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