1 2007-01-23 Ian Lance Taylor <iant@google.com>
3 * Makefile.in (OBJS-common): Reformat, alphabetize, but put
5 (OBJS-archive): Reformat, alphabetize.
6 (OBJS): Change out_object_file to OBJS-md.
8 2007-01-23 Uros Bizjak <ubizjak@gmail.com>
10 * config/i386/i386.md: Use REG_P, MEM_P, CONST_INT_P, LABEL_P,
11 JUMP_P and CALL_P predicates where applicable.
12 * config/i386/i386.c: Ditto.
13 * config/i386/i386.md: Ditto.
14 * config/i386/mmx.md: Ditto.
15 * config/i386/predicates.md: Ditto.
17 2007-01-22 Andreas Schwab <schwab@suse.de>
19 * config/m68k/m68k.h: Fix comment.
21 2007-01-22 Jan Hubicka <jh@suse.cz>
23 * passes.c (init_optimization_passes): Do not rerun
24 pass_early_warn_uninitialized.
26 2007-01-22 Richard Guenther <rguenther@suse.de>
28 PR tree-optimization/30038
29 * tree-ssa-math-opts.c (maybe_record_sincos): New static helper
31 (execute_cse_sincos_1): Likewise.
32 (execute_cse_sincos): Likewise.
33 (gate_cse_sincos): Likewise.
34 (pass_cse_sincos): New pass CSEing sin() and cos() calls using
35 the cexpi() canonicalization of sincos().
36 * tree-pass.h (pass_cse_sincos): Declare.
37 * passes.c (init_optimization_passes): New pass pas_cse_sincos.
39 2007-01-21 Eric Botcazou <ebotcazou@libertysurf.fr>
41 PR rtl-optimization/29329
42 * combine.c (replaced_rhs_insn): Rename to i2mod.
43 (replaced_rhs_value): Rename to i2mod_new_rhs.
44 (i2mod_old_rhs): New global variable.
45 (combine_instructions): Adjust for above change. Save a copy of
46 the old RHS into i2mod_old_rhs when the contents of a REG_EQUAL
47 note are substituted in the second instruction.
48 (distribute_notes) <REG_DEAD>: Adjust for above change. Do not
49 ditch the note if it pertains to the second eliminated register
50 and this register is mentioned in i2mod_old_rhs.
53 2006-09-12 Eric Botcazou <ebotcazou@libertysurf.fr>
55 * combine.c (distribute_notes) <REG_DEAD>: Do not consider SETs past
56 the insn to which the note was originally attached.
58 2007-01-21 Jan Hubicka <jh@suse.cz>
60 * ipa-inline.c (inlining_mode): Comment, move up.
61 (cgraph_decide_inlining_incrementally): Do not perform inlining
62 itself; fix handling of flattening of self recursive functions.
63 (cgraph_find_cycles): Remove.
64 (cgraph_flatten_node): Remove.
65 (cgraph_decide_inlining): Use incremental inliner to handle flattening.
66 (try_inline): New function.
67 (cgraph_early_inlining): Update call of
68 cgraph_decide_inlining_incrementally. Apply inlining here.
69 (apply_inline): Update call of cgraph_decide_inlining_incrementally.
71 2007-01-21 Dirk Mueller <dmueller@suse.de>
74 * tree-vrp.c (check_array_bounds): do not warn
75 about ADDR_EXPR's of ARRAY_REF's which are immediately
76 used in binary expressions.
78 2007-01-21 Ira Rosen <irar@il.ibm.com>
80 * tree-vectorizer.h (struct _stmt_vec_info): Add new field
81 read_write_dep and macros for its access.
82 * tree-vectorizer.c (new_stmt_vec_info): Initialize the new field.
83 * tree-vect-analyze.c (vect_analyze_data_ref_dependence): Remove
84 argument, call vect_check_interleaving for every independent pair of
85 data-refs. Mark loads that access the same memory location as a store
87 (vect_check_dependences): Remove.
88 (vect_analyze_data_ref_dependences): Remove vect_check_dependences
89 call, fix the call to vect_analyze_data_ref_dependence.
90 (vect_analyze_data_ref_access): For statements that access the same
91 data-ref, check that they are not stores; for loads, check that there
92 is no store that access the same location.
94 2007-01-20 Roger Sayle <roger@eyesopen.com>
95 Joseph Myers <joseph@codesourcery.com>
97 * doc/invoke.texi (-fdump-rtl-combine): Fix under/overfull hbox.
98 (-fdump-rtl-stack): Likewise.
99 (-fno-signed-zeros): Use @minus{} for a minus sign. Correct typo.
100 (-mcheck-zero-division, -mcpu): Fix under/overfull hbox.
101 (-mpt-fixed): Use @minus{} for minus sign.
102 (Using Precompiled Headers): Fix under/overfull hbox.
104 2007-01-20 Jan Hubicka <jh@suse.cz>
106 * tree-flow.h (struct stmt_ann_d): Move references_memory to proper
107 place within annotation.
109 2007-01-20 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
111 * pa.c (output_move_double): Change array size of xoperands to 4.
113 2007-01-20 Jan Hubicka <jh@suse.cz>
115 * tree-tailcall.c (adjust_return_value): Do not use RESULT_DECL
118 2007-01-19 Ian Lance Taylor <iant@google.com>
120 * expmed.c (expand_divmod) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]: Cast
121 constant to unsigned HOST_WIDE_INT before negating.
123 2007-01-19 Ian Lance Taylor <iant@google.com>
125 * tree-ssa-operands.h (struct vuse_vec_d): Change num_vuse field
127 (VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Use unsigned comparison.
128 (VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
129 (SET_VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Likewise.
130 (SET_VUSE_ELEMENT_VAR) [ENABLE_CHECKING]: Likewise.
131 (SET_VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
132 (realloc_vdef, realloc_vuse): Change second parameter to
134 (ssa_operand_iterator_d): Change vuse_index and mayuse_index
136 * tree-ssa-operands.c (realloc_vop): Change num_elem parameter to
137 unsigned. Change x and lim locals to unsigned.
138 (realloc_vdef, realloc_vuse): Change num_elem parameter to
140 (finalize_ssa_vuse_ops): Change old_i local to unsigned.
141 (copy_virtual_operands): Change i and n locals to unsigned.
143 2007-01-19 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
146 * builtins.c (fold_builtin_1): Handle builtin fdim.
148 2007-01-20 Jan Hubicka <jh@suse.cz>
150 * tree-ssa.c (init_tree_ssa): Do not call init_alias_heapvars.
151 * tree-ssa-structalias.c (compute_points_to_sets): Do call
153 (init_alias_heapvars): Initialize only when not already initialized.
154 (delete_alias_heapvars): Set heapvar_for_stmt to NULL.
156 2007-01-19 Roger Sayle <roger@eyesopen.com>
158 * common.opt (fsigned-zeros): New command line option.
159 * flags.h (HONOR_SIGNED_ZEROS): Control via flag_signed_zeros instead
160 of flag_unsafe_math_optimizations.
161 * opts.c (set_fast_math_flags): The -ffast-math command line option
162 implies -fno-signed-zeros.
163 (fast_math_flags_set_p): Likewise.
165 * doc/invoke.texi: Document new -fno-signed-zeros option, and update
166 the documentation of -ffast-math appropriately. Wrap long lines.
168 2007-01-19 Steve Ellcey <sje@cup.hp.com>
170 * system.h (ASM_MAKE_LABEL_LINKONCE): Poison.
171 * varasm.c (globalize_decl): Remove ASM_MAKE_LABEL_LINKONCE ifdef.
173 2007-01-19 Tomas Bily <tbily@suse.cz>
175 * cgraphunit.c (cgraph_finalize_function): Updating of pid
177 (tree_init_ic_make_global_vars): New function
178 (tree_init_edge_profiler): call of tree_init_ic_make_global_vars
179 (tree_gen_ic_profiler): New function
180 (tree_gen_ic_func_profiler): New function
181 (tree_profiling): Added calling of tree_gen_ic_func_profiler
182 (tree_profile_hooks): Added hook for indirec/virtual calls
183 * value-prof.c (tree_find_values_to_profile): New case for
185 (tree_values_to_profile): Call for determining indirect/virtual
187 (tree_indirect_call_to_profile): New function
188 (tree_ic_transform): New function
189 (tree_ic): New function
190 (find_func_by_pid): New function
191 (init_pid_map): New function
192 (tree_value_profile_transformations): Added check for
193 indirect/virtual call transformation
194 * value-prof.h (enum hist_type): New counter type for
195 indirect/virtual calls
196 (profile_hooks): Added new hook for profiling indirect/virtual
198 * profile.c (instrument_values): New case for indirect/virtual
200 * gcov-io.h (GCOV_LAST_VALUE_COUNTER): Changed to 6
201 (GCOV_COUNTER_V_INDIR): New counter type
202 (GCOV_COUNTER_NAMES): New name of counter "indirect" added
203 (GCOV_MERGE_FUNCTIONS): New merge function for indirect/virtual
205 * cgraph.c: Definition of cgraph_max_pid
206 (cgraph_create_node): Default init of pid attribute
207 * cgraph.h: Declaration of cgraph_max_pid
208 (struct cgraph_node): Added pid attribute
209 * libgcov.c (__gcov_indirect_call_profiler): New function
210 (__gcov_one_value_profiler_body): New function
211 (__gcov_one_value_profiler): Body was moved to
212 __gcov_one_value_profiler_body and calls it
214 2007-01-19 Basile Starynkevitch <basile@starynkevitch.net>
216 * doc/gty.texi (Options): Document the mark_hook option to GTY.
217 * gengtype.c (write_types_data, write_func_for_structure,
218 write_types, ggc_wtd, pch_wtd): Add skip_hooks to
219 write_types_data, ggc_wtd, pch_wtd for processing mark_hook.
220 (walk_type, write_func_for_structure): Generate the mark_hook if
223 2007-01-19 Jan Hubicka <jh@suse.cz>
225 * ipa-inline.c (cgraph_decide_inlining_incrementally): Instead of
226 'early' argument take inlining mode argument specifying whether to
227 inline for size/speeed or all functions; add support for flattening;
229 (cgraph_early_inlining): Update call of decide_inlining_incrementally.
231 2007-01-19 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
234 * toplev.c (warn_deprecated_use): Use %qD instead of %qs to print
235 the name of the declared identifier.
237 2007-01-19 Dirk Mueller <dmueller@suse.de>
239 * config/i386.h (CONDITIONAL_REGISTER_USAGE): Store
240 result of PIC_OFFSET_TABLE_REGNUM in temporary variable to avoid
241 duplicate evaluation.
243 2007-01-19 Uros Bizjak <ubizjak@gmail.com>
245 * config/i386/i386.md (acos<mode>2): Rename from acossf2 and acosdf2.
246 Macroize expander using X87MODEF12 mode macro. Extend operand 1
247 to XFMode, use acosxf2 and truncate result to requested mode.
248 Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
249 (asin<mode>2): Similarly, with asin expanders.
250 (*fscalexf4_i387): Rename from *fscalexf4.
251 (expNcorexf3): New expander.
252 (expxf2, exp10xf2, exp2xf2): Use expNcorexf3 expander.
253 (exp<mode>2): Rename from expsf2 and expdf2. Macroize expander using
254 X87MODEF12 mode macro. Extend operand 1 to XFMode, use expxf2 and
255 truncate result to requested mode. Use SSE_FLOAT_MODE_P to disable
256 patterns for SSE math.
257 (exp10<mode>2): Similarly, with exp10 expanders.
258 (exp2<mode>2): Similarly, with exp2 expanders.
259 (expm1<mode>2): Similarly, with expm1 expanders.
260 (ldexp<mode>3): Similarly, with ldexp expanders.
261 (log<mode>2, log10<mode>2, log2<mode>2, log1p<mode>2, logb<mode>2):
262 Use gen_truncxf<mode>2_i387_noop to truncate result.
264 2007-01-19 Richard Sandiford <richard@codesourcery.com>
266 * config/m68k/m68k.h (PREFERRED_STACK_BOUNDARY): Define to 32
267 for ColdFire targets.
269 2007-01-19 Nathan Sidwell <nathan@codesourcery.com>
270 Richard Sandiford <richard@codesourcery.com>
272 * config/m68k/m68k.h (M68K_STATIC_CHAIN_REG_NAME): New macro.
273 (INITIALIZE_TRAMPOLINE): Use STATIC_CHAIN_REGNUM.
274 (__transfer_from_trampoline): Use M68K_STATIC_CHAIN_REG_NAME.
275 * config/m68k/m68kelf.h (STATIC_CHAIN_REGNUM): Override.
276 (M68K_STATIC_CHAIN_REG_NAME): Likewise.
277 * config/m68k/netbsd-elf.h (M68K_STATIC_CHAIN_REG_NAME): Likewise.
279 2007-01-19 Richard Sandiford <richard@codesourcery.com>
281 * config/m68k/m68k.md (adddi_dilshr32): Rename to...
282 (*adddi_dilshr32): ...this. Fix formatting. Remove commented-out
283 non-canonical pattern. Restrict to !TARGET_COLDFIRE.
284 (*adddi_dilshr32_cf): New pattern.
285 (adddi3, subdi3): Remove first alternatives.
287 2007-01-19 Richard Sandiford <richard@codesourcery.com>
289 * config/m68k/m68k.c (notice_update_cc): If an SFmode move is
290 implemented using move.l, do not use its cc result for floating-point
293 2007-01-19 Richard Sandiford <richard@codesourcery.com>
295 * config/m68k/m68k.h (EXTRA_CONSTRAINT): Stop the 'T' constraint
296 from accepting 's' constraints if flag_pic.
298 2007-01-19 Richard Sandiford <richard@codesourcery.com>
300 * config/m68k/m68k.md (bordered, bunordered, buneq, bunge, bungt)
301 (bunle, bunlt, bltgt, bordered_rev, bunordered_rev, buneq_rev)
302 (bunge_rev, bungt_rev, bunle_rev, bunlt_rev, bltgt_rev): Change
303 condition from TARGET_68881 to TARGET_HARD_FLOAT.
305 2007-01-19 Sandra Loosemore <sandra@codesourcery.com>
307 * longlong.h (count_leading_zeros, COUNT_LEADING_ZEROS_0): Add
308 ColdFire alternatives.
309 * config/m68k/m68k.h (CLZ_DEFINED_VALUE_AT_ZERO): New macro.
310 * config/m68k/m68k.md (clzsi2): Define for ColdFire
311 architectures that support ff1 instruction.
313 2007-01-19 Richard Sandiford <richard@codesourcery.com>
314 Julian Brown <julian@codesourcery.com>
316 * config/m68k/m68k.h (CONST_OK_FOR_LETTER_P): Add an 'R' case.
317 * config/m68k/m68k.md (*movsi_cfv4): Fold into...
318 (*movsi_cf): ...here. Remove unnecessary 'R' from 'Rg'.
321 2007-01-19 Richard Sandiford <richard@codesourcery.com>
323 * config/m68k/m68k-protos.h (valid_mov3q_const): Take a HOST_WIDE_INT
325 (output_move_const_into_data_reg, output_move_simode_const): Delete.
326 * config/m68k/m68k.c (const_method, const_int_cost): Take a
327 HOST_WIDE_INT instead of an rtx.
328 (m68k_rtx_costs): Update call accordingly.
329 (output_move_const_into_data_reg): Likewise. Fix formatting.
330 (valid_mov3q_const): Take a HOST_WIDE_INT instead of an rtx.
332 (output_move_simode_const): Update calls after above changes.
333 Rework to use automatic variables and predicates like MEM_P.
334 * config/m68k/m68k.md (pushexthisi_const): Update call to
337 2007-01-19 Dirk Mueller <dmueller@suse.de>
339 * tree-ssa-alias.c (perform_var_substitution): Fix typo
342 2007-01-19 Richard Guenther <rguenther@suse.de>
344 * builtins.c (expand_builtin_cexpi): Fall back to expanding
345 via cexp in case sincos is not available.
347 2007-01-19 Richard Guenther <rguenther@suse.de>
349 * doc/tm.texi (TARGET_HAS_SINCOS): Document new target macro.
350 * defaults.h (TARGET_HAS_SINCOS): Default to off.
351 * config/linux.h (TARGET_HAS_SINCOS): Set to on if we have glibc.
352 * config/alpha/linux.h (TARGET_HAS_SINCOS): Likewise.
353 * config/sparc/linux.h (TARGET_HAS_SINCOS): Likewise.
354 * config/sparc/linux64.h (TARGET_HAS_SINCOS): Likewise.
355 * config/rs6000/linux.h (TARGET_HAS_SINCOS): Likewise.
356 * config/rs6000/linux64.h (TARGET_HAS_SINCOS): Likewise.
358 2007-01-19 Uros Bizjak <ubizjak@gmail.com>
360 * config/i386/i386.md (*fpatanxf3_i387, fpatan_extend<mode>xf3_i387):
362 (atan2sf3_1, atan2df3_1, atan2xf3_1): Remove insn patterns.
363 (atan2xf3): Directly generate RTL pattern.
364 (atan2<mode>3): Rename from atan2sf3 and atan2df3 and macroize insn
365 patterns using X87MODEF12 mode macro. Use fpatan_extend<mode>xf3_i387
366 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
367 disable patterns for SSE math.
368 (atan<mode>2): Rename from atansf2 and atandf2 and macroize insn
369 patterns using X87MODEF12 mode macro. Use fpatan_extend<mode>xf3_i387
370 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
371 disable patterns for SSE math.
373 2007-01-19 Alexandre Oliva <aoliva@redhat.com>
375 * libgcc-std.ver: Fix typo in %inherit for GCC_4.3.0.
377 2007-01-18 Roger Sayle <roger@eyesopen.com>
379 * fold-const.c (fold_unary) <VIEW_CONVERT_EXPR>: Optimize away a
380 VIEW_CONVERT_EXPR to the same type as it's operand.
382 2007-01-18 David Edelsohn <edelsohn@gnu.org>
384 * config/rs6000/darwin-ldouble.c: Only build _SOFT_FLOAT if
385 configured for long double 128.
387 2007-01-18 Mike Stump <mrs@apple.com>
389 * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Fix build
392 2007-01-18 Michael Meissner <michael.meissner@amd.com>
394 * i386.c (ix86_compute_frame_layout): Make fprintf's in #if 0 code
397 2007-01-18 Jan Hubicka <jh@suse.cz>
399 * tree-ssa-operands.c (vop_free_bucket_size): Never return value
400 greater than NUM_VOP_FREE_BUCKETS.
402 2007-01-18 Daniel Berlin <dberlin@dberlin.org>
404 * tree-ssa-structalias.c: Update comments.
405 (ptabitmap_obstack): Removed.
407 (oldpta_obstack): Ditto.
408 (stats): Add a few members.
409 (struct variable_info): Remove node, complex, address_taken, and
410 indirect_target members. Add oldsolution member.
411 (new_var_info): Do not initialize removed members.
412 (constraint_expr_type): Remove INCLUDES.
413 (constraint_graph): Add size, implicit_preds, rep,
414 indirect_cycles, eq_rep, label, direct_nodes, and complex members.
415 (FIRST_REF_NODE): New macro.
416 (LAST_REF_NODE): Ditto.
417 (FIRST_ADDR_NODE): Ditto.
418 (find): New function.
420 (dump_constraint): Do not handle INCLUDES.
421 (insert_into_complex): Do not insert duplicate constraints.
422 (condense_varmap_nodes): Renamed and rewritten into ...
423 (merge_node_constraints): This. Also fix bug in handling of
424 offseted copy constraints.
425 (clear_edges_for_node): No longer need to deal with preds at all,
426 or removing associated preds/succs.
427 (merge_graph_nodes): Deal with indirect_cycles.
428 Don't deal with predecessors.
429 (add_implicit_graph_edge): New function.
430 (add_pred_graph_edge): Ditto.
431 (add_graph_edge): Don't deal with predecessors.
432 (build_constraint_graph): Removed.
433 (build_pred_graph): New function.
434 (build_succ_graph): Ditto.
435 (struct scc_info): Removed in_component. Added roots, dfs, and
436 node_mapping. Remove visited_index, unification_queue.
437 (scc_visit): Deal with union-find we do now.
438 Deal with cycles with REF nodes.
439 (collapse_nodes): Renamed and rewritten to ...
441 (process_unification_queue): Removed.
442 (topo_visit): Cleanup
443 (do_da_constraint): Use find.
444 (do_sd_constraint): Ditto.
445 (do_ds_constraint): Ditto.
446 (do_complex_constraint): Ditto.
447 (init_scc_info): Update for removed and added members.
448 (find_and_collapse_graph_cycles): Renamed and rewritten into ...
449 (find_indirect_cycles): This.
450 (equivalence_class): New variable.
451 (label_visit): New function.
452 (perform_variable_substitution): Rewritten.
453 (free_var_substitution_info): New function.
454 (find_equivalent_node): Ditto.
455 (move_complex_constraints): Ditto.
456 (eliminate_indirect_cycles): Ditto.
457 (solve_graph): Only propagate changed bits.
458 Use indirect cycle elimination.
460 (tree_id_t): Rename to tree_vi_t, delete id member, add vi member.
461 (tree_id_eq): Renamed to ...
462 (tree_vi_eq): This. Update for member change
463 (insert_id_for_tree): Renamed and rewritten to ...
464 (insert_vi_for_tree): This.
465 (lookup_id_for_tree): Renamed and rewritten to ...
466 (lookup_vi_for_tree): This.
467 (get_id_for_tree): Renamed and rewritten to ...
468 (get_vi_for_tree): Ditto.
469 (get_constraint_exp_from_ssa_var): Update to use get_vi_for_tree.
470 (process_constraint): Don't handle INCLUDES.
471 Remove special ADDRESSOF case.
472 (find_func_aliases): Rewrite to use vi functions instead of id
474 (create_function_info_for): Ditto.
475 (create_variable_info_for): Ditto.
476 (intra_create_variable_infos): Ditto.
477 (merge_smts_into): Ditto.
478 (find_what_p_points_to): Ditto.
479 (init_base_vars): Ditto.
480 (init_alias_vars): Ditto.
481 (remove_preds_and_fake_succs): New function.
482 (dump_sa_points_to_info): Dump new stats.
483 (dump_solution_for_var): Use find.
484 (set_used_smts): Fix formatting.
485 (compute_points_to_sets): Updated for new functions.
486 (ipa_pta_execute): Ditto.
488 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
489 Richard Sandiford <richard@codesourcery.com>
491 * doc/tm.texi (TARGET_FUNCTION_VALUE): Expand documentation of
493 * calls.c (expand_call): If the return value is a PARALLEL,
494 extract its first member.
495 * config/m68k/linux.h (FUNCTION_EXTRA_EPILOGUE): Remove.
496 * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
497 use FUNCTION_EXTRA_EPILOGUE.
498 (m68k_function_value): Return a PARALLEL if the return value
499 is of a pointer type.
500 * config/m68k/netbsd-elf.h (current_function_returns_pointer)
501 (FUNCTION_EXTRA_EPILOGUE): Remove.
502 * config/m68k/m68k.md (D0_REG): New constant.
504 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
506 * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
507 output a NOP for empty epilogues.
509 2007-01-18 Richard Sandiford <richard@codesourcery.com>
511 * config/m68k/m68k.c (m68k_use_return_insn): Update comments
512 before function. Extend register save check to include all
513 registers, not just integer ones.
515 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
517 * config/m68k/m68k-protos.h (use_return_insn): Rename to...
518 (m68k_use_return_insn): ...this.
519 * config/m68k/m68k.h (USE_RETURN_INSN): Delete.
520 * config/m68k/m68k.c (use_return_insn): Rename to...
521 (m68k_use_return_insn): ...this.
522 * config/m68k/m68k.md (return): Use m68k_use_return_insn instead
525 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
527 * config/m68k/fpgnulib.c (__truncdfsf2): Implement round to
528 nearest even, fix denormal rounding overflow.
530 2007-01-18 Richard Sandiford <richard@codesourcery.com>
532 * config/m68k/m68k.md (movsf_cf_hard): Use fsmove instead of
533 f%$smove and f%$move.
534 (movdf_cf_hard): Use fdmove for cases 0 and 3 and fmove for case 1.
535 (extendsfdf2_cf): Use fdmove instead of f%&move.
536 (truncdfsf2_cf): Use fsmove instead of f%$smove.
537 (add<mode>3_cf, sub<mode>3_cf): Use <FP:prec> instead of <FP:round>.
539 2007-01-18 Richard Sandiford <richard@codesourcery.com>
541 * config/m68k/m68k.md (movdf_cf_hard): Use output_move_double for
544 2007-01-18 Richard Sandiford <richard@codesourcery.com>
546 * real.h (real_format): Add a canonical_nan_lsbs_set field.
547 (coldfire_single_format): Declare.
548 (coldfire_double_format): Likewise.
549 * real.c (encode_ieee_single): Use canonical_nan_lsbs_set instead
550 of qnan_msb_set to determine the lower bits of a canonical
552 (encode_ieee_double): Likewise.
553 (encode_ieee_quad): Likewise.
554 (ieee_single_format): Initialize canonical_nan_lsbs_set.
555 (mips_single_format): Likewise.
556 (ieee_double_format): Likewise.
557 (mips_double_format): Likewise.
558 (ieee_extended_motorola_format): Likewise.
559 (ieee_extended_intel_96_format): Likewise.
560 (ieee_extended_intel_128_format): Likewise.
561 (ieee_extended_intel_96_round_53_format): Likewise.
562 (ibm_extended_format): Likewise.
563 (mips_extended_format): Likewise.
564 (ieee_quad_format): Likewise.
565 (mips_quad_format): Likewise.
566 (vax_f_format): Likewise.
567 (vax_d_format): Likewise.
568 (vax_g_format): Likewise.
569 (i370_single_format): Likewise.
570 (i370_double_format): Likewise.
571 (decimal_single_format): Likewise.
572 (decimal_double_format): Likewise.
573 (decimal_quad_format): Likewise.
574 (c4x_single_format): Likewise.
575 (c4x_extended_format): Likewise.
576 (real_internal_format): Likewise.
577 (coldfire_single_format): New real_format.
578 (coldfire_double_format): Likewise.
579 * config/pdp11/pdp11.c (pdp11_f_format): Initialize
580 canonical_nan_lsbs_set.
581 (pdp11_d_format): Likewise.
582 * config/m68k/m68k.c (override_options): Override REAL_FORMAT_MODE
583 if TARGET_COLDFIRE_CPU.
585 2007-01-18 Richard Sandiford <richard@codesourcery.com>
587 * config/m68k/m68k-protos.h (m68k_output_pic_call): Delete.
588 (output_call, m68k_legitimize_call_address): Declare.
589 * config/m68k/m68k.h (EXTRA_CONSTRAINT): Remove unnecessary
590 parenthesees. Add support for a 'W' constraint.
591 (LEGITIMATE_PIC_OPERAND_P): Remove SYMBOL_REF_FLAG handling.
592 (PRINT_OPERAND_PUNCT_VALID_P): Remove comment about 'o'.
593 (m68k_symbolic_call, m68k_symbolic_jump): Declare.
594 * config/m68k/m68k.c (m68k_symbolic_call, m68k_symbolic_jump): New
596 (override_options): Initialize them. Do not set flag_no_function_cse
597 for TARGET_ID_SHARED_LIBRARY.
598 (m68k_output_pic_call): Delete.
599 (m68k_legitimize_call_address): New function.
600 (print_operand): Remove the %o prefix. Handle the %p prefix.
601 (output_call): New function.
602 (m68k_output_mi_thunk): Use m68k_symbolic_jump. Always load the
603 target address from the GOT if symbolic jumps are not allowed.
604 * config/m68k/m68k.md (call, general_operand): Do not set
605 SYMBOL_REF_FLAG. Use m68k_legitimize_call_address instead.
606 Merge separate flag_pic and !flag_pic define_insns into...
607 (*call, *call_value): ...these new patterns. Match the address
608 rather than the containing MEM and require it to be a call_operand.
609 Use output_call to generate the asm template.
610 * config/m68k/predicates.md (const_call_operand): New predicate.
611 (call_operand): Likewise.
613 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
615 * config/m68k/m68k.h (REGISTER_MOVE_COST): Simplify definition.
616 (STACK_GROWS_DOWNWARD): Define to 1.
617 (FUNCTION_VALUE, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P): Equivocate
618 comments, emphasizing that these values are only defaults.
619 * config/m68k/linux.h (LINK_SPEC): Fix formatting in #undef.
620 * config/m68k/m68k.c (const_method): Remove trailing whitespace.
622 2007-01-18 Richard Sandiford <richard@codesourcery.com>
624 * config/m68k/m68k.md (cmpsi): Remove outdated flag_pic handling.
626 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
627 Richard Sandiford <richard@codesourcery.com>
629 * config/m68k/m68k.h (DATA_REGNO_P, ADDRESS_REGNO_P, INT_REGNO_P)
630 (FP_REGNO_P): New macros.
631 (REGNO_OK_FOR_INDEX_P, REGNO_OK_FOR_BASE_P, REGNO_OK_FOR_DATA_P)
632 (REGNO_OK_FOR_FP_P, REG_OK_FOR_INDEX_P, REG_OK_FOR_BASE_P): Use them.
633 (EH_RETURN_STACKADJ_RTX): Use A0_REG.
634 * config/m68k/m68k.c (m68k_regno_mode_ok): Use the new REGNO macros.
636 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
638 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
639 (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-rtems*): Add t-floatlib
641 * config/m68k/t-floatlib: New file, extracting common code from...
642 * config/m68k/t-m68kbare, config/m68k/t-m68kelf,
643 * config/m68k/t-uclinux: Here.
644 * config/m68k/fpgnulib.c: Do not compile extendeed precision
645 routines on ColdFire targets.
647 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
649 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
650 (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*)
651 (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-linux*)
652 (m68k-*-rtems*): Use tm_file rather than m68k/m68k.h and
653 explicitly set MOTOROLA to 1.
654 * config/m68k/m68k.h (MOTOROLA): Simplify definition accordingly.
656 2007-01-18 Richard Sandiford <richard@codesourcery.com>
657 Nathan Sidwell <nathan@codesourcery.com>
659 * config/m68k/m68k.h (PCC_STATIC_STRUCT_RETURN): Delete.
660 (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Add MOTOROLA cases.
661 * config/m68k/coff.h (REGISTER_PREFIX_MD): Delete.
662 * config/m68k/m68020-elf.h (LIB_SPEC): Delete.
663 * config/m68k/m68k-none.h (CC1_SPEC, CPP_SUBTARGET_SPEC): Delete.
664 * config/m68k/m68kelf.h (IMMEDIATE_PREFIX, REGISTER_PREFIX_MD)
665 (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Delete.
666 (NO_DOLLAR_IN_LABEL, PCC_STATIC_STRUCT_RETURN): Don't undefine.
667 (BSS_ASM_OP, ASM_OUTPUT_SKIP, ASM_OUTPUT_ASCII): Delete.
668 * config/m68k/m68kemb.h (PCC_STATIC_STRUCT_RETURN): Don't undefine.
669 * config/m68k/linux.h (SIZE_TYPE, PTRDIFF_TYPE, WCHAR_TYPE)
670 (WCHAR_TYPE_SIZE, TARGET_OBJFMT_CPP_BUILTINS): Delete.
671 (TARGET_OS_CPP_BUILTINS): Don't define mc68000 and mc68020 here.
672 (DBX_REGISTER_NUMBER): Delete.
673 * config/m68k/netbsd-elf.h (IMMEDIATE_PREFIX): Delete.
674 (PCC_STATIC_STRUCT_RETURN): Don't undefine.
675 * config/m68k/openbsd.h (PCC_STATIC_STRUCT_RETURN): Define.
677 2007-01-18 Richard Sandiford <richard@codesourcery.com>
679 * config.gcc (m68k-*-uclinux*): Add flat.h to $tm_file.
680 * config/flat.h: New file.
681 * crtstuff.c (USE_PT_GNU_EH_FRAME): Don't define if
683 * config/m68k/m68k.h (ASM_PREFERRED_EH_DATA_FORMAT): Do not use
684 indirect references for -msep-data or -mid-shared-library.
685 Do not use PC-relative code addresses either.
686 * config/m68k/m68k.c (override_options): Restrict -fPIC error
688 * config/m68k/uclinux.h (STARTFILE_SPEC): Define. Use Scrt1.o
689 for shared libraries and crt1.o for executables. Use crti.o and
691 (ENDFILE_SPEC): Use crtend.o and crtn.o.
692 (LIB_SPEC): Suppress -Rlibc.gdb if -static-libc is given.
693 Do not add -elf2flt or -shared-lib-id options here.
694 (LINK_SPEC): Define. Pass -elf2flt if no -elf2flt option is given.
695 Pass -shared-lib-id if -mid-shared-library, taking the library
696 identifier from -mshared-library-id if given, otherwise
698 (EH_FRAME_IN_DATA_SECTION): Do not undefine.
699 (INIT_SECTION_ASM_OP, FINI_SECTION_ASM_OP): Likewise.
700 (TARGET_OS_CPP_BUILTINS): Define __GXX_MERGED_TYPEINFO_NAMES=0
701 and __GXX_TYPEINFO_EQUALITY_INLINE=0 if -mid-shared-library.
702 (DRIVER_SELF_SPECS): Map unadorned PIC options to -msep-data.
703 * config/m68k/t-uclinux (EXTRA_MULTILIB_PARTS): Add crtbegin.o
705 * config/m68k/lb1sf68.asm (PICCALL): Use an lea and pc-relative
706 jump sequence for ISA A and ISA A+.
709 2007-01-18 Richard Sandiford <richard@codesourcery.com>
711 * config.gcc (m68k-*-uclinux*): Base the port on the common
712 and m68k GNU/Linux files rather than on the generic ELF ones.
713 * config/m68k/uclinux.h (TARGET_VERSION): Override.
714 (TARGET_OS_CPP_BUILTINS): Use LINUX_TARGET_OS_CPP_BUILTINS.
716 2007-01-18 Julian Brown <julian@codesourcery.com>
717 Richard Sandiford <richard@codesourcery.com>
719 * config/m68k/m68k.h (LONG_DOUBLE_TYPE_SIZE): Make 64-bit on ColdFire.
720 (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
721 * config/m68k/netbsd-elf.h (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Undefine
723 * config/m68k/uclinux-oldabi.h (LONG_DOUBLE_TYPE_SIZE): Redefine to
725 (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
727 2007-01-18 Richard Sandiford <richard@codesourcery.com>
729 * doc/install.texi: Document m68k-uclinuxoldabi.
730 * config.gcc (m68k-*-uclinuxoldabi): New configuration.
731 * config/m68k/uclinux-oldabi.h: New file, copied from
732 config/m68k/uclinux.h.
734 2007-01-18 Kaz Kojima <kkojima@rr.iij4u.or.jp>
736 * config/m32r/m32r-protos.h (m32r_expand_epilogue): Declare it.
737 * config/m32r/m32r.c (m32r_setup_incoming_varargs): Use gen_frame_mem.
738 (m32r_compute_frame_size): Use unsigned for regno.
739 (m32r_reload_lr): Use gen_frame_mem.
741 (m32r_output_function_epilogue): Don't output the function epilogue
743 (m32r_expand_epilogue): New.
744 (direct_return): Return false if the function has the interrupt
746 (m32r_hard_regno_rename_ok): Remove code for the textual epilogue.
747 * config/m32r/m32r.md (epilogue): New expander.
748 (return_lr, return_rte): New insns.
749 (return): Make it expander.
750 (return_normal): New expander.
752 2007-01-18 Josh Conner <jconner@apple.com>
755 * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Add
756 support for UNLE, UNLT, UNGE, and UNGT.
758 2007-01-18 Jan Hubicka <jh@suse.cz>
760 * tree-vrp.c (finalize_jump_threads): Do not call cleanup_cfg by hand.
761 * tree-tailcall (add_virtual_phis): Likewise.
762 (optimize_tail_call): Return TODOs.
763 (execute_tail_calls): Return TODOs.
764 * tree-ssa-ccp (execute_fold_all_builtins): Do cleanup_cfg via TODO.
765 * tree-cfgcleanup.c (cleanup_tree_cfg_loop): Return if something
767 * tree-ssa-forwprop.c (tree_ssa_forward_propagate_single_use_value):
768 Cleanup cfg using TODO.
769 * tree-flow.h (cleanup_tree_cfg_loop): Update prototype.
770 * passes.c (execute_function_todo): When cleanup did something, remove
772 * tree-cfg.c (pass_build_cfg): Add cleanup_cfg TODO.
773 (make_edges): Don't cleanup_cfg.
775 2007-01-18 Uros Bizjak <ubizjak@gmail.com>
777 * reg-stack.c (subst_stack_regs_pat) [UNSPEC_SINCOS_COS,
778 UNSPEC_XTRACT_FRACT]: Use generic code for instructions that
779 operate on the top of stack.
780 [UNSPEC_SINCOS_SIN, UNSPEC_XTRACT_EXP, UNSPEC_TAN]: Rewrite
781 register handling of instructions that output to the second
783 [UNSPEC_TAN_ONE, UNSPEC_TAN_TAN]: Remove.
784 (move_for_stack_reg): Special-case check for dead destination
785 stack slot for constant load of 1.0 inside UNSPEC_TAN.
787 * config/i386/i386.md (UNSPEC_TAN): New constant.
788 (UNSPEC_TAN_ONE, UNSPEC_TAN_TAN): Remove.
789 (fptanxf4_i387, fptan_extend<mode>xf4_i387): New patterns
790 to correctly model move of constant 1.0 to top stack slot.
791 (*tandf3_1, *tansf3_1, *tanxf3_1): Remove insn patterns.
792 (unnamed peephole2 pattern): Remove corresponding peephole2
793 pattern that optimizes tan insn and loading of constant 1.0.
794 (tanxf2): Use fptanxf4_i387.
795 (tan<mode>2): Rename from tansf2 and tandf2 and macroize insn
796 patterns using X87MODEF12 mode macro. Use fptan_extend<mode>xf4_i387
797 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
798 disable patterns for SSE math.
799 (sincos<mode>3): Use truncxf<mode>2_i387_noop for truncation.
800 (fyl2x_extend<mode>xf3_i387): Use X87MODEF12 for operand 1.
802 2007-01-18 Dirk Mueller <dmueller@suse.de>
803 Richard Guenther <rguenther@suse.de>
806 * doc/invoke.texi (Warray-bounds): Document -Warray-bounds.
807 * common.opt (Warray-bounds): Add new warning option.
808 * c-opts.c (c_common_handle_option): Define -Warray-bounds
810 * Makefile.in: make tree-vrp.o depend on toplev.h
811 * tree-vrp.c (vrp_finalize): Call check_array_refs if -Warray-bounds
813 (check_array_refs, check_array_bounds, check_array_ref): New.
815 2007-01-18 Jan Hubicka <jh@suse.cz>
817 * tree-ssa-ccp.c (ccp_finalize): Return if something changed.
818 (execute_ssa_ccp): Return flags conditionally.
819 * tree-ssa-propagate.c (substitue_and_fold): Return if something was
821 * tree-ssa-propagate.h (substitute_and_fold): Update prototype.
823 2007-01-18 Steven Bosscher <steven@gcc.gnu.org>
825 * cfgcleanup.c (cleanup_cfg): Detect cfglayout mode and set
826 the CLEANUP_CFGLAYOUT flag when in cfglayout mode.
828 * Makefile.c (GTFILES): Add cfglayout.h.
829 * gengtype.c (open_base_files): Likewise.
830 * cfglayout.c (cfg_layout_function_footer,
831 cfg_layout_function_header) Reindent to make gengtype happy.
832 * cfglayout.h (cfg_layout_function_footer,
833 cfg_layout_function_header): Add GTY(()) marker.
835 * ifcvt.c (noce_try_sign_mask): Make sure INSN_B is non-null.
837 2007-01-18 Ben Elliston <bje@au.ibm.com>
839 * genautomata.c (write_automata): Include xstrerror output in the
840 error message if writing the DFA description file fails.
842 2007-01-17 H.J. Lu <hongjiu.lu@intel.com>
844 * config/mips/mips-protos.h (mips_output_external): Make it
846 * config/mips/iris.h (TARGET_ASM_EXTERNAL_LIBCALL): Removed.
847 * config/mips/mips.c (irix_output_external_libcall): Likewise.
848 (extern_list): Likewise.
849 (extern_head): Likewise.
850 (TARGET_ASM_FILE_END): Likewise.
851 (mips_file_end): Likewise.
852 (mips_output_external): Rewritten.
854 2007-01-18 Ben Elliston <bje@au.ibm.com>
856 * genpreds.c (write_insn_preds_c): Only write out the function
857 body for regclass_for_constraint if we have register constraints.
859 2007-01-17 Tom Tromey <tromey@redhat.com>
861 * doc/sourcebuild.texi (libgcj Tests): Use sourceware.org.
862 * doc/install.texi (Testing): Use sourceware.org.
863 (Binaries): Likewise.
864 (Specific): Likewise.
865 * doc/contrib.texi (Contributors): Use sourceware.org.
867 2007-01-17 Anatoly Sokolov <aesok@post.ru>
869 * config/avr/avr.h (AVR_HAVE_LPMX): New macro.
870 (AVR_ENHANCED): Rename to ...
871 (AVR_HAVE_MUL): ... new.
872 (avr_enhanced_p): Rename to ...
873 (avr_have_mul_p): ... new.
874 (TARGET_CPU_CPP_BUILTINS): Use 'avr_have_mul_p' instead of
875 'avr_enhanced_p' for "__AVR_ENHANCED__". Define "__AVR_HAVE_MUL__".
876 * config/avr/avr.c (avr_enhanced_p): Rename to ...
877 (avr_have_mul_p): ... new.
878 (base_arch_s): Rename 'enhanced' to 'have_mul'.
879 (avr_override_options): Use 'avr_have_mul_p' and 'have_mul' instead of
880 'avr_enhanced_p' and 'enhanced'.
881 (ashlhi3_out, ashrhi3_out, lshrhi3_out, avr_rtx_costs): Use
882 AVR_HAVE_MUL instead of AVR_ENHANCED.
883 * avr.md (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
884 (mulqi3, *mulqi3_enh, *mulqi3_call, mulqihi3, umulqihi3, mulhi3,
885 *mulhi3_enh, *mulhi3_call, mulsi3, *mulsi3_call): Use AVR_HAVE_MUL
886 instead of AVR_ENHANCED.
887 (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
888 * libgcc.S: Use __AVR_HAVE_MUL__ instead of __AVR_ENHANCED__.
889 (__tablejump__): Use __AVR_HAVE_LPMX__ instead of __AVR_ENHANCED__.
891 2007-01-17 Ian Lance Taylor <iant@google.com>
893 * vec.h (VEC_reserve_exact): Define.
894 (vec_gc_p_reserve_exact): Declare.
895 (vec_gc_o_reserve_exact): Declare.
896 (vec_heap_p_reserve_exact): Declare.
897 (vec_heap_o_reserve_exact): Declare.
898 (VEC_OP (T,A,reserve_exact)): New static inline function, three
900 (VEC_OP (T,A,reserve)) [all versions]: Remove handling of
902 (VEC_OP (T,A,alloc)) [all versions]: Call ...reserve_exact.
903 (VEC_OP (T,A,copy)) [all versions]: Likewise.
904 (VEC_OP (T,a,safe_grow)) [all versions]: Likewise.
905 * vec.c (calculate_allocation): Add exact parameter. Change all
907 (vec_gc_o_reserve_1): New static function, from vec_gc_o_reserve.
908 (vec_gc_p_reserve, vec_gc_o_reserve): Call vec_gc_o_reserve_1.
909 (vec_gc_p_reserve_exact, vec_gc_o_reserve_exact): New functions.
910 (vec_heap_o_reserve_1): New static function, from vec_heap_o_reserve.
911 (vec_heap_p_reserve, vec_heap_o_reserve): Call vec_heap_o_reserve_1.
912 (vec_heap_p_reserve_exact): New function.
913 (vec_heap_o_reserve_exact): New function.
915 2007-01-17 Jan Hubicka <jh@suse.cz>
917 * ipa-type-escape.c (look_for_casts): Revamp using handled_component_p.
919 2007-01-17 Eric Christopher <echristo@apple.com>
921 * config.gcc: Support core2 processor.
923 2007-01-16 Jan Hubicka <jh@suse.cz>
925 * tree-ssanames.c (release_dead_ssa_names): Instead of ggc_freeing
926 the names, just unlink the chain so we don't crash on dangling pointers
929 2007-01-16 Jan Hubicka <jh@suse.cz>
931 * cgraph.h (cgraph_decide_inlining_incrementally): Kill.
932 * tree-pass.h: Reorder to make IPA passes appear toegher.
933 (pass_early_inline, pass_inline_parameters, pass_apply_inline): Declare.
934 * cgraphunit.c (cgraph_finalize_function): Do not compute inling
935 parameters, do not call early inliner.
936 * ipa-inline.c: Update comments. Include tree-flow.h
937 (cgraph_decide_inlining): Do not compute inlining parameters.
938 (cgraph_decide_inlining_incrementally): Return TODOs; assume to
939 be called with function context set up.
940 (pass_ipa_inline): Remove unreachable functions before pass.
941 (cgraph_early_inlining): Simplify assuming to be called from the
943 (pass_early_inline): New pass.
944 (cgraph_gate_ipa_early_inlining): New gate.
945 (pass_ipa_early_inline): Turn into simple wrapper.
946 (compute_inline_parameters): New function.
947 (gate_inline_passes): New gate.
948 (pass_inline_parameters): New pass.
949 (apply_inline): Move here from tree-optimize.c
950 (pass_apply_inline): New pass.
951 * ipa.c (cgraph_remove_unreachable_nodes): Verify cgraph after
953 * tree-inline.c (optimize_inline_calls): Return TODOs rather than
955 (tree_function_versioning): Do not allocate dummy struct function.
956 * tree-inline.h (optimize_inline_calls): Update prototype.
957 * tree-optimize.c (execute_fixup_cfg): Export.
958 (pass_fixup_cfg): Remove
959 (tree_rest_of_compilation): Do not apply inlines.
960 * tree-flow.h (execute_fixup_cfg): Declare.
961 * Makefile.in (gt-passes.c): New.
962 * passes.c: Include gt-passes.h
963 (init_optimization_passes): New passes.
964 (nnodes, order): New static vars.
965 (do_per_function_toporder): New function.
966 (execute_one_pass): Dump current pass here.
967 (execute_ipa_pass_list): Don't dump current pass here.
969 2007-01-16 Janis Johnson <janis187@us.ibm.com>
971 * config/dfp-bit.c (dfp_compare_op): Return separate value for NaN.
972 (DFP_NE, DFP_LE, DFP_GE): Return false for NaN.
974 2007-01-16 David Edelsohn <edelsohn@gnu.org>
976 * config/rs6000/darwin-ldouble.c: Build file for SOFT_FLOAT.
977 (strong_alias): Define.
978 (__gcc_qmul): Provide non-FMA for soft-float.
994 * config/rs6000/rs6000.c (rs6000_init_libfuncs): Initialize
995 soft-float functions.
996 * config/rs6000/libgcc-ppc-glibc.ver: Version soft-float symbols.
997 * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Do not warn
998 about long double soft float.
1000 2007-01-16 Dorit Nuzman <dorit@il.ibm.com>
1001 Tehila Meyzels <tehila@il.ibm.com>
1003 * tree-vectorizer.h (is_pattern_stmt_p): New.
1004 * tree-vect-analyze.c (vect_determine_vectorization_factor): Fix
1005 formatting (tabs instead of spaces). Cleanup and clarify setting
1006 of STMT_VINFO_VECTYPE. Call is_pattern_stmt_p.
1007 * tree-vect-transform.c (vect_get_vec_def_for_operand): Fix typo.
1008 (vectorizable_type_demotion): Check that types are integral.
1009 (vectorizable_type_promotion): Likewise.
1010 (vectorizable_store): Fix typo. Eliminate new-line at end of
1013 2007-01-16 Jan Hubicka <jh@suse.cz>
1015 * tree-ssanames.c (release_dead_ssa_names): Remove invalidated
1018 2007-01-15 Eric Christopher <echristo@apple.com>
1020 * ifcvt.c: Include vec.h, vecprim.h.
1021 (check_cond_move_block): New argument regs.
1022 Reorganize. Add registers used to regs.
1023 (cond_move_process_if_block): Use regs set above as
1026 2007-01-15 Eric Christopher <echristo@apple.com>
1028 * config/darwin.h: Update copyright.
1029 (TARGET_OPTION_TRANSLATE_TABLE): Add umbrella.
1030 (LINK_COMMAND_SPEC): Add -u.
1031 (LINK_SPEC): Fix umbrella for above.
1033 2007-01-15 Joseph S. Myers <joseph@codesourcery.com>
1035 * config/soft-fp/op-common.h, config/soft-fp/op-4.h: Update from
1038 2007-01-15 Tom Tromey <tromey@redhat.com>
1040 * doc/sourcebuild.texi (libgcj Tests): Don't mention jacks.
1041 * doc/install.texi (Testing): Don't mention jacks.
1042 (Configuration): Document --enable-java-maintainer-mode. Move
1043 --with-java-home to libgcj-specific section. Document
1045 (Prerequisites): Mention --enable-java-maintainer-mode, ecj1.
1047 2007-01-15 Jan Hubicka <jh@suse.cz>
1049 * tree-ssa-dce.c (DCE_TODOs): New.
1050 (propagate_necessity): Return if something changed.
1051 (eliminate_unnecessary_stmts): Likewise.
1052 (perform_tree_ssa_dce): Return TODO flags when needed.
1053 (pass_dce, pass_dce_loop, pass_cd_dce): Remove TODO flags.
1055 2007-01-15 Uros Bizjak <ubizjak@gmail.com>
1057 * config/i386/i386.md (fyl2xxf3_i387): Rename from fyl2x_xf3.
1058 (fyl2x_extend<mode>xf3_i387): New insn pattern.
1059 (log<mode>2): Rename from logsf2 and logdf2 and macroize insn
1060 patterns using X87MODEF12 mode macro. Extend operand 1
1061 to XFmode. Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
1062 (log10<mode>2): Ditto.
1063 (log2<mode>2): Ditto.
1064 (log1p<mode>2): Ditto.
1065 (logb<mode>2): Ditto.
1066 (fyl2xp1xf3_i387): Rename from fyl2xp1_xf3.
1067 (fyl2xp1_extend<mode>xf3_i387): New insn pattern.
1068 (*fxtractxf3_i387): Rename from *fxtractxf3.
1069 (fxtract_extend<mode>xf3_i387): New insn pattern.
1070 (ilogbsi2): Use match_dup 3, not match_operand:XF 3.
1071 * config/i386/i386.c (ix86_emit_i387_log1p): Use gen_fyl2xp1xf3_i387()
1072 and gen_fyl2xxf3_i387().
1074 2007-01-14 Zdenek Dvorak <dvorakz@suse.cz>
1076 * loop-unswitch.c (unswitch_loop): Do not call fix_loop_placement.
1077 * cfgloopmanip.c (fix_loop_placement): Made static. Use
1078 get_loop_exit_edges. Changed return type to bool.
1079 * cfgloop.h (fix_loop_placement): Declaration removed.
1081 2007-01-14 Dorit Nuzman <dorit@il.ibm.com>
1083 * param.h (MIN_VECT_LOOP_BOUND): New.
1084 * params.def (MIN_VECT_LOOP_BOUND): New.
1085 * tree-vectorizer.c (slpeel_tree_peel_loop_to_edge): Takes another
1086 argument - minimum threshold for number of iterations.
1087 * tree-vectorizer.h (slpeel_tree_peel_loop_to_edge): Add another
1088 argument to declaration.
1089 * tree-vect-analyze.c (vect_analyze_operations): Check value of
1090 MIN_VECT_LOOP_BOUND.
1091 * tree-vect-transform.c (vect_do_peeling_for_loop_bound): Call
1092 slpeel_tree_peel_loop_to_edge with additional argument.
1093 (vect_do_peeling_for_alignment): Likewise.
1094 * doc/invoke.texi (min-vect-loop-bound): Document new param option.
1096 2007-01-14 Uros Bizjak <ubizjak@gmail.com>
1099 * config/i386/i386.c (print_operand) ['z']: Output 'b' for
1102 2007-01-14 Jan Hubicka <jh@suse.cz>
1104 * tree-dfa.c (remove_referenced_var): New function.
1105 * tree-ssa-live.c (remove_unused_locals): Walk referenced vars and
1106 prune referenced vars list too.
1107 * tree-flow.h (remove_referenced_var): Declare.
1109 2007-01-14 Jan Hubicka <jh@suse.cz>
1111 * tree-eh.c (add_stmt_to_eh_region_fn): Do not add call_exprs
1113 (remove_stmt_from_eh_region_fn): Do not remove call_exprs.
1114 (verify_eh_throw_stmt_node, verify_eh_throw_table_statements): Kill.
1115 * except.h (verify_eh_throw_table_statements): Kill prototype.
1116 * cfgexpand.c (expand_gimple_basic_block): Propagate Eh regions
1117 into call exrepssions.
1118 * tree-optimize.c (execute_free_cfg_annotatiosn): Do not call
1120 * tree-cfg.c: Include pointer-set.h.
1121 (verify_node_sharing): Work on pointer set.
1122 (verify_eh_throw_stmt_node): New.
1123 (verify_stmts): Use pointers sets, verify throw_stmt.
1125 2007-01-13 Zdenek Dvorak <dvorakz@suse.cz>
1127 * ipa-reference.c (analyze_function): Consider also addresses taken
1130 2007-01-12 Roger Sayle <roger@eyesopen.com>
1132 * c-typeck.c (null_pointer_constant_p): Replace use of
1133 TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
1134 (build_c_cast): Likewise.
1136 2007-01-12 Roger Sayle <roger@eyesopen.com>
1138 * tree.h (force_fit_type_double): Remove unused final argument.
1139 * c-common.c (constant_expression_warning): Replace use of
1140 TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
1141 (convert_and_check): Likewise.
1142 (shorten_compare): Update call to force_fit_type_double.
1143 (c_common_truthvalue_conversion) <INTEGER_CST>: Use integer_zerop.
1144 * convert.c (convert_to_pointer): Update call to
1145 force_fit_type_double.
1146 * fold-const.c (force_fit_type_double): Remove overflowed_const
1148 (int_const_binop, fold_convert_const_int_from_int,
1149 fold_convert_const_int_from_real, fold_div_compare,
1150 fold_sign_changed_comparison, fold_unary, fold_negate_const,
1151 fold_abs_const, fold_not_const): Remove the final argument from
1152 calls to force_fit_type_double.
1154 2007-01-12 Andrew Pinski <andrew_pinski@playstation.sony.com>
1156 * configure.ac: Set insn to "nop" for spu-*-* also.
1157 * configure: Regenerate.
1159 2007-01-12 Olga Golovanevsky <olga@il.ibm.com>
1161 * builtins.def : Add BUILT_IN_FREE.
1163 2007-01-12 Jan Hubicka <jh@suse.cz>
1165 PR tree-optimization/30443
1166 * tree-inline.c (tree_function_versioning): Do not optimize when
1167 cloning for inlining.
1169 2007-01-12 Zdenek Dvorak <dvorakz@suse.cz>
1171 * doc/loop.texi: Document recording of loop exits.
1172 * cfgloopmanip.c (loopify, duplicate_loop): Use alloc_loop.
1173 (update_single_exits_after_duplication,
1174 update_single_exit_for_duplicated_loop,
1175 update_single_exit_for_duplicated_loops): Removed.
1176 (duplicate_loop_to_header_edge): Do not call
1177 update_single_exits_after_duplication and
1178 update_single_exit_for_duplicated_loops.
1179 (loop_version): Do not update single_exit information.
1180 (fix_loop_structure): Use record_loop_exits instead of
1181 mark_single_exit_loops.
1182 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update
1183 the lists of loop exits.
1184 * cfghooks.c (redirect_edge_and_branch, redirect_edge_and_branch_force,
1185 split_edge, merge_blocks): Update the lists of loop exits.
1186 * modulo-sched.c (sms_schedule): Pass LOOPS_HAVE_RECORDED_EXITS to
1187 loop_optimizer_init.
1188 * loop-init.c (loop_optimizer_init): Call record_loop_exits instead
1189 of mark_single_exit_loops.
1190 (loop_optimizer_finalize): Call release_recorded_exits.
1191 * tree-ssa-loop.c (tree_loop_optimizer_init): Pass
1192 LOOPS_HAVE_RECORDED_EXITS to loop_optimizer_init.
1193 * tree-vectorizer.c (slpeel_tree_duplicate_loop_to_edge_cfg): Do not
1194 update single exit information.
1195 * lambda-code.c (perfect_nestify): Ditto.
1196 * cfgloop.c (flow_loop_free): Destroy the list of exits of the loop.
1197 (mark_single_exit_loops): Removed.
1198 (alloc_loop, loop_exit_hash, loop_exit_eq, loop_exit_free,
1199 get_exit_descriptions, rescan_loop_exit, record_loop_exits,
1200 dump_recorded_exit, dump_recorded_exits, release_recorded_exits): New
1202 (get_loop_exit_edges, single_exit): Use recorded exit lists.
1203 (add_bb_to_loop, remove_bb_from_loops): Update the lists of loop exits.
1204 (verify_loop_structure): Verify consistency of the exit lists.
1205 (flow_loops_find): Use alloc_loop. Initialize exits hash.
1206 (set_single_exit): Removed.
1207 * cfgloop.h (struct loop_exit): New function.
1208 (struct loop): single_exit_ field replaced by exits field.
1209 (LOOPS_HAVE_MARKED_SINGLE_EXITS): Replaced by LOOPS_HAVE_RECORDED_EXITS.
1210 (struct loops): Added exits hash.
1211 (mark_single_exit_loops, set_single_exit): Declaration removed.
1212 (release_recorded_exits, record_loop_exits, rescan_loop_exit): Declare.
1214 2007-01-12 Richard Sandiford <richard@codesourcery.com>
1216 * doc/invoke.texi: Avoid use of @headitem.
1218 2007-01-12 Richard Sandiford <richard@codesourcery.com>
1220 * cse.c (cse_insn): Move HAVE_CC0 code after declarations.
1222 2007-01-12 Richard Sandiford <richard@codesourcery.com>
1224 * doc/install.texi: Fix m68k-*-* anchor and add m68k-*-* to the
1227 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
1228 Richard Sandiford <richard@codesourcery.com>
1230 * doc/invoke.texi: Document -mno-bitfield, -mno-rtd and -mno-short.
1231 * config/m68k/m68k.opt: Resort options.
1232 (mbitfield, mrtd, mshort): Remove RejectNegative properties.
1234 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
1235 Richard Sandiford <richard@codesourcery.com>
1237 * doc/invoke.texi: Document the macros that are defined by
1238 m68k's -mtune and -mhard-float options.
1239 * config/m68k/m68k-protos.h (m68k_cpp_cpu_ident) Declare.
1240 (m68k_cpp_cpu_family): Likewise.
1241 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Add a full set
1242 of __ucfv*__ macros. Define __mcffpu__ if generating code for
1243 ColdFire FPUs. Define __mcf_cpu_* and __mcf_family_* macros.
1244 * config/m68k/m68k.c (m68k_cpp_cpu_ident): New function.
1245 (m68k_cpp_cpu_family): Likewise.
1247 2007-01-12 Richard Sandiford <richard@codesourcery.com>
1249 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Treat all mc68*
1250 macros besides mc68000 as tuning macros. Use a switch statement
1251 to set them and mcpu32.
1253 2007-01-12 Julian Brown <julian@codesourcery.com>
1255 * config/m68k/m68k.h: Use TARGET_68040 instead of TARGET_68040_ONLY.
1256 (TARGET_68040_ONLY): Rename to...
1257 (TARGET_68040): ...this.
1258 * config/m68k/m68k.c: Use TARGET_68040 instead of TARGET_68040_ONLY.
1259 * config/m68k/m68k.md: Likewise.
1261 2007-01-12 Julian Brown <julian@codesourcery.com>
1262 Nathan Sidwell <nathan@codesourcery.com>
1263 Richard Sandiford <richard@codesourcery.com>
1265 * config.gcc (m680[012]0-*-*, m68k*-*-*): Set m68k_cpu_ident to
1266 the -mcpu= argument associated with the --with-cpu setting.
1267 Define M68K_DEFAULT_TUNE to the default -mtune= option,
1268 if different from the one implied by the -mcpu setting.
1269 Accept --with-cpu=FOO if FOO is listed in m68k-devices.def,
1270 using mcpu=FOO as the default CPU option. Set target_cpu_default2.
1271 * doc/invoke.texi: Mention ColdFire in the introduction to the
1272 m68k options. Document the new -march, -mcpu, -mtune, -mdiv,
1273 -mno-div and -mhard-float options. Make -m68881 a synonym for
1274 -mhard-float. Document the previously-undocumented -m5206e,
1275 -m528x, -m5307 and -m5407 options. Tweak the existing option
1276 documentation for consistency.
1277 * doc/install.texi: Mention new --with-cpu arguments.
1278 * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Only use the
1279 default CPU if neither -mcpu nor -march are specified.
1280 (ASM_CPU_SPEC): Pass down -mcpu and -march options.
1281 (TARGET_CPU_CPP_BUILTINS): Set __mcfisa*__ macros from
1282 TARGET_ISA*. Set the legacy __mcf*__ cpu macros in the same way,
1283 using m68k_tune to decide between families that implement the
1284 same ISA. Use m68k_tune to set __mcfv4e__.
1285 (FL_BITFIELD, FL_68881, FL_COLDFIRE, FL_CF_HWDIV, FL_CF_MAC)
1286 (FL_CF_EMAC, FL_CF_EMAC_B, FL_CF_USP, FL_CF_FPU, FL_ISA_68000)
1287 (FL_ISA_68010, FL_ISA_68020, FL_ISA_68040, FL_ISA_A, FL_ISA_B)
1288 (FL_ISA_C, FL_ISA_MMU): New macros.
1289 (MASK_COLDFIRE): Delete.
1290 (TARGET_68010, TARGET_68020, TARGET_68040_ONLY, TARGET_COLDFIRE)
1291 (TARGET_ISAB): Redefine in terms of m68k_cpu_flags.
1292 (TARGET_68881, TARGET_COLDFIRE_FPU): Redefine in terms of m68k_fpu.
1293 (TARGET_HARD_FLOAT): Do not define here.
1294 (TARGET_ISAAPLUS, TARGET_ISAC): New macros.
1295 (TUNE_68000): New macro.
1296 (TUNE_68000_10): Redefine in terms of TUNE_68000 and TUNE_68010.
1297 (TUNE_68010, TUNE_68030, TUNE_68040, TUNE_68060, TUNE_CPU32)
1298 (TUNE_CFV2): Redefine in terms of m68k_tune.
1299 (uarch_type, target_device, fpu_type): New enums.
1300 (m68k_cpu, m68k_tune, m68k_fpu, m68k_cpu_flags): Declare.
1301 * config/m68k/m68k.c (TARGET_DEFAULT): Remove MASK_68881.
1302 (FL_FOR_isa_00, FL_FOR_isa_10, FL_FOR_isa_20, FL_FOR_isa_40)
1303 (FL_FOR_isa_cpu32, FL_FOR_isa_a, FL_FOR_isa_aplus, FL_FOR_isa_b)
1304 (FL_FOR_isa_c): New macros.
1305 (m68k_isa): New enum.
1306 (m68k_target_selection): New structure.
1307 (all_devices, all_isas, all_microarchs): New tables.
1308 (m68k_cpu_entry, m68k_arch_entry, m68k_tune_entry, m68k_cpu)
1309 (m68k_tune, m68k_fpu, m68k_cpu_flags): New variables.
1310 (MASK_ALL_CPU_BITS): Delete.
1311 (m68k_find_selection): New function.
1312 (m68k_handle_option): Handle -mcpu=, -march= and -mtune=.
1313 Map the legacy target options to a combination of the new ones.
1314 (override_options): Set m68k_cpu, m68k_tune, m68k_fpu and
1315 m68k_cpu_flags. Handle M68K_DEFAULT_TUNE. Use m68k_cpu_flags
1316 to derive default MASK_BITFIELD, MASK_CF_HWDIV and MASK_HARD_FLOAT
1318 * config/m68k/m68k.opt (m5200, m5206e, m528x, m5307, m5407, mcfv4e)
1319 (m68010, m68020, m68020-40, m68020-60, m68030, m68040): Remove Mask
1321 (m68881, msoft-float): Change mask from 68881 to HARD_FLOAT.
1322 (march=, mcpu=, mdiv, mhard-float, mtune=): New options.
1323 * config/m68k/m68k-devices.def: New file.
1325 2007-01-12 Richard Sandiford <richard@codesourcery.com>
1326 Nathan Sidwell <nathan@codesourcery.com>
1328 * config/m68k/m68k.h (ASM_CPU_SPEC, ASM_SPEC, EXTRA_SPECS)
1329 (SUBTARGET_EXTRA_SPECS): New macros.
1330 * config/m68k/linux.h (ASM_SPEC): Remove CPU flags;
1331 use %(asm_cpu_spec) instead.
1332 * config/m68k/m68k-none.h (ASM_SPEC): Likewise.
1333 * config/m68k/openbsd.h (ASM_SPEC): Likewise.
1334 * config/m68k/netbsd-elf.h (ASM_SPEC): Likewise.
1335 (EXTRA_SPECS): Rename to...
1336 (SUBTARGET_EXTRA_SPECS): ...this.
1338 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
1339 Richard Sandiford <richard@codesourcery.com>
1340 Julian Brown <julian@codesourcery.com>
1342 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
1343 (m68k-*-uclinux*, m68k-*-linux*, m68k-*-rtems*): Set default_m68k_cpu
1344 to the configuration's default CPU.
1345 (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*): Likewise.
1346 Remove default masks.
1347 (m680[012]0-*-*): Set the default with_cpu to the first part of
1349 (m68k*-*-*): Set the default with_cpu to m$default_m68k_cpu.
1350 (m68k*-*-linux): Extend the --with-cpu handling to...
1351 (m680[012]0-*-*, m68k*-*-*): ...these configurations. Allow m68000
1352 and m68010. Don't set target_cpu_default2.
1353 * doc/install.texi: Document --with-cpu for m68k.
1354 * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Define.
1355 * config/m68k/m68k-none.h (TARGET_CPU_DEFAULT, M68K_CPU_m68k)
1356 (M68K_CPU_m68000, M68K_CPU_m68010, M68K_CPU_m68020, M68K_CPU_m68030)
1357 (M68K_CPU_m68040, M68K_CPU_m68302, M68K_CPU_m68332, TARGET_DEFAULT)
1358 (ASM_CPU_DEFAULT_SPEC, CC1_CPU_DEFAULT_SPEC): Delete.
1359 (ASM_SPEC): Remove use of %(asm_cpu_default).
1360 (EXTRA_SPECS, SUBTARGET_EXTRA_SPECS, MULTILIB_DEFAULTS): Delete.
1361 * config/m68k/linux.h (TARGET_DEFAULT): Delete.
1362 (CPP_SPEC): Merge definitions. Do not handle __HAVE_68881__ here.
1363 * config/m68k/netbsd-elf.h (TARGET_OS_CPP_BUILTINS): Define
1364 __HAVE_FPU__ if TARGET_HARD_FLOAT.
1365 (TARGET_DEFAULT): Delete.
1366 (EXTRA_SPECS): Delete cpp_cpu_default_spec, cpp_cpu_spec,
1367 cpp_fpu_spec, asm_default_spec and netbsd_cpp_spec.
1368 (CPP_CPU_SPEC): Delete.
1369 (TARGET_VERSION): Merge definitions, using TARGET_68010 to pick
1370 the appropriate string.
1371 (CPP_CPU_DEFAULT_SPEC, ASM_DEFAULT_SPEC, CPP_FPU_SPEC): Delete.
1372 (CPP_SPEC): Define to NETBSD_CPP_SPEC.
1373 (ASM_SPEC): Don't use %(asm_default_spec).
1374 * config/m68k/m68k.c (TARGET_DEFAULT_TARGET_FLAGS): Remove
1375 TARGET_DEFAULT and add MASK_68881.
1376 * config/m68k/m68k.md: Remove mention of TARGET_DEFAULT from comments.
1378 2007-01-12 Richard Sandiford <richard@codesourcery.com>
1380 * config.gcc (m68010-*-netbsdelf*): Add MASK_68010.
1381 (m68k*-*-netbsdelf*, m68k*-*-openbsd*, m68k*-linux*): Add
1382 MASK_68010 alongside MASK_68020.
1383 * doc/invoke.texi: Document -m68010.
1384 * config/m68k/m68k.opt (m68010): New.
1385 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Define mc68010
1387 (TUNE_68010): New macro.
1388 * config/m68k/m68k-none.h (M68K_CPU_m68k, M68K_CPU_m68010)
1389 (M68K_CPU_m68020, M68K_CPU_m68030, M68K_CPU_m68040)
1390 (M68K_CPU_m68332): Add MASK_68010.
1391 * config/m68k/linux.h (TARGET_DEFAULT): Add MASK_68010 to
1392 fallback definition.
1393 * config/m68k/netbsd-elf.h (CPP_CPU_SPEC): Remove now-redundant
1395 * config/m68k/m68k.c (MASK_ALL_CPU_BITS): Add MASK_68010.
1396 (m68k_handle_option): Handle OPT_m68010. Add MASK_68010
1397 to all entries that use MASK_68020.
1398 (output_move_simode_const, output_move_himode, output_move_qimode)
1399 (output_move_stricthi, output_move_strictqi): Use TARGET_68010
1400 instead of TARGET_68020 to select clr behavior. Remove comment
1401 about there being no TARGET_68010.
1402 * config/m68k/m68k.md: Likewise throughout.
1404 2007-01-12 Julian Brown <julian@codesourcery.com>
1406 * config/m68k/m68k.h (TARGET_ISAB): New macro.
1407 * config/m68k/m68k.c: Use TARGET_ISAB rather than TARGET_CFV4.
1408 * config/m68k/m68k.md: Likewise.
1410 2007-01-12 Julian Brown <julian@codesourcery.com>
1412 * config/m68k/m68k.h (LEGITIMATE_INDEX_P, LEGITIMIZE_ADDRESS): Use
1413 TARGET_COLDFIRE_FPU instead of TARGET_CFV4E.
1415 2007-01-12 Julian Brown <julian@codesourcery.com>
1417 * config/m68k/m68k.h (TUNE_68040_60): New macro.
1418 * config/m68k/m68k.c (standard_68881_constant_p): Use it.
1419 * config/m68k/m68k.md: Likewise.
1421 2007-01-12 Julian Brown <julian@codesourcery.com>
1422 Richard Sandiford <richard@codesourcery.com>
1424 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Use TUNE_68030
1425 instead of TARGET_68030, TUNE_68040 instead of TARGET_68040,
1426 TUNE_68060 instead of TARGET_68060 and TUNE_CPU32 instead of
1428 (TARGET_CPU32): Rename to...
1429 (TUNE_CPU32): ...this.
1430 (TUNE_68000_10, TUNE_68030, TUNE_68040, TUNE_68060)
1431 (TUNE_CFV2): New macros.
1432 * config/m68k/netbsd-elf.h (LONG_DOUBLE_TYPE_SIZE): Simplify;
1433 remove conditions that are implied by TARGET_68020.
1434 * config/m68k/m68k.c (m68k_output_function_prologue): Use TUNE_68040
1435 instead of TARGET_68040 and TUNE_CPU32 instead of TARGET_CPU32.
1436 (m68k_output_function_epilogue): Likewise.
1437 (m68k_rtx_costs): Likewise. Use TUNE_68060 instead of TARGET_68060
1438 and TUNE_CFV2 instead of TARGET_5200. Use TUNE_68000_10 instead of
1439 "!TARGET_68020 && !TARGET_COLDFIRE" to choose between 68000 and
1440 non-68000 timings. Refactor multiplication and division costs.
1441 (output_addsi3): Use TUNE_68040 instead of TARGET_68040 and
1442 TUNE_CPU32 instead of TARGET_CPU32.
1443 (standard_68881_constant_p): Use TUNE_68040 instead of TARGET_68040
1444 and TUNE_68060 instead of TARGET_68060.
1445 * config/m68k/m68k.md: Use TUNE_68040 instead of TARGET_68040,
1446 TUNE_68060 instead of TARGET_68060, and TUNE_CPU32 instead of
1448 (movsi_const0): Use TUNE_68000_10 rather than "!TARGET_68020
1449 && !TARGET_COLDFIRE" to choose between moveq and clr.
1450 Likewise in the unnamed movsf pattern.
1451 (ashlsi_17_24, lshrsi_17_24): Guard with TUNE_68000_10 rather than
1452 "!TARGET_68020 && !TARGET_COLDFIRE". Likewise the unnamed
1455 2007-01-12 Richard Sandiford <richard@codesourcery.com>
1457 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Increase amount
1458 of tabbing before backslashes.
1460 2007-01-11 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
1462 * pa-linux.h (ASM_OUTPUT_INTERNAL_LABEL): Undefine.
1463 * pa.h (ASM_OUTPUT_LABEL): Output colon when using GAS.
1464 (ASM_OUTPUT_INTERNAL_LABEL): Define.
1466 2007-01-11 Zdenek Dvorak <dvorakz@suse.cz>
1468 * tree-ssa-loop-ivopts.c (extract_cond_operands): Split from
1469 find_interesting_uses_cond.
1470 (find_interesting_uses_cond): Use extract_cond_operands.
1471 (rewrite_use_compare): Use extract_cond_operands and
1472 force_gimple_operand_bsi. Do not call update_stmt.
1473 (determine_use_iv_cost_condition): Use extract_cond_operands.
1474 Return cheaper of using original bound and changing the exit bound.
1476 2007-01-11 Zdenek Dvorak <dvorakz@suse.cz>
1478 PR tree-optimization/29516
1479 * tree-ssa-address.c (tree_mem_ref_addr, add_to_parts,
1480 most_expensive_mult_to_index, addr_to_parts,
1481 create_mem_ref, maybe_fold_tmr): Make the type of
1482 fields of TARGET_MEM_REF sizetype.
1483 (move_fixed_address_to_symbol, move_pointer_to_base):
1485 * tree.def (TARGET_MEM_REF): Add comment on types of
1488 2007-01-11 Joseph Myers <joseph@codesourcery.com>
1490 * c-common.c (vector_types_convertible_p): Treat opaque types as
1491 always convertible if they have the same size, but not otherwise.
1493 2007-01-11 Steven Bosscher <steven@gcc.gnu.org>
1495 * ifcvt.c (struct noce_if_info): Add comments to the fields.
1496 Remove the b_unconditional field.
1497 (noce_try_sign_mask): Do not look at b_unconditional.
1498 (noce_process_if_block): Do not use merge_if_blocks. Update
1499 the CFG here. Do not set b_unconditional.
1500 (cond_move_process_if_block): Likewise.
1501 (find_cond_trap): Likewise.
1502 (check_cond_move_block): Require simple jump insns at the end
1505 2007-01-11 Jan Hubicka <jh@suse.cz>
1507 PR tree-optimization/1046
1508 * tree-tailcall.c (suitable_for_tail_call_opt_p): Use TREE_ADDRESSABLE
1509 when alias info is not ready.
1510 (pass_tail_recursion): Do not require aliasing.
1511 * tree-ssa-copyrename.c (pass_rename_ssa_cop): Likewise.
1512 * tree-ssa-ccp.c (pass_ccp, pass_fold_builtins): Likewise.
1513 * tree-ssa-copy.c (pass_copy_prop): Likewise.
1514 * tree-ssa-forwprop.c (pass_forwprop): Likewise.
1515 * tree-ssa-dce.c (pass_dce, pass_dce_loop, pass_cd_dce): Likewise.
1516 * passes.c (init_optimization_passes): Execute rename_ssa_copies,
1517 ccp, forwprop, copy_prop, merge_phi, copy_prop, dce and tail recursion
1519 * tree-ssa-operands.c (add_virtual_operand, get_indirect_ref_operand):
1520 When aliasing is not build, mark statement as volatile.
1522 2007-01-11 Tom Tromey <tromey@redhat.com>
1524 PR preprocessor/15185, PR preprocessor/20989:
1525 * doc/cppopts.texi <-MT>: Update description of algorithm for
1526 computing default target.
1527 <-M, -MD>: Reword "basename" text.
1529 2007-01-11 Roger Sayle <roger@eyesopen.com>
1531 * builtins.c (expand_builtin_pow, expand_builtin_powi,
1532 fold_builtin_cabs, fold_builtin_sqrt, fold_builtin_trunc,
1533 fold_builtin_floor, fold_builtin_ceil, fold_builtin_round,
1534 fold_builtin_int_int_roundingfn, fold_builtin_bitop,
1535 fold_builtin_bswap, real_constp, fold_builtin_pow,
1536 fold_builtin_powi, fold_builtin_signbit, fold_builtin_copysign,
1537 do_mpfr_arg1, do_mpfr_arg2, do_mpfr_arg3, do_mpfr_sincos): Replace
1538 uses of the macro TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
1539 * convert.c (convert_to_pointer): Likewise.
1540 * expr.c (highest_pow2_factor, expand_expr_real_1): Likewise.
1541 * fold-const.c (force_fit_type, fold_negate_expr, int_const_binop,
1542 const_binop, fold_convert_const_int_from_int,
1543 fold_convert_const_int_from_real,
1544 fold_convert_const_real_from_real, sign_bit_p,
1545 optimize_minmax_comparison, extract_muldiv_1, fold_div_compare,
1546 fold_sign_changed_comparison, fold_unary, fold_comparison,
1547 fold_binary, multiple_of_p, tree_Expr_non_zero_p,
1548 fold_negate_const, fold_abs_const, fold_not_const): Likewise.
1549 * print-tree.c (print_node_brief, print_node): Likewise.
1550 * stor-layout.c (place_field, layout_type): Likewise.
1551 * tree-chrec.c (keep_cast): Likewise.
1552 * tree.c (build_vector, build_real, build_real_from_int_cst,
1553 build_complex): Likewise.
1555 2007-01-11 Roger Sayle <roger@eyesopen.com>
1557 * tree.h (TREE_CONSTANT_OVERFLOW): Obsolete. For the time being,
1558 treat TREE_CONSTANT_OVERFLOW as a synonym of TREE_OVERFLOW.
1560 2007-01-11 Paolo Bonzini <bonzini@gnu.org>
1562 * configure.ac (strict1_warn): Rename to strict_warn.
1563 (WERROR, --enable-werror, symlink hacks, stage1_cflags,
1564 cc_set_by_configure, quoted_cc_set_by_configure,
1565 stage_prefix_set_by_configure, quoted_stage_prefix_set_by_configure,
1566 all_boot_languages, all_stagestuff): Remove.
1567 (target_list): Remove bootstrap targets.
1568 * Makefile.in (quickstrap): Unconditionally make a synonym of all.
1569 (BOOT_LANGUAGES, STAGE1_CFLAGS, STAGE1_CHECKING,
1570 REMAKEFLAGS, FLAGS_TO_PASS, PREPEND_DOTDOT_TO_RELATIVE_PATHS,
1571 SUBDIR_FLAGS_TO_PASS, WERROR_FLAGS, STRICT2_WARN, LANG_STAGESTUFF,
1572 VOL_FILES, POSTSTAGE1_FLAGS_TO_PASS, STAGE2_FLAGS_TO_PASS,
1573 STAGEPROFILE_FLAGS_TO_PASS, STAGEFEEDBACK_FLAGS_TO_PASS, stage1_build,
1574 stage1_copy, stage2_build, stage2_copy, stageprofile_build,
1575 stageprofile_copy, stage3_build, stage3_copy, stagefeedback_build,
1576 stagefeedback_copy, stage4_build, clean_s1, clean_sw, bootstrap,
1577 bootstrap-lean, bootstrap2, bootstrap2-lean, bootstrap3,
1578 bootstrap3-lean, bootstrap4, bootstrap4-lean, unstage1, unstage2,
1579 unstage3, unstage4, unstageprofile, unstagefeedback, restage, restage2,
1580 restage3, restage4, restageprofile, restagefeedback, bubbleestrap,
1581 cleanstrap, unstrap, restrap, *compare, *compare3, *compare4,
1582 *compare-lean, *compare3-lean, *compare4-lean, stage1-start, stage1,
1583 stage2-start, stage2, stage3-start, stage3, stage4-start, stage4,
1584 stageprofile-start, stageprofile, stagefeedback-start, stagefeedback,
1585 risky-stage1, risky-stage2, risky-stage3, risky-stage4): Remove.
1586 (ORDINARY_FLAGS_TO_PASS): Rename to FLAGS_TO_PASS.
1587 (STAGECOPYSTUFF, STAGEMOVESTUFF): Consolidate into MOSTLYCLEANFILES.
1588 (mostlyclean): Adjust.
1589 (clean, distclean): Don't mention bootstrap stuff.
1590 * configure: Regenerate.
1591 * ada/config-lang.in, cp/config-lang.in, forttran/config-lang.in,
1592 java/config-lang.in, objc/config-lang.in, objcp/config-lang.in,
1593 treelang/config-lang.in (stagestuff): Remove.
1594 * doc/sourcebuild.texi (stage1, stage2, stage3, stage4,
1595 stageprofile, stagefeedback, stagestuff): Remove mention.
1597 2007-01-11 Nick Clifton <nickc@redhat.com>
1599 * config/mcore/predicates.md (mcore_general_movesrc_operand):
1601 (mcore_general_movdst_operand): Do not accept CONST_INTs.
1602 (mcore_arith_K_S_operand): Run the test for the S constraint not
1603 the test for the M constraint.
1604 (mcore_addsub_operand): Do not accept integer values that are
1605 larger than 32 bits.
1606 * config/mcore/mcore.md: Remove unused constraints from split.
1607 (andsi3): Use HOST_WIDE_INT instead of int to hold an INTVAL.
1609 (allocate_stack): Likewise.
1610 * config/mcore/mcore.c (mcore_print_operand): Restrict output of P
1611 operands to 32 bits.
1612 (mcore_const_costs): Use HOST_WIDE_INT instead of int to hold an
1614 (mcore_and_cost, mcore_modify_comparison, const_ok_for_mcore,
1615 mcore_const_ok_for_inline, mcore_const_trick_uses_not,
1616 try_constant_tricks, mcore_num_ones, mcore_num_zeros,
1617 mcore_output_bclri, mcore_output_andn, output_inline_const,
1618 mcore_output_move, mcore_output_movedouble): Likewise.
1619 (mcore_output_cmov): Use CONST_OK_FOR_M and CONST_OK_FOR_N.
1620 (output_inline_const): Likewise.
1621 (output_inline_const): Fix format strings used in sprintf
1623 * config/mcore/mcore-protos.h: Update prototypes for changed
1624 functions in mcore.c.
1625 * config/mcore/mcore.h (CONST_OK_FOR_I): Cast values to
1626 HOST_WIDE_INT and not int.
1627 (CONST_OK_FOR_J, CONST_OK_FOR_K, CONST_OK_FOR_L, CONST_OK_FOR_M,
1628 CONST_OK_FOR_N): Likewise.
1629 (LEGITIMATE_CONSTANT_P): Also check CONSTANT_P.
1630 (GO_IF_LEGITIMATE_INDEX): Use HOST_WIDE_INT instead of int to hold
1633 2007-01-10 Jan Hubicka <jh@suse.cz>
1635 * tree-vrp.c (remove_range_assertions): Release defs.
1636 * tree-ssa-loop-ivopts.c (rmeove_statement): Likewise.
1637 * tree-ssa-dom.c (remove_stmt_or_phi): Likewise.
1639 2007-01-10 Paul Brook <paul@codesourcery.com>
1641 * config/arm/arm.c (arm_rtx_costs_1): Handle mutiply-subtract.
1642 * config/arm/arm.md (mulsi3subsi): New insn.
1644 2007-01-10 Zdenek Dvorak <dvorakz@suse.cz>
1646 * tree-ssa-loop-manip.c (tree_unroll_loop): Make it a wrapper over ...
1647 (tree_transform_and_unroll_loop): New.
1648 * tree-flow.h (transform_callback, tree_transform_and_unroll_loop):
1651 2007-01-10 Robert Kennedy <jimbob@google.com>
1653 * fold-const.c (fold_comparison): Fold comparisons like (x *
1654 1000 < 0) to (x < 0).
1656 2007-01-10 Ian Lance Taylor <iant@google.com>
1658 * tree-pretty-print.c (dump_generic_node): Print parentheses when
1659 operands have the same priority.
1661 2007-01-10 Tom Tromey <tromey@redhat.com>
1663 * fold-const.c (fold_truthop): Don't check can_use_bit_fields_p.
1664 (fold_binary): Likewise.
1665 * langhooks.c (lhd_can_use_bit_fields_p): Removed.
1666 * langhooks-def.h (lhd_can_use_bit_fields_p): Removed.
1667 (LANG_HOOKS_CAN_USE_BIT_FIELDS_P): Removed.
1668 (LANG_HOOKS_INITIALIZER): Remove LANG_HOOKS_CAN_USE_BIT_FIELDS_P.
1669 * langhooks.h (struct lang_hooks): Removed field
1670 'can_use_bit_fields_p'.
1672 2007-01-10 Ralf Corsépius <ralf.corsepius@rtems.org>
1674 * config/bfin/t-bfin, config/bfin/t-bfin-elf: Remove GCC_CFLAGS.
1676 2007-01-10 Razya Ladelsky <razya@il.ibm.com>
1678 * function.c (get_last_funcdef_no): New function.
1679 * function.h (get_last_funcdef_no): Declare.
1680 * tree-inline.c (initialize_cfun): Add initialization.
1681 (tree_function_versioning): Cleanup.
1683 2007-01-10 Jan Hubicka <jh@suse.cz>
1685 * tree-inline.c (setup_one_parameter): Do not propagate into abnormal
1688 2007-01-10 Sa Liu <saliu@de.ibm.com>
1689 Ben Elliston <bje@au.ibm.com>
1691 * spu.h (STACK_SAVE_AREA): Use VOIDmode for SAVE_FUNCTION, SImode
1692 for SAVE_NONLOCAL and Pmode for any other save level.
1693 * spu-protos.h (spu_restore_stack_block): Declare.
1694 * spu.md (save_stack_block): Remove.
1695 (restore_stack_block): Call spu_restore_stack_block.
1696 * spu.c (spu_restore_stack_block): New function.
1697 (spu_expand_epilogue): Remove old comment.
1699 2007-01-09 Zdenek Dvorak <dvorakz@suse.cz>
1701 PR tree-optimization/30322
1702 * tree-ssa-loop-ivopts.c (fold_affine_expr, iv_value): Removed.
1703 (cand_value_at): Return the value as aff_tree.
1704 (may_eliminate_iv): Convert the bound from aff_tree to tree.
1705 * tree-affine.c (aff_combination_add_cst, aff_combination_add_product,
1706 aff_combination_mult): New functions.
1707 (aff_combination_add): Use aff_combination_add_cst.
1708 (aff_combination_convert): Allow conversions to a wider type.
1709 (tree_to_aff_combination): Handle BIT_NOT_EXPR.
1710 * tree-affine.h (aff_combination_mult): Declare.
1712 2007-01-09 Carlos O'Donell <carlos@codesourcery.com>
1714 * doc/tm.texi: Update documentation to reflect reality of exec
1715 and start file search behaviours. Update copyright year.
1716 * doc/invoke.texi: Explain how GCC_EXEC_PREFIX is used to find
1717 header file directories.
1719 2007-01-09 Uros Bizjak <ubizjak@gmail.com>
1721 * config/i386/i386.md (*sinxf2): Rename to *sinxf2_i387.
1722 (*cosxf2): Rename to cosxf2_i387.
1723 (*sindf2, *sinsf2): Extend operand 1 to XFmode. Macroize patterns
1724 using X87MODEF12 mode macro. Rename patterns to
1725 *sin_extend<mode>xf2_i387. Use SSE_FLOAT_MODE_P to disable patterns
1727 (*cosdf2, *cossf2): Ditto.
1728 (sincosdf3, sincossf3): Ditto. Rewrite corresponding splitters
1729 to match extended input operands.
1730 (sincos<mode>3): New expander.
1731 (*sinextendsfdf2, *cosextendsfdf2, *sincosextendsfdf3): Remove
1732 insn patterns and corresponding splitters.
1734 2007-01-09 Kaz Kojima <kkojima@gcc.gnu.org>
1736 * config/sh/t-linux (TARGET_LIBGCC2_CFLAGS): Delete.
1737 (SHLIB_MAPFILES, SHLIB_LINK, SHLIB_INSTALL): Likewise.
1739 2007-01-09 Nicolas Pitre <nico@cam.org>
1742 * arm/ieee754-df.S (Lad_s): Also test the low word of X for zero.
1744 2007-01-08 Geoffrey Keating <geoffk@apple.com>
1746 * target.h (struct gcc_target): New field library_rtti_comdat.
1747 * target-def.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): New.
1748 (TARGET_CXX): Add TARGET_CXX_LIBRARY_RTTI_COMDAT.
1749 * doc/tm.texi (C++ ABI): Document TARGET_CXX_LIBRARY_RTTI_COMDAT.
1750 * config/darwin.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
1752 2007-01-08 Geoffrey Keating <geoffk@apple.com>
1754 * doc/invoke.texi (Optimize Options): Correct description of -O0.
1756 2007-01-08 Richard Guenther <rguenther@suse.de>
1758 * tree.h (force_fit_type_double): Export.
1759 (force_fit_type): Remove.
1760 * fold-const.c (force_fit_type_double): New function.
1761 (force_fit_type): Remove.
1762 (int_const_binop): Use it.
1763 (fold_convert_const_int_from_int): Likewise.
1764 (fold_convert_const_int_from_real): Likewise.
1765 (fold_div_compare): Likewise.
1766 (fold_sign_changed_comparison): Likewise.
1767 (fold_unary): Likewise.
1768 (fold_negate_const): Likewise.
1769 (fold_abs_const): Likewise.
1770 (fold_not_const): Likewise.
1771 * c-common.c (shorten_compare): Use force_fit_type_double.
1772 * convert.c (convert_to_pointer): Likewise.
1774 2007-01-08 Richard Guenther <rguenther@suse.de>
1776 * tree.h (build_int_cst_wide_type): Export.
1777 * tree.c (build_int_cst_wide_type): New function.
1778 (build_int_cst_wide): Fix comment.
1779 * builtins.c (fold_builtin_object_size): Use build_int_cst
1780 to build -1 or 0 of the correct type. Use fit_double_type
1781 to check for overflow.
1782 * fold-const.c (optimize_bit_field_compare): Use build_int_cst_type
1784 (decode_field_reference): Likewise.
1785 (all_ones_mask_p): Likewise.
1786 (native_interpret_int): Use build_int_cst_wide_type.
1787 (fold_binary): Use build_int_cst_type to build an all-ones
1789 * stor-layout.c (set_sizetype): Use build_int_cst_wide_type.
1791 2007-01-08 Daniel Jacobowitz <dan@codesourcery.com>
1793 * config/pa/t-pa64 (libgcc_stub.a): Use $(T).
1795 2007-01-09 Ben Elliston <bje@au.ibm.com>
1797 * genautomata.c (STATS_OPTION): New option.
1798 (stats_flag): New flag.
1799 (gen_automata_option): Handle it.
1800 (initiate_automaton_gen): Ditto.
1801 (write_automata): Output statistics only if stats_flag is
1802 set. Likewise, output time statistics only if time_flag is set.
1803 * doc/md.texi (Processor pipeline description): Document new flag.
1805 2007-01-08 Richard Guenther <rguenther@suse.de>
1807 * builtins.c (fold_builtin_int_roundingfn): Use fit_double_type.
1808 * tree.c (build_int_cst_type): Likewise.
1809 (size_in_bytes): Don't call force_fit_type on the result.
1810 (int_fits_type_p): Use fit_double_type.
1811 * fold-const.c (fit_double_type): New function.
1812 (force_fit_type): Use it.
1813 * tree.h (fit_double_type): Export.
1815 2007-01-08 Jan Hubicka <jh@suse.cz>
1817 * tree-vectorizer.c (gate_increase_alignment): Fix return type.
1818 * ipa.c (function_and_variable_visibility): Fix return type.
1820 2007-01-08 Richard Guenther <rguenther@suse.de>
1822 * tree-ssa-ccp.c (maybe_fold_offset_to_array_ref): Use type
1823 of offset to build the index.
1824 * tree-pretty-print.c (dump_generic_node): Don't build negated
1825 const just for printing.
1826 * c-pretty-print.c (pp_c_integer_constant): Likewise.
1827 * builtins.c (fold_builtin_int_roundingfn): Check if result
1828 fits the type by using force_fit_type and comparing the result.
1829 * predict.c (predict_loops): Use compare_tree_int for comparison.
1830 * tree.c (build_int_cst): Fall back to integer_type_node for
1832 (build_int_cst_wide): Assert type is non-null.
1834 2007-01-08 Roberto Costa <roberto.costa@st.com>
1836 * tree-vrp.c (extract_range_from_cond_expr): New.
1837 (extract_range_from_expr): Handle COND_EXPR nodes used as expressions.
1838 * tree-ssa-ccp.c (get_maxval_strlen): Handle COND_EXPR nodes used
1840 (fold_stmt): Bug fix, avoid infinite recursion when folding COND_EXPRs.
1841 * tree-ssa-forwprop.c (simplify_cond, forward_propagate_into_cond,
1842 tree_ssa_forward_propagate_single_use_vars): Handle COND_EXPR nodes
1843 used as expressions.
1844 * tree-object-size.c (cond_expr_object_size): New.
1845 (collect_object_sizes_for): Handle COND_EXPR nodes used as expressions.
1847 2007-01-08 Jan Hubicka <jh@suse.cz>
1849 * tree-ssa-forwprop.c (forward_propagate_into_cond,
1850 tree_ssa_forward_propagate_single_use_va): Release defs of propagated
1853 2007-01-08 Richard Guenther <rguenther@suse.de>
1855 PR tree-optimization/23603
1856 * tree-vrp.c (set_value_range_to_truthvalue): New function.
1857 (extract_range_from_binary): Fall back to truthvalue instead of
1858 varying for TRUTH_*_EXPR.
1859 (extract_range_from_comparison): Fall back to truthvalue instead of
1861 (vrp_visit_phi_node): Don't adjust new range bounds to +INF/-INF
1862 if all visited PHI values were constant.
1864 2007-01-08 Jan Hubicka <jh@suse.cz>
1866 * cgraphunit.c (cgraph_process_new_functions): Reset reachable flag.
1867 (cgraph_analyze_function): break out from ...
1868 (cgraph_finalize_compilation_unit): ... here.
1869 (cgraph_expand_function): Remove forgoten commented out line.
1870 (cgraph_optimize): Analyze functions.
1872 2007-01-08 Jan Hubicka <jh@suse.cz>
1874 * tree-pas.h (TODO_remove_function): New flag.
1875 (TODO_update*): Renumber.
1876 (pass_ipa_increase_alignment,
1877 pass_ipa_function_and_variable_visibility): New passes.
1878 * cgraphunit.c (cgraph_increase_alignment): Move to tree-vectorizer.c
1879 (cgraph_function_and_variable_visibility): Move to ipa.c
1880 (cgraph_optimize): Don't call cgraph_function_and_variable_visibility,
1881 cgraph_increase_alignment.
1882 * ipa-inline.c (cgraph_decide_inlining): Don't push timevar.
1883 (cgraph_decide_inlining_incrementally): Push TV_INTEGRATION before
1884 calling tree-inline.
1885 (cgraph_early_inlining): Do not call cgraph_remove_unreachable_nodes.
1886 (pass_ipa_inline, pass_early_ipa_inlining): Set TODO_remove_functions
1887 * tree-vectorizer.c (increase_alignment): Move here from cgraphunit.c
1888 (gate_increase_alignment): New function.
1889 (pass_ipa_increase_alignment): New pass.
1890 * ipa.c: Inline tree-pass.h and timevar.h
1891 (function_and_variable_visibility): Move here from cgraphunit.c
1892 * tree-optimize.c (pass_early_local_passes): Add TODO_remove_functions.
1893 * passes.c (init_optimization_passes): Add the two new passes.
1894 (execute_todo): Handle cgraph_remove_functions.
1896 2007-01-08 Nick Clifton <nickc@redhat.com>
1898 * config/frv/predicates.md (reg_or_0_operand): Accept
1901 2007-01-08 Ralf Corsépius <ralf.corsepius@rtems.org>
1903 * config/bfin/rtems.h, config/bfin/t-rtems: New.
1904 * config.gcc: Add bfin*-rtems*.
1906 2007-01-08 Mark Shinwell <shinwell@codesourcery.com>
1908 * c.opt: Add -flax-vector-conversions.
1909 * c-typeck.c (convert_for_assignment): Pass flag to
1910 vector_types_convertible_p to allow emission of note.
1911 (digest_init): Likewise.
1912 * c-opts.c: Handle -flax-vector-conversions.
1913 * c-common.c (flag_lax_vector_conversions): New.
1914 (vector_types_convertible_p): Unless -flax-vector conversions
1915 has been passed, disallow conversions between vectors with
1916 differing numbers of subparts and/or element types. If such
1917 a conversion is disallowed, possibly emit a note on the first
1918 occasion only to inform the user of -flax-vector-conversions.
1919 The new last argument specifies this.
1920 * c-common.h (flag_lax_vector_conversions): New.
1921 (vector_types_convertible_p): Add extra argument.
1922 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Use
1923 char_type_node for V*QI type vectors.
1924 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins):
1925 Update to satisfy new typechecking rules.
1926 * config/rs6000/altivec.h (vec_cmple): Use vec_cmpge, for both
1928 * doc/invoke.texi (C Dialect Options): Document
1929 -flax-vector-conversions.
1931 2007-01-08 Mark Shinwell <shinwell@codesourcery.com>
1933 PR tree-optimization/29877
1934 * tree-ssa-ter.c (is_replaceable_p): Deem assignments with
1935 a register variable on the RHS to not be replaceable.
1937 2007-01-08 Chen Liqin <liqin@sunnorth.com.cn>
1938 * config/score/t-score-elf (MULTILIB_OPTIONS): Change.
1939 * config/score/predicates.md (const_uimm5, sr0_operand, const_simm12,
1940 const_simm15, const_pow2, const_npow2): Added.
1941 * config/score/misc.md (insv, extv, extzv, movmemsi,
1942 move_lbu_a/b, mov_lhu_a/b etc): Added and fix some bug.
1943 * config/score/score.c (score_address_cost, score_select_cc_mode):
1945 Change CONST_OK_FOR_LETTER_P/EXTRA_CONSTRAINT define.
1946 Update score_rtx_costs for MACRO TARGET_RTX_COSTS.
1947 Update score_print_operand.
1948 * config/score/score.h (DATA_ALIGNMENT, SELECT_CC_MODE): Added.
1949 Adjust register allocate order and update some macro define.
1950 * config/score/score-mdaux.c (mdx_unaligned_load, mdx_unsigned_store,
1951 mdx_block_move_straight, mdx_block_move_loop_head,
1952 mdx_block_move_loop_body, mdx_block_move_loop_foot, mdx_block_move_loop,
1953 mdx_block_move): Added.
1954 (mdx_movsicc, mdp_select_add_imm, mdp_select, mds_zero_extract_andi,
1955 mdp_limm): Updated and fix some bug and typo.
1956 * config/score/score.md (movqi/hi/si, add/sub/zero/ext): Updated.
1957 (movsf, movdf, doloop_end): Added.
1959 2007-01-08 Kazu Hirata <kazu@codesourcery.com>
1961 * config/arm/arm.c, config/arm/arm.h, config/arm/arm.md,
1962 config/arm/thumb2.md: Fix comment typos.
1963 * doc/extend.texi: Fix a typo.
1965 2007-01-07 Eric Christopher <echristo@apple.com>
1967 * configure.ac: Check for __stack_chk_fail for darwin.
1968 * configure: Regenerate.
1970 2007-01-07 Richard Guenther <rguenther@suse.de>
1972 * tree-vrp.c (extract_range_from_assert): CSE calls to
1973 compare_values where possible.
1974 (extract_range_from_unary_expr): Likewise.
1976 2007-01-07 Anatoly Sokolov <aesok@post.ru>
1978 * config/avr/avr-protos.h (call_insn_operand): Delete prototype.
1979 * config/avr/avr.c (call_insn_operand): Delete function.
1980 * config/avr/avr.md (*pushqi, *pushhi, *pushsi, *pushsf): Use REG_SP
1981 instead of register number. Use predicates.
1982 * config/avr/predicates.md (const0_operand, reg_or_0_operand,
1983 call_insn_operand): Add.
1985 2007-01-06 Jan Hubicka <jh@suse.cz>
1987 * tree-pass.h (pass_build_cgraph_edges): Declare.
1988 * cgraphunit.c (record_refernece): Move to cgraphbuild.c
1989 (visited_nodes): Remove.
1990 (cgraph_create_edges): Move to cgraphbuild.c; rename to
1991 build_cgrpah_edges; make visited_nodes local.
1992 (cgraph_process_new_functions): DO not call initialize_inline_failed.
1993 (record_references_in_initializer): Move to cgraphbuild.c
1994 (initialize_inline_failed, rebuild_cgraph_edges,
1995 pass_rebuild_cgraph_edges): Move to cgraphbuild.c.
1996 (verify_cgraph_node): Make visited_nodes local.
1997 (cgraph_analyze_function): Do not call cgraph_create_edges and
1998 initialize_inline_failed.
1999 (cgraph_expand_function): Do not call cgraph_lower_function;
2000 assert that function is already lowered.
2001 * Makefile.in (cgraphbuild.o): New.
2002 * passes.c (init_optimization_passes): Add pass_build_cgraph_edges
2003 at the end of lowering passes.
2005 2007-01-06 Steven Bosscher <steven@gcc.gnu.org>
2007 * ifcvt.c (cond_move_convert_if_block): New function, code
2008 factored out from...
2009 (cond_move_process_if_block): ...here. Call the new function
2010 on the THEN and ELSE blocks.
2011 (merge_if_block): Do not copy global_live_at_end, merge_blocks
2012 already takes care of this.
2014 2007-01-05 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
2017 * tree.h (TREE_OVERFLOW_P): New.
2018 * c-typeck.c (parser_build_unary_op): Warn only if result
2019 overflowed and operands did not.
2020 (parser_build_binary_op): Likewise.
2021 (convert_for_assignment): Remove redundant overflow_warning.
2022 * c-common.c (overflow_warning): Don't check or set TREE_OVERFLOW.
2024 2007-01-05 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
2026 * c-typeck.c (store_init_value): Split over two lines to follow
2027 the GNU coding style.
2029 2007-01-05 Benjamin Kosnik <bkoz@redhat.com>
2031 * c-cppbuiltin.c (c_cpp_builtins): __GXX_EXPERIMENTAL_CPP0X__ to
2032 __GXX_EXPERIMENTAL_CXX0X__.
2033 * doc/cpp.texi: Same.
2035 2007-01-05 Richard Guenther <rguenther@suse.de>
2038 * tree.c (get_narrower): Do not construct COMPONENT_REFs
2039 with mismatched types. Instead explicitly build a
2040 conversion NOP_EXPR.
2042 2007-01-05 Ian Lance Taylor <iant@google.com>
2044 * c-common.c (decl_with_nonnull_addr_p): New function.
2045 (c_common_truthvalue_conversion): Call it.
2046 * c-typeck.c (build_binary_op): Likewise.
2047 * c-common.h (decl_with_nonnull_addr_p): Declare.
2049 2007-01-05 Jakub Jelinek <jakub@redhat.com>
2052 * libgcc2.c (__divdc3): Compare c and d against 0.0 instead of
2055 2007-01-05 Joel Brobecker <brobecker@adacore.com>
2057 * doc/install.texi (Final install): Document the fact that
2058 the GNAT runtime should not be stripped.
2060 2007-01-04 Jan Hubicka <jh@suse.cz>
2062 * tree-inline.c (fold_marked_statements): Update operand caches
2063 and EH after folding
2065 2007-01-04 Ian Lance Taylor <iant@google.com>
2067 * c-common.c (check_function_nonnull): Whitespace fix.
2069 2007-01-04 Jan Hubicka <jh@suse.cz>
2071 * tree-optimize.c (execute_fixup_cfg): Correct previously mistakely
2072 comitted older version of patch.
2073 (pass_fixup_cfg): Add TODOs to verify flow and statements, dump
2074 function, celanup cfg and collect garbage.
2076 2007-01-04 Mike Stump <mrs@apple.com>
2078 * Makefile.in (mostlyclean): Don't remove libgcc anymore.
2081 2007-01-04 Eric Christopher <echristo@apple.com>
2083 * libgcc2.c (__bswapsi2): Use SItype.
2084 (__bswapdi2): Use DItype.
2085 * libgcc2.h: Update for above.
2087 2007-01-04 Paul Brook <paul@codesourcery.com>
2089 * config/arm/arm.md (arm_mulsi3, thumb_mulsi3, mulsi3_compare0,
2090 mulsi_compare0_scratch, mulsi3addsi, mulsi3addsi_compare0,
2091 mulsi3addsi_compare0_scratch, mulsidi3adddi, mulsidi3,
2092 umulsidi3, umulsidi3adddi, smulsi3_highpart,
2093 umulsi3_highpart): Make conditional on !arm_arch6.
2094 (arm_mulsi3_v6, thumb_mulsi3_v6, mulsi3_compare0_v6,
2095 mulsi_compare0_scratch_v6, mulsi3addsi_v6, mulsi3addsi_compare0_v6,
2096 mulsi3addsi_compare0_scratch_v6, mulsidi3adddi_v6, mulsidi3_v6,
2097 umulsidi3_v6, umulsidi3adddi_v6, smulsi3_highpart_v6,
2098 umulsi3_highpart_v6): New insns.
2100 2007-01-04 Roger Sayle <roger@eyesopen.com>
2102 * fold-const.c (fold_convert): When casting an expression to void,
2103 fold_ignored_result may discover a GIMPLE_MODIFY_STMT which doesn't
2104 have a type. Instead of attempting to build a NOP_EXPR, return
2105 these "special" trees directly.
2107 2007-01-04 Joseph Myers <joseph@codesourcery.com>
2109 * config/rs6000/rs6000.c (rs6000_rtx_costs): Make adjustment for
2110 MULT inside MINUS as either argument. Use rs6000_cost->dmul -
2111 rs6000_cost->fp not 0 as adjustment for outer NEG.
2113 2007-01-04 Jan Hubicka <jh@suse.cz>
2115 * cgraph.c (cgraph_release_function_body): New function.
2116 (cgraph_remove_node): Use it.
2117 * cgraph.h (cgraph_release_function_body): Declare.
2118 * cgraphunit.c (cgraph_expand_function): Use it.
2119 * ipa.c (cgraph_remove_unreahchable_nodes): Use it.
2120 * tree-ssa.c (delete_tree_ssa): Allow to be called before aliasing
2121 is initialized and while compilation of other function is running.
2122 * tree-optimize.c (execute_free_cfg_annotations): Move code to clear
2123 statement CFG annotations from here to ...
2124 * tree-cfg.c (delete_tree_cfg_annotations): ... here.
2126 2007-01-04 Zdenek Dvorak <dvorakz@suse.cz>
2128 * cfgloop.h (enum li_flags): Make the constants powers of two.
2130 2007-01-04 Jan Hubicka <jh@suse.cz>
2132 * tree-inline.c (copy_bb): Insert new statements to statements_to_fold
2134 (fold_marked_statements): New function.
2135 (optimize_inline_calls, tree_function_versioning): Fold new statements.
2136 * tree-inline.h (copy_body_data): Add statements_to_fold.
2138 2007-01-03 Daniel Jacobowitz <dan@codesourcery.com>
2140 * config.gcc: Mention libgcc/config.host.
2141 * Makefile.in: Update comments mentioning libgcc.
2142 (LIBGCC, INSTALL_LIBGCC, GCC_PARTS, mklibgcc): Delete.
2143 (all.cross, start.encap, rest.encap, rest.cross): Update
2144 dependencies for libgcc move.
2145 (libgcc.mk, LIBGCC_DEPS, libgcov.a, libgcc.a, stmp-multilib)
2146 (clean-target, clean-target-libgcc): Delete.
2147 (srcdirify, GCC_EXTRA_PARTS): New macros.
2148 (libgcc-support, libgcc.mvars): New rules.
2149 (distclean): Remove mention of mklibgcc.
2150 (install): Don't reference INSTALL_LIBGCC.
2151 (install-common): Don't reference EXTRA_PARTS.
2152 (install-libgcc, install-multilib): Delete rules.
2153 * mklibgcc.in: Delete file.
2154 * doc/configfiles.texi: Don't mention mklibgcc.
2156 * config/i386/t-darwin (SHLIB_VERPFX): Delete (moved to libgcc).
2157 * config/i386/t-darwin64 (SHLIB_VERPFX): Likewise.
2158 * config/rs6000/t-darwin (SHLIB_VERPFX): Likewise.
2159 * config/rs6000/t-ppccomm (TARGET_LIBGCC2_CFLAGS, SHLIB_MAPFILES)
2160 (mklibgcc, ldblspecs): Likewise.
2162 * config/i386/t-nwld (libgcc.def, libc.def, libpcre.def)
2163 (posixpre.def): Use $(T).
2164 (SHLIB_EXT, SHLIB_NAME, SHLIB_SLIBDIR_QUAL, SHLIB_DEF, SHLIB_MAP)
2165 (SHLIB_SRC, SHLIB_INSTALL): Delete.
2166 (SHLIB_LINK): Make dummy.
2167 * config/t-slibgcc-darwin: Delete contents except for dummy SHLIB_LINK.
2169 * config/frv/t-linux (EXTRA_MULTILIB_PARTS): Clear.
2171 * config/alpha/t-crtfm: Use $(T) in rules for EXTRA_PARTS.
2172 * config/alpha/t-vms, config/alpha/t-vms64, config/fr30/t-fr30,
2173 config/i386/t-rtems-i386, config/ia64/t-ia64, config/rs6000/t-beos,
2174 config/rs6000/t-newas, config/sparc/t-elf: Likewise.
2176 * configure.ac (all_outputs): Remove mklibgcc.
2177 * configure: Regenerated.
2179 2007-01-03 Josh Conner <jconner@apple.com>
2182 * calls.c (compute_argument_addresses): Set stack and stack_slot
2183 for partial args, too.
2184 (store_one_arg): Use locate.size.constant for the size when
2185 generating a save_area.
2187 2007-01-03 Robert Kennedy <jimbob@google.com>
2189 * tree-cfg.c (tree_merge_blocks): Release SSA_NAME phi results
2190 whose definitions are deleted due to basic block merging.
2192 2007-01-03 Paul Brook <paul@codesourcery.com>
2195 * config/arm/arm.c (output_return_instruction): Pop PC in interrupt
2197 (use_return_insn): Return 0 for Thumb interrupt functions.
2198 (print_multi_reg): Add rfe argument for IRQ returns.
2199 (arm_output_epilogue): Pop interrupt return address directly into PC.
2200 (arm_expand_prologue): Only adjust IRQ return address in Arm mode.
2202 2007-01-03 Paul Brook <paul@codesourcery.com>
2204 Merge from sourcerygxx-4_1.
2205 * config/arm/thumb2.md: New file.
2206 * config/arm/elf.h (JUMP_TABLES_IN_TEXT_SECTION): Return True for
2208 * config/arm/coff.h (JUMP_TABLES_IN_TEXT_SECTION): Ditto.
2209 * config/arm/aout.h (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
2210 (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump tables.
2211 * config/arm/aof.h (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump
2213 (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
2214 * config/arm/ieee754-df.S: Use macros for Thumb-2/Unified asm
2216 * config/arm/ieee754-sf.S: Ditto.
2217 * config/arm/arm.c (thumb_base_register_rtx_p): Rename...
2218 (thumb1_base_register_rtx_p): ... to this.
2219 (thumb_index_register_rtx_p): Rename...
2220 (thumb1_index_register_rtx_p): ... to this.
2221 (thumb_output_function_prologue): Rename...
2222 (thumb1_output_function_prologue): ... to this.
2223 (thumb_legitimate_address_p): Rename...
2224 (thumb1_legitimate_address_p): ... to this.
2225 (thumb_rtx_costs): Rename...
2226 (thumb1_rtx_costs): ... to this.
2227 (thumb_compute_save_reg_mask): Rename...
2228 (thumb1_compute_save_reg_mask): ... to this.
2229 (thumb_final_prescan_insn): Rename...
2230 (thumb1_final_prescan_insn): ... to this.
2231 (thumb_expand_epilogue): Rename...
2232 (thumb1_expand_epilogue): ... to this.
2233 (arm_unwind_emit_stm): Rename...
2234 (arm_unwind_emit_sequence): ... to this.
2235 (thumb2_legitimate_index_p, thumb2_legitimate_address_p,
2236 thumb1_compute_save_reg_mask, arm_dwarf_handle_frame_unspec,
2237 thumb2_index_mul_operand, output_move_vfp, arm_shift_nmem,
2238 arm_save_coproc_regs, thumb_set_frame_pointer, arm_print_condition,
2239 thumb2_final_prescan_insn, thumb2_asm_output_opcode, arm_output_shift,
2240 thumb2_output_casesi): New functions.
2241 (TARGET_DWARF_HANDLE_FRAME_UNSPEC): Define.
2242 (FL_THUMB2, FL_NOTM, FL_DIV, FL_FOR_ARCH6T2, FL_FOR_ARCH7,
2243 FL_FOR_ARCH7A, FL_FOR_ARCH7R, FL_FOR_ARCH7M, ARM_LSL_NAME,
2244 THUMB2_WORK_REGS): Define.
2245 (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv, arm_condexec_count,
2246 arm_condexec_mask, arm_condexec_masklen)): New variables.
2247 (all_architectures): Add armv6t2, armv7, armv7a, armv7r and armv7m.
2248 (arm_override_options): Check new CPU capabilities.
2249 Set new architecture flag variables.
2250 (arm_isr_value): Handle v7m interrupt functions.
2251 (user_return_insn): Return 0 for v7m interrupt functions. Handle
2253 (const_ok_for_arm): Handle Thumb-2 constants.
2254 (arm_gen_constant): Ditto. Use movw when available.
2255 (arm_function_ok_for_sibcall): Return false for v7m interrupt
2257 (legitimize_pic_address, arm_call_tls_get_addr): Handle Thumb-2.
2258 (thumb_find_work_register, arm_load_pic_register,
2259 legitimize_tls_address, arm_address_cost, load_multiple_sequence,
2260 emit_ldm_seq, emit_stm_seq, arm_select_cc_mode, get_jump_table_size,
2261 print_multi_reg, output_mov_long_double_fpa_from_arm,
2262 output_mov_long_double_arm_from_fpa, output_mov_double_fpa_from_arm,
2263 output_mov_double_fpa_from_arm, output_move_double,
2264 arm_compute_save_reg_mask, arm_compute_save_reg0_reg12_mask,
2265 output_return_instruction, arm_output_function_prologue,
2266 arm_output_epilogue, arm_get_frame_offsets, arm_regno_class,
2267 arm_output_mi_thunk, thumb_set_return_address): Ditto.
2268 (arm_expand_prologue): Handle Thumb-2. Use arm_save_coproc_regs.
2269 (arm_coproc_mem_operand): Allow POST_INC/PRE_DEC.
2270 (arithmetic_instr, shift_op): Use arm_shift_nmem.
2271 (arm_print_operand): Use arm_print_condition. Handle '(', ')', '.',
2273 (arm_final_prescan_insn): Use extract_constrain_insn_cached.
2274 (thumb_expand_prologue): Use thumb_set_frame_pointer.
2275 (arm_file_start): Output directive for unified syntax.
2276 (arm_unwind_emit_set): Handle stack alignment instruction.
2277 * config/arm/lib1funcs.asm: Remove default for __ARM_ARCH__.
2278 Add v6t2, v7, v7a, v7r and v7m.
2279 (RETLDM): Add Thumb-2 code.
2280 (do_it, shift1, do_push, do_pop, COND, THUMB_SYNTAX): New macros.
2281 * config/arm/arm.h (TARGET_CPU_CPP_BUILTINS): Define __thumb2__.
2282 (TARGET_THUMB1, TARGET_32BIT, TARGET_THUMB2, TARGET_DSP_MULTIPLY,
2283 TARGET_INT_SIMD, TARGET_UNIFIED_ASM, ARM_FT_STACKALIGN, IS_STACKALIGN,
2284 THUMB2_TRAMPOLINE_TEMPLATE, TRAMPOLINE_ADJUST_ADDRESS,
2285 ASM_OUTPUT_OPCODE, THUMB2_GO_IF_LEGITIMATE_ADDRESS,
2286 THUMB2_LEGITIMIZE_ADDRESS, CASE_VECTOR_PC_RELATIVE,
2287 CASE_VECTOR_SHORTEN_MODE, ADDR_VEC_ALIGN, ASM_OUTPUT_CASE_END,
2288 ADJUST_INSN_LENGTH): Define.
2289 (TARGET_REALLY_IWMMXT, TARGET_IWMMXT_ABI, CONDITIONAL_REGISTER_USAGE,
2290 STATIC_CHAIN_REGNUM, HARD_REGNO_NREGS, INDEX_REG_CLASS,
2291 BASE_REG_CLASS, MODE_BASE_REG_CLASS, SMALL_REGISTER_CLASSES,
2292 PREFERRED_RELOAD_CLASS, SECONDARY_OUTPUT_RELOAD_CLASS,
2293 SECONDARY_INPUT_RELOAD_CLASS, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P,
2294 TRAMPOLINE_SIZE, INITIALIZE_TRAMPOLINE, HAVE_PRE_INCREMENT,
2295 HAVE_POST_DECREMENT, HAVE_PRE_DECREMENT, HAVE_PRE_MODIFY_DISP,
2296 HAVE_POST_MODIFY_DISP, HAVE_PRE_MODIFY_REG, HAVE_POST_MODIFY_REG,
2297 REGNO_MODE_OK_FOR_BASE_P, LEGITIMATE_CONSTANT_P,
2298 REG_MODE_OK_FOR_BASE_P, REG_OK_FOR_INDEX_P, GO_IF_LEGITIMATE_ADDRESS,
2299 LEGITIMIZE_ADDRESS, THUMB2_LEGITIMIZE_ADDRESS,
2300 GO_IF_MODE_DEPENDENT_ADDRESS, MEMORY_MOVE_COST, BRANCH_COST,
2301 ASM_APP_OFF, ASM_OUTPUT_CASE_LABEL, ARM_DECLARE_FUNCTION_NAME,
2302 FINAL_PRESCAN_INSN, PRINT_OPERAND_PUNCT_VALID_P,
2303 PRINT_OPERAND_ADDRESS): Adjust for Thumb-2.
2304 (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv): New declarations.
2305 * config/arm/arm-cores.def: Add arm1156t2-s, cortex-a8, cortex-r4 and
2307 * config/arm/arm-tune.md: Regenerate.
2308 * config/arm/arm-protos.h: Update prototypes.
2309 * config/arm/vfp.md: Enable patterns for Thumb-2.
2310 (arm_movsi_vfp): Add movw alternative. Use output_move_vfp.
2311 (arm_movdi_vfp, movsf_vfp, movdf_vfp): Use output_move_vfp.
2312 (thumb2_movsi_vfp, thumb2_movdi_vfp, thumb2_movsf_vfp,
2313 thumb2_movdf_vfp, thumb2_movsfcc_vfp, thumb2_movdfcc_vfp): New.
2314 * config/arm/libunwind.S: Add Thumb-2 code.
2315 * config/arm/constraints.md: Update include Thumb-2.
2316 * config/arm/ieee754-sf.S: Add Thumb-2/Unified asm support.
2317 * config/arm/ieee754-df.S: Ditto.
2318 * config/arm/bpabi.S: Ditto.
2319 * config/arm/t-arm (MD_INCLUDES): Add thumb2.md.
2320 * config/arm/predicates.md (low_register_operand,
2321 low_reg_or_int_operand, thumb_16bit_operator): New.
2322 (thumb_cmp_operand, thumb_cmpneg_operand): Rename...
2323 (thumb1_cmp_operand, thumb1_cmpneg_operand): ... to this.
2324 * config/arm/t-arm-elf: Add armv7 multilib.
2325 * config/arm/arm.md: Update patterns for Thumb-2 and Unified asm.
2327 (UNSPEC_STACK_ALIGN, ce_count): New.
2328 (arm_incscc, arm_decscc, arm_umaxsi3, arm_uminsi3,
2329 arm_zero_extendsidi2, arm_zero_extendqidi2): New
2331 * config/arm/fpa.md: Update patterns for Thumb-2 and Unified asm.
2332 (thumb2_movsf_fpa, thumb2_movdf_fpa, thumb2_movxf_fpa,
2333 thumb2_movsfcc_fpa, thumb2_movdfcc_fpa): New insns.
2334 * config/arm/cirrus.md: Update patterns for Thumb-2 and Unified asm.
2335 (cirrus_thumb2_movdi, cirrus_thumb2_movsi_insn,
2336 thumb2_cirrus_movsf_hard_insn, thumb2_cirrus_movdf_hard_insn): New
2338 * doc/extend.texi: Document ARMv7-M interrupt functions.
2339 * doc/invoke.texi: Document Thumb-2 new cores+architectures.
2341 2007-01-03 Jakub Jelinek <jakub@redhat.com>
2343 * unwind-dw2.c (SIGNAL_FRAME_BIT, EXTENDED_CONTEXT_BIT): Define.
2344 (struct _Unwind_Context): Rename args_size to flags, remove
2345 signal_frame field, add a new args_size field and version field.
2346 (_Unwind_IsSignalFrame, _Unwind_SetSignalFrame,
2347 _Unwind_IsExtendedContext): New inline functions.
2348 (_Unwind_GetGR, _Unwind_SetGR, _Unwind_GetGRPtr, _Unwind_SetGRPtr):
2349 Assume by_value array is only present if _Unwind_IsExtendedContext.
2350 (_Unwind_GetIPInfo, execute_cfa_program, uw_frame_state_for): Use
2351 _Unwind_IsSignalFrame.
2352 (__frame_state_for): Initialize context.flags to EXTENDED_CONTEXT_BIT.
2353 (uw_update_context_1): Use _Unwind_SetSignalFrame.
2354 (uw_init_context_1): Initialize context->flags to
2355 EXTENDED_CONTEXT_BIT.
2356 * config/rs6000/linux-unwind.h (frob_update_context): Use
2357 _Unwind_SetSignalFrame.
2359 2007-01-03 Andrew Pinski <andrew_pinski@playstation.sony.com>
2362 * gimplify.c (gimplify_modify_expr_complex_part): Move below
2363 tree_to_gimple_tuple. Call tree_to_gimple_tuple when we need
2366 2007-01-03 Kazu Hirata <kazu@codesourcery.com>
2368 * config/i386/i386.h (NON_STACK_REG_P, REGNO_OK_FOR_SIREG_P,
2369 REGNO_OK_FOR_DIREG_P, REWRITE_ADDRESS, ASM_OPERAND_LETTER,
2370 RET, AT_SP): Remove.
2371 * config/i386/i386.md (*sse_prologue_save_insn): Use return
2374 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
2375 * cfgbuild.c (find_basic_blocks): Likewise.
2376 * cfgrtl.c (rtl_create_basic_block): Likewise.
2377 * function.c (temp_slots_at_level): Likewise.
2378 * reg-stack.c (stack_regs_mentioned): Likewise.
2379 * regclass.c (allocate_reg_info): Likewise.
2380 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
2381 set_bb_for_stmt, move_block_to_fn): Likewise.
2382 * tree-complex.c (tree_lower_complex): Likewise.
2383 * vec.h (VEC_safe_grow_cleared): New.
2385 * cgraphunit.c, tree-ssa-alias.c: Fix comment typos.
2387 2007-01-03 Zdenek Dvorak <dvorakz@suse.cz>
2389 * loop-unswitch.c (unswitch_loop): Pass probabilities to loopify.
2390 * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Pass probabilities
2392 * cfgloopmanip.c (scale_loop_frequencies): Export.
2393 (loopify): Scale the frequencies by prescribed coefficients.
2394 (set_zero_probability): New function.
2395 (duplicate_loop_to_header_edge): Improve updating of frequencies.
2396 (lv_adjust_loop_entry_edge, loop_version): Set probabilities
2397 and frequencies according to arguments.
2398 * tree-ssa-loop-manip.c (tree_unroll_loop): Set probabilities
2400 * cfg.c (scale_bbs_frequencies_int): Allow scaling the frequencies up.
2401 * modulo-sched.c (sms_schedule): Set probabilities for entering
2402 versioned loop correctly.
2403 * tree-vect-transform.c (vect_transform_loop): Ditto.
2404 * cfgloop.h (loopify, loop_version): Declaration changed.
2405 (scale_loop_frequencies): Declared.
2407 2007-01-02 Jan Hubicka <jh@suse.cz>
2409 * cgraph.c: Include tree-flow.h
2410 (cgraph_add_new-function): Handle IPA_SSA mode; execute
2412 * cgraph.h (enum cgraph_state): Add CGRAPH_STATE_IPA_SSA.
2413 * tree-pass.h (pass_all_early_optimizations): Declare.
2414 * cgraphunit.c (cgraph_process_new_functions): Add IPA_SSA; execute
2416 (cgraph_analyze_function): Do early_local_passes.
2417 * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
2418 Do not add referenced vars.
2419 * tree-optimize.c (gate_all_optimizations): Do not execute when not in
2421 (gate_all_early_local_passes): New gate.
2422 (pass_early_local_passes): Use new gate.
2423 (execute_early_local_optimizations): New functions.
2424 (gate_all_early_optimizations): New gate.
2425 (pass_all_early_optimizations): New pass.
2426 (execute_free_datastructures): Free SSA only when initialized.
2427 (gate_init_datastructures): Init only when optimizing.
2428 (tree_lowering_passes): Do early local passes when called late.
2429 * tree-profile.c (do_tree_profiling): Don't profile functions added
2431 (do_early_tree_profiling, pass_early_tree_profile): Kill.
2432 * tree-cfg.c (update_modified_stmts): Do not update when operands are
2434 * passes.c (init_optimizations_passes): Reorder so we go into SSA
2435 during early_local_passes.
2436 * Makefile.in (cgraph.o): Add dependency on tree-flow.h.
2439 2007-01-02 Carlos O'Donell <carlos@codesourcery.com>
2441 * Makefile.in: Update copyright year.
2443 2007-01-02 Carlos O'Donell <carlos@codesourcery.com>
2445 * Makefile.in: Export GCC_EXEC_PREFIX before calling $(RUNTEST)
2446 in $(lang_checks) and check-consistency targets.
2448 2007-01-02 Jan Hubicka <jh@suse.cz>
2450 * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
2451 Do not add referenced vars.
2452 * tree-cfg.c (update_modified_stmts): Do not update when SSA operands
2454 * passes.c (init_optimization_passes): Put mudflap_2 after
2455 free_datastructures.
2457 2007-01-02 Jan Hubicka <jh@suse.cz>
2459 * tree-optimize (execute_fixup_cfg): Set after_inlining flag.
2460 Set NOTHROW flag on call statements proved to be nothrow.
2461 Update statement of local calls so new pure/const functions are
2462 updated. Update_ssa when in ssa form. Mark PHI nodes of nonlocal
2464 (tree_rest_of_compilation): Register hooks and initialize bitmap
2465 early. Do not set after_inlining flag.
2467 2007-01-02 Steve Ellcey <sje@cup.hp.com>
2469 * sbitmap.c (HOST_BITS_PER_LONG_LONG): Change to
2470 HOST_BITS_PER_LONGLONG
2472 2007-01-02 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
2475 * c-typeck.c (store_init_value): Don't emit pedantic overflow
2476 warning for non-static initializers.
2478 2007-01-02 Steven Bosscher <steven@gcc.gnu.org>
2480 * config/alpha/alpha.md, arm/arm.c, darwin.c, frv/frv.md,
2481 m32r/m32r.c, m32r/m32r.c, mn10300/mn10300.md, pa/pa.c,
2482 rs6000/rs6000.c, s390/s390.md, sh/sh.md, sparc/sparc.c:
2483 Always use set_unique_reg_note to add REG_EQUAL notes.
2485 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
2488 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
2490 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
2491 * cfgbuild.c (find_basic_blocks): Likewise.
2492 * cfgrtl.c (rtl_create_basic_block): Likewise.
2493 * function.c (temp_slots_at_level): Likewise.
2494 * reg-stack.c (stack_regs_mentioned): Likewise.
2495 * regclass.c (allocate_reg_info): Likewise.
2496 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
2497 set_bb_for_stmt, move_block_to_fn): Likewise.
2498 * tree-complex.c (tree_lower_complex): Likewise.
2499 * vec.h (VEC_safe_grow_cleared): New.
2501 2007-01-02 Ian Lance Taylor <iant@google.com>
2503 * c-common.c (c_common_truthvalue_conversion): When warning about
2504 using an assignment as a truth value, set TREE_NO_WARNING.
2506 2007-01-02 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
2509 * c.opt (Wold-style-declaration): New.
2510 * doc/invoke.texi (C-only Warning Options): New.
2511 (Wold-style-declaration): Document it.
2512 (Wextra): Enabled by -Wextra.
2513 * c-opts.c (c_common_post_options): Enabled by -Wextra.
2514 * c-decl.c (declspecs_add_scspec): Replace -Wextra with
2515 -Wold-style-declaration.
2517 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
2519 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
2520 * cfgbuild.c (find_basic_blocks): Likewise.
2521 * cfgrtl.c (rtl_create_basic_block): Likewise.
2522 * function.c (temp_slots_at_level): Likewise.
2523 * reg-stack.c (stack_regs_mentioned): Likewise.
2524 * regclass.c (allocate_reg_info): Likewise.
2525 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
2526 set_bb_for_stmt, move_block_to_fn): Likewise.
2527 * tree-complex.c (tree_lower_complex): Likewise.
2528 * vec.h (VEC_safe_grow_cleared): New.
2530 2007-01-02 Douglas Gregor <doug.gregor@gmail.com>
2532 * c-common.c(c_common_nodes_and_builtins): Since variants of
2533 void_type_node get built before it is given a name, we need to
2534 give those variants the name, too.
2535 (complete_array_type): We need to work with the canonical main
2536 type of the array, from which we will build the qualified version.
2537 * params.def (PARAM_VERIFY_CANONICAL_TYPES): New.
2538 * print-tree.c (print_node): Display canonical type information
2540 * stor-layout.c (layout_type): When we don't know the
2541 alignment of a type for which we're building an array, we end up
2542 guessing wrong, so make the type require structural equality.
2543 * tree.c (make_node_stat): When we build a new type, it is its
2545 (build_type_attribute_qual_variant): When building an attribute
2546 variant, its canonical type is the non-attribute variant. However,
2547 if the attributes are target-dependent and they differ, we need to
2548 use structural equality checks for this type.
2549 (build_qualified_type): A qualified type is not equivalent to its
2550 unqualified variant; set the canonical type appropriately.
2551 (build_distinct_type_copy): When building a distinct type from
2552 another type, the new type is its own canonical type.
2553 (build_variant_type_copy): When building a new type variant, we
2554 assume that it is equivalent to the original type.
2555 (build_pointer_type_for_mode): When building a pointer type, also
2556 build a canonical type pointer.
2557 (build_reference_type_for_mode): When building a reference type,
2558 also build a canonical type reference.
2559 (build_index_type): When we can't hash an index type (e.g.,
2560 because its maximum value is negative), the index type requires
2561 structural equality tests.
2562 (build_array_type): Build the canonical form of an array type.
2563 (build_function_type): Function types require structural equality,
2564 because they contain default arguments, attributes, etc.
2565 (build_method_type_directly): Ditto for method types.
2566 (build_offset_type): Build the canonical offset type.
2567 (build_complex_type): Build the canonical vector type.
2568 (make_vector_type): Build the canonical vector type.
2569 * tree.h (TYPE_CANONICAL): New.
2570 (TYPE_STRUCTURAL_EQUALITY_P): New.
2571 (SET_TYPE_STRUCTURAL_EQUALITY): New.
2572 (struct tree_type): Added "canonical" field.
2573 * params.h (VERIFY_CANONICAL_TYPES): New.
2574 * doc/c-tree.texi (TYPE_CANONICAL): Document.
2575 (TYPE_STRUCTURAL_EQUALITY_P): Document.
2576 (SET_TYPE_STRUCTURAL_EQUALITY): Document.
2577 * doc/invoke.texi (verify-canonical-types): Document --param
2578 parameter for verifying canonical types.
2580 2007-01-02 Joseph Myers <joseph@codesourcery.com>
2582 * config.gcc (powerpc-*-eabispe*, powerpc-*-eabisimaltivec*,
2583 powerpc-*-eabisim*, powerpc-*-eabialtivec*, powerpc-*-eabi*,
2584 powerpc-*-rtems*, powerpc-wrs-vxworks, powerpc-wrs-vxworksae,
2585 powerpcle-*-eabisim*, powerpcle-*-eabi*): Add rs6000/e500.h to
2587 * config/rs6000/e500.h: New.
2588 * config/rs6000/eabi.h (TARGET_SPE_ABI, TARGET_SPE, TARGET_E500,
2589 TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE, TARGET_E500_DOUBLE):
2591 * config/rs6000/linuxspe.h (TARGET_SPE_ABI, TARGET_SPE,
2592 TARGET_E500, TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE,
2593 TARGET_E500_DOUBLE): Remove.
2594 * config/rs6000/vxworks.h (TARGET_SPE_ABI, TARGET_SPE,
2595 TARGET_E500, TARGET_ISEL, TARGET_FPRS): Remove.
2596 * config/rs6000/rs6000.h (CHECK_E500_OPTIONS): Define.
2597 * config/rs6000/rs6000.c (rs6000_override_options): Use
2600 2007-01-02 Joseph Myers <joseph@codesourcery.com>
2602 * config/rs6000/rs6000.c (print_operand): Check (TARGET_SPE ||
2603 TARGET_E500_DOUBLE), not TARGET_E500, for %y.
2604 (rs6000_generate_compare, rs6000_emit_sCOND, output_cbranch,
2605 rs6000_emit_cmove): Don't check TARGET_E500.
2606 * config/rs6000/rs6000.md (bunordered, bordered, sunordered,
2607 sordered): Don't check TARGET_E500.
2609 2007-01-01 Eric Christopher <echristo@apple.com>
2611 * config/mips/mips.c (mips_regno_mode_ok_for_base_p): Use
2612 HARD_REGISTER_NUM_P.
2614 2007-01-01 Roger Sayle <roger@eyesopen.com>
2616 * fold-const.c (fold_binary) <EQ_EXPR>: Fold "(X^C1) eq/ne C2" into
2617 "X eq/ne (C1^C2)". Fold "(X^Z) eq/ne (Y^Z)" as "X eq/ne Y" when Z
2618 has no side-effects. Fold "(X^C1) eq/ne (Y^C2)" as "(X^(C1^C2))
2621 2007-01-01 Mike Stump <mrs@apple.com>
2623 * configure.ac: Remove support for building with Apple's gcc-3.1.
2625 2007-01-02 Joseph Myers <joseph@codesourcery.com>
2628 * caller-save.c (add_stored_regs): Only handle SUBREGs if inner
2629 REG is a hard register. Do not modify REG before calling
2631 * rtlanal.c (subreg_get_info): Don't assert size of XMODE is a
2632 multiple of the size of YMODE for certain lowpart cases.
2634 2007-01-01 Andrew Pinski <pinskia@gmail.com>
2637 * gimplify (voidify_wrapper_expr): Update for
2640 2007-01-01 Andreas Schwab <schwab@suse.de>
2643 * config/ia64/ia64.c (ia64_compute_frame_size): Account space for
2644 save of BR0 in extra_spill_size instead of spill_size.
2645 (ia64_expand_prologue): Save BR0 outside of the gr/br/fr spill
2647 (ia64_expand_epilogue): Restore BR0 from its new location.
2649 2007-01-01 Andrew Pinski <pinskia@gmail.com>
2651 * gimplify.c (gimplify_init_constructor <case VECTOR_TYPE>):
2652 Use a temporary variable if the left hand side is not a gimple
2655 2007-01-01 Andrew Pinski <pinskia@gmail.com>
2657 * gimplify.c (gimplify_return_expr): Make the temporary variable
2658 for the return expression, a gimple register variable.
2660 2007-01-01 Jan Hubicka <jh@suse.cz>
2662 * emit-rtl.c (emit_copy_of_insn_after): Do not call copy_insn_1 for
2665 2007-01-01 Mike Stump <mrs@apple.com>
2667 * configure.ac (HAVE_GAS_LITERAL16): Add autoconf check for
2669 * config/darwin.c (machopic_select_rtx_section): Use
2671 (darwin_mergeable_constant_section): Likewise.
2672 * configure: Regenerate.
2673 * config.in: Regenerate.
2675 2007-01-01 Jan Hubicka <jh@suse.cz>
2676 Andrew Pinski <pinskia@gmail.com>
2678 * cgraphunit.c (cgraph_optimize): Call cgraph_add_new_functions
2679 before starting IPA passes.