Makefile.in (passes.o, [...]): Add dependencies.
[gcc.git] / gcc / ChangeLog
1 2007-02-09 Jan Hubicka <jh@suse.cz>
2
3 * Makefile.in (passes.o, ipa-inline.o): Add dependencies.
4 * cgraphbuild.c (build_cgraph_edges): Compute frequencies.
5 (rebuild_cgraph_edges): Likewise.
6 * cgraph.c (cgraph_set_call_stmt): Add new argument frequency.
7 (dump_cgraph_node): Dump frequencies.
8 (cgraph_clone_edge): Add frequency scales.
9 (cgraph_clone_node): Add freuqnecy.
10 * cgraph.h (cgraph_edge): Add freuqnecy argument.
11 (CGRAPH_FREQ_BASE, CGRAPH_FREQ_MAX): New constants.
12 (cgraph_create_edge, cgraph_clone_edge, cgraph_clone_node): Update.
13 * tree-pass.h (TODO_rebuild_frequencies): New constant.
14 * cgraphunit.c (verify_cgraph_node): Verify frequencies.
15 (cgraph_copy_node_for_versioning): Update call of cgraph_clone_edge.
16 (save_inline_function_body): Likewise.
17 * ipa-inline.c: inluce rtl.h
18 (cgraph_clone_inlined_nods): Update call of cgraph_clone_node.
19 (cgraph_edge_badness): Use frequencies.
20 (cgraph_decide_recursive_inlining): Update clonning.
21 (cgraph_decide_inlining_of_small_function): Dump frequency.
22 * predict.c (estimate_bb_frequencies): Export.
23 * predict.h (estimate_bb_frequencies): Declare.
24 * tree-inline.c (copy_bb): Watch overflows.
25 (expand_call_inline): Update call of cgraph_create_edge.
26 (optimize_inline_calls): Use TODO flags to update frequnecies.
27 * passes.h: Include predict.h
28 (init_optimization_passes): Move profile ahead.
29 (execute_function_todo): Handle TODO_rebuild_frequencies.
30
31 2007-02-09 Roger Sayle <roger@eyesopen.com>
32
33 * config/alpha/alpha.c (emit_insxl): Force the first operand of
34 the insbl or inswl pattern into a register.
35
36 2007-02-09 Roger Sayle <roger@eyesopen.com>
37
38 * config/ia64/ia64.md (bswapdi2): New define_insn.
39
40 2007-02-09 Richard Henderson <rth@redhat.com>
41
42 * config/i386/constraints.md (Ym): New constraint.
43 * config/i386/i386.md (movsi_1): Change Y2 to Yi constraints.
44 (movdi_1_rex64): Split sse and xmm general register moves from
45 memory move alternatives. Use conditional register constraints.
46 (movsf_1, movdf_integer): Likewise.
47 (zero_extendsidi2_32, zero_extendsidi2_rex64): Likewise.
48 (movdf_integer_rex64): New.
49 (pushsf_rex64): Fix output constraints.
50 * config/i386/sse.md (sse2_loadld): Split rm alternative, use Yi.
51 (sse2_stored): Likewise.
52 (sse2_storeq_rex64): New.
53 * config/i386/i386.c (x86_inter_unit_moves): Enable for not
54 amd and not generic.
55 (ix86_secondary_memory_needed): Don't bypass TARGET_INTER_UNIT_MOVES
56 for optimize_size. Remove SF/DFmode hack.
57
58 2007-02-09 Dwarakanath Rajagopal <dwarak.rajagopal@amd.com>
59
60 * config/i386/driver-i386.c: Turn on -mtune=native for AMDFAM10.
61 (bit_SSE4a): New.
62
63 2007-02-09 Nathan Sidwell <nathan@codesourcery.com>
64 Richard Sandiford <richard@codesourcery.com>
65
66 * config.gcc (m68010-*-netbsdelf*, m68k*-*-netbsdelf*)
67 (m68k*-*-openbsd*, m68k-*-linux*): Set default_cf_cpu.
68 (m68k-*-aout*, m68k-*-coff*, m68k-*-uclinux*, m68k-*-rtems*): Add
69 m68k/t-mlib to tmake_file.
70 (m68020-*-elf*, m68k-*-elf*): Likewise. Add t-m68kbare as well.
71 (m68k*-*-*): Use --with-arch to pick a default for --with-cpu.
72 (m680[012]0-*-*, m68k*-*-*): Add support for --with-arch.
73 Allow it to be cf or m68k. Set m68k_arch_family. If that
74 variable is not empty, add t-$m68k_arch_family to tmake_file.
75 Add t-mlibs to tmake_file.
76 * doc/install.texi: Document --with-arch=m68k and --with-arch=cf.
77 * config/m68k/t-cf: New file.
78 * config/m68k/t-m68k: Likewise.
79 * config/m68k/t-mlibs: Likewise.
80 * config/m68k/t-m68kbare (MULTILIB_OPTIONS, MULTILIB_DIRNAMES)
81 (MULTILIB_MATCHES, MULTILIB_EXCEPTIONS): Delete.
82 (M68K_MLIB_DIRNAMES, M68K_MLIB_OPTIONS): Define.
83 * config/m68k/t-m68kelf (MULTILIB_OPTIONS, MULTILIB_DIRNAMES)
84 (MULTILIB_MATCHES, MULTILIB_EXCEPTIONS, LIBGCC, INSTALL_LIBGCC):
85 Delete.
86 * config/m68k/t-openbsd (MULTILIB_OPTIONS, LIBGCC): Delete.
87 (INSTALL_LIBGCC): Delete.
88 (M68K_MLIB_DIRNAMES, M68K_MLIB_OPTIONS): Define.
89 * config/m68k/t-rtems (MULTILIB_OPTIONS, MULTILIB_DIRNAMES)
90 (MULTILIB_MATCHES, MULTILIB_EXCEPTIONS): Delete.
91 (M68K_MLIB_CPU): Define.
92 * config/m68k/t-uclinux (MULTILIB_OPTIONS, MULTILIB_DIRNAMES)
93 (MULTILIB_MATCHES, MULTILIB_EXCEPTIONS): Delete.
94 (M68K_MLIB_CPU, M68K_MLIB_OPTIONS, M68K_MLIB_DIRNAMES): Define.
95
96 2007-02-09 Zdenek Dvorak <dvorakz@suse.cz>
97 Richard Guenther <rguenther@suse.de>
98
99 PR middle-end/23361
100 * fold-const.c (fold_comparison): Handle obfuscated comparisons
101 against INT_MIN/INT_MAX.
102 * tree-ssa-loop-ivcanon.c (remove_empty_loop): Print to dump
103 file if a loop is removed.
104
105 2007-02-09 Joseph Myers <joseph@codesourcery.com>
106
107 * calls.c (store_one_arg): Pass correct alignment to
108 emit_push_insn for non-BLKmode values.
109 * expr.c (emit_push_insn): If STRICT_ALIGNMENT, copy to an
110 unaligned stack slot via a suitably aligned slot.
111
112 2007-02-08 DJ Delorie <dj@redhat.com>
113
114 * config/m32c/m32c.c (m32c_unpend_compare): Add default to silence
115 warnings.
116 (legal_subregs): Use unsigned char, make const.
117 (m32c_illegal_subreg_p): Use ARRAY_SIZE. Delete unused variables.
118
119 2007-02-08 Paul Brook <paul@codesourcery.com>
120
121 * config/arm/lib1funcs.asm (RETLDM): Pop directly into PC when no
122 special interworking needed.
123
124 2007-02-08 Harsha Jagasia <harsha.jagasia@amd.com>
125
126 * config/i386/xmmintrin.h: Make inclusion of emmintrin.h
127 conditional to __SSE2__.
128 (Entries below should have been added to first ChangeLog
129 entry for amdfam10 dated 2007-02-05)
130 * config/i386/emmintrin.h: Generate #error if __SSE2__ is not
131 defined.
132 * config/i386/pmmintrin.h: Generate #error if __SSE3__ is not
133 defined.
134 * config/i386/tmmintrin.h: Generate #error if __SSSE3__ is not
135 defined.
136
137 2007-02-08 DJ Delorie <dj@redhat.com>
138
139 * config/m32c/m32c-protos.h (m32c_illegal_subreg_p): New.
140 * config/m32c/m32c.c (legal_subregs): New.
141 (m32c_illegal_subreg_p): New.
142 * config/m32c/predicates.md (m32c_any_operand): Use it to reject
143 unsupported subregs of hard regs.
144
145 2007-02-08 Jan Hubicka <jh@suse.cz>
146
147 * tree-cfg.c (bsi_replace): Shortcut when replacing the statement with
148 the same one; always update histograms.
149
150 2007-02-08 Diego Novillo <dnovillo@redhat.com>
151
152 * passes.c (init_optimization_passes): Tidy comment.
153
154 2007-02-08 Roger Sayle <roger@eyesopen.com>
155
156 * simplify-rtx.c (simplify_unary_operation_1) <POPCOUNT>: We can
157 strip zero_extend, bswap and rotates from POCOUNT's argument.
158 <PARITY>: Likewise, we can strip not, bswap, sign_extend,
159 zero_extend and rotates from PARITY's argument.
160 <BSWAP>: A byte-swap followed by a byte-swap is an identity.
161 (simplify_const_unary_operation) <BSWAP>: Evaluate the byte-swap
162 of an integer constant at compile-time.
163
164 2007-02-08 Diego Novillo <dnovillo@redhat.com>
165
166 PR 30562
167 * tree-flow.h (struct var_ann_d): Remove field 'is_used'.
168 Update all users.
169 * tree-ssa-alias.c (compute_is_aliased): Remove. Update all
170 users.
171 (init_alias_info):
172 * tree-ssa-live.c (remove_unused_locals): Do not remove
173 TREE_ADDRESSABLE variables.
174 * tree-ssa-structalias.c (compute_points_to_sets): Tidy.
175 * tree-ssa-operands.c (add_virtual_operand): Remove argument
176 FOR_CLOBBER. Update all users.
177 If VAR has an associated alias set, add a virtual operand for
178 it if no alias is found to conflict with the memory reference.
179
180 2007-02-07 Jan Hubicka <jh@suse.cz>
181 Robert Kidd <rkidd@crhc.uiuc.edu>
182
183 * value-prof.c (visit_hist, free_hist): Return 1 instead of 0.
184
185 2007-02-07 Ian Lance Taylor <iant@google.com>
186
187 * lower-subreg.c (simple_move): Reject PARTIAL_INT modes.
188
189 2007-02-07 Roger Sayle <roger@eyesopen.com>
190
191 * config/rs6000/rs6000.md (ctz<mode>2, ffs<mode>2, popcount<mode>2,
192 parity<mode>2, smulsi3_highpart, abstf2_internal, allocate_stack,
193 tablejumpdi, movsi_to_cr_one): Remove constraints from
194 define_expand's match_operands.
195
196 2007-02-07 Roger Sayle <roger@eyesopen.com>
197
198 * global.c (compute_regsets): Move declatation of "i" inside of
199 #ifdef ELIMINABLE_REGS to avoid unused variable bootstrap failure.
200
201 2007-02-07 Jakub Jelinek <jakub@redhat.com>
202
203 PR c++/30703
204 * gimplify.c (gimplify_scan_omp_clauses): Remove special casing
205 of INDIRECT_REF <RESULT_DECL>.
206
207 * config/i386/i386.c (override_options): Set PTA_SSSE3 for core2.
208
209 2007-02-06 J"orn Rennecke <joern.rennecke@arc.com>
210 Kaz Kojima <kkojima@gcc.gnu.org>
211
212 PR target/29746
213 * config/sh/sh.c (expand_cbranchdi4): Use scratch register
214 properly.
215 (sh_initialize_trampoline): Add parentheses to avoid a warning.
216
217 2007-02-06 Zdenek Dvorak <dvorakz@suse.cz>
218
219 * doc/loop.texi: Document possibility not to perform disambiguation
220 of loops with multiple latches.
221 * cfgloopmanip.c (alp_enum_p): Removed.
222 (add_loop): Handle subloops. Use get_loop_body_with_size.
223 (create_preheader): Do not allow ENTRY_BLOCK_PTR to be preheader.
224 * cfghooks.c (redirect_edge_and_branch_force): Set dominator for
225 the new forwarder block.
226 (make_forwarder_block): Only call new_bb_cbk if it is not NULL.
227 Handle the case latch is NULL.
228 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Avoid cfg modifications
229 when marking loop exits.
230 * ifcvt.c (if_convert): Ditto. Mark loop exits even if cfg cannot
231 be modified.
232 * loop-init.c (loop_optimizer_init): Do not modify cfg. Call
233 disambiguate_loops_with_multiple_latches.
234 * tree-cfgcleanup.c (cleanup_tree_cfg_loop): Calculate dominators
235 before fix_loop_structure.
236 * cfgloop.c: Include pointer-set.h and output.h.
237 (canonicalize_loop_headers, HEADER_BLOCK, LATCH_EDGE,
238 update_latch_info, mfb_keep_just, mfb_keep_nonlatch): Removed.
239 (get_loop_latch_edges, find_subloop_latch_edge_by_profile,
240 find_subloop_latch_edge_by_ivs, find_subloop_latch_edge,
241 mfb_redirect_edges_in_set, form_subloop, merge_latch_edges,
242 disambiguate_multiple_latches, get_loop_body_with_size,
243 disambiguate_loops_with_multiple_latches): New functions.
244 (flow_loop_dump): Dump multiple latch edges.
245 (flow_loop_nodes_find): Handle loops with multiple latches.
246 (flow_loops_find): Ditto. Do not call canonicalize_loop_headers.
247 (glb_enum_p): Modified.
248 (get_loop_body): Use get_loop_body_with_size.
249 * cfgloop.h (LOOPS_HAVE_RECORDED_EXITS): New flag.
250 (AVOID_CFG_MODIFICATIONS): New constant.
251 (disambiguate_loops_with_multiple_latches, add_loop,
252 get_loop_body_with_size): Declare.
253 * Makefile.in (cfgloop.o): Add pointer-set.h and output.h.
254
255 2007-02-06 Seongbae Park <seongbae.park@gmail.com>
256
257 PR inline-asm/28686
258 * global.c (compute_regsets): New function.
259 (global_alloc): Refactored ELIMINABLE_REGSET
260 and NO_GLOBAL_ALLOC_REGS computation out.
261 (rest_of_handle_global_alloc): Call compute_regsets()
262 for non-optimizing case.
263
264 2007-02-06 Richard Henderson <rth@redhat.com>
265
266 * config/i386/constraints.md (Y2): Rename from Y.
267 (Yi): New constraint.
268 * config/i386/i386.md (movsi_1, movdi_2, pushdf_nointeger,
269 pushdf_integer, movdf_nointeger, movdf_integer, zero_extendsidi2_32,
270 zero_extendsidi2_rex64, truncxfdf2_mixed): Change Y constraints to Y2.
271 (extendsfdf2_mixed, extendsfdf2_sse, truncdfsf_fast_mixed,
272 truncdfsf_fast_sse, truncdfsf_mixed, fix_truncdfdi_sse,
273 fix_truncdfsi_sse, floatsidf2_mixed, floatsidf2_sse,
274 floatdidf2_mixed, floatdidf2_sse, absnegdf2_mixed,
275 absnegdf2_sse, sse_setccdf, fop_df_comm_mixed, fop_df_comm_sse,
276 fop_df_1_mixed, fop_df_1_sse): Change Y constraints to x.
277 * config/i386/mmx.md (mov<MMXMODEI>_internal_rex64,
278 mov<MMXMODEI>_internal, movv2sf_internal_rex64, movv2sf_internal,
279 vec_extractv2si_1): Change Y constraints to Y2.
280 * config/i386/sse.md (vec_setv4sf_0, vec_concatv2df, vec_dupv4si,
281 vec_dupv2di, sse2_concatv2si, vec_concatv4si_1, vec_concatv2di):
282 Change Y constraints to Y2.
283 (sse2_loadld): Change Y constraints to x.
284
285 2007-02-06 Roger Sayle <roger@eyesopen.com>
286
287 * config/rs6000/rs6000.md (popcount<mode>2): Rewrite.
288 (parity<mode>2): New define_expand using rs6000_emit_parity.
289 * config/rs6000/rs6000.c (rs6000_emit_popcount,
290 rs6000_emit_parity): New functions.
291 * config/rs6000/rs6000-protos.h (rs6000_emit_popcount,
292 rs6000_emit_parity): Prototype here.
293
294 2007-02-06 Ian Lance Taylor <iant@google.com>
295
296 * lower-subreg.c (simple_move_operand): Reject CONST.
297 (resolve_clobber): Call validate_change rather than directly
298 assigning to XEXP (pat, 0).
299
300 2006-02-06 Paolo Bonzini <bonzini@gnu.org>
301
302 * Makefile.in (tree-ssa-loop-ivopts.o): Add pointer-set.h dependency.
303 (tree-ssa-reassoc.o): Add pointer-set.h dependency.
304 (tree-cfg.o): Remove hashtab.h dependency.
305
306 * tree-ssa-loop-ivopts.c: Include pointer-set.h.
307 (struct ivopts_data): Change niters to pointer_map_t.
308 (struct nfe_cache_elt, nfe_hash, nfe_eq): Delete.
309 (niter_for_exit): Create pointer_map on demand. Change for
310 pointer_map API.
311 (tree_ssa_iv_optimize_init): Initialize data->niters to NULL.
312 (free_loop_data): Destroy data->niters if created and reset field.
313 (tree_ssa_iv_optimize_finalize): Don't delete data->niters here.
314 (tree_ssa_iv_optimize_loop): Check for presence of stale data.
315
316 * tree-ssa-reassoc.c: Include pointer-set.h.
317 (bb_rank): Change to long *.
318 (operand_rank): Change to pointer_map_t.
319 (find_operand_rank): Return long, -1 if not found. Declare as inline.
320 (insert_operand_rank): Accept long.
321 (operand_entry_hash, operand_entry_eq): Remove.
322 (get_rank): Return long. Adjust for changes above.
323 (init_reassoc): Change rank type to long. Adjust creation of bb_rank
324 and operand_rank.
325 (fini_reassoc): Delete operand_rank with pointer_map_destroy.
326
327 * tree-ssa-structalias.c (vi_for_tree): Change to pointer_map.
328 (struct tree_vi, tree_vi_t, tree_vi_hash, tree_vi_eq): Delete.
329 (insert_vi_for_tree): Rewrite for pointer_map API. Assert argument
330 is not NULL.
331 (lookup_vi_for_tree): Rewrite for pointer_map API. Return varinfo_t
332 directly since it cannot be NULL.
333 (get_vi_for_tree): Rewrite for pointer_map API.
334 (find_what_p_points_to): Adjust for change to lookup_vi_for_tree.
335 (init_alias_vars): Create vi_for_tree as pointer_map.
336 (delete_points_to_sets): Delete vi_for_tree using pointer_map_destroy.
337
338 * tree-cfg.c: Don't include hashtab.h.
339 (edge_to_cases): Declare as pointer_map.
340 (struct edge_to_cases_elt, edge_to_cases_hash, edge_to_cases_eq):
341 Delete.
342 (edge_to_cases_cleanup): Rewrite as pointer_map_traverse callback.
343 (start_recording_case_labels): Create edge_to_cases as pointer_map.
344 (end_recoding_case_labels): Cleanup edge_to_cases manually before
345 destroying it.
346 (record_switch_edge): Delete.
347 (get_cases_for_edge): Adjust for pointer_map API, inline
348 record_switch_edge (rewritten for new API), remove goto.
349
350 2006-02-06 Paolo Bonzini <bonzini@gnu.org>
351
352 * Makefile.in (tree-nested.o): Add pointer-set.h dependency.
353 * tree-nested.c: Include pointer-set.h.
354 (var_map_elt, var_map_eq, var_map_hash): Delete.
355 (struct nesting_info): Remove GTY marker. Change the two htab_t's
356 to pointer_map_t's.
357 (nesting_info_bitmap_obstack): New.
358 (lookup_field_for_decl): Adjust for pointer_map API.
359 (lookup_tramp_for_decl): Adjust for pointer_map API.
360 (get_nonlocal_debug_decl): Adjust for pointer_map API.
361 (get_local_debug_decl): Adjust for pointer_map API.
362 (convert_nl_goto_reference): Adjust for pointer_map API.
363 (convert_nl_goto_receiver): Adjust for pointer_map API.
364 (create_nesting_tree): Create outside GGC space. Create bitmap on
365 the new obstack. Create field_map and var_map as pointer_maps.
366 (free_nesting_tree): Adjust for changes to create_nesting_tree.
367 (root): Delete.
368 (lower_nested_functions): Move root here, no need to NULL it.
369 Initialize and release the obstack.
370
371 2007-02-06 Paolo Bonzini <bonzini@gnu.org>
372
373 * tree.c (tree_int_map_hash, tree_int_map_eq, tree_int_map_marked_p):
374 Remove prototypes and make them non-static.
375 (struct tree_int_map): Remove.
376 * tree.h (struct tree_int_map): Move here, turning TO into an
377 unsigned int.
378 (tree_int_map_hash, tree_int_map_eq, tree_int_map_marked_p): Declare.
379
380 * tree.h (TREE_COMPLEXITY): Remove.
381 (struct tree_exp): Remove complexity field.
382 * tree.c (build1_stat): Don't set it.
383
384 2007-02-06 Dorit Nuzman <dorit@il.ibm.com>
385 Victor Kaplansky <victork@il.ibm.com>
386
387 * tree-vectorizer.c (vect_is_simple_use): Support induction.
388 (vect_is_simple_reduction): Support reduction with induction as
389 one of the operands.
390 (vect_is_simple_iv_evolution): Fix formatting.
391 * tree-vect-analyze.c (vect_mark_stmts_to_be_vectorized): Fix
392 formatting. Don't mark induction phis for vectorization.
393 (vect_analyze_scalar_cycles): Analyze all inductions, then reductions.
394 * tree-vect-transform.c (get_initial_def_for_induction): New function.
395 (vect_get_vec_def_for_operand): Support induction.
396 (vect_get_vec_def_for_stmt_copy): Fix formatting and add check for
397 induction case.
398 (vectorizable_reduction): Support reduction with induction as one of
399 the operands.
400 (vectorizable_type_demotion): Use def-type of stmt argument rather
401 than dummy def-type.
402
403 * tree-ssa-loop.c (gate_scev_const_prop): Return the value of
404 flag_tree_scev_cprop.
405 * common.opt (tree-scev-cprop): New flag.
406
407 * tree-vect-transform.c (vect_create_destination_var): Use 'kind' in
408 call to vect_get_new_vect_var.
409
410 2007-02-06 Ira Rosen <irar@il.ibm.com>
411
412 * tree-vect-patterns.c (vect_recog_widen_mult_pattern): Check that
413 vectype is not NULL.
414 (vect_pattern_recog_1): Likewise.
415
416 2007-02-05 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
417
418 * fold-const.c (negate_expr_p): Handle CONJ_EXPR.
419 (fold_negate_expr): Likewise.
420
421 2007-02-05 Alexandre Oliva <aoliva@redhat.com>
422
423 PR debug/30189
424 * dwarf2out.c (modified_type_die): Follow DECL_ORIGINAL_TYPE
425 even if cv-qualification is the same.
426
427 2007-02-05 Geoffrey Keating <geoffk@apple.com>
428
429 * config/rs6000/darwin-tramp.asm (__trampoline_setup): Call
430 __enable_execute_stack on completion.
431
432 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
433
434 * config/i386/athlon.md (athlon_fldxf_k8, athlon_fld_k8,
435 athlon_fstxf_k8, athlon_fst_k8, athlon_fist, athlon_fmov,
436 athlon_fadd_load, athlon_fadd_load_k8, athlon_fadd, athlon_fmul,
437 athlon_fmul_load, athlon_fmul_load_k8, athlon_fsgn,
438 athlon_fdiv_load, athlon_fdiv_load_k8, athlon_fdiv_k8,
439 athlon_fpspc_load, athlon_fpspc, athlon_fcmov_load,
440 athlon_fcmov_load_k8, athlon_fcmov_k8, athlon_fcomi_load_k8,
441 athlon_fcomi, athlon_fcom_load_k8, athlon_fcom): Added amdfam10.
442
443 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
444
445 * config/i386/i386.md (x86_sahf_1, cmpfp_i_mixed, cmpfp_i_sse,
446 cmpfp_i_i387, cmpfp_iu_mixed, cmpfp_iu_sse, cmpfp_iu_387,
447 swapsi, swaphi_1, swapqi_1, swapdi_rex64, fix_truncsfdi_sse,
448 fix_truncdfdi_sse, fix_truncsfsi_sse, fix_truncdfsi_sse,
449 x86_fldcw_1, floatsisf2_mixed, floatsisf2_sse, floatdisf2_mixed,
450 floatdisf2_sse, floatsidf2_mixed, floatsidf2_sse,
451 floatdidf2_mixed, floatdidf2_sse, muldi3_1_rex64, mulsi3_1,
452 mulsi3_1_zext, mulhi3_1, mulqi3_1, umulqihi3_1, mulqihi3_insn,
453 umulditi3_insn, umulsidi3_insn, mulditi3_insn, mulsidi3_insn,
454 umuldi3_highpart_rex64, umulsi3_highpart_insn,
455 umulsi3_highpart_zext, smuldi3_highpart_rex64,
456 smulsi3_highpart_insn, smulsi3_highpart_zext, x86_64_shld,
457 x86_shld_1, x86_64_shrd, sqrtsf2_mixed, sqrtsf2_sse,
458 sqrtsf2_i387, sqrtdf2_mixed, sqrtdf2_sse, sqrtdf2_i387,
459 sqrtextendsfdf2_i387, sqrtxf2, sqrtextendsfxf2_i387,
460 sqrtextenddfxf2_i387): Added amdfam10_decode.
461
462 * config/i386/athlon.md (athlon_idirect_amdfam10,
463 athlon_ivector_amdfam10, athlon_idirect_load_amdfam10,
464 athlon_ivector_load_amdfam10, athlon_idirect_both_amdfam10,
465 athlon_ivector_both_amdfam10, athlon_idirect_store_amdfam10,
466 athlon_ivector_store_amdfam10): New define_insn_reservation.
467 (athlon_idirect_loadmov, athlon_idirect_movstore): Added
468 amdfam10.
469
470 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
471
472 * config/i386/athlon.md (athlon_call_amdfam10,
473 athlon_pop_amdfam10, athlon_lea_amdfam10): New
474 define_insn_reservation.
475 (athlon_branch, athlon_push, athlon_leave_k8, athlon_imul_k8,
476 athlon_imul_k8_DI, athlon_imul_mem_k8, athlon_imul_mem_k8_DI,
477 athlon_idiv, athlon_idiv_mem, athlon_str): Added amdfam10.
478
479 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
480
481 * config/i386/athlon.md (athlon_sseld_amdfam10,
482 athlon_mmxld_amdfam10, athlon_ssest_amdfam10,
483 athlon_mmxssest_short_amdfam10): New define_insn_reservation.
484
485 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
486
487 * config/i386/athlon.md (athlon_sseins_amdfam10): New
488 define_insn_reservation.
489 * config/i386/i386.md (sseins): Added sseins to define_attr type
490 and define_attr unit.
491 * config/i386/sse.md: Set type attribute to sseins for insertq
492 and insertqi.
493
494 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
495
496 * config/i386/athlon.md (sselog_load_amdfam10, sselog_amdfam10,
497 ssecmpvector_load_amdfam10, ssecmpvector_amdfam10,
498 ssecomi_load_amdfam10, ssecomi_amdfam10,
499 sseaddvector_load_amdfam10, sseaddvector_amdfam10): New
500 define_insn_reservation.
501 (ssecmp_load_k8, ssecmp, sseadd_load_k8, seadd): Added amdfam10.
502
503 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
504
505 * config/i386/athlon.md (cvtss2sd_load_amdfam10,
506 cvtss2sd_amdfam10, cvtps2pd_load_amdfam10, cvtps2pd_amdfam10,
507 cvtsi2sd_load_amdfam10, cvtsi2ss_load_amdfam10,
508 cvtsi2sd_amdfam10, cvtsi2ss_amdfam10, cvtsd2ss_load_amdfam10,
509 cvtsd2ss_amdfam10, cvtpd2ps_load_amdfam10, cvtpd2ps_amdfam10,
510 cvtsX2si_load_amdfam10, cvtsX2si_amdfam10): New
511 define_insn_reservation.
512
513 * config/i386/sse.md (cvtsi2ss, cvtsi2ssq, cvtss2si,
514 cvtss2siq, cvttss2si, cvttss2siq, cvtsi2sd, cvtsi2sdq,
515 cvtsd2si, cvtsd2siq, cvttsd2si, cvttsd2siq,
516 cvtpd2dq, cvttpd2dq, cvtsd2ss, cvtss2sd,
517 cvtpd2ps, cvtps2pd): Added amdfam10_decode attribute.
518
519 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
520
521 * config/i386/athlon.md (athlon_ssedivvector_amdfam10,
522 athlon_ssedivvector_load_amdfam10, athlon_ssemulvector_amdfam10,
523 athlon_ssemulvector_load_amdfam10): New define_insn_reservation.
524 (athlon_ssediv, athlon_ssediv_load_k8, athlon_ssemul,
525 athlon_ssemul_load_k8): Added amdfam10.
526
527 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
528
529 * config/i386/i386.h (TARGET_SSE_UNALIGNED_MOVE_OPTIMAL): New macro.
530 (x86_sse_unaligned_move_optimal): New variable.
531
532 * config/i386/i386.c (x86_sse_unaligned_move_optimal): Enable for
533 m_AMDFAM10.
534 (ix86_expand_vector_move_misalign): Add code to generate movupd/movups
535 for unaligned vector SSE double/single precision loads for AMDFAM10.
536
537 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
538
539 * config/i386/i386.h (TARGET_AMDFAM10): New macro.
540 (TARGET_CPU_CPP_BUILTINS): Add code for amdfam10.
541 Define TARGET_CPU_DEFAULT_amdfam10.
542 (TARGET_CPU_DEFAULT_NAMES): Add amdfam10.
543 (processor_type): Add PROCESSOR_AMDFAM10.
544
545 * config/i386/i386.md: Add amdfam10 as a new cpu attribute to match
546 processor_type in config/i386/i386.h.
547 Enable imul peepholes for TARGET_AMDFAM10.
548
549 * config.gcc: Add support for --with-cpu option for amdfam10.
550
551 * config/i386/i386.c (amdfam10_cost): New variable.
552 (m_AMDFAM10): New macro.
553 (m_ATHLON_K8_AMDFAM10): New macro.
554 (x86_use_leave, x86_push_memory, x86_movx, x86_unroll_strlen,
555 x86_cmove, x86_3dnow_a, x86_deep_branch, x86_use_simode_fiop,
556 x86_promote_QImode, x86_integer_DFmode_moves,
557 x86_partial_reg_dependency, x86_memory_mismatch_stall,
558 x86_accumulate_outgoing_args, x86_arch_always_fancy_math_387,
559 x86_sse_partial_reg_dependency, x86_sse_typeless_stores,
560 x86_use_ffreep, x86_use_incdec, x86_four_jump_limit,
561 x86_schedule, x86_use_bt, x86_cmpxchg16b, x86_pad_returns):
562 Enable/disable for amdfam10.
563 (override_options): Add amdfam10_cost to processor_target_table.
564 Set up PROCESSOR_AMDFAM10 for amdfam10 entry in
565 processor_alias_table.
566 (ix86_issue_rate): Add PROCESSOR_AMDFAM10.
567 (ix86_adjust_cost): Add code for amdfam10.
568
569 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
570
571 * config/i386/i386.opt: Add new Advanced Bit Manipulation (-mabm)
572 instruction set feature flag. Add new (-mpopcnt) flag for popcnt
573 instruction. Add new SSE4A (-msse4a) instruction set feature flag.
574 * config/i386/i386.h: Add builtin definition for SSE4A.
575 * config/i386/i386.md: Add support for ABM instructions
576 (popcnt and lzcnt).
577 * config/i386/sse.md: Add support for SSE4A instructions
578 (movntss, movntsd, extrq, insertq).
579 * config/i386/i386.c: Add support for ABM and SSE4A builtins.
580 Add -march=amdfam10 flag.
581 * config/i386/ammintrin.h: Add support for SSE4A intrinsics.
582 * doc/invoke.texi: Add documentation on flags for sse4a, abm, popcnt
583 and amdfam10.
584 * doc/extend.texi: Add documentation for SSE4A builtins.
585
586 2007-02-05 Bob Wilson <bob.wilson@acm.org>
587
588 * config/xtensa/xtensa.c (constantpool_mem_p): Skip over SUBREGs.
589
590 2007-02-05 Richard Guenther <rguenther@suse.de>
591
592 * tree-vectorizer.h (vectorizable_function): Add argument type
593 argument, change return type.
594 * tree-vect-patterns.c (vect_recog_pow_pattern): Adjust caller.
595 * tree-vect-transform.c (vectorizable_function): Handle extra
596 argument, return vectorized function decl.
597 (build_vectorized_function_call): Remove.
598 (vectorizable_call): Handle calls with result and argument types
599 differing. Handle loop vectorization factor correctly.
600 * targhooks.c (default_builtin_vectorized_function): Adjust for
601 extra argument.
602 * targhooks.h (default_builtin_vectorized_function): Likewise.
603 * target.h (builtin_vectorized_function): Add argument type
604 argument.
605 * config/i386/i386.c (ix86_builtin_vectorized_function): Handle
606 extra argument, allow vectorizing of lrintf.
607 * doc/tm.texi (TARGET_VECTORIZE_BUILTIN_VECTORIZED_FUNCTION): Adjust
608 documentation of target hook.
609
610 2007-02-05 Hans-Peter Nilsson <hp@axis.com>
611
612 PR target/30665
613 * config/cris/cris.md ("*andsi_movu", "*andsi_clear", "*andhi_movu")
614 ("*andhi_clear", andu (casesi+45)): For size-changed operand where
615 memory is allowed, require !side_effects_p, not just !MEM_VOLATILE_P.
616
617 2007-02-05 Roger Sayle <roger@eyesopen.com>
618
619 * fold-const.c (fold_unary) <REAL_PART>: Test for availability of
620 BUILT_IN_COS before simplifying REAL_PART(CEXPI)) to COS.
621 <IMAG_PART>: Likewise, check for availability of BUILT_IN_SIN.
622 * builtins.c (fold_builtin_sincos): Check for TARGET_C99_FUNCTIONS
623 before canonicalizing sincos to cexpi.
624 (fold_builtin_cexp): Likewise, for canonicalizing cexp to cexpi.
625
626 2007-02-05 Roger Sayle <roger@eyesopen.com>
627
628 * config/alpha/alpha.c (alpha_add_builtins): New Helper function.
629 Set TREE_READONLY and TREE_NOTHROW directly, not via attributes.
630 (alpha_init_builtins): Use alpha_add_builtins to process tables.
631
632 2007-02-05 Roger Sayle <roger@eyesopen.com>
633
634 * mips-tfile.c (initialize_init_file): Correct endianness test.
635
636 2007-02-05 Kazu Hirata <kazu@codesourcery.com>
637
638 * config/m68k/m68k.md (pushdi-1, pushdi, movsi+1): Don't use
639 the 'y' constraint.
640
641 2007-02-05 Richard Sandiford <richard@codesourcery.com>
642
643 * dwarf2out.c (dwarf2out_frame_debug_expr): Record the register
644 saves in a PARALLEL before the register assignments.
645
646 2007-02-05 Richard Sandiford <richard@codesourcery.com>
647
648 * doc/tm.texi (DWARF_ALT_FRAME_RETURN_COLUMN): Do not require
649 DWARF_FRAME_RETURN_COLUMN to be a general register.
650 * dwarf2out.c (init_return_column_size): New function, split from...
651 (expand_builtin_init_dwarf_reg_sizes): ...here. Allow both
652 DWARF_FRAME_RETURN_COLUMN and DWARF_ALT_FRAME_RETURN_COLUMN
653 to be nongeneral registers.
654 * config/m68k/m68k.h (DWARF_FRAME_REGNUM): Only map FP and
655 integer registers.
656 (DWARF_FRAME_REGISTERS, DWARF_FRAME_RETURN_COLUMN): Define.
657 (DWARF_ALT_FRAME_RETURN_COLUMN): Define.
658
659 2007-02-04 Zdenek Dvorak <dvorakz@suse.cz>
660
661 * cfgcleanup.c (try_optimize_cfg): Avoid removing ENTRY_BLOCK_PTR.
662
663 2007-02-04 Zdenek Dvorak <dvorakz@suse.cz>
664
665 * cfgloopmanip.c (loop_delete_branch_edge): Removed.
666 (remove_path): Use can_remove_branch_p and remove_branch instead
667 of loop_delete_branch_edge.
668 * tree-ssa-loop-manip.c (scale_dominated_blocks_in_loop): New function.
669 (tree_transform_and_unroll_loop): Remove dead branches immediately.
670 Update profile using scale_dominated_blocks_in_loop.
671 * cfghooks.c (can_remove_branch_p, remove_branch): New functions.
672 * cfghooks.h (struct cfg_hooks): Add can_remove_branch_p.
673 (can_remove_branch_p, remove_branch): Declare.
674 * tree-cfg.c (tree_can_remove_branch_p): New function.
675 (tree_cfg_hooks): Add tree_can_remove_branch_p.
676 * cfgrtl.c (rtl_can_remove_branch_p): New function.
677 (rtl_cfg_hooks, cfg_layout_rtl_cfg_hook): Add rtl_can_remove_branch_p.
678
679 2007-02-05 Jan Hubicka <jh@suse.cz>
680
681 PR middle-end/30696
682 * ipa-inline.c (cgraph_clone_inlined_nodes): When there are unanalyzed
683 nodes in cgraph, don't remove offline copy of the function.
684
685 2007-02-04 Jan Hubicka <jh@suse.cz>
686
687 * tree-sra.c (sra_walk_expr): Add linebreaks. BITFIELD_REFs into
688 vectors might cause maybe_lookup_element_for_expr to be called
689 on non-sra-candidate.
690
691 2007-02-04 Kazu Hirata <kazu@codesourcery.com>
692
693 * config/bfin/bfin-modes.def, config/bfin/bfin.c,
694 config/bfin/bfin.md, config/bfin/predicates.md: Follow
695 spelling conventions.
696
697 2007-02-04 Richard Guenther <rguenther@suse.de>
698
699 PR middle-end/30636
700 * fold-const.c (try_move_mult_to_index): Make sure to not
701 overflow one dimension of a multi-dimensional array access.
702
703 2007-02-04 Jan Hubicka <jh@suse.cz>
704
705 * passes.c (init_optimization_passes): Reindent.
706
707 2007-02-04 Jan Hubicka <jh@suse.cz>
708 Eric Botcazou <ebotcazou@adacore.com>
709
710 * tree-optimize.c (has_abnormal_outgoing_edge_p): Move to...
711 (execute_fixup_cfg): Break out the abnormal goto code.
712 * tree-inline.c (has_abnormal_outgoing_edge_p): ...here.
713 (make_nonlocal_label_edges): Move here from execute_fixup_cfg.
714 (optimize_inline_calls): Call make_nonlocal_label_edges.
715
716 2007-02-04 Jan Hubicka <jh@suse.cz>
717
718 * tree-ssa-copyrename.c (copy_rename_partition_coalesce): Return
719 true when something was changed.
720 (rename_ssa_copies): When something was changed, do
721 TODO_remove_unused_locals.
722 * tree-ssa-forwprop.c (tree_ssa_forward_propagate_single_use_value):
723 add TODO_remove_unused_locals when instruction was removed.
724
725 2007-02-04 Jan Hubicka <jh@suse.cz>
726
727 * ipa-inline.c (try_inline): Improve debug output; work on already
728 inline edges too.
729 (cgraph_decide_inlining_incrementally): Indent; improve debug output;
730 call try_inline for already inlined edges too when flattening;
731 inline also functions that make callee growth but overall unit size
732 reduce.
733
734 2007-02-04 Kazu Hirata <kazu@codesourcery.com>
735
736 * config/m32c/bitops.md, config/m32c/jump.md,
737 config/m32c/m32c.c, config/m32c/m32c.h, config/m32r/m32r.c,
738 config/m32r/m32r.h, config/m32r/m32r.md,
739 config/m32r/predicates.md, config/m68hc11/larith.asm,
740 config/m68hc11/m68hc11.c, config/m68hc11/m68hc11.h,
741 config/m68k/m68k.h, config/mcore/mcore.md, config/mips/4k.md,
742 config/mips/mips-protos.h, config/mips/mips.c,
743 config/mips/mips.h, config/mips/mips.md, config/mips/mips16.S,
744 config/mn10300/mn10300.h, config/mn10300/predicates.md,
745 config/mt/mt.c, config/mt/mt.h, config/mt/mt.md: Follow
746 spelling conventions.
747
748 * config/v850/v850.c, config/v850/v850.h, config/v850/v850.md:
749 Follow spelling conventions.
750
751 2007-02-03 Douglas Gregor <doug.gregor@gmail.com>
752
753 * c-opts.c (c_common_post_options): If C++0x mode is enabled, don't
754 warn about C++0x compatibility.
755
756 2007-02-04 Kazu Hirata <kazu@codesourcery.com>
757
758 * config/h8300/h8300.c, config/h8300/h8300.h,
759 config/h8300/h8300.md: Follow spelling conventions.
760
761 2007-02-03 Uros Bizjak <ubizjak@gmail.com>
762
763 PR middle-end/30667
764 * combine.c (try_combine): Do not substitute source operand
765 with constants wider than 2 * HOST_BITS_PER_WIDE_INT.
766
767 2007-02-03 Jan Hubicka <jh@suse.cz>
768
769 PR gcov-profile/30650
770 * value-prof.c (stringop_block_profile): Fix handling of size counter;
771 do not divide by zero for never executed counters.
772 (tree_find_values_to_profile): Fix counters.
773 * gcov-ui.h (GCOV_COUNTER_AVERAGE, GCOV_COUNTER_IOR): Fix comments.
774
775 2007-02-03 Ian Lance Taylor <iant@google.com>
776
777 * lower-subreg.c (simple_move_operand): New static function,
778 broken out of simple_move. Reject LABEL_REF, SYMBOL_REF, and HIGH
779 operands.
780 (simple_move): Call simple_move_operand.
781 (find_decomposable_subregs): Add special handling of MEMs.
782 (can_decompose_p): Rename from cannot_decompose_p. Reverse
783 meaning of return value. If we see a hard register, test whether
784 it can store a word_mode value. Change all callers.
785
786 2007-02-03 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
787
788 * pa.md (addvdi3, addvsi3, subvdi3, subvsi3, negvdi2, negvsi2): New
789 ftrapv insns and expanders.
790 (subdi3): Change define_expand operand 1 to arith11_operand, and
791 operand 2 to reg_or_0_operand. Change constraints of 64-bit insn
792 pattern to handle reg_or_0 operands. Revise 32-bit insn pattern to
793 handle 11-bit constants and reg_or_0 operands in operands 1 and 2,
794 respectively.
795
796 PR middle-end/30174
797 * varasm.c (notice_global_symbol): Treat global objects as weak when
798 flag_shlib is true.
799
800 2007-02-03 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
801
802 * emit-rtl.c (dconstpi): Delete.
803 (dconstsqrt2): New.
804 (init_emit_once): Delete dconstpi and init dconstsqrt2.
805 * real.h (dconstpi): Delete.
806 (dconstsqrt2): New.
807 * builtins.c (fold_builtin_cabs): Use dconstsqrt2.
808 (fold_builtin_hypot): Likewise.
809
810 2007-02-03 Tom Tromey <tromey@redhat.com>
811
812 PR driver/30246
813 * gcc.c (cpp_unique_options): Any of -ggdb3, -gstabs3,
814 -gcoff3, -gxcoff3, -gvms3 implies -dD.
815
816 2007-02-03 Kazu Hirata <kazu@codesourcery.com>
817
818 * c-decl.c, config/avr/avr.c, config/avr/avr.h,
819 config/m68k/m68k.c, config/m68k/netbsd-elf.h,
820 config/mn10300/mn10300.c, config/pdp11/pdp11.h,
821 config/rs6000/cell.md, config/rs6000/darwin.h,
822 config/sh/sh.md, config/sh/sh4-300.md, config/spu/spu.c,
823 config/spu/spu.md, cselib.c, expr.c, haifa-sched.c, hwint.h,
824 jump.c, reload.c, sched-deps.c, sched-int.h, tree-inline.c,
825 tree-profile.c, tree-ssa-live.h, tree-vrp.c: Fix comment
826 typos. Follow spelling conventions.
827 * doc/invoke.texi: Follow spelling conventions.
828
829 2007-02-03 Roger Sayle <roger@eyesopen.com>
830
831 * simplify-rtx.c (simplify_relational_operation_1): Implement some
832 canonicalization transformations that attempt to simplify integer
833 constant comparisons to become comparisons against zero.
834
835 2007-02-03 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
836
837 * builtins.c (fold_builtin_cabs): Fold cabs(x+xi) into
838 fabs(x)*sqrt(2).
839 * fold-const.c (fold_binary): Fix comment typos. Fold complex
840 (x,0)-(0,y) into (x,-y). Likewise (0,y)-(x,0) into (-x,y).
841
842 2007-02-02 Mike Stump <mrs@apple.com>
843
844 * config/darwin9.h (DARWIN_LINKER_GENERATES_ISLANDS): Add.
845 * config/rs6000/rs6000.c (DARWIN_GENERATE_ISLANDS): Add.
846 (output_call): Use DARWIN_GENERATE_ISLANDS to decide when to
847 generate a branch island.
848
849 2007-02-02 Bob Wilson <bob.wilson@acm.org>
850
851 * config/xtensa/xtensa.c (smalloffset_mem_p): Use BASE_REG_P.
852 (xtensa_legitimate_address_p): New.
853 (xtensa_legitimize_address): New.
854 (xtensa_output_addr_const_extra): New.
855 * config/xtensa/xtensa.h (REG_OK_STRICT_FLAG): Define.
856 (BASE_REG_P): New.
857 (REG_OK_FOR_BASE_P): Use BASE_REG_P.
858 (GO_IF_LEGITIMATE_ADDRESS): Move code to xtensa_legitimate_address_p.
859 (LEGITIMIZE_ADDRESS): Move code to xtensa_legitimize_address.
860 (OUTPUT_ADDR_CONST_EXTRA): Move code to xtensa_output_addr_const_extra.
861 * config/xtensa/xtensa-protos.h (xtensa_legitimate_address_p): New.
862 (xtensa_legitimize_address): New.
863 (xtensa_output_addr_const_extra): New.
864
865 2007-02-02 Steve Ellcey <sje@cup.hp.com>
866
867 * config/ia64/ia64.c (ia64_print_operand): Fix compare strings.
868
869 2007-02-02 Ian Lance Taylor <iant@google.com>
870
871 * expmed.c (expand_divmod): Add comment.
872
873 2007-02-02 Kazu Hirata <kazu@codesourcery.com>
874
875 * emit-rtl.c (renumber_insns): Remove.
876 * flags.h: Remove the extern for flag_renumber_insns.
877 * rtl.h: Remove the prototype for renumber_insns.
878 * toplev.c (flag_renumber_insns): Remove.
879
880 2007-02-02 Hui-May Chang <hm.chang@apple.com>
881
882 Revert for x86 darwin:
883 2005-06-19 Uros Bizjak <uros@kss-loka.si>
884
885 * config/i386/i386.c (ix86_function_arg_regno_p): Put back the
886 code before the following patch under TARGET_MACHO.
887 (ix86_function_value_regno_p): Likewise.
888
889 2007-02-02 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
890
891 * fold-const.c (negate_expr_p, fold_negate_expr): Handle
892 COMPLEX_EXPR.
893
894 2007-02-02 Maxim Kuvyrkov <mkuvyrkov@ispras.ru>
895
896 PR target/29682
897 * config/ia64/ia64.c (ia64_speculate_insn): Restrict to memory loads to
898 general or fp registers. Add comments.
899 * config/ia64/ia64.md (reg_pred_prefix): Add comment.
900
901 2007-02-02 Paolo Bonzini <bonzini@gnu.org>
902
903 * pointer-set.c (insert_aux): Only return insertion slot.
904 (pointer_set_insert): Adjust.
905 (pointer_set_traverse, struct pointer_map_t, pointer_map_create,
906 pointer_map_destroy, pointer_map_insert, pointer_map_contains,
907 pointer_map_traverse): New.
908 * pointer-set.h (pointer_set_traverse, struct pointer_map_t,
909 pointer_map_create, pointer_map_destroy, pointer_map_insert,
910 pointer_map_contains, pointer_map_traverse): Declare.
911
912 2007-02-02 Jakub Jelinek <jakub@redhat.com>
913
914 PR middle-end/30473
915 * builtins.c (fold_builtin_sprintf): Do not attempt to optimize
916 sprintf (str, "%s"). Do not optimize sprintf (str, "nopercent", p++).
917
918 2007-02-02 Maxim Kuvyrkov <mkuvyrkov@ispras.ru>
919
920 * sched-int.h (ds_to_dk, dk_to_ds): Declare functions.
921
922 (struct _dep): New type.
923 (dep_t): New typedef.
924 (DEP_PRO, DEP_CON, DEP_KIND): New access macros.
925 (DEP_STATUS): New access macro. The macro with the same name was
926 renamed to DEP_LINK_STATUS.
927 (dep_init): Declare function
928
929 (struct _dep_link): New type.
930 (dep_link_t): New typedef.
931 (DEP_LINK_NODE, DEP_LINK_NEXT, DEP_LINK_PREV_NEXTP): New access macros.
932 (DEP_LINK_DEP, DEP_LINK_PRO, DEP_LINK_CON, DEP_LINK_KIND): New macros.
933 (DEP_LINK_STATUS): New macro.
934 (debug_dep_links): New debug function.
935
936 (struct _deps_list): New type.
937 (deps_list_t): New typedef.
938 (DEPS_LIST_FIRST): New access macro.
939 (FOR_EACH_DEP_LINK): New cycle macro.
940 (create_deps_list, free_deps_list, delete_deps_list): Declare
941 functions.
942 (deps_list_empty_p, debug_deps_list, add_back_dep_to_deps_list): Ditto.
943 (find_link_by_pro_in_deps_list, find_link_by_con_in_deps_list): Ditto.
944 (copy_deps_list_change_con): Ditto.
945
946 (move_dep_link): Declare function.
947
948 (struct _dep_node): New type.
949 (dep_node_t): New typedef.
950 (DEP_NODE_BACK, DEP_NODE_DEP, DEP_NODE_FORW): New access macros.
951
952 (struct haifa_insn_data.back_deps): New field to hold backward
953 dependencies of the insn.
954 (struct haifa_insn_data.depend): Rename to forw_deps. Change its type
955 to deps_list_t.
956 (struct haifa_insn_data.resolved_deps): Rename to resolved_back_deps.
957 Change its type to deps_list_t.
958 (INSN_BACK_DEPS): New access macro to use instead of LOG_LINKS.
959 (INSN_DEPEND): Rename to INSN_FORW_DEPS.
960 (RESOLVED_DEPS): Rename to INSN_RESOLVED_BACK_DEPS.
961
962 (INSN_COST): Move to haifa-sched.c. Use insn_cost () instead.
963
964 (DEP_STATUS): Rename to DEP_LINK_STATUS. Fix typo in the comment.
965
966 (add_forw_dep, delete_back_forw_dep, insn_cost): Update declaration and
967 all callers.
968 (dep_cost): Declare.
969
970 * sched-deps.c (CHECK): New macro to (en/dis)able sanity checks.
971 (ds_to_dk, dk_to_ds): New functions.
972
973 (init_dep_1): New static function.
974 (init_dep): New function.
975 (copy_dep): New static function.
976
977 (dep_link_consistent_p, attach_dep_link, add_to_deps_list): New static
978 functions.
979 (detach_dep_link): New static function.
980 (move_dep_link): New function.
981
982 (dep_links_consistent_p, dump_dep_links): New static functions.
983 (debug_dep_links): New debugging function.
984
985 (deps_obstack, dl_obstack, dn_obstack): New static variables.
986
987 (alloc_deps_list, init_deps_list): New static functions.
988 (create_deps_list): New function.
989 (clear_deps_list): New static function.
990 (free_deps_list, delete_deps_list, deps_list_empty_p): New functions.
991 (deps_list_consistent_p, dump_deps_list): New static functions.
992 (debug_deps_list): New function.
993 (add_back_dep_to_deps_list, find_link_by_pro_in_deps_list): New
994 functions.
995 (find_link_by_con_in_deps_list, copy_deps_list_change_con): Ditto.
996
997 (maybe_add_or_update_back_dep_1, add_or_update_back_dep_1): Update to
998 use new scheduler dependencies lists.
999 (add_back_dep, delete_all_dependences, fixup_sched_groups): Ditto.
1000 (sched_analyze): Ditto. Initialize dependencies lists.
1001 (add_forw_dep, compute_forward_dependences): Update to use new
1002 scheduler dependencies lists.
1003
1004 (init_dependency_caches): Init deps_obstack.
1005 (free_dependency_caches): Free deps_obstack.
1006
1007 (adjust_add_sorted_back_dep, adjust_back_add_forw_dep): Update to use
1008 new scheduler dependencies lists.
1009 (delete_forw_dep, add_or_update_back_forw_dep): Ditto.
1010 (add_back_forw_dep, delete_back_forw_dep): Ditto.
1011
1012 * sched-rgn.c (set_spec_fed, find_conditional_protection, is_pfree):
1013 Update to use new scheduler dependencies lists.
1014 (is_conditionally_protected, is_prisky, add_branch_dependences): Ditto.
1015 (debug_dependencies): Ditto.
1016 (schedule_region): Update comments.
1017
1018 * sched-ebb.c (earliest_block_with_similiar_load): Update to use new
1019 scheduler dependencies lists.
1020 (schedule_ebb): Update comments.
1021
1022 * rtl.def (DEPS_LIST): Remove.
1023
1024 * lists.c (unused_deps_list): Remove.
1025 (free_list): Update assertions.
1026
1027 (alloc_DEPS_LIST, free_DEPS_LIST_list, free_DEPS_LIST_node): Remove.
1028 (remove_free_DEPS_LIST_elem, copy_DEPS_LIST_list): Ditto.
1029
1030 * rtl.h (free_DEPS_LIST_list, alloc_DEPS_LIST): Remove declarations.
1031 (remove_free_DEPS_LIST_elem, copy_DEPS_LIST_list): Ditto.
1032
1033 * haifa-sched.c (comments): Update.
1034 (insn_cost1): Remove. Inline the code into insn_cost ().
1035 (insn_cost): Update to use new scheduler dependencies lists. Move
1036 processing of the dependency cost to dep_cost ().
1037 (dep_cost): New function. Use it instead of insn_cost () when
1038 evaluating cost of the dependency. Use compatible interface to
1039 interact with the target.
1040 (priority): Update to use new scheduler dependencies lists.
1041 (rank_for_schedule): Ditto. Optimize heuristic that prefers the insn
1042 with greater number of insns that depend on the insn.
1043 (schedule_insn): Update to use new scheduler dependencies lists. Add
1044 code to free backward dependencies lists. Inline and optimize code
1045 from resolve_dep () - see PR28071.
1046 (ok_for_early_queue_removal): Update to use new scheduler dependencies
1047 lists. Update call to targetm.sched.is_costly_dependence hook.
1048
1049 (fix_inter_tick, try_ready, fix_tick_ready): Update to use new
1050 scheduler dependencies lists.
1051
1052 (resolve_dep): Remove. Move the logic to schedule_insn ().
1053 (init_h_i_d): Initialize dependencies lists.
1054
1055 (process_insn_depend_be_in_spec): Rename to
1056 process_insn_forw_deps_be_in_spec. Update to use new scheduler
1057 dependencies lists.
1058 (add_to_speculative_block, create_check_block_twin, fix_recovery_deps):
1059 Update to use new scheduler dependencies lists.
1060 (clear_priorities, calc_priorities, add_jump_dependencies): Ditto.
1061
1062 * ddg.c (create_ddg_dependence, create_ddg_dep_no_link): Update to use
1063 new scheduler dependencies lists.
1064 (build_intra_loop_deps): Ditto.
1065
1066 * target.h (struct _dep): Declare to use in
1067 gcc_target.sched.is_costly_dependence.
1068 (struct gcc_target.sched.adjust_cost): Fix typo.
1069 (struct gcc_target.sched.is_costly_dependence): Change signature to use
1070 single dep_t parameter instead of an equivalent triad.
1071 (struct gcc_target.sched.adjust_cost_2): Remove.
1072
1073 * target-def.h (TARGET_SCHED_ADJUST_COST_2): Remove.
1074
1075 * reg-notes.def (DEP_TRUE, DEP_OUTPUT, DEP_ANTI): Update comments.
1076
1077 * doc/tm.texi (TARGET_SCHED_IS_COSTLY_DEPENDENCE): Update
1078 documentation.
1079 (TARGET_SCHED_ADJUST_COST_2): Remove documentation.
1080
1081 * doc/rtl.texi (LOG_LINKS): Remove part about instruction scheduler.
1082 (REG_DEP_TRUE): Document.
1083
1084 * config/ia64/ia64.c (ia64_adjust_cost_2): Rename to ia64_adjust_cost.
1085 Change signature to correspond to the targetm.sched.adjust_cost hook.
1086 Update use in TARGET_SCHED_ADJUST_COST_2.
1087 (TARGET_SCHED_ADJUST_COST_2): Rename to TARGET_SCHED_ADJUST_COST.
1088 (ia64_dependencies_evaluation_hook, ia64_dfa_new_cycle): Update to use
1089 new scheduler dependencies lists.
1090 (ia64_gen_check): Ditto.
1091
1092 * config/mips/mips.c (vr4130_swap_insns_p): Update to use new scheduler
1093 dependencies lists.
1094
1095 * config/rs6000/rs6000.c (rs6000_is_costly_dependence): Change
1096 signature to correspond to the targetm.sched.is_costly_dependence hook.
1097 (is_costly_group): Update to use new scheduler dependencies lists.
1098
1099 * config/spu/spu.c (spu_sched_adjust_cost): Use insn_cost () function
1100 instead of INSN_COST () macro.
1101
1102 2007-02-01 Ian Lance Taylor <iant@google.com>
1103
1104 * lower-subreg.c (resolve_clobber): Handle a subreg of a concatn.
1105
1106 2007-02-01 Guy Martin <gmsoft@gentoo.org>
1107
1108 * pa.md (tp_load): Correct mfctl instruction syntax.
1109
1110 2007-02-01 Geoffrey Keating <geoffk@apple.com>
1111
1112 * config/rs6000/rs6000.c (rs6000_stack_info): Correct
1113 altivec_padding_size calculation on AIX. Improve comment, add
1114 assert to verify that it's right.
1115
1116 * config/rs6000/darwin.h (MD_UNWIND_SUPPORT): Don't define for
1117 64-bit.
1118
1119 2007-2-01 Seongbae Park <seongbae.park@gmail.com>
1120
1121 PR inline-asm/28686
1122 * global.c (global_alloc): Add mising initialization of
1123 ELIMINABLE_REGSET.
1124
1125 2007-02-01 Roger Sayle <roger@eyesopen.com>
1126
1127 * alias.c (init_alias_analysis): Correct whitespace.
1128 * bb-reorder.c (fix_edges_for_rarely_executed_code,
1129 partition_hot_cold_basic_blocks): Likewise.
1130 * builtins.c (expand_builtin_printf, expand_builtin_fprintf,
1131 expand_builtin_sprintf, fold_builtin_carg, fold_builtin_sprintf,
1132 maybe_emit_sprintf_chk_warning, fold_builtin_sprintf_chk,
1133 fold_builtin_snprintf_chk, fold_builtin_printf,
1134 fold_builtin_fprintf, do_mpfr_ckconv, do_mpfr_arg1, do_mpfr_arg2,
1135 do_mpfr_arg3, do_mpfr_sincos): Likewise.
1136 * cfgcleanup.c (cleanup_cfg): Likewise.
1137 * cfgexpand.c (tree_expand_cfg): Likewise.
1138 * fold-const.c (fold_binary) <RDIV_EXPR>: Likewise.
1139 * function.c (get_next_funcdef_no): Likewise.
1140 * gengtype.c (main): Likewise.
1141 * genmodes.c (main): Likewise.
1142 * gcse.c (bypass_conditional_jumps, print_ldst_list): Likewise.
1143 * haifa-sched.c (schedule_block, extend_h_i_d): Likewise.
1144 * ifcvt.c (noce_emit_move_insn): Likewise.
1145 * modulo-sched.c (generate_prolog_epilog, sms_schedule_by_order):
1146 Likewise.
1147 * stor-layout.c (get_best_mode): Likewise.
1148 * tree-ssa-loop-niter.c (get_val_for): Likewise.
1149 * tree-ssa-structalias.c (get_varinfo, get_varinfo_fc,
1150 scc_visit, do_ds_constraint, do_complex_constraint, label_visit,
1151 perform_var_substitution, solve_graph): Likewise.
1152 * tree-vrp.c (vrp_finalize): Likewise.
1153
1154 2007-02-01 Ian Lance Taylor <iant@google.com>
1155
1156 * lower-subreg.c (simplify_gen_subreg_concatn): If we ask for the
1157 high part of a paradoxical subreg, return a constant zero.
1158
1159 2007-02-01 Zdenek Dvorak <dvorakz@suse.cz>
1160
1161 * toplev.c (lang_dependent_init): Call init_set_costs.
1162 * loop-init.c (loop_optimizer_init): Do not call init_set_costs.
1163
1164 2007-02-01 Richard Guenther <rguenther@suse.de>
1165
1166 PR middle-end/30656
1167 * fold-const.c (fold_negate_expr): Allow negating a
1168 constant if overflow does not change.
1169
1170 2007-02-01 Ralf Wildenhues <Ralf.Wildenhues@gmx.de>
1171
1172 * doc/c-tree.texi (Expression trees): Improve markup.
1173 * doc/tm.texi (Register Classes, Addressing Modes)
1174 (Floating Point): Fix spacing after abbreviations. Fix some
1175 typos.
1176
1177 2007-02-01 Ben Elliston <bje@au.ibm.com>
1178
1179 * doc/invoke.texi: Replace "bugfix" with "bug fix" throughout.
1180 * doc/contrib.texi: Likewise.
1181 * doc/install.texi: Likewise.
1182
1183 2007-01-31 Richard Henderson <rth@redhat.com>
1184 Ian Lance Taylor <iant@google.com>
1185
1186 * lower-subreg.c: New file.
1187 * rtl.def (CONCATN): Define.
1188 * passes.c (init_optimization_passes): Add pass_lower_subreg and
1189 pass_lower_subreg2.
1190 * emit-rtl.c (update_reg_offset): New static function, broken out
1191 of gen_rtx_REG_offset.
1192 (gen_rtx_REG_offset): Call update_reg_offset.
1193 (gen_reg_rtx_offset): New function.
1194 * regclass.c: Revert patch of 2006-03-05, restoring
1195 reg_scan_update.
1196 (clear_reg_info_regno): New function.
1197 * dwarf2out.c (concatn_loc_descriptor): New static function.
1198 (loc_descriptor): Handle CONCATN.
1199 * common.opt (fsplit_wide_types): New option.
1200 * opts.c (decode_options): Set flag_split_wide_types when
1201 optimizing.
1202 * timevar.def (TV_LOWER_SUBREG): Define.
1203 * rtl.h (gen_reg_rtx_offset): Declare.
1204 (reg_scan_update): Declare.
1205 * regs.h (clear_reg_info_regno): Declare.
1206 * tree-pass.h (pass_lower_subreg): Declare.
1207 (pass_lower_subreg2): Declare.
1208 * doc/invoke.texi (Option Summary): List -fno-split-wide-types.
1209 (Optimize Options): Add -fsplit-wide-types to -O1 list. Document
1210 -fsplit-wide-types.
1211 * doc/rtl.texi (Regs and Memory): Document concat and concatn.
1212 * Makefile.in (OBJS-common): Add lower-subreg.o.
1213 (lower-subreg.o): New target.
1214
1215 2007-01-31 Kazu Hirata <kazu@codesourcery.com>
1216
1217 * config/sh/sh.h (HAVE_SECONDARY_RELOADS): Remove.
1218
1219 2007-01-31 Anatoly Sokolov <aesok@post.ru>
1220
1221 PR target/19087
1222 * config/avr/avr.c (DWARF2_ADDR_SIZE): Define.
1223
1224 2007-01-31 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1225
1226 PR middle-end/29335
1227 * builtins.c (fold_builtin_sqrt): Use MPFR for constant args.
1228
1229 2007-01-31 Zdenek Dvorak <dvorakz@suse.cz>
1230
1231 * cfgloop.h: Include vec-prim.h.
1232 (enum li_flags): Remove LI_ONLY_OLD.
1233 (loop_iterator): Changed.
1234 (fel_next, fel_init): Iterate over loop tree.
1235 (FOR_EACH_LOOP_BREAK): New macro.
1236 * loop-unswitch.c (unswitch_loops): Do not pass LI_ONLY_OLD to
1237 FOR_EACH_LOOP.
1238 * tree-ssa-loop-unswitch.c (tree_ssa_unswitch_loops): Ditto.
1239 * modulo-sched.c (sms_schedule): Ditto.
1240 * tree-vectorizer.c (vectorize_loops): Ditto.
1241 * doc/loop.texi: Update information on loop numbering and behavior of
1242 FOR_EACH_LOOP wrto new loops.
1243 * tree-scalar-evolution.c (compute_overall_effect_of_inner_loop,
1244 add_to_evolution_1): Test nestedness of loops instead of comparing
1245 their numbers.
1246 * tree-chrec.c (chrec_fold_plus_poly_poly,
1247 chrec_fold_multiply_poly_poly, chrec_evaluate,
1248 hide_evolution_in_other_loops_than_loop, chrec_component_in_loop_num,
1249 reset_evolution_in_loop): Ditto.
1250 * Makefile.in (CFGLOOP_H): Add vecprim.h dependency.
1251
1252 2007-01-31 Dirk Mueller <dmueller@suse.de>
1253
1254 * c-common.c (warn_about_parentheses): Separate warning about
1255 un-parenthized sequence of comparison operators from the one
1256 which is supposed to warn about x <= y <= z.
1257
1258 2007-01-31 Uros Bizjak <ubizjak@gmail.com>
1259
1260 * optabs.h (enum optab_index): Add new OTI_isinf.
1261 (isinf_optab): Define corresponding macro.
1262 * optabs.c (init_optabs): Initialize isinf_optab.
1263 * genopinit.c (optabs): Implement isinf_optab using isinf?f2
1264 patterns.
1265 * builtins.c (mathfn_built_in): Handle BUILT_IN_ISINF{,F,L}.
1266 (expand_builtin_interclass_mathfn): Expand BUILT_IN_ISINF{,F,L}
1267 using isinf_optab.
1268 (expand_builtin): Expand BUILT_IN_ISINF{,F,L} using
1269 expand_builtin_interclass_mathfn.
1270 * reg_stack.c (subst_stack_regs_pat): Handle UNSPEC_FXAM.
1271 * config/i386/i386.md (UNSPEC_FXAM): New constant.
1272 (fxam<mode>2_i387): New insn pattern.
1273 (isinf<mode>2) New expander to implement isinf, isinff and isinfl
1274 built-in functions as x87 inline asm.
1275
1276 2007-01-31 Kazu Hirata <kazu@codesourcery.com>
1277
1278 * gcc/config/arm/unwind-arm.h (_sleb128_t, _uleb128_t): New.
1279
1280 2007-01-30 Eric Christopher <echristo@apple.com>
1281
1282 * config.gcc: Add geode.
1283
1284 2007-01-31 Kazu Hirata <kazu@codesourcery.com>
1285
1286 * cgraphunit.c, config/arm/arm.c, config/m68k/m68k.c,
1287 ipa-inline.c, tree-profile.c, tree-ssa-live.c,
1288 tree-ssa-math-opts.c, tree-ssanames.c, tree-vect-analyze.c,
1289 value-prof.c: Fix comment typos.
1290
1291 2007-01-30 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
1292
1293 PR c++/24745
1294 * doc/invoke.texi (Wpointer-arith): Document warning.
1295
1296 2007-01-30 Janis Johnson
1297
1298 * doc/extend.texi (Decimal Floating Types): Remove decfloat.h from
1299 the list of discrepancies from the draft TR.
1300
1301 2007-01-30 Dirk Mueller <dmueller@suse.de>
1302
1303 PR c++/30601
1304 * doc/invoke.texi (-Wreturn-type): Update description to
1305 match new behavior.
1306
1307 2007-01-30 Richard Sandiford <richard@codesourcery.com>
1308
1309 * cfgrtl.c (try_redirect_by_replacing_jump): Check only_sets_cc0_p.
1310
1311 2007-01-30 Uros Bizjak <ubizjak@gmail.com>
1312
1313 * builtins.c (expand_builtin_int_interclass_roundingfn): New function
1314 to handle optabs that operate on floating point input argument and
1315 output to integer output.
1316 (expand_builtin_mathfn) [BUILT_IN_ILOGB]: Move from here ...
1317 (expand_builtin_interclass_mathfn) [BUILT_IN_ILOGB]: ... to here.
1318 (expand_builtin): Expand BUILT_IN_ILOGB{,F,L} using
1319 expand_builtin_interclass_mathfn ().
1320 * config/i386/i386.md (fxtractxf3_i387): Rename from *fxtractxf3_i387.
1321 (ilogbsi2): Remove.
1322 (ilogbxf2, ilogb<mode>2): New expanders to implement ilogb, ilogbf and
1323 ilogbl built-in functions as x87 intrinsics.
1324
1325 2007-01-30 Richard Guenther <rguenther@suse.de>
1326
1327 PR middle-end/27657
1328 * dwarf2out.c (reference_to_unused): Query varpool if the
1329 variable was output.
1330
1331 2007-01-30 Richard Guenther <rguenther@suse.de>
1332
1333 PR middle-end/30313
1334 * passes.c (execute_one_pass): Reset in_gimple_form to not
1335 confuse non-unit-at-a-time mode.
1336
1337 2007-01-29 Roger Sayle <roger@eyesopen.com>
1338 Richard Guenther <rguenther@suse.de>
1339
1340 * fold-const.c (round_up): Make HIGH an unsigned HOST_WIDE_INT to
1341 avoid undefined behaviour on overflow. Use force_fit_type_double
1342 to construct the constant with the specified TREE_OVERFLOW.
1343
1344 2007-01-29 Janis Johnson <janis187@us.ibm.com>
1345
1346 * config/dfp-bit.c: Add parameterized support for fp exceptions.
1347 * config/dfp-bit.h: Ditto.
1348
1349 2007-01-29 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
1350
1351 * c-decl.c (pop_scope): Replace warnings with call to
1352 warn_for_unused_label.
1353 * c-common.h (warn_for_unused_label): Declare.
1354 * c-common.c (warn_for_unused_label): Define.
1355
1356 2007-01-29 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
1357
1358 * tree-optimize.c (update_inlined_to_pointers): Delete unused
1359 function.
1360
1361 2007-01-29 Janis Johnson <janis187@us.ibm.com>
1362
1363 * Makefile.in (USER_H): Remove decfloat.h.
1364 * ginclude/decfloat.h: Delete, moving contents to ...
1365 * ginclude/float.h: Add support for decimal floating point,
1366 guarded by __STDC_WANT_DEC_FP__.
1367
1368 2007-01-29 Mike Stump <mrs@apple.com>
1369
1370 * doc/gccint.texi (Top): Rename Loop Representation to Loop
1371 Analysis and Representation to resolve case insensitive conflict.
1372 * doc/loop.texi (Loop Analysis and Representation): Likewise.
1373
1374 2007-01-28 Daniel Berlin <dberlin@dberlin.org>
1375
1376 * tree.h (struct tree_memory_tag): Add aliases member.
1377 (MTAG_ALIASES): New macro.
1378 * tree-ssa-alias.c (alias_bitmap_obstack): New variable.
1379 (add_may_alias): Remove pointer-set. Update for may_aliases being
1380 a bitmap.
1381 (mark_aliases_call_clobbered): Update for may_aliases being a
1382 bitmap.
1383 (compute_tag_properties): Ditto.
1384 (create_partition_for): Ditto.
1385 (compute_memory_partitions): Ditto.
1386 (dump_may_aliases_for): Ditto.
1387 (is_aliased_with): Ditto.
1388 (add_may_alias_for_new_tag): Ditto.
1389 (rewrite_alias_set_for): Rewrite for may_aliases being a bitmap.
1390 (compute_is_aliased): New function.
1391 (compute_may_aliases): Call compute_is_aliased).
1392 (init_alias_info): Initialize alias_bitmap_obstack.
1393 (union_alias_set_into): New function.
1394 (compute_flow_sensitive_aliasing): Use union_aliases_into.
1395 (have_common_aliases_p): Rewrite to take two bitmaps and use
1396 intersection.
1397 (compute_flow_insensitive_aliasing): Stop using pointer-sets.
1398 Update for bitmaps.
1399 (finalize_ref_all_pointers): Update for add_may_alias changes.
1400 (new_type_alias): Ditto.
1401 * tree-flow-inline.h (may_aliases): Return a bitmap.
1402 * tree-dfa.c (dump_variable): Check for MTAG_P'ness.
1403 * tree-ssa.c (verify_flow_insensitive_alias_info): Update for
1404 may_aliases being a bitmap.
1405 * tree-flow.h (struct var_ann_d): Remove may_aliases member.
1406 may_aliases now returns a bitmap.
1407 * tree-ssa-structalias.c (merge_smts_into): Update for may_aliases
1408 being a bitmap.
1409 * tree-ssa-operands.c (add_virtual_operand): Update for
1410 may_aliases being a bitmap.
1411
1412 2007-01-29 Daniel Berlin <dberlin@dberlin.org>
1413
1414 PR tree-optimization/30630
1415 * tree-ssa-structalias.c (do_complex_constraint): Mark correct
1416 variable as changed.
1417
1418 2007-01-29 Simon Martin <simartin@users.sourceforge.net>
1419
1420 PR c++/28266
1421 * gimplify.c (gimplify_target_expr): Make sure that the TARGET_EXPR is
1422 expanded only once even if an error occurs.
1423
1424 2007-01-29 Ben Elliston <bje@au.ibm.com>
1425
1426 * gcov-io.h (__gcov_indirect_call_profiler): Declare.
1427 (__gcov_average_profiler): Likewise.
1428 (__gcov_ior_profiler): Likewise.
1429 (__gcov_merge_ior): Likewise.
1430
1431 2007-01-28 Jan Hubicka <jh@suse.cz>
1432
1433 * builtins.c (expand_builtin_memset): Fix typo in my last patch.
1434 * value-prof.c (stringop_block_profile): Likewise.
1435
1436 2007-01-28 Jan Hubicka <jh@suse.cz>
1437
1438 * expr.c (emit_block_move_via_movmem, emit_block_move_via_libcall): Add
1439 variant handling histograms; add wrapper.
1440 (clear_storage_via_libcall): Export.
1441 (emit_block_move_hints): Break out from ...; add histograms.
1442 (emit_block_move): ... this one.
1443 (clear_storage_hints): Break out from ...; add histograms.
1444 (clear_storage): ... this one.
1445 (set_storage_via_memset): Handle histogram.
1446 * expr.h (emit_block_move_via_libcall, emit_block_move_hints): Declare.
1447 (clear_storage_hints, clear_storage_via_libcall): Declare.
1448 (set_storage_via_setmem): Update prototype.
1449 * doc/md.texi (movmem, setmem): Document new arguments.
1450
1451 * value-prof.c (dump_histogram_value, tree_find_values_to_profile): Add
1452 new histograms.
1453 (stringop_block_profile): New global function.
1454 (tree_stringops_values_to_profile): Profile block size and alignment.
1455 * value-prof.h (enum hist_type): add HIST_TYPE_AVERAGE and
1456 HIST_TYPE_IOR.
1457 (struct profile_hooks): Add gen_average_profiler and gen_ior_profiler.
1458 (stringop_block_profile): Declare.
1459 * builtins.c: Include value-prof.h.
1460 (expand_builtin_memcpy, expand_builtin_memset): Pass block profile.
1461 * gcov-ui.h (GCOV_COUNTER_NAMES): Add new counter.
1462 (GCOV_COUNTER_AVERAGE, GCOV_COUNTER_IOR): New constants.
1463 (GCOV_COUNTERS, GCOV_LAST_VALUE_COUNTER): Update.
1464 * profile.c (instrument_values): Add new counters.
1465 * cfgexpand.c (expand_gimple_basic_block): Propagate histograms to
1466 calls.
1467 * tree-profile.c (tree_average_profiler_fn, tree_ior_profiler_fn): New.
1468 (tree_init_edge_profiler): Build new profilers.
1469 (tree_gen_average_profiler, tree_gen_ior_profiler): New.
1470 (pass_tree_profile): Add dump.
1471 (tree_profile_hooks): Update.
1472 * Makefile.in (LIBGCOV): Add new constants.
1473 * libgcov.c (__gcov_merge_ior, __gcov_average_profiler,
1474 __gcov_ior_profiler): New.
1475 * i386.md (movmem/setmem expanders): Add new optional arguments.
1476
1477 2007-01-28 David Edelsohn <edelsohn@gnu.org>
1478
1479 * doc/md.texi (Standard Pattern Names): Document blockage pattern.
1480
1481 2007-01-28 Zdenek Dvorak <dvorakz@suse.cz>
1482
1483 * tree-ssa-loop-unswitch.c: Include tree-inline.h.
1484 (tree_unswitch_single_loop): Pass eni_size_weights to
1485 tree_num_loop_insns.
1486 * tree-ssa-loop-manip.c: Include tree-inline.h.
1487 (can_unroll_loop_p): Pass eni_size_weights to
1488 tree_num_loop_insns.
1489 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p):
1490 Pass eni_size_weights to estimate_num_insns.
1491 * tree.h (init_inline_once): Export.
1492 * toplev.c (backend_init): Call init_inline_once.
1493 * cgraphunit.c (cgraph_process_new_functions,
1494 cgraph_analyze_function): Pass eni_inlining_weights to
1495 estimate_num_insns.
1496 * ipa-inline.c (compute_inline_parameters): Ditto.
1497 * tree-ssa-loop-ivcanon.c (tree_num_loop_insns): Pass weights
1498 to estimate_num_insns.
1499 (try_unroll_loop_completely): Pass eni_size_weights to
1500 tree_num_loop_insns.
1501 * tree-eh.c (decide_copy_try_finally): Pass eni_size_weights
1502 ot estimate_num_insns.
1503 * tree-ssa-loop-prefetch.c: Include tree-inline.h.
1504 (loop_prefetch_arrays): Pass eni_time_weights to tree_num_loop_insns.
1505 * tree-inline.c (eni_inlining_weights, eni_size_weights,
1506 eni_time_weights): New variables.
1507 (init_inline_once): Initialize them.
1508 (struct eni_data): Mew.
1509 (estimate_num_insns_1, estimate_num_insns): Use weights.
1510 * tree-inline.h (struct eni_weights_d): New.
1511 (eni_inlining_weights, eni_size_weights, eni_time_weights): Declare.
1512 (estimate_num_insns): Declaration changed.
1513 * cfgloop.h (tree_num_loop_insns): Declaration changed.
1514 * Makefile.in (tree-ssa-loop-unswitch.o, tree-ssa-loop-prefetch.o,
1515 tree-ssa-loop-manip.o): Add TREE_INLINE_H dependency.
1516
1517 2007-01-28 Zdenek Dvorak <dvorakz@suse.cz>
1518
1519 * tree-data-ref.c (conflict_fn): Assert that the number of affine
1520 relations in the conflict function is valid.
1521
1522 2007-01-27 Ian Lance Taylor <iant@google.com>
1523
1524 * common.opt: Add fstrict-overflow.
1525 * opts.c (decode_options): Set flag_strict_overflow if -O2.
1526 * flags.h (TYPE_OVERFLOW_WRAPS): Define.
1527 (TYPE_OVERFLOW_UNDEFINED): Define.
1528 (TYPE_OVERFLOW_TRAPS): Define. This replaces TYPE_TRAP_SIGNED.
1529 Replace all uses.
1530 * tree.h (TYPE_TRAP_SIGNED): Don't define.
1531 * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_UNDEFINED.
1532 (fold_negate_expr): Likewise.
1533 (make_range): Likewise.
1534 (extract_muldiv_1): Likewise.
1535 (maybe_canonicalize_comparison): Likewise.
1536 (fold_comparison): Likewise.
1537 (fold_binary): Likewise.
1538 (tree_expr_nonnegative_p): Likewise.
1539 (tree_expr_nonzero_p): Likewise.
1540 * tree-vrp.c (compare_values): Likewise.
1541 (extract_range_from_binary_expr): Likewise.
1542 (extract_range_from_unary_expr): Likewise.
1543 * tree-ssa-loop-niter.c (infer_loop_bounds_from_signedness):
1544 Likewise.
1545 (nowrap_type_p): Likewise.
1546 * tree-scalar-evolution.c (simple_iv): Likewise.
1547 * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_WRAPS.
1548 (build_range_check): Likewise.
1549 (extract_muldiv_1): Likewise.
1550 (fold_comparison): Likewise.
1551 * tree-vrp.c (vrp_int_const_binop): Likewise.
1552 (extract_range_from_unary_expr): Likewise.
1553 * convert.c (convert_to_integer): Likewise.
1554 * fold-const.c (fold_negate_expr): Use TYPE_OVERFLOW_TRAPS.
1555 (fold_comparison): Likewise.
1556 (fold_binary): Likewise.
1557 * optabs.c (optab_for_tree_code): Likewise.
1558 * tree-vectorizer.c (vect_is_simple_reduction): Likewise.
1559 * simplify-rtx.c (simplify_const_relational_operation): Check
1560 flag_strict_overflow and flag_trapv.
1561 (simplify_const_relational_operation): Likewise.
1562 * doc/invoke.texi (Option Summary): Mention -fstrict-overflow.
1563 (Optimize Options): Add -fstrict-overflow to -O2 list. Document
1564 -fstrict-overflow.
1565
1566 2007-01-27 Roger Sayle <roger@eyesopen.com>
1567
1568 * tree.c (tree_fold_gcd): Delete.
1569 * tree.h (tree_fold_gcd): Remove prototype.
1570 * tree-data-ref.c (tree_fold_divides_p): Don't use tree_fold_gcd to
1571 test whether one constant integer is a multiple of another. Instead
1572 call int_const_binop with TRUNC_MOD_EXPR and test for a zero result.
1573 * fold-const.c (multiple_of_p): We've determined both TOP and
1574 BOTTOM are integer constants so we can call int_const_binop directly
1575 instead of the more generic const_binop.
1576
1577 2007-01-27 Roger Sayle <roger@eyesopen.com>
1578
1579 * fold-const.c (size_binop): In the fast-paths for X+0, 0+X, X-0 and
1580 1*X check that the constant hasn't overflowed, to preserve the
1581 TREE_OVERFLOW bit.
1582 (round_up): Provide an efficient implementation when rouding-up an
1583 INTEGER_CST to a power-of-two.
1584
1585 2007-01-28 Ralf Wildenhues <Ralf.Wildenhues@gmx.de>
1586
1587 * doc/sourcebuild.texi: Add comma for clarity.
1588 * doc/extend.texi: Fix some typos.
1589 * doc/passes.texi: Likewise.
1590 * doc/cppinternals.texi: Likewise.
1591 * doc/c-tree.texi: Likewise.
1592 * doc/tree-ssa.texi: Likewise.
1593 * doc/install.texi: Likewise.
1594
1595 2007-01-27 Jan Hubicka <jh@suse.cz>
1596
1597 * tree-sra.c (sra_walk_function): Don't rely on aliases being build.
1598 (pass_sra): Do not require alias information.
1599 * passes.c (init_optimization_passes): Add SRA
1600
1601 2007-01-27 Steven Bosscher <steven@gcc.gnu.org>
1602
1603 * tracer.c (rest_of_handle_tracer): We already cleaned
1604 up the CFG in tracer() so don't do it here again.
1605 * cfgcleanup.c (rest_of_handle_jump2): Don't repeat
1606 cleanup_cfg here, either. And don't call renumber_insns.
1607
1608 * cfgrtl.c (rtl_verify_flow_info_1): Don't verify that BB_END
1609 and BB_HEAD are in the insn stream here. Instead make sure
1610 that BB_INSN is valid on all insns. Also, do check here that
1611 there are no pending branch predictions...
1612 (rtl_verify_flow_info): ...instead of doing it here. Checks
1613 for BB_END and BB_HEAD moved from rtl_verify_flow_info_1 to
1614 here.
1615
1616 2007-01-26 Roger Sayle <roger@eyesopen.com>
1617
1618 * config/i386/i386.c (ix86_swap_binary_operands_p): New helper
1619 function to simplify/factorize operand order canonicalization.
1620 (ix86_fixup_binary_operands): Reorganize using the above function.
1621 (ix86_binary_operator_ok): Likewise.
1622
1623 2007-01-27 Jakub Jelinek <jakub@redhat.com>
1624
1625 * genattrtab.c (struct attr_value_list, insn_code_values): Move to
1626 file scope from optimize_attrs.
1627 (simplify_test_exp): If insn_code_values is not NULL, use it to speed
1628 up search.
1629 (optimize_attrs): Clear insn_code_values after freeing it.
1630
1631 2007-01-26 Zdenek Dvorak <dvorakz@suse.cz>
1632
1633 * tree-ssa-address.c (create_mem_ref): Remove ", bsi" from
1634 a parts.base assignment.
1635
1636 2007-01-26 Zdenek Dvorak <dvorakz@suse.cz>
1637
1638 * tree-data-ref.c (dump_subscript): Use dump_conflict_function.
1639 (compute_subscript_distance, initialize_data_dependence_relation,
1640 finalize_ddr_dependent, analyze_ziv_subscript,
1641 analyze_siv_subscript_cst_affine,
1642 compute_overlap_steps_for_affine_univar,
1643 compute_overlap_steps_for_affine_1_2, analyze_subscript_affine_affine,
1644 analyze_siv_subscript, analyze_miv_subscript,
1645 analyze_overlapping_iterations, subscript_dependence_tester_1,
1646 compute_self_dependence, free_dependence_relation): Work
1647 with affine_fn instead of chrecs.
1648 (dump_affine_function, dump_conflict_function, affine_function_equal_p,
1649 common_affine_function, affine_function_base,
1650 affine_function_constant_p, affine_fn_op, affine_fn_plus,
1651 affine_fn_minus, affine_fn_free, conflict_fn_not_known,
1652 conflict_fn_no_dependence, free_conflict_function, free_subscripts,
1653 conflict_fn, affine_fn_cst, affine_fn_univar): New functions.
1654 (all_chrecs_equal_p): Removed.
1655 * tree-data-ref.h (affine_fn, conflict_function): New types.
1656 (struct subscript): Change type of conflicting_iterations_in_a
1657 and conflicting_iterations_in_b.
1658
1659 2007-01-26 Steve Ellcey <sje@cup.hp.com>
1660
1661 PR other/30182
1662 * config/pa/pa.h (TARGET_HPUX_11): New.
1663 * config/pa/pa-hpux11.h (TARGET_HPUX_11): New.
1664 * config/pa/pa.c (pa_init_builtins): Use TARGET_HPUX_11.
1665
1666 2007-01-26 Daniel Berlin <dberlin@dberlin.org>
1667 Richard Guenther <rguenther@suse.de>
1668
1669 * tree-ssa-structalias.c (solve_graph): Handle case
1670 we merged the variable to another.
1671
1672 2007-01-25 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1673
1674 * builtins.c (fold_builtin_1): Treat ccos and ccosh as 'even'
1675 functions.
1676
1677 * fold-const.c (negate_mathfn_p): Treat casin, casinh, catan,
1678 catanh, cproj, csin, csinh, ctan and ctanh as 'odd' functions.
1679
1680 2007-01-25 DJ Delorie <dj@redhat.com>
1681
1682 * config/m32c/m32c.c (m32c_cannot_change_mode_class): We don't
1683 allow changes to modes which don't fit in those registers.
1684
1685 * reload1.c (choose_reload_regs): Check for invalid subregs before
1686 computing their locations, not after.
1687
1688 2007-01-25 Geoffrey Keating <geoffk@apple.com>
1689
1690 PR 25127
1691 * config/rs6000/rs6000.c (first_altivec_reg_to_save): On Darwin,
1692 save Altivec registers in an eh_return function.
1693 (compute_vrsave_mask): Likewise.
1694 (rs6000_stack_info): Correct AIX/Darwin stack alignment computation
1695 for saving Altivec registers.
1696 (rs6000_emit_prologue): Don't allocate stack twice in
1697 eh_return function. Correct expected value of altivec_save_offset
1698 when using save_world. Describe save of R0 to stack when using
1699 save_world. Describe stack pointer adjustment when using
1700 save_world. Remove duplicated eh_return parameter register saving.
1701 Update sp_offset variable after save_world.
1702 * config/rs6000/t-darwin (LIB2FUNCS_STATIC_EXTRA): Remove
1703 darwin-world.asm.
1704 (LIB2FUNCS_EXTRA): Add darwin-world.asm.
1705 * config/rs6000/darwin.h (SUBTARGET_OVERRIDE_OPTIONS): -m64
1706 implies Altivec.
1707
1708 2007-01-25 Steve Ellcey <sje@cup.hp.com>
1709
1710 * config.gcc (ia64*-*-hpux*): Make posix threads the default.
1711
1712 2007-01-25 Steve Ellcey <sje@cup.hp.com>
1713
1714 PR other/30182
1715 * config/pa/pa.c (pa_init_builtins): Set asm names for finite routines.
1716 * config/ia64/ia64.c (ia64_init_builtins): Ditto.
1717
1718 2007-01-25 Richard Guenther <rguenther@suse.de>
1719
1720 * doc/invoke.texi (-Wcoverage-mismatch): Document.
1721 * common.opt (-Wcoverage-mismatch): New warning option.
1722 * coverage.c (get_coverage_counts): Ignore coverage mismatch
1723 if -Wcoverage-mismatch is given.
1724
1725 2007-01-25 Razya Ladelsky <razya@il.ibm.com>
1726
1727 * ipa-cp.c (ipcp_insert_stage, ipcp_driver): Support for SSA.
1728 (ipcp_driver): Change to static definition.
1729 Add dumping of the ifunctions.
1730 (constant_val_insert): Remove unused parameter. Support for SSA.
1731 (ipcp_propagate_const): Support for SSA.
1732 (ipcp_profile_bb_print): Print only analyzed nodes.
1733 (ipcp_replace_map_create): Remove support for Fortran constant
1734 for now.
1735 * ipa-prop.c (ipa_method_modify_stmt,
1736 ipa_callsite_compute_param): Support for SSA.
1737 * ipa-prop.h (ipcp_driver): Remove declaration.
1738 (IS_VALID_TREE_MAP_INDEX): Add define.
1739
1740 2007-01-24 Geoffrey Keating <geoffk@apple.com>
1741
1742 * unwind-dw2.c (execute_stack_op): Handle DW_OP_swap.
1743
1744 2007-01-24 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1745
1746 PR middle-end/30447
1747 * builtins.c (fold_builtin_cabs): Use MPFR to evaluate a
1748 constant argument to cabs and do it without checking for
1749 -funsafe-math-optimizations.
1750
1751 2007-01-24 Douglas Gregor <dgregor@osl.iu.edu>
1752
1753 * c-common.h (RID_FIRST_CXX0X): New.
1754 (RID_LAST_CXX0X): New.
1755 * c-opts.c (c_common_handle_option): -Wc++0x-compat is triggered
1756 by -Wall.
1757 * c.opt (Wc++0x-compat): New.
1758 * doc/invoke.texi (-Wc++0x-compat): Document.
1759
1760 2007-01-24 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1761
1762 * builtins.c (fold_builtin_carg): New.
1763 (fold_builtin_1): Use it.
1764
1765 2007-01-24 Jan Hubicka <jh@suse.cz>
1766
1767 * ipa-inline.c (cgraph_decide_inlining): Initialize initial_insns.
1768
1769 * ipa-inline.c (initial_insns, max_insns): Delete.
1770 (compute_max_insns): New function.
1771 (cgraph_decide_inlining_of_small_function): Use it; take minimal amount
1772 of insns as base for code growth.
1773 (cgraph_decide_inlining): Make initial_insns local; do not compute
1774 max_insns.
1775 * params.def (PARAM_INLINE_UNIT_GROWTH): Set to 60.
1776 * doc/invoke.texi (inline-unit-growth): Update docs.
1777
1778 2007-01-24 Jakub Jelinek <jakub@redhat.com>
1779
1780 * config/i386/i386.h (x86_cmpxchg16b): Remove const.
1781 (TARGET_CMPXCHG16B): Define to x86_cmpxchg16b.
1782 * config/i386/i386.c (x86_cmpxchg16b): Remove const.
1783 (override_options): Add PTA_CX16 flag. Set x86_cmpxchg16b
1784 for CPUs that have PTA_CX16 set.
1785
1786 PR middle-end/27416
1787 * gimplify.c (omp_check_private): New function.
1788 (gimplify_scan_omp_clauses): Use it for
1789 firstprivate/lastprivate/reduction.
1790
1791 PR middle-end/30494
1792 * gimplify.c (omp_add_variable): Don't call omp_notice_variable
1793 on TYPE_SIZE_UNIT for GOVD_LOCAL VLAs.
1794
1795 PR middle-end/30421
1796 * omp-low.c (lower_omp_for_lastprivate): Add dlist argument.
1797 If lower_lastprivate_clauses emits some statements, append them
1798 to dlist rather than body_p and to body_p append an initializer.
1799 (lower_omp_for): Adjust caller.
1800
1801 2007-01-24 Steve Ellcey <sje@cup.hp.com>
1802
1803 * target.h (globalize_decl_name): New.
1804 * target-def.h (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
1805 * output.h (default_globalize_decl_name): New.
1806 * varasm.c (asm_output_bss): Use globalize_decl_name instead of
1807 globalize_label.
1808 (globalize_decl): Ditto.
1809 (default_globalize_decl_name): New.
1810 * config/ia64/ia64.c (ia64_globalize_decl_name): New.
1811 (ia64_handle_version_id_attribute): New.
1812 (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
1813 (ia64_asm_output_external): Use globalize_decl_name instead
1814 of globalize_label.
1815 * doc/extend.texi (version_id): New pragma.
1816 * doc/tm.texi (ARGET_ASM_GLOBALIZE_DECL_NAME): New target hook.
1817
1818 2007-01-24 Andreas Krebbel <krebbel1@de.ibm.com>
1819
1820 * unwind-dw2-fde.c (get_cie_encoding): Replaced _Unwind_Word with
1821 _uleb128_t and _Unwind_SWord with _sleb128_t.
1822 * unwind-dw2.c (extract_cie_info, execute_stack_op, execute_cfa_program,
1823 uw_frame_state_for, uw_update_context_1): Likewise.
1824 * unwind-c.c (parse_lsda_header, PERSONALITY_FUNCTION): Likewise.
1825 * unwind-pe.h (read_uleb128, read_sleb128,
1826 read_encoded_value_with_base): Likewise.
1827 * unwind-generic.h: Define _sleb128_t and _uleb128_t types.
1828
1829 2007-01-24 Richard Guenther <rguenther@suse.de>
1830
1831 * builtins.c (expand_builtin_cexpi): Get the fndecl
1832 for cexp in the correct way.
1833
1834 2007-01-24 Jan Hubicka <jh@suse.cz>
1835
1836 * tree-ssa-dce.c (eliminate_unnecesary_stmts): Remove dead LHS of calls.
1837
1838 2007-01-24 Andreas Krebbel <krebbel1@de.ibm.com>
1839
1840 * c-cppbuiltin.c (builtin_define_type_sizeof): New function.
1841 (c_cpp_builtins): New builtin macros: __SIZEOF_INT__, __SIZEOF_LONG__,
1842 __SIZEOF_LONG_LONG__, __SIZEOF_SHORT__, __SIZEOF_POINTER__,
1843 __SIZEOF_FLOAT__, __SIZEOF_DOUBLE__, __SIZEOF_LONG_DOUBLE__,
1844 __SIZEOF_SIZE_T__, __SIZEOF_WCHAR_T__, __SIZEOF_WINT_T__ and
1845 __SIZEOF_PTRDIFF_T__.
1846 * doc/cpp.texi: Documentation for the new builtin macros added.
1847
1848 2007-01-24 Uros Bizjak <ubizjak@gmail.com>
1849
1850 * config/i386/i386.md (tanxf2, tan<mode>2, atan<mode>2, log<mode>2,
1851 log10<mode>2, log2<mode>2, expxf2, exp10xf2, exp2xf2): Use op2
1852 instead of operands[2] to avoid access past the end of array.
1853
1854 2007-01-24 Richard Sandiford <richard@codesourcery.com>
1855
1856 * reload1.c (emit_reload_insns): Pass the reload register
1857 for a non-spill output reload through forget_old_reloads_1.
1858
1859 2007-01-23 Joseph Myers <joseph@codesourcery.com>
1860
1861 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Define
1862 _SOFT_DOUBLE if doubles use software floating-point.
1863 * config/rs6000/libgcc-ppc-glibc.ver: Export additional long
1864 double functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
1865 * config/rs6000/darwin-ldouble.c: Also compile functions for
1866 hard-float without FPRs. Use fmsub function for all __NO_FPRS__
1867 cases. Compile extra functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
1868 * config/rs6000/linuxspe.h (SUBSUBTARGET_OVERRIDE_OPTIONS): Remove
1869 commented-out long double override.
1870 (CPP_LONGDOUBLE_DEFAULT_SPEC): Likewise.
1871 * config/rs6000/eabispe.h: Likewise.
1872 * config/rs6000/rs6000.c (rs6000_override_options): Don't override
1873 long double for non-SPE.
1874 (rs6000_handle_option): Likewise.
1875 (invalid_e500_subreg): Disallow more subregs involding DImode,
1876 DFmode, TImode or TFmode.
1877 (rs6000_legitimate_offset_address_p): Check TFmode offsets for
1878 E500 double.
1879 (legitimate_lo_sum_address_p): Also check for TFmode for E500
1880 double.
1881 (rs6000_legitimize_address): Also handle TFmode for E500 double.
1882 (rs6000_legitimize_reload_address): Also handle TFmode for E500
1883 double.
1884 (rs6000_legitimate_address): Also check for TFmode for E500
1885 double.
1886 (rs6000_emit_move): Use DFmode subregs of TFmode for E500 double.
1887 (spe_build_register_parallel): Handle TFmode and TCmode.
1888 (rs6000_spe_function_arg): Handle TFmode and TCmode for E500
1889 double.
1890 (function_arg): Handle TFmode and TCmode for E500 double.
1891 (rs6000_init_libfuncs): Initialize extra libfuncs for soft double
1892 in general.
1893 (print_operand): Handle TFmode and TImode for %y.
1894 (rs6000_generate_compare): Handle TFmode comparisons for E500
1895 double.
1896 (spe_func_has_64bit_regs_p): Check for TFmode for E500 double.
1897 (rs6000_function_value): Handle TFmode and TCmode for E500 double.
1898 (rs6000_libcall_value): Handle TFmode and TCmode for E500 double.
1899 * config/rs6000/rs6000.h (CANNOT_CHANGE_MODE_CLASS): Check for
1900 TFmode for E500 double.
1901 * config/rs6000/rs6000.md (FP): Allow TF for E500 double.
1902 (floatsidf2): Enable for E500 double.
1903 (movtf_softfloat): Use rs6000_nonimmediate_operand.
1904 (extenddftf2): Change to extenddftf2_fprs.
1905 (extenddftf2): Call gen_spe_extenddftf2 or gen_extenddftf2_fprs
1906 depending on TARGET_E500_DOUBLE.
1907 (extendsftf2): Enable for E500 double.
1908 (trunctfdf2): Enable for E500 double.
1909 (trunctfsf2): Change to trunctfsf2_fprs.
1910 (trunctfsf2): Call gen_spe_trunctfsf2 or gen_trunctfsf2_fprs
1911 depending on TARGET_E500_DOUBLE.
1912 (floatsitf2): Enable for E500 double.
1913 (fix_trunctfsi2): Change to fix_trunctfsi2_fprs.
1914 (fix_trunctfsi2): Call gen_spe_fix_trunctfsi2 or
1915 gen_fix_trunctfsi2_fprs depending on TARGET_E500_DOUBLE.
1916 (negtf2): Change to negtf2_internal.
1917 (negtf2): New expander.
1918 (abstf2): Enable for E500 double. Call gen_spe_abstf2_tst,
1919 gen_spe_abstf2_cmp or gen_abstf2_internal depending on
1920 TARGET_E500_DOUBLE and flag_unsafe_math_optimizations.
1921 (movdi_internal32): Use rs6000_nonimmediate_operand.
1922 (unnamed splitter): Likewise.
1923 * config/rs6000/spe.md (CMPTFEQ_GPR, TSTTFEQ_GPR, CMPTFGT_GPR,
1924 TSTTFGT_GPR, CMPTFLT_GPR, TSTTFLT_GPR): New unspecs.
1925 (SPE64TF, DITI): New mode macros.
1926 (frob_df_di): Change to frob_<SPE64:mode>_<DITI:mode>; allow more
1927 modes.
1928 (frob_tf_ti): New.
1929 (frob_<mode>_di_2): New.
1930 (frob_tf_di_8_2): New.
1931 (frob_di_df): Change to frob_di_<mode>; allow more modes.
1932 (frob_ti_tf): New.
1933 (frob_di_df_2): Change to frob_<DITI:mode>_<SPE64:mode>_2; allow
1934 more modes.
1935 (frob_ti_<mode>_8_2): New.
1936 (frob_ti_tf_2): New.
1937 (mov_si<mode>_e500_subreg0, mov_si<mode>_e500_subreg0_2,
1938 mov_si<mode>_e500_subreg4, mov_si<mode>_e500_subreg4_2): Allow
1939 TFmode.
1940 (mov_sitf_e500_subreg8, mov_sitf_e500_subreg8_2,
1941 mov_sitf_e500_subreg12, mov_sitf_e500_subreg12_2): New.
1942 (spe_trunctfdf2_internal1, spe_trunctfsf2, spe_extenddftf2,
1943 spe_fix_trunctfsi2, spe_fix_trunctfsi2_internal,
1944 spe_negtf2_internal, spe_abstf2_cmp, spe_abstf2_tst): New.
1945 (cmptfeq_gpr, tsttfeq_gpr, cmptfgt_gpr, tsttfgt_gpr, cmptflt_gpr,
1946 tsttflt_gp): New.
1947
1948 2007-01-23 Ian Lance Taylor <iant@google.com>
1949
1950 * Makefile.in (OBJS-common): Reformat, alphabetize, but put
1951 insn-*.o first.
1952 (OBJS-archive): Reformat, alphabetize.
1953 (OBJS): Change out_object_file to OBJS-md.
1954
1955 2007-01-23 Uros Bizjak <ubizjak@gmail.com>
1956
1957 * config/i386/i386.md: Use REG_P, MEM_P, CONST_INT_P, LABEL_P,
1958 JUMP_P and CALL_P predicates where applicable.
1959 * config/i386/i386.c: Ditto.
1960 * config/i386/i386.md: Ditto.
1961 * config/i386/mmx.md: Ditto.
1962 * config/i386/predicates.md: Ditto.
1963
1964 2007-01-22 Andreas Schwab <schwab@suse.de>
1965
1966 * config/m68k/m68k.h: Fix comment.
1967
1968 2007-01-22 Jan Hubicka <jh@suse.cz>
1969
1970 * passes.c (init_optimization_passes): Do not rerun
1971 pass_early_warn_uninitialized.
1972
1973 2007-01-22 Richard Guenther <rguenther@suse.de>
1974
1975 PR tree-optimization/30038
1976 * tree-ssa-math-opts.c (maybe_record_sincos): New static helper
1977 function.
1978 (execute_cse_sincos_1): Likewise.
1979 (execute_cse_sincos): Likewise.
1980 (gate_cse_sincos): Likewise.
1981 (pass_cse_sincos): New pass CSEing sin() and cos() calls using
1982 the cexpi() canonicalization of sincos().
1983 * tree-pass.h (pass_cse_sincos): Declare.
1984 * passes.c (init_optimization_passes): New pass pas_cse_sincos.
1985
1986 2007-01-21 Eric Botcazou <ebotcazou@libertysurf.fr>
1987
1988 PR rtl-optimization/29329
1989 * combine.c (replaced_rhs_insn): Rename to i2mod.
1990 (replaced_rhs_value): Rename to i2mod_new_rhs.
1991 (i2mod_old_rhs): New global variable.
1992 (combine_instructions): Adjust for above change. Save a copy of
1993 the old RHS into i2mod_old_rhs when the contents of a REG_EQUAL
1994 note are substituted in the second instruction.
1995 (distribute_notes) <REG_DEAD>: Adjust for above change. Do not
1996 ditch the note if it pertains to the second eliminated register
1997 and this register is mentioned in i2mod_old_rhs.
1998
1999 Revert:
2000 2006-09-12 Eric Botcazou <ebotcazou@libertysurf.fr>
2001
2002 * combine.c (distribute_notes) <REG_DEAD>: Do not consider SETs past
2003 the insn to which the note was originally attached.
2004
2005 2007-01-21 Jan Hubicka <jh@suse.cz>
2006
2007 * ipa-inline.c (inlining_mode): Comment, move up.
2008 (cgraph_decide_inlining_incrementally): Do not perform inlining
2009 itself; fix handling of flattening of self recursive functions.
2010 (cgraph_find_cycles): Remove.
2011 (cgraph_flatten_node): Remove.
2012 (cgraph_decide_inlining): Use incremental inliner to handle flattening.
2013 (try_inline): New function.
2014 (cgraph_early_inlining): Update call of
2015 cgraph_decide_inlining_incrementally. Apply inlining here.
2016 (apply_inline): Update call of cgraph_decide_inlining_incrementally.
2017
2018 2007-01-21 Dirk Mueller <dmueller@suse.de>
2019
2020 PR bootstrap/30511
2021 * tree-vrp.c (check_array_bounds): do not warn
2022 about ADDR_EXPR's of ARRAY_REF's which are immediately
2023 used in binary expressions.
2024
2025 2007-01-21 Ira Rosen <irar@il.ibm.com>
2026
2027 * tree-vectorizer.h (struct _stmt_vec_info): Add new field
2028 read_write_dep and macros for its access.
2029 * tree-vectorizer.c (new_stmt_vec_info): Initialize the new field.
2030 * tree-vect-analyze.c (vect_analyze_data_ref_dependence): Remove
2031 argument, call vect_check_interleaving for every independent pair of
2032 data-refs. Mark loads that access the same memory location as a store
2033 in the loop.
2034 (vect_check_dependences): Remove.
2035 (vect_analyze_data_ref_dependences): Remove vect_check_dependences
2036 call, fix the call to vect_analyze_data_ref_dependence.
2037 (vect_analyze_data_ref_access): For statements that access the same
2038 data-ref, check that they are not stores; for loads, check that there
2039 is no store that access the same location.
2040
2041 2007-01-20 Roger Sayle <roger@eyesopen.com>
2042 Joseph Myers <joseph@codesourcery.com>
2043
2044 * doc/invoke.texi (-fdump-rtl-combine): Fix under/overfull hbox.
2045 (-fdump-rtl-stack): Likewise.
2046 (-fno-signed-zeros): Use @minus{} for a minus sign. Correct typo.
2047 (-mcheck-zero-division, -mcpu): Fix under/overfull hbox.
2048 (-mpt-fixed): Use @minus{} for minus sign.
2049 (Using Precompiled Headers): Fix under/overfull hbox.
2050
2051 2007-01-20 Jan Hubicka <jh@suse.cz>
2052
2053 * tree-flow.h (struct stmt_ann_d): Move references_memory to proper
2054 place within annotation.
2055
2056 2007-01-20 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
2057
2058 * pa.c (output_move_double): Change array size of xoperands to 4.
2059
2060 2007-01-20 Jan Hubicka <jh@suse.cz>
2061
2062 * tree-tailcall.c (adjust_return_value): Do not use RESULT_DECL
2063 as temporary.
2064
2065 2007-01-19 Ian Lance Taylor <iant@google.com>
2066
2067 * expmed.c (expand_divmod) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]: Cast
2068 constant to unsigned HOST_WIDE_INT before negating.
2069
2070 2007-01-19 Ian Lance Taylor <iant@google.com>
2071
2072 * tree-ssa-operands.h (struct vuse_vec_d): Change num_vuse field
2073 to unsigned.
2074 (VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Use unsigned comparison.
2075 (VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
2076 (SET_VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Likewise.
2077 (SET_VUSE_ELEMENT_VAR) [ENABLE_CHECKING]: Likewise.
2078 (SET_VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
2079 (realloc_vdef, realloc_vuse): Change second parameter to
2080 unsigned.
2081 (ssa_operand_iterator_d): Change vuse_index and mayuse_index
2082 fields to unsigned.
2083 * tree-ssa-operands.c (realloc_vop): Change num_elem parameter to
2084 unsigned. Change x and lim locals to unsigned.
2085 (realloc_vdef, realloc_vuse): Change num_elem parameter to
2086 unsigned.
2087 (finalize_ssa_vuse_ops): Change old_i local to unsigned.
2088 (copy_virtual_operands): Change i and n locals to unsigned.
2089
2090 2007-01-19 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
2091
2092 PR middle-end/29335
2093 * builtins.c (fold_builtin_1): Handle builtin fdim.
2094
2095 2007-01-20 Jan Hubicka <jh@suse.cz>
2096
2097 * tree-ssa.c (init_tree_ssa): Do not call init_alias_heapvars.
2098 * tree-ssa-structalias.c (compute_points_to_sets): Do call
2099 init_alias_heapvars.
2100 (init_alias_heapvars): Initialize only when not already initialized.
2101 (delete_alias_heapvars): Set heapvar_for_stmt to NULL.
2102
2103 2007-01-19 Roger Sayle <roger@eyesopen.com>
2104
2105 * common.opt (fsigned-zeros): New command line option.
2106 * flags.h (HONOR_SIGNED_ZEROS): Control via flag_signed_zeros instead
2107 of flag_unsafe_math_optimizations.
2108 * opts.c (set_fast_math_flags): The -ffast-math command line option
2109 implies -fno-signed-zeros.
2110 (fast_math_flags_set_p): Likewise.
2111
2112 * doc/invoke.texi: Document new -fno-signed-zeros option, and update
2113 the documentation of -ffast-math appropriately. Wrap long lines.
2114
2115 2007-01-19 Steve Ellcey <sje@cup.hp.com>
2116
2117 * system.h (ASM_MAKE_LABEL_LINKONCE): Poison.
2118 * varasm.c (globalize_decl): Remove ASM_MAKE_LABEL_LINKONCE ifdef.
2119
2120 2007-01-19 Tomas Bily <tbily@suse.cz>
2121
2122 * cgraphunit.c (cgraph_finalize_function): Updating of pid
2123 * tree-profile.c:
2124 (tree_init_ic_make_global_vars): New function
2125 (tree_init_edge_profiler): call of tree_init_ic_make_global_vars
2126 (tree_gen_ic_profiler): New function
2127 (tree_gen_ic_func_profiler): New function
2128 (tree_profiling): Added calling of tree_gen_ic_func_profiler
2129 (tree_profile_hooks): Added hook for indirec/virtual calls
2130 * value-prof.c (tree_find_values_to_profile): New case for
2131 indirect calls
2132 (tree_values_to_profile): Call for determining indirect/virtual
2133 counters
2134 (tree_indirect_call_to_profile): New function
2135 (tree_ic_transform): New function
2136 (tree_ic): New function
2137 (find_func_by_pid): New function
2138 (init_pid_map): New function
2139 (tree_value_profile_transformations): Added check for
2140 indirect/virtual call transformation
2141 * value-prof.h (enum hist_type): New counter type for
2142 indirect/virtual calls
2143 (profile_hooks): Added new hook for profiling indirect/virtual
2144 calls
2145 * profile.c (instrument_values): New case for indirect/virtual
2146 call added
2147 * gcov-io.h (GCOV_LAST_VALUE_COUNTER): Changed to 6
2148 (GCOV_COUNTER_V_INDIR): New counter type
2149 (GCOV_COUNTER_NAMES): New name of counter "indirect" added
2150 (GCOV_MERGE_FUNCTIONS): New merge function for indirect/virtual
2151 call added
2152 * cgraph.c: Definition of cgraph_max_pid
2153 (cgraph_create_node): Default init of pid attribute
2154 * cgraph.h: Declaration of cgraph_max_pid
2155 (struct cgraph_node): Added pid attribute
2156 * libgcov.c (__gcov_indirect_call_profiler): New function
2157 (__gcov_one_value_profiler_body): New function
2158 (__gcov_one_value_profiler): Body was moved to
2159 __gcov_one_value_profiler_body and calls it
2160
2161 2007-01-19 Basile Starynkevitch <basile@starynkevitch.net>
2162
2163 * doc/gty.texi (Options): Document the mark_hook option to GTY.
2164 * gengtype.c (write_types_data, write_func_for_structure,
2165 write_types, ggc_wtd, pch_wtd): Add skip_hooks to
2166 write_types_data, ggc_wtd, pch_wtd for processing mark_hook.
2167 (walk_type, write_func_for_structure): Generate the mark_hook if
2168 needed.
2169
2170 2007-01-19 Jan Hubicka <jh@suse.cz>
2171
2172 * ipa-inline.c (cgraph_decide_inlining_incrementally): Instead of
2173 'early' argument take inlining mode argument specifying whether to
2174 inline for size/speeed or all functions; add support for flattening;
2175 improve dumpting.
2176 (cgraph_early_inlining): Update call of decide_inlining_incrementally.
2177
2178 2007-01-19 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
2179
2180 PR c++/17947
2181 * toplev.c (warn_deprecated_use): Use %qD instead of %qs to print
2182 the name of the declared identifier.
2183
2184 2007-01-19 Dirk Mueller <dmueller@suse.de>
2185
2186 * config/i386.h (CONDITIONAL_REGISTER_USAGE): Store
2187 result of PIC_OFFSET_TABLE_REGNUM in temporary variable to avoid
2188 duplicate evaluation.
2189
2190 2007-01-19 Uros Bizjak <ubizjak@gmail.com>
2191
2192 * config/i386/i386.md (acos<mode>2): Rename from acossf2 and acosdf2.
2193 Macroize expander using X87MODEF12 mode macro. Extend operand 1
2194 to XFMode, use acosxf2 and truncate result to requested mode.
2195 Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
2196 (asin<mode>2): Similarly, with asin expanders.
2197 (*fscalexf4_i387): Rename from *fscalexf4.
2198 (expNcorexf3): New expander.
2199 (expxf2, exp10xf2, exp2xf2): Use expNcorexf3 expander.
2200 (exp<mode>2): Rename from expsf2 and expdf2. Macroize expander using
2201 X87MODEF12 mode macro. Extend operand 1 to XFMode, use expxf2 and
2202 truncate result to requested mode. Use SSE_FLOAT_MODE_P to disable
2203 patterns for SSE math.
2204 (exp10<mode>2): Similarly, with exp10 expanders.
2205 (exp2<mode>2): Similarly, with exp2 expanders.
2206 (expm1<mode>2): Similarly, with expm1 expanders.
2207 (ldexp<mode>3): Similarly, with ldexp expanders.
2208 (log<mode>2, log10<mode>2, log2<mode>2, log1p<mode>2, logb<mode>2):
2209 Use gen_truncxf<mode>2_i387_noop to truncate result.
2210
2211 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2212
2213 * config/m68k/m68k.h (PREFERRED_STACK_BOUNDARY): Define to 32
2214 for ColdFire targets.
2215
2216 2007-01-19 Nathan Sidwell <nathan@codesourcery.com>
2217 Richard Sandiford <richard@codesourcery.com>
2218
2219 * config/m68k/m68k.h (M68K_STATIC_CHAIN_REG_NAME): New macro.
2220 (INITIALIZE_TRAMPOLINE): Use STATIC_CHAIN_REGNUM.
2221 (__transfer_from_trampoline): Use M68K_STATIC_CHAIN_REG_NAME.
2222 * config/m68k/m68kelf.h (STATIC_CHAIN_REGNUM): Override.
2223 (M68K_STATIC_CHAIN_REG_NAME): Likewise.
2224 * config/m68k/netbsd-elf.h (M68K_STATIC_CHAIN_REG_NAME): Likewise.
2225
2226 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2227
2228 * config/m68k/m68k.md (adddi_dilshr32): Rename to...
2229 (*adddi_dilshr32): ...this. Fix formatting. Remove commented-out
2230 non-canonical pattern. Restrict to !TARGET_COLDFIRE.
2231 (*adddi_dilshr32_cf): New pattern.
2232 (adddi3, subdi3): Remove first alternatives.
2233
2234 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2235
2236 * config/m68k/m68k.c (notice_update_cc): If an SFmode move is
2237 implemented using move.l, do not use its cc result for floating-point
2238 comparisons.
2239
2240 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2241
2242 * config/m68k/m68k.h (EXTRA_CONSTRAINT): Stop the 'T' constraint
2243 from accepting 's' constraints if flag_pic.
2244
2245 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2246
2247 * config/m68k/m68k.md (bordered, bunordered, buneq, bunge, bungt)
2248 (bunle, bunlt, bltgt, bordered_rev, bunordered_rev, buneq_rev)
2249 (bunge_rev, bungt_rev, bunle_rev, bunlt_rev, bltgt_rev): Change
2250 condition from TARGET_68881 to TARGET_HARD_FLOAT.
2251
2252 2007-01-19 Sandra Loosemore <sandra@codesourcery.com>
2253
2254 * longlong.h (count_leading_zeros, COUNT_LEADING_ZEROS_0): Add
2255 ColdFire alternatives.
2256 * config/m68k/m68k.h (CLZ_DEFINED_VALUE_AT_ZERO): New macro.
2257 * config/m68k/m68k.md (clzsi2): Define for ColdFire
2258 architectures that support ff1 instruction.
2259
2260 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2261 Julian Brown <julian@codesourcery.com>
2262
2263 * config/m68k/m68k.h (CONST_OK_FOR_LETTER_P): Add an 'R' case.
2264 * config/m68k/m68k.md (*movsi_cfv4): Fold into...
2265 (*movsi_cf): ...here. Remove unnecessary 'R' from 'Rg'.
2266 Add commentary.
2267
2268 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2269
2270 * config/m68k/m68k-protos.h (valid_mov3q_const): Take a HOST_WIDE_INT
2271 and return a bool.
2272 (output_move_const_into_data_reg, output_move_simode_const): Delete.
2273 * config/m68k/m68k.c (const_method, const_int_cost): Take a
2274 HOST_WIDE_INT instead of an rtx.
2275 (m68k_rtx_costs): Update call accordingly.
2276 (output_move_const_into_data_reg): Likewise. Fix formatting.
2277 (valid_mov3q_const): Take a HOST_WIDE_INT instead of an rtx.
2278 Return a bool.
2279 (output_move_simode_const): Update calls after above changes.
2280 Rework to use automatic variables and predicates like MEM_P.
2281 * config/m68k/m68k.md (pushexthisi_const): Update call to
2282 valid_mov3q_const.
2283
2284 2007-01-19 Dirk Mueller <dmueller@suse.de>
2285
2286 * tree-ssa-alias.c (perform_var_substitution): Fix typo
2287 in dump_flags test.
2288
2289 2007-01-19 Richard Guenther <rguenther@suse.de>
2290
2291 * builtins.c (expand_builtin_cexpi): Fall back to expanding
2292 via cexp in case sincos is not available.
2293
2294 2007-01-19 Richard Guenther <rguenther@suse.de>
2295
2296 * doc/tm.texi (TARGET_HAS_SINCOS): Document new target macro.
2297 * defaults.h (TARGET_HAS_SINCOS): Default to off.
2298 * config/linux.h (TARGET_HAS_SINCOS): Set to on if we have glibc.
2299 * config/alpha/linux.h (TARGET_HAS_SINCOS): Likewise.
2300 * config/sparc/linux.h (TARGET_HAS_SINCOS): Likewise.
2301 * config/sparc/linux64.h (TARGET_HAS_SINCOS): Likewise.
2302 * config/rs6000/linux.h (TARGET_HAS_SINCOS): Likewise.
2303 * config/rs6000/linux64.h (TARGET_HAS_SINCOS): Likewise.
2304
2305 2007-01-19 Uros Bizjak <ubizjak@gmail.com>
2306
2307 * config/i386/i386.md (*fpatanxf3_i387, fpatan_extend<mode>xf3_i387):
2308 New insn patterns.
2309 (atan2sf3_1, atan2df3_1, atan2xf3_1): Remove insn patterns.
2310 (atan2xf3): Directly generate RTL pattern.
2311 (atan2<mode>3): Rename from atan2sf3 and atan2df3 and macroize insn
2312 patterns using X87MODEF12 mode macro. Use fpatan_extend<mode>xf3_i387
2313 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
2314 disable patterns for SSE math.
2315 (atan<mode>2): Rename from atansf2 and atandf2 and macroize insn
2316 patterns using X87MODEF12 mode macro. Use fpatan_extend<mode>xf3_i387
2317 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
2318 disable patterns for SSE math.
2319
2320 2007-01-19 Alexandre Oliva <aoliva@redhat.com>
2321
2322 * libgcc-std.ver: Fix typo in %inherit for GCC_4.3.0.
2323
2324 2007-01-18 Roger Sayle <roger@eyesopen.com>
2325
2326 * fold-const.c (fold_unary) <VIEW_CONVERT_EXPR>: Optimize away a
2327 VIEW_CONVERT_EXPR to the same type as it's operand.
2328
2329 2007-01-18 David Edelsohn <edelsohn@gnu.org>
2330
2331 * config/rs6000/darwin-ldouble.c: Only build _SOFT_FLOAT if
2332 configured for long double 128.
2333
2334 2007-01-18 Mike Stump <mrs@apple.com>
2335
2336 * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Fix build
2337 error.
2338
2339 2007-01-18 Michael Meissner <michael.meissner@amd.com>
2340
2341 * i386.c (ix86_compute_frame_layout): Make fprintf's in #if 0 code
2342 type correct.
2343
2344 2007-01-18 Jan Hubicka <jh@suse.cz>
2345
2346 * tree-ssa-operands.c (vop_free_bucket_size): Never return value
2347 greater than NUM_VOP_FREE_BUCKETS.
2348
2349 2007-01-18 Daniel Berlin <dberlin@dberlin.org>
2350
2351 * tree-ssa-structalias.c: Update comments.
2352 (ptabitmap_obstack): Removed.
2353 (pta_obstack): New.
2354 (oldpta_obstack): Ditto.
2355 (stats): Add a few members.
2356 (struct variable_info): Remove node, complex, address_taken, and
2357 indirect_target members. Add oldsolution member.
2358 (new_var_info): Do not initialize removed members.
2359 (constraint_expr_type): Remove INCLUDES.
2360 (constraint_graph): Add size, implicit_preds, rep,
2361 indirect_cycles, eq_rep, label, direct_nodes, and complex members.
2362 (FIRST_REF_NODE): New macro.
2363 (LAST_REF_NODE): Ditto.
2364 (FIRST_ADDR_NODE): Ditto.
2365 (find): New function.
2366 (unite): Ditto.
2367 (dump_constraint): Do not handle INCLUDES.
2368 (insert_into_complex): Do not insert duplicate constraints.
2369 (condense_varmap_nodes): Renamed and rewritten into ...
2370 (merge_node_constraints): This. Also fix bug in handling of
2371 offseted copy constraints.
2372 (clear_edges_for_node): No longer need to deal with preds at all,
2373 or removing associated preds/succs.
2374 (merge_graph_nodes): Deal with indirect_cycles.
2375 Don't deal with predecessors.
2376 (add_implicit_graph_edge): New function.
2377 (add_pred_graph_edge): Ditto.
2378 (add_graph_edge): Don't deal with predecessors.
2379 (build_constraint_graph): Removed.
2380 (build_pred_graph): New function.
2381 (build_succ_graph): Ditto.
2382 (struct scc_info): Removed in_component. Added roots, dfs, and
2383 node_mapping. Remove visited_index, unification_queue.
2384 (scc_visit): Deal with union-find we do now.
2385 Deal with cycles with REF nodes.
2386 (collapse_nodes): Renamed and rewritten to ...
2387 (unify_nodes): This.
2388 (process_unification_queue): Removed.
2389 (topo_visit): Cleanup
2390 (do_da_constraint): Use find.
2391 (do_sd_constraint): Ditto.
2392 (do_ds_constraint): Ditto.
2393 (do_complex_constraint): Ditto.
2394 (init_scc_info): Update for removed and added members.
2395 (find_and_collapse_graph_cycles): Renamed and rewritten into ...
2396 (find_indirect_cycles): This.
2397 (equivalence_class): New variable.
2398 (label_visit): New function.
2399 (perform_variable_substitution): Rewritten.
2400 (free_var_substitution_info): New function.
2401 (find_equivalent_node): Ditto.
2402 (move_complex_constraints): Ditto.
2403 (eliminate_indirect_cycles): Ditto.
2404 (solve_graph): Only propagate changed bits.
2405 Use indirect cycle elimination.
2406 Use find.
2407 (tree_id_t): Rename to tree_vi_t, delete id member, add vi member.
2408 (tree_id_eq): Renamed to ...
2409 (tree_vi_eq): This. Update for member change
2410 (insert_id_for_tree): Renamed and rewritten to ...
2411 (insert_vi_for_tree): This.
2412 (lookup_id_for_tree): Renamed and rewritten to ...
2413 (lookup_vi_for_tree): This.
2414 (get_id_for_tree): Renamed and rewritten to ...
2415 (get_vi_for_tree): Ditto.
2416 (get_constraint_exp_from_ssa_var): Update to use get_vi_for_tree.
2417 (process_constraint): Don't handle INCLUDES.
2418 Remove special ADDRESSOF case.
2419 (find_func_aliases): Rewrite to use vi functions instead of id
2420 ones.
2421 (create_function_info_for): Ditto.
2422 (create_variable_info_for): Ditto.
2423 (intra_create_variable_infos): Ditto.
2424 (merge_smts_into): Ditto.
2425 (find_what_p_points_to): Ditto.
2426 (init_base_vars): Ditto.
2427 (init_alias_vars): Ditto.
2428 (remove_preds_and_fake_succs): New function.
2429 (dump_sa_points_to_info): Dump new stats.
2430 (dump_solution_for_var): Use find.
2431 (set_used_smts): Fix formatting.
2432 (compute_points_to_sets): Updated for new functions.
2433 (ipa_pta_execute): Ditto.
2434
2435 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
2436 Richard Sandiford <richard@codesourcery.com>
2437
2438 * doc/tm.texi (TARGET_FUNCTION_VALUE): Expand documentation of
2439 parallels.
2440 * calls.c (expand_call): If the return value is a PARALLEL,
2441 extract its first member.
2442 * config/m68k/linux.h (FUNCTION_EXTRA_EPILOGUE): Remove.
2443 * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
2444 use FUNCTION_EXTRA_EPILOGUE.
2445 (m68k_function_value): Return a PARALLEL if the return value
2446 is of a pointer type.
2447 * config/m68k/netbsd-elf.h (current_function_returns_pointer)
2448 (FUNCTION_EXTRA_EPILOGUE): Remove.
2449 * config/m68k/m68k.md (D0_REG): New constant.
2450
2451 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
2452
2453 * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
2454 output a NOP for empty epilogues.
2455
2456 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2457
2458 * config/m68k/m68k.c (m68k_use_return_insn): Update comments
2459 before function. Extend register save check to include all
2460 registers, not just integer ones.
2461
2462 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
2463
2464 * config/m68k/m68k-protos.h (use_return_insn): Rename to...
2465 (m68k_use_return_insn): ...this.
2466 * config/m68k/m68k.h (USE_RETURN_INSN): Delete.
2467 * config/m68k/m68k.c (use_return_insn): Rename to...
2468 (m68k_use_return_insn): ...this.
2469 * config/m68k/m68k.md (return): Use m68k_use_return_insn instead
2470 of USE_RETURN_INSN.
2471
2472 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
2473
2474 * config/m68k/fpgnulib.c (__truncdfsf2): Implement round to
2475 nearest even, fix denormal rounding overflow.
2476
2477 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2478
2479 * config/m68k/m68k.md (movsf_cf_hard): Use fsmove instead of
2480 f%$smove and f%$move.
2481 (movdf_cf_hard): Use fdmove for cases 0 and 3 and fmove for case 1.
2482 (extendsfdf2_cf): Use fdmove instead of f%&move.
2483 (truncdfsf2_cf): Use fsmove instead of f%$smove.
2484 (add<mode>3_cf, sub<mode>3_cf): Use <FP:prec> instead of <FP:round>.
2485
2486 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2487
2488 * config/m68k/m68k.md (movdf_cf_hard): Use output_move_double for
2489 GPR<-GPR moves.
2490
2491 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2492
2493 * real.h (real_format): Add a canonical_nan_lsbs_set field.
2494 (coldfire_single_format): Declare.
2495 (coldfire_double_format): Likewise.
2496 * real.c (encode_ieee_single): Use canonical_nan_lsbs_set instead
2497 of qnan_msb_set to determine the lower bits of a canonical
2498 NaN significand.
2499 (encode_ieee_double): Likewise.
2500 (encode_ieee_quad): Likewise.
2501 (ieee_single_format): Initialize canonical_nan_lsbs_set.
2502 (mips_single_format): Likewise.
2503 (ieee_double_format): Likewise.
2504 (mips_double_format): Likewise.
2505 (ieee_extended_motorola_format): Likewise.
2506 (ieee_extended_intel_96_format): Likewise.
2507 (ieee_extended_intel_128_format): Likewise.
2508 (ieee_extended_intel_96_round_53_format): Likewise.
2509 (ibm_extended_format): Likewise.
2510 (mips_extended_format): Likewise.
2511 (ieee_quad_format): Likewise.
2512 (mips_quad_format): Likewise.
2513 (vax_f_format): Likewise.
2514 (vax_d_format): Likewise.
2515 (vax_g_format): Likewise.
2516 (i370_single_format): Likewise.
2517 (i370_double_format): Likewise.
2518 (decimal_single_format): Likewise.
2519 (decimal_double_format): Likewise.
2520 (decimal_quad_format): Likewise.
2521 (c4x_single_format): Likewise.
2522 (c4x_extended_format): Likewise.
2523 (real_internal_format): Likewise.
2524 (coldfire_single_format): New real_format.
2525 (coldfire_double_format): Likewise.
2526 * config/pdp11/pdp11.c (pdp11_f_format): Initialize
2527 canonical_nan_lsbs_set.
2528 (pdp11_d_format): Likewise.
2529 * config/m68k/m68k.c (override_options): Override REAL_FORMAT_MODE
2530 if TARGET_COLDFIRE_CPU.
2531
2532 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2533
2534 * config/m68k/m68k-protos.h (m68k_output_pic_call): Delete.
2535 (output_call, m68k_legitimize_call_address): Declare.
2536 * config/m68k/m68k.h (EXTRA_CONSTRAINT): Remove unnecessary
2537 parenthesees. Add support for a 'W' constraint.
2538 (LEGITIMATE_PIC_OPERAND_P): Remove SYMBOL_REF_FLAG handling.
2539 (PRINT_OPERAND_PUNCT_VALID_P): Remove comment about 'o'.
2540 (m68k_symbolic_call, m68k_symbolic_jump): Declare.
2541 * config/m68k/m68k.c (m68k_symbolic_call, m68k_symbolic_jump): New
2542 variables.
2543 (override_options): Initialize them. Do not set flag_no_function_cse
2544 for TARGET_ID_SHARED_LIBRARY.
2545 (m68k_output_pic_call): Delete.
2546 (m68k_legitimize_call_address): New function.
2547 (print_operand): Remove the %o prefix. Handle the %p prefix.
2548 (output_call): New function.
2549 (m68k_output_mi_thunk): Use m68k_symbolic_jump. Always load the
2550 target address from the GOT if symbolic jumps are not allowed.
2551 * config/m68k/m68k.md (call, general_operand): Do not set
2552 SYMBOL_REF_FLAG. Use m68k_legitimize_call_address instead.
2553 Merge separate flag_pic and !flag_pic define_insns into...
2554 (*call, *call_value): ...these new patterns. Match the address
2555 rather than the containing MEM and require it to be a call_operand.
2556 Use output_call to generate the asm template.
2557 * config/m68k/predicates.md (const_call_operand): New predicate.
2558 (call_operand): Likewise.
2559
2560 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
2561
2562 * config/m68k/m68k.h (REGISTER_MOVE_COST): Simplify definition.
2563 (STACK_GROWS_DOWNWARD): Define to 1.
2564 (FUNCTION_VALUE, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P): Equivocate
2565 comments, emphasizing that these values are only defaults.
2566 * config/m68k/linux.h (LINK_SPEC): Fix formatting in #undef.
2567 * config/m68k/m68k.c (const_method): Remove trailing whitespace.
2568
2569 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2570
2571 * config/m68k/m68k.md (cmpsi): Remove outdated flag_pic handling.
2572
2573 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
2574 Richard Sandiford <richard@codesourcery.com>
2575
2576 * config/m68k/m68k.h (DATA_REGNO_P, ADDRESS_REGNO_P, INT_REGNO_P)
2577 (FP_REGNO_P): New macros.
2578 (REGNO_OK_FOR_INDEX_P, REGNO_OK_FOR_BASE_P, REGNO_OK_FOR_DATA_P)
2579 (REGNO_OK_FOR_FP_P, REG_OK_FOR_INDEX_P, REG_OK_FOR_BASE_P): Use them.
2580 (EH_RETURN_STACKADJ_RTX): Use A0_REG.
2581 * config/m68k/m68k.c (m68k_regno_mode_ok): Use the new REGNO macros.
2582
2583 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
2584
2585 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
2586 (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-rtems*): Add t-floatlib
2587 to $tmake_file.
2588 * config/m68k/t-floatlib: New file, extracting common code from...
2589 * config/m68k/t-m68kbare, config/m68k/t-m68kelf,
2590 * config/m68k/t-uclinux: Here.
2591 * config/m68k/fpgnulib.c: Do not compile extendeed precision
2592 routines on ColdFire targets.
2593
2594 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
2595
2596 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
2597 (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*)
2598 (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-linux*)
2599 (m68k-*-rtems*): Use tm_file rather than m68k/m68k.h and
2600 explicitly set MOTOROLA to 1.
2601 * config/m68k/m68k.h (MOTOROLA): Simplify definition accordingly.
2602
2603 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2604 Nathan Sidwell <nathan@codesourcery.com>
2605
2606 * config/m68k/m68k.h (PCC_STATIC_STRUCT_RETURN): Delete.
2607 (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Add MOTOROLA cases.
2608 * config/m68k/coff.h (REGISTER_PREFIX_MD): Delete.
2609 * config/m68k/m68020-elf.h (LIB_SPEC): Delete.
2610 * config/m68k/m68k-none.h (CC1_SPEC, CPP_SUBTARGET_SPEC): Delete.
2611 * config/m68k/m68kelf.h (IMMEDIATE_PREFIX, REGISTER_PREFIX_MD)
2612 (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Delete.
2613 (NO_DOLLAR_IN_LABEL, PCC_STATIC_STRUCT_RETURN): Don't undefine.
2614 (BSS_ASM_OP, ASM_OUTPUT_SKIP, ASM_OUTPUT_ASCII): Delete.
2615 * config/m68k/m68kemb.h (PCC_STATIC_STRUCT_RETURN): Don't undefine.
2616 * config/m68k/linux.h (SIZE_TYPE, PTRDIFF_TYPE, WCHAR_TYPE)
2617 (WCHAR_TYPE_SIZE, TARGET_OBJFMT_CPP_BUILTINS): Delete.
2618 (TARGET_OS_CPP_BUILTINS): Don't define mc68000 and mc68020 here.
2619 (DBX_REGISTER_NUMBER): Delete.
2620 * config/m68k/netbsd-elf.h (IMMEDIATE_PREFIX): Delete.
2621 (PCC_STATIC_STRUCT_RETURN): Don't undefine.
2622 * config/m68k/openbsd.h (PCC_STATIC_STRUCT_RETURN): Define.
2623
2624 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2625
2626 * config.gcc (m68k-*-uclinux*): Add flat.h to $tm_file.
2627 * config/flat.h: New file.
2628 * crtstuff.c (USE_PT_GNU_EH_FRAME): Don't define if
2629 OBJECT_FORMAT_FLAT.
2630 * config/m68k/m68k.h (ASM_PREFERRED_EH_DATA_FORMAT): Do not use
2631 indirect references for -msep-data or -mid-shared-library.
2632 Do not use PC-relative code addresses either.
2633 * config/m68k/m68k.c (override_options): Restrict -fPIC error
2634 to -mpcrel.
2635 * config/m68k/uclinux.h (STARTFILE_SPEC): Define. Use Scrt1.o
2636 for shared libraries and crt1.o for executables. Use crti.o and
2637 crtbegin.o.
2638 (ENDFILE_SPEC): Use crtend.o and crtn.o.
2639 (LIB_SPEC): Suppress -Rlibc.gdb if -static-libc is given.
2640 Do not add -elf2flt or -shared-lib-id options here.
2641 (LINK_SPEC): Define. Pass -elf2flt if no -elf2flt option is given.
2642 Pass -shared-lib-id if -mid-shared-library, taking the library
2643 identifier from -mshared-library-id if given, otherwise
2644 defaulting to 0.
2645 (EH_FRAME_IN_DATA_SECTION): Do not undefine.
2646 (INIT_SECTION_ASM_OP, FINI_SECTION_ASM_OP): Likewise.
2647 (TARGET_OS_CPP_BUILTINS): Define __GXX_MERGED_TYPEINFO_NAMES=0
2648 and __GXX_TYPEINFO_EQUALITY_INLINE=0 if -mid-shared-library.
2649 (DRIVER_SELF_SPECS): Map unadorned PIC options to -msep-data.
2650 * config/m68k/t-uclinux (EXTRA_MULTILIB_PARTS): Add crtbegin.o
2651 and crtend.o.
2652 * config/m68k/lb1sf68.asm (PICCALL): Use an lea and pc-relative
2653 jump sequence for ISA A and ISA A+.
2654 (PICJUMP): Likewise.
2655
2656 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2657
2658 * config.gcc (m68k-*-uclinux*): Base the port on the common
2659 and m68k GNU/Linux files rather than on the generic ELF ones.
2660 * config/m68k/uclinux.h (TARGET_VERSION): Override.
2661 (TARGET_OS_CPP_BUILTINS): Use LINUX_TARGET_OS_CPP_BUILTINS.
2662
2663 2007-01-18 Julian Brown <julian@codesourcery.com>
2664 Richard Sandiford <richard@codesourcery.com>
2665
2666 * config/m68k/m68k.h (LONG_DOUBLE_TYPE_SIZE): Make 64-bit on ColdFire.
2667 (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
2668 * config/m68k/netbsd-elf.h (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Undefine
2669 before redefining.
2670 * config/m68k/uclinux-oldabi.h (LONG_DOUBLE_TYPE_SIZE): Redefine to
2671 80 unconditionally.
2672 (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
2673
2674 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2675
2676 * doc/install.texi: Document m68k-uclinuxoldabi.
2677 * config.gcc (m68k-*-uclinuxoldabi): New configuration.
2678 * config/m68k/uclinux-oldabi.h: New file, copied from
2679 config/m68k/uclinux.h.
2680
2681 2007-01-18 Kaz Kojima <kkojima@rr.iij4u.or.jp>
2682
2683 * config/m32r/m32r-protos.h (m32r_expand_epilogue): Declare it.
2684 * config/m32r/m32r.c (m32r_setup_incoming_varargs): Use gen_frame_mem.
2685 (m32r_compute_frame_size): Use unsigned for regno.
2686 (m32r_reload_lr): Use gen_frame_mem.
2687 (pop): New.
2688 (m32r_output_function_epilogue): Don't output the function epilogue
2689 textually here.
2690 (m32r_expand_epilogue): New.
2691 (direct_return): Return false if the function has the interrupt
2692 attribute.
2693 (m32r_hard_regno_rename_ok): Remove code for the textual epilogue.
2694 * config/m32r/m32r.md (epilogue): New expander.
2695 (return_lr, return_rte): New insns.
2696 (return): Make it expander.
2697 (return_normal): New expander.
2698
2699 2007-01-18 Josh Conner <jconner@apple.com>
2700
2701 PR target/30485
2702 * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Add
2703 support for UNLE, UNLT, UNGE, and UNGT.
2704
2705 2007-01-18 Jan Hubicka <jh@suse.cz>
2706
2707 * tree-vrp.c (finalize_jump_threads): Do not call cleanup_cfg by hand.
2708 * tree-tailcall (add_virtual_phis): Likewise.
2709 (optimize_tail_call): Return TODOs.
2710 (execute_tail_calls): Return TODOs.
2711 * tree-ssa-ccp (execute_fold_all_builtins): Do cleanup_cfg via TODO.
2712 * tree-cfgcleanup.c (cleanup_tree_cfg_loop): Return if something
2713 changed.
2714 * tree-ssa-forwprop.c (tree_ssa_forward_propagate_single_use_value):
2715 Cleanup cfg using TODO.
2716 * tree-flow.h (cleanup_tree_cfg_loop): Update prototype.
2717 * passes.c (execute_function_todo): When cleanup did something, remove
2718 unused locals.
2719 * tree-cfg.c (pass_build_cfg): Add cleanup_cfg TODO.
2720 (make_edges): Don't cleanup_cfg.
2721
2722 2007-01-18 Uros Bizjak <ubizjak@gmail.com>
2723
2724 * reg-stack.c (subst_stack_regs_pat) [UNSPEC_SINCOS_COS,
2725 UNSPEC_XTRACT_FRACT]: Use generic code for instructions that
2726 operate on the top of stack.
2727 [UNSPEC_SINCOS_SIN, UNSPEC_XTRACT_EXP, UNSPEC_TAN]: Rewrite
2728 register handling of instructions that output to the second
2729 stack slot.
2730 [UNSPEC_TAN_ONE, UNSPEC_TAN_TAN]: Remove.
2731 (move_for_stack_reg): Special-case check for dead destination
2732 stack slot for constant load of 1.0 inside UNSPEC_TAN.
2733
2734 * config/i386/i386.md (UNSPEC_TAN): New constant.
2735 (UNSPEC_TAN_ONE, UNSPEC_TAN_TAN): Remove.
2736 (fptanxf4_i387, fptan_extend<mode>xf4_i387): New patterns
2737 to correctly model move of constant 1.0 to top stack slot.
2738 (*tandf3_1, *tansf3_1, *tanxf3_1): Remove insn patterns.
2739 (unnamed peephole2 pattern): Remove corresponding peephole2
2740 pattern that optimizes tan insn and loading of constant 1.0.
2741 (tanxf2): Use fptanxf4_i387.
2742 (tan<mode>2): Rename from tansf2 and tandf2 and macroize insn
2743 patterns using X87MODEF12 mode macro. Use fptan_extend<mode>xf4_i387
2744 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
2745 disable patterns for SSE math.
2746 (sincos<mode>3): Use truncxf<mode>2_i387_noop for truncation.
2747 (fyl2x_extend<mode>xf3_i387): Use X87MODEF12 for operand 1.
2748
2749 2007-01-18 Dirk Mueller <dmueller@suse.de>
2750 Richard Guenther <rguenther@suse.de>
2751
2752 PR diagnostic/8268
2753 * doc/invoke.texi (Warray-bounds): Document -Warray-bounds.
2754 * common.opt (Warray-bounds): Add new warning option.
2755 * c-opts.c (c_common_handle_option): Define -Warray-bounds
2756 if -Wall is given.
2757 * Makefile.in: make tree-vrp.o depend on toplev.h
2758 * tree-vrp.c (vrp_finalize): Call check_array_refs if -Warray-bounds
2759 is enabled.
2760 (check_array_refs, check_array_bounds, check_array_ref): New.
2761
2762 2007-01-18 Jan Hubicka <jh@suse.cz>
2763
2764 * tree-ssa-ccp.c (ccp_finalize): Return if something changed.
2765 (execute_ssa_ccp): Return flags conditionally.
2766 * tree-ssa-propagate.c (substitue_and_fold): Return if something was
2767 changed.
2768 * tree-ssa-propagate.h (substitute_and_fold): Update prototype.
2769
2770 2007-01-18 Steven Bosscher <steven@gcc.gnu.org>
2771
2772 * cfgcleanup.c (cleanup_cfg): Detect cfglayout mode and set
2773 the CLEANUP_CFGLAYOUT flag when in cfglayout mode.
2774
2775 * Makefile.c (GTFILES): Add cfglayout.h.
2776 * gengtype.c (open_base_files): Likewise.
2777 * cfglayout.c (cfg_layout_function_footer,
2778 cfg_layout_function_header) Reindent to make gengtype happy.
2779 * cfglayout.h (cfg_layout_function_footer,
2780 cfg_layout_function_header): Add GTY(()) marker.
2781
2782 * ifcvt.c (noce_try_sign_mask): Make sure INSN_B is non-null.
2783
2784 2007-01-18 Ben Elliston <bje@au.ibm.com>
2785
2786 * genautomata.c (write_automata): Include xstrerror output in the
2787 error message if writing the DFA description file fails.
2788
2789 2007-01-17 H.J. Lu <hongjiu.lu@intel.com>
2790
2791 * config/mips/mips-protos.h (mips_output_external): Make it
2792 return void.
2793 * config/mips/iris.h (TARGET_ASM_EXTERNAL_LIBCALL): Removed.
2794 * config/mips/mips.c (irix_output_external_libcall): Likewise.
2795 (extern_list): Likewise.
2796 (extern_head): Likewise.
2797 (TARGET_ASM_FILE_END): Likewise.
2798 (mips_file_end): Likewise.
2799 (mips_output_external): Rewritten.
2800
2801 2007-01-18 Ben Elliston <bje@au.ibm.com>
2802
2803 * genpreds.c (write_insn_preds_c): Only write out the function
2804 body for regclass_for_constraint if we have register constraints.
2805
2806 2007-01-17 Tom Tromey <tromey@redhat.com>
2807
2808 * doc/sourcebuild.texi (libgcj Tests): Use sourceware.org.
2809 * doc/install.texi (Testing): Use sourceware.org.
2810 (Binaries): Likewise.
2811 (Specific): Likewise.
2812 * doc/contrib.texi (Contributors): Use sourceware.org.
2813
2814 2007-01-17 Anatoly Sokolov <aesok@post.ru>
2815
2816 * config/avr/avr.h (AVR_HAVE_LPMX): New macro.
2817 (AVR_ENHANCED): Rename to ...
2818 (AVR_HAVE_MUL): ... new.
2819 (avr_enhanced_p): Rename to ...
2820 (avr_have_mul_p): ... new.
2821 (TARGET_CPU_CPP_BUILTINS): Use 'avr_have_mul_p' instead of
2822 'avr_enhanced_p' for "__AVR_ENHANCED__". Define "__AVR_HAVE_MUL__".
2823 * config/avr/avr.c (avr_enhanced_p): Rename to ...
2824 (avr_have_mul_p): ... new.
2825 (base_arch_s): Rename 'enhanced' to 'have_mul'.
2826 (avr_override_options): Use 'avr_have_mul_p' and 'have_mul' instead of
2827 'avr_enhanced_p' and 'enhanced'.
2828 (ashlhi3_out, ashrhi3_out, lshrhi3_out, avr_rtx_costs): Use
2829 AVR_HAVE_MUL instead of AVR_ENHANCED.
2830 * avr.md (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
2831 (mulqi3, *mulqi3_enh, *mulqi3_call, mulqihi3, umulqihi3, mulhi3,
2832 *mulhi3_enh, *mulhi3_call, mulsi3, *mulsi3_call): Use AVR_HAVE_MUL
2833 instead of AVR_ENHANCED.
2834 (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
2835 * libgcc.S: Use __AVR_HAVE_MUL__ instead of __AVR_ENHANCED__.
2836 (__tablejump__): Use __AVR_HAVE_LPMX__ instead of __AVR_ENHANCED__.
2837
2838 2007-01-17 Ian Lance Taylor <iant@google.com>
2839
2840 * vec.h (VEC_reserve_exact): Define.
2841 (vec_gc_p_reserve_exact): Declare.
2842 (vec_gc_o_reserve_exact): Declare.
2843 (vec_heap_p_reserve_exact): Declare.
2844 (vec_heap_o_reserve_exact): Declare.
2845 (VEC_OP (T,A,reserve_exact)): New static inline function, three
2846 versions.
2847 (VEC_OP (T,A,reserve)) [all versions]: Remove handling of
2848 negative parameter.
2849 (VEC_OP (T,A,alloc)) [all versions]: Call ...reserve_exact.
2850 (VEC_OP (T,A,copy)) [all versions]: Likewise.
2851 (VEC_OP (T,a,safe_grow)) [all versions]: Likewise.
2852 * vec.c (calculate_allocation): Add exact parameter. Change all
2853 callers.
2854 (vec_gc_o_reserve_1): New static function, from vec_gc_o_reserve.
2855 (vec_gc_p_reserve, vec_gc_o_reserve): Call vec_gc_o_reserve_1.
2856 (vec_gc_p_reserve_exact, vec_gc_o_reserve_exact): New functions.
2857 (vec_heap_o_reserve_1): New static function, from vec_heap_o_reserve.
2858 (vec_heap_p_reserve, vec_heap_o_reserve): Call vec_heap_o_reserve_1.
2859 (vec_heap_p_reserve_exact): New function.
2860 (vec_heap_o_reserve_exact): New function.
2861
2862 2007-01-17 Jan Hubicka <jh@suse.cz>
2863
2864 * ipa-type-escape.c (look_for_casts): Revamp using handled_component_p.
2865
2866 2007-01-17 Eric Christopher <echristo@apple.com>
2867
2868 * config.gcc: Support core2 processor.
2869
2870 2007-01-16 Jan Hubicka <jh@suse.cz>
2871
2872 * tree-ssanames.c (release_dead_ssa_names): Instead of ggc_freeing
2873 the names, just unlink the chain so we don't crash on dangling pointers
2874 to dead SSA names.
2875
2876 2007-01-16 Jan Hubicka <jh@suse.cz>
2877
2878 * cgraph.h (cgraph_decide_inlining_incrementally): Kill.
2879 * tree-pass.h: Reorder to make IPA passes appear toegher.
2880 (pass_early_inline, pass_inline_parameters, pass_apply_inline): Declare.
2881 * cgraphunit.c (cgraph_finalize_function): Do not compute inling
2882 parameters, do not call early inliner.
2883 * ipa-inline.c: Update comments. Include tree-flow.h
2884 (cgraph_decide_inlining): Do not compute inlining parameters.
2885 (cgraph_decide_inlining_incrementally): Return TODOs; assume to
2886 be called with function context set up.
2887 (pass_ipa_inline): Remove unreachable functions before pass.
2888 (cgraph_early_inlining): Simplify assuming to be called from the
2889 PM as local pass.
2890 (pass_early_inline): New pass.
2891 (cgraph_gate_ipa_early_inlining): New gate.
2892 (pass_ipa_early_inline): Turn into simple wrapper.
2893 (compute_inline_parameters): New function.
2894 (gate_inline_passes): New gate.
2895 (pass_inline_parameters): New pass.
2896 (apply_inline): Move here from tree-optimize.c
2897 (pass_apply_inline): New pass.
2898 * ipa.c (cgraph_remove_unreachable_nodes): Verify cgraph after
2899 transforming.
2900 * tree-inline.c (optimize_inline_calls): Return TODOs rather than
2901 doing them by hand.
2902 (tree_function_versioning): Do not allocate dummy struct function.
2903 * tree-inline.h (optimize_inline_calls): Update prototype.
2904 * tree-optimize.c (execute_fixup_cfg): Export.
2905 (pass_fixup_cfg): Remove
2906 (tree_rest_of_compilation): Do not apply inlines.
2907 * tree-flow.h (execute_fixup_cfg): Declare.
2908 * Makefile.in (gt-passes.c): New.
2909 * passes.c: Include gt-passes.h
2910 (init_optimization_passes): New passes.
2911 (nnodes, order): New static vars.
2912 (do_per_function_toporder): New function.
2913 (execute_one_pass): Dump current pass here.
2914 (execute_ipa_pass_list): Don't dump current pass here.
2915
2916 2007-01-16 Janis Johnson <janis187@us.ibm.com>
2917
2918 * config/dfp-bit.c (dfp_compare_op): Return separate value for NaN.
2919 (DFP_NE, DFP_LE, DFP_GE): Return false for NaN.
2920
2921 2007-01-16 David Edelsohn <edelsohn@gnu.org>
2922
2923 * config/rs6000/darwin-ldouble.c: Build file for SOFT_FLOAT.
2924 (strong_alias): Define.
2925 (__gcc_qmul): Provide non-FMA for soft-float.
2926 (__gcc_qdiv): Same.
2927 (__gcc_qneg): New.
2928 (__gcc_qeq): New.
2929 (__gcc_qle): New.
2930 (__gcc_qge): New.
2931 (__gcc_qunord): New.
2932 (__gcc_stoq): New.
2933 (__gcc_dtoq): New.
2934 (__gcc_qtos): New.
2935 (__gcc_qtod): New.
2936 (__gcc_qtoi): New.
2937 (__gcc_qtou): New.
2938 (__gcc_itoq): New.
2939 (__gcc_utoq): New.
2940 (fmsub): New.
2941 * config/rs6000/rs6000.c (rs6000_init_libfuncs): Initialize
2942 soft-float functions.
2943 * config/rs6000/libgcc-ppc-glibc.ver: Version soft-float symbols.
2944 * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Do not warn
2945 about long double soft float.
2946
2947 2007-01-16 Dorit Nuzman <dorit@il.ibm.com>
2948 Tehila Meyzels <tehila@il.ibm.com>
2949
2950 * tree-vectorizer.h (is_pattern_stmt_p): New.
2951 * tree-vect-analyze.c (vect_determine_vectorization_factor): Fix
2952 formatting (tabs instead of spaces). Cleanup and clarify setting
2953 of STMT_VINFO_VECTYPE. Call is_pattern_stmt_p.
2954 * tree-vect-transform.c (vect_get_vec_def_for_operand): Fix typo.
2955 (vectorizable_type_demotion): Check that types are integral.
2956 (vectorizable_type_promotion): Likewise.
2957 (vectorizable_store): Fix typo. Eliminate new-line at end of
2958 comments.
2959
2960 2007-01-16 Jan Hubicka <jh@suse.cz>
2961
2962 * tree-ssanames.c (release_dead_ssa_names): Remove invalidated
2963 cgraph edges too.
2964
2965 2007-01-15 Eric Christopher <echristo@apple.com>
2966
2967 * ifcvt.c: Include vec.h, vecprim.h.
2968 (check_cond_move_block): New argument regs.
2969 Reorganize. Add registers used to regs.
2970 (cond_move_process_if_block): Use regs set above as
2971 loop bounds.
2972
2973 2007-01-15 Eric Christopher <echristo@apple.com>
2974
2975 * config/darwin.h: Update copyright.
2976 (TARGET_OPTION_TRANSLATE_TABLE): Add umbrella.
2977 (LINK_COMMAND_SPEC): Add -u.
2978 (LINK_SPEC): Fix umbrella for above.
2979
2980 2007-01-15 Joseph S. Myers <joseph@codesourcery.com>
2981
2982 * config/soft-fp/op-common.h, config/soft-fp/op-4.h: Update from
2983 glibc CVS.
2984
2985 2007-01-15 Tom Tromey <tromey@redhat.com>
2986
2987 * doc/sourcebuild.texi (libgcj Tests): Don't mention jacks.
2988 * doc/install.texi (Testing): Don't mention jacks.
2989 (Configuration): Document --enable-java-maintainer-mode. Move
2990 --with-java-home to libgcj-specific section. Document
2991 --with-ecj-jar.
2992 (Prerequisites): Mention --enable-java-maintainer-mode, ecj1.
2993
2994 2007-01-15 Jan Hubicka <jh@suse.cz>
2995
2996 * tree-ssa-dce.c (DCE_TODOs): New.
2997 (propagate_necessity): Return if something changed.
2998 (eliminate_unnecessary_stmts): Likewise.
2999 (perform_tree_ssa_dce): Return TODO flags when needed.
3000 (pass_dce, pass_dce_loop, pass_cd_dce): Remove TODO flags.
3001
3002 2007-01-15 Uros Bizjak <ubizjak@gmail.com>
3003
3004 * config/i386/i386.md (fyl2xxf3_i387): Rename from fyl2x_xf3.
3005 (fyl2x_extend<mode>xf3_i387): New insn pattern.
3006 (log<mode>2): Rename from logsf2 and logdf2 and macroize insn
3007 patterns using X87MODEF12 mode macro. Extend operand 1
3008 to XFmode. Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
3009 (log10<mode>2): Ditto.
3010 (log2<mode>2): Ditto.
3011 (log1p<mode>2): Ditto.
3012 (logb<mode>2): Ditto.
3013 (fyl2xp1xf3_i387): Rename from fyl2xp1_xf3.
3014 (fyl2xp1_extend<mode>xf3_i387): New insn pattern.
3015 (*fxtractxf3_i387): Rename from *fxtractxf3.
3016 (fxtract_extend<mode>xf3_i387): New insn pattern.
3017 (ilogbsi2): Use match_dup 3, not match_operand:XF 3.
3018 * config/i386/i386.c (ix86_emit_i387_log1p): Use gen_fyl2xp1xf3_i387()
3019 and gen_fyl2xxf3_i387().
3020
3021 2007-01-14 Zdenek Dvorak <dvorakz@suse.cz>
3022
3023 * loop-unswitch.c (unswitch_loop): Do not call fix_loop_placement.
3024 * cfgloopmanip.c (fix_loop_placement): Made static. Use
3025 get_loop_exit_edges. Changed return type to bool.
3026 * cfgloop.h (fix_loop_placement): Declaration removed.
3027
3028 2007-01-14 Dorit Nuzman <dorit@il.ibm.com>
3029
3030 * param.h (MIN_VECT_LOOP_BOUND): New.
3031 * params.def (MIN_VECT_LOOP_BOUND): New.
3032 * tree-vectorizer.c (slpeel_tree_peel_loop_to_edge): Takes another
3033 argument - minimum threshold for number of iterations.
3034 * tree-vectorizer.h (slpeel_tree_peel_loop_to_edge): Add another
3035 argument to declaration.
3036 * tree-vect-analyze.c (vect_analyze_operations): Check value of
3037 MIN_VECT_LOOP_BOUND.
3038 * tree-vect-transform.c (vect_do_peeling_for_loop_bound): Call
3039 slpeel_tree_peel_loop_to_edge with additional argument.
3040 (vect_do_peeling_for_alignment): Likewise.
3041 * doc/invoke.texi (min-vect-loop-bound): Document new param option.
3042
3043 2007-01-14 Uros Bizjak <ubizjak@gmail.com>
3044
3045 PR target/30413
3046 * config/i386/i386.c (print_operand) ['z']: Output 'b' for
3047 operands of size 1.
3048
3049 2007-01-14 Jan Hubicka <jh@suse.cz>
3050
3051 * tree-dfa.c (remove_referenced_var): New function.
3052 * tree-ssa-live.c (remove_unused_locals): Walk referenced vars and
3053 prune referenced vars list too.
3054 * tree-flow.h (remove_referenced_var): Declare.
3055
3056 2007-01-14 Jan Hubicka <jh@suse.cz>
3057
3058 * tree-eh.c (add_stmt_to_eh_region_fn): Do not add call_exprs
3059 separately.
3060 (remove_stmt_from_eh_region_fn): Do not remove call_exprs.
3061 (verify_eh_throw_stmt_node, verify_eh_throw_table_statements): Kill.
3062 * except.h (verify_eh_throw_table_statements): Kill prototype.
3063 * cfgexpand.c (expand_gimple_basic_block): Propagate Eh regions
3064 into call exrepssions.
3065 * tree-optimize.c (execute_free_cfg_annotatiosn): Do not call
3066 eh trhow verifier.
3067 * tree-cfg.c: Include pointer-set.h.
3068 (verify_node_sharing): Work on pointer set.
3069 (verify_eh_throw_stmt_node): New.
3070 (verify_stmts): Use pointers sets, verify throw_stmt.
3071
3072 2007-01-13 Zdenek Dvorak <dvorakz@suse.cz>
3073
3074 * ipa-reference.c (analyze_function): Consider also addresses taken
3075 in phi nodes.
3076
3077 2007-01-12 Roger Sayle <roger@eyesopen.com>
3078
3079 * c-typeck.c (null_pointer_constant_p): Replace use of
3080 TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
3081 (build_c_cast): Likewise.
3082
3083 2007-01-12 Roger Sayle <roger@eyesopen.com>
3084
3085 * tree.h (force_fit_type_double): Remove unused final argument.
3086 * c-common.c (constant_expression_warning): Replace use of
3087 TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
3088 (convert_and_check): Likewise.
3089 (shorten_compare): Update call to force_fit_type_double.
3090 (c_common_truthvalue_conversion) <INTEGER_CST>: Use integer_zerop.
3091 * convert.c (convert_to_pointer): Update call to
3092 force_fit_type_double.
3093 * fold-const.c (force_fit_type_double): Remove overflowed_const
3094 argument.
3095 (int_const_binop, fold_convert_const_int_from_int,
3096 fold_convert_const_int_from_real, fold_div_compare,
3097 fold_sign_changed_comparison, fold_unary, fold_negate_const,
3098 fold_abs_const, fold_not_const): Remove the final argument from
3099 calls to force_fit_type_double.
3100
3101 2007-01-12 Andrew Pinski <andrew_pinski@playstation.sony.com>
3102
3103 * configure.ac: Set insn to "nop" for spu-*-* also.
3104 * configure: Regenerate.
3105
3106 2007-01-12 Olga Golovanevsky <olga@il.ibm.com>
3107
3108 * builtins.def : Add BUILT_IN_FREE.
3109
3110 2007-01-12 Jan Hubicka <jh@suse.cz>
3111
3112 PR tree-optimization/30443
3113 * tree-inline.c (tree_function_versioning): Do not optimize when
3114 cloning for inlining.
3115
3116 2007-01-12 Zdenek Dvorak <dvorakz@suse.cz>
3117
3118 * doc/loop.texi: Document recording of loop exits.
3119 * cfgloopmanip.c (loopify, duplicate_loop): Use alloc_loop.
3120 (update_single_exits_after_duplication,
3121 update_single_exit_for_duplicated_loop,
3122 update_single_exit_for_duplicated_loops): Removed.
3123 (duplicate_loop_to_header_edge): Do not call
3124 update_single_exits_after_duplication and
3125 update_single_exit_for_duplicated_loops.
3126 (loop_version): Do not update single_exit information.
3127 (fix_loop_structure): Use record_loop_exits instead of
3128 mark_single_exit_loops.
3129 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update
3130 the lists of loop exits.
3131 * cfghooks.c (redirect_edge_and_branch, redirect_edge_and_branch_force,
3132 split_edge, merge_blocks): Update the lists of loop exits.
3133 * modulo-sched.c (sms_schedule): Pass LOOPS_HAVE_RECORDED_EXITS to
3134 loop_optimizer_init.
3135 * loop-init.c (loop_optimizer_init): Call record_loop_exits instead
3136 of mark_single_exit_loops.
3137 (loop_optimizer_finalize): Call release_recorded_exits.
3138 * tree-ssa-loop.c (tree_loop_optimizer_init): Pass
3139 LOOPS_HAVE_RECORDED_EXITS to loop_optimizer_init.
3140 * tree-vectorizer.c (slpeel_tree_duplicate_loop_to_edge_cfg): Do not
3141 update single exit information.
3142 * lambda-code.c (perfect_nestify): Ditto.
3143 * cfgloop.c (flow_loop_free): Destroy the list of exits of the loop.
3144 (mark_single_exit_loops): Removed.
3145 (alloc_loop, loop_exit_hash, loop_exit_eq, loop_exit_free,
3146 get_exit_descriptions, rescan_loop_exit, record_loop_exits,
3147 dump_recorded_exit, dump_recorded_exits, release_recorded_exits): New
3148 functions.
3149 (get_loop_exit_edges, single_exit): Use recorded exit lists.
3150 (add_bb_to_loop, remove_bb_from_loops): Update the lists of loop exits.
3151 (verify_loop_structure): Verify consistency of the exit lists.
3152 (flow_loops_find): Use alloc_loop. Initialize exits hash.
3153 (set_single_exit): Removed.
3154 * cfgloop.h (struct loop_exit): New function.
3155 (struct loop): single_exit_ field replaced by exits field.
3156 (LOOPS_HAVE_MARKED_SINGLE_EXITS): Replaced by LOOPS_HAVE_RECORDED_EXITS.
3157 (struct loops): Added exits hash.
3158 (mark_single_exit_loops, set_single_exit): Declaration removed.
3159 (release_recorded_exits, record_loop_exits, rescan_loop_exit): Declare.
3160
3161 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3162
3163 * doc/invoke.texi: Avoid use of @headitem.
3164
3165 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3166
3167 * cse.c (cse_insn): Move HAVE_CC0 code after declarations.
3168
3169 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3170
3171 * doc/install.texi: Fix m68k-*-* anchor and add m68k-*-* to the
3172 list of targets.
3173
3174 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
3175 Richard Sandiford <richard@codesourcery.com>
3176
3177 * doc/invoke.texi: Document -mno-bitfield, -mno-rtd and -mno-short.
3178 * config/m68k/m68k.opt: Resort options.
3179 (mbitfield, mrtd, mshort): Remove RejectNegative properties.
3180
3181 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
3182 Richard Sandiford <richard@codesourcery.com>
3183
3184 * doc/invoke.texi: Document the macros that are defined by
3185 m68k's -mtune and -mhard-float options.
3186 * config/m68k/m68k-protos.h (m68k_cpp_cpu_ident) Declare.
3187 (m68k_cpp_cpu_family): Likewise.
3188 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Add a full set
3189 of __ucfv*__ macros. Define __mcffpu__ if generating code for
3190 ColdFire FPUs. Define __mcf_cpu_* and __mcf_family_* macros.
3191 * config/m68k/m68k.c (m68k_cpp_cpu_ident): New function.
3192 (m68k_cpp_cpu_family): Likewise.
3193
3194 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3195
3196 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Treat all mc68*
3197 macros besides mc68000 as tuning macros. Use a switch statement
3198 to set them and mcpu32.
3199
3200 2007-01-12 Julian Brown <julian@codesourcery.com>
3201
3202 * config/m68k/m68k.h: Use TARGET_68040 instead of TARGET_68040_ONLY.
3203 (TARGET_68040_ONLY): Rename to...
3204 (TARGET_68040): ...this.
3205 * config/m68k/m68k.c: Use TARGET_68040 instead of TARGET_68040_ONLY.
3206 * config/m68k/m68k.md: Likewise.
3207
3208 2007-01-12 Julian Brown <julian@codesourcery.com>
3209 Nathan Sidwell <nathan@codesourcery.com>
3210 Richard Sandiford <richard@codesourcery.com>
3211
3212 * config.gcc (m680[012]0-*-*, m68k*-*-*): Set m68k_cpu_ident to
3213 the -mcpu= argument associated with the --with-cpu setting.
3214 Define M68K_DEFAULT_TUNE to the default -mtune= option,
3215 if different from the one implied by the -mcpu setting.
3216 Accept --with-cpu=FOO if FOO is listed in m68k-devices.def,
3217 using mcpu=FOO as the default CPU option. Set target_cpu_default2.
3218 * doc/invoke.texi: Mention ColdFire in the introduction to the
3219 m68k options. Document the new -march, -mcpu, -mtune, -mdiv,
3220 -mno-div and -mhard-float options. Make -m68881 a synonym for
3221 -mhard-float. Document the previously-undocumented -m5206e,
3222 -m528x, -m5307 and -m5407 options. Tweak the existing option
3223 documentation for consistency.
3224 * doc/install.texi: Mention new --with-cpu arguments.
3225 * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Only use the
3226 default CPU if neither -mcpu nor -march are specified.
3227 (ASM_CPU_SPEC): Pass down -mcpu and -march options.
3228 (TARGET_CPU_CPP_BUILTINS): Set __mcfisa*__ macros from
3229 TARGET_ISA*. Set the legacy __mcf*__ cpu macros in the same way,
3230 using m68k_tune to decide between families that implement the
3231 same ISA. Use m68k_tune to set __mcfv4e__.
3232 (FL_BITFIELD, FL_68881, FL_COLDFIRE, FL_CF_HWDIV, FL_CF_MAC)
3233 (FL_CF_EMAC, FL_CF_EMAC_B, FL_CF_USP, FL_CF_FPU, FL_ISA_68000)
3234 (FL_ISA_68010, FL_ISA_68020, FL_ISA_68040, FL_ISA_A, FL_ISA_B)
3235 (FL_ISA_C, FL_ISA_MMU): New macros.
3236 (MASK_COLDFIRE): Delete.
3237 (TARGET_68010, TARGET_68020, TARGET_68040_ONLY, TARGET_COLDFIRE)
3238 (TARGET_ISAB): Redefine in terms of m68k_cpu_flags.
3239 (TARGET_68881, TARGET_COLDFIRE_FPU): Redefine in terms of m68k_fpu.
3240 (TARGET_HARD_FLOAT): Do not define here.
3241 (TARGET_ISAAPLUS, TARGET_ISAC): New macros.
3242 (TUNE_68000): New macro.
3243 (TUNE_68000_10): Redefine in terms of TUNE_68000 and TUNE_68010.
3244 (TUNE_68010, TUNE_68030, TUNE_68040, TUNE_68060, TUNE_CPU32)
3245 (TUNE_CFV2): Redefine in terms of m68k_tune.
3246 (uarch_type, target_device, fpu_type): New enums.
3247 (m68k_cpu, m68k_tune, m68k_fpu, m68k_cpu_flags): Declare.
3248 * config/m68k/m68k.c (TARGET_DEFAULT): Remove MASK_68881.
3249 (FL_FOR_isa_00, FL_FOR_isa_10, FL_FOR_isa_20, FL_FOR_isa_40)
3250 (FL_FOR_isa_cpu32, FL_FOR_isa_a, FL_FOR_isa_aplus, FL_FOR_isa_b)
3251 (FL_FOR_isa_c): New macros.
3252 (m68k_isa): New enum.
3253 (m68k_target_selection): New structure.
3254 (all_devices, all_isas, all_microarchs): New tables.
3255 (m68k_cpu_entry, m68k_arch_entry, m68k_tune_entry, m68k_cpu)
3256 (m68k_tune, m68k_fpu, m68k_cpu_flags): New variables.
3257 (MASK_ALL_CPU_BITS): Delete.
3258 (m68k_find_selection): New function.
3259 (m68k_handle_option): Handle -mcpu=, -march= and -mtune=.
3260 Map the legacy target options to a combination of the new ones.
3261 (override_options): Set m68k_cpu, m68k_tune, m68k_fpu and
3262 m68k_cpu_flags. Handle M68K_DEFAULT_TUNE. Use m68k_cpu_flags
3263 to derive default MASK_BITFIELD, MASK_CF_HWDIV and MASK_HARD_FLOAT
3264 settings.
3265 * config/m68k/m68k.opt (m5200, m5206e, m528x, m5307, m5407, mcfv4e)
3266 (m68010, m68020, m68020-40, m68020-60, m68030, m68040): Remove Mask
3267 properties.
3268 (m68881, msoft-float): Change mask from 68881 to HARD_FLOAT.
3269 (march=, mcpu=, mdiv, mhard-float, mtune=): New options.
3270 * config/m68k/m68k-devices.def: New file.
3271
3272 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3273 Nathan Sidwell <nathan@codesourcery.com>
3274
3275 * config/m68k/m68k.h (ASM_CPU_SPEC, ASM_SPEC, EXTRA_SPECS)
3276 (SUBTARGET_EXTRA_SPECS): New macros.
3277 * config/m68k/linux.h (ASM_SPEC): Remove CPU flags;
3278 use %(asm_cpu_spec) instead.
3279 * config/m68k/m68k-none.h (ASM_SPEC): Likewise.
3280 * config/m68k/openbsd.h (ASM_SPEC): Likewise.
3281 * config/m68k/netbsd-elf.h (ASM_SPEC): Likewise.
3282 (EXTRA_SPECS): Rename to...
3283 (SUBTARGET_EXTRA_SPECS): ...this.
3284
3285 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
3286 Richard Sandiford <richard@codesourcery.com>
3287 Julian Brown <julian@codesourcery.com>
3288
3289 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
3290 (m68k-*-uclinux*, m68k-*-linux*, m68k-*-rtems*): Set default_m68k_cpu
3291 to the configuration's default CPU.
3292 (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*): Likewise.
3293 Remove default masks.
3294 (m680[012]0-*-*): Set the default with_cpu to the first part of
3295 the target name.
3296 (m68k*-*-*): Set the default with_cpu to m$default_m68k_cpu.
3297 (m68k*-*-linux): Extend the --with-cpu handling to...
3298 (m680[012]0-*-*, m68k*-*-*): ...these configurations. Allow m68000
3299 and m68010. Don't set target_cpu_default2.
3300 * doc/install.texi: Document --with-cpu for m68k.
3301 * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Define.
3302 * config/m68k/m68k-none.h (TARGET_CPU_DEFAULT, M68K_CPU_m68k)
3303 (M68K_CPU_m68000, M68K_CPU_m68010, M68K_CPU_m68020, M68K_CPU_m68030)
3304 (M68K_CPU_m68040, M68K_CPU_m68302, M68K_CPU_m68332, TARGET_DEFAULT)
3305 (ASM_CPU_DEFAULT_SPEC, CC1_CPU_DEFAULT_SPEC): Delete.
3306 (ASM_SPEC): Remove use of %(asm_cpu_default).
3307 (EXTRA_SPECS, SUBTARGET_EXTRA_SPECS, MULTILIB_DEFAULTS): Delete.
3308 * config/m68k/linux.h (TARGET_DEFAULT): Delete.
3309 (CPP_SPEC): Merge definitions. Do not handle __HAVE_68881__ here.
3310 * config/m68k/netbsd-elf.h (TARGET_OS_CPP_BUILTINS): Define
3311 __HAVE_FPU__ if TARGET_HARD_FLOAT.
3312 (TARGET_DEFAULT): Delete.
3313 (EXTRA_SPECS): Delete cpp_cpu_default_spec, cpp_cpu_spec,
3314 cpp_fpu_spec, asm_default_spec and netbsd_cpp_spec.
3315 (CPP_CPU_SPEC): Delete.
3316 (TARGET_VERSION): Merge definitions, using TARGET_68010 to pick
3317 the appropriate string.
3318 (CPP_CPU_DEFAULT_SPEC, ASM_DEFAULT_SPEC, CPP_FPU_SPEC): Delete.
3319 (CPP_SPEC): Define to NETBSD_CPP_SPEC.
3320 (ASM_SPEC): Don't use %(asm_default_spec).
3321 * config/m68k/m68k.c (TARGET_DEFAULT_TARGET_FLAGS): Remove
3322 TARGET_DEFAULT and add MASK_68881.
3323 * config/m68k/m68k.md: Remove mention of TARGET_DEFAULT from comments.
3324
3325 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3326
3327 * config.gcc (m68010-*-netbsdelf*): Add MASK_68010.
3328 (m68k*-*-netbsdelf*, m68k*-*-openbsd*, m68k*-linux*): Add
3329 MASK_68010 alongside MASK_68020.
3330 * doc/invoke.texi: Document -m68010.
3331 * config/m68k/m68k.opt (m68010): New.
3332 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Define mc68010
3333 if TUNE_68010.
3334 (TUNE_68010): New macro.
3335 * config/m68k/m68k-none.h (M68K_CPU_m68k, M68K_CPU_m68010)
3336 (M68K_CPU_m68020, M68K_CPU_m68030, M68K_CPU_m68040)
3337 (M68K_CPU_m68332): Add MASK_68010.
3338 * config/m68k/linux.h (TARGET_DEFAULT): Add MASK_68010 to
3339 fallback definition.
3340 * config/m68k/netbsd-elf.h (CPP_CPU_SPEC): Remove now-redundant
3341 defines.
3342 * config/m68k/m68k.c (MASK_ALL_CPU_BITS): Add MASK_68010.
3343 (m68k_handle_option): Handle OPT_m68010. Add MASK_68010
3344 to all entries that use MASK_68020.
3345 (output_move_simode_const, output_move_himode, output_move_qimode)
3346 (output_move_stricthi, output_move_strictqi): Use TARGET_68010
3347 instead of TARGET_68020 to select clr behavior. Remove comment
3348 about there being no TARGET_68010.
3349 * config/m68k/m68k.md: Likewise throughout.
3350
3351 2007-01-12 Julian Brown <julian@codesourcery.com>
3352
3353 * config/m68k/m68k.h (TARGET_ISAB): New macro.
3354 * config/m68k/m68k.c: Use TARGET_ISAB rather than TARGET_CFV4.
3355 * config/m68k/m68k.md: Likewise.
3356
3357 2007-01-12 Julian Brown <julian@codesourcery.com>
3358
3359 * config/m68k/m68k.h (LEGITIMATE_INDEX_P, LEGITIMIZE_ADDRESS): Use
3360 TARGET_COLDFIRE_FPU instead of TARGET_CFV4E.
3361
3362 2007-01-12 Julian Brown <julian@codesourcery.com>
3363
3364 * config/m68k/m68k.h (TUNE_68040_60): New macro.
3365 * config/m68k/m68k.c (standard_68881_constant_p): Use it.
3366 * config/m68k/m68k.md: Likewise.
3367
3368 2007-01-12 Julian Brown <julian@codesourcery.com>
3369 Richard Sandiford <richard@codesourcery.com>
3370
3371 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Use TUNE_68030
3372 instead of TARGET_68030, TUNE_68040 instead of TARGET_68040,
3373 TUNE_68060 instead of TARGET_68060 and TUNE_CPU32 instead of
3374 TARGET_CPU32.
3375 (TARGET_CPU32): Rename to...
3376 (TUNE_CPU32): ...this.
3377 (TUNE_68000_10, TUNE_68030, TUNE_68040, TUNE_68060)
3378 (TUNE_CFV2): New macros.
3379 * config/m68k/netbsd-elf.h (LONG_DOUBLE_TYPE_SIZE): Simplify;
3380 remove conditions that are implied by TARGET_68020.
3381 * config/m68k/m68k.c (m68k_output_function_prologue): Use TUNE_68040
3382 instead of TARGET_68040 and TUNE_CPU32 instead of TARGET_CPU32.
3383 (m68k_output_function_epilogue): Likewise.
3384 (m68k_rtx_costs): Likewise. Use TUNE_68060 instead of TARGET_68060
3385 and TUNE_CFV2 instead of TARGET_5200. Use TUNE_68000_10 instead of
3386 "!TARGET_68020 && !TARGET_COLDFIRE" to choose between 68000 and
3387 non-68000 timings. Refactor multiplication and division costs.
3388 (output_addsi3): Use TUNE_68040 instead of TARGET_68040 and
3389 TUNE_CPU32 instead of TARGET_CPU32.
3390 (standard_68881_constant_p): Use TUNE_68040 instead of TARGET_68040
3391 and TUNE_68060 instead of TARGET_68060.
3392 * config/m68k/m68k.md: Use TUNE_68040 instead of TARGET_68040,
3393 TUNE_68060 instead of TARGET_68060, and TUNE_CPU32 instead of
3394 TARGET_CPU32.
3395 (movsi_const0): Use TUNE_68000_10 rather than "!TARGET_68020
3396 && !TARGET_COLDFIRE" to choose between moveq and clr.
3397 Likewise in the unnamed movsf pattern.
3398 (ashlsi_17_24, lshrsi_17_24): Guard with TUNE_68000_10 rather than
3399 "!TARGET_68020 && !TARGET_COLDFIRE". Likewise the unnamed
3400 ashiftrt pattern.
3401
3402 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3403
3404 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Increase amount
3405 of tabbing before backslashes.
3406
3407 2007-01-11 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
3408
3409 * pa-linux.h (ASM_OUTPUT_INTERNAL_LABEL): Undefine.
3410 * pa.h (ASM_OUTPUT_LABEL): Output colon when using GAS.
3411 (ASM_OUTPUT_INTERNAL_LABEL): Define.
3412
3413 2007-01-11 Zdenek Dvorak <dvorakz@suse.cz>
3414
3415 * tree-ssa-loop-ivopts.c (extract_cond_operands): Split from
3416 find_interesting_uses_cond.
3417 (find_interesting_uses_cond): Use extract_cond_operands.
3418 (rewrite_use_compare): Use extract_cond_operands and
3419 force_gimple_operand_bsi. Do not call update_stmt.
3420 (determine_use_iv_cost_condition): Use extract_cond_operands.
3421 Return cheaper of using original bound and changing the exit bound.
3422
3423 2007-01-11 Zdenek Dvorak <dvorakz@suse.cz>
3424
3425 PR tree-optimization/29516
3426 * tree-ssa-address.c (tree_mem_ref_addr, add_to_parts,
3427 most_expensive_mult_to_index, addr_to_parts,
3428 create_mem_ref, maybe_fold_tmr): Make the type of
3429 fields of TARGET_MEM_REF sizetype.
3430 (move_fixed_address_to_symbol, move_pointer_to_base):
3431 New functions.
3432 * tree.def (TARGET_MEM_REF): Add comment on types of
3433 the operands.
3434
3435 2007-01-11 Joseph Myers <joseph@codesourcery.com>
3436
3437 * c-common.c (vector_types_convertible_p): Treat opaque types as
3438 always convertible if they have the same size, but not otherwise.
3439
3440 2007-01-11 Steven Bosscher <steven@gcc.gnu.org>
3441
3442 * ifcvt.c (struct noce_if_info): Add comments to the fields.
3443 Remove the b_unconditional field.
3444 (noce_try_sign_mask): Do not look at b_unconditional.
3445 (noce_process_if_block): Do not use merge_if_blocks. Update
3446 the CFG here. Do not set b_unconditional.
3447 (cond_move_process_if_block): Likewise.
3448 (find_cond_trap): Likewise.
3449 (check_cond_move_block): Require simple jump insns at the end
3450 of the basic block.
3451
3452 2007-01-11 Jan Hubicka <jh@suse.cz>
3453
3454 PR tree-optimization/1046
3455 * tree-tailcall.c (suitable_for_tail_call_opt_p): Use TREE_ADDRESSABLE
3456 when alias info is not ready.
3457 (pass_tail_recursion): Do not require aliasing.
3458 * tree-ssa-copyrename.c (pass_rename_ssa_cop): Likewise.
3459 * tree-ssa-ccp.c (pass_ccp, pass_fold_builtins): Likewise.
3460 * tree-ssa-copy.c (pass_copy_prop): Likewise.
3461 * tree-ssa-forwprop.c (pass_forwprop): Likewise.
3462 * tree-ssa-dce.c (pass_dce, pass_dce_loop, pass_cd_dce): Likewise.
3463 * passes.c (init_optimization_passes): Execute rename_ssa_copies,
3464 ccp, forwprop, copy_prop, merge_phi, copy_prop, dce and tail recursion
3465 before inlining.
3466 * tree-ssa-operands.c (add_virtual_operand, get_indirect_ref_operand):
3467 When aliasing is not build, mark statement as volatile.
3468
3469 2007-01-11 Tom Tromey <tromey@redhat.com>
3470
3471 PR preprocessor/15185, PR preprocessor/20989:
3472 * doc/cppopts.texi <-MT>: Update description of algorithm for
3473 computing default target.
3474 <-M, -MD>: Reword "basename" text.
3475
3476 2007-01-11 Roger Sayle <roger@eyesopen.com>
3477
3478 * builtins.c (expand_builtin_pow, expand_builtin_powi,
3479 fold_builtin_cabs, fold_builtin_sqrt, fold_builtin_trunc,
3480 fold_builtin_floor, fold_builtin_ceil, fold_builtin_round,
3481 fold_builtin_int_int_roundingfn, fold_builtin_bitop,
3482 fold_builtin_bswap, real_constp, fold_builtin_pow,
3483 fold_builtin_powi, fold_builtin_signbit, fold_builtin_copysign,
3484 do_mpfr_arg1, do_mpfr_arg2, do_mpfr_arg3, do_mpfr_sincos): Replace
3485 uses of the macro TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
3486 * convert.c (convert_to_pointer): Likewise.
3487 * expr.c (highest_pow2_factor, expand_expr_real_1): Likewise.
3488 * fold-const.c (force_fit_type, fold_negate_expr, int_const_binop,
3489 const_binop, fold_convert_const_int_from_int,
3490 fold_convert_const_int_from_real,
3491 fold_convert_const_real_from_real, sign_bit_p,
3492 optimize_minmax_comparison, extract_muldiv_1, fold_div_compare,
3493 fold_sign_changed_comparison, fold_unary, fold_comparison,
3494 fold_binary, multiple_of_p, tree_Expr_non_zero_p,
3495 fold_negate_const, fold_abs_const, fold_not_const): Likewise.
3496 * print-tree.c (print_node_brief, print_node): Likewise.
3497 * stor-layout.c (place_field, layout_type): Likewise.
3498 * tree-chrec.c (keep_cast): Likewise.
3499 * tree.c (build_vector, build_real, build_real_from_int_cst,
3500 build_complex): Likewise.
3501
3502 2007-01-11 Roger Sayle <roger@eyesopen.com>
3503
3504 * tree.h (TREE_CONSTANT_OVERFLOW): Obsolete. For the time being,
3505 treat TREE_CONSTANT_OVERFLOW as a synonym of TREE_OVERFLOW.
3506
3507 2007-01-11 Paolo Bonzini <bonzini@gnu.org>
3508
3509 * configure.ac (strict1_warn): Rename to strict_warn.
3510 (WERROR, --enable-werror, symlink hacks, stage1_cflags,
3511 cc_set_by_configure, quoted_cc_set_by_configure,
3512 stage_prefix_set_by_configure, quoted_stage_prefix_set_by_configure,
3513 all_boot_languages, all_stagestuff): Remove.
3514 (target_list): Remove bootstrap targets.
3515 * Makefile.in (quickstrap): Unconditionally make a synonym of all.
3516 (BOOT_LANGUAGES, STAGE1_CFLAGS, STAGE1_CHECKING,
3517 REMAKEFLAGS, FLAGS_TO_PASS, PREPEND_DOTDOT_TO_RELATIVE_PATHS,
3518 SUBDIR_FLAGS_TO_PASS, WERROR_FLAGS, STRICT2_WARN, LANG_STAGESTUFF,
3519 VOL_FILES, POSTSTAGE1_FLAGS_TO_PASS, STAGE2_FLAGS_TO_PASS,
3520 STAGEPROFILE_FLAGS_TO_PASS, STAGEFEEDBACK_FLAGS_TO_PASS, stage1_build,
3521 stage1_copy, stage2_build, stage2_copy, stageprofile_build,
3522 stageprofile_copy, stage3_build, stage3_copy, stagefeedback_build,
3523 stagefeedback_copy, stage4_build, clean_s1, clean_sw, bootstrap,
3524 bootstrap-lean, bootstrap2, bootstrap2-lean, bootstrap3,
3525 bootstrap3-lean, bootstrap4, bootstrap4-lean, unstage1, unstage2,
3526 unstage3, unstage4, unstageprofile, unstagefeedback, restage, restage2,
3527 restage3, restage4, restageprofile, restagefeedback, bubbleestrap,
3528 cleanstrap, unstrap, restrap, *compare, *compare3, *compare4,
3529 *compare-lean, *compare3-lean, *compare4-lean, stage1-start, stage1,
3530 stage2-start, stage2, stage3-start, stage3, stage4-start, stage4,
3531 stageprofile-start, stageprofile, stagefeedback-start, stagefeedback,
3532 risky-stage1, risky-stage2, risky-stage3, risky-stage4): Remove.
3533 (ORDINARY_FLAGS_TO_PASS): Rename to FLAGS_TO_PASS.
3534 (STAGECOPYSTUFF, STAGEMOVESTUFF): Consolidate into MOSTLYCLEANFILES.
3535 (mostlyclean): Adjust.
3536 (clean, distclean): Don't mention bootstrap stuff.
3537 * configure: Regenerate.
3538 * ada/config-lang.in, cp/config-lang.in, forttran/config-lang.in,
3539 java/config-lang.in, objc/config-lang.in, objcp/config-lang.in,
3540 treelang/config-lang.in (stagestuff): Remove.
3541 * doc/sourcebuild.texi (stage1, stage2, stage3, stage4,
3542 stageprofile, stagefeedback, stagestuff): Remove mention.
3543
3544 2007-01-11 Nick Clifton <nickc@redhat.com>
3545
3546 * config/mcore/predicates.md (mcore_general_movesrc_operand):
3547 Accept CONSTs.
3548 (mcore_general_movdst_operand): Do not accept CONST_INTs.
3549 (mcore_arith_K_S_operand): Run the test for the S constraint not
3550 the test for the M constraint.
3551 (mcore_addsub_operand): Do not accept integer values that are
3552 larger than 32 bits.
3553 * config/mcore/mcore.md: Remove unused constraints from split.
3554 (andsi3): Use HOST_WIDE_INT instead of int to hold an INTVAL.
3555 (addsi3): Likewise.
3556 (allocate_stack): Likewise.
3557 * config/mcore/mcore.c (mcore_print_operand): Restrict output of P
3558 operands to 32 bits.
3559 (mcore_const_costs): Use HOST_WIDE_INT instead of int to hold an
3560 INTVAL.
3561 (mcore_and_cost, mcore_modify_comparison, const_ok_for_mcore,
3562 mcore_const_ok_for_inline, mcore_const_trick_uses_not,
3563 try_constant_tricks, mcore_num_ones, mcore_num_zeros,
3564 mcore_output_bclri, mcore_output_andn, output_inline_const,
3565 mcore_output_move, mcore_output_movedouble): Likewise.
3566 (mcore_output_cmov): Use CONST_OK_FOR_M and CONST_OK_FOR_N.
3567 (output_inline_const): Likewise.
3568 (output_inline_const): Fix format strings used in sprintf
3569 statements.
3570 * config/mcore/mcore-protos.h: Update prototypes for changed
3571 functions in mcore.c.
3572 * config/mcore/mcore.h (CONST_OK_FOR_I): Cast values to
3573 HOST_WIDE_INT and not int.
3574 (CONST_OK_FOR_J, CONST_OK_FOR_K, CONST_OK_FOR_L, CONST_OK_FOR_M,
3575 CONST_OK_FOR_N): Likewise.
3576 (LEGITIMATE_CONSTANT_P): Also check CONSTANT_P.
3577 (GO_IF_LEGITIMATE_INDEX): Use HOST_WIDE_INT instead of int to hold
3578 an INTVAL.
3579
3580 2007-01-10 Jan Hubicka <jh@suse.cz>
3581
3582 * tree-vrp.c (remove_range_assertions): Release defs.
3583 * tree-ssa-loop-ivopts.c (rmeove_statement): Likewise.
3584 * tree-ssa-dom.c (remove_stmt_or_phi): Likewise.
3585
3586 2007-01-10 Paul Brook <paul@codesourcery.com>
3587
3588 * config/arm/arm.c (arm_rtx_costs_1): Handle mutiply-subtract.
3589 * config/arm/arm.md (mulsi3subsi): New insn.
3590
3591 2007-01-10 Zdenek Dvorak <dvorakz@suse.cz>
3592
3593 * tree-ssa-loop-manip.c (tree_unroll_loop): Make it a wrapper over ...
3594 (tree_transform_and_unroll_loop): New.
3595 * tree-flow.h (transform_callback, tree_transform_and_unroll_loop):
3596 Declare.
3597
3598 2007-01-10 Robert Kennedy <jimbob@google.com>
3599
3600 * fold-const.c (fold_comparison): Fold comparisons like (x *
3601 1000 < 0) to (x < 0).
3602
3603 2007-01-10 Ian Lance Taylor <iant@google.com>
3604
3605 * tree-pretty-print.c (dump_generic_node): Print parentheses when
3606 operands have the same priority.
3607
3608 2007-01-10 Tom Tromey <tromey@redhat.com>
3609
3610 * fold-const.c (fold_truthop): Don't check can_use_bit_fields_p.
3611 (fold_binary): Likewise.
3612 * langhooks.c (lhd_can_use_bit_fields_p): Removed.
3613 * langhooks-def.h (lhd_can_use_bit_fields_p): Removed.
3614 (LANG_HOOKS_CAN_USE_BIT_FIELDS_P): Removed.
3615 (LANG_HOOKS_INITIALIZER): Remove LANG_HOOKS_CAN_USE_BIT_FIELDS_P.
3616 * langhooks.h (struct lang_hooks): Removed field
3617 'can_use_bit_fields_p'.
3618
3619 2007-01-10 Ralf Corsépius <ralf.corsepius@rtems.org>
3620
3621 * config/bfin/t-bfin, config/bfin/t-bfin-elf: Remove GCC_CFLAGS.
3622
3623 2007-01-10 Razya Ladelsky <razya@il.ibm.com>
3624
3625 * function.c (get_last_funcdef_no): New function.
3626 * function.h (get_last_funcdef_no): Declare.
3627 * tree-inline.c (initialize_cfun): Add initialization.
3628 (tree_function_versioning): Cleanup.
3629
3630 2007-01-10 Jan Hubicka <jh@suse.cz>
3631
3632 * tree-inline.c (setup_one_parameter): Do not propagate into abnormal
3633 PHIs.
3634
3635 2007-01-10 Sa Liu <saliu@de.ibm.com>
3636 Ben Elliston <bje@au.ibm.com>
3637
3638 * spu.h (STACK_SAVE_AREA): Use VOIDmode for SAVE_FUNCTION, SImode
3639 for SAVE_NONLOCAL and Pmode for any other save level.
3640 * spu-protos.h (spu_restore_stack_block): Declare.
3641 * spu.md (save_stack_block): Remove.
3642 (restore_stack_block): Call spu_restore_stack_block.
3643 * spu.c (spu_restore_stack_block): New function.
3644 (spu_expand_epilogue): Remove old comment.
3645
3646 2007-01-09 Zdenek Dvorak <dvorakz@suse.cz>
3647
3648 PR tree-optimization/30322
3649 * tree-ssa-loop-ivopts.c (fold_affine_expr, iv_value): Removed.
3650 (cand_value_at): Return the value as aff_tree.
3651 (may_eliminate_iv): Convert the bound from aff_tree to tree.
3652 * tree-affine.c (aff_combination_add_cst, aff_combination_add_product,
3653 aff_combination_mult): New functions.
3654 (aff_combination_add): Use aff_combination_add_cst.
3655 (aff_combination_convert): Allow conversions to a wider type.
3656 (tree_to_aff_combination): Handle BIT_NOT_EXPR.
3657 * tree-affine.h (aff_combination_mult): Declare.
3658
3659 2007-01-09 Carlos O'Donell <carlos@codesourcery.com>
3660
3661 * doc/tm.texi: Update documentation to reflect reality of exec
3662 and start file search behaviours. Update copyright year.
3663 * doc/invoke.texi: Explain how GCC_EXEC_PREFIX is used to find
3664 header file directories.
3665
3666 2007-01-09 Uros Bizjak <ubizjak@gmail.com>
3667
3668 * config/i386/i386.md (*sinxf2): Rename to *sinxf2_i387.
3669 (*cosxf2): Rename to cosxf2_i387.
3670 (*sindf2, *sinsf2): Extend operand 1 to XFmode. Macroize patterns
3671 using X87MODEF12 mode macro. Rename patterns to
3672 *sin_extend<mode>xf2_i387. Use SSE_FLOAT_MODE_P to disable patterns
3673 for SSE math.
3674 (*cosdf2, *cossf2): Ditto.
3675 (sincosdf3, sincossf3): Ditto. Rewrite corresponding splitters
3676 to match extended input operands.
3677 (sincos<mode>3): New expander.
3678 (*sinextendsfdf2, *cosextendsfdf2, *sincosextendsfdf3): Remove
3679 insn patterns and corresponding splitters.
3680
3681 2007-01-09 Kaz Kojima <kkojima@gcc.gnu.org>
3682
3683 * config/sh/t-linux (TARGET_LIBGCC2_CFLAGS): Delete.
3684 (SHLIB_MAPFILES, SHLIB_LINK, SHLIB_INSTALL): Likewise.
3685
3686 2007-01-09 Nicolas Pitre <nico@cam.org>
3687
3688 PR target/30173
3689 * arm/ieee754-df.S (Lad_s): Also test the low word of X for zero.
3690
3691 2007-01-08 Geoffrey Keating <geoffk@apple.com>
3692
3693 * target.h (struct gcc_target): New field library_rtti_comdat.
3694 * target-def.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): New.
3695 (TARGET_CXX): Add TARGET_CXX_LIBRARY_RTTI_COMDAT.
3696 * doc/tm.texi (C++ ABI): Document TARGET_CXX_LIBRARY_RTTI_COMDAT.
3697 * config/darwin.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
3698
3699 2007-01-08 Geoffrey Keating <geoffk@apple.com>
3700
3701 * doc/invoke.texi (Optimize Options): Correct description of -O0.
3702
3703 2007-01-08 Richard Guenther <rguenther@suse.de>
3704
3705 * tree.h (force_fit_type_double): Export.
3706 (force_fit_type): Remove.
3707 * fold-const.c (force_fit_type_double): New function.
3708 (force_fit_type): Remove.
3709 (int_const_binop): Use it.
3710 (fold_convert_const_int_from_int): Likewise.
3711 (fold_convert_const_int_from_real): Likewise.
3712 (fold_div_compare): Likewise.
3713 (fold_sign_changed_comparison): Likewise.
3714 (fold_unary): Likewise.
3715 (fold_negate_const): Likewise.
3716 (fold_abs_const): Likewise.
3717 (fold_not_const): Likewise.
3718 * c-common.c (shorten_compare): Use force_fit_type_double.
3719 * convert.c (convert_to_pointer): Likewise.
3720
3721 2007-01-08 Richard Guenther <rguenther@suse.de>
3722
3723 * tree.h (build_int_cst_wide_type): Export.
3724 * tree.c (build_int_cst_wide_type): New function.
3725 (build_int_cst_wide): Fix comment.
3726 * builtins.c (fold_builtin_object_size): Use build_int_cst
3727 to build -1 or 0 of the correct type. Use fit_double_type
3728 to check for overflow.
3729 * fold-const.c (optimize_bit_field_compare): Use build_int_cst_type
3730 to build the mask.
3731 (decode_field_reference): Likewise.
3732 (all_ones_mask_p): Likewise.
3733 (native_interpret_int): Use build_int_cst_wide_type.
3734 (fold_binary): Use build_int_cst_type to build an all-ones
3735 value.
3736 * stor-layout.c (set_sizetype): Use build_int_cst_wide_type.
3737
3738 2007-01-08 Daniel Jacobowitz <dan@codesourcery.com>
3739
3740 * config/pa/t-pa64 (libgcc_stub.a): Use $(T).
3741
3742 2007-01-09 Ben Elliston <bje@au.ibm.com>
3743
3744 * genautomata.c (STATS_OPTION): New option.
3745 (stats_flag): New flag.
3746 (gen_automata_option): Handle it.
3747 (initiate_automaton_gen): Ditto.
3748 (write_automata): Output statistics only if stats_flag is
3749 set. Likewise, output time statistics only if time_flag is set.
3750 * doc/md.texi (Processor pipeline description): Document new flag.
3751
3752 2007-01-08 Richard Guenther <rguenther@suse.de>
3753
3754 * builtins.c (fold_builtin_int_roundingfn): Use fit_double_type.
3755 * tree.c (build_int_cst_type): Likewise.
3756 (size_in_bytes): Don't call force_fit_type on the result.
3757 (int_fits_type_p): Use fit_double_type.
3758 * fold-const.c (fit_double_type): New function.
3759 (force_fit_type): Use it.
3760 * tree.h (fit_double_type): Export.
3761
3762 2007-01-08 Jan Hubicka <jh@suse.cz>
3763
3764 * tree-vectorizer.c (gate_increase_alignment): Fix return type.
3765 * ipa.c (function_and_variable_visibility): Fix return type.
3766
3767 2007-01-08 Richard Guenther <rguenther@suse.de>
3768
3769 * tree-ssa-ccp.c (maybe_fold_offset_to_array_ref): Use type
3770 of offset to build the index.
3771 * tree-pretty-print.c (dump_generic_node): Don't build negated
3772 const just for printing.
3773 * c-pretty-print.c (pp_c_integer_constant): Likewise.
3774 * builtins.c (fold_builtin_int_roundingfn): Check if result
3775 fits the type by using force_fit_type and comparing the result.
3776 * predict.c (predict_loops): Use compare_tree_int for comparison.
3777 * tree.c (build_int_cst): Fall back to integer_type_node for
3778 NULL_TREE type.
3779 (build_int_cst_wide): Assert type is non-null.
3780
3781 2007-01-08 Roberto Costa <roberto.costa@st.com>
3782
3783 * tree-vrp.c (extract_range_from_cond_expr): New.
3784 (extract_range_from_expr): Handle COND_EXPR nodes used as expressions.
3785 * tree-ssa-ccp.c (get_maxval_strlen): Handle COND_EXPR nodes used
3786 as expressions.
3787 (fold_stmt): Bug fix, avoid infinite recursion when folding COND_EXPRs.
3788 * tree-ssa-forwprop.c (simplify_cond, forward_propagate_into_cond,
3789 tree_ssa_forward_propagate_single_use_vars): Handle COND_EXPR nodes
3790 used as expressions.
3791 * tree-object-size.c (cond_expr_object_size): New.
3792 (collect_object_sizes_for): Handle COND_EXPR nodes used as expressions.
3793
3794 2007-01-08 Jan Hubicka <jh@suse.cz>
3795
3796 * tree-ssa-forwprop.c (forward_propagate_into_cond,
3797 tree_ssa_forward_propagate_single_use_va): Release defs of propagated
3798 statement.
3799
3800 2007-01-08 Richard Guenther <rguenther@suse.de>
3801
3802 PR tree-optimization/23603
3803 * tree-vrp.c (set_value_range_to_truthvalue): New function.
3804 (extract_range_from_binary): Fall back to truthvalue instead of
3805 varying for TRUTH_*_EXPR.
3806 (extract_range_from_comparison): Fall back to truthvalue instead of
3807 varying.
3808 (vrp_visit_phi_node): Don't adjust new range bounds to +INF/-INF
3809 if all visited PHI values were constant.
3810
3811 2007-01-08 Jan Hubicka <jh@suse.cz>
3812
3813 * cgraphunit.c (cgraph_process_new_functions): Reset reachable flag.
3814 (cgraph_analyze_function): break out from ...
3815 (cgraph_finalize_compilation_unit): ... here.
3816 (cgraph_expand_function): Remove forgoten commented out line.
3817 (cgraph_optimize): Analyze functions.
3818
3819 2007-01-08 Jan Hubicka <jh@suse.cz>
3820
3821 * tree-pas.h (TODO_remove_function): New flag.
3822 (TODO_update*): Renumber.
3823 (pass_ipa_increase_alignment,
3824 pass_ipa_function_and_variable_visibility): New passes.
3825 * cgraphunit.c (cgraph_increase_alignment): Move to tree-vectorizer.c
3826 (cgraph_function_and_variable_visibility): Move to ipa.c
3827 (cgraph_optimize): Don't call cgraph_function_and_variable_visibility,
3828 cgraph_increase_alignment.
3829 * ipa-inline.c (cgraph_decide_inlining): Don't push timevar.
3830 (cgraph_decide_inlining_incrementally): Push TV_INTEGRATION before
3831 calling tree-inline.
3832 (cgraph_early_inlining): Do not call cgraph_remove_unreachable_nodes.
3833 (pass_ipa_inline, pass_early_ipa_inlining): Set TODO_remove_functions
3834 * tree-vectorizer.c (increase_alignment): Move here from cgraphunit.c
3835 (gate_increase_alignment): New function.
3836 (pass_ipa_increase_alignment): New pass.
3837 * ipa.c: Inline tree-pass.h and timevar.h
3838 (function_and_variable_visibility): Move here from cgraphunit.c
3839 * tree-optimize.c (pass_early_local_passes): Add TODO_remove_functions.
3840 * passes.c (init_optimization_passes): Add the two new passes.
3841 (execute_todo): Handle cgraph_remove_functions.
3842
3843 2007-01-08 Nick Clifton <nickc@redhat.com>
3844
3845 * config/frv/predicates.md (reg_or_0_operand): Accept
3846 CONST_DOUBLEs.
3847
3848 2007-01-08 Ralf Corsépius <ralf.corsepius@rtems.org>
3849
3850 * config/bfin/rtems.h, config/bfin/t-rtems: New.
3851 * config.gcc: Add bfin*-rtems*.
3852
3853 2007-01-08 Mark Shinwell <shinwell@codesourcery.com>
3854
3855 * c.opt: Add -flax-vector-conversions.
3856 * c-typeck.c (convert_for_assignment): Pass flag to
3857 vector_types_convertible_p to allow emission of note.
3858 (digest_init): Likewise.
3859 * c-opts.c: Handle -flax-vector-conversions.
3860 * c-common.c (flag_lax_vector_conversions): New.
3861 (vector_types_convertible_p): Unless -flax-vector conversions
3862 has been passed, disallow conversions between vectors with
3863 differing numbers of subparts and/or element types. If such
3864 a conversion is disallowed, possibly emit a note on the first
3865 occasion only to inform the user of -flax-vector-conversions.
3866 The new last argument specifies this.
3867 * c-common.h (flag_lax_vector_conversions): New.
3868 (vector_types_convertible_p): Add extra argument.
3869 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Use
3870 char_type_node for V*QI type vectors.
3871 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins):
3872 Update to satisfy new typechecking rules.
3873 * config/rs6000/altivec.h (vec_cmple): Use vec_cmpge, for both
3874 C and C++ variants.
3875 * doc/invoke.texi (C Dialect Options): Document
3876 -flax-vector-conversions.
3877
3878 2007-01-08 Mark Shinwell <shinwell@codesourcery.com>
3879
3880 PR tree-optimization/29877
3881 * tree-ssa-ter.c (is_replaceable_p): Deem assignments with
3882 a register variable on the RHS to not be replaceable.
3883
3884 2007-01-08 Chen Liqin <liqin@sunnorth.com.cn>
3885 * config/score/t-score-elf (MULTILIB_OPTIONS): Change.
3886 * config/score/predicates.md (const_uimm5, sr0_operand, const_simm12,
3887 const_simm15, const_pow2, const_npow2): Added.
3888 * config/score/misc.md (insv, extv, extzv, movmemsi,
3889 move_lbu_a/b, mov_lhu_a/b etc): Added and fix some bug.
3890 * config/score/score.c (score_address_cost, score_select_cc_mode):
3891 Added.
3892 Change CONST_OK_FOR_LETTER_P/EXTRA_CONSTRAINT define.
3893 Update score_rtx_costs for MACRO TARGET_RTX_COSTS.
3894 Update score_print_operand.
3895 * config/score/score.h (DATA_ALIGNMENT, SELECT_CC_MODE): Added.
3896 Adjust register allocate order and update some macro define.
3897 * config/score/score-mdaux.c (mdx_unaligned_load, mdx_unsigned_store,
3898 mdx_block_move_straight, mdx_block_move_loop_head,
3899 mdx_block_move_loop_body, mdx_block_move_loop_foot, mdx_block_move_loop,
3900 mdx_block_move): Added.
3901 (mdx_movsicc, mdp_select_add_imm, mdp_select, mds_zero_extract_andi,
3902 mdp_limm): Updated and fix some bug and typo.
3903 * config/score/score.md (movqi/hi/si, add/sub/zero/ext): Updated.
3904 (movsf, movdf, doloop_end): Added.
3905
3906 2007-01-08 Kazu Hirata <kazu@codesourcery.com>
3907
3908 * config/arm/arm.c, config/arm/arm.h, config/arm/arm.md,
3909 config/arm/thumb2.md: Fix comment typos.
3910 * doc/extend.texi: Fix a typo.
3911
3912 2007-01-07 Eric Christopher <echristo@apple.com>
3913
3914 * configure.ac: Check for __stack_chk_fail for darwin.
3915 * configure: Regenerate.
3916
3917 2007-01-07 Richard Guenther <rguenther@suse.de>
3918
3919 * tree-vrp.c (extract_range_from_assert): CSE calls to
3920 compare_values where possible.
3921 (extract_range_from_unary_expr): Likewise.
3922
3923 2007-01-07 Anatoly Sokolov <aesok@post.ru>
3924
3925 * config/avr/avr-protos.h (call_insn_operand): Delete prototype.
3926 * config/avr/avr.c (call_insn_operand): Delete function.
3927 * config/avr/avr.md (*pushqi, *pushhi, *pushsi, *pushsf): Use REG_SP
3928 instead of register number. Use predicates.
3929 * config/avr/predicates.md (const0_operand, reg_or_0_operand,
3930 call_insn_operand): Add.
3931
3932 2007-01-06 Jan Hubicka <jh@suse.cz>
3933
3934 * tree-pass.h (pass_build_cgraph_edges): Declare.
3935 * cgraphunit.c (record_refernece): Move to cgraphbuild.c
3936 (visited_nodes): Remove.
3937 (cgraph_create_edges): Move to cgraphbuild.c; rename to
3938 build_cgrpah_edges; make visited_nodes local.
3939 (cgraph_process_new_functions): DO not call initialize_inline_failed.
3940 (record_references_in_initializer): Move to cgraphbuild.c
3941 (initialize_inline_failed, rebuild_cgraph_edges,
3942 pass_rebuild_cgraph_edges): Move to cgraphbuild.c.
3943 (verify_cgraph_node): Make visited_nodes local.
3944 (cgraph_analyze_function): Do not call cgraph_create_edges and
3945 initialize_inline_failed.
3946 (cgraph_expand_function): Do not call cgraph_lower_function;
3947 assert that function is already lowered.
3948 * Makefile.in (cgraphbuild.o): New.
3949 * passes.c (init_optimization_passes): Add pass_build_cgraph_edges
3950 at the end of lowering passes.
3951
3952 2007-01-06 Steven Bosscher <steven@gcc.gnu.org>
3953
3954 * ifcvt.c (cond_move_convert_if_block): New function, code
3955 factored out from...
3956 (cond_move_process_if_block): ...here. Call the new function
3957 on the THEN and ELSE blocks.
3958 (merge_if_block): Do not copy global_live_at_end, merge_blocks
3959 already takes care of this.
3960
3961 2007-01-05 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
3962
3963 PR c/19978
3964 * tree.h (TREE_OVERFLOW_P): New.
3965 * c-typeck.c (parser_build_unary_op): Warn only if result
3966 overflowed and operands did not.
3967 (parser_build_binary_op): Likewise.
3968 (convert_for_assignment): Remove redundant overflow_warning.
3969 * c-common.c (overflow_warning): Don't check or set TREE_OVERFLOW.
3970
3971 2007-01-05 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
3972
3973 * c-typeck.c (store_init_value): Split over two lines to follow
3974 the GNU coding style.
3975
3976 2007-01-05 Benjamin Kosnik <bkoz@redhat.com>
3977
3978 * c-cppbuiltin.c (c_cpp_builtins): __GXX_EXPERIMENTAL_CPP0X__ to
3979 __GXX_EXPERIMENTAL_CXX0X__.
3980 * doc/cpp.texi: Same.
3981
3982 2007-01-05 Richard Guenther <rguenther@suse.de>
3983
3984 PR middle-end/27826
3985 * tree.c (get_narrower): Do not construct COMPONENT_REFs
3986 with mismatched types. Instead explicitly build a
3987 conversion NOP_EXPR.
3988
3989 2007-01-05 Ian Lance Taylor <iant@google.com>
3990
3991 * c-common.c (decl_with_nonnull_addr_p): New function.
3992 (c_common_truthvalue_conversion): Call it.
3993 * c-typeck.c (build_binary_op): Likewise.
3994 * c-common.h (decl_with_nonnull_addr_p): Declare.
3995
3996 2007-01-05 Jakub Jelinek <jakub@redhat.com>
3997
3998 PR c/30360
3999 * libgcc2.c (__divdc3): Compare c and d against 0.0 instead of
4000 denom against 0.0.
4001
4002 2007-01-05 Joel Brobecker <brobecker@adacore.com>
4003
4004 * doc/install.texi (Final install): Document the fact that
4005 the GNAT runtime should not be stripped.
4006
4007 2007-01-04 Jan Hubicka <jh@suse.cz>
4008
4009 * tree-inline.c (fold_marked_statements): Update operand caches
4010 and EH after folding
4011
4012 2007-01-04 Ian Lance Taylor <iant@google.com>
4013
4014 * c-common.c (check_function_nonnull): Whitespace fix.
4015
4016 2007-01-04 Jan Hubicka <jh@suse.cz>
4017
4018 * tree-optimize.c (execute_fixup_cfg): Correct previously mistakely
4019 comitted older version of patch.
4020 (pass_fixup_cfg): Add TODOs to verify flow and statements, dump
4021 function, celanup cfg and collect garbage.
4022
4023 2007-01-04 Mike Stump <mrs@apple.com>
4024
4025 * Makefile.in (mostlyclean): Don't remove libgcc anymore.
4026 (clean): Likewise.
4027
4028 2007-01-04 Eric Christopher <echristo@apple.com>
4029
4030 * libgcc2.c (__bswapsi2): Use SItype.
4031 (__bswapdi2): Use DItype.
4032 * libgcc2.h: Update for above.
4033
4034 2007-01-04 Paul Brook <paul@codesourcery.com>
4035
4036 * config/arm/arm.md (arm_mulsi3, thumb_mulsi3, mulsi3_compare0,
4037 mulsi_compare0_scratch, mulsi3addsi, mulsi3addsi_compare0,
4038 mulsi3addsi_compare0_scratch, mulsidi3adddi, mulsidi3,
4039 umulsidi3, umulsidi3adddi, smulsi3_highpart,
4040 umulsi3_highpart): Make conditional on !arm_arch6.
4041 (arm_mulsi3_v6, thumb_mulsi3_v6, mulsi3_compare0_v6,
4042 mulsi_compare0_scratch_v6, mulsi3addsi_v6, mulsi3addsi_compare0_v6,
4043 mulsi3addsi_compare0_scratch_v6, mulsidi3adddi_v6, mulsidi3_v6,
4044 umulsidi3_v6, umulsidi3adddi_v6, smulsi3_highpart_v6,
4045 umulsi3_highpart_v6): New insns.
4046
4047 2007-01-04 Roger Sayle <roger@eyesopen.com>
4048
4049 * fold-const.c (fold_convert): When casting an expression to void,
4050 fold_ignored_result may discover a GIMPLE_MODIFY_STMT which doesn't
4051 have a type. Instead of attempting to build a NOP_EXPR, return
4052 these "special" trees directly.
4053
4054 2007-01-04 Joseph Myers <joseph@codesourcery.com>
4055
4056 * config/rs6000/rs6000.c (rs6000_rtx_costs): Make adjustment for
4057 MULT inside MINUS as either argument. Use rs6000_cost->dmul -
4058 rs6000_cost->fp not 0 as adjustment for outer NEG.
4059
4060 2007-01-04 Jan Hubicka <jh@suse.cz>
4061
4062 * cgraph.c (cgraph_release_function_body): New function.
4063 (cgraph_remove_node): Use it.
4064 * cgraph.h (cgraph_release_function_body): Declare.
4065 * cgraphunit.c (cgraph_expand_function): Use it.
4066 * ipa.c (cgraph_remove_unreahchable_nodes): Use it.
4067 * tree-ssa.c (delete_tree_ssa): Allow to be called before aliasing
4068 is initialized and while compilation of other function is running.
4069 * tree-optimize.c (execute_free_cfg_annotations): Move code to clear
4070 statement CFG annotations from here to ...
4071 * tree-cfg.c (delete_tree_cfg_annotations): ... here.
4072
4073 2007-01-04 Zdenek Dvorak <dvorakz@suse.cz>
4074
4075 * cfgloop.h (enum li_flags): Make the constants powers of two.
4076
4077 2007-01-04 Jan Hubicka <jh@suse.cz>
4078
4079 * tree-inline.c (copy_bb): Insert new statements to statements_to_fold
4080 set.
4081 (fold_marked_statements): New function.
4082 (optimize_inline_calls, tree_function_versioning): Fold new statements.
4083 * tree-inline.h (copy_body_data): Add statements_to_fold.
4084
4085 2007-01-03 Daniel Jacobowitz <dan@codesourcery.com>
4086
4087 * config.gcc: Mention libgcc/config.host.
4088 * Makefile.in: Update comments mentioning libgcc.
4089 (LIBGCC, INSTALL_LIBGCC, GCC_PARTS, mklibgcc): Delete.
4090 (all.cross, start.encap, rest.encap, rest.cross): Update
4091 dependencies for libgcc move.
4092 (libgcc.mk, LIBGCC_DEPS, libgcov.a, libgcc.a, stmp-multilib)
4093 (clean-target, clean-target-libgcc): Delete.
4094 (srcdirify, GCC_EXTRA_PARTS): New macros.
4095 (libgcc-support, libgcc.mvars): New rules.
4096 (distclean): Remove mention of mklibgcc.
4097 (install): Don't reference INSTALL_LIBGCC.
4098 (install-common): Don't reference EXTRA_PARTS.
4099 (install-libgcc, install-multilib): Delete rules.
4100 * mklibgcc.in: Delete file.
4101 * doc/configfiles.texi: Don't mention mklibgcc.
4102
4103 * config/i386/t-darwin (SHLIB_VERPFX): Delete (moved to libgcc).
4104 * config/i386/t-darwin64 (SHLIB_VERPFX): Likewise.
4105 * config/rs6000/t-darwin (SHLIB_VERPFX): Likewise.
4106 * config/rs6000/t-ppccomm (TARGET_LIBGCC2_CFLAGS, SHLIB_MAPFILES)
4107 (mklibgcc, ldblspecs): Likewise.
4108
4109 * config/i386/t-nwld (libgcc.def, libc.def, libpcre.def)
4110 (posixpre.def): Use $(T).
4111 (SHLIB_EXT, SHLIB_NAME, SHLIB_SLIBDIR_QUAL, SHLIB_DEF, SHLIB_MAP)
4112 (SHLIB_SRC, SHLIB_INSTALL): Delete.
4113 (SHLIB_LINK): Make dummy.
4114 * config/t-slibgcc-darwin: Delete contents except for dummy SHLIB_LINK.
4115
4116 * config/frv/t-linux (EXTRA_MULTILIB_PARTS): Clear.
4117
4118 * config/alpha/t-crtfm: Use $(T) in rules for EXTRA_PARTS.
4119 * config/alpha/t-vms, config/alpha/t-vms64, config/fr30/t-fr30,
4120 config/i386/t-rtems-i386, config/ia64/t-ia64, config/rs6000/t-beos,
4121 config/rs6000/t-newas, config/sparc/t-elf: Likewise.
4122
4123 * configure.ac (all_outputs): Remove mklibgcc.
4124 * configure: Regenerated.
4125
4126 2007-01-03 Josh Conner <jconner@apple.com>
4127
4128 PR middle-end/29683
4129 * calls.c (compute_argument_addresses): Set stack and stack_slot
4130 for partial args, too.
4131 (store_one_arg): Use locate.size.constant for the size when
4132 generating a save_area.
4133
4134 2007-01-03 Robert Kennedy <jimbob@google.com>
4135
4136 * tree-cfg.c (tree_merge_blocks): Release SSA_NAME phi results
4137 whose definitions are deleted due to basic block merging.
4138
4139 2007-01-03 Paul Brook <paul@codesourcery.com>
4140
4141 PR target/16634
4142 * config/arm/arm.c (output_return_instruction): Pop PC in interrupt
4143 functions.
4144 (use_return_insn): Return 0 for Thumb interrupt functions.
4145 (print_multi_reg): Add rfe argument for IRQ returns.
4146 (arm_output_epilogue): Pop interrupt return address directly into PC.
4147 (arm_expand_prologue): Only adjust IRQ return address in Arm mode.
4148
4149 2007-01-03 Paul Brook <paul@codesourcery.com>
4150
4151 Merge from sourcerygxx-4_1.
4152 * config/arm/thumb2.md: New file.
4153 * config/arm/elf.h (JUMP_TABLES_IN_TEXT_SECTION): Return True for
4154 Thumb-2.
4155 * config/arm/coff.h (JUMP_TABLES_IN_TEXT_SECTION): Ditto.
4156 * config/arm/aout.h (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
4157 (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump tables.
4158 * config/arm/aof.h (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump
4159 tables.
4160 (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
4161 * config/arm/ieee754-df.S: Use macros for Thumb-2/Unified asm
4162 comptibility.
4163 * config/arm/ieee754-sf.S: Ditto.
4164 * config/arm/arm.c (thumb_base_register_rtx_p): Rename...
4165 (thumb1_base_register_rtx_p): ... to this.
4166 (thumb_index_register_rtx_p): Rename...
4167 (thumb1_index_register_rtx_p): ... to this.
4168 (thumb_output_function_prologue): Rename...
4169 (thumb1_output_function_prologue): ... to this.
4170 (thumb_legitimate_address_p): Rename...
4171 (thumb1_legitimate_address_p): ... to this.
4172 (thumb_rtx_costs): Rename...
4173 (thumb1_rtx_costs): ... to this.
4174 (thumb_compute_save_reg_mask): Rename...
4175 (thumb1_compute_save_reg_mask): ... to this.
4176 (thumb_final_prescan_insn): Rename...
4177 (thumb1_final_prescan_insn): ... to this.
4178 (thumb_expand_epilogue): Rename...
4179 (thumb1_expand_epilogue): ... to this.
4180 (arm_unwind_emit_stm): Rename...
4181 (arm_unwind_emit_sequence): ... to this.
4182 (thumb2_legitimate_index_p, thumb2_legitimate_address_p,
4183 thumb1_compute_save_reg_mask, arm_dwarf_handle_frame_unspec,
4184 thumb2_index_mul_operand, output_move_vfp, arm_shift_nmem,
4185 arm_save_coproc_regs, thumb_set_frame_pointer, arm_print_condition,
4186 thumb2_final_prescan_insn, thumb2_asm_output_opcode, arm_output_shift,
4187 thumb2_output_casesi): New functions.
4188 (TARGET_DWARF_HANDLE_FRAME_UNSPEC): Define.
4189 (FL_THUMB2, FL_NOTM, FL_DIV, FL_FOR_ARCH6T2, FL_FOR_ARCH7,
4190 FL_FOR_ARCH7A, FL_FOR_ARCH7R, FL_FOR_ARCH7M, ARM_LSL_NAME,
4191 THUMB2_WORK_REGS): Define.
4192 (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv, arm_condexec_count,
4193 arm_condexec_mask, arm_condexec_masklen)): New variables.
4194 (all_architectures): Add armv6t2, armv7, armv7a, armv7r and armv7m.
4195 (arm_override_options): Check new CPU capabilities.
4196 Set new architecture flag variables.
4197 (arm_isr_value): Handle v7m interrupt functions.
4198 (user_return_insn): Return 0 for v7m interrupt functions. Handle
4199 Thumb-2.
4200 (const_ok_for_arm): Handle Thumb-2 constants.
4201 (arm_gen_constant): Ditto. Use movw when available.
4202 (arm_function_ok_for_sibcall): Return false for v7m interrupt
4203 functions.
4204 (legitimize_pic_address, arm_call_tls_get_addr): Handle Thumb-2.
4205 (thumb_find_work_register, arm_load_pic_register,
4206 legitimize_tls_address, arm_address_cost, load_multiple_sequence,
4207 emit_ldm_seq, emit_stm_seq, arm_select_cc_mode, get_jump_table_size,
4208 print_multi_reg, output_mov_long_double_fpa_from_arm,
4209 output_mov_long_double_arm_from_fpa, output_mov_double_fpa_from_arm,
4210 output_mov_double_fpa_from_arm, output_move_double,
4211 arm_compute_save_reg_mask, arm_compute_save_reg0_reg12_mask,
4212 output_return_instruction, arm_output_function_prologue,
4213 arm_output_epilogue, arm_get_frame_offsets, arm_regno_class,
4214 arm_output_mi_thunk, thumb_set_return_address): Ditto.
4215 (arm_expand_prologue): Handle Thumb-2. Use arm_save_coproc_regs.
4216 (arm_coproc_mem_operand): Allow POST_INC/PRE_DEC.
4217 (arithmetic_instr, shift_op): Use arm_shift_nmem.
4218 (arm_print_operand): Use arm_print_condition. Handle '(', ')', '.',
4219 '!' and 'L'.
4220 (arm_final_prescan_insn): Use extract_constrain_insn_cached.
4221 (thumb_expand_prologue): Use thumb_set_frame_pointer.
4222 (arm_file_start): Output directive for unified syntax.
4223 (arm_unwind_emit_set): Handle stack alignment instruction.
4224 * config/arm/lib1funcs.asm: Remove default for __ARM_ARCH__.
4225 Add v6t2, v7, v7a, v7r and v7m.
4226 (RETLDM): Add Thumb-2 code.
4227 (do_it, shift1, do_push, do_pop, COND, THUMB_SYNTAX): New macros.
4228 * config/arm/arm.h (TARGET_CPU_CPP_BUILTINS): Define __thumb2__.
4229 (TARGET_THUMB1, TARGET_32BIT, TARGET_THUMB2, TARGET_DSP_MULTIPLY,
4230 TARGET_INT_SIMD, TARGET_UNIFIED_ASM, ARM_FT_STACKALIGN, IS_STACKALIGN,
4231 THUMB2_TRAMPOLINE_TEMPLATE, TRAMPOLINE_ADJUST_ADDRESS,
4232 ASM_OUTPUT_OPCODE, THUMB2_GO_IF_LEGITIMATE_ADDRESS,
4233 THUMB2_LEGITIMIZE_ADDRESS, CASE_VECTOR_PC_RELATIVE,
4234 CASE_VECTOR_SHORTEN_MODE, ADDR_VEC_ALIGN, ASM_OUTPUT_CASE_END,
4235 ADJUST_INSN_LENGTH): Define.
4236 (TARGET_REALLY_IWMMXT, TARGET_IWMMXT_ABI, CONDITIONAL_REGISTER_USAGE,
4237 STATIC_CHAIN_REGNUM, HARD_REGNO_NREGS, INDEX_REG_CLASS,
4238 BASE_REG_CLASS, MODE_BASE_REG_CLASS, SMALL_REGISTER_CLASSES,
4239 PREFERRED_RELOAD_CLASS, SECONDARY_OUTPUT_RELOAD_CLASS,
4240 SECONDARY_INPUT_RELOAD_CLASS, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P,
4241 TRAMPOLINE_SIZE, INITIALIZE_TRAMPOLINE, HAVE_PRE_INCREMENT,
4242 HAVE_POST_DECREMENT, HAVE_PRE_DECREMENT, HAVE_PRE_MODIFY_DISP,
4243 HAVE_POST_MODIFY_DISP, HAVE_PRE_MODIFY_REG, HAVE_POST_MODIFY_REG,
4244 REGNO_MODE_OK_FOR_BASE_P, LEGITIMATE_CONSTANT_P,
4245 REG_MODE_OK_FOR_BASE_P, REG_OK_FOR_INDEX_P, GO_IF_LEGITIMATE_ADDRESS,
4246 LEGITIMIZE_ADDRESS, THUMB2_LEGITIMIZE_ADDRESS,
4247 GO_IF_MODE_DEPENDENT_ADDRESS, MEMORY_MOVE_COST, BRANCH_COST,
4248 ASM_APP_OFF, ASM_OUTPUT_CASE_LABEL, ARM_DECLARE_FUNCTION_NAME,
4249 FINAL_PRESCAN_INSN, PRINT_OPERAND_PUNCT_VALID_P,
4250 PRINT_OPERAND_ADDRESS): Adjust for Thumb-2.
4251 (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv): New declarations.
4252 * config/arm/arm-cores.def: Add arm1156t2-s, cortex-a8, cortex-r4 and
4253 cortex-m3.
4254 * config/arm/arm-tune.md: Regenerate.
4255 * config/arm/arm-protos.h: Update prototypes.
4256 * config/arm/vfp.md: Enable patterns for Thumb-2.
4257 (arm_movsi_vfp): Add movw alternative. Use output_move_vfp.
4258 (arm_movdi_vfp, movsf_vfp, movdf_vfp): Use output_move_vfp.
4259 (thumb2_movsi_vfp, thumb2_movdi_vfp, thumb2_movsf_vfp,
4260 thumb2_movdf_vfp, thumb2_movsfcc_vfp, thumb2_movdfcc_vfp): New.
4261 * config/arm/libunwind.S: Add Thumb-2 code.
4262 * config/arm/constraints.md: Update include Thumb-2.
4263 * config/arm/ieee754-sf.S: Add Thumb-2/Unified asm support.
4264 * config/arm/ieee754-df.S: Ditto.
4265 * config/arm/bpabi.S: Ditto.
4266 * config/arm/t-arm (MD_INCLUDES): Add thumb2.md.
4267 * config/arm/predicates.md (low_register_operand,
4268 low_reg_or_int_operand, thumb_16bit_operator): New.
4269 (thumb_cmp_operand, thumb_cmpneg_operand): Rename...
4270 (thumb1_cmp_operand, thumb1_cmpneg_operand): ... to this.
4271 * config/arm/t-arm-elf: Add armv7 multilib.
4272 * config/arm/arm.md: Update patterns for Thumb-2 and Unified asm.
4273 Include thumb2.md.
4274 (UNSPEC_STACK_ALIGN, ce_count): New.
4275 (arm_incscc, arm_decscc, arm_umaxsi3, arm_uminsi3,
4276 arm_zero_extendsidi2, arm_zero_extendqidi2): New
4277 insns/expanders.
4278 * config/arm/fpa.md: Update patterns for Thumb-2 and Unified asm.
4279 (thumb2_movsf_fpa, thumb2_movdf_fpa, thumb2_movxf_fpa,
4280 thumb2_movsfcc_fpa, thumb2_movdfcc_fpa): New insns.
4281 * config/arm/cirrus.md: Update patterns for Thumb-2 and Unified asm.
4282 (cirrus_thumb2_movdi, cirrus_thumb2_movsi_insn,
4283 thumb2_cirrus_movsf_hard_insn, thumb2_cirrus_movdf_hard_insn): New
4284 insns.
4285 * doc/extend.texi: Document ARMv7-M interrupt functions.
4286 * doc/invoke.texi: Document Thumb-2 new cores+architectures.
4287
4288 2007-01-03 Jakub Jelinek <jakub@redhat.com>
4289
4290 * unwind-dw2.c (SIGNAL_FRAME_BIT, EXTENDED_CONTEXT_BIT): Define.
4291 (struct _Unwind_Context): Rename args_size to flags, remove
4292 signal_frame field, add a new args_size field and version field.
4293 (_Unwind_IsSignalFrame, _Unwind_SetSignalFrame,
4294 _Unwind_IsExtendedContext): New inline functions.
4295 (_Unwind_GetGR, _Unwind_SetGR, _Unwind_GetGRPtr, _Unwind_SetGRPtr):
4296 Assume by_value array is only present if _Unwind_IsExtendedContext.
4297 (_Unwind_GetIPInfo, execute_cfa_program, uw_frame_state_for): Use
4298 _Unwind_IsSignalFrame.
4299 (__frame_state_for): Initialize context.flags to EXTENDED_CONTEXT_BIT.
4300 (uw_update_context_1): Use _Unwind_SetSignalFrame.
4301 (uw_init_context_1): Initialize context->flags to
4302 EXTENDED_CONTEXT_BIT.
4303 * config/rs6000/linux-unwind.h (frob_update_context): Use
4304 _Unwind_SetSignalFrame.
4305
4306 2007-01-03 Andrew Pinski <andrew_pinski@playstation.sony.com>
4307
4308 PR middle-end/30353
4309 * gimplify.c (gimplify_modify_expr_complex_part): Move below
4310 tree_to_gimple_tuple. Call tree_to_gimple_tuple when we need
4311 the value.
4312
4313 2007-01-03 Kazu Hirata <kazu@codesourcery.com>
4314
4315 * config/i386/i386.h (NON_STACK_REG_P, REGNO_OK_FOR_SIREG_P,
4316 REGNO_OK_FOR_DIREG_P, REWRITE_ADDRESS, ASM_OPERAND_LETTER,
4317 RET, AT_SP): Remove.
4318 * config/i386/i386.md (*sse_prologue_save_insn): Use return
4319 instead of RET.
4320
4321 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
4322 * cfgbuild.c (find_basic_blocks): Likewise.
4323 * cfgrtl.c (rtl_create_basic_block): Likewise.
4324 * function.c (temp_slots_at_level): Likewise.
4325 * reg-stack.c (stack_regs_mentioned): Likewise.
4326 * regclass.c (allocate_reg_info): Likewise.
4327 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
4328 set_bb_for_stmt, move_block_to_fn): Likewise.
4329 * tree-complex.c (tree_lower_complex): Likewise.
4330 * vec.h (VEC_safe_grow_cleared): New.
4331
4332 * cgraphunit.c, tree-ssa-alias.c: Fix comment typos.
4333
4334 2007-01-03 Zdenek Dvorak <dvorakz@suse.cz>
4335
4336 * loop-unswitch.c (unswitch_loop): Pass probabilities to loopify.
4337 * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Pass probabilities
4338 to loop_version.
4339 * cfgloopmanip.c (scale_loop_frequencies): Export.
4340 (loopify): Scale the frequencies by prescribed coefficients.
4341 (set_zero_probability): New function.
4342 (duplicate_loop_to_header_edge): Improve updating of frequencies.
4343 (lv_adjust_loop_entry_edge, loop_version): Set probabilities
4344 and frequencies according to arguments.
4345 * tree-ssa-loop-manip.c (tree_unroll_loop): Set probabilities
4346 correctly.
4347 * cfg.c (scale_bbs_frequencies_int): Allow scaling the frequencies up.
4348 * modulo-sched.c (sms_schedule): Set probabilities for entering
4349 versioned loop correctly.
4350 * tree-vect-transform.c (vect_transform_loop): Ditto.
4351 * cfgloop.h (loopify, loop_version): Declaration changed.
4352 (scale_loop_frequencies): Declared.
4353
4354 2007-01-02 Jan Hubicka <jh@suse.cz>
4355
4356 * cgraph.c: Include tree-flow.h
4357 (cgraph_add_new-function): Handle IPA_SSA mode; execute
4358 early_local_passes.
4359 * cgraph.h (enum cgraph_state): Add CGRAPH_STATE_IPA_SSA.
4360 * tree-pass.h (pass_all_early_optimizations): Declare.
4361 * cgraphunit.c (cgraph_process_new_functions): Add IPA_SSA; execute
4362 early_local_passes.
4363 (cgraph_analyze_function): Do early_local_passes.
4364 * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
4365 Do not add referenced vars.
4366 * tree-optimize.c (gate_all_optimizations): Do not execute when not in
4367 SSA form.
4368 (gate_all_early_local_passes): New gate.
4369 (pass_early_local_passes): Use new gate.
4370 (execute_early_local_optimizations): New functions.
4371 (gate_all_early_optimizations): New gate.
4372 (pass_all_early_optimizations): New pass.
4373 (execute_free_datastructures): Free SSA only when initialized.
4374 (gate_init_datastructures): Init only when optimizing.
4375 (tree_lowering_passes): Do early local passes when called late.
4376 * tree-profile.c (do_tree_profiling): Don't profile functions added
4377 late.
4378 (do_early_tree_profiling, pass_early_tree_profile): Kill.
4379 * tree-cfg.c (update_modified_stmts): Do not update when operands are
4380 not active.
4381 * passes.c (init_optimizations_passes): Reorder so we go into SSA
4382 during early_local_passes.
4383 * Makefile.in (cgraph.o): Add dependency on tree-flow.h.
4384
4385
4386 2007-01-02 Carlos O'Donell <carlos@codesourcery.com>
4387
4388 * Makefile.in: Update copyright year.
4389
4390 2007-01-02 Carlos O'Donell <carlos@codesourcery.com>
4391
4392 * Makefile.in: Export GCC_EXEC_PREFIX before calling $(RUNTEST)
4393 in $(lang_checks) and check-consistency targets.
4394
4395 2007-01-02 Jan Hubicka <jh@suse.cz>
4396
4397 * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
4398 Do not add referenced vars.
4399 * tree-cfg.c (update_modified_stmts): Do not update when SSA operands
4400 are not active.
4401 * passes.c (init_optimization_passes): Put mudflap_2 after
4402 free_datastructures.
4403
4404 2007-01-02 Jan Hubicka <jh@suse.cz>
4405
4406 * tree-optimize (execute_fixup_cfg): Set after_inlining flag.
4407 Set NOTHROW flag on call statements proved to be nothrow.
4408 Update statement of local calls so new pure/const functions are
4409 updated. Update_ssa when in ssa form. Mark PHI nodes of nonlocal
4410 goto receivers.
4411 (tree_rest_of_compilation): Register hooks and initialize bitmap
4412 early. Do not set after_inlining flag.
4413
4414 2007-01-02 Steve Ellcey <sje@cup.hp.com>
4415
4416 * sbitmap.c (HOST_BITS_PER_LONG_LONG): Change to
4417 HOST_BITS_PER_LONGLONG
4418
4419 2007-01-02 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
4420
4421 PR c/19977
4422 * c-typeck.c (store_init_value): Don't emit pedantic overflow
4423 warning for non-static initializers.
4424
4425 2007-01-02 Steven Bosscher <steven@gcc.gnu.org>
4426
4427 * config/alpha/alpha.md, arm/arm.c, darwin.c, frv/frv.md,
4428 m32r/m32r.c, m32r/m32r.c, mn10300/mn10300.md, pa/pa.c,
4429 rs6000/rs6000.c, s390/s390.md, sh/sh.md, sparc/sparc.c:
4430 Always use set_unique_reg_note to add REG_EQUAL notes.
4431
4432 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
4433
4434 Revert:
4435 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
4436
4437 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
4438 * cfgbuild.c (find_basic_blocks): Likewise.
4439 * cfgrtl.c (rtl_create_basic_block): Likewise.
4440 * function.c (temp_slots_at_level): Likewise.
4441 * reg-stack.c (stack_regs_mentioned): Likewise.
4442 * regclass.c (allocate_reg_info): Likewise.
4443 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
4444 set_bb_for_stmt, move_block_to_fn): Likewise.
4445 * tree-complex.c (tree_lower_complex): Likewise.
4446 * vec.h (VEC_safe_grow_cleared): New.
4447
4448 2007-01-02 Ian Lance Taylor <iant@google.com>
4449
4450 * c-common.c (c_common_truthvalue_conversion): When warning about
4451 using an assignment as a truth value, set TREE_NO_WARNING.
4452
4453 2007-01-02 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
4454
4455 PR middle-end/7651
4456 * c.opt (Wold-style-declaration): New.
4457 * doc/invoke.texi (C-only Warning Options): New.
4458 (Wold-style-declaration): Document it.
4459 (Wextra): Enabled by -Wextra.
4460 * c-opts.c (c_common_post_options): Enabled by -Wextra.
4461 * c-decl.c (declspecs_add_scspec): Replace -Wextra with
4462 -Wold-style-declaration.
4463
4464 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
4465
4466 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
4467 * cfgbuild.c (find_basic_blocks): Likewise.
4468 * cfgrtl.c (rtl_create_basic_block): Likewise.
4469 * function.c (temp_slots_at_level): Likewise.
4470 * reg-stack.c (stack_regs_mentioned): Likewise.
4471 * regclass.c (allocate_reg_info): Likewise.
4472 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
4473 set_bb_for_stmt, move_block_to_fn): Likewise.
4474 * tree-complex.c (tree_lower_complex): Likewise.
4475 * vec.h (VEC_safe_grow_cleared): New.
4476
4477 2007-01-02 Douglas Gregor <doug.gregor@gmail.com>
4478
4479 * c-common.c (c_common_nodes_and_builtins): Since variants of
4480 void_type_node get built before it is given a name, we need to
4481 give those variants the name, too.
4482 (complete_array_type): We need to work with the canonical main
4483 type of the array, from which we will build the qualified version.
4484 * params.def (PARAM_VERIFY_CANONICAL_TYPES): New.
4485 * print-tree.c (print_node): Display canonical type information
4486 for each type.
4487 * stor-layout.c (layout_type): When we don't know the
4488 alignment of a type for which we're building an array, we end up
4489 guessing wrong, so make the type require structural equality.
4490 * tree.c (make_node_stat): When we build a new type, it is its
4491 own canonical type.
4492 (build_type_attribute_qual_variant): When building an attribute
4493 variant, its canonical type is the non-attribute variant. However,
4494 if the attributes are target-dependent and they differ, we need to
4495 use structural equality checks for this type.
4496 (build_qualified_type): A qualified type is not equivalent to its
4497 unqualified variant; set the canonical type appropriately.
4498 (build_distinct_type_copy): When building a distinct type from
4499 another type, the new type is its own canonical type.
4500 (build_variant_type_copy): When building a new type variant, we
4501 assume that it is equivalent to the original type.
4502 (build_pointer_type_for_mode): When building a pointer type, also
4503 build a canonical type pointer.
4504 (build_reference_type_for_mode): When building a reference type,
4505 also build a canonical type reference.
4506 (build_index_type): When we can't hash an index type (e.g.,
4507 because its maximum value is negative), the index type requires
4508 structural equality tests.
4509 (build_array_type): Build the canonical form of an array type.
4510 (build_function_type): Function types require structural equality,
4511 because they contain default arguments, attributes, etc.
4512 (build_method_type_directly): Ditto for method types.
4513 (build_offset_type): Build the canonical offset type.
4514 (build_complex_type): Build the canonical vector type.
4515 (make_vector_type): Build the canonical vector type.
4516 * tree.h (TYPE_CANONICAL): New.
4517 (TYPE_STRUCTURAL_EQUALITY_P): New.
4518 (SET_TYPE_STRUCTURAL_EQUALITY): New.
4519 (struct tree_type): Added "canonical" field.
4520 * params.h (VERIFY_CANONICAL_TYPES): New.
4521 * doc/c-tree.texi (TYPE_CANONICAL): Document.
4522 (TYPE_STRUCTURAL_EQUALITY_P): Document.
4523 (SET_TYPE_STRUCTURAL_EQUALITY): Document.
4524 * doc/invoke.texi (verify-canonical-types): Document --param
4525 parameter for verifying canonical types.
4526
4527 2007-01-02 Joseph Myers <joseph@codesourcery.com>
4528
4529 * config.gcc (powerpc-*-eabispe*, powerpc-*-eabisimaltivec*,
4530 powerpc-*-eabisim*, powerpc-*-eabialtivec*, powerpc-*-eabi*,
4531 powerpc-*-rtems*, powerpc-wrs-vxworks, powerpc-wrs-vxworksae,
4532 powerpcle-*-eabisim*, powerpcle-*-eabi*): Add rs6000/e500.h to
4533 tm_file.
4534 * config/rs6000/e500.h: New.
4535 * config/rs6000/eabi.h (TARGET_SPE_ABI, TARGET_SPE, TARGET_E500,
4536 TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE, TARGET_E500_DOUBLE):
4537 Remove.
4538 * config/rs6000/linuxspe.h (TARGET_SPE_ABI, TARGET_SPE,
4539 TARGET_E500, TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE,
4540 TARGET_E500_DOUBLE): Remove.
4541 * config/rs6000/vxworks.h (TARGET_SPE_ABI, TARGET_SPE,
4542 TARGET_E500, TARGET_ISEL, TARGET_FPRS): Remove.
4543 * config/rs6000/rs6000.h (CHECK_E500_OPTIONS): Define.
4544 * config/rs6000/rs6000.c (rs6000_override_options): Use
4545 CHECK_E500_OPTIONS.
4546
4547 2007-01-02 Joseph Myers <joseph@codesourcery.com>
4548
4549 * config/rs6000/rs6000.c (print_operand): Check (TARGET_SPE ||
4550 TARGET_E500_DOUBLE), not TARGET_E500, for %y.
4551 (rs6000_generate_compare, rs6000_emit_sCOND, output_cbranch,
4552 rs6000_emit_cmove): Don't check TARGET_E500.
4553 * config/rs6000/rs6000.md (bunordered, bordered, sunordered,
4554 sordered): Don't check TARGET_E500.
4555
4556 2007-01-01 Eric Christopher <echristo@apple.com>
4557
4558 * config/mips/mips.c (mips_regno_mode_ok_for_base_p): Use
4559 HARD_REGISTER_NUM_P.
4560
4561 2007-01-01 Roger Sayle <roger@eyesopen.com>
4562
4563 * fold-const.c (fold_binary) <EQ_EXPR>: Fold "(X^C1) eq/ne C2" into
4564 "X eq/ne (C1^C2)". Fold "(X^Z) eq/ne (Y^Z)" as "X eq/ne Y" when Z
4565 has no side-effects. Fold "(X^C1) eq/ne (Y^C2)" as "(X^(C1^C2))
4566 eq/ne Y".
4567
4568 2007-01-01 Mike Stump <mrs@apple.com>
4569
4570 * configure.ac: Remove support for building with Apple's gcc-3.1.
4571
4572 2007-01-02 Joseph Myers <joseph@codesourcery.com>
4573
4574 PR middle-end/30311
4575 * caller-save.c (add_stored_regs): Only handle SUBREGs if inner
4576 REG is a hard register. Do not modify REG before calling
4577 subreg_nregs.
4578 * rtlanal.c (subreg_get_info): Don't assert size of XMODE is a
4579 multiple of the size of YMODE for certain lowpart cases.
4580
4581 2007-01-01 Andrew Pinski <pinskia@gmail.com>
4582
4583 PR middle-end/30253
4584 * gimplify (voidify_wrapper_expr): Update for
4585 GIMPLE_MODIFY_STMT.
4586
4587 2007-01-01 Andreas Schwab <schwab@suse.de>
4588
4589 PR target/29166
4590 * config/ia64/ia64.c (ia64_compute_frame_size): Account space for
4591 save of BR0 in extra_spill_size instead of spill_size.
4592 (ia64_expand_prologue): Save BR0 outside of the gr/br/fr spill
4593 area.
4594 (ia64_expand_epilogue): Restore BR0 from its new location.
4595
4596 2007-01-01 Andrew Pinski <pinskia@gmail.com>
4597
4598 * gimplify.c (gimplify_init_constructor <case VECTOR_TYPE>):
4599 Use a temporary variable if the left hand side is not a gimple
4600 register.
4601
4602 2007-01-01 Andrew Pinski <pinskia@gmail.com>
4603
4604 * gimplify.c (gimplify_return_expr): Make the temporary variable
4605 for the return expression, a gimple register variable.
4606
4607 2007-01-01 Jan Hubicka <jh@suse.cz>
4608
4609 * emit-rtl.c (emit_copy_of_insn_after): Do not call copy_insn_1 for
4610 INSN_LIST.
4611
4612 2007-01-01 Mike Stump <mrs@apple.com>
4613
4614 * configure.ac (HAVE_GAS_LITERAL16): Add autoconf check for
4615 .literal16.
4616 * config/darwin.c (machopic_select_rtx_section): Use
4617 HAVE_GAS_LITERAL16.
4618 (darwin_mergeable_constant_section): Likewise.
4619 * configure: Regenerate.
4620 * config.in: Regenerate.
4621
4622 2007-01-01 Jan Hubicka <jh@suse.cz>
4623 Andrew Pinski <pinskia@gmail.com>
4624
4625 * cgraphunit.c (cgraph_optimize): Call cgraph_add_new_functions
4626 before starting IPA passes.