1 2007-02-05 Bob Wilson <bob.wilson@acm.org>
3 * config/xtensa/xtensa.c (constantpool_mem_p): Skip over SUBREGs.
5 2007-02-05 Richard Guenther <rguenther@suse.de>
7 * tree-vectorizer.h (vectorizable_function): Add argument type
8 argument, change return type.
9 * tree-vect-patterns.c (vect_recog_pow_pattern): Adjust caller.
10 * tree-vect-transform.c (vectorizable_function): Handle extra
11 argument, return vectorized function decl.
12 (build_vectorized_function_call): Remove.
13 (vectorizable_call): Handle calls with result and argument types
14 differing. Handle loop vectorization factor correctly.
15 * targhooks.c (default_builtin_vectorized_function): Adjust for
17 * targhooks.h (default_builtin_vectorized_function): Likewise.
18 * target.h (builtin_vectorized_function): Add argument type
20 * config/i386/i386.c (ix86_builtin_vectorized_function): Handle
21 extra argument, allow vectorizing of lrintf.
22 * doc/tm.texi (TARGET_VECTORIZE_BUILTIN_VECTORIZED_FUNCTION): Adjust
23 documentation of target hook.
25 2007-02-05 Hans-Peter Nilsson <hp@axis.com>
28 * config/cris/cris.md ("*andsi_movu", "*andsi_clear", "*andhi_movu")
29 ("*andhi_clear", andu (casesi+45)): For size-changed operand where
30 memory is allowed, require !side_effects_p, not just !MEM_VOLATILE_P.
32 2007-02-05 Roger Sayle <roger@eyesopen.com>
34 * fold-const.c (fold_unary) <REAL_PART>: Test for availability of
35 BUILT_IN_COS before simplifying REAL_PART(CEXPI)) to COS.
36 <IMAG_PART>: Likewise, check for availability of BUILT_IN_SIN.
37 * builtins.c (fold_builtin_sincos): Check for TARGET_C99_FUNCTIONS
38 before canonicalizing sincos to cexpi.
39 (fold_builtin_cexp): Likewise, for canonicalizing cexp to cexpi.
41 2007-02-05 Roger Sayle <roger@eyesopen.com>
43 * config/alpha/alpha.c (alpha_add_builtins): New Helper function.
44 Set TREE_READONLY and TREE_NOTHROW directly, not via attributes.
45 (alpha_init_builtins): Use alpha_add_builtins to process tables.
47 2007-02-05 Roger Sayle <roger@eyesopen.com>
49 * mips-tfile.c (initialize_init_file): Correct endianness test.
51 2007-02-05 Kazu Hirata <kazu@codesourcery.com>
53 * config/m68k/m68k.md (pushdi-1, pushdi, movsi+1): Don't use
56 2007-02-05 Richard Sandiford <richard@codesourcery.com>
58 * dwarf2out.c (dwarf2out_frame_debug_expr): Record the register
59 saves in a PARALLEL before the register assignments.
61 2007-02-05 Richard Sandiford <richard@codesourcery.com>
63 * doc/tm.texi (DWARF_ALT_FRAME_RETURN_COLUMN): Do not require
64 DWARF_FRAME_RETURN_COLUMN to be a general register.
65 * dwarf2out.c (init_return_column_size): New function, split from...
66 (expand_builtin_init_dwarf_reg_sizes): ...here. Allow both
67 DWARF_FRAME_RETURN_COLUMN and DWARF_ALT_FRAME_RETURN_COLUMN
68 to be nongeneral registers.
69 * config/m68k/m68k.h (DWARF_FRAME_REGNUM): Only map FP and
71 (DWARF_FRAME_REGISTERS, DWARF_FRAME_RETURN_COLUMN): Define.
72 (DWARF_ALT_FRAME_RETURN_COLUMN): Define.
74 2007-02-04 Zdenek Dvorak <dvorakz@suse.cz>
76 * cfgcleanup.c (try_optimize_cfg): Avoid removing ENTRY_BLOCK_PTR.
78 2007-02-04 Zdenek Dvorak <dvorakz@suse.cz>
80 * cfgloopmanip.c (loop_delete_branch_edge): Removed.
81 (remove_path): Use can_remove_branch_p and remove_branch instead
82 of loop_delete_branch_edge.
83 * tree-ssa-loop-manip.c (scale_dominated_blocks_in_loop): New function.
84 (tree_transform_and_unroll_loop): Remove dead branches immediately.
85 Update profile using scale_dominated_blocks_in_loop.
86 * cfghooks.c (can_remove_branch_p, remove_branch): New functions.
87 * cfghooks.h (struct cfg_hooks): Add can_remove_branch_p.
88 (can_remove_branch_p, remove_branch): Declare.
89 * tree-cfg.c (tree_can_remove_branch_p): New function.
90 (tree_cfg_hooks): Add tree_can_remove_branch_p.
91 * cfgrtl.c (rtl_can_remove_branch_p): New function.
92 (rtl_cfg_hooks, cfg_layout_rtl_cfg_hook): Add rtl_can_remove_branch_p.
94 2007-02-05 Jan Hubicka <jh@suse.cz>
97 * ipa-inline.c (cgraph_clone_inlined_nodes): When there are unanalyzed
98 nodes in cgraph, don't remove offline copy of the function.
100 2007-02-04 Jan Hubicka <jh@suse.cz>
102 * tree-sra.c (sra_walk_expr): Add linebreaks. BITFIELD_REFs into
103 vectors might cause maybe_lookup_element_for_expr to be called
104 on non-sra-candidate.
106 2007-02-04 Kazu Hirata <kazu@codesourcery.com>
108 * config/bfin/bfin-modes.def, config/bfin/bfin.c,
109 config/bfin/bfin.md, config/bfin/predicates.md: Follow
110 spelling conventions.
112 2007-02-04 Richard Guenther <rguenther@suse.de>
115 * fold-const.c (try_move_mult_to_index): Make sure to not
116 overflow one dimension of a multi-dimensional array access.
118 2007-02-04 Jan Hubicka <jh@suse.cz>
120 * passes.c (init_optimization_passes): Reindent.
122 2007-02-04 Jan Hubicka <jh@suse.cz>
123 Eric Botcazou <ebotcazou@adacore.com>
125 * tree-optimize.c (has_abnormal_outgoing_edge_p): Move to...
126 (execute_fixup_cfg): Break out the abnormal goto code.
127 * tree-inline.c (has_abnormal_outgoing_edge_p): ...here.
128 (make_nonlocal_label_edges): Move here from execute_fixup_cfg.
129 (optimize_inline_calls): Call make_nonlocal_label_edges.
131 2007-02-04 Jan Hubicka <jh@suse.cz>
133 * tree-ssa-copyrename.c (copy_rename_partition_coalesce): Return
134 true when something was changed.
135 (rename_ssa_copies): When something was changed, do
136 TODO_remove_unused_locals.
137 * tree-ssa-forwprop.c (tree_ssa_forward_propagate_single_use_value):
138 add TODO_remove_unused_locals when instruction was removed.
140 2007-02-04 Jan Hubicka <jh@suse.cz>
142 * ipa-inline.c (try_inline): Improve debug output; work on already
144 (cgraph_decide_inlining_incrementally): Indent; improve debug output;
145 call try_inline for already inlined edges too when flattening;
146 inline also functions that make callee growth but overall unit size
149 2007-02-04 Kazu Hirata <kazu@codesourcery.com>
151 * config/m32c/bitops.md, config/m32c/jump.md,
152 config/m32c/m32c.c, config/m32c/m32c.h, config/m32r/m32r.c,
153 config/m32r/m32r.h, config/m32r/m32r.md,
154 config/m32r/predicates.md, config/m68hc11/larith.asm,
155 config/m68hc11/m68hc11.c, config/m68hc11/m68hc11.h,
156 config/m68k/m68k.h, config/mcore/mcore.md, config/mips/4k.md,
157 config/mips/mips-protos.h, config/mips/mips.c,
158 config/mips/mips.h, config/mips/mips.md, config/mips/mips16.S,
159 config/mn10300/mn10300.h, config/mn10300/predicates.md,
160 config/mt/mt.c, config/mt/mt.h, config/mt/mt.md: Follow
161 spelling conventions.
163 * config/v850/v850.c, config/v850/v850.h, config/v850/v850.md:
164 Follow spelling conventions.
166 2007-02-03 Douglas Gregor <doug.gregor@gmail.com>
168 * c-opts.c (c_common_post_options): If C++0x mode is enabled, don't
169 warn about C++0x compatibility.
171 2007-02-04 Kazu Hirata <kazu@codesourcery.com>
173 * config/h8300/h8300.c, config/h8300/h8300.h,
174 config/h8300/h8300.md: Follow spelling conventions.
176 2007-02-03 Uros Bizjak <ubizjak@gmail.com>
179 * combine.c (try_combine): Do not substitute source operand
180 with constants wider than 2 * HOST_BITS_PER_WIDE_INT.
182 2007-02-03 Jan Hubicka <jh@suse.cz>
184 PR gcov-profile/30650
185 * value-prof.c (stringop_block_profile): Fix handling of size counter;
186 do not divide by zero for never executed counters.
187 (tree_find_values_to_profile): Fix counters.
188 * gcov-ui.h (GCOV_COUNTER_AVERAGE, GCOV_COUNTER_IOR): Fix comments.
190 2007-02-03 Ian Lance Taylor <iant@google.com>
192 * lower-subreg.c (simple_move_operand): New static function,
193 broken out of simple_move. Reject LABEL_REF, SYMBOL_REF, and HIGH
195 (simple_move): Call simple_move_operand.
196 (find_decomposable_subregs): Add special handling of MEMs.
197 (can_decompose_p): Rename from cannot_decompose_p. Reverse
198 meaning of return value. If we see a hard register, test whether
199 it can store a word_mode value. Change all callers.
201 2007-02-03 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
203 * pa.md (addvdi3, addvsi3, subvdi3, subvsi3, negvdi2, negvsi2): New
204 ftrapv insns and expanders.
205 (subdi3): Change define_expand operand 1 to arith11_operand, and
206 operand 2 to reg_or_0_operand. Change constraints of 64-bit insn
207 pattern to handle reg_or_0 operands. Revise 32-bit insn pattern to
208 handle 11-bit constants and reg_or_0 operands in operands 1 and 2,
212 * varasm.c (notice_global_symbol): Treat global objects as weak when
215 2007-02-03 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
217 * emit-rtl.c (dconstpi): Delete.
219 (init_emit_once): Delete dconstpi and init dconstsqrt2.
220 * real.h (dconstpi): Delete.
222 * builtins.c (fold_builtin_cabs): Use dconstsqrt2.
223 (fold_builtin_hypot): Likewise.
225 2007-02-03 Tom Tromey <tromey@redhat.com>
228 * gcc.c (cpp_unique_options): Any of -ggdb3, -gstabs3,
229 -gcoff3, -gxcoff3, -gvms3 implies -dD.
231 2007-02-03 Kazu Hirata <kazu@codesourcery.com>
233 * c-decl.c, config/avr/avr.c, config/avr/avr.h,
234 config/m68k/m68k.c, config/m68k/netbsd-elf.h,
235 config/mn10300/mn10300.c, config/pdp11/pdp11.h,
236 config/rs6000/cell.md, config/rs6000/darwin.h,
237 config/sh/sh.md, config/sh/sh4-300.md, config/spu/spu.c,
238 config/spu/spu.md, cselib.c, expr.c, haifa-sched.c, hwint.h,
239 jump.c, reload.c, sched-deps.c, sched-int.h, tree-inline.c,
240 tree-profile.c, tree-ssa-live.h, tree-vrp.c: Fix comment
241 typos. Follow spelling conventions.
242 * doc/invoke.texi: Follow spelling conventions.
244 2007-02-03 Roger Sayle <roger@eyesopen.com>
246 * simplify-rtx.c (simplify_relational_operation_1): Implement some
247 canonicalization transformations that attempt to simplify integer
248 constant comparisons to become comparisons against zero.
250 2007-02-03 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
252 * builtins.c (fold_builtin_cabs): Fold cabs(x+xi) into
254 * fold-const.c (fold_binary): Fix comment typos. Fold complex
255 (x,0)-(0,y) into (x,-y). Likewise (0,y)-(x,0) into (-x,y).
257 2007-02-02 Mike Stump <mrs@apple.com>
259 * config/darwin9.h (DARWIN_LINKER_GENERATES_ISLANDS): Add.
260 * config/rs6000/rs6000.c (DARWIN_GENERATE_ISLANDS): Add.
261 (output_call): Use DARWIN_GENERATE_ISLANDS to decide when to
262 generate a branch island.
264 2007-02-02 Bob Wilson <bob.wilson@acm.org>
266 * config/xtensa/xtensa.c (smalloffset_mem_p): Use BASE_REG_P.
267 (xtensa_legitimate_address_p): New.
268 (xtensa_legitimize_address): New.
269 (xtensa_output_addr_const_extra): New.
270 * config/xtensa/xtensa.h (REG_OK_STRICT_FLAG): Define.
272 (REG_OK_FOR_BASE_P): Use BASE_REG_P.
273 (GO_IF_LEGITIMATE_ADDRESS): Move code to xtensa_legitimate_address_p.
274 (LEGITIMIZE_ADDRESS): Move code to xtensa_legitimize_address.
275 (OUTPUT_ADDR_CONST_EXTRA): Move code to xtensa_output_addr_const_extra.
276 * config/xtensa/xtensa-protos.h (xtensa_legitimate_address_p): New.
277 (xtensa_legitimize_address): New.
278 (xtensa_output_addr_const_extra): New.
280 2007-02-02 Steve Ellcey <sje@cup.hp.com>
282 * config/ia64/ia64.c (ia64_print_operand): Fix compare strings.
284 2007-02-02 Ian Lance Taylor <iant@google.com>
286 * expmed.c (expand_divmod): Add comment.
288 2007-02-02 Kazu Hirata <kazu@codesourcery.com>
290 * emit-rtl.c (renumber_insns): Remove.
291 * flags.h: Remove the extern for flag_renumber_insns.
292 * rtl.h: Remove the prototype for renumber_insns.
293 * toplev.c (flag_renumber_insns): Remove.
295 2007-02-02 Hui-May Chang <hm.chang@apple.com>
297 Revert for x86 darwin:
298 2005-06-19 Uros Bizjak <uros@kss-loka.si>
300 * config/i386/i386.c (ix86_function_arg_regno_p): Put back the
301 code before the following patch under TARGET_MACHO.
302 (ix86_function_value_regno_p): Likewise.
304 2007-02-02 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
306 * fold-const.c (negate_expr_p, fold_negate_expr): Handle
309 2007-02-02 Maxim Kuvyrkov <mkuvyrkov@ispras.ru>
312 * config/ia64/ia64.c (ia64_speculate_insn): Restrict to memory loads to
313 general or fp registers. Add comments.
314 * config/ia64/ia64.md (reg_pred_prefix): Add comment.
316 2007-02-02 Paolo Bonzini <bonzini@gnu.org>
318 * pointer-set.c (insert_aux): Only return insertion slot.
319 (pointer_set_insert): Adjust.
320 (pointer_set_traverse, struct pointer_map_t, pointer_map_create,
321 pointer_map_destroy, pointer_map_insert, pointer_map_contains,
322 pointer_map_traverse): New.
323 * pointer-set.h (pointer_set_traverse, struct pointer_map_t,
324 pointer_map_create, pointer_map_destroy, pointer_map_insert,
325 pointer_map_contains, pointer_map_traverse): Declare.
327 2007-02-02 Jakub Jelinek <jakub@redhat.com>
330 * builtins.c (fold_builtin_sprintf): Do not attempt to optimize
331 sprintf (str, "%s"). Do not optimize sprintf (str, "nopercent", p++).
333 2007-02-02 Maxim Kuvyrkov <mkuvyrkov@ispras.ru>
335 * sched-int.h (ds_to_dk, dk_to_ds): Declare functions.
337 (struct _dep): New type.
338 (dep_t): New typedef.
339 (DEP_PRO, DEP_CON, DEP_KIND): New access macros.
340 (DEP_STATUS): New access macro. The macro with the same name was
341 renamed to DEP_LINK_STATUS.
342 (dep_init): Declare function
344 (struct _dep_link): New type.
345 (dep_link_t): New typedef.
346 (DEP_LINK_NODE, DEP_LINK_NEXT, DEP_LINK_PREV_NEXTP): New access macros.
347 (DEP_LINK_DEP, DEP_LINK_PRO, DEP_LINK_CON, DEP_LINK_KIND): New macros.
348 (DEP_LINK_STATUS): New macro.
349 (debug_dep_links): New debug function.
351 (struct _deps_list): New type.
352 (deps_list_t): New typedef.
353 (DEPS_LIST_FIRST): New access macro.
354 (FOR_EACH_DEP_LINK): New cycle macro.
355 (create_deps_list, free_deps_list, delete_deps_list): Declare
357 (deps_list_empty_p, debug_deps_list, add_back_dep_to_deps_list): Ditto.
358 (find_link_by_pro_in_deps_list, find_link_by_con_in_deps_list): Ditto.
359 (copy_deps_list_change_con): Ditto.
361 (move_dep_link): Declare function.
363 (struct _dep_node): New type.
364 (dep_node_t): New typedef.
365 (DEP_NODE_BACK, DEP_NODE_DEP, DEP_NODE_FORW): New access macros.
367 (struct haifa_insn_data.back_deps): New field to hold backward
368 dependencies of the insn.
369 (struct haifa_insn_data.depend): Rename to forw_deps. Change its type
371 (struct haifa_insn_data.resolved_deps): Rename to resolved_back_deps.
372 Change its type to deps_list_t.
373 (INSN_BACK_DEPS): New access macro to use instead of LOG_LINKS.
374 (INSN_DEPEND): Rename to INSN_FORW_DEPS.
375 (RESOLVED_DEPS): Rename to INSN_RESOLVED_BACK_DEPS.
377 (INSN_COST): Move to haifa-sched.c. Use insn_cost () instead.
379 (DEP_STATUS): Rename to DEP_LINK_STATUS. Fix typo in the comment.
381 (add_forw_dep, delete_back_forw_dep, insn_cost): Update declaration and
385 * sched-deps.c (CHECK): New macro to (en/dis)able sanity checks.
386 (ds_to_dk, dk_to_ds): New functions.
388 (init_dep_1): New static function.
389 (init_dep): New function.
390 (copy_dep): New static function.
392 (dep_link_consistent_p, attach_dep_link, add_to_deps_list): New static
394 (detach_dep_link): New static function.
395 (move_dep_link): New function.
397 (dep_links_consistent_p, dump_dep_links): New static functions.
398 (debug_dep_links): New debugging function.
400 (deps_obstack, dl_obstack, dn_obstack): New static variables.
402 (alloc_deps_list, init_deps_list): New static functions.
403 (create_deps_list): New function.
404 (clear_deps_list): New static function.
405 (free_deps_list, delete_deps_list, deps_list_empty_p): New functions.
406 (deps_list_consistent_p, dump_deps_list): New static functions.
407 (debug_deps_list): New function.
408 (add_back_dep_to_deps_list, find_link_by_pro_in_deps_list): New
410 (find_link_by_con_in_deps_list, copy_deps_list_change_con): Ditto.
412 (maybe_add_or_update_back_dep_1, add_or_update_back_dep_1): Update to
413 use new scheduler dependencies lists.
414 (add_back_dep, delete_all_dependences, fixup_sched_groups): Ditto.
415 (sched_analyze): Ditto. Initialize dependencies lists.
416 (add_forw_dep, compute_forward_dependences): Update to use new
417 scheduler dependencies lists.
419 (init_dependency_caches): Init deps_obstack.
420 (free_dependency_caches): Free deps_obstack.
422 (adjust_add_sorted_back_dep, adjust_back_add_forw_dep): Update to use
423 new scheduler dependencies lists.
424 (delete_forw_dep, add_or_update_back_forw_dep): Ditto.
425 (add_back_forw_dep, delete_back_forw_dep): Ditto.
427 * sched-rgn.c (set_spec_fed, find_conditional_protection, is_pfree):
428 Update to use new scheduler dependencies lists.
429 (is_conditionally_protected, is_prisky, add_branch_dependences): Ditto.
430 (debug_dependencies): Ditto.
431 (schedule_region): Update comments.
433 * sched-ebb.c (earliest_block_with_similiar_load): Update to use new
434 scheduler dependencies lists.
435 (schedule_ebb): Update comments.
437 * rtl.def (DEPS_LIST): Remove.
439 * lists.c (unused_deps_list): Remove.
440 (free_list): Update assertions.
442 (alloc_DEPS_LIST, free_DEPS_LIST_list, free_DEPS_LIST_node): Remove.
443 (remove_free_DEPS_LIST_elem, copy_DEPS_LIST_list): Ditto.
445 * rtl.h (free_DEPS_LIST_list, alloc_DEPS_LIST): Remove declarations.
446 (remove_free_DEPS_LIST_elem, copy_DEPS_LIST_list): Ditto.
448 * haifa-sched.c (comments): Update.
449 (insn_cost1): Remove. Inline the code into insn_cost ().
450 (insn_cost): Update to use new scheduler dependencies lists. Move
451 processing of the dependency cost to dep_cost ().
452 (dep_cost): New function. Use it instead of insn_cost () when
453 evaluating cost of the dependency. Use compatible interface to
454 interact with the target.
455 (priority): Update to use new scheduler dependencies lists.
456 (rank_for_schedule): Ditto. Optimize heuristic that prefers the insn
457 with greater number of insns that depend on the insn.
458 (schedule_insn): Update to use new scheduler dependencies lists. Add
459 code to free backward dependencies lists. Inline and optimize code
460 from resolve_dep () - see PR28071.
461 (ok_for_early_queue_removal): Update to use new scheduler dependencies
462 lists. Update call to targetm.sched.is_costly_dependence hook.
464 (fix_inter_tick, try_ready, fix_tick_ready): Update to use new
465 scheduler dependencies lists.
467 (resolve_dep): Remove. Move the logic to schedule_insn ().
468 (init_h_i_d): Initialize dependencies lists.
470 (process_insn_depend_be_in_spec): Rename to
471 process_insn_forw_deps_be_in_spec. Update to use new scheduler
473 (add_to_speculative_block, create_check_block_twin, fix_recovery_deps):
474 Update to use new scheduler dependencies lists.
475 (clear_priorities, calc_priorities, add_jump_dependencies): Ditto.
477 * ddg.c (create_ddg_dependence, create_ddg_dep_no_link): Update to use
478 new scheduler dependencies lists.
479 (build_intra_loop_deps): Ditto.
481 * target.h (struct _dep): Declare to use in
482 gcc_target.sched.is_costly_dependence.
483 (struct gcc_target.sched.adjust_cost): Fix typo.
484 (struct gcc_target.sched.is_costly_dependence): Change signature to use
485 single dep_t parameter instead of an equivalent triad.
486 (struct gcc_target.sched.adjust_cost_2): Remove.
488 * target-def.h (TARGET_SCHED_ADJUST_COST_2): Remove.
490 * reg-notes.def (DEP_TRUE, DEP_OUTPUT, DEP_ANTI): Update comments.
492 * doc/tm.texi (TARGET_SCHED_IS_COSTLY_DEPENDENCE): Update
494 (TARGET_SCHED_ADJUST_COST_2): Remove documentation.
496 * doc/rtl.texi (LOG_LINKS): Remove part about instruction scheduler.
497 (REG_DEP_TRUE): Document.
499 * config/ia64/ia64.c (ia64_adjust_cost_2): Rename to ia64_adjust_cost.
500 Change signature to correspond to the targetm.sched.adjust_cost hook.
501 Update use in TARGET_SCHED_ADJUST_COST_2.
502 (TARGET_SCHED_ADJUST_COST_2): Rename to TARGET_SCHED_ADJUST_COST.
503 (ia64_dependencies_evaluation_hook, ia64_dfa_new_cycle): Update to use
504 new scheduler dependencies lists.
505 (ia64_gen_check): Ditto.
507 * config/mips/mips.c (vr4130_swap_insns_p): Update to use new scheduler
510 * config/rs6000/rs6000.c (rs6000_is_costly_dependence): Change
511 signature to correspond to the targetm.sched.is_costly_dependence hook.
512 (is_costly_group): Update to use new scheduler dependencies lists.
514 * config/spu/spu.c (spu_sched_adjust_cost): Use insn_cost () function
515 instead of INSN_COST () macro.
517 2007-02-01 Ian Lance Taylor <iant@google.com>
519 * lower-subreg.c (resolve_clobber): Handle a subreg of a concatn.
521 2007-02-01 Guy Martin <gmsoft@gentoo.org>
523 * pa.md (tp_load): Correct mfctl instruction syntax.
525 2007-02-01 Geoffrey Keating <geoffk@apple.com>
527 * config/rs6000/rs6000.c (rs6000_stack_info): Correct
528 altivec_padding_size calculation on AIX. Improve comment, add
529 assert to verify that it's right.
531 * config/rs6000/darwin.h (MD_UNWIND_SUPPORT): Don't define for
534 2007-2-01 Seongbae Park <seongbae.park@gmail.com>
537 * global.c (global_alloc): Add mising initialization of
540 2007-02-01 Roger Sayle <roger@eyesopen.com>
542 * alias.c (init_alias_analysis): Correct whitespace.
543 * bb-reorder.c (fix_edges_for_rarely_executed_code,
544 partition_hot_cold_basic_blocks): Likewise.
545 * builtins.c (expand_builtin_printf, expand_builtin_fprintf,
546 expand_builtin_sprintf, fold_builtin_carg, fold_builtin_sprintf,
547 maybe_emit_sprintf_chk_warning, fold_builtin_sprintf_chk,
548 fold_builtin_snprintf_chk, fold_builtin_printf,
549 fold_builtin_fprintf, do_mpfr_ckconv, do_mpfr_arg1, do_mpfr_arg2,
550 do_mpfr_arg3, do_mpfr_sincos): Likewise.
551 * cfgcleanup.c (cleanup_cfg): Likewise.
552 * cfgexpand.c (tree_expand_cfg): Likewise.
553 * fold-const.c (fold_binary) <RDIV_EXPR>: Likewise.
554 * function.c (get_next_funcdef_no): Likewise.
555 * gengtype.c (main): Likewise.
556 * genmodes.c (main): Likewise.
557 * gcse.c (bypass_conditional_jumps, print_ldst_list): Likewise.
558 * haifa-sched.c (schedule_block, extend_h_i_d): Likewise.
559 * ifcvt.c (noce_emit_move_insn): Likewise.
560 * modulo-sched.c (generate_prolog_epilog, sms_schedule_by_order):
562 * stor-layout.c (get_best_mode): Likewise.
563 * tree-ssa-loop-niter.c (get_val_for): Likewise.
564 * tree-ssa-structalias.c (get_varinfo, get_varinfo_fc,
565 scc_visit, do_ds_constraint, do_complex_constraint, label_visit,
566 perform_var_substitution, solve_graph): Likewise.
567 * tree-vrp.c (vrp_finalize): Likewise.
569 2007-02-01 Ian Lance Taylor <iant@google.com>
571 * lower-subreg.c (simplify_gen_subreg_concatn): If we ask for the
572 high part of a paradoxical subreg, return a constant zero.
574 2007-02-01 Zdenek Dvorak <dvorakz@suse.cz>
576 * toplev.c (lang_dependent_init): Call init_set_costs.
577 * loop-init.c (loop_optimizer_init): Do not call init_set_costs.
579 2007-02-01 Richard Guenther <rguenther@suse.de>
582 * fold-const.c (fold_negate_expr): Allow negating a
583 constant if overflow does not change.
585 2007-02-01 Ralf Wildenhues <Ralf.Wildenhues@gmx.de>
587 * doc/c-tree.texi (Expression trees): Improve markup.
588 * doc/tm.texi (Register Classes, Addressing Modes)
589 (Floating Point): Fix spacing after abbreviations. Fix some
592 2007-02-01 Ben Elliston <bje@au.ibm.com>
594 * doc/invoke.texi: Replace "bugfix" with "bug fix" throughout.
595 * doc/contrib.texi: Likewise.
596 * doc/install.texi: Likewise.
598 2007-01-31 Richard Henderson <rth@redhat.com>
599 Ian Lance Taylor <iant@google.com>
601 * lower-subreg.c: New file.
602 * rtl.def (CONCATN): Define.
603 * passes.c (init_optimization_passes): Add pass_lower_subreg and
605 * emit-rtl.c (update_reg_offset): New static function, broken out
606 of gen_rtx_REG_offset.
607 (gen_rtx_REG_offset): Call update_reg_offset.
608 (gen_reg_rtx_offset): New function.
609 * regclass.c: Revert patch of 2006-03-05, restoring
611 (clear_reg_info_regno): New function.
612 * dwarf2out.c (concatn_loc_descriptor): New static function.
613 (loc_descriptor): Handle CONCATN.
614 * common.opt (fsplit_wide_types): New option.
615 * opts.c (decode_options): Set flag_split_wide_types when
617 * timevar.def (TV_LOWER_SUBREG): Define.
618 * rtl.h (gen_reg_rtx_offset): Declare.
619 (reg_scan_update): Declare.
620 * regs.h (clear_reg_info_regno): Declare.
621 * tree-pass.h (pass_lower_subreg): Declare.
622 (pass_lower_subreg2): Declare.
623 * doc/invoke.texi (Option Summary): List -fno-split-wide-types.
624 (Optimize Options): Add -fsplit-wide-types to -O1 list. Document
626 * doc/rtl.texi (Regs and Memory): Document concat and concatn.
627 * Makefile.in (OBJS-common): Add lower-subreg.o.
628 (lower-subreg.o): New target.
630 2007-01-31 Kazu Hirata <kazu@codesourcery.com>
632 * config/sh/sh.h (HAVE_SECONDARY_RELOADS): Remove.
634 2007-01-31 Anatoly Sokolov <aesok@post.ru>
637 * config/avr/avr.c (DWARF2_ADDR_SIZE): Define.
639 2007-01-31 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
642 * builtins.c (fold_builtin_sqrt): Use MPFR for constant args.
644 2007-01-31 Zdenek Dvorak <dvorakz@suse.cz>
646 * cfgloop.h: Include vec-prim.h.
647 (enum li_flags): Remove LI_ONLY_OLD.
648 (loop_iterator): Changed.
649 (fel_next, fel_init): Iterate over loop tree.
650 (FOR_EACH_LOOP_BREAK): New macro.
651 * loop-unswitch.c (unswitch_loops): Do not pass LI_ONLY_OLD to
653 * tree-ssa-loop-unswitch.c (tree_ssa_unswitch_loops): Ditto.
654 * modulo-sched.c (sms_schedule): Ditto.
655 * tree-vectorizer.c (vectorize_loops): Ditto.
656 * doc/loop.texi: Update information on loop numbering and behavior of
657 FOR_EACH_LOOP wrto new loops.
658 * tree-scalar-evolution.c (compute_overall_effect_of_inner_loop,
659 add_to_evolution_1): Test nestedness of loops instead of comparing
661 * tree-chrec.c (chrec_fold_plus_poly_poly,
662 chrec_fold_multiply_poly_poly, chrec_evaluate,
663 hide_evolution_in_other_loops_than_loop, chrec_component_in_loop_num,
664 reset_evolution_in_loop): Ditto.
665 * Makefile.in (CFGLOOP_H): Add vecprim.h dependency.
667 2007-01-31 Dirk Mueller <dmueller@suse.de>
669 * c-common.c (warn_about_parentheses): Separate warning about
670 un-parenthized sequence of comparison operators from the one
671 which is supposed to warn about x <= y <= z.
673 2007-01-31 Uros Bizjak <ubizjak@gmail.com>
675 * optabs.h (enum optab_index): Add new OTI_isinf.
676 (isinf_optab): Define corresponding macro.
677 * optabs.c (init_optabs): Initialize isinf_optab.
678 * genopinit.c (optabs): Implement isinf_optab using isinf?f2
680 * builtins.c (mathfn_built_in): Handle BUILT_IN_ISINF{,F,L}.
681 (expand_builtin_interclass_mathfn): Expand BUILT_IN_ISINF{,F,L}
683 (expand_builtin): Expand BUILT_IN_ISINF{,F,L} using
684 expand_builtin_interclass_mathfn.
685 * reg_stack.c (subst_stack_regs_pat): Handle UNSPEC_FXAM.
686 * config/i386/i386.md (UNSPEC_FXAM): New constant.
687 (fxam<mode>2_i387): New insn pattern.
688 (isinf<mode>2) New expander to implement isinf, isinff and isinfl
689 built-in functions as x87 inline asm.
691 2007-01-31 Kazu Hirata <kazu@codesourcery.com>
693 * gcc/config/arm/unwind-arm.h (_sleb128_t, _uleb128_t): New.
695 2007-01-30 Eric Christopher <echristo@apple.com>
697 * config.gcc: Add geode.
699 2007-01-31 Kazu Hirata <kazu@codesourcery.com>
701 * cgraphunit.c, config/arm/arm.c, config/m68k/m68k.c,
702 ipa-inline.c, tree-profile.c, tree-ssa-live.c,
703 tree-ssa-math-opts.c, tree-ssanames.c, tree-vect-analyze.c,
704 value-prof.c: Fix comment typos.
706 2007-01-30 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
709 * doc/invoke.texi (Wpointer-arith): Document warning.
711 2007-01-30 Janis Johnson
713 * doc/extend.texi (Decimal Floating Types): Remove decfloat.h from
714 the list of discrepancies from the draft TR.
716 2007-01-30 Dirk Mueller <dmueller@suse.de>
719 * doc/invoke.texi (-Wreturn-type): Update description to
722 2007-01-30 Richard Sandiford <richard@codesourcery.com>
724 * cfgrtl.c (try_redirect_by_replacing_jump): Check only_sets_cc0_p.
726 2007-01-30 Uros Bizjak <ubizjak@gmail.com>
728 * builtins.c (expand_builtin_int_interclass_roundingfn): New function
729 to handle optabs that operate on floating point input argument and
730 output to integer output.
731 (expand_builtin_mathfn) [BUILT_IN_ILOGB]: Move from here ...
732 (expand_builtin_interclass_mathfn) [BUILT_IN_ILOGB]: ... to here.
733 (expand_builtin): Expand BUILT_IN_ILOGB{,F,L} using
734 expand_builtin_interclass_mathfn ().
735 * config/i386/i386.md (fxtractxf3_i387): Rename from *fxtractxf3_i387.
737 (ilogbxf2, ilogb<mode>2): New expanders to implement ilogb, ilogbf and
738 ilogbl built-in functions as x87 intrinsics.
740 2007-01-30 Richard Guenther <rguenther@suse.de>
743 * dwarf2out.c (reference_to_unused): Query varpool if the
746 2007-01-30 Richard Guenther <rguenther@suse.de>
749 * passes.c (execute_one_pass): Reset in_gimple_form to not
750 confuse non-unit-at-a-time mode.
752 2007-01-29 Roger Sayle <roger@eyesopen.com>
753 Richard Guenther <rguenther@suse.de>
755 * fold-const.c (round_up): Make HIGH an unsigned HOST_WIDE_INT to
756 avoid undefined behaviour on overflow. Use force_fit_type_double
757 to construct the constant with the specified TREE_OVERFLOW.
759 2007-01-29 Janis Johnson <janis187@us.ibm.com>
761 * config/dfp-bit.c: Add parameterized support for fp exceptions.
762 * config/dfp-bit.h: Ditto.
764 2007-01-29 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
766 * c-decl.c (pop_scope): Replace warnings with call to
767 warn_for_unused_label.
768 * c-common.h (warn_for_unused_label): Declare.
769 * c-common.c (warn_for_unused_label): Define.
771 2007-01-29 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
773 * tree-optimize.c (update_inlined_to_pointers): Delete unused
776 2007-01-29 Janis Johnson <janis187@us.ibm.com>
778 * Makefile.in (USER_H): Remove decfloat.h.
779 * ginclude/decfloat.h: Delete, moving contents to ...
780 * ginclude/float.h: Add support for decimal floating point,
781 guarded by __STDC_WANT_DEC_FP__.
783 2007-01-29 Mike Stump <mrs@apple.com>
785 * doc/gccint.texi (Top): Rename Loop Representation to Loop
786 Analysis and Representation to resolve case insensitive conflict.
787 * doc/loop.texi (Loop Analysis and Representation): Likewise.
789 2007-01-28 Daniel Berlin <dberlin@dberlin.org>
791 * tree.h (struct tree_memory_tag): Add aliases member.
792 (MTAG_ALIASES): New macro.
793 * tree-ssa-alias.c (alias_bitmap_obstack): New variable.
794 (add_may_alias): Remove pointer-set. Update for may_aliases being
796 (mark_aliases_call_clobbered): Update for may_aliases being a
798 (compute_tag_properties): Ditto.
799 (create_partition_for): Ditto.
800 (compute_memory_partitions): Ditto.
801 (dump_may_aliases_for): Ditto.
802 (is_aliased_with): Ditto.
803 (add_may_alias_for_new_tag): Ditto.
804 (rewrite_alias_set_for): Rewrite for may_aliases being a bitmap.
805 (compute_is_aliased): New function.
806 (compute_may_aliases): Call compute_is_aliased).
807 (init_alias_info): Initialize alias_bitmap_obstack.
808 (union_alias_set_into): New function.
809 (compute_flow_sensitive_aliasing): Use union_aliases_into.
810 (have_common_aliases_p): Rewrite to take two bitmaps and use
812 (compute_flow_insensitive_aliasing): Stop using pointer-sets.
814 (finalize_ref_all_pointers): Update for add_may_alias changes.
815 (new_type_alias): Ditto.
816 * tree-flow-inline.h (may_aliases): Return a bitmap.
817 * tree-dfa.c (dump_variable): Check for MTAG_P'ness.
818 * tree-ssa.c (verify_flow_insensitive_alias_info): Update for
819 may_aliases being a bitmap.
820 * tree-flow.h (struct var_ann_d): Remove may_aliases member.
821 may_aliases now returns a bitmap.
822 * tree-ssa-structalias.c (merge_smts_into): Update for may_aliases
824 * tree-ssa-operands.c (add_virtual_operand): Update for
825 may_aliases being a bitmap.
827 2007-01-29 Daniel Berlin <dberlin@dberlin.org>
829 PR tree-optimization/30630
830 * tree-ssa-structalias.c (do_complex_constraint): Mark correct
833 2007-01-29 Simon Martin <simartin@users.sourceforge.net>
836 * gimplify.c (gimplify_target_expr): Make sure that the TARGET_EXPR is
837 expanded only once even if an error occurs.
839 2007-01-29 Ben Elliston <bje@au.ibm.com>
841 * gcov-io.h (__gcov_indirect_call_profiler): Declare.
842 (__gcov_average_profiler): Likewise.
843 (__gcov_ior_profiler): Likewise.
844 (__gcov_merge_ior): Likewise.
846 2007-01-28 Jan Hubicka <jh@suse.cz>
848 * builtins.c (expand_builtin_memset): Fix typo in my last patch.
849 * value-prof.c (stringop_block_profile): Likewise.
851 2007-01-28 Jan Hubicka <jh@suse.cz>
853 * expr.c (emit_block_move_via_movmem, emit_block_move_via_libcall): Add
854 variant handling histograms; add wrapper.
855 (clear_storage_via_libcall): Export.
856 (emit_block_move_hints): Break out from ...; add histograms.
857 (emit_block_move): ... this one.
858 (clear_storage_hints): Break out from ...; add histograms.
859 (clear_storage): ... this one.
860 (set_storage_via_memset): Handle histogram.
861 * expr.h (emit_block_move_via_libcall, emit_block_move_hints): Declare.
862 (clear_storage_hints, clear_storage_via_libcall): Declare.
863 (set_storage_via_setmem): Update prototype.
864 * doc/md.texi (movmem, setmem): Document new arguments.
866 * value-prof.c (dump_histogram_value, tree_find_values_to_profile): Add
868 (stringop_block_profile): New global function.
869 (tree_stringops_values_to_profile): Profile block size and alignment.
870 * value-prof.h (enum hist_type): add HIST_TYPE_AVERAGE and
872 (struct profile_hooks): Add gen_average_profiler and gen_ior_profiler.
873 (stringop_block_profile): Declare.
874 * builtins.c: Include value-prof.h.
875 (expand_builtin_memcpy, expand_builtin_memset): Pass block profile.
876 * gcov-ui.h (GCOV_COUNTER_NAMES): Add new counter.
877 (GCOV_COUNTER_AVERAGE, GCOV_COUNTER_IOR): New constants.
878 (GCOV_COUNTERS, GCOV_LAST_VALUE_COUNTER): Update.
879 * profile.c (instrument_values): Add new counters.
880 * cfgexpand.c (expand_gimple_basic_block): Propagate histograms to
882 * tree-profile.c (tree_average_profiler_fn, tree_ior_profiler_fn): New.
883 (tree_init_edge_profiler): Build new profilers.
884 (tree_gen_average_profiler, tree_gen_ior_profiler): New.
885 (pass_tree_profile): Add dump.
886 (tree_profile_hooks): Update.
887 * Makefile.in (LIBGCOV): Add new constants.
888 * libgcov.c (__gcov_merge_ior, __gcov_average_profiler,
889 __gcov_ior_profiler): New.
890 * i386.md (movmem/setmem expanders): Add new optional arguments.
892 2007-01-28 David Edelsohn <edelsohn@gnu.org>
894 * doc/md.texi (Standard Pattern Names): Document blockage pattern.
896 2007-01-28 Zdenek Dvorak <dvorakz@suse.cz>
898 * tree-ssa-loop-unswitch.c: Include tree-inline.h.
899 (tree_unswitch_single_loop): Pass eni_size_weights to
901 * tree-ssa-loop-manip.c: Include tree-inline.h.
902 (can_unroll_loop_p): Pass eni_size_weights to
904 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p):
905 Pass eni_size_weights to estimate_num_insns.
906 * tree.h (init_inline_once): Export.
907 * toplev.c (backend_init): Call init_inline_once.
908 * cgraphunit.c (cgraph_process_new_functions,
909 cgraph_analyze_function): Pass eni_inlining_weights to
911 * ipa-inline.c (compute_inline_parameters): Ditto.
912 * tree-ssa-loop-ivcanon.c (tree_num_loop_insns): Pass weights
913 to estimate_num_insns.
914 (try_unroll_loop_completely): Pass eni_size_weights to
916 * tree-eh.c (decide_copy_try_finally): Pass eni_size_weights
917 ot estimate_num_insns.
918 * tree-ssa-loop-prefetch.c: Include tree-inline.h.
919 (loop_prefetch_arrays): Pass eni_time_weights to tree_num_loop_insns.
920 * tree-inline.c (eni_inlining_weights, eni_size_weights,
921 eni_time_weights): New variables.
922 (init_inline_once): Initialize them.
923 (struct eni_data): Mew.
924 (estimate_num_insns_1, estimate_num_insns): Use weights.
925 * tree-inline.h (struct eni_weights_d): New.
926 (eni_inlining_weights, eni_size_weights, eni_time_weights): Declare.
927 (estimate_num_insns): Declaration changed.
928 * cfgloop.h (tree_num_loop_insns): Declaration changed.
929 * Makefile.in (tree-ssa-loop-unswitch.o, tree-ssa-loop-prefetch.o,
930 tree-ssa-loop-manip.o): Add TREE_INLINE_H dependency.
932 2007-01-28 Zdenek Dvorak <dvorakz@suse.cz>
934 * tree-data-ref.c (conflict_fn): Assert that the number of affine
935 relations in the conflict function is valid.
937 2007-01-27 Ian Lance Taylor <iant@google.com>
939 * common.opt: Add fstrict-overflow.
940 * opts.c (decode_options): Set flag_strict_overflow if -O2.
941 * flags.h (TYPE_OVERFLOW_WRAPS): Define.
942 (TYPE_OVERFLOW_UNDEFINED): Define.
943 (TYPE_OVERFLOW_TRAPS): Define. This replaces TYPE_TRAP_SIGNED.
945 * tree.h (TYPE_TRAP_SIGNED): Don't define.
946 * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_UNDEFINED.
947 (fold_negate_expr): Likewise.
948 (make_range): Likewise.
949 (extract_muldiv_1): Likewise.
950 (maybe_canonicalize_comparison): Likewise.
951 (fold_comparison): Likewise.
952 (fold_binary): Likewise.
953 (tree_expr_nonnegative_p): Likewise.
954 (tree_expr_nonzero_p): Likewise.
955 * tree-vrp.c (compare_values): Likewise.
956 (extract_range_from_binary_expr): Likewise.
957 (extract_range_from_unary_expr): Likewise.
958 * tree-ssa-loop-niter.c (infer_loop_bounds_from_signedness):
960 (nowrap_type_p): Likewise.
961 * tree-scalar-evolution.c (simple_iv): Likewise.
962 * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_WRAPS.
963 (build_range_check): Likewise.
964 (extract_muldiv_1): Likewise.
965 (fold_comparison): Likewise.
966 * tree-vrp.c (vrp_int_const_binop): Likewise.
967 (extract_range_from_unary_expr): Likewise.
968 * convert.c (convert_to_integer): Likewise.
969 * fold-const.c (fold_negate_expr): Use TYPE_OVERFLOW_TRAPS.
970 (fold_comparison): Likewise.
971 (fold_binary): Likewise.
972 * optabs.c (optab_for_tree_code): Likewise.
973 * tree-vectorizer.c (vect_is_simple_reduction): Likewise.
974 * simplify-rtx.c (simplify_const_relational_operation): Check
975 flag_strict_overflow and flag_trapv.
976 (simplify_const_relational_operation): Likewise.
977 * doc/invoke.texi (Option Summary): Mention -fstrict-overflow.
978 (Optimize Options): Add -fstrict-overflow to -O2 list. Document
981 2007-01-27 Roger Sayle <roger@eyesopen.com>
983 * tree.c (tree_fold_gcd): Delete.
984 * tree.h (tree_fold_gcd): Remove prototype.
985 * tree-data-ref.c (tree_fold_divides_p): Don't use tree_fold_gcd to
986 test whether one constant integer is a multiple of another. Instead
987 call int_const_binop with TRUNC_MOD_EXPR and test for a zero result.
988 * fold-const.c (multiple_of_p): We've determined both TOP and
989 BOTTOM are integer constants so we can call int_const_binop directly
990 instead of the more generic const_binop.
992 2007-01-27 Roger Sayle <roger@eyesopen.com>
994 * fold-const.c (size_binop): In the fast-paths for X+0, 0+X, X-0 and
995 1*X check that the constant hasn't overflowed, to preserve the
997 (round_up): Provide an efficient implementation when rouding-up an
998 INTEGER_CST to a power-of-two.
1000 2007-01-28 Ralf Wildenhues <Ralf.Wildenhues@gmx.de>
1002 * doc/sourcebuild.texi: Add comma for clarity.
1003 * doc/extend.texi: Fix some typos.
1004 * doc/passes.texi: Likewise.
1005 * doc/cppinternals.texi: Likewise.
1006 * doc/c-tree.texi: Likewise.
1007 * doc/tree-ssa.texi: Likewise.
1008 * doc/install.texi: Likewise.
1010 2007-01-27 Jan Hubicka <jh@suse.cz>
1012 * tree-sra.c (sra_walk_function): Don't rely on aliases being build.
1013 (pass_sra): Do not require alias information.
1014 * passes.c (init_optimization_passes): Add SRA
1016 2007-01-27 Steven Bosscher <steven@gcc.gnu.org>
1018 * tracer.c (rest_of_handle_tracer): We already cleaned
1019 up the CFG in tracer() so don't do it here again.
1020 * cfgcleanup.c (rest_of_handle_jump2): Don't repeat
1021 cleanup_cfg here, either. And don't call renumber_insns.
1023 * cfgrtl.c (rtl_verify_flow_info_1): Don't verify that BB_END
1024 and BB_HEAD are in the insn stream here. Instead make sure
1025 that BB_INSN is valid on all insns. Also, do check here that
1026 there are no pending branch predictions...
1027 (rtl_verify_flow_info): ...instead of doing it here. Checks
1028 for BB_END and BB_HEAD moved from rtl_verify_flow_info_1 to
1031 2007-01-26 Roger Sayle <roger@eyesopen.com>
1033 * config/i386/i386.c (ix86_swap_binary_operands_p): New helper
1034 function to simplify/factorize operand order canonicalization.
1035 (ix86_fixup_binary_operands): Reorganize using the above function.
1036 (ix86_binary_operator_ok): Likewise.
1038 2007-01-27 Jakub Jelinek <jakub@redhat.com>
1040 * genattrtab.c (struct attr_value_list, insn_code_values): Move to
1041 file scope from optimize_attrs.
1042 (simplify_test_exp): If insn_code_values is not NULL, use it to speed
1044 (optimize_attrs): Clear insn_code_values after freeing it.
1046 2007-01-26 Zdenek Dvorak <dvorakz@suse.cz>
1048 * tree-ssa-address.c (create_mem_ref): Remove ", bsi" from
1049 a parts.base assignment.
1051 2007-01-26 Zdenek Dvorak <dvorakz@suse.cz>
1053 * tree-data-ref.c (dump_subscript): Use dump_conflict_function.
1054 (compute_subscript_distance, initialize_data_dependence_relation,
1055 finalize_ddr_dependent, analyze_ziv_subscript,
1056 analyze_siv_subscript_cst_affine,
1057 compute_overlap_steps_for_affine_univar,
1058 compute_overlap_steps_for_affine_1_2, analyze_subscript_affine_affine,
1059 analyze_siv_subscript, analyze_miv_subscript,
1060 analyze_overlapping_iterations, subscript_dependence_tester_1,
1061 compute_self_dependence, free_dependence_relation): Work
1062 with affine_fn instead of chrecs.
1063 (dump_affine_function, dump_conflict_function, affine_function_equal_p,
1064 common_affine_function, affine_function_base,
1065 affine_function_constant_p, affine_fn_op, affine_fn_plus,
1066 affine_fn_minus, affine_fn_free, conflict_fn_not_known,
1067 conflict_fn_no_dependence, free_conflict_function, free_subscripts,
1068 conflict_fn, affine_fn_cst, affine_fn_univar): New functions.
1069 (all_chrecs_equal_p): Removed.
1070 * tree-data-ref.h (affine_fn, conflict_function): New types.
1071 (struct subscript): Change type of conflicting_iterations_in_a
1072 and conflicting_iterations_in_b.
1074 2007-01-26 Steve Ellcey <sje@cup.hp.com>
1077 * config/pa/pa.h (TARGET_HPUX_11): New.
1078 * config/pa/pa-hpux11.h (TARGET_HPUX_11): New.
1079 * config/pa/pa.c (pa_init_builtins): Use TARGET_HPUX_11.
1081 2007-01-26 Daniel Berlin <dberlin@dberlin.org>
1082 Richard Guenther <rguenther@suse.de>
1084 * tree-ssa-structalias.c (solve_graph): Handle case
1085 we merged the variable to another.
1087 2007-01-25 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1089 * builtins.c (fold_builtin_1): Treat ccos and ccosh as 'even'
1092 * fold-const.c (negate_mathfn_p): Treat casin, casinh, catan,
1093 catanh, cproj, csin, csinh, ctan and ctanh as 'odd' functions.
1095 2007-01-25 DJ Delorie <dj@redhat.com>
1097 * config/m32c/m32c.c (m32c_cannot_change_mode_class): We don't
1098 allow changes to modes which don't fit in those registers.
1100 * reload1.c (choose_reload_regs): Check for invalid subregs before
1101 computing their locations, not after.
1103 2007-01-25 Geoffrey Keating <geoffk@apple.com>
1106 * config/rs6000/rs6000.c (first_altivec_reg_to_save): On Darwin,
1107 save Altivec registers in an eh_return function.
1108 (compute_vrsave_mask): Likewise.
1109 (rs6000_stack_info): Correct AIX/Darwin stack alignment computation
1110 for saving Altivec registers.
1111 (rs6000_emit_prologue): Don't allocate stack twice in
1112 eh_return function. Correct expected value of altivec_save_offset
1113 when using save_world. Describe save of R0 to stack when using
1114 save_world. Describe stack pointer adjustment when using
1115 save_world. Remove duplicated eh_return parameter register saving.
1116 Update sp_offset variable after save_world.
1117 * config/rs6000/t-darwin (LIB2FUNCS_STATIC_EXTRA): Remove
1119 (LIB2FUNCS_EXTRA): Add darwin-world.asm.
1120 * config/rs6000/darwin.h (SUBTARGET_OVERRIDE_OPTIONS): -m64
1123 2007-01-25 Steve Ellcey <sje@cup.hp.com>
1125 * config.gcc (ia64*-*-hpux*): Make posix threads the default.
1127 2007-01-25 Steve Ellcey <sje@cup.hp.com>
1130 * config/pa/pa.c (pa_init_builtins): Set asm names for finite routines.
1131 * config/ia64/ia64.c (ia64_init_builtins): Ditto.
1133 2007-01-25 Richard Guenther <rguenther@suse.de>
1135 * doc/invoke.texi (-Wcoverage-mismatch): Document.
1136 * common.opt (-Wcoverage-mismatch): New warning option.
1137 * coverage.c (get_coverage_counts): Ignore coverage mismatch
1138 if -Wcoverage-mismatch is given.
1140 2007-01-25 Razya Ladelsky <razya@il.ibm.com>
1142 * ipa-cp.c (ipcp_insert_stage, ipcp_driver): Support for SSA.
1143 (ipcp_driver): Change to static definition.
1144 Add dumping of the ifunctions.
1145 (constant_val_insert): Remove unused parameter. Support for SSA.
1146 (ipcp_propagate_const): Support for SSA.
1147 (ipcp_profile_bb_print): Print only analyzed nodes.
1148 (ipcp_replace_map_create): Remove support for Fortran constant
1150 * ipa-prop.c (ipa_method_modify_stmt,
1151 ipa_callsite_compute_param): Support for SSA.
1152 * ipa-prop.h (ipcp_driver): Remove declaration.
1153 (IS_VALID_TREE_MAP_INDEX): Add define.
1155 2007-01-24 Geoffrey Keating <geoffk@apple.com>
1157 * unwind-dw2.c (execute_stack_op): Handle DW_OP_swap.
1159 2007-01-24 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1162 * builtins.c (fold_builtin_cabs): Use MPFR to evaluate a
1163 constant argument to cabs and do it without checking for
1164 -funsafe-math-optimizations.
1166 2007-01-24 Douglas Gregor <dgregor@osl.iu.edu>
1168 * c-common.h (RID_FIRST_CXX0X): New.
1169 (RID_LAST_CXX0X): New.
1170 * c-opts.c (c_common_handle_option): -Wc++0x-compat is triggered
1172 * c.opt (Wc++0x-compat): New.
1173 * doc/invoke.texi (-Wc++0x-compat): Document.
1175 2007-01-24 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1177 * builtins.c (fold_builtin_carg): New.
1178 (fold_builtin_1): Use it.
1180 2007-01-24 Jan Hubicka <jh@suse.cz>
1182 * ipa-inline.c (cgraph_decide_inlining): Initialize initial_insns.
1184 * ipa-inline.c (initial_insns, max_insns): Delete.
1185 (compute_max_insns): New function.
1186 (cgraph_decide_inlining_of_small_function): Use it; take minimal amount
1187 of insns as base for code growth.
1188 (cgraph_decide_inlining): Make initial_insns local; do not compute
1190 * params.def (PARAM_INLINE_UNIT_GROWTH): Set to 60.
1191 * doc/invoke.texi (inline-unit-growth): Update docs.
1193 2007-01-24 Jakub Jelinek <jakub@redhat.com>
1195 * config/i386/i386.h (x86_cmpxchg16b): Remove const.
1196 (TARGET_CMPXCHG16B): Define to x86_cmpxchg16b.
1197 * config/i386/i386.c (x86_cmpxchg16b): Remove const.
1198 (override_options): Add PTA_CX16 flag. Set x86_cmpxchg16b
1199 for CPUs that have PTA_CX16 set.
1202 * gimplify.c (omp_check_private): New function.
1203 (gimplify_scan_omp_clauses): Use it for
1204 firstprivate/lastprivate/reduction.
1207 * gimplify.c (omp_add_variable): Don't call omp_notice_variable
1208 on TYPE_SIZE_UNIT for GOVD_LOCAL VLAs.
1211 * omp-low.c (lower_omp_for_lastprivate): Add dlist argument.
1212 If lower_lastprivate_clauses emits some statements, append them
1213 to dlist rather than body_p and to body_p append an initializer.
1214 (lower_omp_for): Adjust caller.
1216 2007-01-24 Steve Ellcey <sje@cup.hp.com>
1218 * target.h (globalize_decl_name): New.
1219 * target-def.h (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
1220 * output.h (default_globalize_decl_name): New.
1221 * varasm.c (asm_output_bss): Use globalize_decl_name instead of
1223 (globalize_decl): Ditto.
1224 (default_globalize_decl_name): New.
1225 * config/ia64/ia64.c (ia64_globalize_decl_name): New.
1226 (ia64_handle_version_id_attribute): New.
1227 (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
1228 (ia64_asm_output_external): Use globalize_decl_name instead
1230 * doc/extend.texi (version_id): New pragma.
1231 * doc/tm.texi (ARGET_ASM_GLOBALIZE_DECL_NAME): New target hook.
1233 2007-01-24 Andreas Krebbel <krebbel1@de.ibm.com>
1235 * unwind-dw2-fde.c (get_cie_encoding): Replaced _Unwind_Word with
1236 _uleb128_t and _Unwind_SWord with _sleb128_t.
1237 * unwind-dw2.c (extract_cie_info, execute_stack_op, execute_cfa_program,
1238 uw_frame_state_for, uw_update_context_1): Likewise.
1239 * unwind-c.c (parse_lsda_header, PERSONALITY_FUNCTION): Likewise.
1240 * unwind-pe.h (read_uleb128, read_sleb128,
1241 read_encoded_value_with_base): Likewise.
1242 * unwind-generic.h: Define _sleb128_t and _uleb128_t types.
1244 2007-01-24 Richard Guenther <rguenther@suse.de>
1246 * builtins.c (expand_builtin_cexpi): Get the fndecl
1247 for cexp in the correct way.
1249 2007-01-24 Jan Hubicka <jh@suse.cz>
1251 * tree-ssa-dce.c (eliminate_unnecesary_stmts): Remove dead LHS of calls.
1253 2007-01-24 Andreas Krebbel <krebbel1@de.ibm.com>
1255 * c-cppbuiltin.c (builtin_define_type_sizeof): New function.
1256 (c_cpp_builtins): New builtin macros: __SIZEOF_INT__, __SIZEOF_LONG__,
1257 __SIZEOF_LONG_LONG__, __SIZEOF_SHORT__, __SIZEOF_POINTER__,
1258 __SIZEOF_FLOAT__, __SIZEOF_DOUBLE__, __SIZEOF_LONG_DOUBLE__,
1259 __SIZEOF_SIZE_T__, __SIZEOF_WCHAR_T__, __SIZEOF_WINT_T__ and
1260 __SIZEOF_PTRDIFF_T__.
1261 * doc/cpp.texi: Documentation for the new builtin macros added.
1263 2007-01-24 Uros Bizjak <ubizjak@gmail.com>
1265 * config/i386/i386.md (tanxf2, tan<mode>2, atan<mode>2, log<mode>2,
1266 log10<mode>2, log2<mode>2, expxf2, exp10xf2, exp2xf2): Use op2
1267 instead of operands[2] to avoid access past the end of array.
1269 2007-01-24 Richard Sandiford <richard@codesourcery.com>
1271 * reload1.c (emit_reload_insns): Pass the reload register
1272 for a non-spill output reload through forget_old_reloads_1.
1274 2007-01-23 Joseph Myers <joseph@codesourcery.com>
1276 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Define
1277 _SOFT_DOUBLE if doubles use software floating-point.
1278 * config/rs6000/libgcc-ppc-glibc.ver: Export additional long
1279 double functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
1280 * config/rs6000/darwin-ldouble.c: Also compile functions for
1281 hard-float without FPRs. Use fmsub function for all __NO_FPRS__
1282 cases. Compile extra functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
1283 * config/rs6000/linuxspe.h (SUBSUBTARGET_OVERRIDE_OPTIONS): Remove
1284 commented-out long double override.
1285 (CPP_LONGDOUBLE_DEFAULT_SPEC): Likewise.
1286 * config/rs6000/eabispe.h: Likewise.
1287 * config/rs6000/rs6000.c (rs6000_override_options): Don't override
1288 long double for non-SPE.
1289 (rs6000_handle_option): Likewise.
1290 (invalid_e500_subreg): Disallow more subregs involding DImode,
1291 DFmode, TImode or TFmode.
1292 (rs6000_legitimate_offset_address_p): Check TFmode offsets for
1294 (legitimate_lo_sum_address_p): Also check for TFmode for E500
1296 (rs6000_legitimize_address): Also handle TFmode for E500 double.
1297 (rs6000_legitimize_reload_address): Also handle TFmode for E500
1299 (rs6000_legitimate_address): Also check for TFmode for E500
1301 (rs6000_emit_move): Use DFmode subregs of TFmode for E500 double.
1302 (spe_build_register_parallel): Handle TFmode and TCmode.
1303 (rs6000_spe_function_arg): Handle TFmode and TCmode for E500
1305 (function_arg): Handle TFmode and TCmode for E500 double.
1306 (rs6000_init_libfuncs): Initialize extra libfuncs for soft double
1308 (print_operand): Handle TFmode and TImode for %y.
1309 (rs6000_generate_compare): Handle TFmode comparisons for E500
1311 (spe_func_has_64bit_regs_p): Check for TFmode for E500 double.
1312 (rs6000_function_value): Handle TFmode and TCmode for E500 double.
1313 (rs6000_libcall_value): Handle TFmode and TCmode for E500 double.
1314 * config/rs6000/rs6000.h (CANNOT_CHANGE_MODE_CLASS): Check for
1315 TFmode for E500 double.
1316 * config/rs6000/rs6000.md (FP): Allow TF for E500 double.
1317 (floatsidf2): Enable for E500 double.
1318 (movtf_softfloat): Use rs6000_nonimmediate_operand.
1319 (extenddftf2): Change to extenddftf2_fprs.
1320 (extenddftf2): Call gen_spe_extenddftf2 or gen_extenddftf2_fprs
1321 depending on TARGET_E500_DOUBLE.
1322 (extendsftf2): Enable for E500 double.
1323 (trunctfdf2): Enable for E500 double.
1324 (trunctfsf2): Change to trunctfsf2_fprs.
1325 (trunctfsf2): Call gen_spe_trunctfsf2 or gen_trunctfsf2_fprs
1326 depending on TARGET_E500_DOUBLE.
1327 (floatsitf2): Enable for E500 double.
1328 (fix_trunctfsi2): Change to fix_trunctfsi2_fprs.
1329 (fix_trunctfsi2): Call gen_spe_fix_trunctfsi2 or
1330 gen_fix_trunctfsi2_fprs depending on TARGET_E500_DOUBLE.
1331 (negtf2): Change to negtf2_internal.
1332 (negtf2): New expander.
1333 (abstf2): Enable for E500 double. Call gen_spe_abstf2_tst,
1334 gen_spe_abstf2_cmp or gen_abstf2_internal depending on
1335 TARGET_E500_DOUBLE and flag_unsafe_math_optimizations.
1336 (movdi_internal32): Use rs6000_nonimmediate_operand.
1337 (unnamed splitter): Likewise.
1338 * config/rs6000/spe.md (CMPTFEQ_GPR, TSTTFEQ_GPR, CMPTFGT_GPR,
1339 TSTTFGT_GPR, CMPTFLT_GPR, TSTTFLT_GPR): New unspecs.
1340 (SPE64TF, DITI): New mode macros.
1341 (frob_df_di): Change to frob_<SPE64:mode>_<DITI:mode>; allow more
1344 (frob_<mode>_di_2): New.
1345 (frob_tf_di_8_2): New.
1346 (frob_di_df): Change to frob_di_<mode>; allow more modes.
1348 (frob_di_df_2): Change to frob_<DITI:mode>_<SPE64:mode>_2; allow
1350 (frob_ti_<mode>_8_2): New.
1351 (frob_ti_tf_2): New.
1352 (mov_si<mode>_e500_subreg0, mov_si<mode>_e500_subreg0_2,
1353 mov_si<mode>_e500_subreg4, mov_si<mode>_e500_subreg4_2): Allow
1355 (mov_sitf_e500_subreg8, mov_sitf_e500_subreg8_2,
1356 mov_sitf_e500_subreg12, mov_sitf_e500_subreg12_2): New.
1357 (spe_trunctfdf2_internal1, spe_trunctfsf2, spe_extenddftf2,
1358 spe_fix_trunctfsi2, spe_fix_trunctfsi2_internal,
1359 spe_negtf2_internal, spe_abstf2_cmp, spe_abstf2_tst): New.
1360 (cmptfeq_gpr, tsttfeq_gpr, cmptfgt_gpr, tsttfgt_gpr, cmptflt_gpr,
1363 2007-01-23 Ian Lance Taylor <iant@google.com>
1365 * Makefile.in (OBJS-common): Reformat, alphabetize, but put
1367 (OBJS-archive): Reformat, alphabetize.
1368 (OBJS): Change out_object_file to OBJS-md.
1370 2007-01-23 Uros Bizjak <ubizjak@gmail.com>
1372 * config/i386/i386.md: Use REG_P, MEM_P, CONST_INT_P, LABEL_P,
1373 JUMP_P and CALL_P predicates where applicable.
1374 * config/i386/i386.c: Ditto.
1375 * config/i386/i386.md: Ditto.
1376 * config/i386/mmx.md: Ditto.
1377 * config/i386/predicates.md: Ditto.
1379 2007-01-22 Andreas Schwab <schwab@suse.de>
1381 * config/m68k/m68k.h: Fix comment.
1383 2007-01-22 Jan Hubicka <jh@suse.cz>
1385 * passes.c (init_optimization_passes): Do not rerun
1386 pass_early_warn_uninitialized.
1388 2007-01-22 Richard Guenther <rguenther@suse.de>
1390 PR tree-optimization/30038
1391 * tree-ssa-math-opts.c (maybe_record_sincos): New static helper
1393 (execute_cse_sincos_1): Likewise.
1394 (execute_cse_sincos): Likewise.
1395 (gate_cse_sincos): Likewise.
1396 (pass_cse_sincos): New pass CSEing sin() and cos() calls using
1397 the cexpi() canonicalization of sincos().
1398 * tree-pass.h (pass_cse_sincos): Declare.
1399 * passes.c (init_optimization_passes): New pass pas_cse_sincos.
1401 2007-01-21 Eric Botcazou <ebotcazou@libertysurf.fr>
1403 PR rtl-optimization/29329
1404 * combine.c (replaced_rhs_insn): Rename to i2mod.
1405 (replaced_rhs_value): Rename to i2mod_new_rhs.
1406 (i2mod_old_rhs): New global variable.
1407 (combine_instructions): Adjust for above change. Save a copy of
1408 the old RHS into i2mod_old_rhs when the contents of a REG_EQUAL
1409 note are substituted in the second instruction.
1410 (distribute_notes) <REG_DEAD>: Adjust for above change. Do not
1411 ditch the note if it pertains to the second eliminated register
1412 and this register is mentioned in i2mod_old_rhs.
1415 2006-09-12 Eric Botcazou <ebotcazou@libertysurf.fr>
1417 * combine.c (distribute_notes) <REG_DEAD>: Do not consider SETs past
1418 the insn to which the note was originally attached.
1420 2007-01-21 Jan Hubicka <jh@suse.cz>
1422 * ipa-inline.c (inlining_mode): Comment, move up.
1423 (cgraph_decide_inlining_incrementally): Do not perform inlining
1424 itself; fix handling of flattening of self recursive functions.
1425 (cgraph_find_cycles): Remove.
1426 (cgraph_flatten_node): Remove.
1427 (cgraph_decide_inlining): Use incremental inliner to handle flattening.
1428 (try_inline): New function.
1429 (cgraph_early_inlining): Update call of
1430 cgraph_decide_inlining_incrementally. Apply inlining here.
1431 (apply_inline): Update call of cgraph_decide_inlining_incrementally.
1433 2007-01-21 Dirk Mueller <dmueller@suse.de>
1436 * tree-vrp.c (check_array_bounds): do not warn
1437 about ADDR_EXPR's of ARRAY_REF's which are immediately
1438 used in binary expressions.
1440 2007-01-21 Ira Rosen <irar@il.ibm.com>
1442 * tree-vectorizer.h (struct _stmt_vec_info): Add new field
1443 read_write_dep and macros for its access.
1444 * tree-vectorizer.c (new_stmt_vec_info): Initialize the new field.
1445 * tree-vect-analyze.c (vect_analyze_data_ref_dependence): Remove
1446 argument, call vect_check_interleaving for every independent pair of
1447 data-refs. Mark loads that access the same memory location as a store
1449 (vect_check_dependences): Remove.
1450 (vect_analyze_data_ref_dependences): Remove vect_check_dependences
1451 call, fix the call to vect_analyze_data_ref_dependence.
1452 (vect_analyze_data_ref_access): For statements that access the same
1453 data-ref, check that they are not stores; for loads, check that there
1454 is no store that access the same location.
1456 2007-01-20 Roger Sayle <roger@eyesopen.com>
1457 Joseph Myers <joseph@codesourcery.com>
1459 * doc/invoke.texi (-fdump-rtl-combine): Fix under/overfull hbox.
1460 (-fdump-rtl-stack): Likewise.
1461 (-fno-signed-zeros): Use @minus{} for a minus sign. Correct typo.
1462 (-mcheck-zero-division, -mcpu): Fix under/overfull hbox.
1463 (-mpt-fixed): Use @minus{} for minus sign.
1464 (Using Precompiled Headers): Fix under/overfull hbox.
1466 2007-01-20 Jan Hubicka <jh@suse.cz>
1468 * tree-flow.h (struct stmt_ann_d): Move references_memory to proper
1469 place within annotation.
1471 2007-01-20 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
1473 * pa.c (output_move_double): Change array size of xoperands to 4.
1475 2007-01-20 Jan Hubicka <jh@suse.cz>
1477 * tree-tailcall.c (adjust_return_value): Do not use RESULT_DECL
1480 2007-01-19 Ian Lance Taylor <iant@google.com>
1482 * expmed.c (expand_divmod) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]: Cast
1483 constant to unsigned HOST_WIDE_INT before negating.
1485 2007-01-19 Ian Lance Taylor <iant@google.com>
1487 * tree-ssa-operands.h (struct vuse_vec_d): Change num_vuse field
1489 (VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Use unsigned comparison.
1490 (VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
1491 (SET_VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Likewise.
1492 (SET_VUSE_ELEMENT_VAR) [ENABLE_CHECKING]: Likewise.
1493 (SET_VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
1494 (realloc_vdef, realloc_vuse): Change second parameter to
1496 (ssa_operand_iterator_d): Change vuse_index and mayuse_index
1498 * tree-ssa-operands.c (realloc_vop): Change num_elem parameter to
1499 unsigned. Change x and lim locals to unsigned.
1500 (realloc_vdef, realloc_vuse): Change num_elem parameter to
1502 (finalize_ssa_vuse_ops): Change old_i local to unsigned.
1503 (copy_virtual_operands): Change i and n locals to unsigned.
1505 2007-01-19 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1508 * builtins.c (fold_builtin_1): Handle builtin fdim.
1510 2007-01-20 Jan Hubicka <jh@suse.cz>
1512 * tree-ssa.c (init_tree_ssa): Do not call init_alias_heapvars.
1513 * tree-ssa-structalias.c (compute_points_to_sets): Do call
1514 init_alias_heapvars.
1515 (init_alias_heapvars): Initialize only when not already initialized.
1516 (delete_alias_heapvars): Set heapvar_for_stmt to NULL.
1518 2007-01-19 Roger Sayle <roger@eyesopen.com>
1520 * common.opt (fsigned-zeros): New command line option.
1521 * flags.h (HONOR_SIGNED_ZEROS): Control via flag_signed_zeros instead
1522 of flag_unsafe_math_optimizations.
1523 * opts.c (set_fast_math_flags): The -ffast-math command line option
1524 implies -fno-signed-zeros.
1525 (fast_math_flags_set_p): Likewise.
1527 * doc/invoke.texi: Document new -fno-signed-zeros option, and update
1528 the documentation of -ffast-math appropriately. Wrap long lines.
1530 2007-01-19 Steve Ellcey <sje@cup.hp.com>
1532 * system.h (ASM_MAKE_LABEL_LINKONCE): Poison.
1533 * varasm.c (globalize_decl): Remove ASM_MAKE_LABEL_LINKONCE ifdef.
1535 2007-01-19 Tomas Bily <tbily@suse.cz>
1537 * cgraphunit.c (cgraph_finalize_function): Updating of pid
1539 (tree_init_ic_make_global_vars): New function
1540 (tree_init_edge_profiler): call of tree_init_ic_make_global_vars
1541 (tree_gen_ic_profiler): New function
1542 (tree_gen_ic_func_profiler): New function
1543 (tree_profiling): Added calling of tree_gen_ic_func_profiler
1544 (tree_profile_hooks): Added hook for indirec/virtual calls
1545 * value-prof.c (tree_find_values_to_profile): New case for
1547 (tree_values_to_profile): Call for determining indirect/virtual
1549 (tree_indirect_call_to_profile): New function
1550 (tree_ic_transform): New function
1551 (tree_ic): New function
1552 (find_func_by_pid): New function
1553 (init_pid_map): New function
1554 (tree_value_profile_transformations): Added check for
1555 indirect/virtual call transformation
1556 * value-prof.h (enum hist_type): New counter type for
1557 indirect/virtual calls
1558 (profile_hooks): Added new hook for profiling indirect/virtual
1560 * profile.c (instrument_values): New case for indirect/virtual
1562 * gcov-io.h (GCOV_LAST_VALUE_COUNTER): Changed to 6
1563 (GCOV_COUNTER_V_INDIR): New counter type
1564 (GCOV_COUNTER_NAMES): New name of counter "indirect" added
1565 (GCOV_MERGE_FUNCTIONS): New merge function for indirect/virtual
1567 * cgraph.c: Definition of cgraph_max_pid
1568 (cgraph_create_node): Default init of pid attribute
1569 * cgraph.h: Declaration of cgraph_max_pid
1570 (struct cgraph_node): Added pid attribute
1571 * libgcov.c (__gcov_indirect_call_profiler): New function
1572 (__gcov_one_value_profiler_body): New function
1573 (__gcov_one_value_profiler): Body was moved to
1574 __gcov_one_value_profiler_body and calls it
1576 2007-01-19 Basile Starynkevitch <basile@starynkevitch.net>
1578 * doc/gty.texi (Options): Document the mark_hook option to GTY.
1579 * gengtype.c (write_types_data, write_func_for_structure,
1580 write_types, ggc_wtd, pch_wtd): Add skip_hooks to
1581 write_types_data, ggc_wtd, pch_wtd for processing mark_hook.
1582 (walk_type, write_func_for_structure): Generate the mark_hook if
1585 2007-01-19 Jan Hubicka <jh@suse.cz>
1587 * ipa-inline.c (cgraph_decide_inlining_incrementally): Instead of
1588 'early' argument take inlining mode argument specifying whether to
1589 inline for size/speeed or all functions; add support for flattening;
1591 (cgraph_early_inlining): Update call of decide_inlining_incrementally.
1593 2007-01-19 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
1596 * toplev.c (warn_deprecated_use): Use %qD instead of %qs to print
1597 the name of the declared identifier.
1599 2007-01-19 Dirk Mueller <dmueller@suse.de>
1601 * config/i386.h (CONDITIONAL_REGISTER_USAGE): Store
1602 result of PIC_OFFSET_TABLE_REGNUM in temporary variable to avoid
1603 duplicate evaluation.
1605 2007-01-19 Uros Bizjak <ubizjak@gmail.com>
1607 * config/i386/i386.md (acos<mode>2): Rename from acossf2 and acosdf2.
1608 Macroize expander using X87MODEF12 mode macro. Extend operand 1
1609 to XFMode, use acosxf2 and truncate result to requested mode.
1610 Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
1611 (asin<mode>2): Similarly, with asin expanders.
1612 (*fscalexf4_i387): Rename from *fscalexf4.
1613 (expNcorexf3): New expander.
1614 (expxf2, exp10xf2, exp2xf2): Use expNcorexf3 expander.
1615 (exp<mode>2): Rename from expsf2 and expdf2. Macroize expander using
1616 X87MODEF12 mode macro. Extend operand 1 to XFMode, use expxf2 and
1617 truncate result to requested mode. Use SSE_FLOAT_MODE_P to disable
1618 patterns for SSE math.
1619 (exp10<mode>2): Similarly, with exp10 expanders.
1620 (exp2<mode>2): Similarly, with exp2 expanders.
1621 (expm1<mode>2): Similarly, with expm1 expanders.
1622 (ldexp<mode>3): Similarly, with ldexp expanders.
1623 (log<mode>2, log10<mode>2, log2<mode>2, log1p<mode>2, logb<mode>2):
1624 Use gen_truncxf<mode>2_i387_noop to truncate result.
1626 2007-01-19 Richard Sandiford <richard@codesourcery.com>
1628 * config/m68k/m68k.h (PREFERRED_STACK_BOUNDARY): Define to 32
1629 for ColdFire targets.
1631 2007-01-19 Nathan Sidwell <nathan@codesourcery.com>
1632 Richard Sandiford <richard@codesourcery.com>
1634 * config/m68k/m68k.h (M68K_STATIC_CHAIN_REG_NAME): New macro.
1635 (INITIALIZE_TRAMPOLINE): Use STATIC_CHAIN_REGNUM.
1636 (__transfer_from_trampoline): Use M68K_STATIC_CHAIN_REG_NAME.
1637 * config/m68k/m68kelf.h (STATIC_CHAIN_REGNUM): Override.
1638 (M68K_STATIC_CHAIN_REG_NAME): Likewise.
1639 * config/m68k/netbsd-elf.h (M68K_STATIC_CHAIN_REG_NAME): Likewise.
1641 2007-01-19 Richard Sandiford <richard@codesourcery.com>
1643 * config/m68k/m68k.md (adddi_dilshr32): Rename to...
1644 (*adddi_dilshr32): ...this. Fix formatting. Remove commented-out
1645 non-canonical pattern. Restrict to !TARGET_COLDFIRE.
1646 (*adddi_dilshr32_cf): New pattern.
1647 (adddi3, subdi3): Remove first alternatives.
1649 2007-01-19 Richard Sandiford <richard@codesourcery.com>
1651 * config/m68k/m68k.c (notice_update_cc): If an SFmode move is
1652 implemented using move.l, do not use its cc result for floating-point
1655 2007-01-19 Richard Sandiford <richard@codesourcery.com>
1657 * config/m68k/m68k.h (EXTRA_CONSTRAINT): Stop the 'T' constraint
1658 from accepting 's' constraints if flag_pic.
1660 2007-01-19 Richard Sandiford <richard@codesourcery.com>
1662 * config/m68k/m68k.md (bordered, bunordered, buneq, bunge, bungt)
1663 (bunle, bunlt, bltgt, bordered_rev, bunordered_rev, buneq_rev)
1664 (bunge_rev, bungt_rev, bunle_rev, bunlt_rev, bltgt_rev): Change
1665 condition from TARGET_68881 to TARGET_HARD_FLOAT.
1667 2007-01-19 Sandra Loosemore <sandra@codesourcery.com>
1669 * longlong.h (count_leading_zeros, COUNT_LEADING_ZEROS_0): Add
1670 ColdFire alternatives.
1671 * config/m68k/m68k.h (CLZ_DEFINED_VALUE_AT_ZERO): New macro.
1672 * config/m68k/m68k.md (clzsi2): Define for ColdFire
1673 architectures that support ff1 instruction.
1675 2007-01-19 Richard Sandiford <richard@codesourcery.com>
1676 Julian Brown <julian@codesourcery.com>
1678 * config/m68k/m68k.h (CONST_OK_FOR_LETTER_P): Add an 'R' case.
1679 * config/m68k/m68k.md (*movsi_cfv4): Fold into...
1680 (*movsi_cf): ...here. Remove unnecessary 'R' from 'Rg'.
1683 2007-01-19 Richard Sandiford <richard@codesourcery.com>
1685 * config/m68k/m68k-protos.h (valid_mov3q_const): Take a HOST_WIDE_INT
1687 (output_move_const_into_data_reg, output_move_simode_const): Delete.
1688 * config/m68k/m68k.c (const_method, const_int_cost): Take a
1689 HOST_WIDE_INT instead of an rtx.
1690 (m68k_rtx_costs): Update call accordingly.
1691 (output_move_const_into_data_reg): Likewise. Fix formatting.
1692 (valid_mov3q_const): Take a HOST_WIDE_INT instead of an rtx.
1694 (output_move_simode_const): Update calls after above changes.
1695 Rework to use automatic variables and predicates like MEM_P.
1696 * config/m68k/m68k.md (pushexthisi_const): Update call to
1699 2007-01-19 Dirk Mueller <dmueller@suse.de>
1701 * tree-ssa-alias.c (perform_var_substitution): Fix typo
1704 2007-01-19 Richard Guenther <rguenther@suse.de>
1706 * builtins.c (expand_builtin_cexpi): Fall back to expanding
1707 via cexp in case sincos is not available.
1709 2007-01-19 Richard Guenther <rguenther@suse.de>
1711 * doc/tm.texi (TARGET_HAS_SINCOS): Document new target macro.
1712 * defaults.h (TARGET_HAS_SINCOS): Default to off.
1713 * config/linux.h (TARGET_HAS_SINCOS): Set to on if we have glibc.
1714 * config/alpha/linux.h (TARGET_HAS_SINCOS): Likewise.
1715 * config/sparc/linux.h (TARGET_HAS_SINCOS): Likewise.
1716 * config/sparc/linux64.h (TARGET_HAS_SINCOS): Likewise.
1717 * config/rs6000/linux.h (TARGET_HAS_SINCOS): Likewise.
1718 * config/rs6000/linux64.h (TARGET_HAS_SINCOS): Likewise.
1720 2007-01-19 Uros Bizjak <ubizjak@gmail.com>
1722 * config/i386/i386.md (*fpatanxf3_i387, fpatan_extend<mode>xf3_i387):
1724 (atan2sf3_1, atan2df3_1, atan2xf3_1): Remove insn patterns.
1725 (atan2xf3): Directly generate RTL pattern.
1726 (atan2<mode>3): Rename from atan2sf3 and atan2df3 and macroize insn
1727 patterns using X87MODEF12 mode macro. Use fpatan_extend<mode>xf3_i387
1728 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
1729 disable patterns for SSE math.
1730 (atan<mode>2): Rename from atansf2 and atandf2 and macroize insn
1731 patterns using X87MODEF12 mode macro. Use fpatan_extend<mode>xf3_i387
1732 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
1733 disable patterns for SSE math.
1735 2007-01-19 Alexandre Oliva <aoliva@redhat.com>
1737 * libgcc-std.ver: Fix typo in %inherit for GCC_4.3.0.
1739 2007-01-18 Roger Sayle <roger@eyesopen.com>
1741 * fold-const.c (fold_unary) <VIEW_CONVERT_EXPR>: Optimize away a
1742 VIEW_CONVERT_EXPR to the same type as it's operand.
1744 2007-01-18 David Edelsohn <edelsohn@gnu.org>
1746 * config/rs6000/darwin-ldouble.c: Only build _SOFT_FLOAT if
1747 configured for long double 128.
1749 2007-01-18 Mike Stump <mrs@apple.com>
1751 * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Fix build
1754 2007-01-18 Michael Meissner <michael.meissner@amd.com>
1756 * i386.c (ix86_compute_frame_layout): Make fprintf's in #if 0 code
1759 2007-01-18 Jan Hubicka <jh@suse.cz>
1761 * tree-ssa-operands.c (vop_free_bucket_size): Never return value
1762 greater than NUM_VOP_FREE_BUCKETS.
1764 2007-01-18 Daniel Berlin <dberlin@dberlin.org>
1766 * tree-ssa-structalias.c: Update comments.
1767 (ptabitmap_obstack): Removed.
1769 (oldpta_obstack): Ditto.
1770 (stats): Add a few members.
1771 (struct variable_info): Remove node, complex, address_taken, and
1772 indirect_target members. Add oldsolution member.
1773 (new_var_info): Do not initialize removed members.
1774 (constraint_expr_type): Remove INCLUDES.
1775 (constraint_graph): Add size, implicit_preds, rep,
1776 indirect_cycles, eq_rep, label, direct_nodes, and complex members.
1777 (FIRST_REF_NODE): New macro.
1778 (LAST_REF_NODE): Ditto.
1779 (FIRST_ADDR_NODE): Ditto.
1780 (find): New function.
1782 (dump_constraint): Do not handle INCLUDES.
1783 (insert_into_complex): Do not insert duplicate constraints.
1784 (condense_varmap_nodes): Renamed and rewritten into ...
1785 (merge_node_constraints): This. Also fix bug in handling of
1786 offseted copy constraints.
1787 (clear_edges_for_node): No longer need to deal with preds at all,
1788 or removing associated preds/succs.
1789 (merge_graph_nodes): Deal with indirect_cycles.
1790 Don't deal with predecessors.
1791 (add_implicit_graph_edge): New function.
1792 (add_pred_graph_edge): Ditto.
1793 (add_graph_edge): Don't deal with predecessors.
1794 (build_constraint_graph): Removed.
1795 (build_pred_graph): New function.
1796 (build_succ_graph): Ditto.
1797 (struct scc_info): Removed in_component. Added roots, dfs, and
1798 node_mapping. Remove visited_index, unification_queue.
1799 (scc_visit): Deal with union-find we do now.
1800 Deal with cycles with REF nodes.
1801 (collapse_nodes): Renamed and rewritten to ...
1802 (unify_nodes): This.
1803 (process_unification_queue): Removed.
1804 (topo_visit): Cleanup
1805 (do_da_constraint): Use find.
1806 (do_sd_constraint): Ditto.
1807 (do_ds_constraint): Ditto.
1808 (do_complex_constraint): Ditto.
1809 (init_scc_info): Update for removed and added members.
1810 (find_and_collapse_graph_cycles): Renamed and rewritten into ...
1811 (find_indirect_cycles): This.
1812 (equivalence_class): New variable.
1813 (label_visit): New function.
1814 (perform_variable_substitution): Rewritten.
1815 (free_var_substitution_info): New function.
1816 (find_equivalent_node): Ditto.
1817 (move_complex_constraints): Ditto.
1818 (eliminate_indirect_cycles): Ditto.
1819 (solve_graph): Only propagate changed bits.
1820 Use indirect cycle elimination.
1822 (tree_id_t): Rename to tree_vi_t, delete id member, add vi member.
1823 (tree_id_eq): Renamed to ...
1824 (tree_vi_eq): This. Update for member change
1825 (insert_id_for_tree): Renamed and rewritten to ...
1826 (insert_vi_for_tree): This.
1827 (lookup_id_for_tree): Renamed and rewritten to ...
1828 (lookup_vi_for_tree): This.
1829 (get_id_for_tree): Renamed and rewritten to ...
1830 (get_vi_for_tree): Ditto.
1831 (get_constraint_exp_from_ssa_var): Update to use get_vi_for_tree.
1832 (process_constraint): Don't handle INCLUDES.
1833 Remove special ADDRESSOF case.
1834 (find_func_aliases): Rewrite to use vi functions instead of id
1836 (create_function_info_for): Ditto.
1837 (create_variable_info_for): Ditto.
1838 (intra_create_variable_infos): Ditto.
1839 (merge_smts_into): Ditto.
1840 (find_what_p_points_to): Ditto.
1841 (init_base_vars): Ditto.
1842 (init_alias_vars): Ditto.
1843 (remove_preds_and_fake_succs): New function.
1844 (dump_sa_points_to_info): Dump new stats.
1845 (dump_solution_for_var): Use find.
1846 (set_used_smts): Fix formatting.
1847 (compute_points_to_sets): Updated for new functions.
1848 (ipa_pta_execute): Ditto.
1850 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
1851 Richard Sandiford <richard@codesourcery.com>
1853 * doc/tm.texi (TARGET_FUNCTION_VALUE): Expand documentation of
1855 * calls.c (expand_call): If the return value is a PARALLEL,
1856 extract its first member.
1857 * config/m68k/linux.h (FUNCTION_EXTRA_EPILOGUE): Remove.
1858 * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
1859 use FUNCTION_EXTRA_EPILOGUE.
1860 (m68k_function_value): Return a PARALLEL if the return value
1861 is of a pointer type.
1862 * config/m68k/netbsd-elf.h (current_function_returns_pointer)
1863 (FUNCTION_EXTRA_EPILOGUE): Remove.
1864 * config/m68k/m68k.md (D0_REG): New constant.
1866 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
1868 * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
1869 output a NOP for empty epilogues.
1871 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1873 * config/m68k/m68k.c (m68k_use_return_insn): Update comments
1874 before function. Extend register save check to include all
1875 registers, not just integer ones.
1877 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
1879 * config/m68k/m68k-protos.h (use_return_insn): Rename to...
1880 (m68k_use_return_insn): ...this.
1881 * config/m68k/m68k.h (USE_RETURN_INSN): Delete.
1882 * config/m68k/m68k.c (use_return_insn): Rename to...
1883 (m68k_use_return_insn): ...this.
1884 * config/m68k/m68k.md (return): Use m68k_use_return_insn instead
1887 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
1889 * config/m68k/fpgnulib.c (__truncdfsf2): Implement round to
1890 nearest even, fix denormal rounding overflow.
1892 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1894 * config/m68k/m68k.md (movsf_cf_hard): Use fsmove instead of
1895 f%$smove and f%$move.
1896 (movdf_cf_hard): Use fdmove for cases 0 and 3 and fmove for case 1.
1897 (extendsfdf2_cf): Use fdmove instead of f%&move.
1898 (truncdfsf2_cf): Use fsmove instead of f%$smove.
1899 (add<mode>3_cf, sub<mode>3_cf): Use <FP:prec> instead of <FP:round>.
1901 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1903 * config/m68k/m68k.md (movdf_cf_hard): Use output_move_double for
1906 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1908 * real.h (real_format): Add a canonical_nan_lsbs_set field.
1909 (coldfire_single_format): Declare.
1910 (coldfire_double_format): Likewise.
1911 * real.c (encode_ieee_single): Use canonical_nan_lsbs_set instead
1912 of qnan_msb_set to determine the lower bits of a canonical
1914 (encode_ieee_double): Likewise.
1915 (encode_ieee_quad): Likewise.
1916 (ieee_single_format): Initialize canonical_nan_lsbs_set.
1917 (mips_single_format): Likewise.
1918 (ieee_double_format): Likewise.
1919 (mips_double_format): Likewise.
1920 (ieee_extended_motorola_format): Likewise.
1921 (ieee_extended_intel_96_format): Likewise.
1922 (ieee_extended_intel_128_format): Likewise.
1923 (ieee_extended_intel_96_round_53_format): Likewise.
1924 (ibm_extended_format): Likewise.
1925 (mips_extended_format): Likewise.
1926 (ieee_quad_format): Likewise.
1927 (mips_quad_format): Likewise.
1928 (vax_f_format): Likewise.
1929 (vax_d_format): Likewise.
1930 (vax_g_format): Likewise.
1931 (i370_single_format): Likewise.
1932 (i370_double_format): Likewise.
1933 (decimal_single_format): Likewise.
1934 (decimal_double_format): Likewise.
1935 (decimal_quad_format): Likewise.
1936 (c4x_single_format): Likewise.
1937 (c4x_extended_format): Likewise.
1938 (real_internal_format): Likewise.
1939 (coldfire_single_format): New real_format.
1940 (coldfire_double_format): Likewise.
1941 * config/pdp11/pdp11.c (pdp11_f_format): Initialize
1942 canonical_nan_lsbs_set.
1943 (pdp11_d_format): Likewise.
1944 * config/m68k/m68k.c (override_options): Override REAL_FORMAT_MODE
1945 if TARGET_COLDFIRE_CPU.
1947 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1949 * config/m68k/m68k-protos.h (m68k_output_pic_call): Delete.
1950 (output_call, m68k_legitimize_call_address): Declare.
1951 * config/m68k/m68k.h (EXTRA_CONSTRAINT): Remove unnecessary
1952 parenthesees. Add support for a 'W' constraint.
1953 (LEGITIMATE_PIC_OPERAND_P): Remove SYMBOL_REF_FLAG handling.
1954 (PRINT_OPERAND_PUNCT_VALID_P): Remove comment about 'o'.
1955 (m68k_symbolic_call, m68k_symbolic_jump): Declare.
1956 * config/m68k/m68k.c (m68k_symbolic_call, m68k_symbolic_jump): New
1958 (override_options): Initialize them. Do not set flag_no_function_cse
1959 for TARGET_ID_SHARED_LIBRARY.
1960 (m68k_output_pic_call): Delete.
1961 (m68k_legitimize_call_address): New function.
1962 (print_operand): Remove the %o prefix. Handle the %p prefix.
1963 (output_call): New function.
1964 (m68k_output_mi_thunk): Use m68k_symbolic_jump. Always load the
1965 target address from the GOT if symbolic jumps are not allowed.
1966 * config/m68k/m68k.md (call, general_operand): Do not set
1967 SYMBOL_REF_FLAG. Use m68k_legitimize_call_address instead.
1968 Merge separate flag_pic and !flag_pic define_insns into...
1969 (*call, *call_value): ...these new patterns. Match the address
1970 rather than the containing MEM and require it to be a call_operand.
1971 Use output_call to generate the asm template.
1972 * config/m68k/predicates.md (const_call_operand): New predicate.
1973 (call_operand): Likewise.
1975 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
1977 * config/m68k/m68k.h (REGISTER_MOVE_COST): Simplify definition.
1978 (STACK_GROWS_DOWNWARD): Define to 1.
1979 (FUNCTION_VALUE, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P): Equivocate
1980 comments, emphasizing that these values are only defaults.
1981 * config/m68k/linux.h (LINK_SPEC): Fix formatting in #undef.
1982 * config/m68k/m68k.c (const_method): Remove trailing whitespace.
1984 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1986 * config/m68k/m68k.md (cmpsi): Remove outdated flag_pic handling.
1988 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
1989 Richard Sandiford <richard@codesourcery.com>
1991 * config/m68k/m68k.h (DATA_REGNO_P, ADDRESS_REGNO_P, INT_REGNO_P)
1992 (FP_REGNO_P): New macros.
1993 (REGNO_OK_FOR_INDEX_P, REGNO_OK_FOR_BASE_P, REGNO_OK_FOR_DATA_P)
1994 (REGNO_OK_FOR_FP_P, REG_OK_FOR_INDEX_P, REG_OK_FOR_BASE_P): Use them.
1995 (EH_RETURN_STACKADJ_RTX): Use A0_REG.
1996 * config/m68k/m68k.c (m68k_regno_mode_ok): Use the new REGNO macros.
1998 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
2000 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
2001 (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-rtems*): Add t-floatlib
2003 * config/m68k/t-floatlib: New file, extracting common code from...
2004 * config/m68k/t-m68kbare, config/m68k/t-m68kelf,
2005 * config/m68k/t-uclinux: Here.
2006 * config/m68k/fpgnulib.c: Do not compile extendeed precision
2007 routines on ColdFire targets.
2009 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
2011 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
2012 (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*)
2013 (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-linux*)
2014 (m68k-*-rtems*): Use tm_file rather than m68k/m68k.h and
2015 explicitly set MOTOROLA to 1.
2016 * config/m68k/m68k.h (MOTOROLA): Simplify definition accordingly.
2018 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2019 Nathan Sidwell <nathan@codesourcery.com>
2021 * config/m68k/m68k.h (PCC_STATIC_STRUCT_RETURN): Delete.
2022 (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Add MOTOROLA cases.
2023 * config/m68k/coff.h (REGISTER_PREFIX_MD): Delete.
2024 * config/m68k/m68020-elf.h (LIB_SPEC): Delete.
2025 * config/m68k/m68k-none.h (CC1_SPEC, CPP_SUBTARGET_SPEC): Delete.
2026 * config/m68k/m68kelf.h (IMMEDIATE_PREFIX, REGISTER_PREFIX_MD)
2027 (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Delete.
2028 (NO_DOLLAR_IN_LABEL, PCC_STATIC_STRUCT_RETURN): Don't undefine.
2029 (BSS_ASM_OP, ASM_OUTPUT_SKIP, ASM_OUTPUT_ASCII): Delete.
2030 * config/m68k/m68kemb.h (PCC_STATIC_STRUCT_RETURN): Don't undefine.
2031 * config/m68k/linux.h (SIZE_TYPE, PTRDIFF_TYPE, WCHAR_TYPE)
2032 (WCHAR_TYPE_SIZE, TARGET_OBJFMT_CPP_BUILTINS): Delete.
2033 (TARGET_OS_CPP_BUILTINS): Don't define mc68000 and mc68020 here.
2034 (DBX_REGISTER_NUMBER): Delete.
2035 * config/m68k/netbsd-elf.h (IMMEDIATE_PREFIX): Delete.
2036 (PCC_STATIC_STRUCT_RETURN): Don't undefine.
2037 * config/m68k/openbsd.h (PCC_STATIC_STRUCT_RETURN): Define.
2039 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2041 * config.gcc (m68k-*-uclinux*): Add flat.h to $tm_file.
2042 * config/flat.h: New file.
2043 * crtstuff.c (USE_PT_GNU_EH_FRAME): Don't define if
2045 * config/m68k/m68k.h (ASM_PREFERRED_EH_DATA_FORMAT): Do not use
2046 indirect references for -msep-data or -mid-shared-library.
2047 Do not use PC-relative code addresses either.
2048 * config/m68k/m68k.c (override_options): Restrict -fPIC error
2050 * config/m68k/uclinux.h (STARTFILE_SPEC): Define. Use Scrt1.o
2051 for shared libraries and crt1.o for executables. Use crti.o and
2053 (ENDFILE_SPEC): Use crtend.o and crtn.o.
2054 (LIB_SPEC): Suppress -Rlibc.gdb if -static-libc is given.
2055 Do not add -elf2flt or -shared-lib-id options here.
2056 (LINK_SPEC): Define. Pass -elf2flt if no -elf2flt option is given.
2057 Pass -shared-lib-id if -mid-shared-library, taking the library
2058 identifier from -mshared-library-id if given, otherwise
2060 (EH_FRAME_IN_DATA_SECTION): Do not undefine.
2061 (INIT_SECTION_ASM_OP, FINI_SECTION_ASM_OP): Likewise.
2062 (TARGET_OS_CPP_BUILTINS): Define __GXX_MERGED_TYPEINFO_NAMES=0
2063 and __GXX_TYPEINFO_EQUALITY_INLINE=0 if -mid-shared-library.
2064 (DRIVER_SELF_SPECS): Map unadorned PIC options to -msep-data.
2065 * config/m68k/t-uclinux (EXTRA_MULTILIB_PARTS): Add crtbegin.o
2067 * config/m68k/lb1sf68.asm (PICCALL): Use an lea and pc-relative
2068 jump sequence for ISA A and ISA A+.
2069 (PICJUMP): Likewise.
2071 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2073 * config.gcc (m68k-*-uclinux*): Base the port on the common
2074 and m68k GNU/Linux files rather than on the generic ELF ones.
2075 * config/m68k/uclinux.h (TARGET_VERSION): Override.
2076 (TARGET_OS_CPP_BUILTINS): Use LINUX_TARGET_OS_CPP_BUILTINS.
2078 2007-01-18 Julian Brown <julian@codesourcery.com>
2079 Richard Sandiford <richard@codesourcery.com>
2081 * config/m68k/m68k.h (LONG_DOUBLE_TYPE_SIZE): Make 64-bit on ColdFire.
2082 (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
2083 * config/m68k/netbsd-elf.h (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Undefine
2085 * config/m68k/uclinux-oldabi.h (LONG_DOUBLE_TYPE_SIZE): Redefine to
2087 (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
2089 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2091 * doc/install.texi: Document m68k-uclinuxoldabi.
2092 * config.gcc (m68k-*-uclinuxoldabi): New configuration.
2093 * config/m68k/uclinux-oldabi.h: New file, copied from
2094 config/m68k/uclinux.h.
2096 2007-01-18 Kaz Kojima <kkojima@rr.iij4u.or.jp>
2098 * config/m32r/m32r-protos.h (m32r_expand_epilogue): Declare it.
2099 * config/m32r/m32r.c (m32r_setup_incoming_varargs): Use gen_frame_mem.
2100 (m32r_compute_frame_size): Use unsigned for regno.
2101 (m32r_reload_lr): Use gen_frame_mem.
2103 (m32r_output_function_epilogue): Don't output the function epilogue
2105 (m32r_expand_epilogue): New.
2106 (direct_return): Return false if the function has the interrupt
2108 (m32r_hard_regno_rename_ok): Remove code for the textual epilogue.
2109 * config/m32r/m32r.md (epilogue): New expander.
2110 (return_lr, return_rte): New insns.
2111 (return): Make it expander.
2112 (return_normal): New expander.
2114 2007-01-18 Josh Conner <jconner@apple.com>
2117 * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Add
2118 support for UNLE, UNLT, UNGE, and UNGT.
2120 2007-01-18 Jan Hubicka <jh@suse.cz>
2122 * tree-vrp.c (finalize_jump_threads): Do not call cleanup_cfg by hand.
2123 * tree-tailcall (add_virtual_phis): Likewise.
2124 (optimize_tail_call): Return TODOs.
2125 (execute_tail_calls): Return TODOs.
2126 * tree-ssa-ccp (execute_fold_all_builtins): Do cleanup_cfg via TODO.
2127 * tree-cfgcleanup.c (cleanup_tree_cfg_loop): Return if something
2129 * tree-ssa-forwprop.c (tree_ssa_forward_propagate_single_use_value):
2130 Cleanup cfg using TODO.
2131 * tree-flow.h (cleanup_tree_cfg_loop): Update prototype.
2132 * passes.c (execute_function_todo): When cleanup did something, remove
2134 * tree-cfg.c (pass_build_cfg): Add cleanup_cfg TODO.
2135 (make_edges): Don't cleanup_cfg.
2137 2007-01-18 Uros Bizjak <ubizjak@gmail.com>
2139 * reg-stack.c (subst_stack_regs_pat) [UNSPEC_SINCOS_COS,
2140 UNSPEC_XTRACT_FRACT]: Use generic code for instructions that
2141 operate on the top of stack.
2142 [UNSPEC_SINCOS_SIN, UNSPEC_XTRACT_EXP, UNSPEC_TAN]: Rewrite
2143 register handling of instructions that output to the second
2145 [UNSPEC_TAN_ONE, UNSPEC_TAN_TAN]: Remove.
2146 (move_for_stack_reg): Special-case check for dead destination
2147 stack slot for constant load of 1.0 inside UNSPEC_TAN.
2149 * config/i386/i386.md (UNSPEC_TAN): New constant.
2150 (UNSPEC_TAN_ONE, UNSPEC_TAN_TAN): Remove.
2151 (fptanxf4_i387, fptan_extend<mode>xf4_i387): New patterns
2152 to correctly model move of constant 1.0 to top stack slot.
2153 (*tandf3_1, *tansf3_1, *tanxf3_1): Remove insn patterns.
2154 (unnamed peephole2 pattern): Remove corresponding peephole2
2155 pattern that optimizes tan insn and loading of constant 1.0.
2156 (tanxf2): Use fptanxf4_i387.
2157 (tan<mode>2): Rename from tansf2 and tandf2 and macroize insn
2158 patterns using X87MODEF12 mode macro. Use fptan_extend<mode>xf4_i387
2159 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
2160 disable patterns for SSE math.
2161 (sincos<mode>3): Use truncxf<mode>2_i387_noop for truncation.
2162 (fyl2x_extend<mode>xf3_i387): Use X87MODEF12 for operand 1.
2164 2007-01-18 Dirk Mueller <dmueller@suse.de>
2165 Richard Guenther <rguenther@suse.de>
2168 * doc/invoke.texi (Warray-bounds): Document -Warray-bounds.
2169 * common.opt (Warray-bounds): Add new warning option.
2170 * c-opts.c (c_common_handle_option): Define -Warray-bounds
2172 * Makefile.in: make tree-vrp.o depend on toplev.h
2173 * tree-vrp.c (vrp_finalize): Call check_array_refs if -Warray-bounds
2175 (check_array_refs, check_array_bounds, check_array_ref): New.
2177 2007-01-18 Jan Hubicka <jh@suse.cz>
2179 * tree-ssa-ccp.c (ccp_finalize): Return if something changed.
2180 (execute_ssa_ccp): Return flags conditionally.
2181 * tree-ssa-propagate.c (substitue_and_fold): Return if something was
2183 * tree-ssa-propagate.h (substitute_and_fold): Update prototype.
2185 2007-01-18 Steven Bosscher <steven@gcc.gnu.org>
2187 * cfgcleanup.c (cleanup_cfg): Detect cfglayout mode and set
2188 the CLEANUP_CFGLAYOUT flag when in cfglayout mode.
2190 * Makefile.c (GTFILES): Add cfglayout.h.
2191 * gengtype.c (open_base_files): Likewise.
2192 * cfglayout.c (cfg_layout_function_footer,
2193 cfg_layout_function_header) Reindent to make gengtype happy.
2194 * cfglayout.h (cfg_layout_function_footer,
2195 cfg_layout_function_header): Add GTY(()) marker.
2197 * ifcvt.c (noce_try_sign_mask): Make sure INSN_B is non-null.
2199 2007-01-18 Ben Elliston <bje@au.ibm.com>
2201 * genautomata.c (write_automata): Include xstrerror output in the
2202 error message if writing the DFA description file fails.
2204 2007-01-17 H.J. Lu <hongjiu.lu@intel.com>
2206 * config/mips/mips-protos.h (mips_output_external): Make it
2208 * config/mips/iris.h (TARGET_ASM_EXTERNAL_LIBCALL): Removed.
2209 * config/mips/mips.c (irix_output_external_libcall): Likewise.
2210 (extern_list): Likewise.
2211 (extern_head): Likewise.
2212 (TARGET_ASM_FILE_END): Likewise.
2213 (mips_file_end): Likewise.
2214 (mips_output_external): Rewritten.
2216 2007-01-18 Ben Elliston <bje@au.ibm.com>
2218 * genpreds.c (write_insn_preds_c): Only write out the function
2219 body for regclass_for_constraint if we have register constraints.
2221 2007-01-17 Tom Tromey <tromey@redhat.com>
2223 * doc/sourcebuild.texi (libgcj Tests): Use sourceware.org.
2224 * doc/install.texi (Testing): Use sourceware.org.
2225 (Binaries): Likewise.
2226 (Specific): Likewise.
2227 * doc/contrib.texi (Contributors): Use sourceware.org.
2229 2007-01-17 Anatoly Sokolov <aesok@post.ru>
2231 * config/avr/avr.h (AVR_HAVE_LPMX): New macro.
2232 (AVR_ENHANCED): Rename to ...
2233 (AVR_HAVE_MUL): ... new.
2234 (avr_enhanced_p): Rename to ...
2235 (avr_have_mul_p): ... new.
2236 (TARGET_CPU_CPP_BUILTINS): Use 'avr_have_mul_p' instead of
2237 'avr_enhanced_p' for "__AVR_ENHANCED__". Define "__AVR_HAVE_MUL__".
2238 * config/avr/avr.c (avr_enhanced_p): Rename to ...
2239 (avr_have_mul_p): ... new.
2240 (base_arch_s): Rename 'enhanced' to 'have_mul'.
2241 (avr_override_options): Use 'avr_have_mul_p' and 'have_mul' instead of
2242 'avr_enhanced_p' and 'enhanced'.
2243 (ashlhi3_out, ashrhi3_out, lshrhi3_out, avr_rtx_costs): Use
2244 AVR_HAVE_MUL instead of AVR_ENHANCED.
2245 * avr.md (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
2246 (mulqi3, *mulqi3_enh, *mulqi3_call, mulqihi3, umulqihi3, mulhi3,
2247 *mulhi3_enh, *mulhi3_call, mulsi3, *mulsi3_call): Use AVR_HAVE_MUL
2248 instead of AVR_ENHANCED.
2249 (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
2250 * libgcc.S: Use __AVR_HAVE_MUL__ instead of __AVR_ENHANCED__.
2251 (__tablejump__): Use __AVR_HAVE_LPMX__ instead of __AVR_ENHANCED__.
2253 2007-01-17 Ian Lance Taylor <iant@google.com>
2255 * vec.h (VEC_reserve_exact): Define.
2256 (vec_gc_p_reserve_exact): Declare.
2257 (vec_gc_o_reserve_exact): Declare.
2258 (vec_heap_p_reserve_exact): Declare.
2259 (vec_heap_o_reserve_exact): Declare.
2260 (VEC_OP (T,A,reserve_exact)): New static inline function, three
2262 (VEC_OP (T,A,reserve)) [all versions]: Remove handling of
2264 (VEC_OP (T,A,alloc)) [all versions]: Call ...reserve_exact.
2265 (VEC_OP (T,A,copy)) [all versions]: Likewise.
2266 (VEC_OP (T,a,safe_grow)) [all versions]: Likewise.
2267 * vec.c (calculate_allocation): Add exact parameter. Change all
2269 (vec_gc_o_reserve_1): New static function, from vec_gc_o_reserve.
2270 (vec_gc_p_reserve, vec_gc_o_reserve): Call vec_gc_o_reserve_1.
2271 (vec_gc_p_reserve_exact, vec_gc_o_reserve_exact): New functions.
2272 (vec_heap_o_reserve_1): New static function, from vec_heap_o_reserve.
2273 (vec_heap_p_reserve, vec_heap_o_reserve): Call vec_heap_o_reserve_1.
2274 (vec_heap_p_reserve_exact): New function.
2275 (vec_heap_o_reserve_exact): New function.
2277 2007-01-17 Jan Hubicka <jh@suse.cz>
2279 * ipa-type-escape.c (look_for_casts): Revamp using handled_component_p.
2281 2007-01-17 Eric Christopher <echristo@apple.com>
2283 * config.gcc: Support core2 processor.
2285 2007-01-16 Jan Hubicka <jh@suse.cz>
2287 * tree-ssanames.c (release_dead_ssa_names): Instead of ggc_freeing
2288 the names, just unlink the chain so we don't crash on dangling pointers
2291 2007-01-16 Jan Hubicka <jh@suse.cz>
2293 * cgraph.h (cgraph_decide_inlining_incrementally): Kill.
2294 * tree-pass.h: Reorder to make IPA passes appear toegher.
2295 (pass_early_inline, pass_inline_parameters, pass_apply_inline): Declare.
2296 * cgraphunit.c (cgraph_finalize_function): Do not compute inling
2297 parameters, do not call early inliner.
2298 * ipa-inline.c: Update comments. Include tree-flow.h
2299 (cgraph_decide_inlining): Do not compute inlining parameters.
2300 (cgraph_decide_inlining_incrementally): Return TODOs; assume to
2301 be called with function context set up.
2302 (pass_ipa_inline): Remove unreachable functions before pass.
2303 (cgraph_early_inlining): Simplify assuming to be called from the
2305 (pass_early_inline): New pass.
2306 (cgraph_gate_ipa_early_inlining): New gate.
2307 (pass_ipa_early_inline): Turn into simple wrapper.
2308 (compute_inline_parameters): New function.
2309 (gate_inline_passes): New gate.
2310 (pass_inline_parameters): New pass.
2311 (apply_inline): Move here from tree-optimize.c
2312 (pass_apply_inline): New pass.
2313 * ipa.c (cgraph_remove_unreachable_nodes): Verify cgraph after
2315 * tree-inline.c (optimize_inline_calls): Return TODOs rather than
2317 (tree_function_versioning): Do not allocate dummy struct function.
2318 * tree-inline.h (optimize_inline_calls): Update prototype.
2319 * tree-optimize.c (execute_fixup_cfg): Export.
2320 (pass_fixup_cfg): Remove
2321 (tree_rest_of_compilation): Do not apply inlines.
2322 * tree-flow.h (execute_fixup_cfg): Declare.
2323 * Makefile.in (gt-passes.c): New.
2324 * passes.c: Include gt-passes.h
2325 (init_optimization_passes): New passes.
2326 (nnodes, order): New static vars.
2327 (do_per_function_toporder): New function.
2328 (execute_one_pass): Dump current pass here.
2329 (execute_ipa_pass_list): Don't dump current pass here.
2331 2007-01-16 Janis Johnson <janis187@us.ibm.com>
2333 * config/dfp-bit.c (dfp_compare_op): Return separate value for NaN.
2334 (DFP_NE, DFP_LE, DFP_GE): Return false for NaN.
2336 2007-01-16 David Edelsohn <edelsohn@gnu.org>
2338 * config/rs6000/darwin-ldouble.c: Build file for SOFT_FLOAT.
2339 (strong_alias): Define.
2340 (__gcc_qmul): Provide non-FMA for soft-float.
2346 (__gcc_qunord): New.
2356 * config/rs6000/rs6000.c (rs6000_init_libfuncs): Initialize
2357 soft-float functions.
2358 * config/rs6000/libgcc-ppc-glibc.ver: Version soft-float symbols.
2359 * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Do not warn
2360 about long double soft float.
2362 2007-01-16 Dorit Nuzman <dorit@il.ibm.com>
2363 Tehila Meyzels <tehila@il.ibm.com>
2365 * tree-vectorizer.h (is_pattern_stmt_p): New.
2366 * tree-vect-analyze.c (vect_determine_vectorization_factor): Fix
2367 formatting (tabs instead of spaces). Cleanup and clarify setting
2368 of STMT_VINFO_VECTYPE. Call is_pattern_stmt_p.
2369 * tree-vect-transform.c (vect_get_vec_def_for_operand): Fix typo.
2370 (vectorizable_type_demotion): Check that types are integral.
2371 (vectorizable_type_promotion): Likewise.
2372 (vectorizable_store): Fix typo. Eliminate new-line at end of
2375 2007-01-16 Jan Hubicka <jh@suse.cz>
2377 * tree-ssanames.c (release_dead_ssa_names): Remove invalidated
2380 2007-01-15 Eric Christopher <echristo@apple.com>
2382 * ifcvt.c: Include vec.h, vecprim.h.
2383 (check_cond_move_block): New argument regs.
2384 Reorganize. Add registers used to regs.
2385 (cond_move_process_if_block): Use regs set above as
2388 2007-01-15 Eric Christopher <echristo@apple.com>
2390 * config/darwin.h: Update copyright.
2391 (TARGET_OPTION_TRANSLATE_TABLE): Add umbrella.
2392 (LINK_COMMAND_SPEC): Add -u.
2393 (LINK_SPEC): Fix umbrella for above.
2395 2007-01-15 Joseph S. Myers <joseph@codesourcery.com>
2397 * config/soft-fp/op-common.h, config/soft-fp/op-4.h: Update from
2400 2007-01-15 Tom Tromey <tromey@redhat.com>
2402 * doc/sourcebuild.texi (libgcj Tests): Don't mention jacks.
2403 * doc/install.texi (Testing): Don't mention jacks.
2404 (Configuration): Document --enable-java-maintainer-mode. Move
2405 --with-java-home to libgcj-specific section. Document
2407 (Prerequisites): Mention --enable-java-maintainer-mode, ecj1.
2409 2007-01-15 Jan Hubicka <jh@suse.cz>
2411 * tree-ssa-dce.c (DCE_TODOs): New.
2412 (propagate_necessity): Return if something changed.
2413 (eliminate_unnecessary_stmts): Likewise.
2414 (perform_tree_ssa_dce): Return TODO flags when needed.
2415 (pass_dce, pass_dce_loop, pass_cd_dce): Remove TODO flags.
2417 2007-01-15 Uros Bizjak <ubizjak@gmail.com>
2419 * config/i386/i386.md (fyl2xxf3_i387): Rename from fyl2x_xf3.
2420 (fyl2x_extend<mode>xf3_i387): New insn pattern.
2421 (log<mode>2): Rename from logsf2 and logdf2 and macroize insn
2422 patterns using X87MODEF12 mode macro. Extend operand 1
2423 to XFmode. Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
2424 (log10<mode>2): Ditto.
2425 (log2<mode>2): Ditto.
2426 (log1p<mode>2): Ditto.
2427 (logb<mode>2): Ditto.
2428 (fyl2xp1xf3_i387): Rename from fyl2xp1_xf3.
2429 (fyl2xp1_extend<mode>xf3_i387): New insn pattern.
2430 (*fxtractxf3_i387): Rename from *fxtractxf3.
2431 (fxtract_extend<mode>xf3_i387): New insn pattern.
2432 (ilogbsi2): Use match_dup 3, not match_operand:XF 3.
2433 * config/i386/i386.c (ix86_emit_i387_log1p): Use gen_fyl2xp1xf3_i387()
2434 and gen_fyl2xxf3_i387().
2436 2007-01-14 Zdenek Dvorak <dvorakz@suse.cz>
2438 * loop-unswitch.c (unswitch_loop): Do not call fix_loop_placement.
2439 * cfgloopmanip.c (fix_loop_placement): Made static. Use
2440 get_loop_exit_edges. Changed return type to bool.
2441 * cfgloop.h (fix_loop_placement): Declaration removed.
2443 2007-01-14 Dorit Nuzman <dorit@il.ibm.com>
2445 * param.h (MIN_VECT_LOOP_BOUND): New.
2446 * params.def (MIN_VECT_LOOP_BOUND): New.
2447 * tree-vectorizer.c (slpeel_tree_peel_loop_to_edge): Takes another
2448 argument - minimum threshold for number of iterations.
2449 * tree-vectorizer.h (slpeel_tree_peel_loop_to_edge): Add another
2450 argument to declaration.
2451 * tree-vect-analyze.c (vect_analyze_operations): Check value of
2452 MIN_VECT_LOOP_BOUND.
2453 * tree-vect-transform.c (vect_do_peeling_for_loop_bound): Call
2454 slpeel_tree_peel_loop_to_edge with additional argument.
2455 (vect_do_peeling_for_alignment): Likewise.
2456 * doc/invoke.texi (min-vect-loop-bound): Document new param option.
2458 2007-01-14 Uros Bizjak <ubizjak@gmail.com>
2461 * config/i386/i386.c (print_operand) ['z']: Output 'b' for
2464 2007-01-14 Jan Hubicka <jh@suse.cz>
2466 * tree-dfa.c (remove_referenced_var): New function.
2467 * tree-ssa-live.c (remove_unused_locals): Walk referenced vars and
2468 prune referenced vars list too.
2469 * tree-flow.h (remove_referenced_var): Declare.
2471 2007-01-14 Jan Hubicka <jh@suse.cz>
2473 * tree-eh.c (add_stmt_to_eh_region_fn): Do not add call_exprs
2475 (remove_stmt_from_eh_region_fn): Do not remove call_exprs.
2476 (verify_eh_throw_stmt_node, verify_eh_throw_table_statements): Kill.
2477 * except.h (verify_eh_throw_table_statements): Kill prototype.
2478 * cfgexpand.c (expand_gimple_basic_block): Propagate Eh regions
2479 into call exrepssions.
2480 * tree-optimize.c (execute_free_cfg_annotatiosn): Do not call
2482 * tree-cfg.c: Include pointer-set.h.
2483 (verify_node_sharing): Work on pointer set.
2484 (verify_eh_throw_stmt_node): New.
2485 (verify_stmts): Use pointers sets, verify throw_stmt.
2487 2007-01-13 Zdenek Dvorak <dvorakz@suse.cz>
2489 * ipa-reference.c (analyze_function): Consider also addresses taken
2492 2007-01-12 Roger Sayle <roger@eyesopen.com>
2494 * c-typeck.c (null_pointer_constant_p): Replace use of
2495 TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
2496 (build_c_cast): Likewise.
2498 2007-01-12 Roger Sayle <roger@eyesopen.com>
2500 * tree.h (force_fit_type_double): Remove unused final argument.
2501 * c-common.c (constant_expression_warning): Replace use of
2502 TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
2503 (convert_and_check): Likewise.
2504 (shorten_compare): Update call to force_fit_type_double.
2505 (c_common_truthvalue_conversion) <INTEGER_CST>: Use integer_zerop.
2506 * convert.c (convert_to_pointer): Update call to
2507 force_fit_type_double.
2508 * fold-const.c (force_fit_type_double): Remove overflowed_const
2510 (int_const_binop, fold_convert_const_int_from_int,
2511 fold_convert_const_int_from_real, fold_div_compare,
2512 fold_sign_changed_comparison, fold_unary, fold_negate_const,
2513 fold_abs_const, fold_not_const): Remove the final argument from
2514 calls to force_fit_type_double.
2516 2007-01-12 Andrew Pinski <andrew_pinski@playstation.sony.com>
2518 * configure.ac: Set insn to "nop" for spu-*-* also.
2519 * configure: Regenerate.
2521 2007-01-12 Olga Golovanevsky <olga@il.ibm.com>
2523 * builtins.def : Add BUILT_IN_FREE.
2525 2007-01-12 Jan Hubicka <jh@suse.cz>
2527 PR tree-optimization/30443
2528 * tree-inline.c (tree_function_versioning): Do not optimize when
2529 cloning for inlining.
2531 2007-01-12 Zdenek Dvorak <dvorakz@suse.cz>
2533 * doc/loop.texi: Document recording of loop exits.
2534 * cfgloopmanip.c (loopify, duplicate_loop): Use alloc_loop.
2535 (update_single_exits_after_duplication,
2536 update_single_exit_for_duplicated_loop,
2537 update_single_exit_for_duplicated_loops): Removed.
2538 (duplicate_loop_to_header_edge): Do not call
2539 update_single_exits_after_duplication and
2540 update_single_exit_for_duplicated_loops.
2541 (loop_version): Do not update single_exit information.
2542 (fix_loop_structure): Use record_loop_exits instead of
2543 mark_single_exit_loops.
2544 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update
2545 the lists of loop exits.
2546 * cfghooks.c (redirect_edge_and_branch, redirect_edge_and_branch_force,
2547 split_edge, merge_blocks): Update the lists of loop exits.
2548 * modulo-sched.c (sms_schedule): Pass LOOPS_HAVE_RECORDED_EXITS to
2549 loop_optimizer_init.
2550 * loop-init.c (loop_optimizer_init): Call record_loop_exits instead
2551 of mark_single_exit_loops.
2552 (loop_optimizer_finalize): Call release_recorded_exits.
2553 * tree-ssa-loop.c (tree_loop_optimizer_init): Pass
2554 LOOPS_HAVE_RECORDED_EXITS to loop_optimizer_init.
2555 * tree-vectorizer.c (slpeel_tree_duplicate_loop_to_edge_cfg): Do not
2556 update single exit information.
2557 * lambda-code.c (perfect_nestify): Ditto.
2558 * cfgloop.c (flow_loop_free): Destroy the list of exits of the loop.
2559 (mark_single_exit_loops): Removed.
2560 (alloc_loop, loop_exit_hash, loop_exit_eq, loop_exit_free,
2561 get_exit_descriptions, rescan_loop_exit, record_loop_exits,
2562 dump_recorded_exit, dump_recorded_exits, release_recorded_exits): New
2564 (get_loop_exit_edges, single_exit): Use recorded exit lists.
2565 (add_bb_to_loop, remove_bb_from_loops): Update the lists of loop exits.
2566 (verify_loop_structure): Verify consistency of the exit lists.
2567 (flow_loops_find): Use alloc_loop. Initialize exits hash.
2568 (set_single_exit): Removed.
2569 * cfgloop.h (struct loop_exit): New function.
2570 (struct loop): single_exit_ field replaced by exits field.
2571 (LOOPS_HAVE_MARKED_SINGLE_EXITS): Replaced by LOOPS_HAVE_RECORDED_EXITS.
2572 (struct loops): Added exits hash.
2573 (mark_single_exit_loops, set_single_exit): Declaration removed.
2574 (release_recorded_exits, record_loop_exits, rescan_loop_exit): Declare.
2576 2007-01-12 Richard Sandiford <richard@codesourcery.com>
2578 * doc/invoke.texi: Avoid use of @headitem.
2580 2007-01-12 Richard Sandiford <richard@codesourcery.com>
2582 * cse.c (cse_insn): Move HAVE_CC0 code after declarations.
2584 2007-01-12 Richard Sandiford <richard@codesourcery.com>
2586 * doc/install.texi: Fix m68k-*-* anchor and add m68k-*-* to the
2589 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
2590 Richard Sandiford <richard@codesourcery.com>
2592 * doc/invoke.texi: Document -mno-bitfield, -mno-rtd and -mno-short.
2593 * config/m68k/m68k.opt: Resort options.
2594 (mbitfield, mrtd, mshort): Remove RejectNegative properties.
2596 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
2597 Richard Sandiford <richard@codesourcery.com>
2599 * doc/invoke.texi: Document the macros that are defined by
2600 m68k's -mtune and -mhard-float options.
2601 * config/m68k/m68k-protos.h (m68k_cpp_cpu_ident) Declare.
2602 (m68k_cpp_cpu_family): Likewise.
2603 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Add a full set
2604 of __ucfv*__ macros. Define __mcffpu__ if generating code for
2605 ColdFire FPUs. Define __mcf_cpu_* and __mcf_family_* macros.
2606 * config/m68k/m68k.c (m68k_cpp_cpu_ident): New function.
2607 (m68k_cpp_cpu_family): Likewise.
2609 2007-01-12 Richard Sandiford <richard@codesourcery.com>
2611 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Treat all mc68*
2612 macros besides mc68000 as tuning macros. Use a switch statement
2613 to set them and mcpu32.
2615 2007-01-12 Julian Brown <julian@codesourcery.com>
2617 * config/m68k/m68k.h: Use TARGET_68040 instead of TARGET_68040_ONLY.
2618 (TARGET_68040_ONLY): Rename to...
2619 (TARGET_68040): ...this.
2620 * config/m68k/m68k.c: Use TARGET_68040 instead of TARGET_68040_ONLY.
2621 * config/m68k/m68k.md: Likewise.
2623 2007-01-12 Julian Brown <julian@codesourcery.com>
2624 Nathan Sidwell <nathan@codesourcery.com>
2625 Richard Sandiford <richard@codesourcery.com>
2627 * config.gcc (m680[012]0-*-*, m68k*-*-*): Set m68k_cpu_ident to
2628 the -mcpu= argument associated with the --with-cpu setting.
2629 Define M68K_DEFAULT_TUNE to the default -mtune= option,
2630 if different from the one implied by the -mcpu setting.
2631 Accept --with-cpu=FOO if FOO is listed in m68k-devices.def,
2632 using mcpu=FOO as the default CPU option. Set target_cpu_default2.
2633 * doc/invoke.texi: Mention ColdFire in the introduction to the
2634 m68k options. Document the new -march, -mcpu, -mtune, -mdiv,
2635 -mno-div and -mhard-float options. Make -m68881 a synonym for
2636 -mhard-float. Document the previously-undocumented -m5206e,
2637 -m528x, -m5307 and -m5407 options. Tweak the existing option
2638 documentation for consistency.
2639 * doc/install.texi: Mention new --with-cpu arguments.
2640 * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Only use the
2641 default CPU if neither -mcpu nor -march are specified.
2642 (ASM_CPU_SPEC): Pass down -mcpu and -march options.
2643 (TARGET_CPU_CPP_BUILTINS): Set __mcfisa*__ macros from
2644 TARGET_ISA*. Set the legacy __mcf*__ cpu macros in the same way,
2645 using m68k_tune to decide between families that implement the
2646 same ISA. Use m68k_tune to set __mcfv4e__.
2647 (FL_BITFIELD, FL_68881, FL_COLDFIRE, FL_CF_HWDIV, FL_CF_MAC)
2648 (FL_CF_EMAC, FL_CF_EMAC_B, FL_CF_USP, FL_CF_FPU, FL_ISA_68000)
2649 (FL_ISA_68010, FL_ISA_68020, FL_ISA_68040, FL_ISA_A, FL_ISA_B)
2650 (FL_ISA_C, FL_ISA_MMU): New macros.
2651 (MASK_COLDFIRE): Delete.
2652 (TARGET_68010, TARGET_68020, TARGET_68040_ONLY, TARGET_COLDFIRE)
2653 (TARGET_ISAB): Redefine in terms of m68k_cpu_flags.
2654 (TARGET_68881, TARGET_COLDFIRE_FPU): Redefine in terms of m68k_fpu.
2655 (TARGET_HARD_FLOAT): Do not define here.
2656 (TARGET_ISAAPLUS, TARGET_ISAC): New macros.
2657 (TUNE_68000): New macro.
2658 (TUNE_68000_10): Redefine in terms of TUNE_68000 and TUNE_68010.
2659 (TUNE_68010, TUNE_68030, TUNE_68040, TUNE_68060, TUNE_CPU32)
2660 (TUNE_CFV2): Redefine in terms of m68k_tune.
2661 (uarch_type, target_device, fpu_type): New enums.
2662 (m68k_cpu, m68k_tune, m68k_fpu, m68k_cpu_flags): Declare.
2663 * config/m68k/m68k.c (TARGET_DEFAULT): Remove MASK_68881.
2664 (FL_FOR_isa_00, FL_FOR_isa_10, FL_FOR_isa_20, FL_FOR_isa_40)
2665 (FL_FOR_isa_cpu32, FL_FOR_isa_a, FL_FOR_isa_aplus, FL_FOR_isa_b)
2666 (FL_FOR_isa_c): New macros.
2667 (m68k_isa): New enum.
2668 (m68k_target_selection): New structure.
2669 (all_devices, all_isas, all_microarchs): New tables.
2670 (m68k_cpu_entry, m68k_arch_entry, m68k_tune_entry, m68k_cpu)
2671 (m68k_tune, m68k_fpu, m68k_cpu_flags): New variables.
2672 (MASK_ALL_CPU_BITS): Delete.
2673 (m68k_find_selection): New function.
2674 (m68k_handle_option): Handle -mcpu=, -march= and -mtune=.
2675 Map the legacy target options to a combination of the new ones.
2676 (override_options): Set m68k_cpu, m68k_tune, m68k_fpu and
2677 m68k_cpu_flags. Handle M68K_DEFAULT_TUNE. Use m68k_cpu_flags
2678 to derive default MASK_BITFIELD, MASK_CF_HWDIV and MASK_HARD_FLOAT
2680 * config/m68k/m68k.opt (m5200, m5206e, m528x, m5307, m5407, mcfv4e)
2681 (m68010, m68020, m68020-40, m68020-60, m68030, m68040): Remove Mask
2683 (m68881, msoft-float): Change mask from 68881 to HARD_FLOAT.
2684 (march=, mcpu=, mdiv, mhard-float, mtune=): New options.
2685 * config/m68k/m68k-devices.def: New file.
2687 2007-01-12 Richard Sandiford <richard@codesourcery.com>
2688 Nathan Sidwell <nathan@codesourcery.com>
2690 * config/m68k/m68k.h (ASM_CPU_SPEC, ASM_SPEC, EXTRA_SPECS)
2691 (SUBTARGET_EXTRA_SPECS): New macros.
2692 * config/m68k/linux.h (ASM_SPEC): Remove CPU flags;
2693 use %(asm_cpu_spec) instead.
2694 * config/m68k/m68k-none.h (ASM_SPEC): Likewise.
2695 * config/m68k/openbsd.h (ASM_SPEC): Likewise.
2696 * config/m68k/netbsd-elf.h (ASM_SPEC): Likewise.
2697 (EXTRA_SPECS): Rename to...
2698 (SUBTARGET_EXTRA_SPECS): ...this.
2700 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
2701 Richard Sandiford <richard@codesourcery.com>
2702 Julian Brown <julian@codesourcery.com>
2704 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
2705 (m68k-*-uclinux*, m68k-*-linux*, m68k-*-rtems*): Set default_m68k_cpu
2706 to the configuration's default CPU.
2707 (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*): Likewise.
2708 Remove default masks.
2709 (m680[012]0-*-*): Set the default with_cpu to the first part of
2711 (m68k*-*-*): Set the default with_cpu to m$default_m68k_cpu.
2712 (m68k*-*-linux): Extend the --with-cpu handling to...
2713 (m680[012]0-*-*, m68k*-*-*): ...these configurations. Allow m68000
2714 and m68010. Don't set target_cpu_default2.
2715 * doc/install.texi: Document --with-cpu for m68k.
2716 * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Define.
2717 * config/m68k/m68k-none.h (TARGET_CPU_DEFAULT, M68K_CPU_m68k)
2718 (M68K_CPU_m68000, M68K_CPU_m68010, M68K_CPU_m68020, M68K_CPU_m68030)
2719 (M68K_CPU_m68040, M68K_CPU_m68302, M68K_CPU_m68332, TARGET_DEFAULT)
2720 (ASM_CPU_DEFAULT_SPEC, CC1_CPU_DEFAULT_SPEC): Delete.
2721 (ASM_SPEC): Remove use of %(asm_cpu_default).
2722 (EXTRA_SPECS, SUBTARGET_EXTRA_SPECS, MULTILIB_DEFAULTS): Delete.
2723 * config/m68k/linux.h (TARGET_DEFAULT): Delete.
2724 (CPP_SPEC): Merge definitions. Do not handle __HAVE_68881__ here.
2725 * config/m68k/netbsd-elf.h (TARGET_OS_CPP_BUILTINS): Define
2726 __HAVE_FPU__ if TARGET_HARD_FLOAT.
2727 (TARGET_DEFAULT): Delete.
2728 (EXTRA_SPECS): Delete cpp_cpu_default_spec, cpp_cpu_spec,
2729 cpp_fpu_spec, asm_default_spec and netbsd_cpp_spec.
2730 (CPP_CPU_SPEC): Delete.
2731 (TARGET_VERSION): Merge definitions, using TARGET_68010 to pick
2732 the appropriate string.
2733 (CPP_CPU_DEFAULT_SPEC, ASM_DEFAULT_SPEC, CPP_FPU_SPEC): Delete.
2734 (CPP_SPEC): Define to NETBSD_CPP_SPEC.
2735 (ASM_SPEC): Don't use %(asm_default_spec).
2736 * config/m68k/m68k.c (TARGET_DEFAULT_TARGET_FLAGS): Remove
2737 TARGET_DEFAULT and add MASK_68881.
2738 * config/m68k/m68k.md: Remove mention of TARGET_DEFAULT from comments.
2740 2007-01-12 Richard Sandiford <richard@codesourcery.com>
2742 * config.gcc (m68010-*-netbsdelf*): Add MASK_68010.
2743 (m68k*-*-netbsdelf*, m68k*-*-openbsd*, m68k*-linux*): Add
2744 MASK_68010 alongside MASK_68020.
2745 * doc/invoke.texi: Document -m68010.
2746 * config/m68k/m68k.opt (m68010): New.
2747 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Define mc68010
2749 (TUNE_68010): New macro.
2750 * config/m68k/m68k-none.h (M68K_CPU_m68k, M68K_CPU_m68010)
2751 (M68K_CPU_m68020, M68K_CPU_m68030, M68K_CPU_m68040)
2752 (M68K_CPU_m68332): Add MASK_68010.
2753 * config/m68k/linux.h (TARGET_DEFAULT): Add MASK_68010 to
2754 fallback definition.
2755 * config/m68k/netbsd-elf.h (CPP_CPU_SPEC): Remove now-redundant
2757 * config/m68k/m68k.c (MASK_ALL_CPU_BITS): Add MASK_68010.
2758 (m68k_handle_option): Handle OPT_m68010. Add MASK_68010
2759 to all entries that use MASK_68020.
2760 (output_move_simode_const, output_move_himode, output_move_qimode)
2761 (output_move_stricthi, output_move_strictqi): Use TARGET_68010
2762 instead of TARGET_68020 to select clr behavior. Remove comment
2763 about there being no TARGET_68010.
2764 * config/m68k/m68k.md: Likewise throughout.
2766 2007-01-12 Julian Brown <julian@codesourcery.com>
2768 * config/m68k/m68k.h (TARGET_ISAB): New macro.
2769 * config/m68k/m68k.c: Use TARGET_ISAB rather than TARGET_CFV4.
2770 * config/m68k/m68k.md: Likewise.
2772 2007-01-12 Julian Brown <julian@codesourcery.com>
2774 * config/m68k/m68k.h (LEGITIMATE_INDEX_P, LEGITIMIZE_ADDRESS): Use
2775 TARGET_COLDFIRE_FPU instead of TARGET_CFV4E.
2777 2007-01-12 Julian Brown <julian@codesourcery.com>
2779 * config/m68k/m68k.h (TUNE_68040_60): New macro.
2780 * config/m68k/m68k.c (standard_68881_constant_p): Use it.
2781 * config/m68k/m68k.md: Likewise.
2783 2007-01-12 Julian Brown <julian@codesourcery.com>
2784 Richard Sandiford <richard@codesourcery.com>
2786 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Use TUNE_68030
2787 instead of TARGET_68030, TUNE_68040 instead of TARGET_68040,
2788 TUNE_68060 instead of TARGET_68060 and TUNE_CPU32 instead of
2790 (TARGET_CPU32): Rename to...
2791 (TUNE_CPU32): ...this.
2792 (TUNE_68000_10, TUNE_68030, TUNE_68040, TUNE_68060)
2793 (TUNE_CFV2): New macros.
2794 * config/m68k/netbsd-elf.h (LONG_DOUBLE_TYPE_SIZE): Simplify;
2795 remove conditions that are implied by TARGET_68020.
2796 * config/m68k/m68k.c (m68k_output_function_prologue): Use TUNE_68040
2797 instead of TARGET_68040 and TUNE_CPU32 instead of TARGET_CPU32.
2798 (m68k_output_function_epilogue): Likewise.
2799 (m68k_rtx_costs): Likewise. Use TUNE_68060 instead of TARGET_68060
2800 and TUNE_CFV2 instead of TARGET_5200. Use TUNE_68000_10 instead of
2801 "!TARGET_68020 && !TARGET_COLDFIRE" to choose between 68000 and
2802 non-68000 timings. Refactor multiplication and division costs.
2803 (output_addsi3): Use TUNE_68040 instead of TARGET_68040 and
2804 TUNE_CPU32 instead of TARGET_CPU32.
2805 (standard_68881_constant_p): Use TUNE_68040 instead of TARGET_68040
2806 and TUNE_68060 instead of TARGET_68060.
2807 * config/m68k/m68k.md: Use TUNE_68040 instead of TARGET_68040,
2808 TUNE_68060 instead of TARGET_68060, and TUNE_CPU32 instead of
2810 (movsi_const0): Use TUNE_68000_10 rather than "!TARGET_68020
2811 && !TARGET_COLDFIRE" to choose between moveq and clr.
2812 Likewise in the unnamed movsf pattern.
2813 (ashlsi_17_24, lshrsi_17_24): Guard with TUNE_68000_10 rather than
2814 "!TARGET_68020 && !TARGET_COLDFIRE". Likewise the unnamed
2817 2007-01-12 Richard Sandiford <richard@codesourcery.com>
2819 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Increase amount
2820 of tabbing before backslashes.
2822 2007-01-11 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
2824 * pa-linux.h (ASM_OUTPUT_INTERNAL_LABEL): Undefine.
2825 * pa.h (ASM_OUTPUT_LABEL): Output colon when using GAS.
2826 (ASM_OUTPUT_INTERNAL_LABEL): Define.
2828 2007-01-11 Zdenek Dvorak <dvorakz@suse.cz>
2830 * tree-ssa-loop-ivopts.c (extract_cond_operands): Split from
2831 find_interesting_uses_cond.
2832 (find_interesting_uses_cond): Use extract_cond_operands.
2833 (rewrite_use_compare): Use extract_cond_operands and
2834 force_gimple_operand_bsi. Do not call update_stmt.
2835 (determine_use_iv_cost_condition): Use extract_cond_operands.
2836 Return cheaper of using original bound and changing the exit bound.
2838 2007-01-11 Zdenek Dvorak <dvorakz@suse.cz>
2840 PR tree-optimization/29516
2841 * tree-ssa-address.c (tree_mem_ref_addr, add_to_parts,
2842 most_expensive_mult_to_index, addr_to_parts,
2843 create_mem_ref, maybe_fold_tmr): Make the type of
2844 fields of TARGET_MEM_REF sizetype.
2845 (move_fixed_address_to_symbol, move_pointer_to_base):
2847 * tree.def (TARGET_MEM_REF): Add comment on types of
2850 2007-01-11 Joseph Myers <joseph@codesourcery.com>
2852 * c-common.c (vector_types_convertible_p): Treat opaque types as
2853 always convertible if they have the same size, but not otherwise.
2855 2007-01-11 Steven Bosscher <steven@gcc.gnu.org>
2857 * ifcvt.c (struct noce_if_info): Add comments to the fields.
2858 Remove the b_unconditional field.
2859 (noce_try_sign_mask): Do not look at b_unconditional.
2860 (noce_process_if_block): Do not use merge_if_blocks. Update
2861 the CFG here. Do not set b_unconditional.
2862 (cond_move_process_if_block): Likewise.
2863 (find_cond_trap): Likewise.
2864 (check_cond_move_block): Require simple jump insns at the end
2867 2007-01-11 Jan Hubicka <jh@suse.cz>
2869 PR tree-optimization/1046
2870 * tree-tailcall.c (suitable_for_tail_call_opt_p): Use TREE_ADDRESSABLE
2871 when alias info is not ready.
2872 (pass_tail_recursion): Do not require aliasing.
2873 * tree-ssa-copyrename.c (pass_rename_ssa_cop): Likewise.
2874 * tree-ssa-ccp.c (pass_ccp, pass_fold_builtins): Likewise.
2875 * tree-ssa-copy.c (pass_copy_prop): Likewise.
2876 * tree-ssa-forwprop.c (pass_forwprop): Likewise.
2877 * tree-ssa-dce.c (pass_dce, pass_dce_loop, pass_cd_dce): Likewise.
2878 * passes.c (init_optimization_passes): Execute rename_ssa_copies,
2879 ccp, forwprop, copy_prop, merge_phi, copy_prop, dce and tail recursion
2881 * tree-ssa-operands.c (add_virtual_operand, get_indirect_ref_operand):
2882 When aliasing is not build, mark statement as volatile.
2884 2007-01-11 Tom Tromey <tromey@redhat.com>
2886 PR preprocessor/15185, PR preprocessor/20989:
2887 * doc/cppopts.texi <-MT>: Update description of algorithm for
2888 computing default target.
2889 <-M, -MD>: Reword "basename" text.
2891 2007-01-11 Roger Sayle <roger@eyesopen.com>
2893 * builtins.c (expand_builtin_pow, expand_builtin_powi,
2894 fold_builtin_cabs, fold_builtin_sqrt, fold_builtin_trunc,
2895 fold_builtin_floor, fold_builtin_ceil, fold_builtin_round,
2896 fold_builtin_int_int_roundingfn, fold_builtin_bitop,
2897 fold_builtin_bswap, real_constp, fold_builtin_pow,
2898 fold_builtin_powi, fold_builtin_signbit, fold_builtin_copysign,
2899 do_mpfr_arg1, do_mpfr_arg2, do_mpfr_arg3, do_mpfr_sincos): Replace
2900 uses of the macro TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
2901 * convert.c (convert_to_pointer): Likewise.
2902 * expr.c (highest_pow2_factor, expand_expr_real_1): Likewise.
2903 * fold-const.c (force_fit_type, fold_negate_expr, int_const_binop,
2904 const_binop, fold_convert_const_int_from_int,
2905 fold_convert_const_int_from_real,
2906 fold_convert_const_real_from_real, sign_bit_p,
2907 optimize_minmax_comparison, extract_muldiv_1, fold_div_compare,
2908 fold_sign_changed_comparison, fold_unary, fold_comparison,
2909 fold_binary, multiple_of_p, tree_Expr_non_zero_p,
2910 fold_negate_const, fold_abs_const, fold_not_const): Likewise.
2911 * print-tree.c (print_node_brief, print_node): Likewise.
2912 * stor-layout.c (place_field, layout_type): Likewise.
2913 * tree-chrec.c (keep_cast): Likewise.
2914 * tree.c (build_vector, build_real, build_real_from_int_cst,
2915 build_complex): Likewise.
2917 2007-01-11 Roger Sayle <roger@eyesopen.com>
2919 * tree.h (TREE_CONSTANT_OVERFLOW): Obsolete. For the time being,
2920 treat TREE_CONSTANT_OVERFLOW as a synonym of TREE_OVERFLOW.
2922 2007-01-11 Paolo Bonzini <bonzini@gnu.org>
2924 * configure.ac (strict1_warn): Rename to strict_warn.
2925 (WERROR, --enable-werror, symlink hacks, stage1_cflags,
2926 cc_set_by_configure, quoted_cc_set_by_configure,
2927 stage_prefix_set_by_configure, quoted_stage_prefix_set_by_configure,
2928 all_boot_languages, all_stagestuff): Remove.
2929 (target_list): Remove bootstrap targets.
2930 * Makefile.in (quickstrap): Unconditionally make a synonym of all.
2931 (BOOT_LANGUAGES, STAGE1_CFLAGS, STAGE1_CHECKING,
2932 REMAKEFLAGS, FLAGS_TO_PASS, PREPEND_DOTDOT_TO_RELATIVE_PATHS,
2933 SUBDIR_FLAGS_TO_PASS, WERROR_FLAGS, STRICT2_WARN, LANG_STAGESTUFF,
2934 VOL_FILES, POSTSTAGE1_FLAGS_TO_PASS, STAGE2_FLAGS_TO_PASS,
2935 STAGEPROFILE_FLAGS_TO_PASS, STAGEFEEDBACK_FLAGS_TO_PASS, stage1_build,
2936 stage1_copy, stage2_build, stage2_copy, stageprofile_build,
2937 stageprofile_copy, stage3_build, stage3_copy, stagefeedback_build,
2938 stagefeedback_copy, stage4_build, clean_s1, clean_sw, bootstrap,
2939 bootstrap-lean, bootstrap2, bootstrap2-lean, bootstrap3,
2940 bootstrap3-lean, bootstrap4, bootstrap4-lean, unstage1, unstage2,
2941 unstage3, unstage4, unstageprofile, unstagefeedback, restage, restage2,
2942 restage3, restage4, restageprofile, restagefeedback, bubbleestrap,
2943 cleanstrap, unstrap, restrap, *compare, *compare3, *compare4,
2944 *compare-lean, *compare3-lean, *compare4-lean, stage1-start, stage1,
2945 stage2-start, stage2, stage3-start, stage3, stage4-start, stage4,
2946 stageprofile-start, stageprofile, stagefeedback-start, stagefeedback,
2947 risky-stage1, risky-stage2, risky-stage3, risky-stage4): Remove.
2948 (ORDINARY_FLAGS_TO_PASS): Rename to FLAGS_TO_PASS.
2949 (STAGECOPYSTUFF, STAGEMOVESTUFF): Consolidate into MOSTLYCLEANFILES.
2950 (mostlyclean): Adjust.
2951 (clean, distclean): Don't mention bootstrap stuff.
2952 * configure: Regenerate.
2953 * ada/config-lang.in, cp/config-lang.in, forttran/config-lang.in,
2954 java/config-lang.in, objc/config-lang.in, objcp/config-lang.in,
2955 treelang/config-lang.in (stagestuff): Remove.
2956 * doc/sourcebuild.texi (stage1, stage2, stage3, stage4,
2957 stageprofile, stagefeedback, stagestuff): Remove mention.
2959 2007-01-11 Nick Clifton <nickc@redhat.com>
2961 * config/mcore/predicates.md (mcore_general_movesrc_operand):
2963 (mcore_general_movdst_operand): Do not accept CONST_INTs.
2964 (mcore_arith_K_S_operand): Run the test for the S constraint not
2965 the test for the M constraint.
2966 (mcore_addsub_operand): Do not accept integer values that are
2967 larger than 32 bits.
2968 * config/mcore/mcore.md: Remove unused constraints from split.
2969 (andsi3): Use HOST_WIDE_INT instead of int to hold an INTVAL.
2971 (allocate_stack): Likewise.
2972 * config/mcore/mcore.c (mcore_print_operand): Restrict output of P
2973 operands to 32 bits.
2974 (mcore_const_costs): Use HOST_WIDE_INT instead of int to hold an
2976 (mcore_and_cost, mcore_modify_comparison, const_ok_for_mcore,
2977 mcore_const_ok_for_inline, mcore_const_trick_uses_not,
2978 try_constant_tricks, mcore_num_ones, mcore_num_zeros,
2979 mcore_output_bclri, mcore_output_andn, output_inline_const,
2980 mcore_output_move, mcore_output_movedouble): Likewise.
2981 (mcore_output_cmov): Use CONST_OK_FOR_M and CONST_OK_FOR_N.
2982 (output_inline_const): Likewise.
2983 (output_inline_const): Fix format strings used in sprintf
2985 * config/mcore/mcore-protos.h: Update prototypes for changed
2986 functions in mcore.c.
2987 * config/mcore/mcore.h (CONST_OK_FOR_I): Cast values to
2988 HOST_WIDE_INT and not int.
2989 (CONST_OK_FOR_J, CONST_OK_FOR_K, CONST_OK_FOR_L, CONST_OK_FOR_M,
2990 CONST_OK_FOR_N): Likewise.
2991 (LEGITIMATE_CONSTANT_P): Also check CONSTANT_P.
2992 (GO_IF_LEGITIMATE_INDEX): Use HOST_WIDE_INT instead of int to hold
2995 2007-01-10 Jan Hubicka <jh@suse.cz>
2997 * tree-vrp.c (remove_range_assertions): Release defs.
2998 * tree-ssa-loop-ivopts.c (rmeove_statement): Likewise.
2999 * tree-ssa-dom.c (remove_stmt_or_phi): Likewise.
3001 2007-01-10 Paul Brook <paul@codesourcery.com>
3003 * config/arm/arm.c (arm_rtx_costs_1): Handle mutiply-subtract.
3004 * config/arm/arm.md (mulsi3subsi): New insn.
3006 2007-01-10 Zdenek Dvorak <dvorakz@suse.cz>
3008 * tree-ssa-loop-manip.c (tree_unroll_loop): Make it a wrapper over ...
3009 (tree_transform_and_unroll_loop): New.
3010 * tree-flow.h (transform_callback, tree_transform_and_unroll_loop):
3013 2007-01-10 Robert Kennedy <jimbob@google.com>
3015 * fold-const.c (fold_comparison): Fold comparisons like (x *
3016 1000 < 0) to (x < 0).
3018 2007-01-10 Ian Lance Taylor <iant@google.com>
3020 * tree-pretty-print.c (dump_generic_node): Print parentheses when
3021 operands have the same priority.
3023 2007-01-10 Tom Tromey <tromey@redhat.com>
3025 * fold-const.c (fold_truthop): Don't check can_use_bit_fields_p.
3026 (fold_binary): Likewise.
3027 * langhooks.c (lhd_can_use_bit_fields_p): Removed.
3028 * langhooks-def.h (lhd_can_use_bit_fields_p): Removed.
3029 (LANG_HOOKS_CAN_USE_BIT_FIELDS_P): Removed.
3030 (LANG_HOOKS_INITIALIZER): Remove LANG_HOOKS_CAN_USE_BIT_FIELDS_P.
3031 * langhooks.h (struct lang_hooks): Removed field
3032 'can_use_bit_fields_p'.
3034 2007-01-10 Ralf Corsépius <ralf.corsepius@rtems.org>
3036 * config/bfin/t-bfin, config/bfin/t-bfin-elf: Remove GCC_CFLAGS.
3038 2007-01-10 Razya Ladelsky <razya@il.ibm.com>
3040 * function.c (get_last_funcdef_no): New function.
3041 * function.h (get_last_funcdef_no): Declare.
3042 * tree-inline.c (initialize_cfun): Add initialization.
3043 (tree_function_versioning): Cleanup.
3045 2007-01-10 Jan Hubicka <jh@suse.cz>
3047 * tree-inline.c (setup_one_parameter): Do not propagate into abnormal
3050 2007-01-10 Sa Liu <saliu@de.ibm.com>
3051 Ben Elliston <bje@au.ibm.com>
3053 * spu.h (STACK_SAVE_AREA): Use VOIDmode for SAVE_FUNCTION, SImode
3054 for SAVE_NONLOCAL and Pmode for any other save level.
3055 * spu-protos.h (spu_restore_stack_block): Declare.
3056 * spu.md (save_stack_block): Remove.
3057 (restore_stack_block): Call spu_restore_stack_block.
3058 * spu.c (spu_restore_stack_block): New function.
3059 (spu_expand_epilogue): Remove old comment.
3061 2007-01-09 Zdenek Dvorak <dvorakz@suse.cz>
3063 PR tree-optimization/30322
3064 * tree-ssa-loop-ivopts.c (fold_affine_expr, iv_value): Removed.
3065 (cand_value_at): Return the value as aff_tree.
3066 (may_eliminate_iv): Convert the bound from aff_tree to tree.
3067 * tree-affine.c (aff_combination_add_cst, aff_combination_add_product,
3068 aff_combination_mult): New functions.
3069 (aff_combination_add): Use aff_combination_add_cst.
3070 (aff_combination_convert): Allow conversions to a wider type.
3071 (tree_to_aff_combination): Handle BIT_NOT_EXPR.
3072 * tree-affine.h (aff_combination_mult): Declare.
3074 2007-01-09 Carlos O'Donell <carlos@codesourcery.com>
3076 * doc/tm.texi: Update documentation to reflect reality of exec
3077 and start file search behaviours. Update copyright year.
3078 * doc/invoke.texi: Explain how GCC_EXEC_PREFIX is used to find
3079 header file directories.
3081 2007-01-09 Uros Bizjak <ubizjak@gmail.com>
3083 * config/i386/i386.md (*sinxf2): Rename to *sinxf2_i387.
3084 (*cosxf2): Rename to cosxf2_i387.
3085 (*sindf2, *sinsf2): Extend operand 1 to XFmode. Macroize patterns
3086 using X87MODEF12 mode macro. Rename patterns to
3087 *sin_extend<mode>xf2_i387. Use SSE_FLOAT_MODE_P to disable patterns
3089 (*cosdf2, *cossf2): Ditto.
3090 (sincosdf3, sincossf3): Ditto. Rewrite corresponding splitters
3091 to match extended input operands.
3092 (sincos<mode>3): New expander.
3093 (*sinextendsfdf2, *cosextendsfdf2, *sincosextendsfdf3): Remove
3094 insn patterns and corresponding splitters.
3096 2007-01-09 Kaz Kojima <kkojima@gcc.gnu.org>
3098 * config/sh/t-linux (TARGET_LIBGCC2_CFLAGS): Delete.
3099 (SHLIB_MAPFILES, SHLIB_LINK, SHLIB_INSTALL): Likewise.
3101 2007-01-09 Nicolas Pitre <nico@cam.org>
3104 * arm/ieee754-df.S (Lad_s): Also test the low word of X for zero.
3106 2007-01-08 Geoffrey Keating <geoffk@apple.com>
3108 * target.h (struct gcc_target): New field library_rtti_comdat.
3109 * target-def.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): New.
3110 (TARGET_CXX): Add TARGET_CXX_LIBRARY_RTTI_COMDAT.
3111 * doc/tm.texi (C++ ABI): Document TARGET_CXX_LIBRARY_RTTI_COMDAT.
3112 * config/darwin.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
3114 2007-01-08 Geoffrey Keating <geoffk@apple.com>
3116 * doc/invoke.texi (Optimize Options): Correct description of -O0.
3118 2007-01-08 Richard Guenther <rguenther@suse.de>
3120 * tree.h (force_fit_type_double): Export.
3121 (force_fit_type): Remove.
3122 * fold-const.c (force_fit_type_double): New function.
3123 (force_fit_type): Remove.
3124 (int_const_binop): Use it.
3125 (fold_convert_const_int_from_int): Likewise.
3126 (fold_convert_const_int_from_real): Likewise.
3127 (fold_div_compare): Likewise.
3128 (fold_sign_changed_comparison): Likewise.
3129 (fold_unary): Likewise.
3130 (fold_negate_const): Likewise.
3131 (fold_abs_const): Likewise.
3132 (fold_not_const): Likewise.
3133 * c-common.c (shorten_compare): Use force_fit_type_double.
3134 * convert.c (convert_to_pointer): Likewise.
3136 2007-01-08 Richard Guenther <rguenther@suse.de>
3138 * tree.h (build_int_cst_wide_type): Export.
3139 * tree.c (build_int_cst_wide_type): New function.
3140 (build_int_cst_wide): Fix comment.
3141 * builtins.c (fold_builtin_object_size): Use build_int_cst
3142 to build -1 or 0 of the correct type. Use fit_double_type
3143 to check for overflow.
3144 * fold-const.c (optimize_bit_field_compare): Use build_int_cst_type
3146 (decode_field_reference): Likewise.
3147 (all_ones_mask_p): Likewise.
3148 (native_interpret_int): Use build_int_cst_wide_type.
3149 (fold_binary): Use build_int_cst_type to build an all-ones
3151 * stor-layout.c (set_sizetype): Use build_int_cst_wide_type.
3153 2007-01-08 Daniel Jacobowitz <dan@codesourcery.com>
3155 * config/pa/t-pa64 (libgcc_stub.a): Use $(T).
3157 2007-01-09 Ben Elliston <bje@au.ibm.com>
3159 * genautomata.c (STATS_OPTION): New option.
3160 (stats_flag): New flag.
3161 (gen_automata_option): Handle it.
3162 (initiate_automaton_gen): Ditto.
3163 (write_automata): Output statistics only if stats_flag is
3164 set. Likewise, output time statistics only if time_flag is set.
3165 * doc/md.texi (Processor pipeline description): Document new flag.
3167 2007-01-08 Richard Guenther <rguenther@suse.de>
3169 * builtins.c (fold_builtin_int_roundingfn): Use fit_double_type.
3170 * tree.c (build_int_cst_type): Likewise.
3171 (size_in_bytes): Don't call force_fit_type on the result.
3172 (int_fits_type_p): Use fit_double_type.
3173 * fold-const.c (fit_double_type): New function.
3174 (force_fit_type): Use it.
3175 * tree.h (fit_double_type): Export.
3177 2007-01-08 Jan Hubicka <jh@suse.cz>
3179 * tree-vectorizer.c (gate_increase_alignment): Fix return type.
3180 * ipa.c (function_and_variable_visibility): Fix return type.
3182 2007-01-08 Richard Guenther <rguenther@suse.de>
3184 * tree-ssa-ccp.c (maybe_fold_offset_to_array_ref): Use type
3185 of offset to build the index.
3186 * tree-pretty-print.c (dump_generic_node): Don't build negated
3187 const just for printing.
3188 * c-pretty-print.c (pp_c_integer_constant): Likewise.
3189 * builtins.c (fold_builtin_int_roundingfn): Check if result
3190 fits the type by using force_fit_type and comparing the result.
3191 * predict.c (predict_loops): Use compare_tree_int for comparison.
3192 * tree.c (build_int_cst): Fall back to integer_type_node for
3194 (build_int_cst_wide): Assert type is non-null.
3196 2007-01-08 Roberto Costa <roberto.costa@st.com>
3198 * tree-vrp.c (extract_range_from_cond_expr): New.
3199 (extract_range_from_expr): Handle COND_EXPR nodes used as expressions.
3200 * tree-ssa-ccp.c (get_maxval_strlen): Handle COND_EXPR nodes used
3202 (fold_stmt): Bug fix, avoid infinite recursion when folding COND_EXPRs.
3203 * tree-ssa-forwprop.c (simplify_cond, forward_propagate_into_cond,
3204 tree_ssa_forward_propagate_single_use_vars): Handle COND_EXPR nodes
3205 used as expressions.
3206 * tree-object-size.c (cond_expr_object_size): New.
3207 (collect_object_sizes_for): Handle COND_EXPR nodes used as expressions.
3209 2007-01-08 Jan Hubicka <jh@suse.cz>
3211 * tree-ssa-forwprop.c (forward_propagate_into_cond,
3212 tree_ssa_forward_propagate_single_use_va): Release defs of propagated
3215 2007-01-08 Richard Guenther <rguenther@suse.de>
3217 PR tree-optimization/23603
3218 * tree-vrp.c (set_value_range_to_truthvalue): New function.
3219 (extract_range_from_binary): Fall back to truthvalue instead of
3220 varying for TRUTH_*_EXPR.
3221 (extract_range_from_comparison): Fall back to truthvalue instead of
3223 (vrp_visit_phi_node): Don't adjust new range bounds to +INF/-INF
3224 if all visited PHI values were constant.
3226 2007-01-08 Jan Hubicka <jh@suse.cz>
3228 * cgraphunit.c (cgraph_process_new_functions): Reset reachable flag.
3229 (cgraph_analyze_function): break out from ...
3230 (cgraph_finalize_compilation_unit): ... here.
3231 (cgraph_expand_function): Remove forgoten commented out line.
3232 (cgraph_optimize): Analyze functions.
3234 2007-01-08 Jan Hubicka <jh@suse.cz>
3236 * tree-pas.h (TODO_remove_function): New flag.
3237 (TODO_update*): Renumber.
3238 (pass_ipa_increase_alignment,
3239 pass_ipa_function_and_variable_visibility): New passes.
3240 * cgraphunit.c (cgraph_increase_alignment): Move to tree-vectorizer.c
3241 (cgraph_function_and_variable_visibility): Move to ipa.c
3242 (cgraph_optimize): Don't call cgraph_function_and_variable_visibility,
3243 cgraph_increase_alignment.
3244 * ipa-inline.c (cgraph_decide_inlining): Don't push timevar.
3245 (cgraph_decide_inlining_incrementally): Push TV_INTEGRATION before
3246 calling tree-inline.
3247 (cgraph_early_inlining): Do not call cgraph_remove_unreachable_nodes.
3248 (pass_ipa_inline, pass_early_ipa_inlining): Set TODO_remove_functions
3249 * tree-vectorizer.c (increase_alignment): Move here from cgraphunit.c
3250 (gate_increase_alignment): New function.
3251 (pass_ipa_increase_alignment): New pass.
3252 * ipa.c: Inline tree-pass.h and timevar.h
3253 (function_and_variable_visibility): Move here from cgraphunit.c
3254 * tree-optimize.c (pass_early_local_passes): Add TODO_remove_functions.
3255 * passes.c (init_optimization_passes): Add the two new passes.
3256 (execute_todo): Handle cgraph_remove_functions.
3258 2007-01-08 Nick Clifton <nickc@redhat.com>
3260 * config/frv/predicates.md (reg_or_0_operand): Accept
3263 2007-01-08 Ralf Corsépius <ralf.corsepius@rtems.org>
3265 * config/bfin/rtems.h, config/bfin/t-rtems: New.
3266 * config.gcc: Add bfin*-rtems*.
3268 2007-01-08 Mark Shinwell <shinwell@codesourcery.com>
3270 * c.opt: Add -flax-vector-conversions.
3271 * c-typeck.c (convert_for_assignment): Pass flag to
3272 vector_types_convertible_p to allow emission of note.
3273 (digest_init): Likewise.
3274 * c-opts.c: Handle -flax-vector-conversions.
3275 * c-common.c (flag_lax_vector_conversions): New.
3276 (vector_types_convertible_p): Unless -flax-vector conversions
3277 has been passed, disallow conversions between vectors with
3278 differing numbers of subparts and/or element types. If such
3279 a conversion is disallowed, possibly emit a note on the first
3280 occasion only to inform the user of -flax-vector-conversions.
3281 The new last argument specifies this.
3282 * c-common.h (flag_lax_vector_conversions): New.
3283 (vector_types_convertible_p): Add extra argument.
3284 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Use
3285 char_type_node for V*QI type vectors.
3286 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins):
3287 Update to satisfy new typechecking rules.
3288 * config/rs6000/altivec.h (vec_cmple): Use vec_cmpge, for both
3290 * doc/invoke.texi (C Dialect Options): Document
3291 -flax-vector-conversions.
3293 2007-01-08 Mark Shinwell <shinwell@codesourcery.com>
3295 PR tree-optimization/29877
3296 * tree-ssa-ter.c (is_replaceable_p): Deem assignments with
3297 a register variable on the RHS to not be replaceable.
3299 2007-01-08 Chen Liqin <liqin@sunnorth.com.cn>
3300 * config/score/t-score-elf (MULTILIB_OPTIONS): Change.
3301 * config/score/predicates.md (const_uimm5, sr0_operand, const_simm12,
3302 const_simm15, const_pow2, const_npow2): Added.
3303 * config/score/misc.md (insv, extv, extzv, movmemsi,
3304 move_lbu_a/b, mov_lhu_a/b etc): Added and fix some bug.
3305 * config/score/score.c (score_address_cost, score_select_cc_mode):
3307 Change CONST_OK_FOR_LETTER_P/EXTRA_CONSTRAINT define.
3308 Update score_rtx_costs for MACRO TARGET_RTX_COSTS.
3309 Update score_print_operand.
3310 * config/score/score.h (DATA_ALIGNMENT, SELECT_CC_MODE): Added.
3311 Adjust register allocate order and update some macro define.
3312 * config/score/score-mdaux.c (mdx_unaligned_load, mdx_unsigned_store,
3313 mdx_block_move_straight, mdx_block_move_loop_head,
3314 mdx_block_move_loop_body, mdx_block_move_loop_foot, mdx_block_move_loop,
3315 mdx_block_move): Added.
3316 (mdx_movsicc, mdp_select_add_imm, mdp_select, mds_zero_extract_andi,
3317 mdp_limm): Updated and fix some bug and typo.
3318 * config/score/score.md (movqi/hi/si, add/sub/zero/ext): Updated.
3319 (movsf, movdf, doloop_end): Added.
3321 2007-01-08 Kazu Hirata <kazu@codesourcery.com>
3323 * config/arm/arm.c, config/arm/arm.h, config/arm/arm.md,
3324 config/arm/thumb2.md: Fix comment typos.
3325 * doc/extend.texi: Fix a typo.
3327 2007-01-07 Eric Christopher <echristo@apple.com>
3329 * configure.ac: Check for __stack_chk_fail for darwin.
3330 * configure: Regenerate.
3332 2007-01-07 Richard Guenther <rguenther@suse.de>
3334 * tree-vrp.c (extract_range_from_assert): CSE calls to
3335 compare_values where possible.
3336 (extract_range_from_unary_expr): Likewise.
3338 2007-01-07 Anatoly Sokolov <aesok@post.ru>
3340 * config/avr/avr-protos.h (call_insn_operand): Delete prototype.
3341 * config/avr/avr.c (call_insn_operand): Delete function.
3342 * config/avr/avr.md (*pushqi, *pushhi, *pushsi, *pushsf): Use REG_SP
3343 instead of register number. Use predicates.
3344 * config/avr/predicates.md (const0_operand, reg_or_0_operand,
3345 call_insn_operand): Add.
3347 2007-01-06 Jan Hubicka <jh@suse.cz>
3349 * tree-pass.h (pass_build_cgraph_edges): Declare.
3350 * cgraphunit.c (record_refernece): Move to cgraphbuild.c
3351 (visited_nodes): Remove.
3352 (cgraph_create_edges): Move to cgraphbuild.c; rename to
3353 build_cgrpah_edges; make visited_nodes local.
3354 (cgraph_process_new_functions): DO not call initialize_inline_failed.
3355 (record_references_in_initializer): Move to cgraphbuild.c
3356 (initialize_inline_failed, rebuild_cgraph_edges,
3357 pass_rebuild_cgraph_edges): Move to cgraphbuild.c.
3358 (verify_cgraph_node): Make visited_nodes local.
3359 (cgraph_analyze_function): Do not call cgraph_create_edges and
3360 initialize_inline_failed.
3361 (cgraph_expand_function): Do not call cgraph_lower_function;
3362 assert that function is already lowered.
3363 * Makefile.in (cgraphbuild.o): New.
3364 * passes.c (init_optimization_passes): Add pass_build_cgraph_edges
3365 at the end of lowering passes.
3367 2007-01-06 Steven Bosscher <steven@gcc.gnu.org>
3369 * ifcvt.c (cond_move_convert_if_block): New function, code
3370 factored out from...
3371 (cond_move_process_if_block): ...here. Call the new function
3372 on the THEN and ELSE blocks.
3373 (merge_if_block): Do not copy global_live_at_end, merge_blocks
3374 already takes care of this.
3376 2007-01-05 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
3379 * tree.h (TREE_OVERFLOW_P): New.
3380 * c-typeck.c (parser_build_unary_op): Warn only if result
3381 overflowed and operands did not.
3382 (parser_build_binary_op): Likewise.
3383 (convert_for_assignment): Remove redundant overflow_warning.
3384 * c-common.c (overflow_warning): Don't check or set TREE_OVERFLOW.
3386 2007-01-05 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
3388 * c-typeck.c (store_init_value): Split over two lines to follow
3389 the GNU coding style.
3391 2007-01-05 Benjamin Kosnik <bkoz@redhat.com>
3393 * c-cppbuiltin.c (c_cpp_builtins): __GXX_EXPERIMENTAL_CPP0X__ to
3394 __GXX_EXPERIMENTAL_CXX0X__.
3395 * doc/cpp.texi: Same.
3397 2007-01-05 Richard Guenther <rguenther@suse.de>
3400 * tree.c (get_narrower): Do not construct COMPONENT_REFs
3401 with mismatched types. Instead explicitly build a
3402 conversion NOP_EXPR.
3404 2007-01-05 Ian Lance Taylor <iant@google.com>
3406 * c-common.c (decl_with_nonnull_addr_p): New function.
3407 (c_common_truthvalue_conversion): Call it.
3408 * c-typeck.c (build_binary_op): Likewise.
3409 * c-common.h (decl_with_nonnull_addr_p): Declare.
3411 2007-01-05 Jakub Jelinek <jakub@redhat.com>
3414 * libgcc2.c (__divdc3): Compare c and d against 0.0 instead of
3417 2007-01-05 Joel Brobecker <brobecker@adacore.com>
3419 * doc/install.texi (Final install): Document the fact that
3420 the GNAT runtime should not be stripped.
3422 2007-01-04 Jan Hubicka <jh@suse.cz>
3424 * tree-inline.c (fold_marked_statements): Update operand caches
3425 and EH after folding
3427 2007-01-04 Ian Lance Taylor <iant@google.com>
3429 * c-common.c (check_function_nonnull): Whitespace fix.
3431 2007-01-04 Jan Hubicka <jh@suse.cz>
3433 * tree-optimize.c (execute_fixup_cfg): Correct previously mistakely
3434 comitted older version of patch.
3435 (pass_fixup_cfg): Add TODOs to verify flow and statements, dump
3436 function, celanup cfg and collect garbage.
3438 2007-01-04 Mike Stump <mrs@apple.com>
3440 * Makefile.in (mostlyclean): Don't remove libgcc anymore.
3443 2007-01-04 Eric Christopher <echristo@apple.com>
3445 * libgcc2.c (__bswapsi2): Use SItype.
3446 (__bswapdi2): Use DItype.
3447 * libgcc2.h: Update for above.
3449 2007-01-04 Paul Brook <paul@codesourcery.com>
3451 * config/arm/arm.md (arm_mulsi3, thumb_mulsi3, mulsi3_compare0,
3452 mulsi_compare0_scratch, mulsi3addsi, mulsi3addsi_compare0,
3453 mulsi3addsi_compare0_scratch, mulsidi3adddi, mulsidi3,
3454 umulsidi3, umulsidi3adddi, smulsi3_highpart,
3455 umulsi3_highpart): Make conditional on !arm_arch6.
3456 (arm_mulsi3_v6, thumb_mulsi3_v6, mulsi3_compare0_v6,
3457 mulsi_compare0_scratch_v6, mulsi3addsi_v6, mulsi3addsi_compare0_v6,
3458 mulsi3addsi_compare0_scratch_v6, mulsidi3adddi_v6, mulsidi3_v6,
3459 umulsidi3_v6, umulsidi3adddi_v6, smulsi3_highpart_v6,
3460 umulsi3_highpart_v6): New insns.
3462 2007-01-04 Roger Sayle <roger@eyesopen.com>
3464 * fold-const.c (fold_convert): When casting an expression to void,
3465 fold_ignored_result may discover a GIMPLE_MODIFY_STMT which doesn't
3466 have a type. Instead of attempting to build a NOP_EXPR, return
3467 these "special" trees directly.
3469 2007-01-04 Joseph Myers <joseph@codesourcery.com>
3471 * config/rs6000/rs6000.c (rs6000_rtx_costs): Make adjustment for
3472 MULT inside MINUS as either argument. Use rs6000_cost->dmul -
3473 rs6000_cost->fp not 0 as adjustment for outer NEG.
3475 2007-01-04 Jan Hubicka <jh@suse.cz>
3477 * cgraph.c (cgraph_release_function_body): New function.
3478 (cgraph_remove_node): Use it.
3479 * cgraph.h (cgraph_release_function_body): Declare.
3480 * cgraphunit.c (cgraph_expand_function): Use it.
3481 * ipa.c (cgraph_remove_unreahchable_nodes): Use it.
3482 * tree-ssa.c (delete_tree_ssa): Allow to be called before aliasing
3483 is initialized and while compilation of other function is running.
3484 * tree-optimize.c (execute_free_cfg_annotations): Move code to clear
3485 statement CFG annotations from here to ...
3486 * tree-cfg.c (delete_tree_cfg_annotations): ... here.
3488 2007-01-04 Zdenek Dvorak <dvorakz@suse.cz>
3490 * cfgloop.h (enum li_flags): Make the constants powers of two.
3492 2007-01-04 Jan Hubicka <jh@suse.cz>
3494 * tree-inline.c (copy_bb): Insert new statements to statements_to_fold
3496 (fold_marked_statements): New function.
3497 (optimize_inline_calls, tree_function_versioning): Fold new statements.
3498 * tree-inline.h (copy_body_data): Add statements_to_fold.
3500 2007-01-03 Daniel Jacobowitz <dan@codesourcery.com>
3502 * config.gcc: Mention libgcc/config.host.
3503 * Makefile.in: Update comments mentioning libgcc.
3504 (LIBGCC, INSTALL_LIBGCC, GCC_PARTS, mklibgcc): Delete.
3505 (all.cross, start.encap, rest.encap, rest.cross): Update
3506 dependencies for libgcc move.
3507 (libgcc.mk, LIBGCC_DEPS, libgcov.a, libgcc.a, stmp-multilib)
3508 (clean-target, clean-target-libgcc): Delete.
3509 (srcdirify, GCC_EXTRA_PARTS): New macros.
3510 (libgcc-support, libgcc.mvars): New rules.
3511 (distclean): Remove mention of mklibgcc.
3512 (install): Don't reference INSTALL_LIBGCC.
3513 (install-common): Don't reference EXTRA_PARTS.
3514 (install-libgcc, install-multilib): Delete rules.
3515 * mklibgcc.in: Delete file.
3516 * doc/configfiles.texi: Don't mention mklibgcc.
3518 * config/i386/t-darwin (SHLIB_VERPFX): Delete (moved to libgcc).
3519 * config/i386/t-darwin64 (SHLIB_VERPFX): Likewise.
3520 * config/rs6000/t-darwin (SHLIB_VERPFX): Likewise.
3521 * config/rs6000/t-ppccomm (TARGET_LIBGCC2_CFLAGS, SHLIB_MAPFILES)
3522 (mklibgcc, ldblspecs): Likewise.
3524 * config/i386/t-nwld (libgcc.def, libc.def, libpcre.def)
3525 (posixpre.def): Use $(T).
3526 (SHLIB_EXT, SHLIB_NAME, SHLIB_SLIBDIR_QUAL, SHLIB_DEF, SHLIB_MAP)
3527 (SHLIB_SRC, SHLIB_INSTALL): Delete.
3528 (SHLIB_LINK): Make dummy.
3529 * config/t-slibgcc-darwin: Delete contents except for dummy SHLIB_LINK.
3531 * config/frv/t-linux (EXTRA_MULTILIB_PARTS): Clear.
3533 * config/alpha/t-crtfm: Use $(T) in rules for EXTRA_PARTS.
3534 * config/alpha/t-vms, config/alpha/t-vms64, config/fr30/t-fr30,
3535 config/i386/t-rtems-i386, config/ia64/t-ia64, config/rs6000/t-beos,
3536 config/rs6000/t-newas, config/sparc/t-elf: Likewise.
3538 * configure.ac (all_outputs): Remove mklibgcc.
3539 * configure: Regenerated.
3541 2007-01-03 Josh Conner <jconner@apple.com>
3544 * calls.c (compute_argument_addresses): Set stack and stack_slot
3545 for partial args, too.
3546 (store_one_arg): Use locate.size.constant for the size when
3547 generating a save_area.
3549 2007-01-03 Robert Kennedy <jimbob@google.com>
3551 * tree-cfg.c (tree_merge_blocks): Release SSA_NAME phi results
3552 whose definitions are deleted due to basic block merging.
3554 2007-01-03 Paul Brook <paul@codesourcery.com>
3557 * config/arm/arm.c (output_return_instruction): Pop PC in interrupt
3559 (use_return_insn): Return 0 for Thumb interrupt functions.
3560 (print_multi_reg): Add rfe argument for IRQ returns.
3561 (arm_output_epilogue): Pop interrupt return address directly into PC.
3562 (arm_expand_prologue): Only adjust IRQ return address in Arm mode.
3564 2007-01-03 Paul Brook <paul@codesourcery.com>
3566 Merge from sourcerygxx-4_1.
3567 * config/arm/thumb2.md: New file.
3568 * config/arm/elf.h (JUMP_TABLES_IN_TEXT_SECTION): Return True for
3570 * config/arm/coff.h (JUMP_TABLES_IN_TEXT_SECTION): Ditto.
3571 * config/arm/aout.h (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
3572 (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump tables.
3573 * config/arm/aof.h (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump
3575 (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
3576 * config/arm/ieee754-df.S: Use macros for Thumb-2/Unified asm
3578 * config/arm/ieee754-sf.S: Ditto.
3579 * config/arm/arm.c (thumb_base_register_rtx_p): Rename...
3580 (thumb1_base_register_rtx_p): ... to this.
3581 (thumb_index_register_rtx_p): Rename...
3582 (thumb1_index_register_rtx_p): ... to this.
3583 (thumb_output_function_prologue): Rename...
3584 (thumb1_output_function_prologue): ... to this.
3585 (thumb_legitimate_address_p): Rename...
3586 (thumb1_legitimate_address_p): ... to this.
3587 (thumb_rtx_costs): Rename...
3588 (thumb1_rtx_costs): ... to this.
3589 (thumb_compute_save_reg_mask): Rename...
3590 (thumb1_compute_save_reg_mask): ... to this.
3591 (thumb_final_prescan_insn): Rename...
3592 (thumb1_final_prescan_insn): ... to this.
3593 (thumb_expand_epilogue): Rename...
3594 (thumb1_expand_epilogue): ... to this.
3595 (arm_unwind_emit_stm): Rename...
3596 (arm_unwind_emit_sequence): ... to this.
3597 (thumb2_legitimate_index_p, thumb2_legitimate_address_p,
3598 thumb1_compute_save_reg_mask, arm_dwarf_handle_frame_unspec,
3599 thumb2_index_mul_operand, output_move_vfp, arm_shift_nmem,
3600 arm_save_coproc_regs, thumb_set_frame_pointer, arm_print_condition,
3601 thumb2_final_prescan_insn, thumb2_asm_output_opcode, arm_output_shift,
3602 thumb2_output_casesi): New functions.
3603 (TARGET_DWARF_HANDLE_FRAME_UNSPEC): Define.
3604 (FL_THUMB2, FL_NOTM, FL_DIV, FL_FOR_ARCH6T2, FL_FOR_ARCH7,
3605 FL_FOR_ARCH7A, FL_FOR_ARCH7R, FL_FOR_ARCH7M, ARM_LSL_NAME,
3606 THUMB2_WORK_REGS): Define.
3607 (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv, arm_condexec_count,
3608 arm_condexec_mask, arm_condexec_masklen)): New variables.
3609 (all_architectures): Add armv6t2, armv7, armv7a, armv7r and armv7m.
3610 (arm_override_options): Check new CPU capabilities.
3611 Set new architecture flag variables.
3612 (arm_isr_value): Handle v7m interrupt functions.
3613 (user_return_insn): Return 0 for v7m interrupt functions. Handle
3615 (const_ok_for_arm): Handle Thumb-2 constants.
3616 (arm_gen_constant): Ditto. Use movw when available.
3617 (arm_function_ok_for_sibcall): Return false for v7m interrupt
3619 (legitimize_pic_address, arm_call_tls_get_addr): Handle Thumb-2.
3620 (thumb_find_work_register, arm_load_pic_register,
3621 legitimize_tls_address, arm_address_cost, load_multiple_sequence,
3622 emit_ldm_seq, emit_stm_seq, arm_select_cc_mode, get_jump_table_size,
3623 print_multi_reg, output_mov_long_double_fpa_from_arm,
3624 output_mov_long_double_arm_from_fpa, output_mov_double_fpa_from_arm,
3625 output_mov_double_fpa_from_arm, output_move_double,
3626 arm_compute_save_reg_mask, arm_compute_save_reg0_reg12_mask,
3627 output_return_instruction, arm_output_function_prologue,
3628 arm_output_epilogue, arm_get_frame_offsets, arm_regno_class,
3629 arm_output_mi_thunk, thumb_set_return_address): Ditto.
3630 (arm_expand_prologue): Handle Thumb-2. Use arm_save_coproc_regs.
3631 (arm_coproc_mem_operand): Allow POST_INC/PRE_DEC.
3632 (arithmetic_instr, shift_op): Use arm_shift_nmem.
3633 (arm_print_operand): Use arm_print_condition. Handle '(', ')', '.',
3635 (arm_final_prescan_insn): Use extract_constrain_insn_cached.
3636 (thumb_expand_prologue): Use thumb_set_frame_pointer.
3637 (arm_file_start): Output directive for unified syntax.
3638 (arm_unwind_emit_set): Handle stack alignment instruction.
3639 * config/arm/lib1funcs.asm: Remove default for __ARM_ARCH__.
3640 Add v6t2, v7, v7a, v7r and v7m.
3641 (RETLDM): Add Thumb-2 code.
3642 (do_it, shift1, do_push, do_pop, COND, THUMB_SYNTAX): New macros.
3643 * config/arm/arm.h (TARGET_CPU_CPP_BUILTINS): Define __thumb2__.
3644 (TARGET_THUMB1, TARGET_32BIT, TARGET_THUMB2, TARGET_DSP_MULTIPLY,
3645 TARGET_INT_SIMD, TARGET_UNIFIED_ASM, ARM_FT_STACKALIGN, IS_STACKALIGN,
3646 THUMB2_TRAMPOLINE_TEMPLATE, TRAMPOLINE_ADJUST_ADDRESS,
3647 ASM_OUTPUT_OPCODE, THUMB2_GO_IF_LEGITIMATE_ADDRESS,
3648 THUMB2_LEGITIMIZE_ADDRESS, CASE_VECTOR_PC_RELATIVE,
3649 CASE_VECTOR_SHORTEN_MODE, ADDR_VEC_ALIGN, ASM_OUTPUT_CASE_END,
3650 ADJUST_INSN_LENGTH): Define.
3651 (TARGET_REALLY_IWMMXT, TARGET_IWMMXT_ABI, CONDITIONAL_REGISTER_USAGE,
3652 STATIC_CHAIN_REGNUM, HARD_REGNO_NREGS, INDEX_REG_CLASS,
3653 BASE_REG_CLASS, MODE_BASE_REG_CLASS, SMALL_REGISTER_CLASSES,
3654 PREFERRED_RELOAD_CLASS, SECONDARY_OUTPUT_RELOAD_CLASS,
3655 SECONDARY_INPUT_RELOAD_CLASS, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P,
3656 TRAMPOLINE_SIZE, INITIALIZE_TRAMPOLINE, HAVE_PRE_INCREMENT,
3657 HAVE_POST_DECREMENT, HAVE_PRE_DECREMENT, HAVE_PRE_MODIFY_DISP,
3658 HAVE_POST_MODIFY_DISP, HAVE_PRE_MODIFY_REG, HAVE_POST_MODIFY_REG,
3659 REGNO_MODE_OK_FOR_BASE_P, LEGITIMATE_CONSTANT_P,
3660 REG_MODE_OK_FOR_BASE_P, REG_OK_FOR_INDEX_P, GO_IF_LEGITIMATE_ADDRESS,
3661 LEGITIMIZE_ADDRESS, THUMB2_LEGITIMIZE_ADDRESS,
3662 GO_IF_MODE_DEPENDENT_ADDRESS, MEMORY_MOVE_COST, BRANCH_COST,
3663 ASM_APP_OFF, ASM_OUTPUT_CASE_LABEL, ARM_DECLARE_FUNCTION_NAME,
3664 FINAL_PRESCAN_INSN, PRINT_OPERAND_PUNCT_VALID_P,
3665 PRINT_OPERAND_ADDRESS): Adjust for Thumb-2.
3666 (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv): New declarations.
3667 * config/arm/arm-cores.def: Add arm1156t2-s, cortex-a8, cortex-r4 and
3669 * config/arm/arm-tune.md: Regenerate.
3670 * config/arm/arm-protos.h: Update prototypes.
3671 * config/arm/vfp.md: Enable patterns for Thumb-2.
3672 (arm_movsi_vfp): Add movw alternative. Use output_move_vfp.
3673 (arm_movdi_vfp, movsf_vfp, movdf_vfp): Use output_move_vfp.
3674 (thumb2_movsi_vfp, thumb2_movdi_vfp, thumb2_movsf_vfp,
3675 thumb2_movdf_vfp, thumb2_movsfcc_vfp, thumb2_movdfcc_vfp): New.
3676 * config/arm/libunwind.S: Add Thumb-2 code.
3677 * config/arm/constraints.md: Update include Thumb-2.
3678 * config/arm/ieee754-sf.S: Add Thumb-2/Unified asm support.
3679 * config/arm/ieee754-df.S: Ditto.
3680 * config/arm/bpabi.S: Ditto.
3681 * config/arm/t-arm (MD_INCLUDES): Add thumb2.md.
3682 * config/arm/predicates.md (low_register_operand,
3683 low_reg_or_int_operand, thumb_16bit_operator): New.
3684 (thumb_cmp_operand, thumb_cmpneg_operand): Rename...
3685 (thumb1_cmp_operand, thumb1_cmpneg_operand): ... to this.
3686 * config/arm/t-arm-elf: Add armv7 multilib.
3687 * config/arm/arm.md: Update patterns for Thumb-2 and Unified asm.
3689 (UNSPEC_STACK_ALIGN, ce_count): New.
3690 (arm_incscc, arm_decscc, arm_umaxsi3, arm_uminsi3,
3691 arm_zero_extendsidi2, arm_zero_extendqidi2): New
3693 * config/arm/fpa.md: Update patterns for Thumb-2 and Unified asm.
3694 (thumb2_movsf_fpa, thumb2_movdf_fpa, thumb2_movxf_fpa,
3695 thumb2_movsfcc_fpa, thumb2_movdfcc_fpa): New insns.
3696 * config/arm/cirrus.md: Update patterns for Thumb-2 and Unified asm.
3697 (cirrus_thumb2_movdi, cirrus_thumb2_movsi_insn,
3698 thumb2_cirrus_movsf_hard_insn, thumb2_cirrus_movdf_hard_insn): New
3700 * doc/extend.texi: Document ARMv7-M interrupt functions.
3701 * doc/invoke.texi: Document Thumb-2 new cores+architectures.
3703 2007-01-03 Jakub Jelinek <jakub@redhat.com>
3705 * unwind-dw2.c (SIGNAL_FRAME_BIT, EXTENDED_CONTEXT_BIT): Define.
3706 (struct _Unwind_Context): Rename args_size to flags, remove
3707 signal_frame field, add a new args_size field and version field.
3708 (_Unwind_IsSignalFrame, _Unwind_SetSignalFrame,
3709 _Unwind_IsExtendedContext): New inline functions.
3710 (_Unwind_GetGR, _Unwind_SetGR, _Unwind_GetGRPtr, _Unwind_SetGRPtr):
3711 Assume by_value array is only present if _Unwind_IsExtendedContext.
3712 (_Unwind_GetIPInfo, execute_cfa_program, uw_frame_state_for): Use
3713 _Unwind_IsSignalFrame.
3714 (__frame_state_for): Initialize context.flags to EXTENDED_CONTEXT_BIT.
3715 (uw_update_context_1): Use _Unwind_SetSignalFrame.
3716 (uw_init_context_1): Initialize context->flags to
3717 EXTENDED_CONTEXT_BIT.
3718 * config/rs6000/linux-unwind.h (frob_update_context): Use
3719 _Unwind_SetSignalFrame.
3721 2007-01-03 Andrew Pinski <andrew_pinski@playstation.sony.com>
3724 * gimplify.c (gimplify_modify_expr_complex_part): Move below
3725 tree_to_gimple_tuple. Call tree_to_gimple_tuple when we need
3728 2007-01-03 Kazu Hirata <kazu@codesourcery.com>
3730 * config/i386/i386.h (NON_STACK_REG_P, REGNO_OK_FOR_SIREG_P,
3731 REGNO_OK_FOR_DIREG_P, REWRITE_ADDRESS, ASM_OPERAND_LETTER,
3732 RET, AT_SP): Remove.
3733 * config/i386/i386.md (*sse_prologue_save_insn): Use return
3736 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
3737 * cfgbuild.c (find_basic_blocks): Likewise.
3738 * cfgrtl.c (rtl_create_basic_block): Likewise.
3739 * function.c (temp_slots_at_level): Likewise.
3740 * reg-stack.c (stack_regs_mentioned): Likewise.
3741 * regclass.c (allocate_reg_info): Likewise.
3742 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
3743 set_bb_for_stmt, move_block_to_fn): Likewise.
3744 * tree-complex.c (tree_lower_complex): Likewise.
3745 * vec.h (VEC_safe_grow_cleared): New.
3747 * cgraphunit.c, tree-ssa-alias.c: Fix comment typos.
3749 2007-01-03 Zdenek Dvorak <dvorakz@suse.cz>
3751 * loop-unswitch.c (unswitch_loop): Pass probabilities to loopify.
3752 * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Pass probabilities
3754 * cfgloopmanip.c (scale_loop_frequencies): Export.
3755 (loopify): Scale the frequencies by prescribed coefficients.
3756 (set_zero_probability): New function.
3757 (duplicate_loop_to_header_edge): Improve updating of frequencies.
3758 (lv_adjust_loop_entry_edge, loop_version): Set probabilities
3759 and frequencies according to arguments.
3760 * tree-ssa-loop-manip.c (tree_unroll_loop): Set probabilities
3762 * cfg.c (scale_bbs_frequencies_int): Allow scaling the frequencies up.
3763 * modulo-sched.c (sms_schedule): Set probabilities for entering
3764 versioned loop correctly.
3765 * tree-vect-transform.c (vect_transform_loop): Ditto.
3766 * cfgloop.h (loopify, loop_version): Declaration changed.
3767 (scale_loop_frequencies): Declared.
3769 2007-01-02 Jan Hubicka <jh@suse.cz>
3771 * cgraph.c: Include tree-flow.h
3772 (cgraph_add_new-function): Handle IPA_SSA mode; execute
3774 * cgraph.h (enum cgraph_state): Add CGRAPH_STATE_IPA_SSA.
3775 * tree-pass.h (pass_all_early_optimizations): Declare.
3776 * cgraphunit.c (cgraph_process_new_functions): Add IPA_SSA; execute
3778 (cgraph_analyze_function): Do early_local_passes.
3779 * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
3780 Do not add referenced vars.
3781 * tree-optimize.c (gate_all_optimizations): Do not execute when not in
3783 (gate_all_early_local_passes): New gate.
3784 (pass_early_local_passes): Use new gate.
3785 (execute_early_local_optimizations): New functions.
3786 (gate_all_early_optimizations): New gate.
3787 (pass_all_early_optimizations): New pass.
3788 (execute_free_datastructures): Free SSA only when initialized.
3789 (gate_init_datastructures): Init only when optimizing.
3790 (tree_lowering_passes): Do early local passes when called late.
3791 * tree-profile.c (do_tree_profiling): Don't profile functions added
3793 (do_early_tree_profiling, pass_early_tree_profile): Kill.
3794 * tree-cfg.c (update_modified_stmts): Do not update when operands are
3796 * passes.c (init_optimizations_passes): Reorder so we go into SSA
3797 during early_local_passes.
3798 * Makefile.in (cgraph.o): Add dependency on tree-flow.h.
3801 2007-01-02 Carlos O'Donell <carlos@codesourcery.com>
3803 * Makefile.in: Update copyright year.
3805 2007-01-02 Carlos O'Donell <carlos@codesourcery.com>
3807 * Makefile.in: Export GCC_EXEC_PREFIX before calling $(RUNTEST)
3808 in $(lang_checks) and check-consistency targets.
3810 2007-01-02 Jan Hubicka <jh@suse.cz>
3812 * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
3813 Do not add referenced vars.
3814 * tree-cfg.c (update_modified_stmts): Do not update when SSA operands
3816 * passes.c (init_optimization_passes): Put mudflap_2 after
3817 free_datastructures.
3819 2007-01-02 Jan Hubicka <jh@suse.cz>
3821 * tree-optimize (execute_fixup_cfg): Set after_inlining flag.
3822 Set NOTHROW flag on call statements proved to be nothrow.
3823 Update statement of local calls so new pure/const functions are
3824 updated. Update_ssa when in ssa form. Mark PHI nodes of nonlocal
3826 (tree_rest_of_compilation): Register hooks and initialize bitmap
3827 early. Do not set after_inlining flag.
3829 2007-01-02 Steve Ellcey <sje@cup.hp.com>
3831 * sbitmap.c (HOST_BITS_PER_LONG_LONG): Change to
3832 HOST_BITS_PER_LONGLONG
3834 2007-01-02 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
3837 * c-typeck.c (store_init_value): Don't emit pedantic overflow
3838 warning for non-static initializers.
3840 2007-01-02 Steven Bosscher <steven@gcc.gnu.org>
3842 * config/alpha/alpha.md, arm/arm.c, darwin.c, frv/frv.md,
3843 m32r/m32r.c, m32r/m32r.c, mn10300/mn10300.md, pa/pa.c,
3844 rs6000/rs6000.c, s390/s390.md, sh/sh.md, sparc/sparc.c:
3845 Always use set_unique_reg_note to add REG_EQUAL notes.
3847 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
3850 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
3852 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
3853 * cfgbuild.c (find_basic_blocks): Likewise.
3854 * cfgrtl.c (rtl_create_basic_block): Likewise.
3855 * function.c (temp_slots_at_level): Likewise.
3856 * reg-stack.c (stack_regs_mentioned): Likewise.
3857 * regclass.c (allocate_reg_info): Likewise.
3858 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
3859 set_bb_for_stmt, move_block_to_fn): Likewise.
3860 * tree-complex.c (tree_lower_complex): Likewise.
3861 * vec.h (VEC_safe_grow_cleared): New.
3863 2007-01-02 Ian Lance Taylor <iant@google.com>
3865 * c-common.c (c_common_truthvalue_conversion): When warning about
3866 using an assignment as a truth value, set TREE_NO_WARNING.
3868 2007-01-02 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
3871 * c.opt (Wold-style-declaration): New.
3872 * doc/invoke.texi (C-only Warning Options): New.
3873 (Wold-style-declaration): Document it.
3874 (Wextra): Enabled by -Wextra.
3875 * c-opts.c (c_common_post_options): Enabled by -Wextra.
3876 * c-decl.c (declspecs_add_scspec): Replace -Wextra with
3877 -Wold-style-declaration.
3879 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
3881 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
3882 * cfgbuild.c (find_basic_blocks): Likewise.
3883 * cfgrtl.c (rtl_create_basic_block): Likewise.
3884 * function.c (temp_slots_at_level): Likewise.
3885 * reg-stack.c (stack_regs_mentioned): Likewise.
3886 * regclass.c (allocate_reg_info): Likewise.
3887 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
3888 set_bb_for_stmt, move_block_to_fn): Likewise.
3889 * tree-complex.c (tree_lower_complex): Likewise.
3890 * vec.h (VEC_safe_grow_cleared): New.
3892 2007-01-02 Douglas Gregor <doug.gregor@gmail.com>
3894 * c-common.c (c_common_nodes_and_builtins): Since variants of
3895 void_type_node get built before it is given a name, we need to
3896 give those variants the name, too.
3897 (complete_array_type): We need to work with the canonical main
3898 type of the array, from which we will build the qualified version.
3899 * params.def (PARAM_VERIFY_CANONICAL_TYPES): New.
3900 * print-tree.c (print_node): Display canonical type information
3902 * stor-layout.c (layout_type): When we don't know the
3903 alignment of a type for which we're building an array, we end up
3904 guessing wrong, so make the type require structural equality.
3905 * tree.c (make_node_stat): When we build a new type, it is its
3907 (build_type_attribute_qual_variant): When building an attribute
3908 variant, its canonical type is the non-attribute variant. However,
3909 if the attributes are target-dependent and they differ, we need to
3910 use structural equality checks for this type.
3911 (build_qualified_type): A qualified type is not equivalent to its
3912 unqualified variant; set the canonical type appropriately.
3913 (build_distinct_type_copy): When building a distinct type from
3914 another type, the new type is its own canonical type.
3915 (build_variant_type_copy): When building a new type variant, we
3916 assume that it is equivalent to the original type.
3917 (build_pointer_type_for_mode): When building a pointer type, also
3918 build a canonical type pointer.
3919 (build_reference_type_for_mode): When building a reference type,
3920 also build a canonical type reference.
3921 (build_index_type): When we can't hash an index type (e.g.,
3922 because its maximum value is negative), the index type requires
3923 structural equality tests.
3924 (build_array_type): Build the canonical form of an array type.
3925 (build_function_type): Function types require structural equality,
3926 because they contain default arguments, attributes, etc.
3927 (build_method_type_directly): Ditto for method types.
3928 (build_offset_type): Build the canonical offset type.
3929 (build_complex_type): Build the canonical vector type.
3930 (make_vector_type): Build the canonical vector type.
3931 * tree.h (TYPE_CANONICAL): New.
3932 (TYPE_STRUCTURAL_EQUALITY_P): New.
3933 (SET_TYPE_STRUCTURAL_EQUALITY): New.
3934 (struct tree_type): Added "canonical" field.
3935 * params.h (VERIFY_CANONICAL_TYPES): New.
3936 * doc/c-tree.texi (TYPE_CANONICAL): Document.
3937 (TYPE_STRUCTURAL_EQUALITY_P): Document.
3938 (SET_TYPE_STRUCTURAL_EQUALITY): Document.
3939 * doc/invoke.texi (verify-canonical-types): Document --param
3940 parameter for verifying canonical types.
3942 2007-01-02 Joseph Myers <joseph@codesourcery.com>
3944 * config.gcc (powerpc-*-eabispe*, powerpc-*-eabisimaltivec*,
3945 powerpc-*-eabisim*, powerpc-*-eabialtivec*, powerpc-*-eabi*,
3946 powerpc-*-rtems*, powerpc-wrs-vxworks, powerpc-wrs-vxworksae,
3947 powerpcle-*-eabisim*, powerpcle-*-eabi*): Add rs6000/e500.h to
3949 * config/rs6000/e500.h: New.
3950 * config/rs6000/eabi.h (TARGET_SPE_ABI, TARGET_SPE, TARGET_E500,
3951 TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE, TARGET_E500_DOUBLE):
3953 * config/rs6000/linuxspe.h (TARGET_SPE_ABI, TARGET_SPE,
3954 TARGET_E500, TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE,
3955 TARGET_E500_DOUBLE): Remove.
3956 * config/rs6000/vxworks.h (TARGET_SPE_ABI, TARGET_SPE,
3957 TARGET_E500, TARGET_ISEL, TARGET_FPRS): Remove.
3958 * config/rs6000/rs6000.h (CHECK_E500_OPTIONS): Define.
3959 * config/rs6000/rs6000.c (rs6000_override_options): Use
3962 2007-01-02 Joseph Myers <joseph@codesourcery.com>
3964 * config/rs6000/rs6000.c (print_operand): Check (TARGET_SPE ||
3965 TARGET_E500_DOUBLE), not TARGET_E500, for %y.
3966 (rs6000_generate_compare, rs6000_emit_sCOND, output_cbranch,
3967 rs6000_emit_cmove): Don't check TARGET_E500.
3968 * config/rs6000/rs6000.md (bunordered, bordered, sunordered,
3969 sordered): Don't check TARGET_E500.
3971 2007-01-01 Eric Christopher <echristo@apple.com>
3973 * config/mips/mips.c (mips_regno_mode_ok_for_base_p): Use
3974 HARD_REGISTER_NUM_P.
3976 2007-01-01 Roger Sayle <roger@eyesopen.com>
3978 * fold-const.c (fold_binary) <EQ_EXPR>: Fold "(X^C1) eq/ne C2" into
3979 "X eq/ne (C1^C2)". Fold "(X^Z) eq/ne (Y^Z)" as "X eq/ne Y" when Z
3980 has no side-effects. Fold "(X^C1) eq/ne (Y^C2)" as "(X^(C1^C2))
3983 2007-01-01 Mike Stump <mrs@apple.com>
3985 * configure.ac: Remove support for building with Apple's gcc-3.1.
3987 2007-01-02 Joseph Myers <joseph@codesourcery.com>
3990 * caller-save.c (add_stored_regs): Only handle SUBREGs if inner
3991 REG is a hard register. Do not modify REG before calling
3993 * rtlanal.c (subreg_get_info): Don't assert size of XMODE is a
3994 multiple of the size of YMODE for certain lowpart cases.
3996 2007-01-01 Andrew Pinski <pinskia@gmail.com>
3999 * gimplify (voidify_wrapper_expr): Update for
4002 2007-01-01 Andreas Schwab <schwab@suse.de>
4005 * config/ia64/ia64.c (ia64_compute_frame_size): Account space for
4006 save of BR0 in extra_spill_size instead of spill_size.
4007 (ia64_expand_prologue): Save BR0 outside of the gr/br/fr spill
4009 (ia64_expand_epilogue): Restore BR0 from its new location.
4011 2007-01-01 Andrew Pinski <pinskia@gmail.com>
4013 * gimplify.c (gimplify_init_constructor <case VECTOR_TYPE>):
4014 Use a temporary variable if the left hand side is not a gimple
4017 2007-01-01 Andrew Pinski <pinskia@gmail.com>
4019 * gimplify.c (gimplify_return_expr): Make the temporary variable
4020 for the return expression, a gimple register variable.
4022 2007-01-01 Jan Hubicka <jh@suse.cz>
4024 * emit-rtl.c (emit_copy_of_insn_after): Do not call copy_insn_1 for
4027 2007-01-01 Mike Stump <mrs@apple.com>
4029 * configure.ac (HAVE_GAS_LITERAL16): Add autoconf check for
4031 * config/darwin.c (machopic_select_rtx_section): Use
4033 (darwin_mergeable_constant_section): Likewise.
4034 * configure: Regenerate.
4035 * config.in: Regenerate.
4037 2007-01-01 Jan Hubicka <jh@suse.cz>
4038 Andrew Pinski <pinskia@gmail.com>
4040 * cgraphunit.c (cgraph_optimize): Call cgraph_add_new_functions
4041 before starting IPA passes.