julian@codesourcery.com>
2096 Nathan Sidwell <nathan@codesourcery.com>
2097 Richard Sandiford <richard@codesourcery.com>
2099 * config.gcc (m680[012]0-*-*, m68k*-*-*): Set m68k_cpu_ident to
2100 the -mcpu= argument associated with the --with-cpu setting.
2101 Define M68K_DEFAULT_TUNE to the default -mtune= option,
2102 if different from the one implied by the -mcpu setting.
2103 Accept --with-cpu=FOO if FOO is listed in m68k-devices.def,
2104 using mcpu=FOO as the default CPU option. Set target_cpu_default2.
2105 * doc/invoke.texi: Mention ColdFire in the introduction to the
2106 m68k options. Document the new -march, -mcpu, -mtune, -mdiv,
2107 -mno-div and -mhard-float options. Make -m68881 a synonym for
2108 -mhard-float. Document the previously-undocumented -m5206e,
2109 -m528x, -m5307 and -m5407 options. Tweak the existing option
2110 documentation for consistency.
2111 * doc/install.texi: Mention new --with-cpu arguments.
2112 * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Only use the
2113 default CPU if neither -mcpu nor -march are specified.
2114 (ASM_CPU_SPEC): Pass down -mcpu and -march options.
2115 (TARGET_CPU_CPP_BUILTINS): Set __mcfisa*__ macros from
2116 TARGET_ISA*. Set the legacy __mcf*__ cpu macros in the same way,
2117 using m68k_tune to decide between families that implement the
2118 same ISA. Use m68k_tune to set __mcfv4e__.
2119 (FL_BITFIELD, FL_68881, FL_COLDFIRE, FL_CF_HWDIV, FL_CF_MAC)
2120 (FL_CF_EMAC, FL_CF_EMAC_B, FL_CF_USP, FL_CF_FPU, FL_ISA_68000)
2121 (FL_ISA_68010, FL_ISA_68020, FL_ISA_68040, FL_ISA_A, FL_ISA_B)
2122 (FL_ISA_C, FL_ISA_MMU): New macros.
2123 (MASK_COLDFIRE): Delete.
2124 (TARGET_68010, TARGET_68020, TARGET_68040_ONLY, TARGET_COLDFIRE)
2125 (TARGET_ISAB): Redefine in terms of m68k_cpu_flags.
2126 (TARGET_68881, TARGET_COLDFIRE_FPU): Redefine in terms of m68k_fpu.
2127 (TARGET_HARD_FLOAT): Do not define here.
2128 (TARGET_ISAAPLUS, TARGET_ISAC): New macros.
2129 (TUNE_68000): New macro.
2130 (TUNE_68000_10): Redefine in terms of TUNE_68000 and TUNE_68010.
2131 (TUNE_68010, TUNE_68030, TUNE_68040, TUNE_68060, TUNE_CPU32)
2132 (TUNE_CFV2): Redefine in terms of m68k_tune.
2133 (uarch_type, target_device, fpu_type): New enums.
2134 (m68k_cpu, m68k_tune, m68k_fpu, m68k_cpu_flags): Declare.
2135 * config/m68k/m68k.c (TARGET_DEFAULT): Remove MASK_68881.
2136 (FL_FOR_isa_00, FL_FOR_isa_10, FL_FOR_isa_20, FL_FOR_isa_40)
2137 (FL_FOR_isa_cpu32, FL_FOR_isa_a, FL_FOR_isa_aplus, FL_FOR_isa_b)
2138 (FL_FOR_isa_c): New macros.
2139 (m68k_isa): New enum.
2140 (m68k_target_selection): New structure.
2141 (all_devices, all_isas, all_microarchs): New tables.
2142 (m68k_cpu_entry, m68k_arch_entry, m68k_tune_entry, m68k_cpu)
2143 (m68k_tune, m68k_fpu, m68k_cpu_flags): New variables.
2144 (MASK_ALL_CPU_BITS): Delete.
2145 (m68k_find_selection): New function.
2146 (m68k_handle_option): Handle -mcpu=, -march= and -mtune=.
2147 Map the legacy target options to a combination of the new ones.
2148 (override_options): Set m68k_cpu, m68k_tune, m68k_fpu and
2149 m68k_cpu_flags. Handle M68K_DEFAULT_TUNE. Use m68k_cpu_flags
2150 to derive default MASK_BITFIELD, MASK_CF_HWDIV and MASK_HARD_FLOAT
2151 settings.
2152 * config/m68k/m68k.opt (m5200, m5206e, m528x, m5307, m5407, mcfv4e)
2153 (m68010, m68020, m68020-40, m68020-60, m68030, m68040): Remove Mask
2154 properties.
2155 (m68881, msoft-float): Change mask from 68881 to HARD_FLOAT.
2156 (march=, mcpu=, mdiv, mhard-float, mtune=): New options.
2157 * config/m68k/m68k-devices.def: New file.
2159 2007-01-12 Richard Sandiford <richard@codesourcery.com>
2160 Nathan Sidwell <nathan@codesourcery.com>
2162 * config/m68k/m68k.h (ASM_CPU_SPEC, ASM_SPEC, EXTRA_SPECS)
2163 (SUBTARGET_EXTRA_SPECS): New macros.
2164 * config/m68k/linux.h (ASM_SPEC): Remove CPU flags;
2165 use %(asm_cpu_spec) instead.
2166 * config/m68k/m68k-none.h (ASM_SPEC): Likewise.
2167 * config/m68k/openbsd.h (ASM_SPEC): Likewise.
2168 * config/m68k/netbsd-elf.h (ASM_SPEC): Likewise.
2169 (EXTRA_SPECS): Rename to...
2170 (SUBTARGET_EXTRA_SPECS): ...this.
2172 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
2173 Richard Sandiford <richard@codesourcery.com>
2174 Julian Brown <julian@codesourcery.com>
2176 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
2177 (m68k-*-uclinux*, m68k-*-linux*, m68k-*-rtems*): Set default_m68k_cpu
2178 to the configuration's default CPU.
2179 (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*): Likewise.
2180 Remove default masks.
2181 (m680[012]0-*-*): Set the default with_cpu to the first part of
2182 the target name.
2183 (m68k*-*-*): Set the default with_cpu to m$default_m68k_cpu.
2184 (m68k*-*-linux): Extend the --with-cpu handling to...
2185 (m680[012]0-*-*, m68k*-*-*): ...these configurations. Allow m68000
2186 and m68010. Don't set target_cpu_default2.
2187 * doc/install.texi: Document --with-cpu for m68k.
2188 * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Define.
2189 * config/m68k/m68k-none.h (TARGET_CPU_DEFAULT, M68K_CPU_m68k)
2190 (M68K_CPU_m68000, M68K_CPU_m68010, M68K_CPU_m68020, M68K_CPU_m68030)
2191 (M68K_CPU_m68040, M68K_CPU_m68302, M68K_CPU_m68332, TARGET_DEFAULT)
2192 (ASM_CPU_DEFAULT_SPEC, CC1_CPU_DEFAULT_SPEC): Delete.
2193 (ASM_SPEC): Remove use of %(asm_cpu_default).
2194 (EXTRA_SPECS, SUBTARGET_EXTRA_SPECS, MULTILIB_DEFAULTS): Delete.
2195 * config/m68k/linux.h (TARGET_DEFAULT): Delete.
2196 (CPP_SPEC): Merge definitions. Do not handle __HAVE_68881__ here.
2197 * config/m68k/netbsd-elf.h (TARGET_OS_CPP_BUILTINS): Define
2198 __HAVE_FPU__ if TARGET_HARD_FLOAT.
2199 (TARGET_DEFAULT): Delete.
2200 (EXTRA_SPECS): Delete cpp_cpu_default_spec, cpp_cpu_spec,
2201 cpp_fpu_spec, asm_default_spec and netbsd_cpp_spec.
2202 (CPP_CPU_SPEC): Delete.
2203 (TARGET_VERSION): Merge definitions, using TARGET_68010 to pick
2204 the appropriate string.
2205 (CPP_CPU_DEFAULT_SPEC, ASM_DEFAULT_SPEC, CPP_FPU_SPEC): Delete.
2206 (CPP_SPEC): Define to NETBSD_CPP_SPEC.
2207 (ASM_SPEC): Don't use %(asm_default_spec).
2208 * config/m68k/m68k.c (TARGET_DEFAULT_TARGET_FLAGS): Remove
2209 TARGET_DEFAULT and add MASK_68881.
2210 * config/m68k/m68k.md: Remove mention of TARGET_DEFAULT from comments.
2212 2007-01-12 Richard Sandiford <richard@codesourcery.com>
2214 * config.gcc (m68010-*-netbsdelf*): Add MASK_68010.
2215 (m68k*-*-netbsdelf*, m68k*-*-openbsd*, m68k*-linux*): Add
2216 MASK_68010 alongside MASK_68020.
2217 * doc/invoke.texi: Document -m68010.
2218 * config/m68k/m68k.opt (m68010): New.
2219 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Define mc68010
2220 if TUNE_68010.
2221 (TUNE_68010): New macro.
2222 * config/m68k/m68k-none.h (M68K_CPU_m68k, M68K_CPU_m68010)
2223 (M68K_CPU_m68020, M68K_CPU_m68030, M68K_CPU_m68040)
2224 (M68K_CPU_m68332): Add MASK_68010.
2225 * config/m68k/linux.h (TARGET_DEFAULT): Add MASK_68010 to
2226 fallback definition.
2227 * config/m68k/netbsd-elf.h (CPP_CPU_SPEC): Remove now-redundant
2228 defines.
2229 * config/m68k/m68k.c (MASK_ALL_CPU_BITS): Add MASK_68010.
2230 (m68k_handle_option): Handle OPT_m68010. Add MASK_68010
2231 to all entries that use MASK_68020.
2232 (output_move_simode_const, output_move_himode, output_move_qimode)
2233 (output_move_stricthi, output_move_strictqi): Use TARGET_68010
2234 instead of TARGET_68020 to select clr behavior. Remove comment
2235 about there being no TARGET_68010.
2236 * config/m68k/m68k.md: Likewise throughout.
2238 2007-01-12 Julian Brown <julian@codesourcery.com>
2240 * config/m68k/m68k.h (TARGET_ISAB): New macro.
2241 * config/m68k/m68k.c: Use TARGET_ISAB rather than TARGET_CFV4.
2242 * config/m68k/m68k.md: Likewise.
2244 2007-01-12 Julian Brown <julian@codesourcery.com>
2246 * config/m68k/m68k.h (LEGITIMATE_INDEX_P, LEGITIMIZE_ADDRESS): Use
2247 TARGET_COLDFIRE_FPU instead of TARGET_CFV4E.
2249 2007-01-12 Julian Brown <julian@codesourcery.com>
2251 * config/m68k/m68k.h (TUNE_68040_60): New macro.
2252 * config/m68k/m68k.c (standard_68881_constant_p): Use it.
2253 * config/m68k/m68k.md: Likewise.
2255 2007-01-12 Julian Brown <julian@codesourcery.com>
2256 Richard Sandiford <richard@codesourcery.com>
2258 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Use TUNE_68030
2259 instead of TARGET_68030, TUNE_68040 instead of TARGET_68040,
2260 TUNE_68060 instead of TARGET_68060 and TUNE_CPU32 instead of
2261 TARGET_CPU32.
2262 (TARGET_CPU32): Rename to...
2263 (TUNE_CPU32): ...this.
2264 (TUNE_68000_10, TUNE_68030, TUNE_68040, TUNE_68060)
2265 (TUNE_CFV2): New macros.
2266 * config/m68k/netbsd-elf.h (LONG_DOUBLE_TYPE_SIZE): Simplify;
2267 remove conditions that are implied by TARGET_68020.
2268 * config/m68k/m68k.c (m68k_output_function_prologue): Use TUNE_68040
2269 instead of TARGET_68040 and TUNE_CPU32 instead of TARGET_CPU32.
2270 (m68k_output_function_epilogue): Likewise.
2271 (m68k_rtx_costs): Likewise. Use TUNE_68060 instead of TARGET_68060
2272 and TUNE_CFV2 instead of TARGET_5200. Use TUNE_68000_10 instead of
2273 "!TARGET_68020 && !TARGET_COLDFIRE" to choose between 68000 and
2274 non-68000 timings. Refactor multiplication and division costs.
2275 (output_addsi3): Use TUNE_68040 instead of TARGET_68040 and
2276 TUNE_CPU32 instead of TARGET_CPU32.
2277 (standard_68881_constant_p): Use TUNE_68040 instead of TARGET_68040
2278 and TUNE_68060 instead of TARGET_68060.
2279 * config/m68k/m68k.md: Use TUNE_68040 instead of TARGET_68040,
2280 TUNE_68060 instead of TARGET_68060, and TUNE_CPU32 instead of
2281 TARGET_CPU32.
2282 (movsi_const0): Use TUNE_68000_10 rather than "!TARGET_68020
2283 && !TARGET_COLDFIRE" to choose between moveq and clr.
2284 Likewise in the unnamed movsf pattern.
2285 (ashlsi_17_24, lshrsi_17_24): Guard with TUNE_68000_10 rather than
2286 "!TARGET_68020 && !TARGET_COLDFIRE". Likewise the unnamed
2287 ashiftrt pattern.
2289 2007-01-12 Richard Sandiford <richard@codesourcery.com>
2291 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Increase amount
2292 of tabbing before backslashes.
2294 2007-01-11 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
2296 * pa-linux.h (ASM_OUTPUT_INTERNAL_LABEL): Undefine.
2297 * pa.h (ASM_OUTPUT_LABEL): Output colon when using GAS.
2298 (ASM_OUTPUT_INTERNAL_LABEL): Define.
2300 2007-01-11 Zdenek Dvorak <dvorakz@suse.cz>
2302 * tree-ssa-loop-ivopts.c (extract_cond_operands): Split from
2303 find_interesting_uses_cond.
2304 (find_interesting_uses_cond): Use extract_cond_operands.
2305 (rewrite_use_compare): Use extract_cond_operands and
2306 force_gimple_operand_bsi. Do not call update_stmt.
2307 (determine_use_iv_cost_condition): Use extract_cond_operands.
2308 Return cheaper of using original bound and changing the exit bound.
2310 2007-01-11 Zdenek Dvorak <dvorakz@suse.cz>
2312 PR tree-optimization/29516
2313 * tree-ssa-address.c (tree_mem_ref_addr, add_to_parts,
2314 most_expensive_mult_to_index, addr_to_parts,
2315 create_mem_ref, maybe_fold_tmr): Make the type of
2316 fields of TARGET_MEM_REF sizetype.
2317 (move_fixed_address_to_symbol, move_pointer_to_base):
2318 New functions.
2319 * tree.def (TARGET_MEM_REF): Add comment on types of
2320 the operands.
2322 2007-01-11 Joseph Myers <joseph@codesourcery.com>
2324 * c-common.c (vector_types_convertible_p): Treat opaque types as
2325 always convertible if they have the same size, but not otherwise.
2327 2007-01-11 Steven Bosscher <steven@gcc.gnu.org>
2329 * ifcvt.c (struct noce_if_info): Add comments to the fields.
2330 Remove the b_unconditional field.
2331 (noce_try_sign_mask): Do not look at b_unconditional.
2332 (noce_process_if_block): Do not use merge_if_blocks. Update
2333 the CFG here. Do not set b_unconditional.
2334 (cond_move_process_if_block): Likewise.
2335 (find_cond_trap): Likewise.
2336 (check_cond_move_block): Require simple jump insns at the end
2337 of the basic block.
2339 2007-01-11 Jan Hubicka <jh@suse.cz>
2341 PR tree-optimization/1046
2342 * tree-tailcall.c (suitable_for_tail_call_opt_p): Use TREE_ADDRESSABLE
2343 when alias info is not ready.
2344 (pass_tail_recursion): Do not require aliasing.
2345 * tree-ssa-copyrename.c (pass_rename_ssa_cop): Likewise.
2346 * tree-ssa-ccp.c (pass_ccp, pass_fold_builtins): Likewise.
2347 * tree-ssa-copy.c (pass_copy_prop): Likewise.
2348 * tree-ssa-forwprop.c (pass_forwprop): Likewise.
2349 * tree-ssa-dce.c (pass_dce, pass_dce_loop, pass_cd_dce): Likewise.
2350 * passes.c (init_optimization_passes): Execute rename_ssa_copies,
2351 ccp, forwprop, copy_prop, merge_phi, copy_prop, dce and tail recursion
2352 before inlining.
2353 * tree-ssa-operands.c (add_virtual_operand, get_indirect_ref_operand):
2354 When aliasing is not build, mark statement as volatile.
2356 2007-01-11 Tom Tromey <tromey@redhat.com>
2358 PR preprocessor/15185, PR preprocessor/20989:
2359 * doc/cppopts.texi <-MT>: Update description of algorithm for
2360 computing default target.
2361 <-M, -MD>: Reword "basename" text.
2363 2007-01-11 Roger Sayle <roger@eyesopen.com>
2365 * builtins.c (expand_builtin_pow, expand_builtin_powi,
2366 fold_builtin_cabs, fold_builtin_sqrt, fold_builtin_trunc,
2367 fold_builtin_floor, fold_builtin_ceil, fold_builtin_round,
2368 fold_builtin_int_int_roundingfn, fold_builtin_bitop,
2369 fold_builtin_bswap, real_constp, fold_builtin_pow,
2370 fold_builtin_powi, fold_builtin_signbit, fold_builtin_copysign,
2371 do_mpfr_arg1, do_mpfr_arg2, do_mpfr_arg3, do_mpfr_sincos): Replace
2372 uses of the macro TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
2373 * convert.c (convert_to_pointer): Likewise.
2374 * expr.c (highest_pow2_factor, expand_expr_real_1): Likewise.
2375 * fold-const.c (force_fit_type, fold_negate_expr, int_const_binop,
2376 const_binop, fold_convert_const_int_from_int,
2377 fold_convert_const_int_from_real,
2378 fold_convert_const_real_from_real, sign_bit_p,
2379 optimize_minmax_comparison, extract_muldiv_1, fold_div_compare,
2380 fold_sign_changed_comparison, fold_unary, fold_comparison,
2381 fold_binary, multiple_of_p, tree_Expr_non_zero_p,
2382 fold_negate_const, fold_abs_const, fold_not_const): Likewise.
2383 * print-tree.c (print_node_brief, print_node): Likewise.
2384 * stor-layout.c (place_field, layout_type): Likewise.
2385 * tree-chrec.c (keep_cast): Likewise.
2386 * tree.c (build_vector, build_real, build_real_from_int_cst,
2387 build_complex): Likewise.
2389 2007-01-11 Roger Sayle <roger@eyesopen.com>
2391 * tree.h (TREE_CONSTANT_OVERFLOW): Obsolete. For the time being,
2392 treat TREE_CONSTANT_OVERFLOW as a synonym of TREE_OVERFLOW.
2394 2007-01-11 Paolo Bonzini <bonzini@gnu.org>
2396 * configure.ac (strict1_warn): Rename to strict_warn.
2397 (WERROR, --enable-werror, symlink hacks, stage1_cflags,
2398 cc_set_by_configure, quoted_cc_set_by_configure,
2399 stage_prefix_set_by_configure, quoted_stage_prefix_set_by_configure,
2400 all_boot_languages, all_stagestuff): Remove.
2401 (target_list): Remove bootstrap targets.
2402 * Makefile.in (quickstrap): Unconditionally make a synonym of all.
2403 (BOOT_LANGUAGES, STAGE1_CFLAGS, STAGE1_CHECKING,
2404 REMAKEFLAGS, FLAGS_TO_PASS, PREPEND_DOTDOT_TO_RELATIVE_PATHS,
2405 SUBDIR_FLAGS_TO_PASS, WERROR_FLAGS, STRICT2_WARN, LANG_STAGESTUFF,
2406 VOL_FILES, POSTSTAGE1_FLAGS_TO_PASS, STAGE2_FLAGS_TO_PASS,
2407 STAGEPROFILE_FLAGS_TO_PASS, STAGEFEEDBACK_FLAGS_TO_PASS, stage1_build,
2408 stage1_copy, stage2_build, stage2_copy, stageprofile_build,
2409 stageprofile_copy, stage3_build, stage3_copy, stagefeedback_build,
2410 stagefeedback_copy, stage4_build, clean_s1, clean_sw, bootstrap,
2411 bootstrap-lean, bootstrap2, bootstrap2-lean, bootstrap3,
2412 bootstrap3-lean, bootstrap4, bootstrap4-lean, unstage1, unstage2,
2413 unstage3, unstage4, unstageprofile, unstagefeedback, restage, restage2,
2414 restage3, restage4, restageprofile, restagefeedback, bubbleestrap,
2415 cleanstrap, unstrap, restrap, *compare, *compare3, *compare4,
2416 *compare-lean, *compare3-lean, *compare4-lean, stage1-start, stage1,
2417 stage2-start, stage2, stage3-start, stage3, stage4-start, stage4,
2418 stageprofile-start, stageprofile, stagefeedback-start, stagefeedback,
2419 risky-stage1, risky-stage2, risky-stage3, risky-stage4): Remove.
2420 (ORDINARY_FLAGS_TO_PASS): Rename to FLAGS_TO_PASS.
2421 (STAGECOPYSTUFF, STAGEMOVESTUFF): Consolidate into MOSTLYCLEANFILES.
2422 (mostlyclean): Adjust.
2423 (clean, distclean): Don't mention bootstrap stuff.
2424 * configure: Regenerate.
2425 * ada/config-lang.in, cp/config-lang.in, forttran/config-lang.in,
2426 java/config-lang.in, objc/config-lang.in, objcp/config-lang.in,
2427 treelang/config-lang.in (stagestuff): Remove.
2428 * doc/sourcebuild.texi (stage1, stage2, stage3, stage4,
2429 stageprofile, stagefeedback, stagestuff): Remove mention.
2431 2007-01-11 Nick Clifton <nickc@redhat.com>
2433 * config/mcore/predicates.md (mcore_general_movesrc_operand):
2434 Accept CONSTs.
2435 (mcore_general_movdst_operand): Do not accept CONST_INTs.
2436 (mcore_arith_K_S_operand): Run the test for the S constraint not
2437 the test for the M constraint.
2438 (mcore_addsub_operand): Do not accept integer values that are
2439 larger than 32 bits.
2440 * config/mcore/mcore.md: Remove unused constraints from split.
2441 (andsi3): Use HOST_WIDE_INT instead of int to hold an INTVAL.
2442 (addsi3): Likewise.
2443 (allocate_stack): Likewise.
2444 * config/mcore/mcore.c (mcore_print_operand): Restrict output of P
2445 operands to 32 bits.
2446 (mcore_const_costs): Use HOST_WIDE_INT instead of int to hold an
2447 INTVAL.
2448 (mcore_and_cost, mcore_modify_comparison, const_ok_for_mcore,
2449 mcore_const_ok_for_inline, mcore_const_trick_uses_not,
2450 try_constant_tricks, mcore_num_ones, mcore_num_zeros,
2451 mcore_output_bclri, mcore_output_andn, output_inline_const,
2452 mcore_output_move, mcore_output_movedouble): Likewise.
2453 (mcore_output_cmov): Use CONST_OK_FOR_M and CONST_OK_FOR_N.
2454 (output_inline_const): Likewise.
2455 (output_inline_const): Fix format strings used in sprintf
2456 statements.
2457 * config/mcore/mcore-protos.h: Update prototypes for changed
2458 functions in mcore.c.
2459 * config/mcore/mcore.h (CONST_OK_FOR_I): Cast values to
2460 HOST_WIDE_INT and not int.
2461 (CONST_OK_FOR_J, CONST_OK_FOR_K, CONST_OK_FOR_L, CONST_OK_FOR_M,
2462 CONST_OK_FOR_N): Likewise.
2463 (LEGITIMATE_CONSTANT_P): Also check CONSTANT_P.
2464 (GO_IF_LEGITIMATE_INDEX): Use HOST_WIDE_INT instead of int to hold
2465 an INTVAL.
2467 2007-01-10 Jan Hubicka <jh@suse.cz>
2469 * tree-vrp.c (remove_range_assertions): Release defs.
2470 * tree-ssa-loop-ivopts.c (rmeove_statement): Likewise.
2471 * tree-ssa-dom.c (remove_stmt_or_phi): Likewise.
2473 2007-01-10 Paul Brook <paul@codesourcery.com>
2475 * config/arm/arm.c (arm_rtx_costs_1): Handle mutiply-subtract.
2476 * config/arm/arm.md (mulsi3subsi): New insn.
2478 2007-01-10 Zdenek Dvorak <dvorakz@suse.cz>
2480 * tree-ssa-loop-manip.c (tree_unroll_loop): Make it a wrapper over ...
2481 (tree_transform_and_unroll_loop): New.
2482 * tree-flow.h (transform_callback, tree_transform_and_unroll_loop):
2483 Declare.
2485 2007-01-10 Robert Kennedy <jimbob@google.com>
2487 * fold-const.c (fold_comparison): Fold comparisons like (x *
2488 1000 < 0) to (x < 0).
2490 2007-01-10 Ian Lance Taylor <iant@google.com>
2492 * tree-pretty-print.c (dump_generic_node): Print parentheses when
2493 operands have the same priority.
2495 2007-01-10 Tom Tromey <tromey@redhat.com>
2497 * fold-const.c (fold_truthop): Don't check can_use_bit_fields_p.
2498 (fold_binary): Likewise.
2499 * langhooks.c (lhd_can_use_bit_fields_p): Removed.
2500 * langhooks-def.h (lhd_can_use_bit_fields_p): Removed.
2501 (LANG_HOOKS_CAN_USE_BIT_FIELDS_P): Removed.
2502 (LANG_HOOKS_INITIALIZER): Remove LANG_HOOKS_CAN_USE_BIT_FIELDS_P.
2503 * langhooks.h (struct lang_hooks): Removed field
2504 'can_use_bit_fields_p'.
2506 2007-01-10 Ralf Corsépius <ralf.corsepius@rtems.org>
2508 * config/bfin/t-bfin, config/bfin/t-bfin-elf: Remove GCC_CFLAGS.
2510 2007-01-10 Razya Ladelsky <razya@il.ibm.com>
2512 * function.c (get_last_funcdef_no): New function.
2513 * function.h (get_last_funcdef_no): Declare.
2514 * tree-inline.c (initialize_cfun): Add initialization.
2515 (tree_function_versioning): Cleanup.
2517 2007-01-10 Jan Hubicka <jh@suse.cz>
2519 * tree-inline.c (setup_one_parameter): Do not propagate into abnormal
2520 PHIs.
2522 2007-01-10 Sa Liu <saliu@de.ibm.com>
2523 Ben Elliston <bje@au.ibm.com>
2525 * spu.h (STACK_SAVE_AREA): Use VOIDmode for SAVE_FUNCTION, SImode
2526 for SAVE_NONLOCAL and Pmode for any other save level.
2527 * spu-protos.h (spu_restore_stack_block): Declare.
2528 * spu.md (save_stack_block): Remove.
2529 (restore_stack_block): Call spu_restore_stack_block.
2530 * spu.c (spu_restore_stack_block): New function.
2531 (spu_expand_epilogue): Remove old comment.
2533 2007-01-09 Zdenek Dvorak <dvorakz@suse.cz>
2535 PR tree-optimization/30322
2536 * tree-ssa-loop-ivopts.c (fold_affine_expr, iv_value): Removed.
2537 (cand_value_at): Return the value as aff_tree.
2538 (may_eliminate_iv): Convert the bound from aff_tree to tree.
2539 * tree-affine.c (aff_combination_add_cst, aff_combination_add_product,
2540 aff_combination_mult): New functions.
2541 (aff_combination_add): Use aff_combination_add_cst.
2542 (aff_combination_convert): Allow conversions to a wider type.
2543 (tree_to_aff_combination): Handle BIT_NOT_EXPR.
2544 * tree-affine.h (aff_combination_mult): Declare.
2546 2007-01-09 Carlos O'Donell <carlos@codesourcery.com>
2548 * doc/tm.texi: Update documentation to reflect reality of exec
2549 and start file search behaviours. Update copyright year.
2550 * doc/invoke.texi: Explain how GCC_EXEC_PREFIX is used to find
2551 header file directories.
2553 2007-01-09 Uros Bizjak <ubizjak@gmail.com>
2555 * config/i386/i386.md (*sinxf2): Rename to *sinxf2_i387.
2556 (*cosxf2): Rename to cosxf2_i387.
2557 (*sindf2, *sinsf2): Extend operand 1 to XFmode. Macroize patterns
2558 using X87MODEF12 mode macro. Rename patterns to
2559 *sin_extend<mode>xf2_i387. Use SSE_FLOAT_MODE_P to disable patterns
2560 for SSE math.
2561 (*cosdf2, *cossf2): Ditto.
2562 (sincosdf3, sincossf3): Ditto. Rewrite corresponding splitters
2563 to match extended input operands.
2564 (sincos<mode>3): New expander.
2565 (*sinextendsfdf2, *cosextendsfdf2, *sincosextendsfdf3): Remove
2566 insn patterns and corresponding splitters.
2568 2007-01-09 Kaz Kojima <kkojima@gcc.gnu.org>
2570 * config/sh/t-linux (TARGET_LIBGCC2_CFLAGS): Delete.
2571 (SHLIB_MAPFILES, SHLIB_LINK, SHLIB_INSTALL): Likewise.
2573 2007-01-09 Nicolas Pitre <nico@cam.org>
2575 PR target/30173
2576 * arm/ieee754-df.S (Lad_s): Also test the low word of X for zero.
2578 2007-01-08 Geoffrey Keating <geoffk@apple.com>
2580 * target.h (struct gcc_target): New field library_rtti_comdat.
2581 * target-def.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): New.
2582 (TARGET_CXX): Add TARGET_CXX_LIBRARY_RTTI_COMDAT.
2583 * doc/tm.texi (C++ ABI): Document TARGET_CXX_LIBRARY_RTTI_COMDAT.
2584 * config/darwin.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
2586 2007-01-08 Geoffrey Keating <geoffk@apple.com>
2588 * doc/invoke.texi (Optimize Options): Correct description of -O0.
2590 2007-01-08 Richard Guenther <rguenther@suse.de>
2592 * tree.h (force_fit_type_double): Export.
2593 (force_fit_type): Remove.
2594 * fold-const.c (force_fit_type_double): New function.
2595 (force_fit_type): Remove.
2596 (int_const_binop): Use it.
2597 (fold_convert_const_int_from_int): Likewise.
2598 (fold_convert_const_int_from_real): Likewise.
2599 (fold_div_compare): Likewise.
2600 (fold_sign_changed_comparison): Likewise.
2601 (fold_unary): Likewise.
2602 (fold_negate_const): Likewise.
2603 (fold_abs_const): Likewise.
2604 (fold_not_const): Likewise.
2605 * c-common.c (shorten_compare): Use force_fit_type_double.
2606 * convert.c (convert_to_pointer): Likewise.
2608 2007-01-08 Richard Guenther <rguenther@suse.de>
2610 * tree.h (build_int_cst_wide_type): Export.
2611 * tree.c (build_int_cst_wide_type): New function.
2612 (build_int_cst_wide): Fix comment.
2613 * builtins.c (fold_builtin_object_size): Use build_int_cst
2614 to build -1 or 0 of the correct type. Use fit_double_type
2615 to check for overflow.
2616 * fold-const.c (optimize_bit_field_compare): Use build_int_cst_type
2617 to build the mask.
2618 (decode_field_reference): Likewise.
2619 (all_ones_mask_p): Likewise.
2620 (native_interpret_int): Use build_int_cst_wide_type.
2621 (fold_binary): Use build_int_cst_type to build an all-ones
2622 value.
2623 * stor-layout.c (set_sizetype): Use build_int_cst_wide_type.
2625 2007-01-08 Daniel Jacobowitz <dan@codesourcery.com>
2627 * config/pa/t-pa64 (libgcc_stub.a): Use $(T).
2629 2007-01-09 Ben Elliston <bje@au.ibm.com>
2631 * genautomata.c (STATS_OPTION): New option.
2632 (stats_flag): New flag.
2633 (gen_automata_option): Handle it.
2634 (initiate_automaton_gen): Ditto.
2635 (write_automata): Output statistics only if stats_flag is
2636 set. Likewise, output time statistics only if time_flag is set.
2637 * doc/md.texi (Processor pipeline description): Document new flag.
2639 2007-01-08 Richard Guenther <rguenther@suse.de>
2641 * builtins.c (fold_builtin_int_roundingfn): Use fit_double_type.
2642 * tree.c (build_int_cst_type): Likewise.
2643 (size_in_bytes): Don't call force_fit_type on the result.
2644 (int_fits_type_p): Use fit_double_type.
2645 * fold-const.c (fit_double_type): New function.
2646 (force_fit_type): Use it.
2647 * tree.h (fit_double_type): Export.
2649 2007-01-08 Jan Hubicka <jh@suse.cz>
2651 * tree-vectorizer.c (gate_increase_alignment): Fix return type.
2652 * ipa.c (function_and_variable_visibility): Fix return type.
2654 2007-01-08 Richard Guenther <rguenther@suse.de>
2656 * tree-ssa-ccp.c (maybe_fold_offset_to_array_ref): Use type
2657 of offset to build the index.
2658 * tree-pretty-print.c (dump_generic_node): Don't build negated
2659 const just for printing.
2660 * c-pretty-print.c (pp_c_integer_constant): Likewise.
2661 * builtins.c (fold_builtin_int_roundingfn): Check if result
2662 fits the type by using force_fit_type and comparing the result.
2663 * predict.c (predict_loops): Use compare_tree_int for comparison.
2664 * tree.c (build_int_cst): Fall back to integer_type_node for
2665 NULL_TREE type.
2666 (build_int_cst_wide): Assert type is non-null.
2668 2007-01-08 Roberto Costa <roberto.costa@st.com>
2670 * tree-vrp.c (extract_range_from_cond_expr): New.
2671 (extract_range_from_expr): Handle COND_EXPR nodes used as expressions.
2672 * tree-ssa-ccp.c (get_maxval_strlen): Handle COND_EXPR nodes used
2673 as expressions.
2674 (fold_stmt): Bug fix, avoid infinite recursion when folding COND_EXPRs.
2675 * tree-ssa-forwprop.c (simplify_cond, forward_propagate_into_cond,
2676 tree_ssa_forward_propagate_single_use_vars): Handle COND_EXPR nodes
2677 used as expressions.
2678 * tree-object-size.c (cond_expr_object_size): New.
2679 (collect_object_sizes_for): Handle COND_EXPR nodes used as expressions.
2681 2007-01-08 Jan Hubicka <jh@suse.cz>
2683 * tree-ssa-forwprop.c (forward_propagate_into_cond,
2684 tree_ssa_forward_propagate_single_use_va): Release defs of propagated
2685 statement.
2687 2007-01-08 Richard Guenther <rguenther@suse.de>
2689 PR tree-optimization/23603
2690 * tree-vrp.c (set_value_range_to_truthvalue): New function.
2691 (extract_range_from_binary): Fall back to truthvalue instead of
2692 varying for TRUTH_*_EXPR.
2693 (extract_range_from_comparison): Fall back to truthvalue instead of
2694 varying.
2695 (vrp_visit_phi_node): Don't adjust new range bounds to +INF/-INF
2696 if all visited PHI values were constant.
2698 2007-01-08 Jan Hubicka <jh@suse.cz>
2700 * cgraphunit.c (cgraph_process_new_functions): Reset reachable flag.
2701 (cgraph_analyze_function): break out from ...
2702 (cgraph_finalize_compilation_unit): ... here.
2703 (cgraph_expand_function): Remove forgoten commented out line.
2704 (cgraph_optimize): Analyze functions.
2706 2007-01-08 Jan Hubicka <jh@suse.cz>
2708 * tree-pas.h (TODO_remove_function): New flag.
2709 (TODO_update*): Renumber.
2710 (pass_ipa_increase_alignment,
2711 pass_ipa_function_and_variable_visibility): New passes.
2712 * cgraphunit.c (cgraph_increase_alignment): Move to tree-vectorizer.c
2713 (cgraph_function_and_variable_visibility): Move to ipa.c
2714 (cgraph_optimize): Don't call cgraph_function_and_variable_visibility,
2715 cgraph_increase_alignment.
2716 * ipa-inline.c (cgraph_decide_inlining): Don't push timevar.
2717 (cgraph_decide_inlining_incrementally): Push TV_INTEGRATION before
2718 calling tree-inline.
2719 (cgraph_early_inlining): Do not call cgraph_remove_unreachable_nodes.
2720 (pass_ipa_inline, pass_early_ipa_inlining): Set TODO_remove_functions
2721 * tree-vectorizer.c (increase_alignment): Move here from cgraphunit.c
2722 (gate_increase_alignment): New function.
2723 (pass_ipa_increase_alignment): New pass.
2724 * ipa.c: Inline tree-pass.h and timevar.h
2725 (function_and_variable_visibility): Move here from cgraphunit.c
2726 * tree-optimize.c (pass_early_local_passes): Add TODO_remove_functions.
2727 * passes.c (init_optimization_passes): Add the two new passes.
2728 (execute_todo): Handle cgraph_remove_functions.
2730 2007-01-08 Nick Clifton <nickc@redhat.com>
2732 * config/frv/predicates.md (reg_or_0_operand): Accept
2733 CONST_DOUBLEs.
2735 2007-01-08 Ralf Corsépius <ralf.corsepius@rtems.org>
2737 * config/bfin/rtems.h, config/bfin/t-rtems: New.
2738 * config.gcc: Add bfin*-rtems*.
2740 2007-01-08 Mark Shinwell <shinwell@codesourcery.com>
2742 * c.opt: Add -flax-vector-conversions.
2743 * c-typeck.c (convert_for_assignment): Pass flag to
2744 vector_types_convertible_p to allow emission of note.
2745 (digest_init): Likewise.
2746 * c-opts.c: Handle -flax-vector-conversions.
2747 * c-common.c (flag_lax_vector_conversions): New.
2748 (vector_types_convertible_p): Unless -flax-vector conversions
2749 has been passed, disallow conversions between vectors with
2750 differing numbers of subparts and/or element types. If such
2751 a conversion is disallowed, possibly emit a note on the first
2752 occasion only to inform the user of -flax-vector-conversions.
2753 The new last argument specifies this.
2754 * c-common.h (flag_lax_vector_conversions): New.
2755 (vector_types_convertible_p): Add extra argument.
2756 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Use
2757 char_type_node for V*QI type vectors.
2758 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins):
2759 Update to satisfy new typechecking rules.
2760 * config/rs6000/altivec.h (vec_cmple): Use vec_cmpge, for both
2761 C and C++ variants.
2762 * doc/invoke.texi (C Dialect Options): Document
2763 -flax-vector-conversions.
2765 2007-01-08 Mark Shinwell <shinwell@codesourcery.com>
2767 PR tree-optimization/29877
2768 * tree-ssa-ter.c (is_replaceable_p): Deem assignments with
2769 a register variable on the RHS to not be replaceable.
2771 2007-01-08 Chen Liqin <liqin@sunnorth.com.cn>
2772 * config/score/t-score-elf (MULTILIB_OPTIONS): Change.
2773 * config/score/predicates.md (const_uimm5, sr0_operand, const_simm12,
2774 const_simm15, const_pow2, const_npow2): Added.
2775 * config/score/misc.md (insv, extv, extzv, movmemsi,
2776 move_lbu_a/b, mov_lhu_a/b etc): Added and fix some bug.
2777 * config/score/score.c (score_address_cost, score_select_cc_mode):
2778 Added.
2779 Change CONST_OK_FOR_LETTER_P/EXTRA_CONSTRAINT define.
2780 Update score_rtx_costs for MACRO TARGET_RTX_COSTS.
2781 Update score_print_operand.
2782 * config/score/score.h (DATA_ALIGNMENT, SELECT_CC_MODE): Added.
2783 Adjust register allocate order and update some macro define.
2784 * config/score/score-mdaux.c (mdx_unaligned_load, mdx_unsigned_store,
2785 mdx_block_move_straight, mdx_block_move_loop_head,
2786 mdx_block_move_loop_body, mdx_block_move_loop_foot, mdx_block_move_loop,
2787 mdx_block_move): Added.
2788 (mdx_movsicc, mdp_select_add_imm, mdp_select, mds_zero_extract_andi,
2789 mdp_limm): Updated and fix some bug and typo.
2790 * config/score/score.md (movqi/hi/si, add/sub/zero/ext): Updated.
2791 (movsf, movdf, doloop_end): Added.
2793 2007-01-08 Kazu Hirata <kazu@codesourcery.com>
2795 * config/arm/arm.c, config/arm/arm.h, config/arm/arm.md,
2796 config/arm/thumb2.md: Fix comment typos.
2797 * doc/extend.texi: Fix a typo.
2799 2007-01-07 Eric Christopher <echristo@apple.com>
2801 * configure.ac: Check for __stack_chk_fail for darwin.
2802 * configure: Regenerate.
2804 2007-01-07 Richard Guenther <rguenther@suse.de>
2806 * tree-vrp.c (extract_range_from_assert): CSE calls to
2807 compare_values where possible.
2808 (extract_range_from_unary_expr): Likewise.
2810 2007-01-07 Anatoly Sokolov <aesok@post.ru>
2812 * config/avr/avr-protos.h (call_insn_operand): Delete prototype.
2813 * config/avr/avr.c (call_insn_operand): Delete function.
2814 * config/avr/avr.md (*pushqi, *pushhi, *pushsi, *pushsf): Use REG_SP
2815 instead of register number. Use predicates.
2816 * config/avr/predicates.md (const0_operand, reg_or_0_operand,
2817 call_insn_operand): Add.
2819 2007-01-06 Jan Hubicka <jh@suse.cz>
2821 * tree-pass.h (pass_build_cgraph_edges): Declare.
2822 * cgraphunit.c (record_refernece): Move to cgraphbuild.c
2823 (visited_nodes): Remove.
2824 (cgraph_create_edges): Move to cgraphbuild.c; rename to
2825 build_cgrpah_edges; make visited_nodes local.
2826 (cgraph_process_new_functions): DO not call initialize_inline_failed.
2827 (record_references_in_initializer): Move to cgraphbuild.c
2828 (initialize_inline_failed, rebuild_cgraph_edges,
2829 pass_rebuild_cgraph_edges): Move to cgraphbuild.c.
2830 (verify_cgraph_node): Make visited_nodes local.
2831 (cgraph_analyze_function): Do not call cgraph_create_edges and
2832 initialize_inline_failed.
2833 (cgraph_expand_function): Do not call cgraph_lower_function;
2834 assert that function is already lowered.
2835 * Makefile.in (cgraphbuild.o): New.
2836 * passes.c (init_optimization_passes): Add pass_build_cgraph_edges
2837 at the end of lowering passes.
2839 2007-01-06 Steven Bosscher <steven@gcc.gnu.org>
2841 * ifcvt.c (cond_move_convert_if_block): New function, code
2842 factored out from...
2843 (cond_move_process_if_block): ...here. Call the new function
2844 on the THEN and ELSE blocks.
2845 (merge_if_block): Do not copy global_live_at_end, merge_blocks
2846 already takes care of this.
2848 2007-01-05 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
2850 PR c/19978
2851 * tree.h (TREE_OVERFLOW_P): New.
2852 * c-typeck.c (parser_build_unary_op): Warn only if result
2853 overflowed and operands did not.
2854 (parser_build_binary_op): Likewise.
2855 (convert_for_assignment): Remove redundant overflow_warning.
2856 * c-common.c (overflow_warning): Don't check or set TREE_OVERFLOW.
2858 2007-01-05 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
2860 * c-typeck.c (store_init_value): Split over two lines to follow
2861 the GNU coding style.
2863 2007-01-05 Benjamin Kosnik <bkoz@redhat.com>
2865 * c-cppbuiltin.c (c_cpp_builtins): __GXX_EXPERIMENTAL_CPP0X__ to
2866 __GXX_EXPERIMENTAL_CXX0X__.
2867 * doc/cpp.texi: Same.
2869 2007-01-05 Richard Guenther <rguenther@suse.de>
2871 PR middle-end/27826
2872 * tree.c (get_narrower): Do not construct COMPONENT_REFs
2873 with mismatched types. Instead explicitly build a
2874 conversion NOP_EXPR.
2876 2007-01-05 Ian Lance Taylor <iant@google.com>
2878 * c-common.c (decl_with_nonnull_addr_p): New function.
2879 (c_common_truthvalue_conversion): Call it.
2880 * c-typeck.c (build_binary_op): Likewise.
2881 * c-common.h (decl_with_nonnull_addr_p): Declare.
2883 2007-01-05 Jakub Jelinek <jakub@redhat.com>
2885 PR c/30360
2886 * libgcc2.c (__divdc3): Compare c and d against 0.0 instead of
2887 denom against 0.0.
2889 2007-01-05 Joel Brobecker <brobecker@adacore.com>
2891 * doc/install.texi (Final install): Document the fact that
2892 the GNAT runtime should not be stripped.
2894 2007-01-04 Jan Hubicka <jh@suse.cz>
2896 * tree-inline.c (fold_marked_statements): Update operand caches
2897 and EH after folding
2899 2007-01-04 Ian Lance Taylor <iant@google.com>
2901 * c-common.c (check_function_nonnull): Whitespace fix.
2903 2007-01-04 Jan Hubicka <jh@suse.cz>
2905 * tree-optimize.c (execute_fixup_cfg): Correct previously mistakely
2906 comitted older version of patch.
2907 (pass_fixup_cfg): Add TODOs to verify flow and statements, dump
2908 function, celanup cfg and collect garbage.
2910 2007-01-04 Mike Stump <mrs@apple.com>
2912 * Makefile.in (mostlyclean): Don't remove libgcc anymore.
2913 (clean): Likewise.
2915 2007-01-04 Eric Christopher <echristo@apple.com>
2917 * libgcc2.c (__bswapsi2): Use SItype.
2918 (__bswapdi2): Use DItype.
2919 * libgcc2.h: Update for above.
2921 2007-01-04 Paul Brook <paul@codesourcery.com>
2923 * config/arm/arm.md (arm_mulsi3, thumb_mulsi3, mulsi3_compare0,
2924 mulsi_compare0_scratch, mulsi3addsi, mulsi3addsi_compare0,
2925 mulsi3addsi_compare0_scratch, mulsidi3adddi, mulsidi3,
2926 umulsidi3, umulsidi3adddi, smulsi3_highpart,
2927 umulsi3_highpart): Make conditional on !arm_arch6.
2928 (arm_mulsi3_v6, thumb_mulsi3_v6, mulsi3_compare0_v6,
2929 mulsi_compare0_scratch_v6, mulsi3addsi_v6, mulsi3addsi_compare0_v6,
2930 mulsi3addsi_compare0_scratch_v6, mulsidi3adddi_v6, mulsidi3_v6,
2931 umulsidi3_v6, umulsidi3adddi_v6, smulsi3_highpart_v6,
2932 umulsi3_highpart_v6): New insns.
2934 2007-01-04 Roger Sayle <roger@eyesopen.com>
2936 * fold-const.c (fold_convert): When casting an expression to void,
2937 fold_ignored_result may discover a GIMPLE_MODIFY_STMT which doesn't
2938 have a type. Instead of attempting to build a NOP_EXPR, return
2939 these "special" trees directly.
2941 2007-01-04 Joseph Myers <joseph@codesourcery.com>
2943 * config/rs6000/rs6000.c (rs6000_rtx_costs): Make adjustment for
2944 MULT inside MINUS as either argument. Use rs6000_cost->dmul -
2945 rs6000_cost->fp not 0 as adjustment for outer NEG.
2947 2007-01-04 Jan Hubicka <jh@suse.cz>
2949 * cgraph.c (cgraph_release_function_body): New function.
2950 (cgraph_remove_node): Use it.
2951 * cgraph.h (cgraph_release_function_body): Declare.
2952 * cgraphunit.c (cgraph_expand_function): Use it.
2953 * ipa.c (cgraph_remove_unreahchable_nodes): Use it.
2954 * tree-ssa.c (delete_tree_ssa): Allow to be called before aliasing
2955 is initialized and while compilation of other function is running.
2956 * tree-optimize.c (execute_free_cfg_annotations): Move code to clear
2957 statement CFG annotations from here to ...
2958 * tree-cfg.c (delete_tree_cfg_annotations): ... here.
2960 2007-01-04 Zdenek Dvorak <dvorakz@suse.cz>
2962 * cfgloop.h (enum li_flags): Make the constants powers of two.
2964 2007-01-04 Jan Hubicka <jh@suse.cz>
2966 * tree-inline.c (copy_bb): Insert new statements to statements_to_fold
2967 set.
2968 (fold_marked_statements): New function.
2969 (optimize_inline_calls, tree_function_versioning): Fold new statements.
2970 * tree-inline.h (copy_body_data): Add statements_to_fold.
2972 2007-01-03 Daniel Jacobowitz <dan@codesourcery.com>
2974 * config.gcc: Mention libgcc/config.host.
2975 * Makefile.in: Update comments mentioning libgcc.
2976 (LIBGCC, INSTALL_LIBGCC, GCC_PARTS, mklibgcc): Delete.
2977 (all.cross, start.encap, rest.encap, rest.cross): Update
2978 dependencies for libgcc move.
2979 (libgcc.mk, LIBGCC_DEPS, libgcov.a, libgcc.a, stmp-multilib)
2980 (clean-target, clean-target-libgcc): Delete.
2981 (srcdirify, GCC_EXTRA_PARTS): New macros.
2982 (libgcc-support, libgcc.mvars): New rules.
2983 (distclean): Remove mention of mklibgcc.
2984 (install): Don't reference INSTALL_LIBGCC.
2985 (install-common): Don't reference EXTRA_PARTS.
2986 (install-libgcc, install-multilib): Delete rules.
2987 * mklibgcc.in: Delete file.
2988 * doc/configfiles.texi: Don't mention mklibgcc.
2990 * config/i386/t-darwin (SHLIB_VERPFX): Delete (moved to libgcc).
2991 * config/i386/t-darwin64 (SHLIB_VERPFX): Likewise.
2992 * config/rs6000/t-darwin (SHLIB_VERPFX): Likewise.
2993 * config/rs6000/t-ppccomm (TARGET_LIBGCC2_CFLAGS, SHLIB_MAPFILES)
2994 (mklibgcc, ldblspecs): Likewise.
2996 * config/i386/t-nwld (libgcc.def, libc.def, libpcre.def)
2997 (posixpre.def): Use $(T).
2998 (SHLIB_EXT, SHLIB_NAME, SHLIB_SLIBDIR_QUAL, SHLIB_DEF, SHLIB_MAP)
2999 (SHLIB_SRC, SHLIB_INSTALL): Delete.
3000 (SHLIB_LINK): Make dummy.
3001 * config/t-slibgcc-darwin: Delete contents except for dummy SHLIB_LINK.
3003 * config/frv/t-linux (EXTRA_MULTILIB_PARTS): Clear.
3005 * config/alpha/t-crtfm: Use $(T) in rules for EXTRA_PARTS.
3006 * config/alpha/t-vms, config/alpha/t-vms64, config/fr30/t-fr30,
3007 config/i386/t-rtems-i386, config/ia64/t-ia64, config/rs6000/t-beos,
3008 config/rs6000/t-newas, config/sparc/t-elf: Likewise.
3010 * configure.ac (all_outputs): Remove mklibgcc.
3011 * configure: Regenerated.
3013 2007-01-03 Josh Conner <jconner@apple.com>
3015 PR middle-end/29683
3016 * calls.c (compute_argument_addresses): Set stack and stack_slot
3017 for partial args, too.
3018 (store_one_arg): Use locate.size.constant for the size when
3019 generating a save_area.
3021 2007-01-03 Robert Kennedy <jimbob@google.com>
3023 * tree-cfg.c (tree_merge_blocks): Release SSA_NAME phi results
3024 whose definitions are deleted due to basic block merging.
3026 2007-01-03 Paul Brook <paul@codesourcery.com>
3028 PR target/16634
3029 * config/arm/arm.c (output_return_instruction): Pop PC in interrupt
3030 functions.
3031 (use_return_insn): Return 0 for Thumb interrupt functions.
3032 (print_multi_reg): Add rfe argument for IRQ returns.
3033 (arm_output_epilogue): Pop interrupt return address directly into PC.
3034 (arm_expand_prologue): Only adjust IRQ return address in Arm mode.
3036 2007-01-03 Paul Brook <paul@codesourcery.com>
3038 Merge from sourcerygxx-4_1.
3039 * config/arm/thumb2.md: New file.
3040 * config/arm/elf.h (JUMP_TABLES_IN_TEXT_SECTION): Return True for
3041 Thumb-2.
3042 * config/arm/coff.h (JUMP_TABLES_IN_TEXT_SECTION): Ditto.
3043 * config/arm/aout.h (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
3044 (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump tables.
3045 * config/arm/aof.h (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump
3046 tables.
3047 (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
3048 * config/arm/ieee754-df.S: Use macros for Thumb-2/Unified asm
3049 comptibility.
3050 * config/arm/ieee754-sf.S: Ditto.
3051 * config/arm/arm.c (thumb_base_register_rtx_p): Rename...
3052 (thumb1_base_register_rtx_p): ... to this.
3053 (thumb_index_register_rtx_p): Rename...
3054 (thumb1_index_register_rtx_p): ... to this.
3055 (thumb_output_function_prologue): Rename...
3056 (thumb1_output_function_prologue): ... to this.
3057 (thumb_legitimate_address_p): Rename...
3058 (thumb1_legitimate_address_p): ... to this.
3059 (thumb_rtx_costs): Rename...
3060 (thumb1_rtx_costs): ... to this.
3061 (thumb_compute_save_reg_mask): Rename...
3062 (thumb1_compute_save_reg_mask): ... to this.
3063 (thumb_final_prescan_insn): Rename...
3064 (thumb1_final_prescan_insn): ... to this.
3065 (thumb_expand_epilogue): Rename...
3066 (thumb1_expand_epilogue): ... to this.
3067 (arm_unwind_emit_stm): Rename...
3068 (arm_unwind_emit_sequence): ... to this.
3069 (thumb2_legitimate_index_p, thumb2_legitimate_address_p,
3070 thumb1_compute_save_reg_mask, arm_dwarf_handle_frame_unspec,
3071 thumb2_index_mul_operand, output_move_vfp, arm_shift_nmem,
3072 arm_save_coproc_regs, thumb_set_frame_pointer, arm_print_condition,
3073 thumb2_final_prescan_insn, thumb2_asm_output_opcode, arm_output_shift,
3074 thumb2_output_casesi): New functions.
3075 (TARGET_DWARF_HANDLE_FRAME_UNSPEC): Define.
3076 (FL_THUMB2, FL_NOTM, FL_DIV, FL_FOR_ARCH6T2, FL_FOR_ARCH7,
3077 FL_FOR_ARCH7A, FL_FOR_ARCH7R, FL_FOR_ARCH7M, ARM_LSL_NAME,
3078 THUMB2_WORK_REGS): Define.
3079 (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv, arm_condexec_count,
3080 arm_condexec_mask, arm_condexec_masklen)): New variables.
3081 (all_architectures): Add armv6t2, armv7, armv7a, armv7r and armv7m.
3082 (arm_override_options): Check new CPU capabilities.
3083 Set new architecture flag variables.
3084 (arm_isr_value): Handle v7m interrupt functions.
3085 (user_return_insn): Return 0 for v7m interrupt functions. Handle
3086 Thumb-2.
3087 (const_ok_for_arm): Handle Thumb-2 constants.
3088 (arm_gen_constant): Ditto. Use movw when available.
3089 (arm_function_ok_for_sibcall): Return false for v7m interrupt
3090 functions.
3091 (legitimize_pic_address, arm_call_tls_get_addr): Handle Thumb-2.
3092 (thumb_find_work_register, arm_load_pic_register,
3093 legitimize_tls_address, arm_address_cost, load_multiple_sequence,
3094 emit_ldm_seq, emit_stm_seq, arm_select_cc_mode, get_jump_table_size,
3095 print_multi_reg, output_mov_long_double_fpa_from_arm,
3096 output_mov_long_double_arm_from_fpa, output_mov_double_fpa_from_arm,
3097 output_mov_double_fpa_from_arm, output_move_double,
3098 arm_compute_save_reg_mask, arm_compute_save_reg0_reg12_mask,
3099 output_return_instruction, arm_output_function_prologue,
3100 arm_output_epilogue, arm_get_frame_offsets, arm_regno_class,
3101 arm_output_mi_thunk, thumb_set_return_address): Ditto.
3102 (arm_expand_prologue): Handle Thumb-2. Use arm_save_coproc_regs.
3103 (arm_coproc_mem_operand): Allow POST_INC/PRE_DEC.
3104 (arithmetic_instr, shift_op): Use arm_shift_nmem.
3105 (arm_print_operand): Use arm_print_condition. Handle '(', ')', '.',
3106 '!' and 'L'.
3107 (arm_final_prescan_insn): Use extract_constrain_insn_cached.
3108 (thumb_expand_prologue): Use thumb_set_frame_pointer.
3109 (arm_file_start): Output directive for unified syntax.
3110 (arm_unwind_emit_set): Handle stack alignment instruction.
3111 * config/arm/lib1funcs.asm: Remove default for __ARM_ARCH__.
3112 Add v6t2, v7, v7a, v7r and v7m.
3113 (RETLDM): Add Thumb-2 code.
3114 (do_it, shift1, do_push, do_pop, COND, THUMB_SYNTAX): New macros.
3115 * config/arm/arm.h (TARGET_CPU_CPP_BUILTINS): Define __thumb2__.
3116 (TARGET_THUMB1, TARGET_32BIT, TARGET_THUMB2, TARGET_DSP_MULTIPLY,
3117 TARGET_INT_SIMD, TARGET_UNIFIED_ASM, ARM_FT_STACKALIGN, IS_STACKALIGN,
3118 THUMB2_TRAMPOLINE_TEMPLATE, TRAMPOLINE_ADJUST_ADDRESS,
3119 ASM_OUTPUT_OPCODE, THUMB2_GO_IF_LEGITIMATE_ADDRESS,
3120 THUMB2_LEGITIMIZE_ADDRESS, CASE_VECTOR_PC_RELATIVE,
3121 CASE_VECTOR_SHORTEN_MODE, ADDR_VEC_ALIGN, ASM_OUTPUT_CASE_END,
3122 ADJUST_INSN_LENGTH): Define.
3123 (TARGET_REALLY_IWMMXT, TARGET_IWMMXT_ABI, CONDITIONAL_REGISTER_USAGE,
3124 STATIC_CHAIN_REGNUM, HARD_REGNO_NREGS, INDEX_REG_CLASS,
3125 BASE_REG_CLASS, MODE_BASE_REG_CLASS, SMALL_REGISTER_CLASSES,
3126 PREFERRED_RELOAD_CLASS, SECONDARY_OUTPUT_RELOAD_CLASS,
3127 SECONDARY_INPUT_RELOAD_CLASS, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P,
3128 TRAMPOLINE_SIZE, INITIALIZE_TRAMPOLINE, HAVE_PRE_INCREMENT,
3129 HAVE_POST_DECREMENT, HAVE_PRE_DECREMENT, HAVE_PRE_MODIFY_DISP,
3130 HAVE_POST_MODIFY_DISP, HAVE_PRE_MODIFY_REG, HAVE_POST_MODIFY_REG,
3131 REGNO_MODE_OK_FOR_BASE_P, LEGITIMATE_CONSTANT_P,
3132 REG_MODE_OK_FOR_BASE_P, REG_OK_FOR_INDEX_P, GO_IF_LEGITIMATE_ADDRESS,
3133 LEGITIMIZE_ADDRESS, THUMB2_LEGITIMIZE_ADDRESS,
3134 GO_IF_MODE_DEPENDENT_ADDRESS, MEMORY_MOVE_COST, BRANCH_COST,
3135 ASM_APP_OFF, ASM_OUTPUT_CASE_LABEL, ARM_DECLARE_FUNCTION_NAME,
3136 FINAL_PRESCAN_INSN, PRINT_OPERAND_PUNCT_VALID_P,
3137 PRINT_OPERAND_ADDRESS): Adjust for Thumb-2.
3138 (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv): New declarations.
3139 * config/arm/arm-cores.def: Add arm1156t2-s, cortex-a8, cortex-r4 and
3140 cortex-m3.
3141 * config/arm/arm-tune.md: Regenerate.
3142 * config/arm/arm-protos.h: Update prototypes.
3143 * config/arm/vfp.md: Enable patterns for Thumb-2.
3144 (arm_movsi_vfp): Add movw alternative. Use output_move_vfp.
3145 (arm_movdi_vfp, movsf_vfp, movdf_vfp): Use output_move_vfp.
3146 (thumb2_movsi_vfp, thumb2_movdi_vfp, thumb2_movsf_vfp,
3147 thumb2_movdf_vfp, thumb2_movsfcc_vfp, thumb2_movdfcc_vfp): New.
3148 * config/arm/libunwind.S: Add Thumb-2 code.
3149 * config/arm/constraints.md: Update include Thumb-2.
3150 * config/arm/ieee754-sf.S: Add Thumb-2/Unified asm support.
3151 * config/arm/ieee754-df.S: Ditto.
3152 * config/arm/bpabi.S: Ditto.
3153 * config/arm/t-arm (MD_INCLUDES): Add thumb2.md.
3154 * config/arm/predicates.md (low_register_operand,
3155 low_reg_or_int_operand, thumb_16bit_operator): New.
3156 (thumb_cmp_operand, thumb_cmpneg_operand): Rename...
3157 (thumb1_cmp_operand, thumb1_cmpneg_operand): ... to this.
3158 * config/arm/t-arm-elf: Add armv7 multilib.
3159 * config/arm/arm.md: Update patterns for Thumb-2 and Unified asm.
3160 Include thumb2.md.
3161 (UNSPEC_STACK_ALIGN, ce_count): New.
3162 (arm_incscc, arm_decscc, arm_umaxsi3, arm_uminsi3,
3163 arm_zero_extendsidi2, arm_zero_extendqidi2): New
3164 insns/expanders.
3165 * config/arm/fpa.md: Update patterns for Thumb-2 and Unified asm.
3166 (thumb2_movsf_fpa, thumb2_movdf_fpa, thumb2_movxf_fpa,
3167 thumb2_movsfcc_fpa, thumb2_movdfcc_fpa): New insns.
3168 * config/arm/cirrus.md: Update patterns for Thumb-2 and Unified asm.
3169 (cirrus_thumb2_movdi, cirrus_thumb2_movsi_insn,
3170 thumb2_cirrus_movsf_hard_insn, thumb2_cirrus_movdf_hard_insn): New
3171 insns.
3172 * doc/extend.texi: Document ARMv7-M interrupt functions.
3173 * doc/invoke.texi: Document Thumb-2 new cores+architectures.
3175 2007-01-03 Jakub Jelinek <jakub@redhat.com>
3177 * unwind-dw2.c (SIGNAL_FRAME_BIT, EXTENDED_CONTEXT_BIT): Define.
3178 (struct _Unwind_Context): Rename args_size to flags, remove
3179 signal_frame field, add a new args_size field and version field.
3180 (_Unwind_IsSignalFrame, _Unwind_SetSignalFrame,
3181 _Unwind_IsExtendedContext): New inline functions.
3182 (_Unwind_GetGR, _Unwind_SetGR, _Unwind_GetGRPtr, _Unwind_SetGRPtr):
3183 Assume by_value array is only present if _Unwind_IsExtendedContext.
3184 (_Unwind_GetIPInfo, execute_cfa_program, uw_frame_state_for): Use
3185 _Unwind_IsSignalFrame.
3186 (__frame_state_for): Initialize context.flags to EXTENDED_CONTEXT_BIT.
3187 (uw_update_context_1): Use _Unwind_SetSignalFrame.
3188 (uw_init_context_1): Initialize context->flags to
3189 EXTENDED_CONTEXT_BIT.
3190 * config/rs6000/linux-unwind.h (frob_update_context): Use
3191 _Unwind_SetSignalFrame.
3193 2007-01-03 Andrew Pinski <andrew_pinski@playstation.sony.com>
3195 PR middle-end/30353
3196 * gimplify.c (gimplify_modify_expr_complex_part): Move below
3197 tree_to_gimple_tuple. Call tree_to_gimple_tuple when we need
3198 the value.
3200 2007-01-03 Kazu Hirata <kazu@codesourcery.com>
3202 * config/i386/i386.h (NON_STACK_REG_P, REGNO_OK_FOR_SIREG_P,
3203 REGNO_OK_FOR_DIREG_P, REWRITE_ADDRESS, ASM_OPERAND_LETTER,
3204 RET, AT_SP): Remove.
3205 * config/i386/i386.md (*sse_prologue_save_insn): Use return
3206 instead of RET.
3208 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
3209 * cfgbuild.c (find_basic_blocks): Likewise.
3210 * cfgrtl.c (rtl_create_basic_block): Likewise.
3211 * function.c (temp_slots_at_level): Likewise.
3212 * reg-stack.c (stack_regs_mentioned): Likewise.
3213 * regclass.c (allocate_reg_info): Likewise.
3214 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
3215 set_bb_for_stmt, move_block_to_fn): Likewise.
3216 * tree-complex.c (tree_lower_complex): Likewise.
3217 * vec.h (VEC_safe_grow_cleared): New.
3219 * cgraphunit.c, tree-ssa-alias.c: Fix comment typos.
3221 2007-01-03 Zdenek Dvorak <dvorakz@suse.cz>
3223 * loop-unswitch.c (unswitch_loop): Pass probabilities to loopify.
3224 * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Pass probabilities
3225 to loop_version.
3226 * cfgloopmanip.c (scale_loop_frequencies): Export.
3227 (loopify): Scale the frequencies by prescribed coefficients.
3228 (set_zero_probability): New function.
3229 (duplicate_loop_to_header_edge): Improve updating of frequencies.
3230 (lv_adjust_loop_entry_edge, loop_version): Set probabilities
3231 and frequencies according to arguments.
3232 * tree-ssa-loop-manip.c (tree_unroll_loop): Set probabilities
3233 correctly.
3234 * cfg.c (scale_bbs_frequencies_int): Allow scaling the frequencies up.
3235 * modulo-sched.c (sms_schedule): Set probabilities for entering
3236 versioned loop correctly.
3237 * tree-vect-transform.c (vect_transform_loop): Ditto.
3238 * cfgloop.h (loopify, loop_version): Declaration changed.
3239 (scale_loop_frequencies): Declared.
3241 2007-01-02 Jan Hubicka <jh@suse.cz>
3243 * cgraph.c: Include tree-flow.h
3244 (cgraph_add_new-function): Handle IPA_SSA mode; execute
3245 early_local_passes.
3246 * cgraph.h (enum cgraph_state): Add CGRAPH_STATE_IPA_SSA.
3247 * tree-pass.h (pass_all_early_optimizations): Declare.
3248 * cgraphunit.c (cgraph_process_new_functions): Add IPA_SSA; execute
3249 early_local_passes.
3250 (cgraph_analyze_function): Do early_local_passes.
3251 * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
3252 Do not add referenced vars.
3253 * tree-optimize.c (gate_all_optimizations): Do not execute when not in
3254 SSA form.
3255 (gate_all_early_local_passes): New gate.
3256 (pass_early_local_passes): Use new gate.
3257 (execute_early_local_optimizations): New functions.
3258 (gate_all_early_optimizations): New gate.
3259 (pass_all_early_optimizations): New pass.
3260 (execute_free_datastructures): Free SSA only when initialized.
3261 (gate_init_datastructures): Init only when optimizing.
3262 (tree_lowering_passes): Do early local passes when called late.
3263 * tree-profile.c (do_tree_profiling): Don't profile functions added
3264 late.
3265 (do_early_tree_profiling, pass_early_tree_profile): Kill.
3266 * tree-cfg.c (update_modified_stmts): Do not update when operands are
3267 not active.
3268 * passes.c (init_optimizations_passes): Reorder so we go into SSA
3269 during early_local_passes.
3270 * Makefile.in (cgraph.o): Add dependency on tree-flow.h.
3273 2007-01-02 Carlos O'Donell <carlos@codesourcery.com>
3275 * Makefile.in: Update copyright year.
3277 2007-01-02 Carlos O'Donell <carlos@codesourcery.com>
3279 * Makefile.in: Export GCC_EXEC_PREFIX before calling $(RUNTEST)
3280 in $(lang_checks) and check-consistency targets.
3282 2007-01-02 Jan Hubicka <jh@suse.cz>
3284 * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
3285 Do not add referenced vars.
3286 * tree-cfg.c (update_modified_stmts): Do not update when SSA operands
3287 are not active.
3288 * passes.c (init_optimization_passes): Put mudflap_2 after
3289 free_datastructures.
3291 2007-01-02 Jan Hubicka <jh@suse.cz>
3293 * tree-optimize (execute_fixup_cfg): Set after_inlining flag.
3294 Set NOTHROW flag on call statements proved to be nothrow.
3295 Update statement of local calls so new pure/const functions are
3296 updated. Update_ssa when in ssa form. Mark PHI nodes of nonlocal
3297 goto receivers.
3298 (tree_rest_of_compilation): Register hooks and initialize bitmap
3299 early. Do not set after_inlining flag.
3301 2007-01-02 Steve Ellcey <sje@cup.hp.com>
3303 * sbitmap.c (HOST_BITS_PER_LONG_LONG): Change to
3304 HOST_BITS_PER_LONGLONG
3306 2007-01-02 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
3308 PR c/19977
3309 * c-typeck.c (store_init_value): Don't emit pedantic overflow
3310 warning for non-static initializers.
3312 2007-01-02 Steven Bosscher <steven@gcc.gnu.org>
3314 * config/alpha/alpha.md, arm/arm.c, darwin.c, frv/frv.md,
3315 m32r/m32r.c, m32r/m32r.c, mn10300/mn10300.md, pa/pa.c,
3316 rs6000/rs6000.c, s390/s390.md, sh/sh.md, sparc/sparc.c:
3317 Always use set_unique_reg_note to add REG_EQUAL notes.
3319 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
3321 Revert:
3322 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
3324 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
3325 * cfgbuild.c (find_basic_blocks): Likewise.
3326 * cfgrtl.c (rtl_create_basic_block): Likewise.
3327 * function.c (temp_slots_at_level): Likewise.
3328 * reg-stack.c (stack_regs_mentioned): Likewise.
3329 * regclass.c (allocate_reg_info): Likewise.
3330 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
3331 set_bb_for_stmt, move_block_to_fn): Likewise.
3332 * tree-complex.c (tree_lower_complex): Likewise.
3333 * vec.h (VEC_safe_grow_cleared): New.
3335 2007-01-02 Ian Lance Taylor <iant@google.com>
3337 * c-common.c (c_common_truthvalue_conversion): When warning about
3338 using an assignment as a truth value, set TREE_NO_WARNING.
3340 2007-01-02 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
3342 PR middle-end/7651
3343 * c.opt (Wold-style-declaration): New.
3344 * doc/invoke.texi (C-only Warning Options): New.
3345 (Wold-style-declaration): Document it.
3346 (Wextra): Enabled by -Wextra.
3347 * c-opts.c (c_common_post_options): Enabled by -Wextra.
3348 * c-decl.c (declspecs_add_scspec): Replace -Wextra with
3349 -Wold-style-declaration.
3351 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
3353 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
3354 * cfgbuild.c (find_basic_blocks): Likewise.
3355 * cfgrtl.c (rtl_create_basic_block): Likewise.
3356 * function.c (temp_slots_at_level): Likewise.
3357 * reg-stack.c (stack_regs_mentioned): Likewise.
3358 * regclass.c (allocate_reg_info): Likewise.
3359 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
3360 set_bb_for_stmt, move_block_to_fn): Likewise.
3361 * tree-complex.c (tree_lower_complex): Likewise.
3362 * vec.h (VEC_safe_grow_cleared): New.
3364 2007-01-02 Douglas Gregor <doug.gregor@gmail.com>
3366 * c-common.c (c_common_nodes_and_builtins): Since variants of
3367 void_type_node get built before it is given a name, we need to
3368 give those variants the name, too.
3369 (complete_array_type): We need to work with the canonical main
3370 type of the array, from which we will build the qualified version.
3371 * params.def (PARAM_VERIFY_CANONICAL_TYPES): New.
3372 * print-tree.c (print_node): Display canonical type information
3373 for each type.
3374 * stor-layout.c (layout_type): When we don't know the
3375 alignment of a type for which we're building an array, we end up
3376 guessing wrong, so make the type require structural equality.
3377 * tree.c (make_node_stat): When we build a new type, it is its
3378 own canonical type.
3379 (build_type_attribute_qual_variant): When building an attribute
3380 variant, its canonical type is the non-attribute variant. However,
3381 if the attributes are target-dependent and they differ, we need to
3382 use structural equality checks for this type.
3383 (build_qualified_type): A qualified type is not equivalent to its
3384 unqualified variant; set the canonical type appropriately.
3385 (build_distinct_type_copy): When building a distinct type from
3386 another type, the new type is its own canonical type.
3387 (build_variant_type_copy): When building a new type variant, we
3388 assume that it is equivalent to the original type.
3389 (build_pointer_type_for_mode): When building a pointer type, also
3390 build a canonical type pointer.
3391 (build_reference_type_for_mode): When building a reference type,
3392 also build a canonical type reference.
3393 (build_index_type): When we can't hash an index type (e.g.,
3394 because its maximum value is negative), the index type requires
3395 structural equality tests.
3396 (build_array_type): Build the canonical form of an array type.
3397 (build_function_type): Function types require structural equality,
3398 because they contain default arguments, attributes, etc.
3399 (build_method_type_directly): Ditto for method types.
3400 (build_offset_type): Build the canonical offset type.
3401 (build_complex_type): Build the canonical vector type.
3402 (make_vector_type): Build the canonical vector type.
3403 * tree.h (TYPE_CANONICAL): New.
3404 (TYPE_STRUCTURAL_EQUALITY_P): New.
3405 (SET_TYPE_STRUCTURAL_EQUALITY): New.
3406 (struct tree_type): Added "canonical" field.
3407 * params.h (VERIFY_CANONICAL_TYPES): New.
3408 * doc/c-tree.texi (TYPE_CANONICAL): Document.
3409 (TYPE_STRUCTURAL_EQUALITY_P): Document.
3410 (SET_TYPE_STRUCTURAL_EQUALITY): Document.
3411 * doc/invoke.texi (verify-canonical-types): Document --param
3412 parameter for verifying canonical types.
3414 2007-01-02 Joseph Myers <joseph@codesourcery.com>
3416 * config.gcc (powerpc-*-eabispe*, powerpc-*-eabisimaltivec*,
3417 powerpc-*-eabisim*, powerpc-*-eabialtivec*, powerpc-*-eabi*,
3418 powerpc-*-rtems*, powerpc-wrs-vxworks, powerpc-wrs-vxworksae,
3419 powerpcle-*-eabisim*, powerpcle-*-eabi*): Add rs6000/e500.h to
3420 tm_file.
3421 * config/rs6000/e500.h: New.
3422 * config/rs6000/eabi.h (TARGET_SPE_ABI, TARGET_SPE, TARGET_E500,
3423 TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE, TARGET_E500_DOUBLE):
3424 Remove.
3425 * config/rs6000/linuxspe.h (TARGET_SPE_ABI, TARGET_SPE,
3426 TARGET_E500, TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE,
3427 TARGET_E500_DOUBLE): Remove.
3428 * config/rs6000/vxworks.h (TARGET_SPE_ABI, TARGET_SPE,
3429 TARGET_E500, TARGET_ISEL, TARGET_FPRS): Remove.
3430 * config/rs6000/rs6000.h (CHECK_E500_OPTIONS): Define.
3431 * config/rs6000/rs6000.c (rs6000_override_options): Use
3432 CHECK_E500_OPTIONS.
3434 2007-01-02 Joseph Myers <joseph@codesourcery.com>
3436 * config/rs6000/rs6000.c (print_operand): Check (TARGET_SPE ||
3437 TARGET_E500_DOUBLE), not TARGET_E500, for %y.
3438 (rs6000_generate_compare, rs6000_emit_sCOND, output_cbranch,
3439 rs6000_emit_cmove): Don't check TARGET_E500.
3440 * config/rs6000/rs6000.md (bunordered, bordered, sunordered,
3441 sordered): Don't check TARGET_E500.
3443 2007-01-01 Eric Christopher <echristo@apple.com>
3445 * config/mips/mips.c (mips_regno_mode_ok_for_base_p): Use
3446 HARD_REGISTER_NUM_P.
3448 2007-01-01 Roger Sayle <roger@eyesopen.com>
3450 * fold-const.c (fold_binary) <EQ_EXPR>: Fold "(X^C1) eq/ne C2" into
3451 "X eq/ne (C1^C2)". Fold "(X^Z) eq/ne (Y^Z)" as "X eq/ne Y" when Z
3452 has no side-effects. Fold "(X^C1) eq/ne (Y^C2)" as "(X^(C1^C2))
3453 eq/ne Y".
3455 2007-01-01 Mike Stump <mrs@apple.com>
3457 * configure.ac: Remove support for building with Apple's gcc-3.1.
3459 2007-01-02 Joseph Myers <joseph@codesourcery.com>
3461 PR middle-end/30311
3462 * caller-save.c (add_stored_regs): Only handle SUBREGs if inner
3463 REG is a hard register. Do not modify REG before calling
3464 subreg_nregs.
3465 * rtlanal.c (subreg_get_info): Don't assert size of XMODE is a
3466 multiple of the size of YMODE for certain lowpart cases.
3468 2007-01-01 Andrew Pinski <pinskia@gmail.com>
3470 PR middle-end/30253
3471 * gimplify (voidify_wrapper_expr): Update for
3472 GIMPLE_MODIFY_STMT.
3474 2007-01-01 Andreas Schwab <schwab@suse.de>
3476 PR target/29166
3477 * config/ia64/ia64.c (ia64_compute_frame_size): Account space for
3478 save of BR0 in extra_spill_size instead of spill_size.
3479 (ia64_expand_prologue): Save BR0 outside of the gr/br/fr spill
3480 area.
3481 (ia64_expand_epilogue): Restore BR0 from its new location.
3483 2007-01-01 Andrew Pinski <pinskia@gmail.com>
3485 * gimplify.c (gimplify_init_constructor <case VECTOR_TYPE>):
3486 Use a temporary variable if the left hand side is not a gimple
3487 register.
3489 2007-01-01 Andrew Pinski <pinskia@gmail.com>
3491 * gimplify.c (gimplify_return_expr): Make the temporary variable
3492 for the return expression, a gimple register variable.
3494 2007-01-01 Jan Hubicka <jh@suse.cz>
3496 * emit-rtl.c (emit_copy_of_insn_after): Do not call copy_insn_1 for
3497 INSN_LIST.
3499 2007-01-01 Mike Stump <mrs@apple.com>
3501 * configure.ac (HAVE_GAS_LITERAL16): Add autoconf check for
3502 .literal16.
3503 * config/darwin.c (machopic_select_rtx_section): Use
3504 HAVE_GAS_LITERAL16.
3505 (darwin_mergeable_constant_section): Likewise.
3506 * configure: Regenerate.
3507 * config.in: Regenerate.
3509 2007-01-01 Jan Hubicka <jh@suse.cz>
3510 Andrew Pinski <pinskia@gmail.com>
3512 * cgraphunit.c (cgraph_optimize): Call cgraph_add_new_functions
3513 before starting IPA passes.