1 2007-02-11 Roger Sayle <roger@eyesopen.com>
2 Matt Thomas <matt@3am-software.com>
4 * simplify-rtx.c (simplify_relational_operation_1): Correct typo.
6 2007-02-11 Roger Sayle <roger@eyesopen.com>
8 * simplify-rtx.c (simplify_relational_operation_1): Optimize
9 comparisons of POPCOUNT against zero.
10 (simplify_const_relational_operation): Likewise.
12 2007-02-11 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
14 * doc/invoke.texi (Wextra): Delete outdated paragraph.
16 2007-02-11 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
18 * dwarf2out.c (root_type): Delete unused function.
20 2007-02-11 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
22 * genattrtab.c (contained_in_p): Delete unused function.
23 (write_expr_attr_cache): Likewise.
25 2007-02-11 Jan Hubicka <jh@suse.cz>
27 * ipa-inline.c (cgraph_edge_badness): Add "else" missing in the
30 2007-02-11 Steven Bosscher <steven@gcc.gnu.org>
32 * fwprop.c (try_fwprop_subst): Use set_unique_reg_note
33 to add the REG_EQ* note.
34 * see.c (see_merge_one_use_extension): Likewise.
35 * local-alloc.c (update_equiv_regs): Likewise. Also don't
36 turn REG_EQUAL notes into REG_EQUIV notes if the target
37 register may have more than one set.
38 * function.c (assign_parm_setup_reg): Use set_unique_reg_note.
39 * gcse.c (try_replace_reg): Likewise.
40 * alias.c (init_alias_analysis): Use find_reg_equal_equiv_note.
41 * calls.c (fixup_tail_calls): Likewise. Abort if there is
42 more than one REG_EQUIV note.
43 * reload1.c (gen_reload): Use set_unique_reg_note.
45 2007-02-11 Uros Bizjak <ubizjak@gmail.com>
47 * config/i386/i386.c (TARGET_VECTORIZE_BUILTIN_CONVERSION): Define.
48 (ix86_builtin_conversion): New function.
50 2007-02-06 Mark Mitchell <mark@codesourcery.com>
53 * tree.h (DECL_REPLACEABLE_P): New macro.
54 * except.c (set_nothrow_function_flags): Likewise.
56 2007-02-11 Tehila Meyzels <tehila@il.ibm.com>
57 Ira Rosen <irar@il.ibm.com>
58 Dorit Nuzman <dorit@il.ibm.com>
60 * doc/tm.texi (TARGET_VECTORIZE_BUILTIN_CONVERSION): New target hook.
61 * targhooks.c (default_builtin_vectorized_conversion): New.
62 * targhooks.h (default_builtin_vectorized_function): New declaration.
63 * target.h (struct vectorize): Add builtin_conversion field.
64 * tree-vectorizer.h (type_conversion_vec_info_type): New enum
65 stmt_vec_info_type value.
66 (vectorizable_conversion): New declaration.
67 * tree-vect-analyze.c (vect_analyze_operations): Add
68 vectorizable_conversion call.
69 * target-def.h (TARGET_VECTORIZE_BUILTIN_CONVERSION): New.
70 * tree-vect-transform.c (vectorizable_conversion): New function.
71 (vect_transform_stmt): Add case for type_conversion_vec_info_type.
72 * tree-vect-generic.c (expand_vector_operations_1): Consider correct
74 * config/rs6000/rs6000.c (rs6000_builtin_conversion): New.
75 (TARGET_VECTORIZE_BUILTIN_CONVERSION): Defined.
76 (rs6000_expand_builtin): Add handling a case of ALTIVEC_BUILTIN_VCFUX or
77 ALTIVEC_BUILTIN_VCFSX.
79 2007-02-10 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
82 * pa.md (movdf): For 64-bit target, fail if operand 1 is a non-zero
83 CONST_DOUBLE and operand 0 is a hard register.
84 (movdi): For 64-bit target, remove code to force CONST_DOUBLE to
85 memory. Fail if operand 1 is a non-zero CONST_INT and operand 0
86 is a hard floating-point register.
88 2007-02-10 Richard Henderson <rth@redhat.com>
89 Jakub Jelinek <jakub@redhat.com>
90 Alexandre Oliva <aoliva@redhat.com>
92 * Makefile.in (libgcc-support, libgcc.mvars): Add emutls.c.
93 * builtin-types.def (BT_WORD): Make unsigned.
94 (BT_FN_VOID_PTR_WORD_WORD_PTR): New.
95 * builtins.def (BUILT_IN_EMUTLS_GET_ADDRESS): New.
96 (BUILT_IN_EMUTLS_REGISTER_COMMON): New.
97 * c-decl.c (grokdeclarator): Don't error if !have_tls.
98 * c-parser.c (c_parser_omp_threadprivate): Likewise.
99 * dwarf2out.c (loc_descriptor_from_tree_1): Don't do anything for
101 * expr.c (emutls_var_address): New.
102 (expand_expr_real_1): Expand emulated tls.
103 (expand_expr_addr_expr_1): Likewise.
104 * libgcc-std.ver: Add __emutls_get_address, __emutls_register_common.
105 * output.h (emutls_finish): Declare.
106 * toplev.c (compile_file): Call it.
107 * tree-ssa-address.c (gen_addr_rtx): Check for const-ness of the
108 address before wrapping in CONST.
109 * varasm.c (emutls_htab, emutls_object_type): New.
110 (EMUTLS_VAR_PREFIX, EMUTLS_TMPL_PREFIX): New.
111 (get_emutls_object_name, get_emutls_object_type): New.
112 (get_emutls_init_templ_addr, emutls_decl): New.
113 (emutls_common_1, emutls_finish): New.
114 (assemble_variable): When emulating tls, swap decls; generate
115 constructor for the emutls objects.
116 (do_assemble_alias): When emulating tls, swap decl and target name.
117 (default_encode_section_info): Don't add SYMBOL_FLAG_TLS_SHIFT
119 * varpool.c (decide_is_variable_needed): Look at force_output.
120 Recurse for emulated tls.
121 (cgraph_varpool_remove_unreferenced_decls): Remove checks redundant
122 with decide_is_variable_needed.
123 * emutls.c: New file.
124 * config/sparc/sol2.h (ASM_DECLARE_OBJECT_NAME): Only emit
125 tls_object for real tls.
127 2007-02-10 Kaz Kojima <kkojima@gcc.gnu.org>
129 PR rtl-optimization/29599
130 * reload1.c (eliminate_regs_in_insn): Take the destination
131 mode into account when computing the offset.
133 2007-02-09 Stuart Hastings <stuart@apple.com>
134 Richard Henderson <rth@redhat.com>
136 * gcc/config/i386/i386.h (TARGET_KEEPS_VECTOR_ALIGNED_STACK): New.
137 * gcc/config/i386/darwin.h: (TARGET_KEEPS_VECTOR_ALIGNED_STACK): New.
138 * gcc/config/i386/i386.md (fixuns_trunc<mode>si2, fixuns_truncsfhi2,
139 fixuns_truncdfhi2): New.
140 (fix_truncsfdi_sse): Call ix86_expand_convert_sign_didf_sse.
141 (floatunsdidf2): Call ix86_expand_convert_uns_didf_sse.
142 (floatunssisf2): Add call to ix86_expand_convert_uns_sisf_sse.
143 (floatunssidf2): Allow nonimmediate source.
144 * gcc/config/i386/sse.md (movdi_to_sse): New.
145 (vec_concatv2di): Drop '*'.
146 * gcc/config/i386/i386-protos.h (ix86_expand_convert_uns_si_sse,
147 ix86_expand_convert_uns_didf_sse, ix86_expand_convert_uns_sidf_sse,
148 ix86_expand_convert_uns_sisf_sse, ix86_expand_convert_sign_didf_sse):
150 * gcc/config/i386/i386.c (ix86_expand_convert_uns_si_sse,
151 ix86_expand_convert_uns_didf_sse, ix86_expand_convert_uns_sidf_sse,
152 ix86_expand_convert_uns_sisf_sse, ix86_expand_convert_sign_didf_sse,
153 ix86_build_const_vector, ix86_expand_vector_init_one_nonzero): New.
154 (ix86_build_signbit_mask): Fix decl of v, refactor to call
155 ix86_build_const_vector.
156 (x86_emit_floatuns): Rewrite.
158 2007-02-10 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
160 * genautomata.c (longest_path_length): Delete unused function.
161 (struct state): Delete unused longest_path_length.
162 (UNDEFINED_LONGEST_PATH_LENGTH): Delete unused macro.
163 (get_free_state): Delete unused.
165 2007-02-09 Jan Hubicka <jh@suse.cz>
167 * params.def (PARAM_INLINE_UNIT_GROWTH): Set to 30.
168 * doc/invoke.texi (inline-unit-growth): Update default value.
170 * Makefile.in (passes.o, ipa-inline.o): Add dependencies.
171 * cgraphbuild.c (build_cgraph_edges): Compute frequencies.
172 (rebuild_cgraph_edges): Likewise.
173 * cgraph.c (cgraph_set_call_stmt): Add new argument frequency.
174 (dump_cgraph_node): Dump frequencies.
175 (cgraph_clone_edge): Add frequency scales.
176 (cgraph_clone_node): Add freuqnecy.
177 * cgraph.h (cgraph_edge): Add freuqnecy argument.
178 (CGRAPH_FREQ_BASE, CGRAPH_FREQ_MAX): New constants.
179 (cgraph_create_edge, cgraph_clone_edge, cgraph_clone_node): Update.
180 * tree-pass.h (TODO_rebuild_frequencies): New constant.
181 * cgraphunit.c (verify_cgraph_node): Verify frequencies.
182 (cgraph_copy_node_for_versioning): Update call of cgraph_clone_edge.
183 (save_inline_function_body): Likewise.
184 * ipa-inline.c: inluce rtl.h
185 (cgraph_clone_inlined_nods): Update call of cgraph_clone_node.
186 (cgraph_edge_badness): Use frequencies.
187 (cgraph_decide_recursive_inlining): Update clonning.
188 (cgraph_decide_inlining_of_small_function): Dump frequency.
189 * predict.c (estimate_bb_frequencies): Export.
190 * predict.h (estimate_bb_frequencies): Declare.
191 * tree-inline.c (copy_bb): Watch overflows.
192 (expand_call_inline): Update call of cgraph_create_edge.
193 (optimize_inline_calls): Use TODO flags to update frequnecies.
194 * passes.h: Include predict.h
195 (init_optimization_passes): Move profile ahead.
196 (execute_function_todo): Handle TODO_rebuild_frequencies.
198 2007-02-09 Roger Sayle <roger@eyesopen.com>
200 * config/alpha/alpha.c (emit_insxl): Force the first operand of
201 the insbl or inswl pattern into a register.
203 2007-02-09 Roger Sayle <roger@eyesopen.com>
205 * config/ia64/ia64.md (bswapdi2): New define_insn.
207 2007-02-09 Richard Henderson <rth@redhat.com>
209 * config/i386/constraints.md (Ym): New constraint.
210 * config/i386/i386.md (movsi_1): Change Y2 to Yi constraints.
211 (movdi_1_rex64): Split sse and xmm general register moves from
212 memory move alternatives. Use conditional register constraints.
213 (movsf_1, movdf_integer): Likewise.
214 (zero_extendsidi2_32, zero_extendsidi2_rex64): Likewise.
215 (movdf_integer_rex64): New.
216 (pushsf_rex64): Fix output constraints.
217 * config/i386/sse.md (sse2_loadld): Split rm alternative, use Yi.
218 (sse2_stored): Likewise.
219 (sse2_storeq_rex64): New.
220 * config/i386/i386.c (x86_inter_unit_moves): Enable for not
222 (ix86_secondary_memory_needed): Don't bypass TARGET_INTER_UNIT_MOVES
223 for optimize_size. Remove SF/DFmode hack.
225 2007-02-09 Dwarakanath Rajagopal <dwarak.rajagopal@amd.com>
227 * config/i386/driver-i386.c: Turn on -mtune=native for AMDFAM10.
230 2007-02-09 Nathan Sidwell <nathan@codesourcery.com>
231 Richard Sandiford <richard@codesourcery.com>
233 * config.gcc (m68010-*-netbsdelf*, m68k*-*-netbsdelf*)
234 (m68k*-*-openbsd*, m68k-*-linux*): Set default_cf_cpu.
235 (m68k-*-aout*, m68k-*-coff*, m68k-*-uclinux*, m68k-*-rtems*): Add
236 m68k/t-mlib to tmake_file.
237 (m68020-*-elf*, m68k-*-elf*): Likewise. Add t-m68kbare as well.
238 (m68k*-*-*): Use --with-arch to pick a default for --with-cpu.
239 (m680[012]0-*-*, m68k*-*-*): Add support for --with-arch.
240 Allow it to be cf or m68k. Set m68k_arch_family. If that
241 variable is not empty, add t-$m68k_arch_family to tmake_file.
242 Add t-mlibs to tmake_file.
243 * doc/install.texi: Document --with-arch=m68k and --with-arch=cf.
244 * config/m68k/t-cf: New file.
245 * config/m68k/t-m68k: Likewise.
246 * config/m68k/t-mlibs: Likewise.
247 * config/m68k/t-m68kbare (MULTILIB_OPTIONS, MULTILIB_DIRNAMES)
248 (MULTILIB_MATCHES, MULTILIB_EXCEPTIONS): Delete.
249 (M68K_MLIB_DIRNAMES, M68K_MLIB_OPTIONS): Define.
250 * config/m68k/t-m68kelf (MULTILIB_OPTIONS, MULTILIB_DIRNAMES)
251 (MULTILIB_MATCHES, MULTILIB_EXCEPTIONS, LIBGCC, INSTALL_LIBGCC):
253 * config/m68k/t-openbsd (MULTILIB_OPTIONS, LIBGCC): Delete.
254 (INSTALL_LIBGCC): Delete.
255 (M68K_MLIB_DIRNAMES, M68K_MLIB_OPTIONS): Define.
256 * config/m68k/t-rtems (MULTILIB_OPTIONS, MULTILIB_DIRNAMES)
257 (MULTILIB_MATCHES, MULTILIB_EXCEPTIONS): Delete.
258 (M68K_MLIB_CPU): Define.
259 * config/m68k/t-uclinux (MULTILIB_OPTIONS, MULTILIB_DIRNAMES)
260 (MULTILIB_MATCHES, MULTILIB_EXCEPTIONS): Delete.
261 (M68K_MLIB_CPU, M68K_MLIB_OPTIONS, M68K_MLIB_DIRNAMES): Define.
263 2007-02-09 Zdenek Dvorak <dvorakz@suse.cz>
264 Richard Guenther <rguenther@suse.de>
267 * fold-const.c (fold_comparison): Handle obfuscated comparisons
268 against INT_MIN/INT_MAX.
269 * tree-ssa-loop-ivcanon.c (remove_empty_loop): Print to dump
270 file if a loop is removed.
272 2007-02-09 Joseph Myers <joseph@codesourcery.com>
274 * calls.c (store_one_arg): Pass correct alignment to
275 emit_push_insn for non-BLKmode values.
276 * expr.c (emit_push_insn): If STRICT_ALIGNMENT, copy to an
277 unaligned stack slot via a suitably aligned slot.
279 2007-02-08 DJ Delorie <dj@redhat.com>
281 * config/m32c/m32c.c (m32c_unpend_compare): Add default to silence
283 (legal_subregs): Use unsigned char, make const.
284 (m32c_illegal_subreg_p): Use ARRAY_SIZE. Delete unused variables.
286 2007-02-08 Paul Brook <paul@codesourcery.com>
288 * config/arm/lib1funcs.asm (RETLDM): Pop directly into PC when no
289 special interworking needed.
291 2007-02-08 Harsha Jagasia <harsha.jagasia@amd.com>
293 * config/i386/xmmintrin.h: Make inclusion of emmintrin.h
294 conditional to __SSE2__.
295 (Entries below should have been added to first ChangeLog
296 entry for amdfam10 dated 2007-02-05)
297 * config/i386/emmintrin.h: Generate #error if __SSE2__ is not
299 * config/i386/pmmintrin.h: Generate #error if __SSE3__ is not
301 * config/i386/tmmintrin.h: Generate #error if __SSSE3__ is not
304 2007-02-08 DJ Delorie <dj@redhat.com>
306 * config/m32c/m32c-protos.h (m32c_illegal_subreg_p): New.
307 * config/m32c/m32c.c (legal_subregs): New.
308 (m32c_illegal_subreg_p): New.
309 * config/m32c/predicates.md (m32c_any_operand): Use it to reject
310 unsupported subregs of hard regs.
312 2007-02-08 Jan Hubicka <jh@suse.cz>
314 * tree-cfg.c (bsi_replace): Shortcut when replacing the statement with
315 the same one; always update histograms.
317 2007-02-08 Diego Novillo <dnovillo@redhat.com>
319 * passes.c (init_optimization_passes): Tidy comment.
321 2007-02-08 Roger Sayle <roger@eyesopen.com>
323 * simplify-rtx.c (simplify_unary_operation_1) <POPCOUNT>: We can
324 strip zero_extend, bswap and rotates from POCOUNT's argument.
325 <PARITY>: Likewise, we can strip not, bswap, sign_extend,
326 zero_extend and rotates from PARITY's argument.
327 <BSWAP>: A byte-swap followed by a byte-swap is an identity.
328 (simplify_const_unary_operation) <BSWAP>: Evaluate the byte-swap
329 of an integer constant at compile-time.
331 2007-02-08 Diego Novillo <dnovillo@redhat.com>
334 * tree-flow.h (struct var_ann_d): Remove field 'is_used'.
336 * tree-ssa-alias.c (compute_is_aliased): Remove. Update all
339 * tree-ssa-live.c (remove_unused_locals): Do not remove
340 TREE_ADDRESSABLE variables.
341 * tree-ssa-structalias.c (compute_points_to_sets): Tidy.
342 * tree-ssa-operands.c (add_virtual_operand): Remove argument
343 FOR_CLOBBER. Update all users.
344 If VAR has an associated alias set, add a virtual operand for
345 it if no alias is found to conflict with the memory reference.
347 2007-02-07 Jan Hubicka <jh@suse.cz>
348 Robert Kidd <rkidd@crhc.uiuc.edu>
350 * value-prof.c (visit_hist, free_hist): Return 1 instead of 0.
352 2007-02-07 Ian Lance Taylor <iant@google.com>
354 * lower-subreg.c (simple_move): Reject PARTIAL_INT modes.
356 2007-02-07 Roger Sayle <roger@eyesopen.com>
358 * config/rs6000/rs6000.md (ctz<mode>2, ffs<mode>2, popcount<mode>2,
359 parity<mode>2, smulsi3_highpart, abstf2_internal, allocate_stack,
360 tablejumpdi, movsi_to_cr_one): Remove constraints from
361 define_expand's match_operands.
363 2007-02-07 Roger Sayle <roger@eyesopen.com>
365 * global.c (compute_regsets): Move declatation of "i" inside of
366 #ifdef ELIMINABLE_REGS to avoid unused variable bootstrap failure.
368 2007-02-07 Jakub Jelinek <jakub@redhat.com>
371 * gimplify.c (gimplify_scan_omp_clauses): Remove special casing
372 of INDIRECT_REF <RESULT_DECL>.
374 * config/i386/i386.c (override_options): Set PTA_SSSE3 for core2.
376 2007-02-06 J"orn Rennecke <joern.rennecke@arc.com>
377 Kaz Kojima <kkojima@gcc.gnu.org>
380 * config/sh/sh.c (expand_cbranchdi4): Use scratch register
382 (sh_initialize_trampoline): Add parentheses to avoid a warning.
384 2007-02-06 Zdenek Dvorak <dvorakz@suse.cz>
386 * doc/loop.texi: Document possibility not to perform disambiguation
387 of loops with multiple latches.
388 * cfgloopmanip.c (alp_enum_p): Removed.
389 (add_loop): Handle subloops. Use get_loop_body_with_size.
390 (create_preheader): Do not allow ENTRY_BLOCK_PTR to be preheader.
391 * cfghooks.c (redirect_edge_and_branch_force): Set dominator for
392 the new forwarder block.
393 (make_forwarder_block): Only call new_bb_cbk if it is not NULL.
394 Handle the case latch is NULL.
395 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Avoid cfg modifications
396 when marking loop exits.
397 * ifcvt.c (if_convert): Ditto. Mark loop exits even if cfg cannot
399 * loop-init.c (loop_optimizer_init): Do not modify cfg. Call
400 disambiguate_loops_with_multiple_latches.
401 * tree-cfgcleanup.c (cleanup_tree_cfg_loop): Calculate dominators
402 before fix_loop_structure.
403 * cfgloop.c: Include pointer-set.h and output.h.
404 (canonicalize_loop_headers, HEADER_BLOCK, LATCH_EDGE,
405 update_latch_info, mfb_keep_just, mfb_keep_nonlatch): Removed.
406 (get_loop_latch_edges, find_subloop_latch_edge_by_profile,
407 find_subloop_latch_edge_by_ivs, find_subloop_latch_edge,
408 mfb_redirect_edges_in_set, form_subloop, merge_latch_edges,
409 disambiguate_multiple_latches, get_loop_body_with_size,
410 disambiguate_loops_with_multiple_latches): New functions.
411 (flow_loop_dump): Dump multiple latch edges.
412 (flow_loop_nodes_find): Handle loops with multiple latches.
413 (flow_loops_find): Ditto. Do not call canonicalize_loop_headers.
414 (glb_enum_p): Modified.
415 (get_loop_body): Use get_loop_body_with_size.
416 * cfgloop.h (LOOPS_HAVE_RECORDED_EXITS): New flag.
417 (AVOID_CFG_MODIFICATIONS): New constant.
418 (disambiguate_loops_with_multiple_latches, add_loop,
419 get_loop_body_with_size): Declare.
420 * Makefile.in (cfgloop.o): Add pointer-set.h and output.h.
422 2007-02-06 Seongbae Park <seongbae.park@gmail.com>
425 * global.c (compute_regsets): New function.
426 (global_alloc): Refactored ELIMINABLE_REGSET
427 and NO_GLOBAL_ALLOC_REGS computation out.
428 (rest_of_handle_global_alloc): Call compute_regsets()
429 for non-optimizing case.
431 2007-02-06 Richard Henderson <rth@redhat.com>
433 * config/i386/constraints.md (Y2): Rename from Y.
434 (Yi): New constraint.
435 * config/i386/i386.md (movsi_1, movdi_2, pushdf_nointeger,
436 pushdf_integer, movdf_nointeger, movdf_integer, zero_extendsidi2_32,
437 zero_extendsidi2_rex64, truncxfdf2_mixed): Change Y constraints to Y2.
438 (extendsfdf2_mixed, extendsfdf2_sse, truncdfsf_fast_mixed,
439 truncdfsf_fast_sse, truncdfsf_mixed, fix_truncdfdi_sse,
440 fix_truncdfsi_sse, floatsidf2_mixed, floatsidf2_sse,
441 floatdidf2_mixed, floatdidf2_sse, absnegdf2_mixed,
442 absnegdf2_sse, sse_setccdf, fop_df_comm_mixed, fop_df_comm_sse,
443 fop_df_1_mixed, fop_df_1_sse): Change Y constraints to x.
444 * config/i386/mmx.md (mov<MMXMODEI>_internal_rex64,
445 mov<MMXMODEI>_internal, movv2sf_internal_rex64, movv2sf_internal,
446 vec_extractv2si_1): Change Y constraints to Y2.
447 * config/i386/sse.md (vec_setv4sf_0, vec_concatv2df, vec_dupv4si,
448 vec_dupv2di, sse2_concatv2si, vec_concatv4si_1, vec_concatv2di):
449 Change Y constraints to Y2.
450 (sse2_loadld): Change Y constraints to x.
452 2007-02-06 Roger Sayle <roger@eyesopen.com>
454 * config/rs6000/rs6000.md (popcount<mode>2): Rewrite.
455 (parity<mode>2): New define_expand using rs6000_emit_parity.
456 * config/rs6000/rs6000.c (rs6000_emit_popcount,
457 rs6000_emit_parity): New functions.
458 * config/rs6000/rs6000-protos.h (rs6000_emit_popcount,
459 rs6000_emit_parity): Prototype here.
461 2007-02-06 Ian Lance Taylor <iant@google.com>
463 * lower-subreg.c (simple_move_operand): Reject CONST.
464 (resolve_clobber): Call validate_change rather than directly
465 assigning to XEXP (pat, 0).
467 2006-02-06 Paolo Bonzini <bonzini@gnu.org>
469 * Makefile.in (tree-ssa-loop-ivopts.o): Add pointer-set.h dependency.
470 (tree-ssa-reassoc.o): Add pointer-set.h dependency.
471 (tree-cfg.o): Remove hashtab.h dependency.
473 * tree-ssa-loop-ivopts.c: Include pointer-set.h.
474 (struct ivopts_data): Change niters to pointer_map_t.
475 (struct nfe_cache_elt, nfe_hash, nfe_eq): Delete.
476 (niter_for_exit): Create pointer_map on demand. Change for
478 (tree_ssa_iv_optimize_init): Initialize data->niters to NULL.
479 (free_loop_data): Destroy data->niters if created and reset field.
480 (tree_ssa_iv_optimize_finalize): Don't delete data->niters here.
481 (tree_ssa_iv_optimize_loop): Check for presence of stale data.
483 * tree-ssa-reassoc.c: Include pointer-set.h.
484 (bb_rank): Change to long *.
485 (operand_rank): Change to pointer_map_t.
486 (find_operand_rank): Return long, -1 if not found. Declare as inline.
487 (insert_operand_rank): Accept long.
488 (operand_entry_hash, operand_entry_eq): Remove.
489 (get_rank): Return long. Adjust for changes above.
490 (init_reassoc): Change rank type to long. Adjust creation of bb_rank
492 (fini_reassoc): Delete operand_rank with pointer_map_destroy.
494 * tree-ssa-structalias.c (vi_for_tree): Change to pointer_map.
495 (struct tree_vi, tree_vi_t, tree_vi_hash, tree_vi_eq): Delete.
496 (insert_vi_for_tree): Rewrite for pointer_map API. Assert argument
498 (lookup_vi_for_tree): Rewrite for pointer_map API. Return varinfo_t
499 directly since it cannot be NULL.
500 (get_vi_for_tree): Rewrite for pointer_map API.
501 (find_what_p_points_to): Adjust for change to lookup_vi_for_tree.
502 (init_alias_vars): Create vi_for_tree as pointer_map.
503 (delete_points_to_sets): Delete vi_for_tree using pointer_map_destroy.
505 * tree-cfg.c: Don't include hashtab.h.
506 (edge_to_cases): Declare as pointer_map.
507 (struct edge_to_cases_elt, edge_to_cases_hash, edge_to_cases_eq):
509 (edge_to_cases_cleanup): Rewrite as pointer_map_traverse callback.
510 (start_recording_case_labels): Create edge_to_cases as pointer_map.
511 (end_recoding_case_labels): Cleanup edge_to_cases manually before
513 (record_switch_edge): Delete.
514 (get_cases_for_edge): Adjust for pointer_map API, inline
515 record_switch_edge (rewritten for new API), remove goto.
517 2006-02-06 Paolo Bonzini <bonzini@gnu.org>
519 * Makefile.in (tree-nested.o): Add pointer-set.h dependency.
520 * tree-nested.c: Include pointer-set.h.
521 (var_map_elt, var_map_eq, var_map_hash): Delete.
522 (struct nesting_info): Remove GTY marker. Change the two htab_t's
524 (nesting_info_bitmap_obstack): New.
525 (lookup_field_for_decl): Adjust for pointer_map API.
526 (lookup_tramp_for_decl): Adjust for pointer_map API.
527 (get_nonlocal_debug_decl): Adjust for pointer_map API.
528 (get_local_debug_decl): Adjust for pointer_map API.
529 (convert_nl_goto_reference): Adjust for pointer_map API.
530 (convert_nl_goto_receiver): Adjust for pointer_map API.
531 (create_nesting_tree): Create outside GGC space. Create bitmap on
532 the new obstack. Create field_map and var_map as pointer_maps.
533 (free_nesting_tree): Adjust for changes to create_nesting_tree.
535 (lower_nested_functions): Move root here, no need to NULL it.
536 Initialize and release the obstack.
538 2007-02-06 Paolo Bonzini <bonzini@gnu.org>
540 * tree.c (tree_int_map_hash, tree_int_map_eq, tree_int_map_marked_p):
541 Remove prototypes and make them non-static.
542 (struct tree_int_map): Remove.
543 * tree.h (struct tree_int_map): Move here, turning TO into an
545 (tree_int_map_hash, tree_int_map_eq, tree_int_map_marked_p): Declare.
547 * tree.h (TREE_COMPLEXITY): Remove.
548 (struct tree_exp): Remove complexity field.
549 * tree.c (build1_stat): Don't set it.
551 2007-02-06 Dorit Nuzman <dorit@il.ibm.com>
552 Victor Kaplansky <victork@il.ibm.com>
554 * tree-vectorizer.c (vect_is_simple_use): Support induction.
555 (vect_is_simple_reduction): Support reduction with induction as
557 (vect_is_simple_iv_evolution): Fix formatting.
558 * tree-vect-analyze.c (vect_mark_stmts_to_be_vectorized): Fix
559 formatting. Don't mark induction phis for vectorization.
560 (vect_analyze_scalar_cycles): Analyze all inductions, then reductions.
561 * tree-vect-transform.c (get_initial_def_for_induction): New function.
562 (vect_get_vec_def_for_operand): Support induction.
563 (vect_get_vec_def_for_stmt_copy): Fix formatting and add check for
565 (vectorizable_reduction): Support reduction with induction as one of
567 (vectorizable_type_demotion): Use def-type of stmt argument rather
570 * tree-ssa-loop.c (gate_scev_const_prop): Return the value of
571 flag_tree_scev_cprop.
572 * common.opt (tree-scev-cprop): New flag.
574 * tree-vect-transform.c (vect_create_destination_var): Use 'kind' in
575 call to vect_get_new_vect_var.
577 2007-02-06 Ira Rosen <irar@il.ibm.com>
579 * tree-vect-patterns.c (vect_recog_widen_mult_pattern): Check that
581 (vect_pattern_recog_1): Likewise.
583 2007-02-05 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
585 * fold-const.c (negate_expr_p): Handle CONJ_EXPR.
586 (fold_negate_expr): Likewise.
588 2007-02-05 Alexandre Oliva <aoliva@redhat.com>
591 * dwarf2out.c (modified_type_die): Follow DECL_ORIGINAL_TYPE
592 even if cv-qualification is the same.
594 2007-02-05 Geoffrey Keating <geoffk@apple.com>
596 * config/rs6000/darwin-tramp.asm (__trampoline_setup): Call
597 __enable_execute_stack on completion.
599 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
601 * config/i386/athlon.md (athlon_fldxf_k8, athlon_fld_k8,
602 athlon_fstxf_k8, athlon_fst_k8, athlon_fist, athlon_fmov,
603 athlon_fadd_load, athlon_fadd_load_k8, athlon_fadd, athlon_fmul,
604 athlon_fmul_load, athlon_fmul_load_k8, athlon_fsgn,
605 athlon_fdiv_load, athlon_fdiv_load_k8, athlon_fdiv_k8,
606 athlon_fpspc_load, athlon_fpspc, athlon_fcmov_load,
607 athlon_fcmov_load_k8, athlon_fcmov_k8, athlon_fcomi_load_k8,
608 athlon_fcomi, athlon_fcom_load_k8, athlon_fcom): Added amdfam10.
610 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
612 * config/i386/i386.md (x86_sahf_1, cmpfp_i_mixed, cmpfp_i_sse,
613 cmpfp_i_i387, cmpfp_iu_mixed, cmpfp_iu_sse, cmpfp_iu_387,
614 swapsi, swaphi_1, swapqi_1, swapdi_rex64, fix_truncsfdi_sse,
615 fix_truncdfdi_sse, fix_truncsfsi_sse, fix_truncdfsi_sse,
616 x86_fldcw_1, floatsisf2_mixed, floatsisf2_sse, floatdisf2_mixed,
617 floatdisf2_sse, floatsidf2_mixed, floatsidf2_sse,
618 floatdidf2_mixed, floatdidf2_sse, muldi3_1_rex64, mulsi3_1,
619 mulsi3_1_zext, mulhi3_1, mulqi3_1, umulqihi3_1, mulqihi3_insn,
620 umulditi3_insn, umulsidi3_insn, mulditi3_insn, mulsidi3_insn,
621 umuldi3_highpart_rex64, umulsi3_highpart_insn,
622 umulsi3_highpart_zext, smuldi3_highpart_rex64,
623 smulsi3_highpart_insn, smulsi3_highpart_zext, x86_64_shld,
624 x86_shld_1, x86_64_shrd, sqrtsf2_mixed, sqrtsf2_sse,
625 sqrtsf2_i387, sqrtdf2_mixed, sqrtdf2_sse, sqrtdf2_i387,
626 sqrtextendsfdf2_i387, sqrtxf2, sqrtextendsfxf2_i387,
627 sqrtextenddfxf2_i387): Added amdfam10_decode.
629 * config/i386/athlon.md (athlon_idirect_amdfam10,
630 athlon_ivector_amdfam10, athlon_idirect_load_amdfam10,
631 athlon_ivector_load_amdfam10, athlon_idirect_both_amdfam10,
632 athlon_ivector_both_amdfam10, athlon_idirect_store_amdfam10,
633 athlon_ivector_store_amdfam10): New define_insn_reservation.
634 (athlon_idirect_loadmov, athlon_idirect_movstore): Added
637 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
639 * config/i386/athlon.md (athlon_call_amdfam10,
640 athlon_pop_amdfam10, athlon_lea_amdfam10): New
641 define_insn_reservation.
642 (athlon_branch, athlon_push, athlon_leave_k8, athlon_imul_k8,
643 athlon_imul_k8_DI, athlon_imul_mem_k8, athlon_imul_mem_k8_DI,
644 athlon_idiv, athlon_idiv_mem, athlon_str): Added amdfam10.
646 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
648 * config/i386/athlon.md (athlon_sseld_amdfam10,
649 athlon_mmxld_amdfam10, athlon_ssest_amdfam10,
650 athlon_mmxssest_short_amdfam10): New define_insn_reservation.
652 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
654 * config/i386/athlon.md (athlon_sseins_amdfam10): New
655 define_insn_reservation.
656 * config/i386/i386.md (sseins): Added sseins to define_attr type
657 and define_attr unit.
658 * config/i386/sse.md: Set type attribute to sseins for insertq
661 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
663 * config/i386/athlon.md (sselog_load_amdfam10, sselog_amdfam10,
664 ssecmpvector_load_amdfam10, ssecmpvector_amdfam10,
665 ssecomi_load_amdfam10, ssecomi_amdfam10,
666 sseaddvector_load_amdfam10, sseaddvector_amdfam10): New
667 define_insn_reservation.
668 (ssecmp_load_k8, ssecmp, sseadd_load_k8, seadd): Added amdfam10.
670 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
672 * config/i386/athlon.md (cvtss2sd_load_amdfam10,
673 cvtss2sd_amdfam10, cvtps2pd_load_amdfam10, cvtps2pd_amdfam10,
674 cvtsi2sd_load_amdfam10, cvtsi2ss_load_amdfam10,
675 cvtsi2sd_amdfam10, cvtsi2ss_amdfam10, cvtsd2ss_load_amdfam10,
676 cvtsd2ss_amdfam10, cvtpd2ps_load_amdfam10, cvtpd2ps_amdfam10,
677 cvtsX2si_load_amdfam10, cvtsX2si_amdfam10): New
678 define_insn_reservation.
680 * config/i386/sse.md (cvtsi2ss, cvtsi2ssq, cvtss2si,
681 cvtss2siq, cvttss2si, cvttss2siq, cvtsi2sd, cvtsi2sdq,
682 cvtsd2si, cvtsd2siq, cvttsd2si, cvttsd2siq,
683 cvtpd2dq, cvttpd2dq, cvtsd2ss, cvtss2sd,
684 cvtpd2ps, cvtps2pd): Added amdfam10_decode attribute.
686 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
688 * config/i386/athlon.md (athlon_ssedivvector_amdfam10,
689 athlon_ssedivvector_load_amdfam10, athlon_ssemulvector_amdfam10,
690 athlon_ssemulvector_load_amdfam10): New define_insn_reservation.
691 (athlon_ssediv, athlon_ssediv_load_k8, athlon_ssemul,
692 athlon_ssemul_load_k8): Added amdfam10.
694 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
696 * config/i386/i386.h (TARGET_SSE_UNALIGNED_MOVE_OPTIMAL): New macro.
697 (x86_sse_unaligned_move_optimal): New variable.
699 * config/i386/i386.c (x86_sse_unaligned_move_optimal): Enable for
701 (ix86_expand_vector_move_misalign): Add code to generate movupd/movups
702 for unaligned vector SSE double/single precision loads for AMDFAM10.
704 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
706 * config/i386/i386.h (TARGET_AMDFAM10): New macro.
707 (TARGET_CPU_CPP_BUILTINS): Add code for amdfam10.
708 Define TARGET_CPU_DEFAULT_amdfam10.
709 (TARGET_CPU_DEFAULT_NAMES): Add amdfam10.
710 (processor_type): Add PROCESSOR_AMDFAM10.
712 * config/i386/i386.md: Add amdfam10 as a new cpu attribute to match
713 processor_type in config/i386/i386.h.
714 Enable imul peepholes for TARGET_AMDFAM10.
716 * config.gcc: Add support for --with-cpu option for amdfam10.
718 * config/i386/i386.c (amdfam10_cost): New variable.
719 (m_AMDFAM10): New macro.
720 (m_ATHLON_K8_AMDFAM10): New macro.
721 (x86_use_leave, x86_push_memory, x86_movx, x86_unroll_strlen,
722 x86_cmove, x86_3dnow_a, x86_deep_branch, x86_use_simode_fiop,
723 x86_promote_QImode, x86_integer_DFmode_moves,
724 x86_partial_reg_dependency, x86_memory_mismatch_stall,
725 x86_accumulate_outgoing_args, x86_arch_always_fancy_math_387,
726 x86_sse_partial_reg_dependency, x86_sse_typeless_stores,
727 x86_use_ffreep, x86_use_incdec, x86_four_jump_limit,
728 x86_schedule, x86_use_bt, x86_cmpxchg16b, x86_pad_returns):
729 Enable/disable for amdfam10.
730 (override_options): Add amdfam10_cost to processor_target_table.
731 Set up PROCESSOR_AMDFAM10 for amdfam10 entry in
732 processor_alias_table.
733 (ix86_issue_rate): Add PROCESSOR_AMDFAM10.
734 (ix86_adjust_cost): Add code for amdfam10.
736 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
738 * config/i386/i386.opt: Add new Advanced Bit Manipulation (-mabm)
739 instruction set feature flag. Add new (-mpopcnt) flag for popcnt
740 instruction. Add new SSE4A (-msse4a) instruction set feature flag.
741 * config/i386/i386.h: Add builtin definition for SSE4A.
742 * config/i386/i386.md: Add support for ABM instructions
744 * config/i386/sse.md: Add support for SSE4A instructions
745 (movntss, movntsd, extrq, insertq).
746 * config/i386/i386.c: Add support for ABM and SSE4A builtins.
747 Add -march=amdfam10 flag.
748 * config/i386/ammintrin.h: Add support for SSE4A intrinsics.
749 * doc/invoke.texi: Add documentation on flags for sse4a, abm, popcnt
751 * doc/extend.texi: Add documentation for SSE4A builtins.
753 2007-02-05 Bob Wilson <bob.wilson@acm.org>
755 * config/xtensa/xtensa.c (constantpool_mem_p): Skip over SUBREGs.
757 2007-02-05 Richard Guenther <rguenther@suse.de>
759 * tree-vectorizer.h (vectorizable_function): Add argument type
760 argument, change return type.
761 * tree-vect-patterns.c (vect_recog_pow_pattern): Adjust caller.
762 * tree-vect-transform.c (vectorizable_function): Handle extra
763 argument, return vectorized function decl.
764 (build_vectorized_function_call): Remove.
765 (vectorizable_call): Handle calls with result and argument types
766 differing. Handle loop vectorization factor correctly.
767 * targhooks.c (default_builtin_vectorized_function): Adjust for
769 * targhooks.h (default_builtin_vectorized_function): Likewise.
770 * target.h (builtin_vectorized_function): Add argument type
772 * config/i386/i386.c (ix86_builtin_vectorized_function): Handle
773 extra argument, allow vectorizing of lrintf.
774 * doc/tm.texi (TARGET_VECTORIZE_BUILTIN_VECTORIZED_FUNCTION): Adjust
775 documentation of target hook.
777 2007-02-05 Hans-Peter Nilsson <hp@axis.com>
780 * config/cris/cris.md ("*andsi_movu", "*andsi_clear", "*andhi_movu")
781 ("*andhi_clear", andu (casesi+45)): For size-changed operand where
782 memory is allowed, require !side_effects_p, not just !MEM_VOLATILE_P.
784 2007-02-05 Roger Sayle <roger@eyesopen.com>
786 * fold-const.c (fold_unary) <REAL_PART>: Test for availability of
787 BUILT_IN_COS before simplifying REAL_PART(CEXPI)) to COS.
788 <IMAG_PART>: Likewise, check for availability of BUILT_IN_SIN.
789 * builtins.c (fold_builtin_sincos): Check for TARGET_C99_FUNCTIONS
790 before canonicalizing sincos to cexpi.
791 (fold_builtin_cexp): Likewise, for canonicalizing cexp to cexpi.
793 2007-02-05 Roger Sayle <roger@eyesopen.com>
795 * config/alpha/alpha.c (alpha_add_builtins): New Helper function.
796 Set TREE_READONLY and TREE_NOTHROW directly, not via attributes.
797 (alpha_init_builtins): Use alpha_add_builtins to process tables.
799 2007-02-05 Roger Sayle <roger@eyesopen.com>
801 * mips-tfile.c (initialize_init_file): Correct endianness test.
803 2007-02-05 Kazu Hirata <kazu@codesourcery.com>
805 * config/m68k/m68k.md (pushdi-1, pushdi, movsi+1): Don't use
808 2007-02-05 Richard Sandiford <richard@codesourcery.com>
810 * dwarf2out.c (dwarf2out_frame_debug_expr): Record the register
811 saves in a PARALLEL before the register assignments.
813 2007-02-05 Richard Sandiford <richard@codesourcery.com>
815 * doc/tm.texi (DWARF_ALT_FRAME_RETURN_COLUMN): Do not require
816 DWARF_FRAME_RETURN_COLUMN to be a general register.
817 * dwarf2out.c (init_return_column_size): New function, split from...
818 (expand_builtin_init_dwarf_reg_sizes): ...here. Allow both
819 DWARF_FRAME_RETURN_COLUMN and DWARF_ALT_FRAME_RETURN_COLUMN
820 to be nongeneral registers.
821 * config/m68k/m68k.h (DWARF_FRAME_REGNUM): Only map FP and
823 (DWARF_FRAME_REGISTERS, DWARF_FRAME_RETURN_COLUMN): Define.
824 (DWARF_ALT_FRAME_RETURN_COLUMN): Define.
826 2007-02-04 Zdenek Dvorak <dvorakz@suse.cz>
828 * cfgcleanup.c (try_optimize_cfg): Avoid removing ENTRY_BLOCK_PTR.
830 2007-02-04 Zdenek Dvorak <dvorakz@suse.cz>
832 * cfgloopmanip.c (loop_delete_branch_edge): Removed.
833 (remove_path): Use can_remove_branch_p and remove_branch instead
834 of loop_delete_branch_edge.
835 * tree-ssa-loop-manip.c (scale_dominated_blocks_in_loop): New function.
836 (tree_transform_and_unroll_loop): Remove dead branches immediately.
837 Update profile using scale_dominated_blocks_in_loop.
838 * cfghooks.c (can_remove_branch_p, remove_branch): New functions.
839 * cfghooks.h (struct cfg_hooks): Add can_remove_branch_p.
840 (can_remove_branch_p, remove_branch): Declare.
841 * tree-cfg.c (tree_can_remove_branch_p): New function.
842 (tree_cfg_hooks): Add tree_can_remove_branch_p.
843 * cfgrtl.c (rtl_can_remove_branch_p): New function.
844 (rtl_cfg_hooks, cfg_layout_rtl_cfg_hook): Add rtl_can_remove_branch_p.
846 2007-02-05 Jan Hubicka <jh@suse.cz>
849 * ipa-inline.c (cgraph_clone_inlined_nodes): When there are unanalyzed
850 nodes in cgraph, don't remove offline copy of the function.
852 2007-02-04 Jan Hubicka <jh@suse.cz>
854 * tree-sra.c (sra_walk_expr): Add linebreaks. BITFIELD_REFs into
855 vectors might cause maybe_lookup_element_for_expr to be called
856 on non-sra-candidate.
858 2007-02-04 Kazu Hirata <kazu@codesourcery.com>
860 * config/bfin/bfin-modes.def, config/bfin/bfin.c,
861 config/bfin/bfin.md, config/bfin/predicates.md: Follow
862 spelling conventions.
864 2007-02-04 Richard Guenther <rguenther@suse.de>
867 * fold-const.c (try_move_mult_to_index): Make sure to not
868 overflow one dimension of a multi-dimensional array access.
870 2007-02-04 Jan Hubicka <jh@suse.cz>
872 * passes.c (init_optimization_passes): Reindent.
874 2007-02-04 Jan Hubicka <jh@suse.cz>
875 Eric Botcazou <ebotcazou@adacore.com>
877 * tree-optimize.c (has_abnormal_outgoing_edge_p): Move to...
878 (execute_fixup_cfg): Break out the abnormal goto code.
879 * tree-inline.c (has_abnormal_outgoing_edge_p): ...here.
880 (make_nonlocal_label_edges): Move here from execute_fixup_cfg.
881 (optimize_inline_calls): Call make_nonlocal_label_edges.
883 2007-02-04 Jan Hubicka <jh@suse.cz>
885 * tree-ssa-copyrename.c (copy_rename_partition_coalesce): Return
886 true when something was changed.
887 (rename_ssa_copies): When something was changed, do
888 TODO_remove_unused_locals.
889 * tree-ssa-forwprop.c (tree_ssa_forward_propagate_single_use_value):
890 add TODO_remove_unused_locals when instruction was removed.
892 2007-02-04 Jan Hubicka <jh@suse.cz>
894 * ipa-inline.c (try_inline): Improve debug output; work on already
896 (cgraph_decide_inlining_incrementally): Indent; improve debug output;
897 call try_inline for already inlined edges too when flattening;
898 inline also functions that make callee growth but overall unit size
901 2007-02-04 Kazu Hirata <kazu@codesourcery.com>
903 * config/m32c/bitops.md, config/m32c/jump.md,
904 config/m32c/m32c.c, config/m32c/m32c.h, config/m32r/m32r.c,
905 config/m32r/m32r.h, config/m32r/m32r.md,
906 config/m32r/predicates.md, config/m68hc11/larith.asm,
907 config/m68hc11/m68hc11.c, config/m68hc11/m68hc11.h,
908 config/m68k/m68k.h, config/mcore/mcore.md, config/mips/4k.md,
909 config/mips/mips-protos.h, config/mips/mips.c,
910 config/mips/mips.h, config/mips/mips.md, config/mips/mips16.S,
911 config/mn10300/mn10300.h, config/mn10300/predicates.md,
912 config/mt/mt.c, config/mt/mt.h, config/mt/mt.md: Follow
913 spelling conventions.
915 * config/v850/v850.c, config/v850/v850.h, config/v850/v850.md:
916 Follow spelling conventions.
918 2007-02-03 Douglas Gregor <doug.gregor@gmail.com>
920 * c-opts.c (c_common_post_options): If C++0x mode is enabled, don't
921 warn about C++0x compatibility.
923 2007-02-04 Kazu Hirata <kazu@codesourcery.com>
925 * config/h8300/h8300.c, config/h8300/h8300.h,
926 config/h8300/h8300.md: Follow spelling conventions.
928 2007-02-03 Uros Bizjak <ubizjak@gmail.com>
931 * combine.c (try_combine): Do not substitute source operand
932 with constants wider than 2 * HOST_BITS_PER_WIDE_INT.
934 2007-02-03 Jan Hubicka <jh@suse.cz>
936 PR gcov-profile/30650
937 * value-prof.c (stringop_block_profile): Fix handling of size counter;
938 do not divide by zero for never executed counters.
939 (tree_find_values_to_profile): Fix counters.
940 * gcov-ui.h (GCOV_COUNTER_AVERAGE, GCOV_COUNTER_IOR): Fix comments.
942 2007-02-03 Ian Lance Taylor <iant@google.com>
944 * lower-subreg.c (simple_move_operand): New static function,
945 broken out of simple_move. Reject LABEL_REF, SYMBOL_REF, and HIGH
947 (simple_move): Call simple_move_operand.
948 (find_decomposable_subregs): Add special handling of MEMs.
949 (can_decompose_p): Rename from cannot_decompose_p. Reverse
950 meaning of return value. If we see a hard register, test whether
951 it can store a word_mode value. Change all callers.
953 2007-02-03 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
955 * pa.md (addvdi3, addvsi3, subvdi3, subvsi3, negvdi2, negvsi2): New
956 ftrapv insns and expanders.
957 (subdi3): Change define_expand operand 1 to arith11_operand, and
958 operand 2 to reg_or_0_operand. Change constraints of 64-bit insn
959 pattern to handle reg_or_0 operands. Revise 32-bit insn pattern to
960 handle 11-bit constants and reg_or_0 operands in operands 1 and 2,
964 * varasm.c (notice_global_symbol): Treat global objects as weak when
967 2007-02-03 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
969 * emit-rtl.c (dconstpi): Delete.
971 (init_emit_once): Delete dconstpi and init dconstsqrt2.
972 * real.h (dconstpi): Delete.
974 * builtins.c (fold_builtin_cabs): Use dconstsqrt2.
975 (fold_builtin_hypot): Likewise.
977 2007-02-03 Tom Tromey <tromey@redhat.com>
980 * gcc.c (cpp_unique_options): Any of -ggdb3, -gstabs3,
981 -gcoff3, -gxcoff3, -gvms3 implies -dD.
983 2007-02-03 Kazu Hirata <kazu@codesourcery.com>
985 * c-decl.c, config/avr/avr.c, config/avr/avr.h,
986 config/m68k/m68k.c, config/m68k/netbsd-elf.h,
987 config/mn10300/mn10300.c, config/pdp11/pdp11.h,
988 config/rs6000/cell.md, config/rs6000/darwin.h,
989 config/sh/sh.md, config/sh/sh4-300.md, config/spu/spu.c,
990 config/spu/spu.md, cselib.c, expr.c, haifa-sched.c, hwint.h,
991 jump.c, reload.c, sched-deps.c, sched-int.h, tree-inline.c,
992 tree-profile.c, tree-ssa-live.h, tree-vrp.c: Fix comment
993 typos. Follow spelling conventions.
994 * doc/invoke.texi: Follow spelling conventions.
996 2007-02-03 Roger Sayle <roger@eyesopen.com>
998 * simplify-rtx.c (simplify_relational_operation_1): Implement some
999 canonicalization transformations that attempt to simplify integer
1000 constant comparisons to become comparisons against zero.
1002 2007-02-03 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1004 * builtins.c (fold_builtin_cabs): Fold cabs(x+xi) into
1006 * fold-const.c (fold_binary): Fix comment typos. Fold complex
1007 (x,0)-(0,y) into (x,-y). Likewise (0,y)-(x,0) into (-x,y).
1009 2007-02-02 Mike Stump <mrs@apple.com>
1011 * config/darwin9.h (DARWIN_LINKER_GENERATES_ISLANDS): Add.
1012 * config/rs6000/rs6000.c (DARWIN_GENERATE_ISLANDS): Add.
1013 (output_call): Use DARWIN_GENERATE_ISLANDS to decide when to
1014 generate a branch island.
1016 2007-02-02 Bob Wilson <bob.wilson@acm.org>
1018 * config/xtensa/xtensa.c (smalloffset_mem_p): Use BASE_REG_P.
1019 (xtensa_legitimate_address_p): New.
1020 (xtensa_legitimize_address): New.
1021 (xtensa_output_addr_const_extra): New.
1022 * config/xtensa/xtensa.h (REG_OK_STRICT_FLAG): Define.
1024 (REG_OK_FOR_BASE_P): Use BASE_REG_P.
1025 (GO_IF_LEGITIMATE_ADDRESS): Move code to xtensa_legitimate_address_p.
1026 (LEGITIMIZE_ADDRESS): Move code to xtensa_legitimize_address.
1027 (OUTPUT_ADDR_CONST_EXTRA): Move code to xtensa_output_addr_const_extra.
1028 * config/xtensa/xtensa-protos.h (xtensa_legitimate_address_p): New.
1029 (xtensa_legitimize_address): New.
1030 (xtensa_output_addr_const_extra): New.
1032 2007-02-02 Steve Ellcey <sje@cup.hp.com>
1034 * config/ia64/ia64.c (ia64_print_operand): Fix compare strings.
1036 2007-02-02 Ian Lance Taylor <iant@google.com>
1038 * expmed.c (expand_divmod): Add comment.
1040 2007-02-02 Kazu Hirata <kazu@codesourcery.com>
1042 * emit-rtl.c (renumber_insns): Remove.
1043 * flags.h: Remove the extern for flag_renumber_insns.
1044 * rtl.h: Remove the prototype for renumber_insns.
1045 * toplev.c (flag_renumber_insns): Remove.
1047 2007-02-02 Hui-May Chang <hm.chang@apple.com>
1049 Revert for x86 darwin:
1050 2005-06-19 Uros Bizjak <uros@kss-loka.si>
1052 * config/i386/i386.c (ix86_function_arg_regno_p): Put back the
1053 code before the following patch under TARGET_MACHO.
1054 (ix86_function_value_regno_p): Likewise.
1056 2007-02-02 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1058 * fold-const.c (negate_expr_p, fold_negate_expr): Handle
1061 2007-02-02 Maxim Kuvyrkov <mkuvyrkov@ispras.ru>
1064 * config/ia64/ia64.c (ia64_speculate_insn): Restrict to memory loads to
1065 general or fp registers. Add comments.
1066 * config/ia64/ia64.md (reg_pred_prefix): Add comment.
1068 2007-02-02 Paolo Bonzini <bonzini@gnu.org>
1070 * pointer-set.c (insert_aux): Only return insertion slot.
1071 (pointer_set_insert): Adjust.
1072 (pointer_set_traverse, struct pointer_map_t, pointer_map_create,
1073 pointer_map_destroy, pointer_map_insert, pointer_map_contains,
1074 pointer_map_traverse): New.
1075 * pointer-set.h (pointer_set_traverse, struct pointer_map_t,
1076 pointer_map_create, pointer_map_destroy, pointer_map_insert,
1077 pointer_map_contains, pointer_map_traverse): Declare.
1079 2007-02-02 Jakub Jelinek <jakub@redhat.com>
1082 * builtins.c (fold_builtin_sprintf): Do not attempt to optimize
1083 sprintf (str, "%s"). Do not optimize sprintf (str, "nopercent", p++).
1085 2007-02-02 Maxim Kuvyrkov <mkuvyrkov@ispras.ru>
1087 * sched-int.h (ds_to_dk, dk_to_ds): Declare functions.
1089 (struct _dep): New type.
1090 (dep_t): New typedef.
1091 (DEP_PRO, DEP_CON, DEP_KIND): New access macros.
1092 (DEP_STATUS): New access macro. The macro with the same name was
1093 renamed to DEP_LINK_STATUS.
1094 (dep_init): Declare function
1096 (struct _dep_link): New type.
1097 (dep_link_t): New typedef.
1098 (DEP_LINK_NODE, DEP_LINK_NEXT, DEP_LINK_PREV_NEXTP): New access macros.
1099 (DEP_LINK_DEP, DEP_LINK_PRO, DEP_LINK_CON, DEP_LINK_KIND): New macros.
1100 (DEP_LINK_STATUS): New macro.
1101 (debug_dep_links): New debug function.
1103 (struct _deps_list): New type.
1104 (deps_list_t): New typedef.
1105 (DEPS_LIST_FIRST): New access macro.
1106 (FOR_EACH_DEP_LINK): New cycle macro.
1107 (create_deps_list, free_deps_list, delete_deps_list): Declare
1109 (deps_list_empty_p, debug_deps_list, add_back_dep_to_deps_list): Ditto.
1110 (find_link_by_pro_in_deps_list, find_link_by_con_in_deps_list): Ditto.
1111 (copy_deps_list_change_con): Ditto.
1113 (move_dep_link): Declare function.
1115 (struct _dep_node): New type.
1116 (dep_node_t): New typedef.
1117 (DEP_NODE_BACK, DEP_NODE_DEP, DEP_NODE_FORW): New access macros.
1119 (struct haifa_insn_data.back_deps): New field to hold backward
1120 dependencies of the insn.
1121 (struct haifa_insn_data.depend): Rename to forw_deps. Change its type
1123 (struct haifa_insn_data.resolved_deps): Rename to resolved_back_deps.
1124 Change its type to deps_list_t.
1125 (INSN_BACK_DEPS): New access macro to use instead of LOG_LINKS.
1126 (INSN_DEPEND): Rename to INSN_FORW_DEPS.
1127 (RESOLVED_DEPS): Rename to INSN_RESOLVED_BACK_DEPS.
1129 (INSN_COST): Move to haifa-sched.c. Use insn_cost () instead.
1131 (DEP_STATUS): Rename to DEP_LINK_STATUS. Fix typo in the comment.
1133 (add_forw_dep, delete_back_forw_dep, insn_cost): Update declaration and
1135 (dep_cost): Declare.
1137 * sched-deps.c (CHECK): New macro to (en/dis)able sanity checks.
1138 (ds_to_dk, dk_to_ds): New functions.
1140 (init_dep_1): New static function.
1141 (init_dep): New function.
1142 (copy_dep): New static function.
1144 (dep_link_consistent_p, attach_dep_link, add_to_deps_list): New static
1146 (detach_dep_link): New static function.
1147 (move_dep_link): New function.
1149 (dep_links_consistent_p, dump_dep_links): New static functions.
1150 (debug_dep_links): New debugging function.
1152 (deps_obstack, dl_obstack, dn_obstack): New static variables.
1154 (alloc_deps_list, init_deps_list): New static functions.
1155 (create_deps_list): New function.
1156 (clear_deps_list): New static function.
1157 (free_deps_list, delete_deps_list, deps_list_empty_p): New functions.
1158 (deps_list_consistent_p, dump_deps_list): New static functions.
1159 (debug_deps_list): New function.
1160 (add_back_dep_to_deps_list, find_link_by_pro_in_deps_list): New
1162 (find_link_by_con_in_deps_list, copy_deps_list_change_con): Ditto.
1164 (maybe_add_or_update_back_dep_1, add_or_update_back_dep_1): Update to
1165 use new scheduler dependencies lists.
1166 (add_back_dep, delete_all_dependences, fixup_sched_groups): Ditto.
1167 (sched_analyze): Ditto. Initialize dependencies lists.
1168 (add_forw_dep, compute_forward_dependences): Update to use new
1169 scheduler dependencies lists.
1171 (init_dependency_caches): Init deps_obstack.
1172 (free_dependency_caches): Free deps_obstack.
1174 (adjust_add_sorted_back_dep, adjust_back_add_forw_dep): Update to use
1175 new scheduler dependencies lists.
1176 (delete_forw_dep, add_or_update_back_forw_dep): Ditto.
1177 (add_back_forw_dep, delete_back_forw_dep): Ditto.
1179 * sched-rgn.c (set_spec_fed, find_conditional_protection, is_pfree):
1180 Update to use new scheduler dependencies lists.
1181 (is_conditionally_protected, is_prisky, add_branch_dependences): Ditto.
1182 (debug_dependencies): Ditto.
1183 (schedule_region): Update comments.
1185 * sched-ebb.c (earliest_block_with_similiar_load): Update to use new
1186 scheduler dependencies lists.
1187 (schedule_ebb): Update comments.
1189 * rtl.def (DEPS_LIST): Remove.
1191 * lists.c (unused_deps_list): Remove.
1192 (free_list): Update assertions.
1194 (alloc_DEPS_LIST, free_DEPS_LIST_list, free_DEPS_LIST_node): Remove.
1195 (remove_free_DEPS_LIST_elem, copy_DEPS_LIST_list): Ditto.
1197 * rtl.h (free_DEPS_LIST_list, alloc_DEPS_LIST): Remove declarations.
1198 (remove_free_DEPS_LIST_elem, copy_DEPS_LIST_list): Ditto.
1200 * haifa-sched.c (comments): Update.
1201 (insn_cost1): Remove. Inline the code into insn_cost ().
1202 (insn_cost): Update to use new scheduler dependencies lists. Move
1203 processing of the dependency cost to dep_cost ().
1204 (dep_cost): New function. Use it instead of insn_cost () when
1205 evaluating cost of the dependency. Use compatible interface to
1206 interact with the target.
1207 (priority): Update to use new scheduler dependencies lists.
1208 (rank_for_schedule): Ditto. Optimize heuristic that prefers the insn
1209 with greater number of insns that depend on the insn.
1210 (schedule_insn): Update to use new scheduler dependencies lists. Add
1211 code to free backward dependencies lists. Inline and optimize code
1212 from resolve_dep () - see PR28071.
1213 (ok_for_early_queue_removal): Update to use new scheduler dependencies
1214 lists. Update call to targetm.sched.is_costly_dependence hook.
1216 (fix_inter_tick, try_ready, fix_tick_ready): Update to use new
1217 scheduler dependencies lists.
1219 (resolve_dep): Remove. Move the logic to schedule_insn ().
1220 (init_h_i_d): Initialize dependencies lists.
1222 (process_insn_depend_be_in_spec): Rename to
1223 process_insn_forw_deps_be_in_spec. Update to use new scheduler
1225 (add_to_speculative_block, create_check_block_twin, fix_recovery_deps):
1226 Update to use new scheduler dependencies lists.
1227 (clear_priorities, calc_priorities, add_jump_dependencies): Ditto.
1229 * ddg.c (create_ddg_dependence, create_ddg_dep_no_link): Update to use
1230 new scheduler dependencies lists.
1231 (build_intra_loop_deps): Ditto.
1233 * target.h (struct _dep): Declare to use in
1234 gcc_target.sched.is_costly_dependence.
1235 (struct gcc_target.sched.adjust_cost): Fix typo.
1236 (struct gcc_target.sched.is_costly_dependence): Change signature to use
1237 single dep_t parameter instead of an equivalent triad.
1238 (struct gcc_target.sched.adjust_cost_2): Remove.
1240 * target-def.h (TARGET_SCHED_ADJUST_COST_2): Remove.
1242 * reg-notes.def (DEP_TRUE, DEP_OUTPUT, DEP_ANTI): Update comments.
1244 * doc/tm.texi (TARGET_SCHED_IS_COSTLY_DEPENDENCE): Update
1246 (TARGET_SCHED_ADJUST_COST_2): Remove documentation.
1248 * doc/rtl.texi (LOG_LINKS): Remove part about instruction scheduler.
1249 (REG_DEP_TRUE): Document.
1251 * config/ia64/ia64.c (ia64_adjust_cost_2): Rename to ia64_adjust_cost.
1252 Change signature to correspond to the targetm.sched.adjust_cost hook.
1253 Update use in TARGET_SCHED_ADJUST_COST_2.
1254 (TARGET_SCHED_ADJUST_COST_2): Rename to TARGET_SCHED_ADJUST_COST.
1255 (ia64_dependencies_evaluation_hook, ia64_dfa_new_cycle): Update to use
1256 new scheduler dependencies lists.
1257 (ia64_gen_check): Ditto.
1259 * config/mips/mips.c (vr4130_swap_insns_p): Update to use new scheduler
1262 * config/rs6000/rs6000.c (rs6000_is_costly_dependence): Change
1263 signature to correspond to the targetm.sched.is_costly_dependence hook.
1264 (is_costly_group): Update to use new scheduler dependencies lists.
1266 * config/spu/spu.c (spu_sched_adjust_cost): Use insn_cost () function
1267 instead of INSN_COST () macro.
1269 2007-02-01 Ian Lance Taylor <iant@google.com>
1271 * lower-subreg.c (resolve_clobber): Handle a subreg of a concatn.
1273 2007-02-01 Guy Martin <gmsoft@gentoo.org>
1275 * pa.md (tp_load): Correct mfctl instruction syntax.
1277 2007-02-01 Geoffrey Keating <geoffk@apple.com>
1279 * config/rs6000/rs6000.c (rs6000_stack_info): Correct
1280 altivec_padding_size calculation on AIX. Improve comment, add
1281 assert to verify that it's right.
1283 * config/rs6000/darwin.h (MD_UNWIND_SUPPORT): Don't define for
1286 2007-2-01 Seongbae Park <seongbae.park@gmail.com>
1289 * global.c (global_alloc): Add mising initialization of
1292 2007-02-01 Roger Sayle <roger@eyesopen.com>
1294 * alias.c (init_alias_analysis): Correct whitespace.
1295 * bb-reorder.c (fix_edges_for_rarely_executed_code,
1296 partition_hot_cold_basic_blocks): Likewise.
1297 * builtins.c (expand_builtin_printf, expand_builtin_fprintf,
1298 expand_builtin_sprintf, fold_builtin_carg, fold_builtin_sprintf,
1299 maybe_emit_sprintf_chk_warning, fold_builtin_sprintf_chk,
1300 fold_builtin_snprintf_chk, fold_builtin_printf,
1301 fold_builtin_fprintf, do_mpfr_ckconv, do_mpfr_arg1, do_mpfr_arg2,
1302 do_mpfr_arg3, do_mpfr_sincos): Likewise.
1303 * cfgcleanup.c (cleanup_cfg): Likewise.
1304 * cfgexpand.c (tree_expand_cfg): Likewise.
1305 * fold-const.c (fold_binary) <RDIV_EXPR>: Likewise.
1306 * function.c (get_next_funcdef_no): Likewise.
1307 * gengtype.c (main): Likewise.
1308 * genmodes.c (main): Likewise.
1309 * gcse.c (bypass_conditional_jumps, print_ldst_list): Likewise.
1310 * haifa-sched.c (schedule_block, extend_h_i_d): Likewise.
1311 * ifcvt.c (noce_emit_move_insn): Likewise.
1312 * modulo-sched.c (generate_prolog_epilog, sms_schedule_by_order):
1314 * stor-layout.c (get_best_mode): Likewise.
1315 * tree-ssa-loop-niter.c (get_val_for): Likewise.
1316 * tree-ssa-structalias.c (get_varinfo, get_varinfo_fc,
1317 scc_visit, do_ds_constraint, do_complex_constraint, label_visit,
1318 perform_var_substitution, solve_graph): Likewise.
1319 * tree-vrp.c (vrp_finalize): Likewise.
1321 2007-02-01 Ian Lance Taylor <iant@google.com>
1323 * lower-subreg.c (simplify_gen_subreg_concatn): If we ask for the
1324 high part of a paradoxical subreg, return a constant zero.
1326 2007-02-01 Zdenek Dvorak <dvorakz@suse.cz>
1328 * toplev.c (lang_dependent_init): Call init_set_costs.
1329 * loop-init.c (loop_optimizer_init): Do not call init_set_costs.
1331 2007-02-01 Richard Guenther <rguenther@suse.de>
1334 * fold-const.c (fold_negate_expr): Allow negating a
1335 constant if overflow does not change.
1337 2007-02-01 Ralf Wildenhues <Ralf.Wildenhues@gmx.de>
1339 * doc/c-tree.texi (Expression trees): Improve markup.
1340 * doc/tm.texi (Register Classes, Addressing Modes)
1341 (Floating Point): Fix spacing after abbreviations. Fix some
1344 2007-02-01 Ben Elliston <bje@au.ibm.com>
1346 * doc/invoke.texi: Replace "bugfix" with "bug fix" throughout.
1347 * doc/contrib.texi: Likewise.
1348 * doc/install.texi: Likewise.
1350 2007-01-31 Richard Henderson <rth@redhat.com>
1351 Ian Lance Taylor <iant@google.com>
1353 * lower-subreg.c: New file.
1354 * rtl.def (CONCATN): Define.
1355 * passes.c (init_optimization_passes): Add pass_lower_subreg and
1357 * emit-rtl.c (update_reg_offset): New static function, broken out
1358 of gen_rtx_REG_offset.
1359 (gen_rtx_REG_offset): Call update_reg_offset.
1360 (gen_reg_rtx_offset): New function.
1361 * regclass.c: Revert patch of 2006-03-05, restoring
1363 (clear_reg_info_regno): New function.
1364 * dwarf2out.c (concatn_loc_descriptor): New static function.
1365 (loc_descriptor): Handle CONCATN.
1366 * common.opt (fsplit_wide_types): New option.
1367 * opts.c (decode_options): Set flag_split_wide_types when
1369 * timevar.def (TV_LOWER_SUBREG): Define.
1370 * rtl.h (gen_reg_rtx_offset): Declare.
1371 (reg_scan_update): Declare.
1372 * regs.h (clear_reg_info_regno): Declare.
1373 * tree-pass.h (pass_lower_subreg): Declare.
1374 (pass_lower_subreg2): Declare.
1375 * doc/invoke.texi (Option Summary): List -fno-split-wide-types.
1376 (Optimize Options): Add -fsplit-wide-types to -O1 list. Document
1378 * doc/rtl.texi (Regs and Memory): Document concat and concatn.
1379 * Makefile.in (OBJS-common): Add lower-subreg.o.
1380 (lower-subreg.o): New target.
1382 2007-01-31 Kazu Hirata <kazu@codesourcery.com>
1384 * config/sh/sh.h (HAVE_SECONDARY_RELOADS): Remove.
1386 2007-01-31 Anatoly Sokolov <aesok@post.ru>
1389 * config/avr/avr.c (DWARF2_ADDR_SIZE): Define.
1391 2007-01-31 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1394 * builtins.c (fold_builtin_sqrt): Use MPFR for constant args.
1396 2007-01-31 Zdenek Dvorak <dvorakz@suse.cz>
1398 * cfgloop.h: Include vec-prim.h.
1399 (enum li_flags): Remove LI_ONLY_OLD.
1400 (loop_iterator): Changed.
1401 (fel_next, fel_init): Iterate over loop tree.
1402 (FOR_EACH_LOOP_BREAK): New macro.
1403 * loop-unswitch.c (unswitch_loops): Do not pass LI_ONLY_OLD to
1405 * tree-ssa-loop-unswitch.c (tree_ssa_unswitch_loops): Ditto.
1406 * modulo-sched.c (sms_schedule): Ditto.
1407 * tree-vectorizer.c (vectorize_loops): Ditto.
1408 * doc/loop.texi: Update information on loop numbering and behavior of
1409 FOR_EACH_LOOP wrto new loops.
1410 * tree-scalar-evolution.c (compute_overall_effect_of_inner_loop,
1411 add_to_evolution_1): Test nestedness of loops instead of comparing
1413 * tree-chrec.c (chrec_fold_plus_poly_poly,
1414 chrec_fold_multiply_poly_poly, chrec_evaluate,
1415 hide_evolution_in_other_loops_than_loop, chrec_component_in_loop_num,
1416 reset_evolution_in_loop): Ditto.
1417 * Makefile.in (CFGLOOP_H): Add vecprim.h dependency.
1419 2007-01-31 Dirk Mueller <dmueller@suse.de>
1421 * c-common.c (warn_about_parentheses): Separate warning about
1422 un-parenthized sequence of comparison operators from the one
1423 which is supposed to warn about x <= y <= z.
1425 2007-01-31 Uros Bizjak <ubizjak@gmail.com>
1427 * optabs.h (enum optab_index): Add new OTI_isinf.
1428 (isinf_optab): Define corresponding macro.
1429 * optabs.c (init_optabs): Initialize isinf_optab.
1430 * genopinit.c (optabs): Implement isinf_optab using isinf?f2
1432 * builtins.c (mathfn_built_in): Handle BUILT_IN_ISINF{,F,L}.
1433 (expand_builtin_interclass_mathfn): Expand BUILT_IN_ISINF{,F,L}
1435 (expand_builtin): Expand BUILT_IN_ISINF{,F,L} using
1436 expand_builtin_interclass_mathfn.
1437 * reg_stack.c (subst_stack_regs_pat): Handle UNSPEC_FXAM.
1438 * config/i386/i386.md (UNSPEC_FXAM): New constant.
1439 (fxam<mode>2_i387): New insn pattern.
1440 (isinf<mode>2) New expander to implement isinf, isinff and isinfl
1441 built-in functions as x87 inline asm.
1443 2007-01-31 Kazu Hirata <kazu@codesourcery.com>
1445 * gcc/config/arm/unwind-arm.h (_sleb128_t, _uleb128_t): New.
1447 2007-01-30 Eric Christopher <echristo@apple.com>
1449 * config.gcc: Add geode.
1451 2007-01-31 Kazu Hirata <kazu@codesourcery.com>
1453 * cgraphunit.c, config/arm/arm.c, config/m68k/m68k.c,
1454 ipa-inline.c, tree-profile.c, tree-ssa-live.c,
1455 tree-ssa-math-opts.c, tree-ssanames.c, tree-vect-analyze.c,
1456 value-prof.c: Fix comment typos.
1458 2007-01-30 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
1461 * doc/invoke.texi (Wpointer-arith): Document warning.
1463 2007-01-30 Janis Johnson <janis187@us.ibm.com>
1465 * doc/extend.texi (Decimal Floating Types): Remove decfloat.h from
1466 the list of discrepancies from the draft TR.
1468 2007-01-30 Dirk Mueller <dmueller@suse.de>
1471 * doc/invoke.texi (-Wreturn-type): Update description to
1474 2007-01-30 Richard Sandiford <richard@codesourcery.com>
1476 * cfgrtl.c (try_redirect_by_replacing_jump): Check only_sets_cc0_p.
1478 2007-01-30 Uros Bizjak <ubizjak@gmail.com>
1480 * builtins.c (expand_builtin_int_interclass_roundingfn): New function
1481 to handle optabs that operate on floating point input argument and
1482 output to integer output.
1483 (expand_builtin_mathfn) [BUILT_IN_ILOGB]: Move from here ...
1484 (expand_builtin_interclass_mathfn) [BUILT_IN_ILOGB]: ... to here.
1485 (expand_builtin): Expand BUILT_IN_ILOGB{,F,L} using
1486 expand_builtin_interclass_mathfn ().
1487 * config/i386/i386.md (fxtractxf3_i387): Rename from *fxtractxf3_i387.
1489 (ilogbxf2, ilogb<mode>2): New expanders to implement ilogb, ilogbf and
1490 ilogbl built-in functions as x87 intrinsics.
1492 2007-01-30 Richard Guenther <rguenther@suse.de>
1495 * dwarf2out.c (reference_to_unused): Query varpool if the
1496 variable was output.
1498 2007-01-30 Richard Guenther <rguenther@suse.de>
1501 * passes.c (execute_one_pass): Reset in_gimple_form to not
1502 confuse non-unit-at-a-time mode.
1504 2007-01-29 Roger Sayle <roger@eyesopen.com>
1505 Richard Guenther <rguenther@suse.de>
1507 * fold-const.c (round_up): Make HIGH an unsigned HOST_WIDE_INT to
1508 avoid undefined behaviour on overflow. Use force_fit_type_double
1509 to construct the constant with the specified TREE_OVERFLOW.
1511 2007-01-29 Janis Johnson <janis187@us.ibm.com>
1513 * config/dfp-bit.c: Add parameterized support for fp exceptions.
1514 * config/dfp-bit.h: Ditto.
1516 2007-01-29 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
1518 * c-decl.c (pop_scope): Replace warnings with call to
1519 warn_for_unused_label.
1520 * c-common.h (warn_for_unused_label): Declare.
1521 * c-common.c (warn_for_unused_label): Define.
1523 2007-01-29 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
1525 * tree-optimize.c (update_inlined_to_pointers): Delete unused
1528 2007-01-29 Janis Johnson <janis187@us.ibm.com>
1530 * Makefile.in (USER_H): Remove decfloat.h.
1531 * ginclude/decfloat.h: Delete, moving contents to ...
1532 * ginclude/float.h: Add support for decimal floating point,
1533 guarded by __STDC_WANT_DEC_FP__.
1535 2007-01-29 Mike Stump <mrs@apple.com>
1537 * doc/gccint.texi (Top): Rename Loop Representation to Loop
1538 Analysis and Representation to resolve case insensitive conflict.
1539 * doc/loop.texi (Loop Analysis and Representation): Likewise.
1541 2007-01-28 Daniel Berlin <dberlin@dberlin.org>
1543 * tree.h (struct tree_memory_tag): Add aliases member.
1544 (MTAG_ALIASES): New macro.
1545 * tree-ssa-alias.c (alias_bitmap_obstack): New variable.
1546 (add_may_alias): Remove pointer-set. Update for may_aliases being
1548 (mark_aliases_call_clobbered): Update for may_aliases being a
1550 (compute_tag_properties): Ditto.
1551 (create_partition_for): Ditto.
1552 (compute_memory_partitions): Ditto.
1553 (dump_may_aliases_for): Ditto.
1554 (is_aliased_with): Ditto.
1555 (add_may_alias_for_new_tag): Ditto.
1556 (rewrite_alias_set_for): Rewrite for may_aliases being a bitmap.
1557 (compute_is_aliased): New function.
1558 (compute_may_aliases): Call compute_is_aliased).
1559 (init_alias_info): Initialize alias_bitmap_obstack.
1560 (union_alias_set_into): New function.
1561 (compute_flow_sensitive_aliasing): Use union_aliases_into.
1562 (have_common_aliases_p): Rewrite to take two bitmaps and use
1564 (compute_flow_insensitive_aliasing): Stop using pointer-sets.
1566 (finalize_ref_all_pointers): Update for add_may_alias changes.
1567 (new_type_alias): Ditto.
1568 * tree-flow-inline.h (may_aliases): Return a bitmap.
1569 * tree-dfa.c (dump_variable): Check for MTAG_P'ness.
1570 * tree-ssa.c (verify_flow_insensitive_alias_info): Update for
1571 may_aliases being a bitmap.
1572 * tree-flow.h (struct var_ann_d): Remove may_aliases member.
1573 may_aliases now returns a bitmap.
1574 * tree-ssa-structalias.c (merge_smts_into): Update for may_aliases
1576 * tree-ssa-operands.c (add_virtual_operand): Update for
1577 may_aliases being a bitmap.
1579 2007-01-29 Daniel Berlin <dberlin@dberlin.org>
1581 PR tree-optimization/30630
1582 * tree-ssa-structalias.c (do_complex_constraint): Mark correct
1583 variable as changed.
1585 2007-01-29 Simon Martin <simartin@users.sourceforge.net>
1588 * gimplify.c (gimplify_target_expr): Make sure that the TARGET_EXPR is
1589 expanded only once even if an error occurs.
1591 2007-01-29 Ben Elliston <bje@au.ibm.com>
1593 * gcov-io.h (__gcov_indirect_call_profiler): Declare.
1594 (__gcov_average_profiler): Likewise.
1595 (__gcov_ior_profiler): Likewise.
1596 (__gcov_merge_ior): Likewise.
1598 2007-01-28 Jan Hubicka <jh@suse.cz>
1600 * builtins.c (expand_builtin_memset): Fix typo in my last patch.
1601 * value-prof.c (stringop_block_profile): Likewise.
1603 2007-01-28 Jan Hubicka <jh@suse.cz>
1605 * expr.c (emit_block_move_via_movmem, emit_block_move_via_libcall): Add
1606 variant handling histograms; add wrapper.
1607 (clear_storage_via_libcall): Export.
1608 (emit_block_move_hints): Break out from ...; add histograms.
1609 (emit_block_move): ... this one.
1610 (clear_storage_hints): Break out from ...; add histograms.
1611 (clear_storage): ... this one.
1612 (set_storage_via_memset): Handle histogram.
1613 * expr.h (emit_block_move_via_libcall, emit_block_move_hints): Declare.
1614 (clear_storage_hints, clear_storage_via_libcall): Declare.
1615 (set_storage_via_setmem): Update prototype.
1616 * doc/md.texi (movmem, setmem): Document new arguments.
1618 * value-prof.c (dump_histogram_value, tree_find_values_to_profile): Add
1620 (stringop_block_profile): New global function.
1621 (tree_stringops_values_to_profile): Profile block size and alignment.
1622 * value-prof.h (enum hist_type): add HIST_TYPE_AVERAGE and
1624 (struct profile_hooks): Add gen_average_profiler and gen_ior_profiler.
1625 (stringop_block_profile): Declare.
1626 * builtins.c: Include value-prof.h.
1627 (expand_builtin_memcpy, expand_builtin_memset): Pass block profile.
1628 * gcov-ui.h (GCOV_COUNTER_NAMES): Add new counter.
1629 (GCOV_COUNTER_AVERAGE, GCOV_COUNTER_IOR): New constants.
1630 (GCOV_COUNTERS, GCOV_LAST_VALUE_COUNTER): Update.
1631 * profile.c (instrument_values): Add new counters.
1632 * cfgexpand.c (expand_gimple_basic_block): Propagate histograms to
1634 * tree-profile.c (tree_average_profiler_fn, tree_ior_profiler_fn): New.
1635 (tree_init_edge_profiler): Build new profilers.
1636 (tree_gen_average_profiler, tree_gen_ior_profiler): New.
1637 (pass_tree_profile): Add dump.
1638 (tree_profile_hooks): Update.
1639 * Makefile.in (LIBGCOV): Add new constants.
1640 * libgcov.c (__gcov_merge_ior, __gcov_average_profiler,
1641 __gcov_ior_profiler): New.
1642 * i386.md (movmem/setmem expanders): Add new optional arguments.
1644 2007-01-28 David Edelsohn <edelsohn@gnu.org>
1646 * doc/md.texi (Standard Pattern Names): Document blockage pattern.
1648 2007-01-28 Zdenek Dvorak <dvorakz@suse.cz>
1650 * tree-ssa-loop-unswitch.c: Include tree-inline.h.
1651 (tree_unswitch_single_loop): Pass eni_size_weights to
1652 tree_num_loop_insns.
1653 * tree-ssa-loop-manip.c: Include tree-inline.h.
1654 (can_unroll_loop_p): Pass eni_size_weights to
1655 tree_num_loop_insns.
1656 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p):
1657 Pass eni_size_weights to estimate_num_insns.
1658 * tree.h (init_inline_once): Export.
1659 * toplev.c (backend_init): Call init_inline_once.
1660 * cgraphunit.c (cgraph_process_new_functions,
1661 cgraph_analyze_function): Pass eni_inlining_weights to
1663 * ipa-inline.c (compute_inline_parameters): Ditto.
1664 * tree-ssa-loop-ivcanon.c (tree_num_loop_insns): Pass weights
1665 to estimate_num_insns.
1666 (try_unroll_loop_completely): Pass eni_size_weights to
1667 tree_num_loop_insns.
1668 * tree-eh.c (decide_copy_try_finally): Pass eni_size_weights
1669 ot estimate_num_insns.
1670 * tree-ssa-loop-prefetch.c: Include tree-inline.h.
1671 (loop_prefetch_arrays): Pass eni_time_weights to tree_num_loop_insns.
1672 * tree-inline.c (eni_inlining_weights, eni_size_weights,
1673 eni_time_weights): New variables.
1674 (init_inline_once): Initialize them.
1675 (struct eni_data): Mew.
1676 (estimate_num_insns_1, estimate_num_insns): Use weights.
1677 * tree-inline.h (struct eni_weights_d): New.
1678 (eni_inlining_weights, eni_size_weights, eni_time_weights): Declare.
1679 (estimate_num_insns): Declaration changed.
1680 * cfgloop.h (tree_num_loop_insns): Declaration changed.
1681 * Makefile.in (tree-ssa-loop-unswitch.o, tree-ssa-loop-prefetch.o,
1682 tree-ssa-loop-manip.o): Add TREE_INLINE_H dependency.
1684 2007-01-28 Zdenek Dvorak <dvorakz@suse.cz>
1686 * tree-data-ref.c (conflict_fn): Assert that the number of affine
1687 relations in the conflict function is valid.
1689 2007-01-27 Ian Lance Taylor <iant@google.com>
1691 * common.opt: Add fstrict-overflow.
1692 * opts.c (decode_options): Set flag_strict_overflow if -O2.
1693 * flags.h (TYPE_OVERFLOW_WRAPS): Define.
1694 (TYPE_OVERFLOW_UNDEFINED): Define.
1695 (TYPE_OVERFLOW_TRAPS): Define. This replaces TYPE_TRAP_SIGNED.
1697 * tree.h (TYPE_TRAP_SIGNED): Don't define.
1698 * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_UNDEFINED.
1699 (fold_negate_expr): Likewise.
1700 (make_range): Likewise.
1701 (extract_muldiv_1): Likewise.
1702 (maybe_canonicalize_comparison): Likewise.
1703 (fold_comparison): Likewise.
1704 (fold_binary): Likewise.
1705 (tree_expr_nonnegative_p): Likewise.
1706 (tree_expr_nonzero_p): Likewise.
1707 * tree-vrp.c (compare_values): Likewise.
1708 (extract_range_from_binary_expr): Likewise.
1709 (extract_range_from_unary_expr): Likewise.
1710 * tree-ssa-loop-niter.c (infer_loop_bounds_from_signedness):
1712 (nowrap_type_p): Likewise.
1713 * tree-scalar-evolution.c (simple_iv): Likewise.
1714 * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_WRAPS.
1715 (build_range_check): Likewise.
1716 (extract_muldiv_1): Likewise.
1717 (fold_comparison): Likewise.
1718 * tree-vrp.c (vrp_int_const_binop): Likewise.
1719 (extract_range_from_unary_expr): Likewise.
1720 * convert.c (convert_to_integer): Likewise.
1721 * fold-const.c (fold_negate_expr): Use TYPE_OVERFLOW_TRAPS.
1722 (fold_comparison): Likewise.
1723 (fold_binary): Likewise.
1724 * optabs.c (optab_for_tree_code): Likewise.
1725 * tree-vectorizer.c (vect_is_simple_reduction): Likewise.
1726 * simplify-rtx.c (simplify_const_relational_operation): Check
1727 flag_strict_overflow and flag_trapv.
1728 (simplify_const_relational_operation): Likewise.
1729 * doc/invoke.texi (Option Summary): Mention -fstrict-overflow.
1730 (Optimize Options): Add -fstrict-overflow to -O2 list. Document
1733 2007-01-27 Roger Sayle <roger@eyesopen.com>
1735 * tree.c (tree_fold_gcd): Delete.
1736 * tree.h (tree_fold_gcd): Remove prototype.
1737 * tree-data-ref.c (tree_fold_divides_p): Don't use tree_fold_gcd to
1738 test whether one constant integer is a multiple of another. Instead
1739 call int_const_binop with TRUNC_MOD_EXPR and test for a zero result.
1740 * fold-const.c (multiple_of_p): We've determined both TOP and
1741 BOTTOM are integer constants so we can call int_const_binop directly
1742 instead of the more generic const_binop.
1744 2007-01-27 Roger Sayle <roger@eyesopen.com>
1746 * fold-const.c (size_binop): In the fast-paths for X+0, 0+X, X-0 and
1747 1*X check that the constant hasn't overflowed, to preserve the
1749 (round_up): Provide an efficient implementation when rouding-up an
1750 INTEGER_CST to a power-of-two.
1752 2007-01-28 Ralf Wildenhues <Ralf.Wildenhues@gmx.de>
1754 * doc/sourcebuild.texi: Add comma for clarity.
1755 * doc/extend.texi: Fix some typos.
1756 * doc/passes.texi: Likewise.
1757 * doc/cppinternals.texi: Likewise.
1758 * doc/c-tree.texi: Likewise.
1759 * doc/tree-ssa.texi: Likewise.
1760 * doc/install.texi: Likewise.
1762 2007-01-27 Jan Hubicka <jh@suse.cz>
1764 * tree-sra.c (sra_walk_function): Don't rely on aliases being build.
1765 (pass_sra): Do not require alias information.
1766 * passes.c (init_optimization_passes): Add SRA
1768 2007-01-27 Steven Bosscher <steven@gcc.gnu.org>
1770 * tracer.c (rest_of_handle_tracer): We already cleaned
1771 up the CFG in tracer() so don't do it here again.
1772 * cfgcleanup.c (rest_of_handle_jump2): Don't repeat
1773 cleanup_cfg here, either. And don't call renumber_insns.
1775 * cfgrtl.c (rtl_verify_flow_info_1): Don't verify that BB_END
1776 and BB_HEAD are in the insn stream here. Instead make sure
1777 that BB_INSN is valid on all insns. Also, do check here that
1778 there are no pending branch predictions...
1779 (rtl_verify_flow_info): ...instead of doing it here. Checks
1780 for BB_END and BB_HEAD moved from rtl_verify_flow_info_1 to
1783 2007-01-26 Roger Sayle <roger@eyesopen.com>
1785 * config/i386/i386.c (ix86_swap_binary_operands_p): New helper
1786 function to simplify/factorize operand order canonicalization.
1787 (ix86_fixup_binary_operands): Reorganize using the above function.
1788 (ix86_binary_operator_ok): Likewise.
1790 2007-01-27 Jakub Jelinek <jakub@redhat.com>
1792 * genattrtab.c (struct attr_value_list, insn_code_values): Move to
1793 file scope from optimize_attrs.
1794 (simplify_test_exp): If insn_code_values is not NULL, use it to speed
1796 (optimize_attrs): Clear insn_code_values after freeing it.
1798 2007-01-26 Zdenek Dvorak <dvorakz@suse.cz>
1800 * tree-ssa-address.c (create_mem_ref): Remove ", bsi" from
1801 a parts.base assignment.
1803 2007-01-26 Zdenek Dvorak <dvorakz@suse.cz>
1805 * tree-data-ref.c (dump_subscript): Use dump_conflict_function.
1806 (compute_subscript_distance, initialize_data_dependence_relation,
1807 finalize_ddr_dependent, analyze_ziv_subscript,
1808 analyze_siv_subscript_cst_affine,
1809 compute_overlap_steps_for_affine_univar,
1810 compute_overlap_steps_for_affine_1_2, analyze_subscript_affine_affine,
1811 analyze_siv_subscript, analyze_miv_subscript,
1812 analyze_overlapping_iterations, subscript_dependence_tester_1,
1813 compute_self_dependence, free_dependence_relation): Work
1814 with affine_fn instead of chrecs.
1815 (dump_affine_function, dump_conflict_function, affine_function_equal_p,
1816 common_affine_function, affine_function_base,
1817 affine_function_constant_p, affine_fn_op, affine_fn_plus,
1818 affine_fn_minus, affine_fn_free, conflict_fn_not_known,
1819 conflict_fn_no_dependence, free_conflict_function, free_subscripts,
1820 conflict_fn, affine_fn_cst, affine_fn_univar): New functions.
1821 (all_chrecs_equal_p): Removed.
1822 * tree-data-ref.h (affine_fn, conflict_function): New types.
1823 (struct subscript): Change type of conflicting_iterations_in_a
1824 and conflicting_iterations_in_b.
1826 2007-01-26 Steve Ellcey <sje@cup.hp.com>
1829 * config/pa/pa.h (TARGET_HPUX_11): New.
1830 * config/pa/pa-hpux11.h (TARGET_HPUX_11): New.
1831 * config/pa/pa.c (pa_init_builtins): Use TARGET_HPUX_11.
1833 2007-01-26 Daniel Berlin <dberlin@dberlin.org>
1834 Richard Guenther <rguenther@suse.de>
1836 * tree-ssa-structalias.c (solve_graph): Handle case
1837 we merged the variable to another.
1839 2007-01-25 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1841 * builtins.c (fold_builtin_1): Treat ccos and ccosh as 'even'
1844 * fold-const.c (negate_mathfn_p): Treat casin, casinh, catan,
1845 catanh, cproj, csin, csinh, ctan and ctanh as 'odd' functions.
1847 2007-01-25 DJ Delorie <dj@redhat.com>
1849 * config/m32c/m32c.c (m32c_cannot_change_mode_class): We don't
1850 allow changes to modes which don't fit in those registers.
1852 * reload1.c (choose_reload_regs): Check for invalid subregs before
1853 computing their locations, not after.
1855 2007-01-25 Geoffrey Keating <geoffk@apple.com>
1858 * config/rs6000/rs6000.c (first_altivec_reg_to_save): On Darwin,
1859 save Altivec registers in an eh_return function.
1860 (compute_vrsave_mask): Likewise.
1861 (rs6000_stack_info): Correct AIX/Darwin stack alignment computation
1862 for saving Altivec registers.
1863 (rs6000_emit_prologue): Don't allocate stack twice in
1864 eh_return function. Correct expected value of altivec_save_offset
1865 when using save_world. Describe save of R0 to stack when using
1866 save_world. Describe stack pointer adjustment when using
1867 save_world. Remove duplicated eh_return parameter register saving.
1868 Update sp_offset variable after save_world.
1869 * config/rs6000/t-darwin (LIB2FUNCS_STATIC_EXTRA): Remove
1871 (LIB2FUNCS_EXTRA): Add darwin-world.asm.
1872 * config/rs6000/darwin.h (SUBTARGET_OVERRIDE_OPTIONS): -m64
1875 2007-01-25 Steve Ellcey <sje@cup.hp.com>
1877 * config.gcc (ia64*-*-hpux*): Make posix threads the default.
1879 2007-01-25 Steve Ellcey <sje@cup.hp.com>
1882 * config/pa/pa.c (pa_init_builtins): Set asm names for finite routines.
1883 * config/ia64/ia64.c (ia64_init_builtins): Ditto.
1885 2007-01-25 Richard Guenther <rguenther@suse.de>
1887 * doc/invoke.texi (-Wcoverage-mismatch): Document.
1888 * common.opt (-Wcoverage-mismatch): New warning option.
1889 * coverage.c (get_coverage_counts): Ignore coverage mismatch
1890 if -Wcoverage-mismatch is given.
1892 2007-01-25 Razya Ladelsky <razya@il.ibm.com>
1894 * ipa-cp.c (ipcp_insert_stage, ipcp_driver): Support for SSA.
1895 (ipcp_driver): Change to static definition.
1896 Add dumping of the ifunctions.
1897 (constant_val_insert): Remove unused parameter. Support for SSA.
1898 (ipcp_propagate_const): Support for SSA.
1899 (ipcp_profile_bb_print): Print only analyzed nodes.
1900 (ipcp_replace_map_create): Remove support for Fortran constant
1902 * ipa-prop.c (ipa_method_modify_stmt,
1903 ipa_callsite_compute_param): Support for SSA.
1904 * ipa-prop.h (ipcp_driver): Remove declaration.
1905 (IS_VALID_TREE_MAP_INDEX): Add define.
1907 2007-01-24 Geoffrey Keating <geoffk@apple.com>
1909 * unwind-dw2.c (execute_stack_op): Handle DW_OP_swap.
1911 2007-01-24 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1914 * builtins.c (fold_builtin_cabs): Use MPFR to evaluate a
1915 constant argument to cabs and do it without checking for
1916 -funsafe-math-optimizations.
1918 2007-01-24 Douglas Gregor <dgregor@osl.iu.edu>
1920 * c-common.h (RID_FIRST_CXX0X): New.
1921 (RID_LAST_CXX0X): New.
1922 * c-opts.c (c_common_handle_option): -Wc++0x-compat is triggered
1924 * c.opt (Wc++0x-compat): New.
1925 * doc/invoke.texi (-Wc++0x-compat): Document.
1927 2007-01-24 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1929 * builtins.c (fold_builtin_carg): New.
1930 (fold_builtin_1): Use it.
1932 2007-01-24 Jan Hubicka <jh@suse.cz>
1934 * ipa-inline.c (cgraph_decide_inlining): Initialize initial_insns.
1936 * ipa-inline.c (initial_insns, max_insns): Delete.
1937 (compute_max_insns): New function.
1938 (cgraph_decide_inlining_of_small_function): Use it; take minimal amount
1939 of insns as base for code growth.
1940 (cgraph_decide_inlining): Make initial_insns local; do not compute
1942 * params.def (PARAM_INLINE_UNIT_GROWTH): Set to 60.
1943 * doc/invoke.texi (inline-unit-growth): Update docs.
1945 2007-01-24 Jakub Jelinek <jakub@redhat.com>
1947 * config/i386/i386.h (x86_cmpxchg16b): Remove const.
1948 (TARGET_CMPXCHG16B): Define to x86_cmpxchg16b.
1949 * config/i386/i386.c (x86_cmpxchg16b): Remove const.
1950 (override_options): Add PTA_CX16 flag. Set x86_cmpxchg16b
1951 for CPUs that have PTA_CX16 set.
1954 * gimplify.c (omp_check_private): New function.
1955 (gimplify_scan_omp_clauses): Use it for
1956 firstprivate/lastprivate/reduction.
1959 * gimplify.c (omp_add_variable): Don't call omp_notice_variable
1960 on TYPE_SIZE_UNIT for GOVD_LOCAL VLAs.
1963 * omp-low.c (lower_omp_for_lastprivate): Add dlist argument.
1964 If lower_lastprivate_clauses emits some statements, append them
1965 to dlist rather than body_p and to body_p append an initializer.
1966 (lower_omp_for): Adjust caller.
1968 2007-01-24 Steve Ellcey <sje@cup.hp.com>
1970 * target.h (globalize_decl_name): New.
1971 * target-def.h (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
1972 * output.h (default_globalize_decl_name): New.
1973 * varasm.c (asm_output_bss): Use globalize_decl_name instead of
1975 (globalize_decl): Ditto.
1976 (default_globalize_decl_name): New.
1977 * config/ia64/ia64.c (ia64_globalize_decl_name): New.
1978 (ia64_handle_version_id_attribute): New.
1979 (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
1980 (ia64_asm_output_external): Use globalize_decl_name instead
1982 * doc/extend.texi (version_id): New pragma.
1983 * doc/tm.texi (ARGET_ASM_GLOBALIZE_DECL_NAME): New target hook.
1985 2007-01-24 Andreas Krebbel <krebbel1@de.ibm.com>
1987 * unwind-dw2-fde.c (get_cie_encoding): Replaced _Unwind_Word with
1988 _uleb128_t and _Unwind_SWord with _sleb128_t.
1989 * unwind-dw2.c (extract_cie_info, execute_stack_op, execute_cfa_program,
1990 uw_frame_state_for, uw_update_context_1): Likewise.
1991 * unwind-c.c (parse_lsda_header, PERSONALITY_FUNCTION): Likewise.
1992 * unwind-pe.h (read_uleb128, read_sleb128,
1993 read_encoded_value_with_base): Likewise.
1994 * unwind-generic.h: Define _sleb128_t and _uleb128_t types.
1996 2007-01-24 Richard Guenther <rguenther@suse.de>
1998 * builtins.c (expand_builtin_cexpi): Get the fndecl
1999 for cexp in the correct way.
2001 2007-01-24 Jan Hubicka <jh@suse.cz>
2003 * tree-ssa-dce.c (eliminate_unnecesary_stmts): Remove dead LHS of calls.
2005 2007-01-24 Andreas Krebbel <krebbel1@de.ibm.com>
2007 * c-cppbuiltin.c (builtin_define_type_sizeof): New function.
2008 (c_cpp_builtins): New builtin macros: __SIZEOF_INT__, __SIZEOF_LONG__,
2009 __SIZEOF_LONG_LONG__, __SIZEOF_SHORT__, __SIZEOF_POINTER__,
2010 __SIZEOF_FLOAT__, __SIZEOF_DOUBLE__, __SIZEOF_LONG_DOUBLE__,
2011 __SIZEOF_SIZE_T__, __SIZEOF_WCHAR_T__, __SIZEOF_WINT_T__ and
2012 __SIZEOF_PTRDIFF_T__.
2013 * doc/cpp.texi: Documentation for the new builtin macros added.
2015 2007-01-24 Uros Bizjak <ubizjak@gmail.com>
2017 * config/i386/i386.md (tanxf2, tan<mode>2, atan<mode>2, log<mode>2,
2018 log10<mode>2, log2<mode>2, expxf2, exp10xf2, exp2xf2): Use op2
2019 instead of operands[2] to avoid access past the end of array.
2021 2007-01-24 Richard Sandiford <richard@codesourcery.com>
2023 * reload1.c (emit_reload_insns): Pass the reload register
2024 for a non-spill output reload through forget_old_reloads_1.
2026 2007-01-23 Joseph Myers <joseph@codesourcery.com>
2028 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Define
2029 _SOFT_DOUBLE if doubles use software floating-point.
2030 * config/rs6000/libgcc-ppc-glibc.ver: Export additional long
2031 double functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
2032 * config/rs6000/darwin-ldouble.c: Also compile functions for
2033 hard-float without FPRs. Use fmsub function for all __NO_FPRS__
2034 cases. Compile extra functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
2035 * config/rs6000/linuxspe.h (SUBSUBTARGET_OVERRIDE_OPTIONS): Remove
2036 commented-out long double override.
2037 (CPP_LONGDOUBLE_DEFAULT_SPEC): Likewise.
2038 * config/rs6000/eabispe.h: Likewise.
2039 * config/rs6000/rs6000.c (rs6000_override_options): Don't override
2040 long double for non-SPE.
2041 (rs6000_handle_option): Likewise.
2042 (invalid_e500_subreg): Disallow more subregs involding DImode,
2043 DFmode, TImode or TFmode.
2044 (rs6000_legitimate_offset_address_p): Check TFmode offsets for
2046 (legitimate_lo_sum_address_p): Also check for TFmode for E500
2048 (rs6000_legitimize_address): Also handle TFmode for E500 double.
2049 (rs6000_legitimize_reload_address): Also handle TFmode for E500
2051 (rs6000_legitimate_address): Also check for TFmode for E500
2053 (rs6000_emit_move): Use DFmode subregs of TFmode for E500 double.
2054 (spe_build_register_parallel): Handle TFmode and TCmode.
2055 (rs6000_spe_function_arg): Handle TFmode and TCmode for E500
2057 (function_arg): Handle TFmode and TCmode for E500 double.
2058 (rs6000_init_libfuncs): Initialize extra libfuncs for soft double
2060 (print_operand): Handle TFmode and TImode for %y.
2061 (rs6000_generate_compare): Handle TFmode comparisons for E500
2063 (spe_func_has_64bit_regs_p): Check for TFmode for E500 double.
2064 (rs6000_function_value): Handle TFmode and TCmode for E500 double.
2065 (rs6000_libcall_value): Handle TFmode and TCmode for E500 double.
2066 * config/rs6000/rs6000.h (CANNOT_CHANGE_MODE_CLASS): Check for
2067 TFmode for E500 double.
2068 * config/rs6000/rs6000.md (FP): Allow TF for E500 double.
2069 (floatsidf2): Enable for E500 double.
2070 (movtf_softfloat): Use rs6000_nonimmediate_operand.
2071 (extenddftf2): Change to extenddftf2_fprs.
2072 (extenddftf2): Call gen_spe_extenddftf2 or gen_extenddftf2_fprs
2073 depending on TARGET_E500_DOUBLE.
2074 (extendsftf2): Enable for E500 double.
2075 (trunctfdf2): Enable for E500 double.
2076 (trunctfsf2): Change to trunctfsf2_fprs.
2077 (trunctfsf2): Call gen_spe_trunctfsf2 or gen_trunctfsf2_fprs
2078 depending on TARGET_E500_DOUBLE.
2079 (floatsitf2): Enable for E500 double.
2080 (fix_trunctfsi2): Change to fix_trunctfsi2_fprs.
2081 (fix_trunctfsi2): Call gen_spe_fix_trunctfsi2 or
2082 gen_fix_trunctfsi2_fprs depending on TARGET_E500_DOUBLE.
2083 (negtf2): Change to negtf2_internal.
2084 (negtf2): New expander.
2085 (abstf2): Enable for E500 double. Call gen_spe_abstf2_tst,
2086 gen_spe_abstf2_cmp or gen_abstf2_internal depending on
2087 TARGET_E500_DOUBLE and flag_unsafe_math_optimizations.
2088 (movdi_internal32): Use rs6000_nonimmediate_operand.
2089 (unnamed splitter): Likewise.
2090 * config/rs6000/spe.md (CMPTFEQ_GPR, TSTTFEQ_GPR, CMPTFGT_GPR,
2091 TSTTFGT_GPR, CMPTFLT_GPR, TSTTFLT_GPR): New unspecs.
2092 (SPE64TF, DITI): New mode macros.
2093 (frob_df_di): Change to frob_<SPE64:mode>_<DITI:mode>; allow more
2096 (frob_<mode>_di_2): New.
2097 (frob_tf_di_8_2): New.
2098 (frob_di_df): Change to frob_di_<mode>; allow more modes.
2100 (frob_di_df_2): Change to frob_<DITI:mode>_<SPE64:mode>_2; allow
2102 (frob_ti_<mode>_8_2): New.
2103 (frob_ti_tf_2): New.
2104 (mov_si<mode>_e500_subreg0, mov_si<mode>_e500_subreg0_2,
2105 mov_si<mode>_e500_subreg4, mov_si<mode>_e500_subreg4_2): Allow
2107 (mov_sitf_e500_subreg8, mov_sitf_e500_subreg8_2,
2108 mov_sitf_e500_subreg12, mov_sitf_e500_subreg12_2): New.
2109 (spe_trunctfdf2_internal1, spe_trunctfsf2, spe_extenddftf2,
2110 spe_fix_trunctfsi2, spe_fix_trunctfsi2_internal,
2111 spe_negtf2_internal, spe_abstf2_cmp, spe_abstf2_tst): New.
2112 (cmptfeq_gpr, tsttfeq_gpr, cmptfgt_gpr, tsttfgt_gpr, cmptflt_gpr,
2115 2007-01-23 Ian Lance Taylor <iant@google.com>
2117 * Makefile.in (OBJS-common): Reformat, alphabetize, but put
2119 (OBJS-archive): Reformat, alphabetize.
2120 (OBJS): Change out_object_file to OBJS-md.
2122 2007-01-23 Uros Bizjak <ubizjak@gmail.com>
2124 * config/i386/i386.md: Use REG_P, MEM_P, CONST_INT_P, LABEL_P,
2125 JUMP_P and CALL_P predicates where applicable.
2126 * config/i386/i386.c: Ditto.
2127 * config/i386/i386.md: Ditto.
2128 * config/i386/mmx.md: Ditto.
2129 * config/i386/predicates.md: Ditto.
2131 2007-01-22 Andreas Schwab <schwab@suse.de>
2133 * config/m68k/m68k.h: Fix comment.
2135 2007-01-22 Jan Hubicka <jh@suse.cz>
2137 * passes.c (init_optimization_passes): Do not rerun
2138 pass_early_warn_uninitialized.
2140 2007-01-22 Richard Guenther <rguenther@suse.de>
2142 PR tree-optimization/30038
2143 * tree-ssa-math-opts.c (maybe_record_sincos): New static helper
2145 (execute_cse_sincos_1): Likewise.
2146 (execute_cse_sincos): Likewise.
2147 (gate_cse_sincos): Likewise.
2148 (pass_cse_sincos): New pass CSEing sin() and cos() calls using
2149 the cexpi() canonicalization of sincos().
2150 * tree-pass.h (pass_cse_sincos): Declare.
2151 * passes.c (init_optimization_passes): New pass pas_cse_sincos.
2153 2007-01-21 Eric Botcazou <ebotcazou@libertysurf.fr>
2155 PR rtl-optimization/29329
2156 * combine.c (replaced_rhs_insn): Rename to i2mod.
2157 (replaced_rhs_value): Rename to i2mod_new_rhs.
2158 (i2mod_old_rhs): New global variable.
2159 (combine_instructions): Adjust for above change. Save a copy of
2160 the old RHS into i2mod_old_rhs when the contents of a REG_EQUAL
2161 note are substituted in the second instruction.
2162 (distribute_notes) <REG_DEAD>: Adjust for above change. Do not
2163 ditch the note if it pertains to the second eliminated register
2164 and this register is mentioned in i2mod_old_rhs.
2167 2006-09-12 Eric Botcazou <ebotcazou@libertysurf.fr>
2169 * combine.c (distribute_notes) <REG_DEAD>: Do not consider SETs past
2170 the insn to which the note was originally attached.
2172 2007-01-21 Jan Hubicka <jh@suse.cz>
2174 * ipa-inline.c (inlining_mode): Comment, move up.
2175 (cgraph_decide_inlining_incrementally): Do not perform inlining
2176 itself; fix handling of flattening of self recursive functions.
2177 (cgraph_find_cycles): Remove.
2178 (cgraph_flatten_node): Remove.
2179 (cgraph_decide_inlining): Use incremental inliner to handle flattening.
2180 (try_inline): New function.
2181 (cgraph_early_inlining): Update call of
2182 cgraph_decide_inlining_incrementally. Apply inlining here.
2183 (apply_inline): Update call of cgraph_decide_inlining_incrementally.
2185 2007-01-21 Dirk Mueller <dmueller@suse.de>
2188 * tree-vrp.c (check_array_bounds): do not warn
2189 about ADDR_EXPR's of ARRAY_REF's which are immediately
2190 used in binary expressions.
2192 2007-01-21 Ira Rosen <irar@il.ibm.com>
2194 * tree-vectorizer.h (struct _stmt_vec_info): Add new field
2195 read_write_dep and macros for its access.
2196 * tree-vectorizer.c (new_stmt_vec_info): Initialize the new field.
2197 * tree-vect-analyze.c (vect_analyze_data_ref_dependence): Remove
2198 argument, call vect_check_interleaving for every independent pair of
2199 data-refs. Mark loads that access the same memory location as a store
2201 (vect_check_dependences): Remove.
2202 (vect_analyze_data_ref_dependences): Remove vect_check_dependences
2203 call, fix the call to vect_analyze_data_ref_dependence.
2204 (vect_analyze_data_ref_access): For statements that access the same
2205 data-ref, check that they are not stores; for loads, check that there
2206 is no store that access the same location.
2208 2007-01-20 Roger Sayle <roger@eyesopen.com>
2209 Joseph Myers <joseph@codesourcery.com>
2211 * doc/invoke.texi (-fdump-rtl-combine): Fix under/overfull hbox.
2212 (-fdump-rtl-stack): Likewise.
2213 (-fno-signed-zeros): Use @minus{} for a minus sign. Correct typo.
2214 (-mcheck-zero-division, -mcpu): Fix under/overfull hbox.
2215 (-mpt-fixed): Use @minus{} for minus sign.
2216 (Using Precompiled Headers): Fix under/overfull hbox.
2218 2007-01-20 Jan Hubicka <jh@suse.cz>
2220 * tree-flow.h (struct stmt_ann_d): Move references_memory to proper
2221 place within annotation.
2223 2007-01-20 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
2225 * pa.c (output_move_double): Change array size of xoperands to 4.
2227 2007-01-20 Jan Hubicka <jh@suse.cz>
2229 * tree-tailcall.c (adjust_return_value): Do not use RESULT_DECL
2232 2007-01-19 Ian Lance Taylor <iant@google.com>
2234 * expmed.c (expand_divmod) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]: Cast
2235 constant to unsigned HOST_WIDE_INT before negating.
2237 2007-01-19 Ian Lance Taylor <iant@google.com>
2239 * tree-ssa-operands.h (struct vuse_vec_d): Change num_vuse field
2241 (VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Use unsigned comparison.
2242 (VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
2243 (SET_VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Likewise.
2244 (SET_VUSE_ELEMENT_VAR) [ENABLE_CHECKING]: Likewise.
2245 (SET_VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
2246 (realloc_vdef, realloc_vuse): Change second parameter to
2248 (ssa_operand_iterator_d): Change vuse_index and mayuse_index
2250 * tree-ssa-operands.c (realloc_vop): Change num_elem parameter to
2251 unsigned. Change x and lim locals to unsigned.
2252 (realloc_vdef, realloc_vuse): Change num_elem parameter to
2254 (finalize_ssa_vuse_ops): Change old_i local to unsigned.
2255 (copy_virtual_operands): Change i and n locals to unsigned.
2257 2007-01-19 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
2260 * builtins.c (fold_builtin_1): Handle builtin fdim.
2262 2007-01-20 Jan Hubicka <jh@suse.cz>
2264 * tree-ssa.c (init_tree_ssa): Do not call init_alias_heapvars.
2265 * tree-ssa-structalias.c (compute_points_to_sets): Do call
2266 init_alias_heapvars.
2267 (init_alias_heapvars): Initialize only when not already initialized.
2268 (delete_alias_heapvars): Set heapvar_for_stmt to NULL.
2270 2007-01-19 Roger Sayle <roger@eyesopen.com>
2272 * common.opt (fsigned-zeros): New command line option.
2273 * flags.h (HONOR_SIGNED_ZEROS): Control via flag_signed_zeros instead
2274 of flag_unsafe_math_optimizations.
2275 * opts.c (set_fast_math_flags): The -ffast-math command line option
2276 implies -fno-signed-zeros.
2277 (fast_math_flags_set_p): Likewise.
2279 * doc/invoke.texi: Document new -fno-signed-zeros option, and update
2280 the documentation of -ffast-math appropriately. Wrap long lines.
2282 2007-01-19 Steve Ellcey <sje@cup.hp.com>
2284 * system.h (ASM_MAKE_LABEL_LINKONCE): Poison.
2285 * varasm.c (globalize_decl): Remove ASM_MAKE_LABEL_LINKONCE ifdef.
2287 2007-01-19 Tomas Bily <tbily@suse.cz>
2289 * cgraphunit.c (cgraph_finalize_function): Updating of pid
2291 (tree_init_ic_make_global_vars): New function
2292 (tree_init_edge_profiler): call of tree_init_ic_make_global_vars
2293 (tree_gen_ic_profiler): New function
2294 (tree_gen_ic_func_profiler): New function
2295 (tree_profiling): Added calling of tree_gen_ic_func_profiler
2296 (tree_profile_hooks): Added hook for indirec/virtual calls
2297 * value-prof.c (tree_find_values_to_profile): New case for
2299 (tree_values_to_profile): Call for determining indirect/virtual
2301 (tree_indirect_call_to_profile): New function
2302 (tree_ic_transform): New function
2303 (tree_ic): New function
2304 (find_func_by_pid): New function
2305 (init_pid_map): New function
2306 (tree_value_profile_transformations): Added check for
2307 indirect/virtual call transformation
2308 * value-prof.h (enum hist_type): New counter type for
2309 indirect/virtual calls
2310 (profile_hooks): Added new hook for profiling indirect/virtual
2312 * profile.c (instrument_values): New case for indirect/virtual
2314 * gcov-io.h (GCOV_LAST_VALUE_COUNTER): Changed to 6
2315 (GCOV_COUNTER_V_INDIR): New counter type
2316 (GCOV_COUNTER_NAMES): New name of counter "indirect" added
2317 (GCOV_MERGE_FUNCTIONS): New merge function for indirect/virtual
2319 * cgraph.c: Definition of cgraph_max_pid
2320 (cgraph_create_node): Default init of pid attribute
2321 * cgraph.h: Declaration of cgraph_max_pid
2322 (struct cgraph_node): Added pid attribute
2323 * libgcov.c (__gcov_indirect_call_profiler): New function
2324 (__gcov_one_value_profiler_body): New function
2325 (__gcov_one_value_profiler): Body was moved to
2326 __gcov_one_value_profiler_body and calls it
2328 2007-01-19 Basile Starynkevitch <basile@starynkevitch.net>
2330 * doc/gty.texi (Options): Document the mark_hook option to GTY.
2331 * gengtype.c (write_types_data, write_func_for_structure,
2332 write_types, ggc_wtd, pch_wtd): Add skip_hooks to
2333 write_types_data, ggc_wtd, pch_wtd for processing mark_hook.
2334 (walk_type, write_func_for_structure): Generate the mark_hook if
2337 2007-01-19 Jan Hubicka <jh@suse.cz>
2339 * ipa-inline.c (cgraph_decide_inlining_incrementally): Instead of
2340 'early' argument take inlining mode argument specifying whether to
2341 inline for size/speeed or all functions; add support for flattening;
2343 (cgraph_early_inlining): Update call of decide_inlining_incrementally.
2345 2007-01-19 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
2348 * toplev.c (warn_deprecated_use): Use %qD instead of %qs to print
2349 the name of the declared identifier.
2351 2007-01-19 Dirk Mueller <dmueller@suse.de>
2353 * config/i386.h (CONDITIONAL_REGISTER_USAGE): Store
2354 result of PIC_OFFSET_TABLE_REGNUM in temporary variable to avoid
2355 duplicate evaluation.
2357 2007-01-19 Uros Bizjak <ubizjak@gmail.com>
2359 * config/i386/i386.md (acos<mode>2): Rename from acossf2 and acosdf2.
2360 Macroize expander using X87MODEF12 mode macro. Extend operand 1
2361 to XFMode, use acosxf2 and truncate result to requested mode.
2362 Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
2363 (asin<mode>2): Similarly, with asin expanders.
2364 (*fscalexf4_i387): Rename from *fscalexf4.
2365 (expNcorexf3): New expander.
2366 (expxf2, exp10xf2, exp2xf2): Use expNcorexf3 expander.
2367 (exp<mode>2): Rename from expsf2 and expdf2. Macroize expander using
2368 X87MODEF12 mode macro. Extend operand 1 to XFMode, use expxf2 and
2369 truncate result to requested mode. Use SSE_FLOAT_MODE_P to disable
2370 patterns for SSE math.
2371 (exp10<mode>2): Similarly, with exp10 expanders.
2372 (exp2<mode>2): Similarly, with exp2 expanders.
2373 (expm1<mode>2): Similarly, with expm1 expanders.
2374 (ldexp<mode>3): Similarly, with ldexp expanders.
2375 (log<mode>2, log10<mode>2, log2<mode>2, log1p<mode>2, logb<mode>2):
2376 Use gen_truncxf<mode>2_i387_noop to truncate result.
2378 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2380 * config/m68k/m68k.h (PREFERRED_STACK_BOUNDARY): Define to 32
2381 for ColdFire targets.
2383 2007-01-19 Nathan Sidwell <nathan@codesourcery.com>
2384 Richard Sandiford <richard@codesourcery.com>
2386 * config/m68k/m68k.h (M68K_STATIC_CHAIN_REG_NAME): New macro.
2387 (INITIALIZE_TRAMPOLINE): Use STATIC_CHAIN_REGNUM.
2388 (__transfer_from_trampoline): Use M68K_STATIC_CHAIN_REG_NAME.
2389 * config/m68k/m68kelf.h (STATIC_CHAIN_REGNUM): Override.
2390 (M68K_STATIC_CHAIN_REG_NAME): Likewise.
2391 * config/m68k/netbsd-elf.h (M68K_STATIC_CHAIN_REG_NAME): Likewise.
2393 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2395 * config/m68k/m68k.md (adddi_dilshr32): Rename to...
2396 (*adddi_dilshr32): ...this. Fix formatting. Remove commented-out
2397 non-canonical pattern. Restrict to !TARGET_COLDFIRE.
2398 (*adddi_dilshr32_cf): New pattern.
2399 (adddi3, subdi3): Remove first alternatives.
2401 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2403 * config/m68k/m68k.c (notice_update_cc): If an SFmode move is
2404 implemented using move.l, do not use its cc result for floating-point
2407 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2409 * config/m68k/m68k.h (EXTRA_CONSTRAINT): Stop the 'T' constraint
2410 from accepting 's' constraints if flag_pic.
2412 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2414 * config/m68k/m68k.md (bordered, bunordered, buneq, bunge, bungt)
2415 (bunle, bunlt, bltgt, bordered_rev, bunordered_rev, buneq_rev)
2416 (bunge_rev, bungt_rev, bunle_rev, bunlt_rev, bltgt_rev): Change
2417 condition from TARGET_68881 to TARGET_HARD_FLOAT.
2419 2007-01-19 Sandra Loosemore <sandra@codesourcery.com>
2421 * longlong.h (count_leading_zeros, COUNT_LEADING_ZEROS_0): Add
2422 ColdFire alternatives.
2423 * config/m68k/m68k.h (CLZ_DEFINED_VALUE_AT_ZERO): New macro.
2424 * config/m68k/m68k.md (clzsi2): Define for ColdFire
2425 architectures that support ff1 instruction.
2427 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2428 Julian Brown <julian@codesourcery.com>
2430 * config/m68k/m68k.h (CONST_OK_FOR_LETTER_P): Add an 'R' case.
2431 * config/m68k/m68k.md (*movsi_cfv4): Fold into...
2432 (*movsi_cf): ...here. Remove unnecessary 'R' from 'Rg'.
2435 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2437 * config/m68k/m68k-protos.h (valid_mov3q_const): Take a HOST_WIDE_INT
2439 (output_move_const_into_data_reg, output_move_simode_const): Delete.
2440 * config/m68k/m68k.c (const_method, const_int_cost): Take a
2441 HOST_WIDE_INT instead of an rtx.
2442 (m68k_rtx_costs): Update call accordingly.
2443 (output_move_const_into_data_reg): Likewise. Fix formatting.
2444 (valid_mov3q_const): Take a HOST_WIDE_INT instead of an rtx.
2446 (output_move_simode_const): Update calls after above changes.
2447 Rework to use automatic variables and predicates like MEM_P.
2448 * config/m68k/m68k.md (pushexthisi_const): Update call to
2451 2007-01-19 Dirk Mueller <dmueller@suse.de>
2453 * tree-ssa-alias.c (perform_var_substitution): Fix typo
2456 2007-01-19 Richard Guenther <rguenther@suse.de>
2458 * builtins.c (expand_builtin_cexpi): Fall back to expanding
2459 via cexp in case sincos is not available.
2461 2007-01-19 Richard Guenther <rguenther@suse.de>
2463 * doc/tm.texi (TARGET_HAS_SINCOS): Document new target macro.
2464 * defaults.h (TARGET_HAS_SINCOS): Default to off.
2465 * config/linux.h (TARGET_HAS_SINCOS): Set to on if we have glibc.
2466 * config/alpha/linux.h (TARGET_HAS_SINCOS): Likewise.
2467 * config/sparc/linux.h (TARGET_HAS_SINCOS): Likewise.
2468 * config/sparc/linux64.h (TARGET_HAS_SINCOS): Likewise.
2469 * config/rs6000/linux.h (TARGET_HAS_SINCOS): Likewise.
2470 * config/rs6000/linux64.h (TARGET_HAS_SINCOS): Likewise.
2472 2007-01-19 Uros Bizjak <ubizjak@gmail.com>
2474 * config/i386/i386.md (*fpatanxf3_i387, fpatan_extend<mode>xf3_i387):
2476 (atan2sf3_1, atan2df3_1, atan2xf3_1): Remove insn patterns.
2477 (atan2xf3): Directly generate RTL pattern.
2478 (atan2<mode>3): Rename from atan2sf3 and atan2df3 and macroize insn
2479 patterns using X87MODEF12 mode macro. Use fpatan_extend<mode>xf3_i387
2480 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
2481 disable patterns for SSE math.
2482 (atan<mode>2): Rename from atansf2 and atandf2 and macroize insn
2483 patterns using X87MODEF12 mode macro. Use fpatan_extend<mode>xf3_i387
2484 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
2485 disable patterns for SSE math.
2487 2007-01-19 Alexandre Oliva <aoliva@redhat.com>
2489 * libgcc-std.ver: Fix typo in %inherit for GCC_4.3.0.
2491 2007-01-18 Roger Sayle <roger@eyesopen.com>
2493 * fold-const.c (fold_unary) <VIEW_CONVERT_EXPR>: Optimize away a
2494 VIEW_CONVERT_EXPR to the same type as it's operand.
2496 2007-01-18 David Edelsohn <edelsohn@gnu.org>
2498 * config/rs6000/darwin-ldouble.c: Only build _SOFT_FLOAT if
2499 configured for long double 128.
2501 2007-01-18 Mike Stump <mrs@apple.com>
2503 * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Fix build
2506 2007-01-18 Michael Meissner <michael.meissner@amd.com>
2508 * i386.c (ix86_compute_frame_layout): Make fprintf's in #if 0 code
2511 2007-01-18 Jan Hubicka <jh@suse.cz>
2513 * tree-ssa-operands.c (vop_free_bucket_size): Never return value
2514 greater than NUM_VOP_FREE_BUCKETS.
2516 2007-01-18 Daniel Berlin <dberlin@dberlin.org>
2518 * tree-ssa-structalias.c: Update comments.
2519 (ptabitmap_obstack): Removed.
2521 (oldpta_obstack): Ditto.
2522 (stats): Add a few members.
2523 (struct variable_info): Remove node, complex, address_taken, and
2524 indirect_target members. Add oldsolution member.
2525 (new_var_info): Do not initialize removed members.
2526 (constraint_expr_type): Remove INCLUDES.
2527 (constraint_graph): Add size, implicit_preds, rep,
2528 indirect_cycles, eq_rep, label, direct_nodes, and complex members.
2529 (FIRST_REF_NODE): New macro.
2530 (LAST_REF_NODE): Ditto.
2531 (FIRST_ADDR_NODE): Ditto.
2532 (find): New function.
2534 (dump_constraint): Do not handle INCLUDES.
2535 (insert_into_complex): Do not insert duplicate constraints.
2536 (condense_varmap_nodes): Renamed and rewritten into ...
2537 (merge_node_constraints): This. Also fix bug in handling of
2538 offseted copy constraints.
2539 (clear_edges_for_node): No longer need to deal with preds at all,
2540 or removing associated preds/succs.
2541 (merge_graph_nodes): Deal with indirect_cycles.
2542 Don't deal with predecessors.
2543 (add_implicit_graph_edge): New function.
2544 (add_pred_graph_edge): Ditto.
2545 (add_graph_edge): Don't deal with predecessors.
2546 (build_constraint_graph): Removed.
2547 (build_pred_graph): New function.
2548 (build_succ_graph): Ditto.
2549 (struct scc_info): Removed in_component. Added roots, dfs, and
2550 node_mapping. Remove visited_index, unification_queue.
2551 (scc_visit): Deal with union-find we do now.
2552 Deal with cycles with REF nodes.
2553 (collapse_nodes): Renamed and rewritten to ...
2554 (unify_nodes): This.
2555 (process_unification_queue): Removed.
2556 (topo_visit): Cleanup
2557 (do_da_constraint): Use find.
2558 (do_sd_constraint): Ditto.
2559 (do_ds_constraint): Ditto.
2560 (do_complex_constraint): Ditto.
2561 (init_scc_info): Update for removed and added members.
2562 (find_and_collapse_graph_cycles): Renamed and rewritten into ...
2563 (find_indirect_cycles): This.
2564 (equivalence_class): New variable.
2565 (label_visit): New function.
2566 (perform_variable_substitution): Rewritten.
2567 (free_var_substitution_info): New function.
2568 (find_equivalent_node): Ditto.
2569 (move_complex_constraints): Ditto.
2570 (eliminate_indirect_cycles): Ditto.
2571 (solve_graph): Only propagate changed bits.
2572 Use indirect cycle elimination.
2574 (tree_id_t): Rename to tree_vi_t, delete id member, add vi member.
2575 (tree_id_eq): Renamed to ...
2576 (tree_vi_eq): This. Update for member change
2577 (insert_id_for_tree): Renamed and rewritten to ...
2578 (insert_vi_for_tree): This.
2579 (lookup_id_for_tree): Renamed and rewritten to ...
2580 (lookup_vi_for_tree): This.
2581 (get_id_for_tree): Renamed and rewritten to ...
2582 (get_vi_for_tree): Ditto.
2583 (get_constraint_exp_from_ssa_var): Update to use get_vi_for_tree.
2584 (process_constraint): Don't handle INCLUDES.
2585 Remove special ADDRESSOF case.
2586 (find_func_aliases): Rewrite to use vi functions instead of id
2588 (create_function_info_for): Ditto.
2589 (create_variable_info_for): Ditto.
2590 (intra_create_variable_infos): Ditto.
2591 (merge_smts_into): Ditto.
2592 (find_what_p_points_to): Ditto.
2593 (init_base_vars): Ditto.
2594 (init_alias_vars): Ditto.
2595 (remove_preds_and_fake_succs): New function.
2596 (dump_sa_points_to_info): Dump new stats.
2597 (dump_solution_for_var): Use find.
2598 (set_used_smts): Fix formatting.
2599 (compute_points_to_sets): Updated for new functions.
2600 (ipa_pta_execute): Ditto.
2602 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
2603 Richard Sandiford <richard@codesourcery.com>
2605 * doc/tm.texi (TARGET_FUNCTION_VALUE): Expand documentation of
2607 * calls.c (expand_call): If the return value is a PARALLEL,
2608 extract its first member.
2609 * config/m68k/linux.h (FUNCTION_EXTRA_EPILOGUE): Remove.
2610 * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
2611 use FUNCTION_EXTRA_EPILOGUE.
2612 (m68k_function_value): Return a PARALLEL if the return value
2613 is of a pointer type.
2614 * config/m68k/netbsd-elf.h (current_function_returns_pointer)
2615 (FUNCTION_EXTRA_EPILOGUE): Remove.
2616 * config/m68k/m68k.md (D0_REG): New constant.
2618 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
2620 * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
2621 output a NOP for empty epilogues.
2623 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2625 * config/m68k/m68k.c (m68k_use_return_insn): Update comments
2626 before function. Extend register save check to include all
2627 registers, not just integer ones.
2629 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
2631 * config/m68k/m68k-protos.h (use_return_insn): Rename to...
2632 (m68k_use_return_insn): ...this.
2633 * config/m68k/m68k.h (USE_RETURN_INSN): Delete.
2634 * config/m68k/m68k.c (use_return_insn): Rename to...
2635 (m68k_use_return_insn): ...this.
2636 * config/m68k/m68k.md (return): Use m68k_use_return_insn instead
2639 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
2641 * config/m68k/fpgnulib.c (__truncdfsf2): Implement round to
2642 nearest even, fix denormal rounding overflow.
2644 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2646 * config/m68k/m68k.md (movsf_cf_hard): Use fsmove instead of
2647 f%$smove and f%$move.
2648 (movdf_cf_hard): Use fdmove for cases 0 and 3 and fmove for case 1.
2649 (extendsfdf2_cf): Use fdmove instead of f%&move.
2650 (truncdfsf2_cf): Use fsmove instead of f%$smove.
2651 (add<mode>3_cf, sub<mode>3_cf): Use <FP:prec> instead of <FP:round>.
2653 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2655 * config/m68k/m68k.md (movdf_cf_hard): Use output_move_double for
2658 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2660 * real.h (real_format): Add a canonical_nan_lsbs_set field.
2661 (coldfire_single_format): Declare.
2662 (coldfire_double_format): Likewise.
2663 * real.c (encode_ieee_single): Use canonical_nan_lsbs_set instead
2664 of qnan_msb_set to determine the lower bits of a canonical
2666 (encode_ieee_double): Likewise.
2667 (encode_ieee_quad): Likewise.
2668 (ieee_single_format): Initialize canonical_nan_lsbs_set.
2669 (mips_single_format): Likewise.
2670 (ieee_double_format): Likewise.
2671 (mips_double_format): Likewise.
2672 (ieee_extended_motorola_format): Likewise.
2673 (ieee_extended_intel_96_format): Likewise.
2674 (ieee_extended_intel_128_format): Likewise.
2675 (ieee_extended_intel_96_round_53_format): Likewise.
2676 (ibm_extended_format): Likewise.
2677 (mips_extended_format): Likewise.
2678 (ieee_quad_format): Likewise.
2679 (mips_quad_format): Likewise.
2680 (vax_f_format): Likewise.
2681 (vax_d_format): Likewise.
2682 (vax_g_format): Likewise.
2683 (i370_single_format): Likewise.
2684 (i370_double_format): Likewise.
2685 (decimal_single_format): Likewise.
2686 (decimal_double_format): Likewise.
2687 (decimal_quad_format): Likewise.
2688 (c4x_single_format): Likewise.
2689 (c4x_extended_format): Likewise.
2690 (real_internal_format): Likewise.
2691 (coldfire_single_format): New real_format.
2692 (coldfire_double_format): Likewise.
2693 * config/pdp11/pdp11.c (pdp11_f_format): Initialize
2694 canonical_nan_lsbs_set.
2695 (pdp11_d_format): Likewise.
2696 * config/m68k/m68k.c (override_options): Override REAL_FORMAT_MODE
2697 if TARGET_COLDFIRE_CPU.
2699 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2701 * config/m68k/m68k-protos.h (m68k_output_pic_call): Delete.
2702 (output_call, m68k_legitimize_call_address): Declare.
2703 * config/m68k/m68k.h (EXTRA_CONSTRAINT): Remove unnecessary
2704 parenthesees. Add support for a 'W' constraint.
2705 (LEGITIMATE_PIC_OPERAND_P): Remove SYMBOL_REF_FLAG handling.
2706 (PRINT_OPERAND_PUNCT_VALID_P): Remove comment about 'o'.
2707 (m68k_symbolic_call, m68k_symbolic_jump): Declare.
2708 * config/m68k/m68k.c (m68k_symbolic_call, m68k_symbolic_jump): New
2710 (override_options): Initialize them. Do not set flag_no_function_cse
2711 for TARGET_ID_SHARED_LIBRARY.
2712 (m68k_output_pic_call): Delete.
2713 (m68k_legitimize_call_address): New function.
2714 (print_operand): Remove the %o prefix. Handle the %p prefix.
2715 (output_call): New function.
2716 (m68k_output_mi_thunk): Use m68k_symbolic_jump. Always load the
2717 target address from the GOT if symbolic jumps are not allowed.
2718 * config/m68k/m68k.md (call, general_operand): Do not set
2719 SYMBOL_REF_FLAG. Use m68k_legitimize_call_address instead.
2720 Merge separate flag_pic and !flag_pic define_insns into...
2721 (*call, *call_value): ...these new patterns. Match the address
2722 rather than the containing MEM and require it to be a call_operand.
2723 Use output_call to generate the asm template.
2724 * config/m68k/predicates.md (const_call_operand): New predicate.
2725 (call_operand): Likewise.
2727 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
2729 * config/m68k/m68k.h (REGISTER_MOVE_COST): Simplify definition.
2730 (STACK_GROWS_DOWNWARD): Define to 1.
2731 (FUNCTION_VALUE, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P): Equivocate
2732 comments, emphasizing that these values are only defaults.
2733 * config/m68k/linux.h (LINK_SPEC): Fix formatting in #undef.
2734 * config/m68k/m68k.c (const_method): Remove trailing whitespace.
2736 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2738 * config/m68k/m68k.md (cmpsi): Remove outdated flag_pic handling.
2740 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
2741 Richard Sandiford <richard@codesourcery.com>
2743 * config/m68k/m68k.h (DATA_REGNO_P, ADDRESS_REGNO_P, INT_REGNO_P)
2744 (FP_REGNO_P): New macros.
2745 (REGNO_OK_FOR_INDEX_P, REGNO_OK_FOR_BASE_P, REGNO_OK_FOR_DATA_P)
2746 (REGNO_OK_FOR_FP_P, REG_OK_FOR_INDEX_P, REG_OK_FOR_BASE_P): Use them.
2747 (EH_RETURN_STACKADJ_RTX): Use A0_REG.
2748 * config/m68k/m68k.c (m68k_regno_mode_ok): Use the new REGNO macros.
2750 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
2752 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
2753 (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-rtems*): Add t-floatlib
2755 * config/m68k/t-floatlib: New file, extracting common code from...
2756 * config/m68k/t-m68kbare, config/m68k/t-m68kelf,
2757 * config/m68k/t-uclinux: Here.
2758 * config/m68k/fpgnulib.c: Do not compile extendeed precision
2759 routines on ColdFire targets.
2761 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
2763 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
2764 (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*)
2765 (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-linux*)
2766 (m68k-*-rtems*): Use tm_file rather than m68k/m68k.h and
2767 explicitly set MOTOROLA to 1.
2768 * config/m68k/m68k.h (MOTOROLA): Simplify definition accordingly.
2770 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2771 Nathan Sidwell <nathan@codesourcery.com>
2773 * config/m68k/m68k.h (PCC_STATIC_STRUCT_RETURN): Delete.
2774 (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Add MOTOROLA cases.
2775 * config/m68k/coff.h (REGISTER_PREFIX_MD): Delete.
2776 * config/m68k/m68020-elf.h (LIB_SPEC): Delete.
2777 * config/m68k/m68k-none.h (CC1_SPEC, CPP_SUBTARGET_SPEC): Delete.
2778 * config/m68k/m68kelf.h (IMMEDIATE_PREFIX, REGISTER_PREFIX_MD)
2779 (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Delete.
2780 (NO_DOLLAR_IN_LABEL, PCC_STATIC_STRUCT_RETURN): Don't undefine.
2781 (BSS_ASM_OP, ASM_OUTPUT_SKIP, ASM_OUTPUT_ASCII): Delete.
2782 * config/m68k/m68kemb.h (PCC_STATIC_STRUCT_RETURN): Don't undefine.
2783 * config/m68k/linux.h (SIZE_TYPE, PTRDIFF_TYPE, WCHAR_TYPE)
2784 (WCHAR_TYPE_SIZE, TARGET_OBJFMT_CPP_BUILTINS): Delete.
2785 (TARGET_OS_CPP_BUILTINS): Don't define mc68000 and mc68020 here.
2786 (DBX_REGISTER_NUMBER): Delete.
2787 * config/m68k/netbsd-elf.h (IMMEDIATE_PREFIX): Delete.
2788 (PCC_STATIC_STRUCT_RETURN): Don't undefine.
2789 * config/m68k/openbsd.h (PCC_STATIC_STRUCT_RETURN): Define.
2791 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2793 * config.gcc (m68k-*-uclinux*): Add flat.h to $tm_file.
2794 * config/flat.h: New file.
2795 * crtstuff.c (USE_PT_GNU_EH_FRAME): Don't define if
2797 * config/m68k/m68k.h (ASM_PREFERRED_EH_DATA_FORMAT): Do not use
2798 indirect references for -msep-data or -mid-shared-library.
2799 Do not use PC-relative code addresses either.
2800 * config/m68k/m68k.c (override_options): Restrict -fPIC error
2802 * config/m68k/uclinux.h (STARTFILE_SPEC): Define. Use Scrt1.o
2803 for shared libraries and crt1.o for executables. Use crti.o and
2805 (ENDFILE_SPEC): Use crtend.o and crtn.o.
2806 (LIB_SPEC): Suppress -Rlibc.gdb if -static-libc is given.
2807 Do not add -elf2flt or -shared-lib-id options here.
2808 (LINK_SPEC): Define. Pass -elf2flt if no -elf2flt option is given.
2809 Pass -shared-lib-id if -mid-shared-library, taking the library
2810 identifier from -mshared-library-id if given, otherwise
2812 (EH_FRAME_IN_DATA_SECTION): Do not undefine.
2813 (INIT_SECTION_ASM_OP, FINI_SECTION_ASM_OP): Likewise.
2814 (TARGET_OS_CPP_BUILTINS): Define __GXX_MERGED_TYPEINFO_NAMES=0
2815 and __GXX_TYPEINFO_EQUALITY_INLINE=0 if -mid-shared-library.
2816 (DRIVER_SELF_SPECS): Map unadorned PIC options to -msep-data.
2817 * config/m68k/t-uclinux (EXTRA_MULTILIB_PARTS): Add crtbegin.o
2819 * config/m68k/lb1sf68.asm (PICCALL): Use an lea and pc-relative
2820 jump sequence for ISA A and ISA A+.
2821 (PICJUMP): Likewise.
2823 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2825 * config.gcc (m68k-*-uclinux*): Base the port on the common
2826 and m68k GNU/Linux files rather than on the generic ELF ones.
2827 * config/m68k/uclinux.h (TARGET_VERSION): Override.
2828 (TARGET_OS_CPP_BUILTINS): Use LINUX_TARGET_OS_CPP_BUILTINS.
2830 2007-01-18 Julian Brown <julian@codesourcery.com>
2831 Richard Sandiford <richard@codesourcery.com>
2833 * config/m68k/m68k.h (LONG_DOUBLE_TYPE_SIZE): Make 64-bit on ColdFire.
2834 (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
2835 * config/m68k/netbsd-elf.h (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Undefine
2837 * config/m68k/uclinux-oldabi.h (LONG_DOUBLE_TYPE_SIZE): Redefine to
2839 (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
2841 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2843 * doc/install.texi: Document m68k-uclinuxoldabi.
2844 * config.gcc (m68k-*-uclinuxoldabi): New configuration.
2845 * config/m68k/uclinux-oldabi.h: New file, copied from
2846 config/m68k/uclinux.h.
2848 2007-01-18 Kaz Kojima <kkojima@rr.iij4u.or.jp>
2850 * config/m32r/m32r-protos.h (m32r_expand_epilogue): Declare it.
2851 * config/m32r/m32r.c (m32r_setup_incoming_varargs): Use gen_frame_mem.
2852 (m32r_compute_frame_size): Use unsigned for regno.
2853 (m32r_reload_lr): Use gen_frame_mem.
2855 (m32r_output_function_epilogue): Don't output the function epilogue
2857 (m32r_expand_epilogue): New.
2858 (direct_return): Return false if the function has the interrupt
2860 (m32r_hard_regno_rename_ok): Remove code for the textual epilogue.
2861 * config/m32r/m32r.md (epilogue): New expander.
2862 (return_lr, return_rte): New insns.
2863 (return): Make it expander.
2864 (return_normal): New expander.
2866 2007-01-18 Josh Conner <jconner@apple.com>
2869 * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Add
2870 support for UNLE, UNLT, UNGE, and UNGT.
2872 2007-01-18 Jan Hubicka <jh@suse.cz>
2874 * tree-vrp.c (finalize_jump_threads): Do not call cleanup_cfg by hand.
2875 * tree-tailcall (add_virtual_phis): Likewise.
2876 (optimize_tail_call): Return TODOs.
2877 (execute_tail_calls): Return TODOs.
2878 * tree-ssa-ccp (execute_fold_all_builtins): Do cleanup_cfg via TODO.
2879 * tree-cfgcleanup.c (cleanup_tree_cfg_loop): Return if something
2881 * tree-ssa-forwprop.c (tree_ssa_forward_propagate_single_use_value):
2882 Cleanup cfg using TODO.
2883 * tree-flow.h (cleanup_tree_cfg_loop): Update prototype.
2884 * passes.c (execute_function_todo): When cleanup did something, remove
2886 * tree-cfg.c (pass_build_cfg): Add cleanup_cfg TODO.
2887 (make_edges): Don't cleanup_cfg.
2889 2007-01-18 Uros Bizjak <ubizjak@gmail.com>
2891 * reg-stack.c (subst_stack_regs_pat) [UNSPEC_SINCOS_COS,
2892 UNSPEC_XTRACT_FRACT]: Use generic code for instructions that
2893 operate on the top of stack.
2894 [UNSPEC_SINCOS_SIN, UNSPEC_XTRACT_EXP, UNSPEC_TAN]: Rewrite
2895 register handling of instructions that output to the second
2897 [UNSPEC_TAN_ONE, UNSPEC_TAN_TAN]: Remove.
2898 (move_for_stack_reg): Special-case check for dead destination
2899 stack slot for constant load of 1.0 inside UNSPEC_TAN.
2901 * config/i386/i386.md (UNSPEC_TAN): New constant.
2902 (UNSPEC_TAN_ONE, UNSPEC_TAN_TAN): Remove.
2903 (fptanxf4_i387, fptan_extend<mode>xf4_i387): New patterns
2904 to correctly model move of constant 1.0 to top stack slot.
2905 (*tandf3_1, *tansf3_1, *tanxf3_1): Remove insn patterns.
2906 (unnamed peephole2 pattern): Remove corresponding peephole2
2907 pattern that optimizes tan insn and loading of constant 1.0.
2908 (tanxf2): Use fptanxf4_i387.
2909 (tan<mode>2): Rename from tansf2 and tandf2 and macroize insn
2910 patterns using X87MODEF12 mode macro. Use fptan_extend<mode>xf4_i387
2911 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
2912 disable patterns for SSE math.
2913 (sincos<mode>3): Use truncxf<mode>2_i387_noop for truncation.
2914 (fyl2x_extend<mode>xf3_i387): Use X87MODEF12 for operand 1.
2916 2007-01-18 Dirk Mueller <dmueller@suse.de>
2917 Richard Guenther <rguenther@suse.de>
2920 * doc/invoke.texi (Warray-bounds): Document -Warray-bounds.
2921 * common.opt (Warray-bounds): Add new warning option.
2922 * c-opts.c (c_common_handle_option): Define -Warray-bounds
2924 * Makefile.in: make tree-vrp.o depend on toplev.h
2925 * tree-vrp.c (vrp_finalize): Call check_array_refs if -Warray-bounds
2927 (check_array_refs, check_array_bounds, check_array_ref): New.
2929 2007-01-18 Jan Hubicka <jh@suse.cz>
2931 * tree-ssa-ccp.c (ccp_finalize): Return if something changed.
2932 (execute_ssa_ccp): Return flags conditionally.
2933 * tree-ssa-propagate.c (substitue_and_fold): Return if something was
2935 * tree-ssa-propagate.h (substitute_and_fold): Update prototype.
2937 2007-01-18 Steven Bosscher <steven@gcc.gnu.org>
2939 * cfgcleanup.c (cleanup_cfg): Detect cfglayout mode and set
2940 the CLEANUP_CFGLAYOUT flag when in cfglayout mode.
2942 * Makefile.c (GTFILES): Add cfglayout.h.
2943 * gengtype.c (open_base_files): Likewise.
2944 * cfglayout.c (cfg_layout_function_footer,
2945 cfg_layout_function_header) Reindent to make gengtype happy.
2946 * cfglayout.h (cfg_layout_function_footer,
2947 cfg_layout_function_header): Add GTY(()) marker.
2949 * ifcvt.c (noce_try_sign_mask): Make sure INSN_B is non-null.
2951 2007-01-18 Ben Elliston <bje@au.ibm.com>
2953 * genautomata.c (write_automata): Include xstrerror output in the
2954 error message if writing the DFA description file fails.
2956 2007-01-17 H.J. Lu <hongjiu.lu@intel.com>
2958 * config/mips/mips-protos.h (mips_output_external): Make it
2960 * config/mips/iris.h (TARGET_ASM_EXTERNAL_LIBCALL): Removed.
2961 * config/mips/mips.c (irix_output_external_libcall): Likewise.
2962 (extern_list): Likewise.
2963 (extern_head): Likewise.
2964 (TARGET_ASM_FILE_END): Likewise.
2965 (mips_file_end): Likewise.
2966 (mips_output_external): Rewritten.
2968 2007-01-18 Ben Elliston <bje@au.ibm.com>
2970 * genpreds.c (write_insn_preds_c): Only write out the function
2971 body for regclass_for_constraint if we have register constraints.
2973 2007-01-17 Tom Tromey <tromey@redhat.com>
2975 * doc/sourcebuild.texi (libgcj Tests): Use sourceware.org.
2976 * doc/install.texi (Testing): Use sourceware.org.
2977 (Binaries): Likewise.
2978 (Specific): Likewise.
2979 * doc/contrib.texi (Contributors): Use sourceware.org.
2981 2007-01-17 Anatoly Sokolov <aesok@post.ru>
2983 * config/avr/avr.h (AVR_HAVE_LPMX): New macro.
2984 (AVR_ENHANCED): Rename to ...
2985 (AVR_HAVE_MUL): ... new.
2986 (avr_enhanced_p): Rename to ...
2987 (avr_have_mul_p): ... new.
2988 (TARGET_CPU_CPP_BUILTINS): Use 'avr_have_mul_p' instead of
2989 'avr_enhanced_p' for "__AVR_ENHANCED__". Define "__AVR_HAVE_MUL__".
2990 * config/avr/avr.c (avr_enhanced_p): Rename to ...
2991 (avr_have_mul_p): ... new.
2992 (base_arch_s): Rename 'enhanced' to 'have_mul'.
2993 (avr_override_options): Use 'avr_have_mul_p' and 'have_mul' instead of
2994 'avr_enhanced_p' and 'enhanced'.
2995 (ashlhi3_out, ashrhi3_out, lshrhi3_out, avr_rtx_costs): Use
2996 AVR_HAVE_MUL instead of AVR_ENHANCED.
2997 * avr.md (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
2998 (mulqi3, *mulqi3_enh, *mulqi3_call, mulqihi3, umulqihi3, mulhi3,
2999 *mulhi3_enh, *mulhi3_call, mulsi3, *mulsi3_call): Use AVR_HAVE_MUL
3000 instead of AVR_ENHANCED.
3001 (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
3002 * libgcc.S: Use __AVR_HAVE_MUL__ instead of __AVR_ENHANCED__.
3003 (__tablejump__): Use __AVR_HAVE_LPMX__ instead of __AVR_ENHANCED__.
3005 2007-01-17 Ian Lance Taylor <iant@google.com>
3007 * vec.h (VEC_reserve_exact): Define.
3008 (vec_gc_p_reserve_exact): Declare.
3009 (vec_gc_o_reserve_exact): Declare.
3010 (vec_heap_p_reserve_exact): Declare.
3011 (vec_heap_o_reserve_exact): Declare.
3012 (VEC_OP (T,A,reserve_exact)): New static inline function, three
3014 (VEC_OP (T,A,reserve)) [all versions]: Remove handling of
3016 (VEC_OP (T,A,alloc)) [all versions]: Call ...reserve_exact.
3017 (VEC_OP (T,A,copy)) [all versions]: Likewise.
3018 (VEC_OP (T,a,safe_grow)) [all versions]: Likewise.
3019 * vec.c (calculate_allocation): Add exact parameter. Change all
3021 (vec_gc_o_reserve_1): New static function, from vec_gc_o_reserve.
3022 (vec_gc_p_reserve, vec_gc_o_reserve): Call vec_gc_o_reserve_1.
3023 (vec_gc_p_reserve_exact, vec_gc_o_reserve_exact): New functions.
3024 (vec_heap_o_reserve_1): New static function, from vec_heap_o_reserve.
3025 (vec_heap_p_reserve, vec_heap_o_reserve): Call vec_heap_o_reserve_1.
3026 (vec_heap_p_reserve_exact): New function.
3027 (vec_heap_o_reserve_exact): New function.
3029 2007-01-17 Jan Hubicka <jh@suse.cz>
3031 * ipa-type-escape.c (look_for_casts): Revamp using handled_component_p.
3033 2007-01-17 Eric Christopher <echristo@apple.com>
3035 * config.gcc: Support core2 processor.
3037 2007-01-16 Jan Hubicka <jh@suse.cz>
3039 * tree-ssanames.c (release_dead_ssa_names): Instead of ggc_freeing
3040 the names, just unlink the chain so we don't crash on dangling pointers
3043 2007-01-16 Jan Hubicka <jh@suse.cz>
3045 * cgraph.h (cgraph_decide_inlining_incrementally): Kill.
3046 * tree-pass.h: Reorder to make IPA passes appear toegher.
3047 (pass_early_inline, pass_inline_parameters, pass_apply_inline): Declare.
3048 * cgraphunit.c (cgraph_finalize_function): Do not compute inling
3049 parameters, do not call early inliner.
3050 * ipa-inline.c: Update comments. Include tree-flow.h
3051 (cgraph_decide_inlining): Do not compute inlining parameters.
3052 (cgraph_decide_inlining_incrementally): Return TODOs; assume to
3053 be called with function context set up.
3054 (pass_ipa_inline): Remove unreachable functions before pass.
3055 (cgraph_early_inlining): Simplify assuming to be called from the
3057 (pass_early_inline): New pass.
3058 (cgraph_gate_ipa_early_inlining): New gate.
3059 (pass_ipa_early_inline): Turn into simple wrapper.
3060 (compute_inline_parameters): New function.
3061 (gate_inline_passes): New gate.
3062 (pass_inline_parameters): New pass.
3063 (apply_inline): Move here from tree-optimize.c
3064 (pass_apply_inline): New pass.
3065 * ipa.c (cgraph_remove_unreachable_nodes): Verify cgraph after
3067 * tree-inline.c (optimize_inline_calls): Return TODOs rather than
3069 (tree_function_versioning): Do not allocate dummy struct function.
3070 * tree-inline.h (optimize_inline_calls): Update prototype.
3071 * tree-optimize.c (execute_fixup_cfg): Export.
3072 (pass_fixup_cfg): Remove
3073 (tree_rest_of_compilation): Do not apply inlines.
3074 * tree-flow.h (execute_fixup_cfg): Declare.
3075 * Makefile.in (gt-passes.c): New.
3076 * passes.c: Include gt-passes.h
3077 (init_optimization_passes): New passes.
3078 (nnodes, order): New static vars.
3079 (do_per_function_toporder): New function.
3080 (execute_one_pass): Dump current pass here.
3081 (execute_ipa_pass_list): Don't dump current pass here.
3083 2007-01-16 Janis Johnson <janis187@us.ibm.com>
3085 * config/dfp-bit.c (dfp_compare_op): Return separate value for NaN.
3086 (DFP_NE, DFP_LE, DFP_GE): Return false for NaN.
3088 2007-01-16 David Edelsohn <edelsohn@gnu.org>
3090 * config/rs6000/darwin-ldouble.c: Build file for SOFT_FLOAT.
3091 (strong_alias): Define.
3092 (__gcc_qmul): Provide non-FMA for soft-float.
3098 (__gcc_qunord): New.
3108 * config/rs6000/rs6000.c (rs6000_init_libfuncs): Initialize
3109 soft-float functions.
3110 * config/rs6000/libgcc-ppc-glibc.ver: Version soft-float symbols.
3111 * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Do not warn
3112 about long double soft float.
3114 2007-01-16 Dorit Nuzman <dorit@il.ibm.com>
3115 Tehila Meyzels <tehila@il.ibm.com>
3117 * tree-vectorizer.h (is_pattern_stmt_p): New.
3118 * tree-vect-analyze.c (vect_determine_vectorization_factor): Fix
3119 formatting (tabs instead of spaces). Cleanup and clarify setting
3120 of STMT_VINFO_VECTYPE. Call is_pattern_stmt_p.
3121 * tree-vect-transform.c (vect_get_vec_def_for_operand): Fix typo.
3122 (vectorizable_type_demotion): Check that types are integral.
3123 (vectorizable_type_promotion): Likewise.
3124 (vectorizable_store): Fix typo. Eliminate new-line at end of
3127 2007-01-16 Jan Hubicka <jh@suse.cz>
3129 * tree-ssanames.c (release_dead_ssa_names): Remove invalidated
3132 2007-01-15 Eric Christopher <echristo@apple.com>
3134 * ifcvt.c: Include vec.h, vecprim.h.
3135 (check_cond_move_block): New argument regs.
3136 Reorganize. Add registers used to regs.
3137 (cond_move_process_if_block): Use regs set above as
3140 2007-01-15 Eric Christopher <echristo@apple.com>
3142 * config/darwin.h: Update copyright.
3143 (TARGET_OPTION_TRANSLATE_TABLE): Add umbrella.
3144 (LINK_COMMAND_SPEC): Add -u.
3145 (LINK_SPEC): Fix umbrella for above.
3147 2007-01-15 Joseph S. Myers <joseph@codesourcery.com>
3149 * config/soft-fp/op-common.h, config/soft-fp/op-4.h: Update from
3152 2007-01-15 Tom Tromey <tromey@redhat.com>
3154 * doc/sourcebuild.texi (libgcj Tests): Don't mention jacks.
3155 * doc/install.texi (Testing): Don't mention jacks.
3156 (Configuration): Document --enable-java-maintainer-mode. Move
3157 --with-java-home to libgcj-specific section. Document
3159 (Prerequisites): Mention --enable-java-maintainer-mode, ecj1.
3161 2007-01-15 Jan Hubicka <jh@suse.cz>
3163 * tree-ssa-dce.c (DCE_TODOs): New.
3164 (propagate_necessity): Return if something changed.
3165 (eliminate_unnecessary_stmts): Likewise.
3166 (perform_tree_ssa_dce): Return TODO flags when needed.
3167 (pass_dce, pass_dce_loop, pass_cd_dce): Remove TODO flags.
3169 2007-01-15 Uros Bizjak <ubizjak@gmail.com>
3171 * config/i386/i386.md (fyl2xxf3_i387): Rename from fyl2x_xf3.
3172 (fyl2x_extend<mode>xf3_i387): New insn pattern.
3173 (log<mode>2): Rename from logsf2 and logdf2 and macroize insn
3174 patterns using X87MODEF12 mode macro. Extend operand 1
3175 to XFmode. Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
3176 (log10<mode>2): Ditto.
3177 (log2<mode>2): Ditto.
3178 (log1p<mode>2): Ditto.
3179 (logb<mode>2): Ditto.
3180 (fyl2xp1xf3_i387): Rename from fyl2xp1_xf3.
3181 (fyl2xp1_extend<mode>xf3_i387): New insn pattern.
3182 (*fxtractxf3_i387): Rename from *fxtractxf3.
3183 (fxtract_extend<mode>xf3_i387): New insn pattern.
3184 (ilogbsi2): Use match_dup 3, not match_operand:XF 3.
3185 * config/i386/i386.c (ix86_emit_i387_log1p): Use gen_fyl2xp1xf3_i387()
3186 and gen_fyl2xxf3_i387().
3188 2007-01-14 Zdenek Dvorak <dvorakz@suse.cz>
3190 * loop-unswitch.c (unswitch_loop): Do not call fix_loop_placement.
3191 * cfgloopmanip.c (fix_loop_placement): Made static. Use
3192 get_loop_exit_edges. Changed return type to bool.
3193 * cfgloop.h (fix_loop_placement): Declaration removed.
3195 2007-01-14 Dorit Nuzman <dorit@il.ibm.com>
3197 * param.h (MIN_VECT_LOOP_BOUND): New.
3198 * params.def (MIN_VECT_LOOP_BOUND): New.
3199 * tree-vectorizer.c (slpeel_tree_peel_loop_to_edge): Takes another
3200 argument - minimum threshold for number of iterations.
3201 * tree-vectorizer.h (slpeel_tree_peel_loop_to_edge): Add another
3202 argument to declaration.
3203 * tree-vect-analyze.c (vect_analyze_operations): Check value of
3204 MIN_VECT_LOOP_BOUND.
3205 * tree-vect-transform.c (vect_do_peeling_for_loop_bound): Call
3206 slpeel_tree_peel_loop_to_edge with additional argument.
3207 (vect_do_peeling_for_alignment): Likewise.
3208 * doc/invoke.texi (min-vect-loop-bound): Document new param option.
3210 2007-01-14 Uros Bizjak <ubizjak@gmail.com>
3213 * config/i386/i386.c (print_operand) ['z']: Output 'b' for
3216 2007-01-14 Jan Hubicka <jh@suse.cz>
3218 * tree-dfa.c (remove_referenced_var): New function.
3219 * tree-ssa-live.c (remove_unused_locals): Walk referenced vars and
3220 prune referenced vars list too.
3221 * tree-flow.h (remove_referenced_var): Declare.
3223 2007-01-14 Jan Hubicka <jh@suse.cz>
3225 * tree-eh.c (add_stmt_to_eh_region_fn): Do not add call_exprs
3227 (remove_stmt_from_eh_region_fn): Do not remove call_exprs.
3228 (verify_eh_throw_stmt_node, verify_eh_throw_table_statements): Kill.
3229 * except.h (verify_eh_throw_table_statements): Kill prototype.
3230 * cfgexpand.c (expand_gimple_basic_block): Propagate Eh regions
3231 into call exrepssions.
3232 * tree-optimize.c (execute_free_cfg_annotatiosn): Do not call
3234 * tree-cfg.c: Include pointer-set.h.
3235 (verify_node_sharing): Work on pointer set.
3236 (verify_eh_throw_stmt_node): New.
3237 (verify_stmts): Use pointers sets, verify throw_stmt.
3239 2007-01-13 Zdenek Dvorak <dvorakz@suse.cz>
3241 * ipa-reference.c (analyze_function): Consider also addresses taken
3244 2007-01-12 Roger Sayle <roger@eyesopen.com>
3246 * c-typeck.c (null_pointer_constant_p): Replace use of
3247 TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
3248 (build_c_cast): Likewise.
3250 2007-01-12 Roger Sayle <roger@eyesopen.com>
3252 * tree.h (force_fit_type_double): Remove unused final argument.
3253 * c-common.c (constant_expression_warning): Replace use of
3254 TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
3255 (convert_and_check): Likewise.
3256 (shorten_compare): Update call to force_fit_type_double.
3257 (c_common_truthvalue_conversion) <INTEGER_CST>: Use integer_zerop.
3258 * convert.c (convert_to_pointer): Update call to
3259 force_fit_type_double.
3260 * fold-const.c (force_fit_type_double): Remove overflowed_const
3262 (int_const_binop, fold_convert_const_int_from_int,
3263 fold_convert_const_int_from_real, fold_div_compare,
3264 fold_sign_changed_comparison, fold_unary, fold_negate_const,
3265 fold_abs_const, fold_not_const): Remove the final argument from
3266 calls to force_fit_type_double.
3268 2007-01-12 Andrew Pinski <andrew_pinski@playstation.sony.com>
3270 * configure.ac: Set insn to "nop" for spu-*-* also.
3271 * configure: Regenerate.
3273 2007-01-12 Olga Golovanevsky <olga@il.ibm.com>
3275 * builtins.def : Add BUILT_IN_FREE.
3277 2007-01-12 Jan Hubicka <jh@suse.cz>
3279 PR tree-optimization/30443
3280 * tree-inline.c (tree_function_versioning): Do not optimize when
3281 cloning for inlining.
3283 2007-01-12 Zdenek Dvorak <dvorakz@suse.cz>
3285 * doc/loop.texi: Document recording of loop exits.
3286 * cfgloopmanip.c (loopify, duplicate_loop): Use alloc_loop.
3287 (update_single_exits_after_duplication,
3288 update_single_exit_for_duplicated_loop,
3289 update_single_exit_for_duplicated_loops): Removed.
3290 (duplicate_loop_to_header_edge): Do not call
3291 update_single_exits_after_duplication and
3292 update_single_exit_for_duplicated_loops.
3293 (loop_version): Do not update single_exit information.
3294 (fix_loop_structure): Use record_loop_exits instead of
3295 mark_single_exit_loops.
3296 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update
3297 the lists of loop exits.
3298 * cfghooks.c (redirect_edge_and_branch, redirect_edge_and_branch_force,
3299 split_edge, merge_blocks): Update the lists of loop exits.
3300 * modulo-sched.c (sms_schedule): Pass LOOPS_HAVE_RECORDED_EXITS to
3301 loop_optimizer_init.
3302 * loop-init.c (loop_optimizer_init): Call record_loop_exits instead
3303 of mark_single_exit_loops.
3304 (loop_optimizer_finalize): Call release_recorded_exits.
3305 * tree-ssa-loop.c (tree_loop_optimizer_init): Pass
3306 LOOPS_HAVE_RECORDED_EXITS to loop_optimizer_init.
3307 * tree-vectorizer.c (slpeel_tree_duplicate_loop_to_edge_cfg): Do not
3308 update single exit information.
3309 * lambda-code.c (perfect_nestify): Ditto.
3310 * cfgloop.c (flow_loop_free): Destroy the list of exits of the loop.
3311 (mark_single_exit_loops): Removed.
3312 (alloc_loop, loop_exit_hash, loop_exit_eq, loop_exit_free,
3313 get_exit_descriptions, rescan_loop_exit, record_loop_exits,
3314 dump_recorded_exit, dump_recorded_exits, release_recorded_exits): New
3316 (get_loop_exit_edges, single_exit): Use recorded exit lists.
3317 (add_bb_to_loop, remove_bb_from_loops): Update the lists of loop exits.
3318 (verify_loop_structure): Verify consistency of the exit lists.
3319 (flow_loops_find): Use alloc_loop. Initialize exits hash.
3320 (set_single_exit): Removed.
3321 * cfgloop.h (struct loop_exit): New function.
3322 (struct loop): single_exit_ field replaced by exits field.
3323 (LOOPS_HAVE_MARKED_SINGLE_EXITS): Replaced by LOOPS_HAVE_RECORDED_EXITS.
3324 (struct loops): Added exits hash.
3325 (mark_single_exit_loops, set_single_exit): Declaration removed.
3326 (release_recorded_exits, record_loop_exits, rescan_loop_exit): Declare.
3328 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3330 * doc/invoke.texi: Avoid use of @headitem.
3332 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3334 * cse.c (cse_insn): Move HAVE_CC0 code after declarations.
3336 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3338 * doc/install.texi: Fix m68k-*-* anchor and add m68k-*-* to the
3341 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
3342 Richard Sandiford <richard@codesourcery.com>
3344 * doc/invoke.texi: Document -mno-bitfield, -mno-rtd and -mno-short.
3345 * config/m68k/m68k.opt: Resort options.
3346 (mbitfield, mrtd, mshort): Remove RejectNegative properties.
3348 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
3349 Richard Sandiford <richard@codesourcery.com>
3351 * doc/invoke.texi: Document the macros that are defined by
3352 m68k's -mtune and -mhard-float options.
3353 * config/m68k/m68k-protos.h (m68k_cpp_cpu_ident) Declare.
3354 (m68k_cpp_cpu_family): Likewise.
3355 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Add a full set
3356 of __ucfv*__ macros. Define __mcffpu__ if generating code for
3357 ColdFire FPUs. Define __mcf_cpu_* and __mcf_family_* macros.
3358 * config/m68k/m68k.c (m68k_cpp_cpu_ident): New function.
3359 (m68k_cpp_cpu_family): Likewise.
3361 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3363 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Treat all mc68*
3364 macros besides mc68000 as tuning macros. Use a switch statement
3365 to set them and mcpu32.
3367 2007-01-12 Julian Brown <julian@codesourcery.com>
3369 * config/m68k/m68k.h: Use TARGET_68040 instead of TARGET_68040_ONLY.
3370 (TARGET_68040_ONLY): Rename to...
3371 (TARGET_68040): ...this.
3372 * config/m68k/m68k.c: Use TARGET_68040 instead of TARGET_68040_ONLY.
3373 * config/m68k/m68k.md: Likewise.
3375 2007-01-12 Julian Brown <julian@codesourcery.com>
3376 Nathan Sidwell <nathan@codesourcery.com>
3377 Richard Sandiford <richard@codesourcery.com>
3379 * config.gcc (m680[012]0-*-*, m68k*-*-*): Set m68k_cpu_ident to
3380 the -mcpu= argument associated with the --with-cpu setting.
3381 Define M68K_DEFAULT_TUNE to the default -mtune= option,
3382 if different from the one implied by the -mcpu setting.
3383 Accept --with-cpu=FOO if FOO is listed in m68k-devices.def,
3384 using mcpu=FOO as the default CPU option. Set target_cpu_default2.
3385 * doc/invoke.texi: Mention ColdFire in the introduction to the
3386 m68k options. Document the new -march, -mcpu, -mtune, -mdiv,
3387 -mno-div and -mhard-float options. Make -m68881 a synonym for
3388 -mhard-float. Document the previously-undocumented -m5206e,
3389 -m528x, -m5307 and -m5407 options. Tweak the existing option
3390 documentation for consistency.
3391 * doc/install.texi: Mention new --with-cpu arguments.
3392 * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Only use the
3393 default CPU if neither -mcpu nor -march are specified.
3394 (ASM_CPU_SPEC): Pass down -mcpu and -march options.
3395 (TARGET_CPU_CPP_BUILTINS): Set __mcfisa*__ macros from
3396 TARGET_ISA*. Set the legacy __mcf*__ cpu macros in the same way,
3397 using m68k_tune to decide between families that implement the
3398 same ISA. Use m68k_tune to set __mcfv4e__.
3399 (FL_BITFIELD, FL_68881, FL_COLDFIRE, FL_CF_HWDIV, FL_CF_MAC)
3400 (FL_CF_EMAC, FL_CF_EMAC_B, FL_CF_USP, FL_CF_FPU, FL_ISA_68000)
3401 (FL_ISA_68010, FL_ISA_68020, FL_ISA_68040, FL_ISA_A, FL_ISA_B)
3402 (FL_ISA_C, FL_ISA_MMU): New macros.
3403 (MASK_COLDFIRE): Delete.
3404 (TARGET_68010, TARGET_68020, TARGET_68040_ONLY, TARGET_COLDFIRE)
3405 (TARGET_ISAB): Redefine in terms of m68k_cpu_flags.
3406 (TARGET_68881, TARGET_COLDFIRE_FPU): Redefine in terms of m68k_fpu.
3407 (TARGET_HARD_FLOAT): Do not define here.
3408 (TARGET_ISAAPLUS, TARGET_ISAC): New macros.
3409 (TUNE_68000): New macro.
3410 (TUNE_68000_10): Redefine in terms of TUNE_68000 and TUNE_68010.
3411 (TUNE_68010, TUNE_68030, TUNE_68040, TUNE_68060, TUNE_CPU32)
3412 (TUNE_CFV2): Redefine in terms of m68k_tune.
3413 (uarch_type, target_device, fpu_type): New enums.
3414 (m68k_cpu, m68k_tune, m68k_fpu, m68k_cpu_flags): Declare.
3415 * config/m68k/m68k.c (TARGET_DEFAULT): Remove MASK_68881.
3416 (FL_FOR_isa_00, FL_FOR_isa_10, FL_FOR_isa_20, FL_FOR_isa_40)
3417 (FL_FOR_isa_cpu32, FL_FOR_isa_a, FL_FOR_isa_aplus, FL_FOR_isa_b)
3418 (FL_FOR_isa_c): New macros.
3419 (m68k_isa): New enum.
3420 (m68k_target_selection): New structure.
3421 (all_devices, all_isas, all_microarchs): New tables.
3422 (m68k_cpu_entry, m68k_arch_entry, m68k_tune_entry, m68k_cpu)
3423 (m68k_tune, m68k_fpu, m68k_cpu_flags): New variables.
3424 (MASK_ALL_CPU_BITS): Delete.
3425 (m68k_find_selection): New function.
3426 (m68k_handle_option): Handle -mcpu=, -march= and -mtune=.
3427 Map the legacy target options to a combination of the new ones.
3428 (override_options): Set m68k_cpu, m68k_tune, m68k_fpu and
3429 m68k_cpu_flags. Handle M68K_DEFAULT_TUNE. Use m68k_cpu_flags
3430 to derive default MASK_BITFIELD, MASK_CF_HWDIV and MASK_HARD_FLOAT
3432 * config/m68k/m68k.opt (m5200, m5206e, m528x, m5307, m5407, mcfv4e)
3433 (m68010, m68020, m68020-40, m68020-60, m68030, m68040): Remove Mask
3435 (m68881, msoft-float): Change mask from 68881 to HARD_FLOAT.
3436 (march=, mcpu=, mdiv, mhard-float, mtune=): New options.
3437 * config/m68k/m68k-devices.def: New file.
3439 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3440 Nathan Sidwell <nathan@codesourcery.com>
3442 * config/m68k/m68k.h (ASM_CPU_SPEC, ASM_SPEC, EXTRA_SPECS)
3443 (SUBTARGET_EXTRA_SPECS): New macros.
3444 * config/m68k/linux.h (ASM_SPEC): Remove CPU flags;
3445 use %(asm_cpu_spec) instead.
3446 * config/m68k/m68k-none.h (ASM_SPEC): Likewise.
3447 * config/m68k/openbsd.h (ASM_SPEC): Likewise.
3448 * config/m68k/netbsd-elf.h (ASM_SPEC): Likewise.
3449 (EXTRA_SPECS): Rename to...
3450 (SUBTARGET_EXTRA_SPECS): ...this.
3452 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
3453 Richard Sandiford <richard@codesourcery.com>
3454 Julian Brown <julian@codesourcery.com>
3456 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
3457 (m68k-*-uclinux*, m68k-*-linux*, m68k-*-rtems*): Set default_m68k_cpu
3458 to the configuration's default CPU.
3459 (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*): Likewise.
3460 Remove default masks.
3461 (m680[012]0-*-*): Set the default with_cpu to the first part of
3463 (m68k*-*-*): Set the default with_cpu to m$default_m68k_cpu.
3464 (m68k*-*-linux): Extend the --with-cpu handling to...
3465 (m680[012]0-*-*, m68k*-*-*): ...these configurations. Allow m68000
3466 and m68010. Don't set target_cpu_default2.
3467 * doc/install.texi: Document --with-cpu for m68k.
3468 * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Define.
3469 * config/m68k/m68k-none.h (TARGET_CPU_DEFAULT, M68K_CPU_m68k)
3470 (M68K_CPU_m68000, M68K_CPU_m68010, M68K_CPU_m68020, M68K_CPU_m68030)
3471 (M68K_CPU_m68040, M68K_CPU_m68302, M68K_CPU_m68332, TARGET_DEFAULT)
3472 (ASM_CPU_DEFAULT_SPEC, CC1_CPU_DEFAULT_SPEC): Delete.
3473 (ASM_SPEC): Remove use of %(asm_cpu_default).
3474 (EXTRA_SPECS, SUBTARGET_EXTRA_SPECS, MULTILIB_DEFAULTS): Delete.
3475 * config/m68k/linux.h (TARGET_DEFAULT): Delete.
3476 (CPP_SPEC): Merge definitions. Do not handle __HAVE_68881__ here.
3477 * config/m68k/netbsd-elf.h (TARGET_OS_CPP_BUILTINS): Define
3478 __HAVE_FPU__ if TARGET_HARD_FLOAT.
3479 (TARGET_DEFAULT): Delete.
3480 (EXTRA_SPECS): Delete cpp_cpu_default_spec, cpp_cpu_spec,
3481 cpp_fpu_spec, asm_default_spec and netbsd_cpp_spec.
3482 (CPP_CPU_SPEC): Delete.
3483 (TARGET_VERSION): Merge definitions, using TARGET_68010 to pick
3484 the appropriate string.
3485 (CPP_CPU_DEFAULT_SPEC, ASM_DEFAULT_SPEC, CPP_FPU_SPEC): Delete.
3486 (CPP_SPEC): Define to NETBSD_CPP_SPEC.
3487 (ASM_SPEC): Don't use %(asm_default_spec).
3488 * config/m68k/m68k.c (TARGET_DEFAULT_TARGET_FLAGS): Remove
3489 TARGET_DEFAULT and add MASK_68881.
3490 * config/m68k/m68k.md: Remove mention of TARGET_DEFAULT from comments.
3492 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3494 * config.gcc (m68010-*-netbsdelf*): Add MASK_68010.
3495 (m68k*-*-netbsdelf*, m68k*-*-openbsd*, m68k*-linux*): Add
3496 MASK_68010 alongside MASK_68020.
3497 * doc/invoke.texi: Document -m68010.
3498 * config/m68k/m68k.opt (m68010): New.
3499 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Define mc68010
3501 (TUNE_68010): New macro.
3502 * config/m68k/m68k-none.h (M68K_CPU_m68k, M68K_CPU_m68010)
3503 (M68K_CPU_m68020, M68K_CPU_m68030, M68K_CPU_m68040)
3504 (M68K_CPU_m68332): Add MASK_68010.
3505 * config/m68k/linux.h (TARGET_DEFAULT): Add MASK_68010 to
3506 fallback definition.
3507 * config/m68k/netbsd-elf.h (CPP_CPU_SPEC): Remove now-redundant
3509 * config/m68k/m68k.c (MASK_ALL_CPU_BITS): Add MASK_68010.
3510 (m68k_handle_option): Handle OPT_m68010. Add MASK_68010
3511 to all entries that use MASK_68020.
3512 (output_move_simode_const, output_move_himode, output_move_qimode)
3513 (output_move_stricthi, output_move_strictqi): Use TARGET_68010
3514 instead of TARGET_68020 to select clr behavior. Remove comment
3515 about there being no TARGET_68010.
3516 * config/m68k/m68k.md: Likewise throughout.
3518 2007-01-12 Julian Brown <julian@codesourcery.com>
3520 * config/m68k/m68k.h (TARGET_ISAB): New macro.
3521 * config/m68k/m68k.c: Use TARGET_ISAB rather than TARGET_CFV4.
3522 * config/m68k/m68k.md: Likewise.
3524 2007-01-12 Julian Brown <julian@codesourcery.com>
3526 * config/m68k/m68k.h (LEGITIMATE_INDEX_P, LEGITIMIZE_ADDRESS): Use
3527 TARGET_COLDFIRE_FPU instead of TARGET_CFV4E.
3529 2007-01-12 Julian Brown <julian@codesourcery.com>
3531 * config/m68k/m68k.h (TUNE_68040_60): New macro.
3532 * config/m68k/m68k.c (standard_68881_constant_p): Use it.
3533 * config/m68k/m68k.md: Likewise.
3535 2007-01-12 Julian Brown <julian@codesourcery.com>
3536 Richard Sandiford <richard@codesourcery.com>
3538 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Use TUNE_68030
3539 instead of TARGET_68030, TUNE_68040 instead of TARGET_68040,
3540 TUNE_68060 instead of TARGET_68060 and TUNE_CPU32 instead of
3542 (TARGET_CPU32): Rename to...
3543 (TUNE_CPU32): ...this.
3544 (TUNE_68000_10, TUNE_68030, TUNE_68040, TUNE_68060)
3545 (TUNE_CFV2): New macros.
3546 * config/m68k/netbsd-elf.h (LONG_DOUBLE_TYPE_SIZE): Simplify;
3547 remove conditions that are implied by TARGET_68020.
3548 * config/m68k/m68k.c (m68k_output_function_prologue): Use TUNE_68040
3549 instead of TARGET_68040 and TUNE_CPU32 instead of TARGET_CPU32.
3550 (m68k_output_function_epilogue): Likewise.
3551 (m68k_rtx_costs): Likewise. Use TUNE_68060 instead of TARGET_68060
3552 and TUNE_CFV2 instead of TARGET_5200. Use TUNE_68000_10 instead of
3553 "!TARGET_68020 && !TARGET_COLDFIRE" to choose between 68000 and
3554 non-68000 timings. Refactor multiplication and division costs.
3555 (output_addsi3): Use TUNE_68040 instead of TARGET_68040 and
3556 TUNE_CPU32 instead of TARGET_CPU32.
3557 (standard_68881_constant_p): Use TUNE_68040 instead of TARGET_68040
3558 and TUNE_68060 instead of TARGET_68060.
3559 * config/m68k/m68k.md: Use TUNE_68040 instead of TARGET_68040,
3560 TUNE_68060 instead of TARGET_68060, and TUNE_CPU32 instead of
3562 (movsi_const0): Use TUNE_68000_10 rather than "!TARGET_68020
3563 && !TARGET_COLDFIRE" to choose between moveq and clr.
3564 Likewise in the unnamed movsf pattern.
3565 (ashlsi_17_24, lshrsi_17_24): Guard with TUNE_68000_10 rather than
3566 "!TARGET_68020 && !TARGET_COLDFIRE". Likewise the unnamed
3569 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3571 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Increase amount
3572 of tabbing before backslashes.
3574 2007-01-11 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
3576 * pa-linux.h (ASM_OUTPUT_INTERNAL_LABEL): Undefine.
3577 * pa.h (ASM_OUTPUT_LABEL): Output colon when using GAS.
3578 (ASM_OUTPUT_INTERNAL_LABEL): Define.
3580 2007-01-11 Zdenek Dvorak <dvorakz@suse.cz>
3582 * tree-ssa-loop-ivopts.c (extract_cond_operands): Split from
3583 find_interesting_uses_cond.
3584 (find_interesting_uses_cond): Use extract_cond_operands.
3585 (rewrite_use_compare): Use extract_cond_operands and
3586 force_gimple_operand_bsi. Do not call update_stmt.
3587 (determine_use_iv_cost_condition): Use extract_cond_operands.
3588 Return cheaper of using original bound and changing the exit bound.
3590 2007-01-11 Zdenek Dvorak <dvorakz@suse.cz>
3592 PR tree-optimization/29516
3593 * tree-ssa-address.c (tree_mem_ref_addr, add_to_parts,
3594 most_expensive_mult_to_index, addr_to_parts,
3595 create_mem_ref, maybe_fold_tmr): Make the type of
3596 fields of TARGET_MEM_REF sizetype.
3597 (move_fixed_address_to_symbol, move_pointer_to_base):
3599 * tree.def (TARGET_MEM_REF): Add comment on types of
3602 2007-01-11 Joseph Myers <joseph@codesourcery.com>
3604 * c-common.c (vector_types_convertible_p): Treat opaque types as
3605 always convertible if they have the same size, but not otherwise.
3607 2007-01-11 Steven Bosscher <steven@gcc.gnu.org>
3609 * ifcvt.c (struct noce_if_info): Add comments to the fields.
3610 Remove the b_unconditional field.
3611 (noce_try_sign_mask): Do not look at b_unconditional.
3612 (noce_process_if_block): Do not use merge_if_blocks. Update
3613 the CFG here. Do not set b_unconditional.
3614 (cond_move_process_if_block): Likewise.
3615 (find_cond_trap): Likewise.
3616 (check_cond_move_block): Require simple jump insns at the end
3619 2007-01-11 Jan Hubicka <jh@suse.cz>
3621 PR tree-optimization/1046
3622 * tree-tailcall.c (suitable_for_tail_call_opt_p): Use TREE_ADDRESSABLE
3623 when alias info is not ready.
3624 (pass_tail_recursion): Do not require aliasing.
3625 * tree-ssa-copyrename.c (pass_rename_ssa_cop): Likewise.
3626 * tree-ssa-ccp.c (pass_ccp, pass_fold_builtins): Likewise.
3627 * tree-ssa-copy.c (pass_copy_prop): Likewise.
3628 * tree-ssa-forwprop.c (pass_forwprop): Likewise.
3629 * tree-ssa-dce.c (pass_dce, pass_dce_loop, pass_cd_dce): Likewise.
3630 * passes.c (init_optimization_passes): Execute rename_ssa_copies,
3631 ccp, forwprop, copy_prop, merge_phi, copy_prop, dce and tail recursion
3633 * tree-ssa-operands.c (add_virtual_operand, get_indirect_ref_operand):
3634 When aliasing is not build, mark statement as volatile.
3636 2007-01-11 Tom Tromey <tromey@redhat.com>
3638 PR preprocessor/15185, PR preprocessor/20989:
3639 * doc/cppopts.texi <-MT>: Update description of algorithm for
3640 computing default target.
3641 <-M, -MD>: Reword "basename" text.
3643 2007-01-11 Roger Sayle <roger@eyesopen.com>
3645 * builtins.c (expand_builtin_pow, expand_builtin_powi,
3646 fold_builtin_cabs, fold_builtin_sqrt, fold_builtin_trunc,
3647 fold_builtin_floor, fold_builtin_ceil, fold_builtin_round,
3648 fold_builtin_int_int_roundingfn, fold_builtin_bitop,
3649 fold_builtin_bswap, real_constp, fold_builtin_pow,
3650 fold_builtin_powi, fold_builtin_signbit, fold_builtin_copysign,
3651 do_mpfr_arg1, do_mpfr_arg2, do_mpfr_arg3, do_mpfr_sincos): Replace
3652 uses of the macro TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
3653 * convert.c (convert_to_pointer): Likewise.
3654 * expr.c (highest_pow2_factor, expand_expr_real_1): Likewise.
3655 * fold-const.c (force_fit_type, fold_negate_expr, int_const_binop,
3656 const_binop, fold_convert_const_int_from_int,
3657 fold_convert_const_int_from_real,
3658 fold_convert_const_real_from_real, sign_bit_p,
3659 optimize_minmax_comparison, extract_muldiv_1, fold_div_compare,
3660 fold_sign_changed_comparison, fold_unary, fold_comparison,
3661 fold_binary, multiple_of_p, tree_Expr_non_zero_p,
3662 fold_negate_const, fold_abs_const, fold_not_const): Likewise.
3663 * print-tree.c (print_node_brief, print_node): Likewise.
3664 * stor-layout.c (place_field, layout_type): Likewise.
3665 * tree-chrec.c (keep_cast): Likewise.
3666 * tree.c (build_vector, build_real, build_real_from_int_cst,
3667 build_complex): Likewise.
3669 2007-01-11 Roger Sayle <roger@eyesopen.com>
3671 * tree.h (TREE_CONSTANT_OVERFLOW): Obsolete. For the time being,
3672 treat TREE_CONSTANT_OVERFLOW as a synonym of TREE_OVERFLOW.
3674 2007-01-11 Paolo Bonzini <bonzini@gnu.org>
3676 * configure.ac (strict1_warn): Rename to strict_warn.
3677 (WERROR, --enable-werror, symlink hacks, stage1_cflags,
3678 cc_set_by_configure, quoted_cc_set_by_configure,
3679 stage_prefix_set_by_configure, quoted_stage_prefix_set_by_configure,
3680 all_boot_languages, all_stagestuff): Remove.
3681 (target_list): Remove bootstrap targets.
3682 * Makefile.in (quickstrap): Unconditionally make a synonym of all.
3683 (BOOT_LANGUAGES, STAGE1_CFLAGS, STAGE1_CHECKING,
3684 REMAKEFLAGS, FLAGS_TO_PASS, PREPEND_DOTDOT_TO_RELATIVE_PATHS,
3685 SUBDIR_FLAGS_TO_PASS, WERROR_FLAGS, STRICT2_WARN, LANG_STAGESTUFF,
3686 VOL_FILES, POSTSTAGE1_FLAGS_TO_PASS, STAGE2_FLAGS_TO_PASS,
3687 STAGEPROFILE_FLAGS_TO_PASS, STAGEFEEDBACK_FLAGS_TO_PASS, stage1_build,
3688 stage1_copy, stage2_build, stage2_copy, stageprofile_build,
3689 stageprofile_copy, stage3_build, stage3_copy, stagefeedback_build,
3690 stagefeedback_copy, stage4_build, clean_s1, clean_sw, bootstrap,
3691 bootstrap-lean, bootstrap2, bootstrap2-lean, bootstrap3,
3692 bootstrap3-lean, bootstrap4, bootstrap4-lean, unstage1, unstage2,
3693 unstage3, unstage4, unstageprofile, unstagefeedback, restage, restage2,
3694 restage3, restage4, restageprofile, restagefeedback, bubbleestrap,
3695 cleanstrap, unstrap, restrap, *compare, *compare3, *compare4,
3696 *compare-lean, *compare3-lean, *compare4-lean, stage1-start, stage1,
3697 stage2-start, stage2, stage3-start, stage3, stage4-start, stage4,
3698 stageprofile-start, stageprofile, stagefeedback-start, stagefeedback,
3699 risky-stage1, risky-stage2, risky-stage3, risky-stage4): Remove.
3700 (ORDINARY_FLAGS_TO_PASS): Rename to FLAGS_TO_PASS.
3701 (STAGECOPYSTUFF, STAGEMOVESTUFF): Consolidate into MOSTLYCLEANFILES.
3702 (mostlyclean): Adjust.
3703 (clean, distclean): Don't mention bootstrap stuff.
3704 * configure: Regenerate.
3705 * ada/config-lang.in, cp/config-lang.in, forttran/config-lang.in,
3706 java/config-lang.in, objc/config-lang.in, objcp/config-lang.in,
3707 treelang/config-lang.in (stagestuff): Remove.
3708 * doc/sourcebuild.texi (stage1, stage2, stage3, stage4,
3709 stageprofile, stagefeedback, stagestuff): Remove mention.
3711 2007-01-11 Nick Clifton <nickc@redhat.com>
3713 * config/mcore/predicates.md (mcore_general_movesrc_operand):
3715 (mcore_general_movdst_operand): Do not accept CONST_INTs.
3716 (mcore_arith_K_S_operand): Run the test for the S constraint not
3717 the test for the M constraint.
3718 (mcore_addsub_operand): Do not accept integer values that are
3719 larger than 32 bits.
3720 * config/mcore/mcore.md: Remove unused constraints from split.
3721 (andsi3): Use HOST_WIDE_INT instead of int to hold an INTVAL.
3723 (allocate_stack): Likewise.
3724 * config/mcore/mcore.c (mcore_print_operand): Restrict output of P
3725 operands to 32 bits.
3726 (mcore_const_costs): Use HOST_WIDE_INT instead of int to hold an
3728 (mcore_and_cost, mcore_modify_comparison, const_ok_for_mcore,
3729 mcore_const_ok_for_inline, mcore_const_trick_uses_not,
3730 try_constant_tricks, mcore_num_ones, mcore_num_zeros,
3731 mcore_output_bclri, mcore_output_andn, output_inline_const,
3732 mcore_output_move, mcore_output_movedouble): Likewise.
3733 (mcore_output_cmov): Use CONST_OK_FOR_M and CONST_OK_FOR_N.
3734 (output_inline_const): Likewise.
3735 (output_inline_const): Fix format strings used in sprintf
3737 * config/mcore/mcore-protos.h: Update prototypes for changed
3738 functions in mcore.c.
3739 * config/mcore/mcore.h (CONST_OK_FOR_I): Cast values to
3740 HOST_WIDE_INT and not int.
3741 (CONST_OK_FOR_J, CONST_OK_FOR_K, CONST_OK_FOR_L, CONST_OK_FOR_M,
3742 CONST_OK_FOR_N): Likewise.
3743 (LEGITIMATE_CONSTANT_P): Also check CONSTANT_P.
3744 (GO_IF_LEGITIMATE_INDEX): Use HOST_WIDE_INT instead of int to hold
3747 2007-01-10 Jan Hubicka <jh@suse.cz>
3749 * tree-vrp.c (remove_range_assertions): Release defs.
3750 * tree-ssa-loop-ivopts.c (rmeove_statement): Likewise.
3751 * tree-ssa-dom.c (remove_stmt_or_phi): Likewise.
3753 2007-01-10 Paul Brook <paul@codesourcery.com>
3755 * config/arm/arm.c (arm_rtx_costs_1): Handle mutiply-subtract.
3756 * config/arm/arm.md (mulsi3subsi): New insn.
3758 2007-01-10 Zdenek Dvorak <dvorakz@suse.cz>
3760 * tree-ssa-loop-manip.c (tree_unroll_loop): Make it a wrapper over ...
3761 (tree_transform_and_unroll_loop): New.
3762 * tree-flow.h (transform_callback, tree_transform_and_unroll_loop):
3765 2007-01-10 Robert Kennedy <jimbob@google.com>
3767 * fold-const.c (fold_comparison): Fold comparisons like (x *
3768 1000 < 0) to (x < 0).
3770 2007-01-10 Ian Lance Taylor <iant@google.com>
3772 * tree-pretty-print.c (dump_generic_node): Print parentheses when
3773 operands have the same priority.
3775 2007-01-10 Tom Tromey <tromey@redhat.com>
3777 * fold-const.c (fold_truthop): Don't check can_use_bit_fields_p.
3778 (fold_binary): Likewise.
3779 * langhooks.c (lhd_can_use_bit_fields_p): Removed.
3780 * langhooks-def.h (lhd_can_use_bit_fields_p): Removed.
3781 (LANG_HOOKS_CAN_USE_BIT_FIELDS_P): Removed.
3782 (LANG_HOOKS_INITIALIZER): Remove LANG_HOOKS_CAN_USE_BIT_FIELDS_P.
3783 * langhooks.h (struct lang_hooks): Removed field
3784 'can_use_bit_fields_p'.
3786 2007-01-10 Ralf Corsépius <ralf.corsepius@rtems.org>
3788 * config/bfin/t-bfin, config/bfin/t-bfin-elf: Remove GCC_CFLAGS.
3790 2007-01-10 Razya Ladelsky <razya@il.ibm.com>
3792 * function.c (get_last_funcdef_no): New function.
3793 * function.h (get_last_funcdef_no): Declare.
3794 * tree-inline.c (initialize_cfun): Add initialization.
3795 (tree_function_versioning): Cleanup.
3797 2007-01-10 Jan Hubicka <jh@suse.cz>
3799 * tree-inline.c (setup_one_parameter): Do not propagate into abnormal
3802 2007-01-10 Sa Liu <saliu@de.ibm.com>
3803 Ben Elliston <bje@au.ibm.com>
3805 * spu.h (STACK_SAVE_AREA): Use VOIDmode for SAVE_FUNCTION, SImode
3806 for SAVE_NONLOCAL and Pmode for any other save level.
3807 * spu-protos.h (spu_restore_stack_block): Declare.
3808 * spu.md (save_stack_block): Remove.
3809 (restore_stack_block): Call spu_restore_stack_block.
3810 * spu.c (spu_restore_stack_block): New function.
3811 (spu_expand_epilogue): Remove old comment.
3813 2007-01-09 Zdenek Dvorak <dvorakz@suse.cz>
3815 PR tree-optimization/30322
3816 * tree-ssa-loop-ivopts.c (fold_affine_expr, iv_value): Removed.
3817 (cand_value_at): Return the value as aff_tree.
3818 (may_eliminate_iv): Convert the bound from aff_tree to tree.
3819 * tree-affine.c (aff_combination_add_cst, aff_combination_add_product,
3820 aff_combination_mult): New functions.
3821 (aff_combination_add): Use aff_combination_add_cst.
3822 (aff_combination_convert): Allow conversions to a wider type.
3823 (tree_to_aff_combination): Handle BIT_NOT_EXPR.
3824 * tree-affine.h (aff_combination_mult): Declare.
3826 2007-01-09 Carlos O'Donell <carlos@codesourcery.com>
3828 * doc/tm.texi: Update documentation to reflect reality of exec
3829 and start file search behaviours. Update copyright year.
3830 * doc/invoke.texi: Explain how GCC_EXEC_PREFIX is used to find
3831 header file directories.
3833 2007-01-09 Uros Bizjak <ubizjak@gmail.com>
3835 * config/i386/i386.md (*sinxf2): Rename to *sinxf2_i387.
3836 (*cosxf2): Rename to cosxf2_i387.
3837 (*sindf2, *sinsf2): Extend operand 1 to XFmode. Macroize patterns
3838 using X87MODEF12 mode macro. Rename patterns to
3839 *sin_extend<mode>xf2_i387. Use SSE_FLOAT_MODE_P to disable patterns
3841 (*cosdf2, *cossf2): Ditto.
3842 (sincosdf3, sincossf3): Ditto. Rewrite corresponding splitters
3843 to match extended input operands.
3844 (sincos<mode>3): New expander.
3845 (*sinextendsfdf2, *cosextendsfdf2, *sincosextendsfdf3): Remove
3846 insn patterns and corresponding splitters.
3848 2007-01-09 Kaz Kojima <kkojima@gcc.gnu.org>
3850 * config/sh/t-linux (TARGET_LIBGCC2_CFLAGS): Delete.
3851 (SHLIB_MAPFILES, SHLIB_LINK, SHLIB_INSTALL): Likewise.
3853 2007-01-09 Nicolas Pitre <nico@cam.org>
3856 * arm/ieee754-df.S (Lad_s): Also test the low word of X for zero.
3858 2007-01-08 Geoffrey Keating <geoffk@apple.com>
3860 * target.h (struct gcc_target): New field library_rtti_comdat.
3861 * target-def.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): New.
3862 (TARGET_CXX): Add TARGET_CXX_LIBRARY_RTTI_COMDAT.
3863 * doc/tm.texi (C++ ABI): Document TARGET_CXX_LIBRARY_RTTI_COMDAT.
3864 * config/darwin.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
3866 2007-01-08 Geoffrey Keating <geoffk@apple.com>
3868 * doc/invoke.texi (Optimize Options): Correct description of -O0.
3870 2007-01-08 Richard Guenther <rguenther@suse.de>
3872 * tree.h (force_fit_type_double): Export.
3873 (force_fit_type): Remove.
3874 * fold-const.c (force_fit_type_double): New function.
3875 (force_fit_type): Remove.
3876 (int_const_binop): Use it.
3877 (fold_convert_const_int_from_int): Likewise.
3878 (fold_convert_const_int_from_real): Likewise.
3879 (fold_div_compare): Likewise.
3880 (fold_sign_changed_comparison): Likewise.
3881 (fold_unary): Likewise.
3882 (fold_negate_const): Likewise.
3883 (fold_abs_const): Likewise.
3884 (fold_not_const): Likewise.
3885 * c-common.c (shorten_compare): Use force_fit_type_double.
3886 * convert.c (convert_to_pointer): Likewise.
3888 2007-01-08 Richard Guenther <rguenther@suse.de>
3890 * tree.h (build_int_cst_wide_type): Export.
3891 * tree.c (build_int_cst_wide_type): New function.
3892 (build_int_cst_wide): Fix comment.
3893 * builtins.c (fold_builtin_object_size): Use build_int_cst
3894 to build -1 or 0 of the correct type. Use fit_double_type
3895 to check for overflow.
3896 * fold-const.c (optimize_bit_field_compare): Use build_int_cst_type
3898 (decode_field_reference): Likewise.
3899 (all_ones_mask_p): Likewise.
3900 (native_interpret_int): Use build_int_cst_wide_type.
3901 (fold_binary): Use build_int_cst_type to build an all-ones
3903 * stor-layout.c (set_sizetype): Use build_int_cst_wide_type.
3905 2007-01-08 Daniel Jacobowitz <dan@codesourcery.com>
3907 * config/pa/t-pa64 (libgcc_stub.a): Use $(T).
3909 2007-01-09 Ben Elliston <bje@au.ibm.com>
3911 * genautomata.c (STATS_OPTION): New option.
3912 (stats_flag): New flag.
3913 (gen_automata_option): Handle it.
3914 (initiate_automaton_gen): Ditto.
3915 (write_automata): Output statistics only if stats_flag is
3916 set. Likewise, output time statistics only if time_flag is set.
3917 * doc/md.texi (Processor pipeline description): Document new flag.
3919 2007-01-08 Richard Guenther <rguenther@suse.de>
3921 * builtins.c (fold_builtin_int_roundingfn): Use fit_double_type.
3922 * tree.c (build_int_cst_type): Likewise.
3923 (size_in_bytes): Don't call force_fit_type on the result.
3924 (int_fits_type_p): Use fit_double_type.
3925 * fold-const.c (fit_double_type): New function.
3926 (force_fit_type): Use it.
3927 * tree.h (fit_double_type): Export.
3929 2007-01-08 Jan Hubicka <jh@suse.cz>
3931 * tree-vectorizer.c (gate_increase_alignment): Fix return type.
3932 * ipa.c (function_and_variable_visibility): Fix return type.
3934 2007-01-08 Richard Guenther <rguenther@suse.de>
3936 * tree-ssa-ccp.c (maybe_fold_offset_to_array_ref): Use type
3937 of offset to build the index.
3938 * tree-pretty-print.c (dump_generic_node): Don't build negated
3939 const just for printing.
3940 * c-pretty-print.c (pp_c_integer_constant): Likewise.
3941 * builtins.c (fold_builtin_int_roundingfn): Check if result
3942 fits the type by using force_fit_type and comparing the result.
3943 * predict.c (predict_loops): Use compare_tree_int for comparison.
3944 * tree.c (build_int_cst): Fall back to integer_type_node for
3946 (build_int_cst_wide): Assert type is non-null.
3948 2007-01-08 Roberto Costa <roberto.costa@st.com>
3950 * tree-vrp.c (extract_range_from_cond_expr): New.
3951 (extract_range_from_expr): Handle COND_EXPR nodes used as expressions.
3952 * tree-ssa-ccp.c (get_maxval_strlen): Handle COND_EXPR nodes used
3954 (fold_stmt): Bug fix, avoid infinite recursion when folding COND_EXPRs.
3955 * tree-ssa-forwprop.c (simplify_cond, forward_propagate_into_cond,
3956 tree_ssa_forward_propagate_single_use_vars): Handle COND_EXPR nodes
3957 used as expressions.
3958 * tree-object-size.c (cond_expr_object_size): New.
3959 (collect_object_sizes_for): Handle COND_EXPR nodes used as expressions.
3961 2007-01-08 Jan Hubicka <jh@suse.cz>
3963 * tree-ssa-forwprop.c (forward_propagate_into_cond,
3964 tree_ssa_forward_propagate_single_use_va): Release defs of propagated
3967 2007-01-08 Richard Guenther <rguenther@suse.de>
3969 PR tree-optimization/23603
3970 * tree-vrp.c (set_value_range_to_truthvalue): New function.
3971 (extract_range_from_binary): Fall back to truthvalue instead of
3972 varying for TRUTH_*_EXPR.
3973 (extract_range_from_comparison): Fall back to truthvalue instead of
3975 (vrp_visit_phi_node): Don't adjust new range bounds to +INF/-INF
3976 if all visited PHI values were constant.
3978 2007-01-08 Jan Hubicka <jh@suse.cz>
3980 * cgraphunit.c (cgraph_process_new_functions): Reset reachable flag.
3981 (cgraph_analyze_function): break out from ...
3982 (cgraph_finalize_compilation_unit): ... here.
3983 (cgraph_expand_function): Remove forgoten commented out line.
3984 (cgraph_optimize): Analyze functions.
3986 2007-01-08 Jan Hubicka <jh@suse.cz>
3988 * tree-pas.h (TODO_remove_function): New flag.
3989 (TODO_update*): Renumber.
3990 (pass_ipa_increase_alignment,
3991 pass_ipa_function_and_variable_visibility): New passes.
3992 * cgraphunit.c (cgraph_increase_alignment): Move to tree-vectorizer.c
3993 (cgraph_function_and_variable_visibility): Move to ipa.c
3994 (cgraph_optimize): Don't call cgraph_function_and_variable_visibility,
3995 cgraph_increase_alignment.
3996 * ipa-inline.c (cgraph_decide_inlining): Don't push timevar.
3997 (cgraph_decide_inlining_incrementally): Push TV_INTEGRATION before
3998 calling tree-inline.
3999 (cgraph_early_inlining): Do not call cgraph_remove_unreachable_nodes.
4000 (pass_ipa_inline, pass_early_ipa_inlining): Set TODO_remove_functions
4001 * tree-vectorizer.c (increase_alignment): Move here from cgraphunit.c
4002 (gate_increase_alignment): New function.
4003 (pass_ipa_increase_alignment): New pass.
4004 * ipa.c: Inline tree-pass.h and timevar.h
4005 (function_and_variable_visibility): Move here from cgraphunit.c
4006 * tree-optimize.c (pass_early_local_passes): Add TODO_remove_functions.
4007 * passes.c (init_optimization_passes): Add the two new passes.
4008 (execute_todo): Handle cgraph_remove_functions.
4010 2007-01-08 Nick Clifton <nickc@redhat.com>
4012 * config/frv/predicates.md (reg_or_0_operand): Accept
4015 2007-01-08 Ralf Corsépius <ralf.corsepius@rtems.org>
4017 * config/bfin/rtems.h, config/bfin/t-rtems: New.
4018 * config.gcc: Add bfin*-rtems*.
4020 2007-01-08 Mark Shinwell <shinwell@codesourcery.com>
4022 * c.opt: Add -flax-vector-conversions.
4023 * c-typeck.c (convert_for_assignment): Pass flag to
4024 vector_types_convertible_p to allow emission of note.
4025 (digest_init): Likewise.
4026 * c-opts.c: Handle -flax-vector-conversions.
4027 * c-common.c (flag_lax_vector_conversions): New.
4028 (vector_types_convertible_p): Unless -flax-vector conversions
4029 has been passed, disallow conversions between vectors with
4030 differing numbers of subparts and/or element types. If such
4031 a conversion is disallowed, possibly emit a note on the first
4032 occasion only to inform the user of -flax-vector-conversions.
4033 The new last argument specifies this.
4034 * c-common.h (flag_lax_vector_conversions): New.
4035 (vector_types_convertible_p): Add extra argument.
4036 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Use
4037 char_type_node for V*QI type vectors.
4038 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins):
4039 Update to satisfy new typechecking rules.
4040 * config/rs6000/altivec.h (vec_cmple): Use vec_cmpge, for both
4042 * doc/invoke.texi (C Dialect Options): Document
4043 -flax-vector-conversions.
4045 2007-01-08 Mark Shinwell <shinwell@codesourcery.com>
4047 PR tree-optimization/29877
4048 * tree-ssa-ter.c (is_replaceable_p): Deem assignments with
4049 a register variable on the RHS to not be replaceable.
4051 2007-01-08 Chen Liqin <liqin@sunnorth.com.cn>
4052 * config/score/t-score-elf (MULTILIB_OPTIONS): Change.
4053 * config/score/predicates.md (const_uimm5, sr0_operand, const_simm12,
4054 const_simm15, const_pow2, const_npow2): Added.
4055 * config/score/misc.md (insv, extv, extzv, movmemsi,
4056 move_lbu_a/b, mov_lhu_a/b etc): Added and fix some bug.
4057 * config/score/score.c (score_address_cost, score_select_cc_mode):
4059 Change CONST_OK_FOR_LETTER_P/EXTRA_CONSTRAINT define.
4060 Update score_rtx_costs for MACRO TARGET_RTX_COSTS.
4061 Update score_print_operand.
4062 * config/score/score.h (DATA_ALIGNMENT, SELECT_CC_MODE): Added.
4063 Adjust register allocate order and update some macro define.
4064 * config/score/score-mdaux.c (mdx_unaligned_load, mdx_unsigned_store,
4065 mdx_block_move_straight, mdx_block_move_loop_head,
4066 mdx_block_move_loop_body, mdx_block_move_loop_foot, mdx_block_move_loop,
4067 mdx_block_move): Added.
4068 (mdx_movsicc, mdp_select_add_imm, mdp_select, mds_zero_extract_andi,
4069 mdp_limm): Updated and fix some bug and typo.
4070 * config/score/score.md (movqi/hi/si, add/sub/zero/ext): Updated.
4071 (movsf, movdf, doloop_end): Added.
4073 2007-01-08 Kazu Hirata <kazu@codesourcery.com>
4075 * config/arm/arm.c, config/arm/arm.h, config/arm/arm.md,
4076 config/arm/thumb2.md: Fix comment typos.
4077 * doc/extend.texi: Fix a typo.
4079 2007-01-07 Eric Christopher <echristo@apple.com>
4081 * configure.ac: Check for __stack_chk_fail for darwin.
4082 * configure: Regenerate.
4084 2007-01-07 Richard Guenther <rguenther@suse.de>
4086 * tree-vrp.c (extract_range_from_assert): CSE calls to
4087 compare_values where possible.
4088 (extract_range_from_unary_expr): Likewise.
4090 2007-01-07 Anatoly Sokolov <aesok@post.ru>
4092 * config/avr/avr-protos.h (call_insn_operand): Delete prototype.
4093 * config/avr/avr.c (call_insn_operand): Delete function.
4094 * config/avr/avr.md (*pushqi, *pushhi, *pushsi, *pushsf): Use REG_SP
4095 instead of register number. Use predicates.
4096 * config/avr/predicates.md (const0_operand, reg_or_0_operand,
4097 call_insn_operand): Add.
4099 2007-01-06 Jan Hubicka <jh@suse.cz>
4101 * tree-pass.h (pass_build_cgraph_edges): Declare.
4102 * cgraphunit.c (record_refernece): Move to cgraphbuild.c
4103 (visited_nodes): Remove.
4104 (cgraph_create_edges): Move to cgraphbuild.c; rename to
4105 build_cgrpah_edges; make visited_nodes local.
4106 (cgraph_process_new_functions): DO not call initialize_inline_failed.
4107 (record_references_in_initializer): Move to cgraphbuild.c
4108 (initialize_inline_failed, rebuild_cgraph_edges,
4109 pass_rebuild_cgraph_edges): Move to cgraphbuild.c.
4110 (verify_cgraph_node): Make visited_nodes local.
4111 (cgraph_analyze_function): Do not call cgraph_create_edges and
4112 initialize_inline_failed.
4113 (cgraph_expand_function): Do not call cgraph_lower_function;
4114 assert that function is already lowered.
4115 * Makefile.in (cgraphbuild.o): New.
4116 * passes.c (init_optimization_passes): Add pass_build_cgraph_edges
4117 at the end of lowering passes.
4119 2007-01-06 Steven Bosscher <steven@gcc.gnu.org>
4121 * ifcvt.c (cond_move_convert_if_block): New function, code
4122 factored out from...
4123 (cond_move_process_if_block): ...here. Call the new function
4124 on the THEN and ELSE blocks.
4125 (merge_if_block): Do not copy global_live_at_end, merge_blocks
4126 already takes care of this.
4128 2007-01-05 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
4131 * tree.h (TREE_OVERFLOW_P): New.
4132 * c-typeck.c (parser_build_unary_op): Warn only if result
4133 overflowed and operands did not.
4134 (parser_build_binary_op): Likewise.
4135 (convert_for_assignment): Remove redundant overflow_warning.
4136 * c-common.c (overflow_warning): Don't check or set TREE_OVERFLOW.
4138 2007-01-05 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
4140 * c-typeck.c (store_init_value): Split over two lines to follow
4141 the GNU coding style.
4143 2007-01-05 Benjamin Kosnik <bkoz@redhat.com>
4145 * c-cppbuiltin.c (c_cpp_builtins): __GXX_EXPERIMENTAL_CPP0X__ to
4146 __GXX_EXPERIMENTAL_CXX0X__.
4147 * doc/cpp.texi: Same.
4149 2007-01-05 Richard Guenther <rguenther@suse.de>
4152 * tree.c (get_narrower): Do not construct COMPONENT_REFs
4153 with mismatched types. Instead explicitly build a
4154 conversion NOP_EXPR.
4156 2007-01-05 Ian Lance Taylor <iant@google.com>
4158 * c-common.c (decl_with_nonnull_addr_p): New function.
4159 (c_common_truthvalue_conversion): Call it.
4160 * c-typeck.c (build_binary_op): Likewise.
4161 * c-common.h (decl_with_nonnull_addr_p): Declare.
4163 2007-01-05 Jakub Jelinek <jakub@redhat.com>
4166 * libgcc2.c (__divdc3): Compare c and d against 0.0 instead of
4169 2007-01-05 Joel Brobecker <brobecker@adacore.com>
4171 * doc/install.texi (Final install): Document the fact that
4172 the GNAT runtime should not be stripped.
4174 2007-01-04 Jan Hubicka <jh@suse.cz>
4176 * tree-inline.c (fold_marked_statements): Update operand caches
4177 and EH after folding
4179 2007-01-04 Ian Lance Taylor <iant@google.com>
4181 * c-common.c (check_function_nonnull): Whitespace fix.
4183 2007-01-04 Jan Hubicka <jh@suse.cz>
4185 * tree-optimize.c (execute_fixup_cfg): Correct previously mistakely
4186 comitted older version of patch.
4187 (pass_fixup_cfg): Add TODOs to verify flow and statements, dump
4188 function, celanup cfg and collect garbage.
4190 2007-01-04 Mike Stump <mrs@apple.com>
4192 * Makefile.in (mostlyclean): Don't remove libgcc anymore.
4195 2007-01-04 Eric Christopher <echristo@apple.com>
4197 * libgcc2.c (__bswapsi2): Use SItype.
4198 (__bswapdi2): Use DItype.
4199 * libgcc2.h: Update for above.
4201 2007-01-04 Paul Brook <paul@codesourcery.com>
4203 * config/arm/arm.md (arm_mulsi3, thumb_mulsi3, mulsi3_compare0,
4204 mulsi_compare0_scratch, mulsi3addsi, mulsi3addsi_compare0,
4205 mulsi3addsi_compare0_scratch, mulsidi3adddi, mulsidi3,
4206 umulsidi3, umulsidi3adddi, smulsi3_highpart,
4207 umulsi3_highpart): Make conditional on !arm_arch6.
4208 (arm_mulsi3_v6, thumb_mulsi3_v6, mulsi3_compare0_v6,
4209 mulsi_compare0_scratch_v6, mulsi3addsi_v6, mulsi3addsi_compare0_v6,
4210 mulsi3addsi_compare0_scratch_v6, mulsidi3adddi_v6, mulsidi3_v6,
4211 umulsidi3_v6, umulsidi3adddi_v6, smulsi3_highpart_v6,
4212 umulsi3_highpart_v6): New insns.
4214 2007-01-04 Roger Sayle <roger@eyesopen.com>
4216 * fold-const.c (fold_convert): When casting an expression to void,
4217 fold_ignored_result may discover a GIMPLE_MODIFY_STMT which doesn't
4218 have a type. Instead of attempting to build a NOP_EXPR, return
4219 these "special" trees directly.
4221 2007-01-04 Joseph Myers <joseph@codesourcery.com>
4223 * config/rs6000/rs6000.c (rs6000_rtx_costs): Make adjustment for
4224 MULT inside MINUS as either argument. Use rs6000_cost->dmul -
4225 rs6000_cost->fp not 0 as adjustment for outer NEG.
4227 2007-01-04 Jan Hubicka <jh@suse.cz>
4229 * cgraph.c (cgraph_release_function_body): New function.
4230 (cgraph_remove_node): Use it.
4231 * cgraph.h (cgraph_release_function_body): Declare.
4232 * cgraphunit.c (cgraph_expand_function): Use it.
4233 * ipa.c (cgraph_remove_unreahchable_nodes): Use it.
4234 * tree-ssa.c (delete_tree_ssa): Allow to be called before aliasing
4235 is initialized and while compilation of other function is running.
4236 * tree-optimize.c (execute_free_cfg_annotations): Move code to clear
4237 statement CFG annotations from here to ...
4238 * tree-cfg.c (delete_tree_cfg_annotations): ... here.
4240 2007-01-04 Zdenek Dvorak <dvorakz@suse.cz>
4242 * cfgloop.h (enum li_flags): Make the constants powers of two.
4244 2007-01-04 Jan Hubicka <jh@suse.cz>
4246 * tree-inline.c (copy_bb): Insert new statements to statements_to_fold
4248 (fold_marked_statements): New function.
4249 (optimize_inline_calls, tree_function_versioning): Fold new statements.
4250 * tree-inline.h (copy_body_data): Add statements_to_fold.
4252 2007-01-03 Daniel Jacobowitz <dan@codesourcery.com>
4254 * config.gcc: Mention libgcc/config.host.
4255 * Makefile.in: Update comments mentioning libgcc.
4256 (LIBGCC, INSTALL_LIBGCC, GCC_PARTS, mklibgcc): Delete.
4257 (all.cross, start.encap, rest.encap, rest.cross): Update
4258 dependencies for libgcc move.
4259 (libgcc.mk, LIBGCC_DEPS, libgcov.a, libgcc.a, stmp-multilib)
4260 (clean-target, clean-target-libgcc): Delete.
4261 (srcdirify, GCC_EXTRA_PARTS): New macros.
4262 (libgcc-support, libgcc.mvars): New rules.
4263 (distclean): Remove mention of mklibgcc.
4264 (install): Don't reference INSTALL_LIBGCC.
4265 (install-common): Don't reference EXTRA_PARTS.
4266 (install-libgcc, install-multilib): Delete rules.
4267 * mklibgcc.in: Delete file.
4268 * doc/configfiles.texi: Don't mention mklibgcc.
4270 * config/i386/t-darwin (SHLIB_VERPFX): Delete (moved to libgcc).
4271 * config/i386/t-darwin64 (SHLIB_VERPFX): Likewise.
4272 * config/rs6000/t-darwin (SHLIB_VERPFX): Likewise.
4273 * config/rs6000/t-ppccomm (TARGET_LIBGCC2_CFLAGS, SHLIB_MAPFILES)
4274 (mklibgcc, ldblspecs): Likewise.
4276 * config/i386/t-nwld (libgcc.def, libc.def, libpcre.def)
4277 (posixpre.def): Use $(T).
4278 (SHLIB_EXT, SHLIB_NAME, SHLIB_SLIBDIR_QUAL, SHLIB_DEF, SHLIB_MAP)
4279 (SHLIB_SRC, SHLIB_INSTALL): Delete.
4280 (SHLIB_LINK): Make dummy.
4281 * config/t-slibgcc-darwin: Delete contents except for dummy SHLIB_LINK.
4283 * config/frv/t-linux (EXTRA_MULTILIB_PARTS): Clear.
4285 * config/alpha/t-crtfm: Use $(T) in rules for EXTRA_PARTS.
4286 * config/alpha/t-vms, config/alpha/t-vms64, config/fr30/t-fr30,
4287 config/i386/t-rtems-i386, config/ia64/t-ia64, config/rs6000/t-beos,
4288 config/rs6000/t-newas, config/sparc/t-elf: Likewise.
4290 * configure.ac (all_outputs): Remove mklibgcc.
4291 * configure: Regenerated.
4293 2007-01-03 Josh Conner <jconner@apple.com>
4296 * calls.c (compute_argument_addresses): Set stack and stack_slot
4297 for partial args, too.
4298 (store_one_arg): Use locate.size.constant for the size when
4299 generating a save_area.
4301 2007-01-03 Robert Kennedy <jimbob@google.com>
4303 * tree-cfg.c (tree_merge_blocks): Release SSA_NAME phi results
4304 whose definitions are deleted due to basic block merging.
4306 2007-01-03 Paul Brook <paul@codesourcery.com>
4309 * config/arm/arm.c (output_return_instruction): Pop PC in interrupt
4311 (use_return_insn): Return 0 for Thumb interrupt functions.
4312 (print_multi_reg): Add rfe argument for IRQ returns.
4313 (arm_output_epilogue): Pop interrupt return address directly into PC.
4314 (arm_expand_prologue): Only adjust IRQ return address in Arm mode.
4316 2007-01-03 Paul Brook <paul@codesourcery.com>
4318 Merge from sourcerygxx-4_1.
4319 * config/arm/thumb2.md: New file.
4320 * config/arm/elf.h (JUMP_TABLES_IN_TEXT_SECTION): Return True for
4322 * config/arm/coff.h (JUMP_TABLES_IN_TEXT_SECTION): Ditto.
4323 * config/arm/aout.h (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
4324 (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump tables.
4325 * config/arm/aof.h (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump
4327 (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
4328 * config/arm/ieee754-df.S: Use macros for Thumb-2/Unified asm
4330 * config/arm/ieee754-sf.S: Ditto.
4331 * config/arm/arm.c (thumb_base_register_rtx_p): Rename...
4332 (thumb1_base_register_rtx_p): ... to this.
4333 (thumb_index_register_rtx_p): Rename...
4334 (thumb1_index_register_rtx_p): ... to this.
4335 (thumb_output_function_prologue): Rename...
4336 (thumb1_output_function_prologue): ... to this.
4337 (thumb_legitimate_address_p): Rename...
4338 (thumb1_legitimate_address_p): ... to this.
4339 (thumb_rtx_costs): Rename...
4340 (thumb1_rtx_costs): ... to this.
4341 (thumb_compute_save_reg_mask): Rename...
4342 (thumb1_compute_save_reg_mask): ... to this.
4343 (thumb_final_prescan_insn): Rename...
4344 (thumb1_final_prescan_insn): ... to this.
4345 (thumb_expand_epilogue): Rename...
4346 (thumb1_expand_epilogue): ... to this.
4347 (arm_unwind_emit_stm): Rename...
4348 (arm_unwind_emit_sequence): ... to this.
4349 (thumb2_legitimate_index_p, thumb2_legitimate_address_p,
4350 thumb1_compute_save_reg_mask, arm_dwarf_handle_frame_unspec,
4351 thumb2_index_mul_operand, output_move_vfp, arm_shift_nmem,
4352 arm_save_coproc_regs, thumb_set_frame_pointer, arm_print_condition,
4353 thumb2_final_prescan_insn, thumb2_asm_output_opcode, arm_output_shift,
4354 thumb2_output_casesi): New functions.
4355 (TARGET_DWARF_HANDLE_FRAME_UNSPEC): Define.
4356 (FL_THUMB2, FL_NOTM, FL_DIV, FL_FOR_ARCH6T2, FL_FOR_ARCH7,
4357 FL_FOR_ARCH7A, FL_FOR_ARCH7R, FL_FOR_ARCH7M, ARM_LSL_NAME,
4358 THUMB2_WORK_REGS): Define.
4359 (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv, arm_condexec_count,
4360 arm_condexec_mask, arm_condexec_masklen)): New variables.
4361 (all_architectures): Add armv6t2, armv7, armv7a, armv7r and armv7m.
4362 (arm_override_options): Check new CPU capabilities.
4363 Set new architecture flag variables.
4364 (arm_isr_value): Handle v7m interrupt functions.
4365 (user_return_insn): Return 0 for v7m interrupt functions. Handle
4367 (const_ok_for_arm): Handle Thumb-2 constants.
4368 (arm_gen_constant): Ditto. Use movw when available.
4369 (arm_function_ok_for_sibcall): Return false for v7m interrupt
4371 (legitimize_pic_address, arm_call_tls_get_addr): Handle Thumb-2.
4372 (thumb_find_work_register, arm_load_pic_register,
4373 legitimize_tls_address, arm_address_cost, load_multiple_sequence,
4374 emit_ldm_seq, emit_stm_seq, arm_select_cc_mode, get_jump_table_size,
4375 print_multi_reg, output_mov_long_double_fpa_from_arm,
4376 output_mov_long_double_arm_from_fpa, output_mov_double_fpa_from_arm,
4377 output_mov_double_fpa_from_arm, output_move_double,
4378 arm_compute_save_reg_mask, arm_compute_save_reg0_reg12_mask,
4379 output_return_instruction, arm_output_function_prologue,
4380 arm_output_epilogue, arm_get_frame_offsets, arm_regno_class,
4381 arm_output_mi_thunk, thumb_set_return_address): Ditto.
4382 (arm_expand_prologue): Handle Thumb-2. Use arm_save_coproc_regs.
4383 (arm_coproc_mem_operand): Allow POST_INC/PRE_DEC.
4384 (arithmetic_instr, shift_op): Use arm_shift_nmem.
4385 (arm_print_operand): Use arm_print_condition. Handle '(', ')', '.',
4387 (arm_final_prescan_insn): Use extract_constrain_insn_cached.
4388 (thumb_expand_prologue): Use thumb_set_frame_pointer.
4389 (arm_file_start): Output directive for unified syntax.
4390 (arm_unwind_emit_set): Handle stack alignment instruction.
4391 * config/arm/lib1funcs.asm: Remove default for __ARM_ARCH__.
4392 Add v6t2, v7, v7a, v7r and v7m.
4393 (RETLDM): Add Thumb-2 code.
4394 (do_it, shift1, do_push, do_pop, COND, THUMB_SYNTAX): New macros.
4395 * config/arm/arm.h (TARGET_CPU_CPP_BUILTINS): Define __thumb2__.
4396 (TARGET_THUMB1, TARGET_32BIT, TARGET_THUMB2, TARGET_DSP_MULTIPLY,
4397 TARGET_INT_SIMD, TARGET_UNIFIED_ASM, ARM_FT_STACKALIGN, IS_STACKALIGN,
4398 THUMB2_TRAMPOLINE_TEMPLATE, TRAMPOLINE_ADJUST_ADDRESS,
4399 ASM_OUTPUT_OPCODE, THUMB2_GO_IF_LEGITIMATE_ADDRESS,
4400 THUMB2_LEGITIMIZE_ADDRESS, CASE_VECTOR_PC_RELATIVE,
4401 CASE_VECTOR_SHORTEN_MODE, ADDR_VEC_ALIGN, ASM_OUTPUT_CASE_END,
4402 ADJUST_INSN_LENGTH): Define.
4403 (TARGET_REALLY_IWMMXT, TARGET_IWMMXT_ABI, CONDITIONAL_REGISTER_USAGE,
4404 STATIC_CHAIN_REGNUM, HARD_REGNO_NREGS, INDEX_REG_CLASS,
4405 BASE_REG_CLASS, MODE_BASE_REG_CLASS, SMALL_REGISTER_CLASSES,
4406 PREFERRED_RELOAD_CLASS, SECONDARY_OUTPUT_RELOAD_CLASS,
4407 SECONDARY_INPUT_RELOAD_CLASS, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P,
4408 TRAMPOLINE_SIZE, INITIALIZE_TRAMPOLINE, HAVE_PRE_INCREMENT,
4409 HAVE_POST_DECREMENT, HAVE_PRE_DECREMENT, HAVE_PRE_MODIFY_DISP,
4410 HAVE_POST_MODIFY_DISP, HAVE_PRE_MODIFY_REG, HAVE_POST_MODIFY_REG,
4411 REGNO_MODE_OK_FOR_BASE_P, LEGITIMATE_CONSTANT_P,
4412 REG_MODE_OK_FOR_BASE_P, REG_OK_FOR_INDEX_P, GO_IF_LEGITIMATE_ADDRESS,
4413 LEGITIMIZE_ADDRESS, THUMB2_LEGITIMIZE_ADDRESS,
4414 GO_IF_MODE_DEPENDENT_ADDRESS, MEMORY_MOVE_COST, BRANCH_COST,
4415 ASM_APP_OFF, ASM_OUTPUT_CASE_LABEL, ARM_DECLARE_FUNCTION_NAME,
4416 FINAL_PRESCAN_INSN, PRINT_OPERAND_PUNCT_VALID_P,
4417 PRINT_OPERAND_ADDRESS): Adjust for Thumb-2.
4418 (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv): New declarations.
4419 * config/arm/arm-cores.def: Add arm1156t2-s, cortex-a8, cortex-r4 and
4421 * config/arm/arm-tune.md: Regenerate.
4422 * config/arm/arm-protos.h: Update prototypes.
4423 * config/arm/vfp.md: Enable patterns for Thumb-2.
4424 (arm_movsi_vfp): Add movw alternative. Use output_move_vfp.
4425 (arm_movdi_vfp, movsf_vfp, movdf_vfp): Use output_move_vfp.
4426 (thumb2_movsi_vfp, thumb2_movdi_vfp, thumb2_movsf_vfp,
4427 thumb2_movdf_vfp, thumb2_movsfcc_vfp, thumb2_movdfcc_vfp): New.
4428 * config/arm/libunwind.S: Add Thumb-2 code.
4429 * config/arm/constraints.md: Update include Thumb-2.
4430 * config/arm/ieee754-sf.S: Add Thumb-2/Unified asm support.
4431 * config/arm/ieee754-df.S: Ditto.
4432 * config/arm/bpabi.S: Ditto.
4433 * config/arm/t-arm (MD_INCLUDES): Add thumb2.md.
4434 * config/arm/predicates.md (low_register_operand,
4435 low_reg_or_int_operand, thumb_16bit_operator): New.
4436 (thumb_cmp_operand, thumb_cmpneg_operand): Rename...
4437 (thumb1_cmp_operand, thumb1_cmpneg_operand): ... to this.
4438 * config/arm/t-arm-elf: Add armv7 multilib.
4439 * config/arm/arm.md: Update patterns for Thumb-2 and Unified asm.
4441 (UNSPEC_STACK_ALIGN, ce_count): New.
4442 (arm_incscc, arm_decscc, arm_umaxsi3, arm_uminsi3,
4443 arm_zero_extendsidi2, arm_zero_extendqidi2): New
4445 * config/arm/fpa.md: Update patterns for Thumb-2 and Unified asm.
4446 (thumb2_movsf_fpa, thumb2_movdf_fpa, thumb2_movxf_fpa,
4447 thumb2_movsfcc_fpa, thumb2_movdfcc_fpa): New insns.
4448 * config/arm/cirrus.md: Update patterns for Thumb-2 and Unified asm.
4449 (cirrus_thumb2_movdi, cirrus_thumb2_movsi_insn,
4450 thumb2_cirrus_movsf_hard_insn, thumb2_cirrus_movdf_hard_insn): New
4452 * doc/extend.texi: Document ARMv7-M interrupt functions.
4453 * doc/invoke.texi: Document Thumb-2 new cores+architectures.
4455 2007-01-03 Jakub Jelinek <jakub@redhat.com>
4457 * unwind-dw2.c (SIGNAL_FRAME_BIT, EXTENDED_CONTEXT_BIT): Define.
4458 (struct _Unwind_Context): Rename args_size to flags, remove
4459 signal_frame field, add a new args_size field and version field.
4460 (_Unwind_IsSignalFrame, _Unwind_SetSignalFrame,
4461 _Unwind_IsExtendedContext): New inline functions.
4462 (_Unwind_GetGR, _Unwind_SetGR, _Unwind_GetGRPtr, _Unwind_SetGRPtr):
4463 Assume by_value array is only present if _Unwind_IsExtendedContext.
4464 (_Unwind_GetIPInfo, execute_cfa_program, uw_frame_state_for): Use
4465 _Unwind_IsSignalFrame.
4466 (__frame_state_for): Initialize context.flags to EXTENDED_CONTEXT_BIT.
4467 (uw_update_context_1): Use _Unwind_SetSignalFrame.
4468 (uw_init_context_1): Initialize context->flags to
4469 EXTENDED_CONTEXT_BIT.
4470 * config/rs6000/linux-unwind.h (frob_update_context): Use
4471 _Unwind_SetSignalFrame.
4473 2007-01-03 Andrew Pinski <andrew_pinski@playstation.sony.com>
4476 * gimplify.c (gimplify_modify_expr_complex_part): Move below
4477 tree_to_gimple_tuple. Call tree_to_gimple_tuple when we need
4480 2007-01-03 Kazu Hirata <kazu@codesourcery.com>
4482 * config/i386/i386.h (NON_STACK_REG_P, REGNO_OK_FOR_SIREG_P,
4483 REGNO_OK_FOR_DIREG_P, REWRITE_ADDRESS, ASM_OPERAND_LETTER,
4484 RET, AT_SP): Remove.
4485 * config/i386/i386.md (*sse_prologue_save_insn): Use return
4488 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
4489 * cfgbuild.c (find_basic_blocks): Likewise.
4490 * cfgrtl.c (rtl_create_basic_block): Likewise.
4491 * function.c (temp_slots_at_level): Likewise.
4492 * reg-stack.c (stack_regs_mentioned): Likewise.
4493 * regclass.c (allocate_reg_info): Likewise.
4494 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
4495 set_bb_for_stmt, move_block_to_fn): Likewise.
4496 * tree-complex.c (tree_lower_complex): Likewise.
4497 * vec.h (VEC_safe_grow_cleared): New.
4499 * cgraphunit.c, tree-ssa-alias.c: Fix comment typos.
4501 2007-01-03 Zdenek Dvorak <dvorakz@suse.cz>
4503 * loop-unswitch.c (unswitch_loop): Pass probabilities to loopify.
4504 * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Pass probabilities
4506 * cfgloopmanip.c (scale_loop_frequencies): Export.
4507 (loopify): Scale the frequencies by prescribed coefficients.
4508 (set_zero_probability): New function.
4509 (duplicate_loop_to_header_edge): Improve updating of frequencies.
4510 (lv_adjust_loop_entry_edge, loop_version): Set probabilities
4511 and frequencies according to arguments.
4512 * tree-ssa-loop-manip.c (tree_unroll_loop): Set probabilities
4514 * cfg.c (scale_bbs_frequencies_int): Allow scaling the frequencies up.
4515 * modulo-sched.c (sms_schedule): Set probabilities for entering
4516 versioned loop correctly.
4517 * tree-vect-transform.c (vect_transform_loop): Ditto.
4518 * cfgloop.h (loopify, loop_version): Declaration changed.
4519 (scale_loop_frequencies): Declared.
4521 2007-01-02 Jan Hubicka <jh@suse.cz>
4523 * cgraph.c: Include tree-flow.h
4524 (cgraph_add_new-function): Handle IPA_SSA mode; execute
4526 * cgraph.h (enum cgraph_state): Add CGRAPH_STATE_IPA_SSA.
4527 * tree-pass.h (pass_all_early_optimizations): Declare.
4528 * cgraphunit.c (cgraph_process_new_functions): Add IPA_SSA; execute
4530 (cgraph_analyze_function): Do early_local_passes.
4531 * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
4532 Do not add referenced vars.
4533 * tree-optimize.c (gate_all_optimizations): Do not execute when not in
4535 (gate_all_early_local_passes): New gate.
4536 (pass_early_local_passes): Use new gate.
4537 (execute_early_local_optimizations): New functions.
4538 (gate_all_early_optimizations): New gate.
4539 (pass_all_early_optimizations): New pass.
4540 (execute_free_datastructures): Free SSA only when initialized.
4541 (gate_init_datastructures): Init only when optimizing.
4542 (tree_lowering_passes): Do early local passes when called late.
4543 * tree-profile.c (do_tree_profiling): Don't profile functions added
4545 (do_early_tree_profiling, pass_early_tree_profile): Kill.
4546 * tree-cfg.c (update_modified_stmts): Do not update when operands are
4548 * passes.c (init_optimizations_passes): Reorder so we go into SSA
4549 during early_local_passes.
4550 * Makefile.in (cgraph.o): Add dependency on tree-flow.h.
4553 2007-01-02 Carlos O'Donell <carlos@codesourcery.com>
4555 * Makefile.in: Update copyright year.
4557 2007-01-02 Carlos O'Donell <carlos@codesourcery.com>
4559 * Makefile.in: Export GCC_EXEC_PREFIX before calling $(RUNTEST)
4560 in $(lang_checks) and check-consistency targets.
4562 2007-01-02 Jan Hubicka <jh@suse.cz>
4564 * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
4565 Do not add referenced vars.
4566 * tree-cfg.c (update_modified_stmts): Do not update when SSA operands
4568 * passes.c (init_optimization_passes): Put mudflap_2 after
4569 free_datastructures.
4571 2007-01-02 Jan Hubicka <jh@suse.cz>
4573 * tree-optimize (execute_fixup_cfg): Set after_inlining flag.
4574 Set NOTHROW flag on call statements proved to be nothrow.
4575 Update statement of local calls so new pure/const functions are
4576 updated. Update_ssa when in ssa form. Mark PHI nodes of nonlocal
4578 (tree_rest_of_compilation): Register hooks and initialize bitmap
4579 early. Do not set after_inlining flag.
4581 2007-01-02 Steve Ellcey <sje@cup.hp.com>
4583 * sbitmap.c (HOST_BITS_PER_LONG_LONG): Change to
4584 HOST_BITS_PER_LONGLONG
4586 2007-01-02 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
4589 * c-typeck.c (store_init_value): Don't emit pedantic overflow
4590 warning for non-static initializers.
4592 2007-01-02 Steven Bosscher <steven@gcc.gnu.org>
4594 * config/alpha/alpha.md, arm/arm.c, darwin.c, frv/frv.md,
4595 m32r/m32r.c, m32r/m32r.c, mn10300/mn10300.md, pa/pa.c,
4596 rs6000/rs6000.c, s390/s390.md, sh/sh.md, sparc/sparc.c:
4597 Always use set_unique_reg_note to add REG_EQUAL notes.
4599 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
4602 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
4604 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
4605 * cfgbuild.c (find_basic_blocks): Likewise.
4606 * cfgrtl.c (rtl_create_basic_block): Likewise.
4607 * function.c (temp_slots_at_level): Likewise.
4608 * reg-stack.c (stack_regs_mentioned): Likewise.
4609 * regclass.c (allocate_reg_info): Likewise.
4610 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
4611 set_bb_for_stmt, move_block_to_fn): Likewise.
4612 * tree-complex.c (tree_lower_complex): Likewise.
4613 * vec.h (VEC_safe_grow_cleared): New.
4615 2007-01-02 Ian Lance Taylor <iant@google.com>
4617 * c-common.c (c_common_truthvalue_conversion): When warning about
4618 using an assignment as a truth value, set TREE_NO_WARNING.
4620 2007-01-02 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
4623 * c.opt (Wold-style-declaration): New.
4624 * doc/invoke.texi (C-only Warning Options): New.
4625 (Wold-style-declaration): Document it.
4626 (Wextra): Enabled by -Wextra.
4627 * c-opts.c (c_common_post_options): Enabled by -Wextra.
4628 * c-decl.c (declspecs_add_scspec): Replace -Wextra with
4629 -Wold-style-declaration.
4631 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
4633 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
4634 * cfgbuild.c (find_basic_blocks): Likewise.
4635 * cfgrtl.c (rtl_create_basic_block): Likewise.
4636 * function.c (temp_slots_at_level): Likewise.
4637 * reg-stack.c (stack_regs_mentioned): Likewise.
4638 * regclass.c (allocate_reg_info): Likewise.
4639 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
4640 set_bb_for_stmt, move_block_to_fn): Likewise.
4641 * tree-complex.c (tree_lower_complex): Likewise.
4642 * vec.h (VEC_safe_grow_cleared): New.
4644 2007-01-02 Douglas Gregor <doug.gregor@gmail.com>
4646 * c-common.c (c_common_nodes_and_builtins): Since variants of
4647 void_type_node get built before it is given a name, we need to
4648 give those variants the name, too.
4649 (complete_array_type): We need to work with the canonical main
4650 type of the array, from which we will build the qualified version.
4651 * params.def (PARAM_VERIFY_CANONICAL_TYPES): New.
4652 * print-tree.c (print_node): Display canonical type information
4654 * stor-layout.c (layout_type): When we don't know the
4655 alignment of a type for which we're building an array, we end up
4656 guessing wrong, so make the type require structural equality.
4657 * tree.c (make_node_stat): When we build a new type, it is its
4659 (build_type_attribute_qual_variant): When building an attribute
4660 variant, its canonical type is the non-attribute variant. However,
4661 if the attributes are target-dependent and they differ, we need to
4662 use structural equality checks for this type.
4663 (build_qualified_type): A qualified type is not equivalent to its
4664 unqualified variant; set the canonical type appropriately.
4665 (build_distinct_type_copy): When building a distinct type from
4666 another type, the new type is its own canonical type.
4667 (build_variant_type_copy): When building a new type variant, we
4668 assume that it is equivalent to the original type.
4669 (build_pointer_type_for_mode): When building a pointer type, also
4670 build a canonical type pointer.
4671 (build_reference_type_for_mode): When building a reference type,
4672 also build a canonical type reference.
4673 (build_index_type): When we can't hash an index type (e.g.,
4674 because its maximum value is negative), the index type requires
4675 structural equality tests.
4676 (build_array_type): Build the canonical form of an array type.
4677 (build_function_type): Function types require structural equality,
4678 because they contain default arguments, attributes, etc.
4679 (build_method_type_directly): Ditto for method types.
4680 (build_offset_type): Build the canonical offset type.
4681 (build_complex_type): Build the canonical vector type.
4682 (make_vector_type): Build the canonical vector type.
4683 * tree.h (TYPE_CANONICAL): New.
4684 (TYPE_STRUCTURAL_EQUALITY_P): New.
4685 (SET_TYPE_STRUCTURAL_EQUALITY): New.
4686 (struct tree_type): Added "canonical" field.
4687 * params.h (VERIFY_CANONICAL_TYPES): New.
4688 * doc/c-tree.texi (TYPE_CANONICAL): Document.
4689 (TYPE_STRUCTURAL_EQUALITY_P): Document.
4690 (SET_TYPE_STRUCTURAL_EQUALITY): Document.
4691 * doc/invoke.texi (verify-canonical-types): Document --param
4692 parameter for verifying canonical types.
4694 2007-01-02 Joseph Myers <joseph@codesourcery.com>
4696 * config.gcc (powerpc-*-eabispe*, powerpc-*-eabisimaltivec*,
4697 powerpc-*-eabisim*, powerpc-*-eabialtivec*, powerpc-*-eabi*,
4698 powerpc-*-rtems*, powerpc-wrs-vxworks, powerpc-wrs-vxworksae,
4699 powerpcle-*-eabisim*, powerpcle-*-eabi*): Add rs6000/e500.h to
4701 * config/rs6000/e500.h: New.
4702 * config/rs6000/eabi.h (TARGET_SPE_ABI, TARGET_SPE, TARGET_E500,
4703 TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE, TARGET_E500_DOUBLE):
4705 * config/rs6000/linuxspe.h (TARGET_SPE_ABI, TARGET_SPE,
4706 TARGET_E500, TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE,
4707 TARGET_E500_DOUBLE): Remove.
4708 * config/rs6000/vxworks.h (TARGET_SPE_ABI, TARGET_SPE,
4709 TARGET_E500, TARGET_ISEL, TARGET_FPRS): Remove.
4710 * config/rs6000/rs6000.h (CHECK_E500_OPTIONS): Define.
4711 * config/rs6000/rs6000.c (rs6000_override_options): Use
4714 2007-01-02 Joseph Myers <joseph@codesourcery.com>
4716 * config/rs6000/rs6000.c (print_operand): Check (TARGET_SPE ||
4717 TARGET_E500_DOUBLE), not TARGET_E500, for %y.
4718 (rs6000_generate_compare, rs6000_emit_sCOND, output_cbranch,
4719 rs6000_emit_cmove): Don't check TARGET_E500.
4720 * config/rs6000/rs6000.md (bunordered, bordered, sunordered,
4721 sordered): Don't check TARGET_E500.
4723 2007-01-01 Eric Christopher <echristo@apple.com>
4725 * config/mips/mips.c (mips_regno_mode_ok_for_base_p): Use
4726 HARD_REGISTER_NUM_P.
4728 2007-01-01 Roger Sayle <roger@eyesopen.com>
4730 * fold-const.c (fold_binary) <EQ_EXPR>: Fold "(X^C1) eq/ne C2" into
4731 "X eq/ne (C1^C2)". Fold "(X^Z) eq/ne (Y^Z)" as "X eq/ne Y" when Z
4732 has no side-effects. Fold "(X^C1) eq/ne (Y^C2)" as "(X^(C1^C2))
4735 2007-01-01 Mike Stump <mrs@apple.com>
4737 * configure.ac: Remove support for building with Apple's gcc-3.1.
4739 2007-01-02 Joseph Myers <joseph@codesourcery.com>
4742 * caller-save.c (add_stored_regs): Only handle SUBREGs if inner
4743 REG is a hard register. Do not modify REG before calling
4745 * rtlanal.c (subreg_get_info): Don't assert size of XMODE is a
4746 multiple of the size of YMODE for certain lowpart cases.
4748 2007-01-01 Andrew Pinski <pinskia@gmail.com>
4751 * gimplify (voidify_wrapper_expr): Update for
4754 2007-01-01 Andreas Schwab <schwab@suse.de>
4757 * config/ia64/ia64.c (ia64_compute_frame_size): Account space for
4758 save of BR0 in extra_spill_size instead of spill_size.
4759 (ia64_expand_prologue): Save BR0 outside of the gr/br/fr spill
4761 (ia64_expand_epilogue): Restore BR0 from its new location.
4763 2007-01-01 Andrew Pinski <pinskia@gmail.com>
4765 * gimplify.c (gimplify_init_constructor <case VECTOR_TYPE>):
4766 Use a temporary variable if the left hand side is not a gimple
4769 2007-01-01 Andrew Pinski <pinskia@gmail.com>
4771 * gimplify.c (gimplify_return_expr): Make the temporary variable
4772 for the return expression, a gimple register variable.
4774 2007-01-01 Jan Hubicka <jh@suse.cz>
4776 * emit-rtl.c (emit_copy_of_insn_after): Do not call copy_insn_1 for
4779 2007-01-01 Mike Stump <mrs@apple.com>
4781 * configure.ac (HAVE_GAS_LITERAL16): Add autoconf check for
4783 * config/darwin.c (machopic_select_rtx_section): Use
4785 (darwin_mergeable_constant_section): Likewise.
4786 * configure: Regenerate.
4787 * config.in: Regenerate.
4789 2007-01-01 Jan Hubicka <jh@suse.cz>
4790 Andrew Pinski <pinskia@gmail.com>
4792 * cgraphunit.c (cgraph_optimize): Call cgraph_add_new_functions
4793 before starting IPA passes.