1 2007-02-12 Dorit Nuzman <dorit@il.ibm.com>
3 PR tree-optimization/29145
4 * tree-data-ref.c (base_addr_differ_p): Make us more conservative
5 in our handling of restrict qualified pointers.
7 2007-02-12 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
10 * doc/invoke.texi (Wunused-value): Update description.
11 (Wextra): Delete item.
12 * opts.c (set_Wextra): Don't use the value of Wextra to set the
13 value of Wunused-value.
14 * c-typeck.c (c_process_expr_stmt): Don't check extra_warnings.
15 (c_finish_stmt_expr): Don't check extra_warnings.
16 (emit_side_effect_warnings): The caller is responsible to check
19 2007-02-11 Roger Sayle <roger@eyesopen.com>
20 Matt Thomas <matt@3am-software.com>
22 * simplify-rtx.c (simplify_relational_operation_1): Correct typo.
24 2007-02-11 Roger Sayle <roger@eyesopen.com>
26 * simplify-rtx.c (simplify_relational_operation_1): Optimize
27 comparisons of POPCOUNT against zero.
28 (simplify_const_relational_operation): Likewise.
30 2007-02-11 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
32 * doc/invoke.texi (Wextra): Delete outdated paragraph.
34 2007-02-11 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
36 * dwarf2out.c (root_type): Delete unused function.
38 2007-02-11 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
40 * genattrtab.c (contained_in_p): Delete unused function.
41 (write_expr_attr_cache): Likewise.
43 2007-02-11 Jan Hubicka <jh@suse.cz>
45 * ipa-inline.c (cgraph_edge_badness): Add "else" missing in the
48 2007-02-11 Steven Bosscher <steven@gcc.gnu.org>
50 * fwprop.c (try_fwprop_subst): Use set_unique_reg_note
51 to add the REG_EQ* note.
52 * see.c (see_merge_one_use_extension): Likewise.
53 * local-alloc.c (update_equiv_regs): Likewise. Also don't
54 turn REG_EQUAL notes into REG_EQUIV notes if the target
55 register may have more than one set.
56 * function.c (assign_parm_setup_reg): Use set_unique_reg_note.
57 * gcse.c (try_replace_reg): Likewise.
58 * alias.c (init_alias_analysis): Use find_reg_equal_equiv_note.
59 * calls.c (fixup_tail_calls): Likewise. Abort if there is
60 more than one REG_EQUIV note.
61 * reload1.c (gen_reload): Use set_unique_reg_note.
63 2007-02-11 Uros Bizjak <ubizjak@gmail.com>
65 * config/i386/i386.c (TARGET_VECTORIZE_BUILTIN_CONVERSION): Define.
66 (ix86_builtin_conversion): New function.
68 2007-02-06 Mark Mitchell <mark@codesourcery.com>
71 * tree.h (DECL_REPLACEABLE_P): New macro.
72 * except.c (set_nothrow_function_flags): Likewise.
74 2007-02-11 Tehila Meyzels <tehila@il.ibm.com>
75 Ira Rosen <irar@il.ibm.com>
76 Dorit Nuzman <dorit@il.ibm.com>
78 * doc/tm.texi (TARGET_VECTORIZE_BUILTIN_CONVERSION): New target hook.
79 * targhooks.c (default_builtin_vectorized_conversion): New.
80 * targhooks.h (default_builtin_vectorized_function): New declaration.
81 * target.h (struct vectorize): Add builtin_conversion field.
82 * tree-vectorizer.h (type_conversion_vec_info_type): New enum
83 stmt_vec_info_type value.
84 (vectorizable_conversion): New declaration.
85 * tree-vect-analyze.c (vect_analyze_operations): Add
86 vectorizable_conversion call.
87 * target-def.h (TARGET_VECTORIZE_BUILTIN_CONVERSION): New.
88 * tree-vect-transform.c (vectorizable_conversion): New function.
89 (vect_transform_stmt): Add case for type_conversion_vec_info_type.
90 * tree-vect-generic.c (expand_vector_operations_1): Consider correct
92 * config/rs6000/rs6000.c (rs6000_builtin_conversion): New.
93 (TARGET_VECTORIZE_BUILTIN_CONVERSION): Defined.
94 (rs6000_expand_builtin): Add handling a case of ALTIVEC_BUILTIN_VCFUX or
95 ALTIVEC_BUILTIN_VCFSX.
97 2007-02-10 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
100 * pa.md (movdf): For 64-bit target, fail if operand 1 is a non-zero
101 CONST_DOUBLE and operand 0 is a hard register.
102 (movdi): For 64-bit target, remove code to force CONST_DOUBLE to
103 memory. Fail if operand 1 is a non-zero CONST_INT and operand 0
104 is a hard floating-point register.
106 2007-02-10 Richard Henderson <rth@redhat.com>
107 Jakub Jelinek <jakub@redhat.com>
108 Alexandre Oliva <aoliva@redhat.com>
110 * Makefile.in (libgcc-support, libgcc.mvars): Add emutls.c.
111 * builtin-types.def (BT_WORD): Make unsigned.
112 (BT_FN_VOID_PTR_WORD_WORD_PTR): New.
113 * builtins.def (BUILT_IN_EMUTLS_GET_ADDRESS): New.
114 (BUILT_IN_EMUTLS_REGISTER_COMMON): New.
115 * c-decl.c (grokdeclarator): Don't error if !have_tls.
116 * c-parser.c (c_parser_omp_threadprivate): Likewise.
117 * dwarf2out.c (loc_descriptor_from_tree_1): Don't do anything for
119 * expr.c (emutls_var_address): New.
120 (expand_expr_real_1): Expand emulated tls.
121 (expand_expr_addr_expr_1): Likewise.
122 * libgcc-std.ver: Add __emutls_get_address, __emutls_register_common.
123 * output.h (emutls_finish): Declare.
124 * toplev.c (compile_file): Call it.
125 * tree-ssa-address.c (gen_addr_rtx): Check for const-ness of the
126 address before wrapping in CONST.
127 * varasm.c (emutls_htab, emutls_object_type): New.
128 (EMUTLS_VAR_PREFIX, EMUTLS_TMPL_PREFIX): New.
129 (get_emutls_object_name, get_emutls_object_type): New.
130 (get_emutls_init_templ_addr, emutls_decl): New.
131 (emutls_common_1, emutls_finish): New.
132 (assemble_variable): When emulating tls, swap decls; generate
133 constructor for the emutls objects.
134 (do_assemble_alias): When emulating tls, swap decl and target name.
135 (default_encode_section_info): Don't add SYMBOL_FLAG_TLS_SHIFT
137 * varpool.c (decide_is_variable_needed): Look at force_output.
138 Recurse for emulated tls.
139 (cgraph_varpool_remove_unreferenced_decls): Remove checks redundant
140 with decide_is_variable_needed.
141 * emutls.c: New file.
142 * config/sparc/sol2.h (ASM_DECLARE_OBJECT_NAME): Only emit
143 tls_object for real tls.
145 2007-02-10 Kaz Kojima <kkojima@gcc.gnu.org>
147 PR rtl-optimization/29599
148 * reload1.c (eliminate_regs_in_insn): Take the destination
149 mode into account when computing the offset.
151 2007-02-09 Stuart Hastings <stuart@apple.com>
152 Richard Henderson <rth@redhat.com>
154 * gcc/config/i386/i386.h (TARGET_KEEPS_VECTOR_ALIGNED_STACK): New.
155 * gcc/config/i386/darwin.h: (TARGET_KEEPS_VECTOR_ALIGNED_STACK): New.
156 * gcc/config/i386/i386.md (fixuns_trunc<mode>si2, fixuns_truncsfhi2,
157 fixuns_truncdfhi2): New.
158 (fix_truncsfdi_sse): Call ix86_expand_convert_sign_didf_sse.
159 (floatunsdidf2): Call ix86_expand_convert_uns_didf_sse.
160 (floatunssisf2): Add call to ix86_expand_convert_uns_sisf_sse.
161 (floatunssidf2): Allow nonimmediate source.
162 * gcc/config/i386/sse.md (movdi_to_sse): New.
163 (vec_concatv2di): Drop '*'.
164 * gcc/config/i386/i386-protos.h (ix86_expand_convert_uns_si_sse,
165 ix86_expand_convert_uns_didf_sse, ix86_expand_convert_uns_sidf_sse,
166 ix86_expand_convert_uns_sisf_sse, ix86_expand_convert_sign_didf_sse):
168 * gcc/config/i386/i386.c (ix86_expand_convert_uns_si_sse,
169 ix86_expand_convert_uns_didf_sse, ix86_expand_convert_uns_sidf_sse,
170 ix86_expand_convert_uns_sisf_sse, ix86_expand_convert_sign_didf_sse,
171 ix86_build_const_vector, ix86_expand_vector_init_one_nonzero): New.
172 (ix86_build_signbit_mask): Fix decl of v, refactor to call
173 ix86_build_const_vector.
174 (x86_emit_floatuns): Rewrite.
176 2007-02-10 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
178 * genautomata.c (longest_path_length): Delete unused function.
179 (struct state): Delete unused longest_path_length.
180 (UNDEFINED_LONGEST_PATH_LENGTH): Delete unused macro.
181 (get_free_state): Delete unused.
183 2007-02-09 Jan Hubicka <jh@suse.cz>
185 * params.def (PARAM_INLINE_UNIT_GROWTH): Set to 30.
186 * doc/invoke.texi (inline-unit-growth): Update default value.
188 * Makefile.in (passes.o, ipa-inline.o): Add dependencies.
189 * cgraphbuild.c (build_cgraph_edges): Compute frequencies.
190 (rebuild_cgraph_edges): Likewise.
191 * cgraph.c (cgraph_set_call_stmt): Add new argument frequency.
192 (dump_cgraph_node): Dump frequencies.
193 (cgraph_clone_edge): Add frequency scales.
194 (cgraph_clone_node): Add freuqnecy.
195 * cgraph.h (cgraph_edge): Add freuqnecy argument.
196 (CGRAPH_FREQ_BASE, CGRAPH_FREQ_MAX): New constants.
197 (cgraph_create_edge, cgraph_clone_edge, cgraph_clone_node): Update.
198 * tree-pass.h (TODO_rebuild_frequencies): New constant.
199 * cgraphunit.c (verify_cgraph_node): Verify frequencies.
200 (cgraph_copy_node_for_versioning): Update call of cgraph_clone_edge.
201 (save_inline_function_body): Likewise.
202 * ipa-inline.c: inluce rtl.h
203 (cgraph_clone_inlined_nods): Update call of cgraph_clone_node.
204 (cgraph_edge_badness): Use frequencies.
205 (cgraph_decide_recursive_inlining): Update clonning.
206 (cgraph_decide_inlining_of_small_function): Dump frequency.
207 * predict.c (estimate_bb_frequencies): Export.
208 * predict.h (estimate_bb_frequencies): Declare.
209 * tree-inline.c (copy_bb): Watch overflows.
210 (expand_call_inline): Update call of cgraph_create_edge.
211 (optimize_inline_calls): Use TODO flags to update frequnecies.
212 * passes.h: Include predict.h
213 (init_optimization_passes): Move profile ahead.
214 (execute_function_todo): Handle TODO_rebuild_frequencies.
216 2007-02-09 Roger Sayle <roger@eyesopen.com>
218 * config/alpha/alpha.c (emit_insxl): Force the first operand of
219 the insbl or inswl pattern into a register.
221 2007-02-09 Roger Sayle <roger@eyesopen.com>
223 * config/ia64/ia64.md (bswapdi2): New define_insn.
225 2007-02-09 Richard Henderson <rth@redhat.com>
227 * config/i386/constraints.md (Ym): New constraint.
228 * config/i386/i386.md (movsi_1): Change Y2 to Yi constraints.
229 (movdi_1_rex64): Split sse and xmm general register moves from
230 memory move alternatives. Use conditional register constraints.
231 (movsf_1, movdf_integer): Likewise.
232 (zero_extendsidi2_32, zero_extendsidi2_rex64): Likewise.
233 (movdf_integer_rex64): New.
234 (pushsf_rex64): Fix output constraints.
235 * config/i386/sse.md (sse2_loadld): Split rm alternative, use Yi.
236 (sse2_stored): Likewise.
237 (sse2_storeq_rex64): New.
238 * config/i386/i386.c (x86_inter_unit_moves): Enable for not
240 (ix86_secondary_memory_needed): Don't bypass TARGET_INTER_UNIT_MOVES
241 for optimize_size. Remove SF/DFmode hack.
243 2007-02-09 Dwarakanath Rajagopal <dwarak.rajagopal@amd.com>
245 * config/i386/driver-i386.c: Turn on -mtune=native for AMDFAM10.
248 2007-02-09 Nathan Sidwell <nathan@codesourcery.com>
249 Richard Sandiford <richard@codesourcery.com>
251 * config.gcc (m68010-*-netbsdelf*, m68k*-*-netbsdelf*)
252 (m68k*-*-openbsd*, m68k-*-linux*): Set default_cf_cpu.
253 (m68k-*-aout*, m68k-*-coff*, m68k-*-uclinux*, m68k-*-rtems*): Add
254 m68k/t-mlib to tmake_file.
255 (m68020-*-elf*, m68k-*-elf*): Likewise. Add t-m68kbare as well.
256 (m68k*-*-*): Use --with-arch to pick a default for --with-cpu.
257 (m680[012]0-*-*, m68k*-*-*): Add support for --with-arch.
258 Allow it to be cf or m68k. Set m68k_arch_family. If that
259 variable is not empty, add t-$m68k_arch_family to tmake_file.
260 Add t-mlibs to tmake_file.
261 * doc/install.texi: Document --with-arch=m68k and --with-arch=cf.
262 * config/m68k/t-cf: New file.
263 * config/m68k/t-m68k: Likewise.
264 * config/m68k/t-mlibs: Likewise.
265 * config/m68k/t-m68kbare (MULTILIB_OPTIONS, MULTILIB_DIRNAMES)
266 (MULTILIB_MATCHES, MULTILIB_EXCEPTIONS): Delete.
267 (M68K_MLIB_DIRNAMES, M68K_MLIB_OPTIONS): Define.
268 * config/m68k/t-m68kelf (MULTILIB_OPTIONS, MULTILIB_DIRNAMES)
269 (MULTILIB_MATCHES, MULTILIB_EXCEPTIONS, LIBGCC, INSTALL_LIBGCC):
271 * config/m68k/t-openbsd (MULTILIB_OPTIONS, LIBGCC): Delete.
272 (INSTALL_LIBGCC): Delete.
273 (M68K_MLIB_DIRNAMES, M68K_MLIB_OPTIONS): Define.
274 * config/m68k/t-rtems (MULTILIB_OPTIONS, MULTILIB_DIRNAMES)
275 (MULTILIB_MATCHES, MULTILIB_EXCEPTIONS): Delete.
276 (M68K_MLIB_CPU): Define.
277 * config/m68k/t-uclinux (MULTILIB_OPTIONS, MULTILIB_DIRNAMES)
278 (MULTILIB_MATCHES, MULTILIB_EXCEPTIONS): Delete.
279 (M68K_MLIB_CPU, M68K_MLIB_OPTIONS, M68K_MLIB_DIRNAMES): Define.
281 2007-02-09 Zdenek Dvorak <dvorakz@suse.cz>
282 Richard Guenther <rguenther@suse.de>
285 * fold-const.c (fold_comparison): Handle obfuscated comparisons
286 against INT_MIN/INT_MAX.
287 * tree-ssa-loop-ivcanon.c (remove_empty_loop): Print to dump
288 file if a loop is removed.
290 2007-02-09 Joseph Myers <joseph@codesourcery.com>
292 * calls.c (store_one_arg): Pass correct alignment to
293 emit_push_insn for non-BLKmode values.
294 * expr.c (emit_push_insn): If STRICT_ALIGNMENT, copy to an
295 unaligned stack slot via a suitably aligned slot.
297 2007-02-08 DJ Delorie <dj@redhat.com>
299 * config/m32c/m32c.c (m32c_unpend_compare): Add default to silence
301 (legal_subregs): Use unsigned char, make const.
302 (m32c_illegal_subreg_p): Use ARRAY_SIZE. Delete unused variables.
304 2007-02-08 Paul Brook <paul@codesourcery.com>
306 * config/arm/lib1funcs.asm (RETLDM): Pop directly into PC when no
307 special interworking needed.
309 2007-02-08 Harsha Jagasia <harsha.jagasia@amd.com>
311 * config/i386/xmmintrin.h: Make inclusion of emmintrin.h
312 conditional to __SSE2__.
313 (Entries below should have been added to first ChangeLog
314 entry for amdfam10 dated 2007-02-05)
315 * config/i386/emmintrin.h: Generate #error if __SSE2__ is not
317 * config/i386/pmmintrin.h: Generate #error if __SSE3__ is not
319 * config/i386/tmmintrin.h: Generate #error if __SSSE3__ is not
322 2007-02-08 DJ Delorie <dj@redhat.com>
324 * config/m32c/m32c-protos.h (m32c_illegal_subreg_p): New.
325 * config/m32c/m32c.c (legal_subregs): New.
326 (m32c_illegal_subreg_p): New.
327 * config/m32c/predicates.md (m32c_any_operand): Use it to reject
328 unsupported subregs of hard regs.
330 2007-02-08 Jan Hubicka <jh@suse.cz>
332 * tree-cfg.c (bsi_replace): Shortcut when replacing the statement with
333 the same one; always update histograms.
335 2007-02-08 Diego Novillo <dnovillo@redhat.com>
337 * passes.c (init_optimization_passes): Tidy comment.
339 2007-02-08 Roger Sayle <roger@eyesopen.com>
341 * simplify-rtx.c (simplify_unary_operation_1) <POPCOUNT>: We can
342 strip zero_extend, bswap and rotates from POCOUNT's argument.
343 <PARITY>: Likewise, we can strip not, bswap, sign_extend,
344 zero_extend and rotates from PARITY's argument.
345 <BSWAP>: A byte-swap followed by a byte-swap is an identity.
346 (simplify_const_unary_operation) <BSWAP>: Evaluate the byte-swap
347 of an integer constant at compile-time.
349 2007-02-08 Diego Novillo <dnovillo@redhat.com>
352 * tree-flow.h (struct var_ann_d): Remove field 'is_used'.
354 * tree-ssa-alias.c (compute_is_aliased): Remove. Update all
357 * tree-ssa-live.c (remove_unused_locals): Do not remove
358 TREE_ADDRESSABLE variables.
359 * tree-ssa-structalias.c (compute_points_to_sets): Tidy.
360 * tree-ssa-operands.c (add_virtual_operand): Remove argument
361 FOR_CLOBBER. Update all users.
362 If VAR has an associated alias set, add a virtual operand for
363 it if no alias is found to conflict with the memory reference.
365 2007-02-07 Jan Hubicka <jh@suse.cz>
366 Robert Kidd <rkidd@crhc.uiuc.edu>
368 * value-prof.c (visit_hist, free_hist): Return 1 instead of 0.
370 2007-02-07 Ian Lance Taylor <iant@google.com>
372 * lower-subreg.c (simple_move): Reject PARTIAL_INT modes.
374 2007-02-07 Roger Sayle <roger@eyesopen.com>
376 * config/rs6000/rs6000.md (ctz<mode>2, ffs<mode>2, popcount<mode>2,
377 parity<mode>2, smulsi3_highpart, abstf2_internal, allocate_stack,
378 tablejumpdi, movsi_to_cr_one): Remove constraints from
379 define_expand's match_operands.
381 2007-02-07 Roger Sayle <roger@eyesopen.com>
383 * global.c (compute_regsets): Move declatation of "i" inside of
384 #ifdef ELIMINABLE_REGS to avoid unused variable bootstrap failure.
386 2007-02-07 Jakub Jelinek <jakub@redhat.com>
389 * gimplify.c (gimplify_scan_omp_clauses): Remove special casing
390 of INDIRECT_REF <RESULT_DECL>.
392 * config/i386/i386.c (override_options): Set PTA_SSSE3 for core2.
394 2007-02-06 J"orn Rennecke <joern.rennecke@arc.com>
395 Kaz Kojima <kkojima@gcc.gnu.org>
398 * config/sh/sh.c (expand_cbranchdi4): Use scratch register
400 (sh_initialize_trampoline): Add parentheses to avoid a warning.
402 2007-02-06 Zdenek Dvorak <dvorakz@suse.cz>
404 * doc/loop.texi: Document possibility not to perform disambiguation
405 of loops with multiple latches.
406 * cfgloopmanip.c (alp_enum_p): Removed.
407 (add_loop): Handle subloops. Use get_loop_body_with_size.
408 (create_preheader): Do not allow ENTRY_BLOCK_PTR to be preheader.
409 * cfghooks.c (redirect_edge_and_branch_force): Set dominator for
410 the new forwarder block.
411 (make_forwarder_block): Only call new_bb_cbk if it is not NULL.
412 Handle the case latch is NULL.
413 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Avoid cfg modifications
414 when marking loop exits.
415 * ifcvt.c (if_convert): Ditto. Mark loop exits even if cfg cannot
417 * loop-init.c (loop_optimizer_init): Do not modify cfg. Call
418 disambiguate_loops_with_multiple_latches.
419 * tree-cfgcleanup.c (cleanup_tree_cfg_loop): Calculate dominators
420 before fix_loop_structure.
421 * cfgloop.c: Include pointer-set.h and output.h.
422 (canonicalize_loop_headers, HEADER_BLOCK, LATCH_EDGE,
423 update_latch_info, mfb_keep_just, mfb_keep_nonlatch): Removed.
424 (get_loop_latch_edges, find_subloop_latch_edge_by_profile,
425 find_subloop_latch_edge_by_ivs, find_subloop_latch_edge,
426 mfb_redirect_edges_in_set, form_subloop, merge_latch_edges,
427 disambiguate_multiple_latches, get_loop_body_with_size,
428 disambiguate_loops_with_multiple_latches): New functions.
429 (flow_loop_dump): Dump multiple latch edges.
430 (flow_loop_nodes_find): Handle loops with multiple latches.
431 (flow_loops_find): Ditto. Do not call canonicalize_loop_headers.
432 (glb_enum_p): Modified.
433 (get_loop_body): Use get_loop_body_with_size.
434 * cfgloop.h (LOOPS_HAVE_RECORDED_EXITS): New flag.
435 (AVOID_CFG_MODIFICATIONS): New constant.
436 (disambiguate_loops_with_multiple_latches, add_loop,
437 get_loop_body_with_size): Declare.
438 * Makefile.in (cfgloop.o): Add pointer-set.h and output.h.
440 2007-02-06 Seongbae Park <seongbae.park@gmail.com>
443 * global.c (compute_regsets): New function.
444 (global_alloc): Refactored ELIMINABLE_REGSET
445 and NO_GLOBAL_ALLOC_REGS computation out.
446 (rest_of_handle_global_alloc): Call compute_regsets()
447 for non-optimizing case.
449 2007-02-06 Richard Henderson <rth@redhat.com>
451 * config/i386/constraints.md (Y2): Rename from Y.
452 (Yi): New constraint.
453 * config/i386/i386.md (movsi_1, movdi_2, pushdf_nointeger,
454 pushdf_integer, movdf_nointeger, movdf_integer, zero_extendsidi2_32,
455 zero_extendsidi2_rex64, truncxfdf2_mixed): Change Y constraints to Y2.
456 (extendsfdf2_mixed, extendsfdf2_sse, truncdfsf_fast_mixed,
457 truncdfsf_fast_sse, truncdfsf_mixed, fix_truncdfdi_sse,
458 fix_truncdfsi_sse, floatsidf2_mixed, floatsidf2_sse,
459 floatdidf2_mixed, floatdidf2_sse, absnegdf2_mixed,
460 absnegdf2_sse, sse_setccdf, fop_df_comm_mixed, fop_df_comm_sse,
461 fop_df_1_mixed, fop_df_1_sse): Change Y constraints to x.
462 * config/i386/mmx.md (mov<MMXMODEI>_internal_rex64,
463 mov<MMXMODEI>_internal, movv2sf_internal_rex64, movv2sf_internal,
464 vec_extractv2si_1): Change Y constraints to Y2.
465 * config/i386/sse.md (vec_setv4sf_0, vec_concatv2df, vec_dupv4si,
466 vec_dupv2di, sse2_concatv2si, vec_concatv4si_1, vec_concatv2di):
467 Change Y constraints to Y2.
468 (sse2_loadld): Change Y constraints to x.
470 2007-02-06 Roger Sayle <roger@eyesopen.com>
472 * config/rs6000/rs6000.md (popcount<mode>2): Rewrite.
473 (parity<mode>2): New define_expand using rs6000_emit_parity.
474 * config/rs6000/rs6000.c (rs6000_emit_popcount,
475 rs6000_emit_parity): New functions.
476 * config/rs6000/rs6000-protos.h (rs6000_emit_popcount,
477 rs6000_emit_parity): Prototype here.
479 2007-02-06 Ian Lance Taylor <iant@google.com>
481 * lower-subreg.c (simple_move_operand): Reject CONST.
482 (resolve_clobber): Call validate_change rather than directly
483 assigning to XEXP (pat, 0).
485 2006-02-06 Paolo Bonzini <bonzini@gnu.org>
487 * Makefile.in (tree-ssa-loop-ivopts.o): Add pointer-set.h dependency.
488 (tree-ssa-reassoc.o): Add pointer-set.h dependency.
489 (tree-cfg.o): Remove hashtab.h dependency.
491 * tree-ssa-loop-ivopts.c: Include pointer-set.h.
492 (struct ivopts_data): Change niters to pointer_map_t.
493 (struct nfe_cache_elt, nfe_hash, nfe_eq): Delete.
494 (niter_for_exit): Create pointer_map on demand. Change for
496 (tree_ssa_iv_optimize_init): Initialize data->niters to NULL.
497 (free_loop_data): Destroy data->niters if created and reset field.
498 (tree_ssa_iv_optimize_finalize): Don't delete data->niters here.
499 (tree_ssa_iv_optimize_loop): Check for presence of stale data.
501 * tree-ssa-reassoc.c: Include pointer-set.h.
502 (bb_rank): Change to long *.
503 (operand_rank): Change to pointer_map_t.
504 (find_operand_rank): Return long, -1 if not found. Declare as inline.
505 (insert_operand_rank): Accept long.
506 (operand_entry_hash, operand_entry_eq): Remove.
507 (get_rank): Return long. Adjust for changes above.
508 (init_reassoc): Change rank type to long. Adjust creation of bb_rank
510 (fini_reassoc): Delete operand_rank with pointer_map_destroy.
512 * tree-ssa-structalias.c (vi_for_tree): Change to pointer_map.
513 (struct tree_vi, tree_vi_t, tree_vi_hash, tree_vi_eq): Delete.
514 (insert_vi_for_tree): Rewrite for pointer_map API. Assert argument
516 (lookup_vi_for_tree): Rewrite for pointer_map API. Return varinfo_t
517 directly since it cannot be NULL.
518 (get_vi_for_tree): Rewrite for pointer_map API.
519 (find_what_p_points_to): Adjust for change to lookup_vi_for_tree.
520 (init_alias_vars): Create vi_for_tree as pointer_map.
521 (delete_points_to_sets): Delete vi_for_tree using pointer_map_destroy.
523 * tree-cfg.c: Don't include hashtab.h.
524 (edge_to_cases): Declare as pointer_map.
525 (struct edge_to_cases_elt, edge_to_cases_hash, edge_to_cases_eq):
527 (edge_to_cases_cleanup): Rewrite as pointer_map_traverse callback.
528 (start_recording_case_labels): Create edge_to_cases as pointer_map.
529 (end_recoding_case_labels): Cleanup edge_to_cases manually before
531 (record_switch_edge): Delete.
532 (get_cases_for_edge): Adjust for pointer_map API, inline
533 record_switch_edge (rewritten for new API), remove goto.
535 2006-02-06 Paolo Bonzini <bonzini@gnu.org>
537 * Makefile.in (tree-nested.o): Add pointer-set.h dependency.
538 * tree-nested.c: Include pointer-set.h.
539 (var_map_elt, var_map_eq, var_map_hash): Delete.
540 (struct nesting_info): Remove GTY marker. Change the two htab_t's
542 (nesting_info_bitmap_obstack): New.
543 (lookup_field_for_decl): Adjust for pointer_map API.
544 (lookup_tramp_for_decl): Adjust for pointer_map API.
545 (get_nonlocal_debug_decl): Adjust for pointer_map API.
546 (get_local_debug_decl): Adjust for pointer_map API.
547 (convert_nl_goto_reference): Adjust for pointer_map API.
548 (convert_nl_goto_receiver): Adjust for pointer_map API.
549 (create_nesting_tree): Create outside GGC space. Create bitmap on
550 the new obstack. Create field_map and var_map as pointer_maps.
551 (free_nesting_tree): Adjust for changes to create_nesting_tree.
553 (lower_nested_functions): Move root here, no need to NULL it.
554 Initialize and release the obstack.
556 2007-02-06 Paolo Bonzini <bonzini@gnu.org>
558 * tree.c (tree_int_map_hash, tree_int_map_eq, tree_int_map_marked_p):
559 Remove prototypes and make them non-static.
560 (struct tree_int_map): Remove.
561 * tree.h (struct tree_int_map): Move here, turning TO into an
563 (tree_int_map_hash, tree_int_map_eq, tree_int_map_marked_p): Declare.
565 * tree.h (TREE_COMPLEXITY): Remove.
566 (struct tree_exp): Remove complexity field.
567 * tree.c (build1_stat): Don't set it.
569 2007-02-06 Dorit Nuzman <dorit@il.ibm.com>
570 Victor Kaplansky <victork@il.ibm.com>
572 * tree-vectorizer.c (vect_is_simple_use): Support induction.
573 (vect_is_simple_reduction): Support reduction with induction as
575 (vect_is_simple_iv_evolution): Fix formatting.
576 * tree-vect-analyze.c (vect_mark_stmts_to_be_vectorized): Fix
577 formatting. Don't mark induction phis for vectorization.
578 (vect_analyze_scalar_cycles): Analyze all inductions, then reductions.
579 * tree-vect-transform.c (get_initial_def_for_induction): New function.
580 (vect_get_vec_def_for_operand): Support induction.
581 (vect_get_vec_def_for_stmt_copy): Fix formatting and add check for
583 (vectorizable_reduction): Support reduction with induction as one of
585 (vectorizable_type_demotion): Use def-type of stmt argument rather
588 * tree-ssa-loop.c (gate_scev_const_prop): Return the value of
589 flag_tree_scev_cprop.
590 * common.opt (tree-scev-cprop): New flag.
592 * tree-vect-transform.c (vect_create_destination_var): Use 'kind' in
593 call to vect_get_new_vect_var.
595 2007-02-06 Ira Rosen <irar@il.ibm.com>
597 * tree-vect-patterns.c (vect_recog_widen_mult_pattern): Check that
599 (vect_pattern_recog_1): Likewise.
601 2007-02-05 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
603 * fold-const.c (negate_expr_p): Handle CONJ_EXPR.
604 (fold_negate_expr): Likewise.
606 2007-02-05 Alexandre Oliva <aoliva@redhat.com>
609 * dwarf2out.c (modified_type_die): Follow DECL_ORIGINAL_TYPE
610 even if cv-qualification is the same.
612 2007-02-05 Geoffrey Keating <geoffk@apple.com>
614 * config/rs6000/darwin-tramp.asm (__trampoline_setup): Call
615 __enable_execute_stack on completion.
617 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
619 * config/i386/athlon.md (athlon_fldxf_k8, athlon_fld_k8,
620 athlon_fstxf_k8, athlon_fst_k8, athlon_fist, athlon_fmov,
621 athlon_fadd_load, athlon_fadd_load_k8, athlon_fadd, athlon_fmul,
622 athlon_fmul_load, athlon_fmul_load_k8, athlon_fsgn,
623 athlon_fdiv_load, athlon_fdiv_load_k8, athlon_fdiv_k8,
624 athlon_fpspc_load, athlon_fpspc, athlon_fcmov_load,
625 athlon_fcmov_load_k8, athlon_fcmov_k8, athlon_fcomi_load_k8,
626 athlon_fcomi, athlon_fcom_load_k8, athlon_fcom): Added amdfam10.
628 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
630 * config/i386/i386.md (x86_sahf_1, cmpfp_i_mixed, cmpfp_i_sse,
631 cmpfp_i_i387, cmpfp_iu_mixed, cmpfp_iu_sse, cmpfp_iu_387,
632 swapsi, swaphi_1, swapqi_1, swapdi_rex64, fix_truncsfdi_sse,
633 fix_truncdfdi_sse, fix_truncsfsi_sse, fix_truncdfsi_sse,
634 x86_fldcw_1, floatsisf2_mixed, floatsisf2_sse, floatdisf2_mixed,
635 floatdisf2_sse, floatsidf2_mixed, floatsidf2_sse,
636 floatdidf2_mixed, floatdidf2_sse, muldi3_1_rex64, mulsi3_1,
637 mulsi3_1_zext, mulhi3_1, mulqi3_1, umulqihi3_1, mulqihi3_insn,
638 umulditi3_insn, umulsidi3_insn, mulditi3_insn, mulsidi3_insn,
639 umuldi3_highpart_rex64, umulsi3_highpart_insn,
640 umulsi3_highpart_zext, smuldi3_highpart_rex64,
641 smulsi3_highpart_insn, smulsi3_highpart_zext, x86_64_shld,
642 x86_shld_1, x86_64_shrd, sqrtsf2_mixed, sqrtsf2_sse,
643 sqrtsf2_i387, sqrtdf2_mixed, sqrtdf2_sse, sqrtdf2_i387,
644 sqrtextendsfdf2_i387, sqrtxf2, sqrtextendsfxf2_i387,
645 sqrtextenddfxf2_i387): Added amdfam10_decode.
647 * config/i386/athlon.md (athlon_idirect_amdfam10,
648 athlon_ivector_amdfam10, athlon_idirect_load_amdfam10,
649 athlon_ivector_load_amdfam10, athlon_idirect_both_amdfam10,
650 athlon_ivector_both_amdfam10, athlon_idirect_store_amdfam10,
651 athlon_ivector_store_amdfam10): New define_insn_reservation.
652 (athlon_idirect_loadmov, athlon_idirect_movstore): Added
655 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
657 * config/i386/athlon.md (athlon_call_amdfam10,
658 athlon_pop_amdfam10, athlon_lea_amdfam10): New
659 define_insn_reservation.
660 (athlon_branch, athlon_push, athlon_leave_k8, athlon_imul_k8,
661 athlon_imul_k8_DI, athlon_imul_mem_k8, athlon_imul_mem_k8_DI,
662 athlon_idiv, athlon_idiv_mem, athlon_str): Added amdfam10.
664 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
666 * config/i386/athlon.md (athlon_sseld_amdfam10,
667 athlon_mmxld_amdfam10, athlon_ssest_amdfam10,
668 athlon_mmxssest_short_amdfam10): New define_insn_reservation.
670 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
672 * config/i386/athlon.md (athlon_sseins_amdfam10): New
673 define_insn_reservation.
674 * config/i386/i386.md (sseins): Added sseins to define_attr type
675 and define_attr unit.
676 * config/i386/sse.md: Set type attribute to sseins for insertq
679 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
681 * config/i386/athlon.md (sselog_load_amdfam10, sselog_amdfam10,
682 ssecmpvector_load_amdfam10, ssecmpvector_amdfam10,
683 ssecomi_load_amdfam10, ssecomi_amdfam10,
684 sseaddvector_load_amdfam10, sseaddvector_amdfam10): New
685 define_insn_reservation.
686 (ssecmp_load_k8, ssecmp, sseadd_load_k8, seadd): Added amdfam10.
688 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
690 * config/i386/athlon.md (cvtss2sd_load_amdfam10,
691 cvtss2sd_amdfam10, cvtps2pd_load_amdfam10, cvtps2pd_amdfam10,
692 cvtsi2sd_load_amdfam10, cvtsi2ss_load_amdfam10,
693 cvtsi2sd_amdfam10, cvtsi2ss_amdfam10, cvtsd2ss_load_amdfam10,
694 cvtsd2ss_amdfam10, cvtpd2ps_load_amdfam10, cvtpd2ps_amdfam10,
695 cvtsX2si_load_amdfam10, cvtsX2si_amdfam10): New
696 define_insn_reservation.
698 * config/i386/sse.md (cvtsi2ss, cvtsi2ssq, cvtss2si,
699 cvtss2siq, cvttss2si, cvttss2siq, cvtsi2sd, cvtsi2sdq,
700 cvtsd2si, cvtsd2siq, cvttsd2si, cvttsd2siq,
701 cvtpd2dq, cvttpd2dq, cvtsd2ss, cvtss2sd,
702 cvtpd2ps, cvtps2pd): Added amdfam10_decode attribute.
704 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
706 * config/i386/athlon.md (athlon_ssedivvector_amdfam10,
707 athlon_ssedivvector_load_amdfam10, athlon_ssemulvector_amdfam10,
708 athlon_ssemulvector_load_amdfam10): New define_insn_reservation.
709 (athlon_ssediv, athlon_ssediv_load_k8, athlon_ssemul,
710 athlon_ssemul_load_k8): Added amdfam10.
712 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
714 * config/i386/i386.h (TARGET_SSE_UNALIGNED_MOVE_OPTIMAL): New macro.
715 (x86_sse_unaligned_move_optimal): New variable.
717 * config/i386/i386.c (x86_sse_unaligned_move_optimal): Enable for
719 (ix86_expand_vector_move_misalign): Add code to generate movupd/movups
720 for unaligned vector SSE double/single precision loads for AMDFAM10.
722 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
724 * config/i386/i386.h (TARGET_AMDFAM10): New macro.
725 (TARGET_CPU_CPP_BUILTINS): Add code for amdfam10.
726 Define TARGET_CPU_DEFAULT_amdfam10.
727 (TARGET_CPU_DEFAULT_NAMES): Add amdfam10.
728 (processor_type): Add PROCESSOR_AMDFAM10.
730 * config/i386/i386.md: Add amdfam10 as a new cpu attribute to match
731 processor_type in config/i386/i386.h.
732 Enable imul peepholes for TARGET_AMDFAM10.
734 * config.gcc: Add support for --with-cpu option for amdfam10.
736 * config/i386/i386.c (amdfam10_cost): New variable.
737 (m_AMDFAM10): New macro.
738 (m_ATHLON_K8_AMDFAM10): New macro.
739 (x86_use_leave, x86_push_memory, x86_movx, x86_unroll_strlen,
740 x86_cmove, x86_3dnow_a, x86_deep_branch, x86_use_simode_fiop,
741 x86_promote_QImode, x86_integer_DFmode_moves,
742 x86_partial_reg_dependency, x86_memory_mismatch_stall,
743 x86_accumulate_outgoing_args, x86_arch_always_fancy_math_387,
744 x86_sse_partial_reg_dependency, x86_sse_typeless_stores,
745 x86_use_ffreep, x86_use_incdec, x86_four_jump_limit,
746 x86_schedule, x86_use_bt, x86_cmpxchg16b, x86_pad_returns):
747 Enable/disable for amdfam10.
748 (override_options): Add amdfam10_cost to processor_target_table.
749 Set up PROCESSOR_AMDFAM10 for amdfam10 entry in
750 processor_alias_table.
751 (ix86_issue_rate): Add PROCESSOR_AMDFAM10.
752 (ix86_adjust_cost): Add code for amdfam10.
754 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
756 * config/i386/i386.opt: Add new Advanced Bit Manipulation (-mabm)
757 instruction set feature flag. Add new (-mpopcnt) flag for popcnt
758 instruction. Add new SSE4A (-msse4a) instruction set feature flag.
759 * config/i386/i386.h: Add builtin definition for SSE4A.
760 * config/i386/i386.md: Add support for ABM instructions
762 * config/i386/sse.md: Add support for SSE4A instructions
763 (movntss, movntsd, extrq, insertq).
764 * config/i386/i386.c: Add support for ABM and SSE4A builtins.
765 Add -march=amdfam10 flag.
766 * config/i386/ammintrin.h: Add support for SSE4A intrinsics.
767 * doc/invoke.texi: Add documentation on flags for sse4a, abm, popcnt
769 * doc/extend.texi: Add documentation for SSE4A builtins.
771 2007-02-05 Bob Wilson <bob.wilson@acm.org>
773 * config/xtensa/xtensa.c (constantpool_mem_p): Skip over SUBREGs.
775 2007-02-05 Richard Guenther <rguenther@suse.de>
777 * tree-vectorizer.h (vectorizable_function): Add argument type
778 argument, change return type.
779 * tree-vect-patterns.c (vect_recog_pow_pattern): Adjust caller.
780 * tree-vect-transform.c (vectorizable_function): Handle extra
781 argument, return vectorized function decl.
782 (build_vectorized_function_call): Remove.
783 (vectorizable_call): Handle calls with result and argument types
784 differing. Handle loop vectorization factor correctly.
785 * targhooks.c (default_builtin_vectorized_function): Adjust for
787 * targhooks.h (default_builtin_vectorized_function): Likewise.
788 * target.h (builtin_vectorized_function): Add argument type
790 * config/i386/i386.c (ix86_builtin_vectorized_function): Handle
791 extra argument, allow vectorizing of lrintf.
792 * doc/tm.texi (TARGET_VECTORIZE_BUILTIN_VECTORIZED_FUNCTION): Adjust
793 documentation of target hook.
795 2007-02-05 Hans-Peter Nilsson <hp@axis.com>
798 * config/cris/cris.md ("*andsi_movu", "*andsi_clear", "*andhi_movu")
799 ("*andhi_clear", andu (casesi+45)): For size-changed operand where
800 memory is allowed, require !side_effects_p, not just !MEM_VOLATILE_P.
802 2007-02-05 Roger Sayle <roger@eyesopen.com>
804 * fold-const.c (fold_unary) <REAL_PART>: Test for availability of
805 BUILT_IN_COS before simplifying REAL_PART(CEXPI)) to COS.
806 <IMAG_PART>: Likewise, check for availability of BUILT_IN_SIN.
807 * builtins.c (fold_builtin_sincos): Check for TARGET_C99_FUNCTIONS
808 before canonicalizing sincos to cexpi.
809 (fold_builtin_cexp): Likewise, for canonicalizing cexp to cexpi.
811 2007-02-05 Roger Sayle <roger@eyesopen.com>
813 * config/alpha/alpha.c (alpha_add_builtins): New Helper function.
814 Set TREE_READONLY and TREE_NOTHROW directly, not via attributes.
815 (alpha_init_builtins): Use alpha_add_builtins to process tables.
817 2007-02-05 Roger Sayle <roger@eyesopen.com>
819 * mips-tfile.c (initialize_init_file): Correct endianness test.
821 2007-02-05 Kazu Hirata <kazu@codesourcery.com>
823 * config/m68k/m68k.md (pushdi-1, pushdi, movsi+1): Don't use
826 2007-02-05 Richard Sandiford <richard@codesourcery.com>
828 * dwarf2out.c (dwarf2out_frame_debug_expr): Record the register
829 saves in a PARALLEL before the register assignments.
831 2007-02-05 Richard Sandiford <richard@codesourcery.com>
833 * doc/tm.texi (DWARF_ALT_FRAME_RETURN_COLUMN): Do not require
834 DWARF_FRAME_RETURN_COLUMN to be a general register.
835 * dwarf2out.c (init_return_column_size): New function, split from...
836 (expand_builtin_init_dwarf_reg_sizes): ...here. Allow both
837 DWARF_FRAME_RETURN_COLUMN and DWARF_ALT_FRAME_RETURN_COLUMN
838 to be nongeneral registers.
839 * config/m68k/m68k.h (DWARF_FRAME_REGNUM): Only map FP and
841 (DWARF_FRAME_REGISTERS, DWARF_FRAME_RETURN_COLUMN): Define.
842 (DWARF_ALT_FRAME_RETURN_COLUMN): Define.
844 2007-02-04 Zdenek Dvorak <dvorakz@suse.cz>
846 * cfgcleanup.c (try_optimize_cfg): Avoid removing ENTRY_BLOCK_PTR.
848 2007-02-04 Zdenek Dvorak <dvorakz@suse.cz>
850 * cfgloopmanip.c (loop_delete_branch_edge): Removed.
851 (remove_path): Use can_remove_branch_p and remove_branch instead
852 of loop_delete_branch_edge.
853 * tree-ssa-loop-manip.c (scale_dominated_blocks_in_loop): New function.
854 (tree_transform_and_unroll_loop): Remove dead branches immediately.
855 Update profile using scale_dominated_blocks_in_loop.
856 * cfghooks.c (can_remove_branch_p, remove_branch): New functions.
857 * cfghooks.h (struct cfg_hooks): Add can_remove_branch_p.
858 (can_remove_branch_p, remove_branch): Declare.
859 * tree-cfg.c (tree_can_remove_branch_p): New function.
860 (tree_cfg_hooks): Add tree_can_remove_branch_p.
861 * cfgrtl.c (rtl_can_remove_branch_p): New function.
862 (rtl_cfg_hooks, cfg_layout_rtl_cfg_hook): Add rtl_can_remove_branch_p.
864 2007-02-05 Jan Hubicka <jh@suse.cz>
867 * ipa-inline.c (cgraph_clone_inlined_nodes): When there are unanalyzed
868 nodes in cgraph, don't remove offline copy of the function.
870 2007-02-04 Jan Hubicka <jh@suse.cz>
872 * tree-sra.c (sra_walk_expr): Add linebreaks. BITFIELD_REFs into
873 vectors might cause maybe_lookup_element_for_expr to be called
874 on non-sra-candidate.
876 2007-02-04 Kazu Hirata <kazu@codesourcery.com>
878 * config/bfin/bfin-modes.def, config/bfin/bfin.c,
879 config/bfin/bfin.md, config/bfin/predicates.md: Follow
880 spelling conventions.
882 2007-02-04 Richard Guenther <rguenther@suse.de>
885 * fold-const.c (try_move_mult_to_index): Make sure to not
886 overflow one dimension of a multi-dimensional array access.
888 2007-02-04 Jan Hubicka <jh@suse.cz>
890 * passes.c (init_optimization_passes): Reindent.
892 2007-02-04 Jan Hubicka <jh@suse.cz>
893 Eric Botcazou <ebotcazou@adacore.com>
895 * tree-optimize.c (has_abnormal_outgoing_edge_p): Move to...
896 (execute_fixup_cfg): Break out the abnormal goto code.
897 * tree-inline.c (has_abnormal_outgoing_edge_p): ...here.
898 (make_nonlocal_label_edges): Move here from execute_fixup_cfg.
899 (optimize_inline_calls): Call make_nonlocal_label_edges.
901 2007-02-04 Jan Hubicka <jh@suse.cz>
903 * tree-ssa-copyrename.c (copy_rename_partition_coalesce): Return
904 true when something was changed.
905 (rename_ssa_copies): When something was changed, do
906 TODO_remove_unused_locals.
907 * tree-ssa-forwprop.c (tree_ssa_forward_propagate_single_use_value):
908 add TODO_remove_unused_locals when instruction was removed.
910 2007-02-04 Jan Hubicka <jh@suse.cz>
912 * ipa-inline.c (try_inline): Improve debug output; work on already
914 (cgraph_decide_inlining_incrementally): Indent; improve debug output;
915 call try_inline for already inlined edges too when flattening;
916 inline also functions that make callee growth but overall unit size
919 2007-02-04 Kazu Hirata <kazu@codesourcery.com>
921 * config/m32c/bitops.md, config/m32c/jump.md,
922 config/m32c/m32c.c, config/m32c/m32c.h, config/m32r/m32r.c,
923 config/m32r/m32r.h, config/m32r/m32r.md,
924 config/m32r/predicates.md, config/m68hc11/larith.asm,
925 config/m68hc11/m68hc11.c, config/m68hc11/m68hc11.h,
926 config/m68k/m68k.h, config/mcore/mcore.md, config/mips/4k.md,
927 config/mips/mips-protos.h, config/mips/mips.c,
928 config/mips/mips.h, config/mips/mips.md, config/mips/mips16.S,
929 config/mn10300/mn10300.h, config/mn10300/predicates.md,
930 config/mt/mt.c, config/mt/mt.h, config/mt/mt.md: Follow
931 spelling conventions.
933 * config/v850/v850.c, config/v850/v850.h, config/v850/v850.md:
934 Follow spelling conventions.
936 2007-02-03 Douglas Gregor <doug.gregor@gmail.com>
938 * c-opts.c (c_common_post_options): If C++0x mode is enabled, don't
939 warn about C++0x compatibility.
941 2007-02-04 Kazu Hirata <kazu@codesourcery.com>
943 * config/h8300/h8300.c, config/h8300/h8300.h,
944 config/h8300/h8300.md: Follow spelling conventions.
946 2007-02-03 Uros Bizjak <ubizjak@gmail.com>
949 * combine.c (try_combine): Do not substitute source operand
950 with constants wider than 2 * HOST_BITS_PER_WIDE_INT.
952 2007-02-03 Jan Hubicka <jh@suse.cz>
954 PR gcov-profile/30650
955 * value-prof.c (stringop_block_profile): Fix handling of size counter;
956 do not divide by zero for never executed counters.
957 (tree_find_values_to_profile): Fix counters.
958 * gcov-ui.h (GCOV_COUNTER_AVERAGE, GCOV_COUNTER_IOR): Fix comments.
960 2007-02-03 Ian Lance Taylor <iant@google.com>
962 * lower-subreg.c (simple_move_operand): New static function,
963 broken out of simple_move. Reject LABEL_REF, SYMBOL_REF, and HIGH
965 (simple_move): Call simple_move_operand.
966 (find_decomposable_subregs): Add special handling of MEMs.
967 (can_decompose_p): Rename from cannot_decompose_p. Reverse
968 meaning of return value. If we see a hard register, test whether
969 it can store a word_mode value. Change all callers.
971 2007-02-03 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
973 * pa.md (addvdi3, addvsi3, subvdi3, subvsi3, negvdi2, negvsi2): New
974 ftrapv insns and expanders.
975 (subdi3): Change define_expand operand 1 to arith11_operand, and
976 operand 2 to reg_or_0_operand. Change constraints of 64-bit insn
977 pattern to handle reg_or_0 operands. Revise 32-bit insn pattern to
978 handle 11-bit constants and reg_or_0 operands in operands 1 and 2,
982 * varasm.c (notice_global_symbol): Treat global objects as weak when
985 2007-02-03 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
987 * emit-rtl.c (dconstpi): Delete.
989 (init_emit_once): Delete dconstpi and init dconstsqrt2.
990 * real.h (dconstpi): Delete.
992 * builtins.c (fold_builtin_cabs): Use dconstsqrt2.
993 (fold_builtin_hypot): Likewise.
995 2007-02-03 Tom Tromey <tromey@redhat.com>
998 * gcc.c (cpp_unique_options): Any of -ggdb3, -gstabs3,
999 -gcoff3, -gxcoff3, -gvms3 implies -dD.
1001 2007-02-03 Kazu Hirata <kazu@codesourcery.com>
1003 * c-decl.c, config/avr/avr.c, config/avr/avr.h,
1004 config/m68k/m68k.c, config/m68k/netbsd-elf.h,
1005 config/mn10300/mn10300.c, config/pdp11/pdp11.h,
1006 config/rs6000/cell.md, config/rs6000/darwin.h,
1007 config/sh/sh.md, config/sh/sh4-300.md, config/spu/spu.c,
1008 config/spu/spu.md, cselib.c, expr.c, haifa-sched.c, hwint.h,
1009 jump.c, reload.c, sched-deps.c, sched-int.h, tree-inline.c,
1010 tree-profile.c, tree-ssa-live.h, tree-vrp.c: Fix comment
1011 typos. Follow spelling conventions.
1012 * doc/invoke.texi: Follow spelling conventions.
1014 2007-02-03 Roger Sayle <roger@eyesopen.com>
1016 * simplify-rtx.c (simplify_relational_operation_1): Implement some
1017 canonicalization transformations that attempt to simplify integer
1018 constant comparisons to become comparisons against zero.
1020 2007-02-03 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1022 * builtins.c (fold_builtin_cabs): Fold cabs(x+xi) into
1024 * fold-const.c (fold_binary): Fix comment typos. Fold complex
1025 (x,0)-(0,y) into (x,-y). Likewise (0,y)-(x,0) into (-x,y).
1027 2007-02-02 Mike Stump <mrs@apple.com>
1029 * config/darwin9.h (DARWIN_LINKER_GENERATES_ISLANDS): Add.
1030 * config/rs6000/rs6000.c (DARWIN_GENERATE_ISLANDS): Add.
1031 (output_call): Use DARWIN_GENERATE_ISLANDS to decide when to
1032 generate a branch island.
1034 2007-02-02 Bob Wilson <bob.wilson@acm.org>
1036 * config/xtensa/xtensa.c (smalloffset_mem_p): Use BASE_REG_P.
1037 (xtensa_legitimate_address_p): New.
1038 (xtensa_legitimize_address): New.
1039 (xtensa_output_addr_const_extra): New.
1040 * config/xtensa/xtensa.h (REG_OK_STRICT_FLAG): Define.
1042 (REG_OK_FOR_BASE_P): Use BASE_REG_P.
1043 (GO_IF_LEGITIMATE_ADDRESS): Move code to xtensa_legitimate_address_p.
1044 (LEGITIMIZE_ADDRESS): Move code to xtensa_legitimize_address.
1045 (OUTPUT_ADDR_CONST_EXTRA): Move code to xtensa_output_addr_const_extra.
1046 * config/xtensa/xtensa-protos.h (xtensa_legitimate_address_p): New.
1047 (xtensa_legitimize_address): New.
1048 (xtensa_output_addr_const_extra): New.
1050 2007-02-02 Steve Ellcey <sje@cup.hp.com>
1052 * config/ia64/ia64.c (ia64_print_operand): Fix compare strings.
1054 2007-02-02 Ian Lance Taylor <iant@google.com>
1056 * expmed.c (expand_divmod): Add comment.
1058 2007-02-02 Kazu Hirata <kazu@codesourcery.com>
1060 * emit-rtl.c (renumber_insns): Remove.
1061 * flags.h: Remove the extern for flag_renumber_insns.
1062 * rtl.h: Remove the prototype for renumber_insns.
1063 * toplev.c (flag_renumber_insns): Remove.
1065 2007-02-02 Hui-May Chang <hm.chang@apple.com>
1067 Revert for x86 darwin:
1068 2005-06-19 Uros Bizjak <uros@kss-loka.si>
1070 * config/i386/i386.c (ix86_function_arg_regno_p): Put back the
1071 code before the following patch under TARGET_MACHO.
1072 (ix86_function_value_regno_p): Likewise.
1074 2007-02-02 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1076 * fold-const.c (negate_expr_p, fold_negate_expr): Handle
1079 2007-02-02 Maxim Kuvyrkov <mkuvyrkov@ispras.ru>
1082 * config/ia64/ia64.c (ia64_speculate_insn): Restrict to memory loads to
1083 general or fp registers. Add comments.
1084 * config/ia64/ia64.md (reg_pred_prefix): Add comment.
1086 2007-02-02 Paolo Bonzini <bonzini@gnu.org>
1088 * pointer-set.c (insert_aux): Only return insertion slot.
1089 (pointer_set_insert): Adjust.
1090 (pointer_set_traverse, struct pointer_map_t, pointer_map_create,
1091 pointer_map_destroy, pointer_map_insert, pointer_map_contains,
1092 pointer_map_traverse): New.
1093 * pointer-set.h (pointer_set_traverse, struct pointer_map_t,
1094 pointer_map_create, pointer_map_destroy, pointer_map_insert,
1095 pointer_map_contains, pointer_map_traverse): Declare.
1097 2007-02-02 Jakub Jelinek <jakub@redhat.com>
1100 * builtins.c (fold_builtin_sprintf): Do not attempt to optimize
1101 sprintf (str, "%s"). Do not optimize sprintf (str, "nopercent", p++).
1103 2007-02-02 Maxim Kuvyrkov <mkuvyrkov@ispras.ru>
1105 * sched-int.h (ds_to_dk, dk_to_ds): Declare functions.
1107 (struct _dep): New type.
1108 (dep_t): New typedef.
1109 (DEP_PRO, DEP_CON, DEP_KIND): New access macros.
1110 (DEP_STATUS): New access macro. The macro with the same name was
1111 renamed to DEP_LINK_STATUS.
1112 (dep_init): Declare function
1114 (struct _dep_link): New type.
1115 (dep_link_t): New typedef.
1116 (DEP_LINK_NODE, DEP_LINK_NEXT, DEP_LINK_PREV_NEXTP): New access macros.
1117 (DEP_LINK_DEP, DEP_LINK_PRO, DEP_LINK_CON, DEP_LINK_KIND): New macros.
1118 (DEP_LINK_STATUS): New macro.
1119 (debug_dep_links): New debug function.
1121 (struct _deps_list): New type.
1122 (deps_list_t): New typedef.
1123 (DEPS_LIST_FIRST): New access macro.
1124 (FOR_EACH_DEP_LINK): New cycle macro.
1125 (create_deps_list, free_deps_list, delete_deps_list): Declare
1127 (deps_list_empty_p, debug_deps_list, add_back_dep_to_deps_list): Ditto.
1128 (find_link_by_pro_in_deps_list, find_link_by_con_in_deps_list): Ditto.
1129 (copy_deps_list_change_con): Ditto.
1131 (move_dep_link): Declare function.
1133 (struct _dep_node): New type.
1134 (dep_node_t): New typedef.
1135 (DEP_NODE_BACK, DEP_NODE_DEP, DEP_NODE_FORW): New access macros.
1137 (struct haifa_insn_data.back_deps): New field to hold backward
1138 dependencies of the insn.
1139 (struct haifa_insn_data.depend): Rename to forw_deps. Change its type
1141 (struct haifa_insn_data.resolved_deps): Rename to resolved_back_deps.
1142 Change its type to deps_list_t.
1143 (INSN_BACK_DEPS): New access macro to use instead of LOG_LINKS.
1144 (INSN_DEPEND): Rename to INSN_FORW_DEPS.
1145 (RESOLVED_DEPS): Rename to INSN_RESOLVED_BACK_DEPS.
1147 (INSN_COST): Move to haifa-sched.c. Use insn_cost () instead.
1149 (DEP_STATUS): Rename to DEP_LINK_STATUS. Fix typo in the comment.
1151 (add_forw_dep, delete_back_forw_dep, insn_cost): Update declaration and
1153 (dep_cost): Declare.
1155 * sched-deps.c (CHECK): New macro to (en/dis)able sanity checks.
1156 (ds_to_dk, dk_to_ds): New functions.
1158 (init_dep_1): New static function.
1159 (init_dep): New function.
1160 (copy_dep): New static function.
1162 (dep_link_consistent_p, attach_dep_link, add_to_deps_list): New static
1164 (detach_dep_link): New static function.
1165 (move_dep_link): New function.
1167 (dep_links_consistent_p, dump_dep_links): New static functions.
1168 (debug_dep_links): New debugging function.
1170 (deps_obstack, dl_obstack, dn_obstack): New static variables.
1172 (alloc_deps_list, init_deps_list): New static functions.
1173 (create_deps_list): New function.
1174 (clear_deps_list): New static function.
1175 (free_deps_list, delete_deps_list, deps_list_empty_p): New functions.
1176 (deps_list_consistent_p, dump_deps_list): New static functions.
1177 (debug_deps_list): New function.
1178 (add_back_dep_to_deps_list, find_link_by_pro_in_deps_list): New
1180 (find_link_by_con_in_deps_list, copy_deps_list_change_con): Ditto.
1182 (maybe_add_or_update_back_dep_1, add_or_update_back_dep_1): Update to
1183 use new scheduler dependencies lists.
1184 (add_back_dep, delete_all_dependences, fixup_sched_groups): Ditto.
1185 (sched_analyze): Ditto. Initialize dependencies lists.
1186 (add_forw_dep, compute_forward_dependences): Update to use new
1187 scheduler dependencies lists.
1189 (init_dependency_caches): Init deps_obstack.
1190 (free_dependency_caches): Free deps_obstack.
1192 (adjust_add_sorted_back_dep, adjust_back_add_forw_dep): Update to use
1193 new scheduler dependencies lists.
1194 (delete_forw_dep, add_or_update_back_forw_dep): Ditto.
1195 (add_back_forw_dep, delete_back_forw_dep): Ditto.
1197 * sched-rgn.c (set_spec_fed, find_conditional_protection, is_pfree):
1198 Update to use new scheduler dependencies lists.
1199 (is_conditionally_protected, is_prisky, add_branch_dependences): Ditto.
1200 (debug_dependencies): Ditto.
1201 (schedule_region): Update comments.
1203 * sched-ebb.c (earliest_block_with_similiar_load): Update to use new
1204 scheduler dependencies lists.
1205 (schedule_ebb): Update comments.
1207 * rtl.def (DEPS_LIST): Remove.
1209 * lists.c (unused_deps_list): Remove.
1210 (free_list): Update assertions.
1212 (alloc_DEPS_LIST, free_DEPS_LIST_list, free_DEPS_LIST_node): Remove.
1213 (remove_free_DEPS_LIST_elem, copy_DEPS_LIST_list): Ditto.
1215 * rtl.h (free_DEPS_LIST_list, alloc_DEPS_LIST): Remove declarations.
1216 (remove_free_DEPS_LIST_elem, copy_DEPS_LIST_list): Ditto.
1218 * haifa-sched.c (comments): Update.
1219 (insn_cost1): Remove. Inline the code into insn_cost ().
1220 (insn_cost): Update to use new scheduler dependencies lists. Move
1221 processing of the dependency cost to dep_cost ().
1222 (dep_cost): New function. Use it instead of insn_cost () when
1223 evaluating cost of the dependency. Use compatible interface to
1224 interact with the target.
1225 (priority): Update to use new scheduler dependencies lists.
1226 (rank_for_schedule): Ditto. Optimize heuristic that prefers the insn
1227 with greater number of insns that depend on the insn.
1228 (schedule_insn): Update to use new scheduler dependencies lists. Add
1229 code to free backward dependencies lists. Inline and optimize code
1230 from resolve_dep () - see PR28071.
1231 (ok_for_early_queue_removal): Update to use new scheduler dependencies
1232 lists. Update call to targetm.sched.is_costly_dependence hook.
1234 (fix_inter_tick, try_ready, fix_tick_ready): Update to use new
1235 scheduler dependencies lists.
1237 (resolve_dep): Remove. Move the logic to schedule_insn ().
1238 (init_h_i_d): Initialize dependencies lists.
1240 (process_insn_depend_be_in_spec): Rename to
1241 process_insn_forw_deps_be_in_spec. Update to use new scheduler
1243 (add_to_speculative_block, create_check_block_twin, fix_recovery_deps):
1244 Update to use new scheduler dependencies lists.
1245 (clear_priorities, calc_priorities, add_jump_dependencies): Ditto.
1247 * ddg.c (create_ddg_dependence, create_ddg_dep_no_link): Update to use
1248 new scheduler dependencies lists.
1249 (build_intra_loop_deps): Ditto.
1251 * target.h (struct _dep): Declare to use in
1252 gcc_target.sched.is_costly_dependence.
1253 (struct gcc_target.sched.adjust_cost): Fix typo.
1254 (struct gcc_target.sched.is_costly_dependence): Change signature to use
1255 single dep_t parameter instead of an equivalent triad.
1256 (struct gcc_target.sched.adjust_cost_2): Remove.
1258 * target-def.h (TARGET_SCHED_ADJUST_COST_2): Remove.
1260 * reg-notes.def (DEP_TRUE, DEP_OUTPUT, DEP_ANTI): Update comments.
1262 * doc/tm.texi (TARGET_SCHED_IS_COSTLY_DEPENDENCE): Update
1264 (TARGET_SCHED_ADJUST_COST_2): Remove documentation.
1266 * doc/rtl.texi (LOG_LINKS): Remove part about instruction scheduler.
1267 (REG_DEP_TRUE): Document.
1269 * config/ia64/ia64.c (ia64_adjust_cost_2): Rename to ia64_adjust_cost.
1270 Change signature to correspond to the targetm.sched.adjust_cost hook.
1271 Update use in TARGET_SCHED_ADJUST_COST_2.
1272 (TARGET_SCHED_ADJUST_COST_2): Rename to TARGET_SCHED_ADJUST_COST.
1273 (ia64_dependencies_evaluation_hook, ia64_dfa_new_cycle): Update to use
1274 new scheduler dependencies lists.
1275 (ia64_gen_check): Ditto.
1277 * config/mips/mips.c (vr4130_swap_insns_p): Update to use new scheduler
1280 * config/rs6000/rs6000.c (rs6000_is_costly_dependence): Change
1281 signature to correspond to the targetm.sched.is_costly_dependence hook.
1282 (is_costly_group): Update to use new scheduler dependencies lists.
1284 * config/spu/spu.c (spu_sched_adjust_cost): Use insn_cost () function
1285 instead of INSN_COST () macro.
1287 2007-02-01 Ian Lance Taylor <iant@google.com>
1289 * lower-subreg.c (resolve_clobber): Handle a subreg of a concatn.
1291 2007-02-01 Guy Martin <gmsoft@gentoo.org>
1293 * pa.md (tp_load): Correct mfctl instruction syntax.
1295 2007-02-01 Geoffrey Keating <geoffk@apple.com>
1297 * config/rs6000/rs6000.c (rs6000_stack_info): Correct
1298 altivec_padding_size calculation on AIX. Improve comment, add
1299 assert to verify that it's right.
1301 * config/rs6000/darwin.h (MD_UNWIND_SUPPORT): Don't define for
1304 2007-2-01 Seongbae Park <seongbae.park@gmail.com>
1307 * global.c (global_alloc): Add mising initialization of
1310 2007-02-01 Roger Sayle <roger@eyesopen.com>
1312 * alias.c (init_alias_analysis): Correct whitespace.
1313 * bb-reorder.c (fix_edges_for_rarely_executed_code,
1314 partition_hot_cold_basic_blocks): Likewise.
1315 * builtins.c (expand_builtin_printf, expand_builtin_fprintf,
1316 expand_builtin_sprintf, fold_builtin_carg, fold_builtin_sprintf,
1317 maybe_emit_sprintf_chk_warning, fold_builtin_sprintf_chk,
1318 fold_builtin_snprintf_chk, fold_builtin_printf,
1319 fold_builtin_fprintf, do_mpfr_ckconv, do_mpfr_arg1, do_mpfr_arg2,
1320 do_mpfr_arg3, do_mpfr_sincos): Likewise.
1321 * cfgcleanup.c (cleanup_cfg): Likewise.
1322 * cfgexpand.c (tree_expand_cfg): Likewise.
1323 * fold-const.c (fold_binary) <RDIV_EXPR>: Likewise.
1324 * function.c (get_next_funcdef_no): Likewise.
1325 * gengtype.c (main): Likewise.
1326 * genmodes.c (main): Likewise.
1327 * gcse.c (bypass_conditional_jumps, print_ldst_list): Likewise.
1328 * haifa-sched.c (schedule_block, extend_h_i_d): Likewise.
1329 * ifcvt.c (noce_emit_move_insn): Likewise.
1330 * modulo-sched.c (generate_prolog_epilog, sms_schedule_by_order):
1332 * stor-layout.c (get_best_mode): Likewise.
1333 * tree-ssa-loop-niter.c (get_val_for): Likewise.
1334 * tree-ssa-structalias.c (get_varinfo, get_varinfo_fc,
1335 scc_visit, do_ds_constraint, do_complex_constraint, label_visit,
1336 perform_var_substitution, solve_graph): Likewise.
1337 * tree-vrp.c (vrp_finalize): Likewise.
1339 2007-02-01 Ian Lance Taylor <iant@google.com>
1341 * lower-subreg.c (simplify_gen_subreg_concatn): If we ask for the
1342 high part of a paradoxical subreg, return a constant zero.
1344 2007-02-01 Zdenek Dvorak <dvorakz@suse.cz>
1346 * toplev.c (lang_dependent_init): Call init_set_costs.
1347 * loop-init.c (loop_optimizer_init): Do not call init_set_costs.
1349 2007-02-01 Richard Guenther <rguenther@suse.de>
1352 * fold-const.c (fold_negate_expr): Allow negating a
1353 constant if overflow does not change.
1355 2007-02-01 Ralf Wildenhues <Ralf.Wildenhues@gmx.de>
1357 * doc/c-tree.texi (Expression trees): Improve markup.
1358 * doc/tm.texi (Register Classes, Addressing Modes)
1359 (Floating Point): Fix spacing after abbreviations. Fix some
1362 2007-02-01 Ben Elliston <bje@au.ibm.com>
1364 * doc/invoke.texi: Replace "bugfix" with "bug fix" throughout.
1365 * doc/contrib.texi: Likewise.
1366 * doc/install.texi: Likewise.
1368 2007-01-31 Richard Henderson <rth@redhat.com>
1369 Ian Lance Taylor <iant@google.com>
1371 * lower-subreg.c: New file.
1372 * rtl.def (CONCATN): Define.
1373 * passes.c (init_optimization_passes): Add pass_lower_subreg and
1375 * emit-rtl.c (update_reg_offset): New static function, broken out
1376 of gen_rtx_REG_offset.
1377 (gen_rtx_REG_offset): Call update_reg_offset.
1378 (gen_reg_rtx_offset): New function.
1379 * regclass.c: Revert patch of 2006-03-05, restoring
1381 (clear_reg_info_regno): New function.
1382 * dwarf2out.c (concatn_loc_descriptor): New static function.
1383 (loc_descriptor): Handle CONCATN.
1384 * common.opt (fsplit_wide_types): New option.
1385 * opts.c (decode_options): Set flag_split_wide_types when
1387 * timevar.def (TV_LOWER_SUBREG): Define.
1388 * rtl.h (gen_reg_rtx_offset): Declare.
1389 (reg_scan_update): Declare.
1390 * regs.h (clear_reg_info_regno): Declare.
1391 * tree-pass.h (pass_lower_subreg): Declare.
1392 (pass_lower_subreg2): Declare.
1393 * doc/invoke.texi (Option Summary): List -fno-split-wide-types.
1394 (Optimize Options): Add -fsplit-wide-types to -O1 list. Document
1396 * doc/rtl.texi (Regs and Memory): Document concat and concatn.
1397 * Makefile.in (OBJS-common): Add lower-subreg.o.
1398 (lower-subreg.o): New target.
1400 2007-01-31 Kazu Hirata <kazu@codesourcery.com>
1402 * config/sh/sh.h (HAVE_SECONDARY_RELOADS): Remove.
1404 2007-01-31 Anatoly Sokolov <aesok@post.ru>
1407 * config/avr/avr.c (DWARF2_ADDR_SIZE): Define.
1409 2007-01-31 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1412 * builtins.c (fold_builtin_sqrt): Use MPFR for constant args.
1414 2007-01-31 Zdenek Dvorak <dvorakz@suse.cz>
1416 * cfgloop.h: Include vec-prim.h.
1417 (enum li_flags): Remove LI_ONLY_OLD.
1418 (loop_iterator): Changed.
1419 (fel_next, fel_init): Iterate over loop tree.
1420 (FOR_EACH_LOOP_BREAK): New macro.
1421 * loop-unswitch.c (unswitch_loops): Do not pass LI_ONLY_OLD to
1423 * tree-ssa-loop-unswitch.c (tree_ssa_unswitch_loops): Ditto.
1424 * modulo-sched.c (sms_schedule): Ditto.
1425 * tree-vectorizer.c (vectorize_loops): Ditto.
1426 * doc/loop.texi: Update information on loop numbering and behavior of
1427 FOR_EACH_LOOP wrto new loops.
1428 * tree-scalar-evolution.c (compute_overall_effect_of_inner_loop,
1429 add_to_evolution_1): Test nestedness of loops instead of comparing
1431 * tree-chrec.c (chrec_fold_plus_poly_poly,
1432 chrec_fold_multiply_poly_poly, chrec_evaluate,
1433 hide_evolution_in_other_loops_than_loop, chrec_component_in_loop_num,
1434 reset_evolution_in_loop): Ditto.
1435 * Makefile.in (CFGLOOP_H): Add vecprim.h dependency.
1437 2007-01-31 Dirk Mueller <dmueller@suse.de>
1439 * c-common.c (warn_about_parentheses): Separate warning about
1440 un-parenthized sequence of comparison operators from the one
1441 which is supposed to warn about x <= y <= z.
1443 2007-01-31 Uros Bizjak <ubizjak@gmail.com>
1445 * optabs.h (enum optab_index): Add new OTI_isinf.
1446 (isinf_optab): Define corresponding macro.
1447 * optabs.c (init_optabs): Initialize isinf_optab.
1448 * genopinit.c (optabs): Implement isinf_optab using isinf?f2
1450 * builtins.c (mathfn_built_in): Handle BUILT_IN_ISINF{,F,L}.
1451 (expand_builtin_interclass_mathfn): Expand BUILT_IN_ISINF{,F,L}
1453 (expand_builtin): Expand BUILT_IN_ISINF{,F,L} using
1454 expand_builtin_interclass_mathfn.
1455 * reg_stack.c (subst_stack_regs_pat): Handle UNSPEC_FXAM.
1456 * config/i386/i386.md (UNSPEC_FXAM): New constant.
1457 (fxam<mode>2_i387): New insn pattern.
1458 (isinf<mode>2) New expander to implement isinf, isinff and isinfl
1459 built-in functions as x87 inline asm.
1461 2007-01-31 Kazu Hirata <kazu@codesourcery.com>
1463 * gcc/config/arm/unwind-arm.h (_sleb128_t, _uleb128_t): New.
1465 2007-01-30 Eric Christopher <echristo@apple.com>
1467 * config.gcc: Add geode.
1469 2007-01-31 Kazu Hirata <kazu@codesourcery.com>
1471 * cgraphunit.c, config/arm/arm.c, config/m68k/m68k.c,
1472 ipa-inline.c, tree-profile.c, tree-ssa-live.c,
1473 tree-ssa-math-opts.c, tree-ssanames.c, tree-vect-analyze.c,
1474 value-prof.c: Fix comment typos.
1476 2007-01-30 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
1479 * doc/invoke.texi (Wpointer-arith): Document warning.
1481 2007-01-30 Janis Johnson <janis187@us.ibm.com>
1483 * doc/extend.texi (Decimal Floating Types): Remove decfloat.h from
1484 the list of discrepancies from the draft TR.
1486 2007-01-30 Dirk Mueller <dmueller@suse.de>
1489 * doc/invoke.texi (-Wreturn-type): Update description to
1492 2007-01-30 Richard Sandiford <richard@codesourcery.com>
1494 * cfgrtl.c (try_redirect_by_replacing_jump): Check only_sets_cc0_p.
1496 2007-01-30 Uros Bizjak <ubizjak@gmail.com>
1498 * builtins.c (expand_builtin_int_interclass_roundingfn): New function
1499 to handle optabs that operate on floating point input argument and
1500 output to integer output.
1501 (expand_builtin_mathfn) [BUILT_IN_ILOGB]: Move from here ...
1502 (expand_builtin_interclass_mathfn) [BUILT_IN_ILOGB]: ... to here.
1503 (expand_builtin): Expand BUILT_IN_ILOGB{,F,L} using
1504 expand_builtin_interclass_mathfn ().
1505 * config/i386/i386.md (fxtractxf3_i387): Rename from *fxtractxf3_i387.
1507 (ilogbxf2, ilogb<mode>2): New expanders to implement ilogb, ilogbf and
1508 ilogbl built-in functions as x87 intrinsics.
1510 2007-01-30 Richard Guenther <rguenther@suse.de>
1513 * dwarf2out.c (reference_to_unused): Query varpool if the
1514 variable was output.
1516 2007-01-30 Richard Guenther <rguenther@suse.de>
1519 * passes.c (execute_one_pass): Reset in_gimple_form to not
1520 confuse non-unit-at-a-time mode.
1522 2007-01-29 Roger Sayle <roger@eyesopen.com>
1523 Richard Guenther <rguenther@suse.de>
1525 * fold-const.c (round_up): Make HIGH an unsigned HOST_WIDE_INT to
1526 avoid undefined behaviour on overflow. Use force_fit_type_double
1527 to construct the constant with the specified TREE_OVERFLOW.
1529 2007-01-29 Janis Johnson <janis187@us.ibm.com>
1531 * config/dfp-bit.c: Add parameterized support for fp exceptions.
1532 * config/dfp-bit.h: Ditto.
1534 2007-01-29 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
1536 * c-decl.c (pop_scope): Replace warnings with call to
1537 warn_for_unused_label.
1538 * c-common.h (warn_for_unused_label): Declare.
1539 * c-common.c (warn_for_unused_label): Define.
1541 2007-01-29 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
1543 * tree-optimize.c (update_inlined_to_pointers): Delete unused
1546 2007-01-29 Janis Johnson <janis187@us.ibm.com>
1548 * Makefile.in (USER_H): Remove decfloat.h.
1549 * ginclude/decfloat.h: Delete, moving contents to ...
1550 * ginclude/float.h: Add support for decimal floating point,
1551 guarded by __STDC_WANT_DEC_FP__.
1553 2007-01-29 Mike Stump <mrs@apple.com>
1555 * doc/gccint.texi (Top): Rename Loop Representation to Loop
1556 Analysis and Representation to resolve case insensitive conflict.
1557 * doc/loop.texi (Loop Analysis and Representation): Likewise.
1559 2007-01-28 Daniel Berlin <dberlin@dberlin.org>
1561 * tree.h (struct tree_memory_tag): Add aliases member.
1562 (MTAG_ALIASES): New macro.
1563 * tree-ssa-alias.c (alias_bitmap_obstack): New variable.
1564 (add_may_alias): Remove pointer-set. Update for may_aliases being
1566 (mark_aliases_call_clobbered): Update for may_aliases being a
1568 (compute_tag_properties): Ditto.
1569 (create_partition_for): Ditto.
1570 (compute_memory_partitions): Ditto.
1571 (dump_may_aliases_for): Ditto.
1572 (is_aliased_with): Ditto.
1573 (add_may_alias_for_new_tag): Ditto.
1574 (rewrite_alias_set_for): Rewrite for may_aliases being a bitmap.
1575 (compute_is_aliased): New function.
1576 (compute_may_aliases): Call compute_is_aliased).
1577 (init_alias_info): Initialize alias_bitmap_obstack.
1578 (union_alias_set_into): New function.
1579 (compute_flow_sensitive_aliasing): Use union_aliases_into.
1580 (have_common_aliases_p): Rewrite to take two bitmaps and use
1582 (compute_flow_insensitive_aliasing): Stop using pointer-sets.
1584 (finalize_ref_all_pointers): Update for add_may_alias changes.
1585 (new_type_alias): Ditto.
1586 * tree-flow-inline.h (may_aliases): Return a bitmap.
1587 * tree-dfa.c (dump_variable): Check for MTAG_P'ness.
1588 * tree-ssa.c (verify_flow_insensitive_alias_info): Update for
1589 may_aliases being a bitmap.
1590 * tree-flow.h (struct var_ann_d): Remove may_aliases member.
1591 may_aliases now returns a bitmap.
1592 * tree-ssa-structalias.c (merge_smts_into): Update for may_aliases
1594 * tree-ssa-operands.c (add_virtual_operand): Update for
1595 may_aliases being a bitmap.
1597 2007-01-29 Daniel Berlin <dberlin@dberlin.org>
1599 PR tree-optimization/30630
1600 * tree-ssa-structalias.c (do_complex_constraint): Mark correct
1601 variable as changed.
1603 2007-01-29 Simon Martin <simartin@users.sourceforge.net>
1606 * gimplify.c (gimplify_target_expr): Make sure that the TARGET_EXPR is
1607 expanded only once even if an error occurs.
1609 2007-01-29 Ben Elliston <bje@au.ibm.com>
1611 * gcov-io.h (__gcov_indirect_call_profiler): Declare.
1612 (__gcov_average_profiler): Likewise.
1613 (__gcov_ior_profiler): Likewise.
1614 (__gcov_merge_ior): Likewise.
1616 2007-01-28 Jan Hubicka <jh@suse.cz>
1618 * builtins.c (expand_builtin_memset): Fix typo in my last patch.
1619 * value-prof.c (stringop_block_profile): Likewise.
1621 2007-01-28 Jan Hubicka <jh@suse.cz>
1623 * expr.c (emit_block_move_via_movmem, emit_block_move_via_libcall): Add
1624 variant handling histograms; add wrapper.
1625 (clear_storage_via_libcall): Export.
1626 (emit_block_move_hints): Break out from ...; add histograms.
1627 (emit_block_move): ... this one.
1628 (clear_storage_hints): Break out from ...; add histograms.
1629 (clear_storage): ... this one.
1630 (set_storage_via_memset): Handle histogram.
1631 * expr.h (emit_block_move_via_libcall, emit_block_move_hints): Declare.
1632 (clear_storage_hints, clear_storage_via_libcall): Declare.
1633 (set_storage_via_setmem): Update prototype.
1634 * doc/md.texi (movmem, setmem): Document new arguments.
1636 * value-prof.c (dump_histogram_value, tree_find_values_to_profile): Add
1638 (stringop_block_profile): New global function.
1639 (tree_stringops_values_to_profile): Profile block size and alignment.
1640 * value-prof.h (enum hist_type): add HIST_TYPE_AVERAGE and
1642 (struct profile_hooks): Add gen_average_profiler and gen_ior_profiler.
1643 (stringop_block_profile): Declare.
1644 * builtins.c: Include value-prof.h.
1645 (expand_builtin_memcpy, expand_builtin_memset): Pass block profile.
1646 * gcov-ui.h (GCOV_COUNTER_NAMES): Add new counter.
1647 (GCOV_COUNTER_AVERAGE, GCOV_COUNTER_IOR): New constants.
1648 (GCOV_COUNTERS, GCOV_LAST_VALUE_COUNTER): Update.
1649 * profile.c (instrument_values): Add new counters.
1650 * cfgexpand.c (expand_gimple_basic_block): Propagate histograms to
1652 * tree-profile.c (tree_average_profiler_fn, tree_ior_profiler_fn): New.
1653 (tree_init_edge_profiler): Build new profilers.
1654 (tree_gen_average_profiler, tree_gen_ior_profiler): New.
1655 (pass_tree_profile): Add dump.
1656 (tree_profile_hooks): Update.
1657 * Makefile.in (LIBGCOV): Add new constants.
1658 * libgcov.c (__gcov_merge_ior, __gcov_average_profiler,
1659 __gcov_ior_profiler): New.
1660 * i386.md (movmem/setmem expanders): Add new optional arguments.
1662 2007-01-28 David Edelsohn <edelsohn@gnu.org>
1664 * doc/md.texi (Standard Pattern Names): Document blockage pattern.
1666 2007-01-28 Zdenek Dvorak <dvorakz@suse.cz>
1668 * tree-ssa-loop-unswitch.c: Include tree-inline.h.
1669 (tree_unswitch_single_loop): Pass eni_size_weights to
1670 tree_num_loop_insns.
1671 * tree-ssa-loop-manip.c: Include tree-inline.h.
1672 (can_unroll_loop_p): Pass eni_size_weights to
1673 tree_num_loop_insns.
1674 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p):
1675 Pass eni_size_weights to estimate_num_insns.
1676 * tree.h (init_inline_once): Export.
1677 * toplev.c (backend_init): Call init_inline_once.
1678 * cgraphunit.c (cgraph_process_new_functions,
1679 cgraph_analyze_function): Pass eni_inlining_weights to
1681 * ipa-inline.c (compute_inline_parameters): Ditto.
1682 * tree-ssa-loop-ivcanon.c (tree_num_loop_insns): Pass weights
1683 to estimate_num_insns.
1684 (try_unroll_loop_completely): Pass eni_size_weights to
1685 tree_num_loop_insns.
1686 * tree-eh.c (decide_copy_try_finally): Pass eni_size_weights
1687 ot estimate_num_insns.
1688 * tree-ssa-loop-prefetch.c: Include tree-inline.h.
1689 (loop_prefetch_arrays): Pass eni_time_weights to tree_num_loop_insns.
1690 * tree-inline.c (eni_inlining_weights, eni_size_weights,
1691 eni_time_weights): New variables.
1692 (init_inline_once): Initialize them.
1693 (struct eni_data): Mew.
1694 (estimate_num_insns_1, estimate_num_insns): Use weights.
1695 * tree-inline.h (struct eni_weights_d): New.
1696 (eni_inlining_weights, eni_size_weights, eni_time_weights): Declare.
1697 (estimate_num_insns): Declaration changed.
1698 * cfgloop.h (tree_num_loop_insns): Declaration changed.
1699 * Makefile.in (tree-ssa-loop-unswitch.o, tree-ssa-loop-prefetch.o,
1700 tree-ssa-loop-manip.o): Add TREE_INLINE_H dependency.
1702 2007-01-28 Zdenek Dvorak <dvorakz@suse.cz>
1704 * tree-data-ref.c (conflict_fn): Assert that the number of affine
1705 relations in the conflict function is valid.
1707 2007-01-27 Ian Lance Taylor <iant@google.com>
1709 * common.opt: Add fstrict-overflow.
1710 * opts.c (decode_options): Set flag_strict_overflow if -O2.
1711 * flags.h (TYPE_OVERFLOW_WRAPS): Define.
1712 (TYPE_OVERFLOW_UNDEFINED): Define.
1713 (TYPE_OVERFLOW_TRAPS): Define. This replaces TYPE_TRAP_SIGNED.
1715 * tree.h (TYPE_TRAP_SIGNED): Don't define.
1716 * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_UNDEFINED.
1717 (fold_negate_expr): Likewise.
1718 (make_range): Likewise.
1719 (extract_muldiv_1): Likewise.
1720 (maybe_canonicalize_comparison): Likewise.
1721 (fold_comparison): Likewise.
1722 (fold_binary): Likewise.
1723 (tree_expr_nonnegative_p): Likewise.
1724 (tree_expr_nonzero_p): Likewise.
1725 * tree-vrp.c (compare_values): Likewise.
1726 (extract_range_from_binary_expr): Likewise.
1727 (extract_range_from_unary_expr): Likewise.
1728 * tree-ssa-loop-niter.c (infer_loop_bounds_from_signedness):
1730 (nowrap_type_p): Likewise.
1731 * tree-scalar-evolution.c (simple_iv): Likewise.
1732 * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_WRAPS.
1733 (build_range_check): Likewise.
1734 (extract_muldiv_1): Likewise.
1735 (fold_comparison): Likewise.
1736 * tree-vrp.c (vrp_int_const_binop): Likewise.
1737 (extract_range_from_unary_expr): Likewise.
1738 * convert.c (convert_to_integer): Likewise.
1739 * fold-const.c (fold_negate_expr): Use TYPE_OVERFLOW_TRAPS.
1740 (fold_comparison): Likewise.
1741 (fold_binary): Likewise.
1742 * optabs.c (optab_for_tree_code): Likewise.
1743 * tree-vectorizer.c (vect_is_simple_reduction): Likewise.
1744 * simplify-rtx.c (simplify_const_relational_operation): Check
1745 flag_strict_overflow and flag_trapv.
1746 (simplify_const_relational_operation): Likewise.
1747 * doc/invoke.texi (Option Summary): Mention -fstrict-overflow.
1748 (Optimize Options): Add -fstrict-overflow to -O2 list. Document
1751 2007-01-27 Roger Sayle <roger@eyesopen.com>
1753 * tree.c (tree_fold_gcd): Delete.
1754 * tree.h (tree_fold_gcd): Remove prototype.
1755 * tree-data-ref.c (tree_fold_divides_p): Don't use tree_fold_gcd to
1756 test whether one constant integer is a multiple of another. Instead
1757 call int_const_binop with TRUNC_MOD_EXPR and test for a zero result.
1758 * fold-const.c (multiple_of_p): We've determined both TOP and
1759 BOTTOM are integer constants so we can call int_const_binop directly
1760 instead of the more generic const_binop.
1762 2007-01-27 Roger Sayle <roger@eyesopen.com>
1764 * fold-const.c (size_binop): In the fast-paths for X+0, 0+X, X-0 and
1765 1*X check that the constant hasn't overflowed, to preserve the
1767 (round_up): Provide an efficient implementation when rouding-up an
1768 INTEGER_CST to a power-of-two.
1770 2007-01-28 Ralf Wildenhues <Ralf.Wildenhues@gmx.de>
1772 * doc/sourcebuild.texi: Add comma for clarity.
1773 * doc/extend.texi: Fix some typos.
1774 * doc/passes.texi: Likewise.
1775 * doc/cppinternals.texi: Likewise.
1776 * doc/c-tree.texi: Likewise.
1777 * doc/tree-ssa.texi: Likewise.
1778 * doc/install.texi: Likewise.
1780 2007-01-27 Jan Hubicka <jh@suse.cz>
1782 * tree-sra.c (sra_walk_function): Don't rely on aliases being build.
1783 (pass_sra): Do not require alias information.
1784 * passes.c (init_optimization_passes): Add SRA
1786 2007-01-27 Steven Bosscher <steven@gcc.gnu.org>
1788 * tracer.c (rest_of_handle_tracer): We already cleaned
1789 up the CFG in tracer() so don't do it here again.
1790 * cfgcleanup.c (rest_of_handle_jump2): Don't repeat
1791 cleanup_cfg here, either. And don't call renumber_insns.
1793 * cfgrtl.c (rtl_verify_flow_info_1): Don't verify that BB_END
1794 and BB_HEAD are in the insn stream here. Instead make sure
1795 that BB_INSN is valid on all insns. Also, do check here that
1796 there are no pending branch predictions...
1797 (rtl_verify_flow_info): ...instead of doing it here. Checks
1798 for BB_END and BB_HEAD moved from rtl_verify_flow_info_1 to
1801 2007-01-26 Roger Sayle <roger@eyesopen.com>
1803 * config/i386/i386.c (ix86_swap_binary_operands_p): New helper
1804 function to simplify/factorize operand order canonicalization.
1805 (ix86_fixup_binary_operands): Reorganize using the above function.
1806 (ix86_binary_operator_ok): Likewise.
1808 2007-01-27 Jakub Jelinek <jakub@redhat.com>
1810 * genattrtab.c (struct attr_value_list, insn_code_values): Move to
1811 file scope from optimize_attrs.
1812 (simplify_test_exp): If insn_code_values is not NULL, use it to speed
1814 (optimize_attrs): Clear insn_code_values after freeing it.
1816 2007-01-26 Zdenek Dvorak <dvorakz@suse.cz>
1818 * tree-ssa-address.c (create_mem_ref): Remove ", bsi" from
1819 a parts.base assignment.
1821 2007-01-26 Zdenek Dvorak <dvorakz@suse.cz>
1823 * tree-data-ref.c (dump_subscript): Use dump_conflict_function.
1824 (compute_subscript_distance, initialize_data_dependence_relation,
1825 finalize_ddr_dependent, analyze_ziv_subscript,
1826 analyze_siv_subscript_cst_affine,
1827 compute_overlap_steps_for_affine_univar,
1828 compute_overlap_steps_for_affine_1_2, analyze_subscript_affine_affine,
1829 analyze_siv_subscript, analyze_miv_subscript,
1830 analyze_overlapping_iterations, subscript_dependence_tester_1,
1831 compute_self_dependence, free_dependence_relation): Work
1832 with affine_fn instead of chrecs.
1833 (dump_affine_function, dump_conflict_function, affine_function_equal_p,
1834 common_affine_function, affine_function_base,
1835 affine_function_constant_p, affine_fn_op, affine_fn_plus,
1836 affine_fn_minus, affine_fn_free, conflict_fn_not_known,
1837 conflict_fn_no_dependence, free_conflict_function, free_subscripts,
1838 conflict_fn, affine_fn_cst, affine_fn_univar): New functions.
1839 (all_chrecs_equal_p): Removed.
1840 * tree-data-ref.h (affine_fn, conflict_function): New types.
1841 (struct subscript): Change type of conflicting_iterations_in_a
1842 and conflicting_iterations_in_b.
1844 2007-01-26 Steve Ellcey <sje@cup.hp.com>
1847 * config/pa/pa.h (TARGET_HPUX_11): New.
1848 * config/pa/pa-hpux11.h (TARGET_HPUX_11): New.
1849 * config/pa/pa.c (pa_init_builtins): Use TARGET_HPUX_11.
1851 2007-01-26 Daniel Berlin <dberlin@dberlin.org>
1852 Richard Guenther <rguenther@suse.de>
1854 * tree-ssa-structalias.c (solve_graph): Handle case
1855 we merged the variable to another.
1857 2007-01-25 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1859 * builtins.c (fold_builtin_1): Treat ccos and ccosh as 'even'
1862 * fold-const.c (negate_mathfn_p): Treat casin, casinh, catan,
1863 catanh, cproj, csin, csinh, ctan and ctanh as 'odd' functions.
1865 2007-01-25 DJ Delorie <dj@redhat.com>
1867 * config/m32c/m32c.c (m32c_cannot_change_mode_class): We don't
1868 allow changes to modes which don't fit in those registers.
1870 * reload1.c (choose_reload_regs): Check for invalid subregs before
1871 computing their locations, not after.
1873 2007-01-25 Geoffrey Keating <geoffk@apple.com>
1876 * config/rs6000/rs6000.c (first_altivec_reg_to_save): On Darwin,
1877 save Altivec registers in an eh_return function.
1878 (compute_vrsave_mask): Likewise.
1879 (rs6000_stack_info): Correct AIX/Darwin stack alignment computation
1880 for saving Altivec registers.
1881 (rs6000_emit_prologue): Don't allocate stack twice in
1882 eh_return function. Correct expected value of altivec_save_offset
1883 when using save_world. Describe save of R0 to stack when using
1884 save_world. Describe stack pointer adjustment when using
1885 save_world. Remove duplicated eh_return parameter register saving.
1886 Update sp_offset variable after save_world.
1887 * config/rs6000/t-darwin (LIB2FUNCS_STATIC_EXTRA): Remove
1889 (LIB2FUNCS_EXTRA): Add darwin-world.asm.
1890 * config/rs6000/darwin.h (SUBTARGET_OVERRIDE_OPTIONS): -m64
1893 2007-01-25 Steve Ellcey <sje@cup.hp.com>
1895 * config.gcc (ia64*-*-hpux*): Make posix threads the default.
1897 2007-01-25 Steve Ellcey <sje@cup.hp.com>
1900 * config/pa/pa.c (pa_init_builtins): Set asm names for finite routines.
1901 * config/ia64/ia64.c (ia64_init_builtins): Ditto.
1903 2007-01-25 Richard Guenther <rguenther@suse.de>
1905 * doc/invoke.texi (-Wcoverage-mismatch): Document.
1906 * common.opt (-Wcoverage-mismatch): New warning option.
1907 * coverage.c (get_coverage_counts): Ignore coverage mismatch
1908 if -Wcoverage-mismatch is given.
1910 2007-01-25 Razya Ladelsky <razya@il.ibm.com>
1912 * ipa-cp.c (ipcp_insert_stage, ipcp_driver): Support for SSA.
1913 (ipcp_driver): Change to static definition.
1914 Add dumping of the ifunctions.
1915 (constant_val_insert): Remove unused parameter. Support for SSA.
1916 (ipcp_propagate_const): Support for SSA.
1917 (ipcp_profile_bb_print): Print only analyzed nodes.
1918 (ipcp_replace_map_create): Remove support for Fortran constant
1920 * ipa-prop.c (ipa_method_modify_stmt,
1921 ipa_callsite_compute_param): Support for SSA.
1922 * ipa-prop.h (ipcp_driver): Remove declaration.
1923 (IS_VALID_TREE_MAP_INDEX): Add define.
1925 2007-01-24 Geoffrey Keating <geoffk@apple.com>
1927 * unwind-dw2.c (execute_stack_op): Handle DW_OP_swap.
1929 2007-01-24 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1932 * builtins.c (fold_builtin_cabs): Use MPFR to evaluate a
1933 constant argument to cabs and do it without checking for
1934 -funsafe-math-optimizations.
1936 2007-01-24 Douglas Gregor <dgregor@osl.iu.edu>
1938 * c-common.h (RID_FIRST_CXX0X): New.
1939 (RID_LAST_CXX0X): New.
1940 * c-opts.c (c_common_handle_option): -Wc++0x-compat is triggered
1942 * c.opt (Wc++0x-compat): New.
1943 * doc/invoke.texi (-Wc++0x-compat): Document.
1945 2007-01-24 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1947 * builtins.c (fold_builtin_carg): New.
1948 (fold_builtin_1): Use it.
1950 2007-01-24 Jan Hubicka <jh@suse.cz>
1952 * ipa-inline.c (cgraph_decide_inlining): Initialize initial_insns.
1954 * ipa-inline.c (initial_insns, max_insns): Delete.
1955 (compute_max_insns): New function.
1956 (cgraph_decide_inlining_of_small_function): Use it; take minimal amount
1957 of insns as base for code growth.
1958 (cgraph_decide_inlining): Make initial_insns local; do not compute
1960 * params.def (PARAM_INLINE_UNIT_GROWTH): Set to 60.
1961 * doc/invoke.texi (inline-unit-growth): Update docs.
1963 2007-01-24 Jakub Jelinek <jakub@redhat.com>
1965 * config/i386/i386.h (x86_cmpxchg16b): Remove const.
1966 (TARGET_CMPXCHG16B): Define to x86_cmpxchg16b.
1967 * config/i386/i386.c (x86_cmpxchg16b): Remove const.
1968 (override_options): Add PTA_CX16 flag. Set x86_cmpxchg16b
1969 for CPUs that have PTA_CX16 set.
1972 * gimplify.c (omp_check_private): New function.
1973 (gimplify_scan_omp_clauses): Use it for
1974 firstprivate/lastprivate/reduction.
1977 * gimplify.c (omp_add_variable): Don't call omp_notice_variable
1978 on TYPE_SIZE_UNIT for GOVD_LOCAL VLAs.
1981 * omp-low.c (lower_omp_for_lastprivate): Add dlist argument.
1982 If lower_lastprivate_clauses emits some statements, append them
1983 to dlist rather than body_p and to body_p append an initializer.
1984 (lower_omp_for): Adjust caller.
1986 2007-01-24 Steve Ellcey <sje@cup.hp.com>
1988 * target.h (globalize_decl_name): New.
1989 * target-def.h (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
1990 * output.h (default_globalize_decl_name): New.
1991 * varasm.c (asm_output_bss): Use globalize_decl_name instead of
1993 (globalize_decl): Ditto.
1994 (default_globalize_decl_name): New.
1995 * config/ia64/ia64.c (ia64_globalize_decl_name): New.
1996 (ia64_handle_version_id_attribute): New.
1997 (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
1998 (ia64_asm_output_external): Use globalize_decl_name instead
2000 * doc/extend.texi (version_id): New pragma.
2001 * doc/tm.texi (ARGET_ASM_GLOBALIZE_DECL_NAME): New target hook.
2003 2007-01-24 Andreas Krebbel <krebbel1@de.ibm.com>
2005 * unwind-dw2-fde.c (get_cie_encoding): Replaced _Unwind_Word with
2006 _uleb128_t and _Unwind_SWord with _sleb128_t.
2007 * unwind-dw2.c (extract_cie_info, execute_stack_op, execute_cfa_program,
2008 uw_frame_state_for, uw_update_context_1): Likewise.
2009 * unwind-c.c (parse_lsda_header, PERSONALITY_FUNCTION): Likewise.
2010 * unwind-pe.h (read_uleb128, read_sleb128,
2011 read_encoded_value_with_base): Likewise.
2012 * unwind-generic.h: Define _sleb128_t and _uleb128_t types.
2014 2007-01-24 Richard Guenther <rguenther@suse.de>
2016 * builtins.c (expand_builtin_cexpi): Get the fndecl
2017 for cexp in the correct way.
2019 2007-01-24 Jan Hubicka <jh@suse.cz>
2021 * tree-ssa-dce.c (eliminate_unnecesary_stmts): Remove dead LHS of calls.
2023 2007-01-24 Andreas Krebbel <krebbel1@de.ibm.com>
2025 * c-cppbuiltin.c (builtin_define_type_sizeof): New function.
2026 (c_cpp_builtins): New builtin macros: __SIZEOF_INT__, __SIZEOF_LONG__,
2027 __SIZEOF_LONG_LONG__, __SIZEOF_SHORT__, __SIZEOF_POINTER__,
2028 __SIZEOF_FLOAT__, __SIZEOF_DOUBLE__, __SIZEOF_LONG_DOUBLE__,
2029 __SIZEOF_SIZE_T__, __SIZEOF_WCHAR_T__, __SIZEOF_WINT_T__ and
2030 __SIZEOF_PTRDIFF_T__.
2031 * doc/cpp.texi: Documentation for the new builtin macros added.
2033 2007-01-24 Uros Bizjak <ubizjak@gmail.com>
2035 * config/i386/i386.md (tanxf2, tan<mode>2, atan<mode>2, log<mode>2,
2036 log10<mode>2, log2<mode>2, expxf2, exp10xf2, exp2xf2): Use op2
2037 instead of operands[2] to avoid access past the end of array.
2039 2007-01-24 Richard Sandiford <richard@codesourcery.com>
2041 * reload1.c (emit_reload_insns): Pass the reload register
2042 for a non-spill output reload through forget_old_reloads_1.
2044 2007-01-23 Joseph Myers <joseph@codesourcery.com>
2046 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Define
2047 _SOFT_DOUBLE if doubles use software floating-point.
2048 * config/rs6000/libgcc-ppc-glibc.ver: Export additional long
2049 double functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
2050 * config/rs6000/darwin-ldouble.c: Also compile functions for
2051 hard-float without FPRs. Use fmsub function for all __NO_FPRS__
2052 cases. Compile extra functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
2053 * config/rs6000/linuxspe.h (SUBSUBTARGET_OVERRIDE_OPTIONS): Remove
2054 commented-out long double override.
2055 (CPP_LONGDOUBLE_DEFAULT_SPEC): Likewise.
2056 * config/rs6000/eabispe.h: Likewise.
2057 * config/rs6000/rs6000.c (rs6000_override_options): Don't override
2058 long double for non-SPE.
2059 (rs6000_handle_option): Likewise.
2060 (invalid_e500_subreg): Disallow more subregs involding DImode,
2061 DFmode, TImode or TFmode.
2062 (rs6000_legitimate_offset_address_p): Check TFmode offsets for
2064 (legitimate_lo_sum_address_p): Also check for TFmode for E500
2066 (rs6000_legitimize_address): Also handle TFmode for E500 double.
2067 (rs6000_legitimize_reload_address): Also handle TFmode for E500
2069 (rs6000_legitimate_address): Also check for TFmode for E500
2071 (rs6000_emit_move): Use DFmode subregs of TFmode for E500 double.
2072 (spe_build_register_parallel): Handle TFmode and TCmode.
2073 (rs6000_spe_function_arg): Handle TFmode and TCmode for E500
2075 (function_arg): Handle TFmode and TCmode for E500 double.
2076 (rs6000_init_libfuncs): Initialize extra libfuncs for soft double
2078 (print_operand): Handle TFmode and TImode for %y.
2079 (rs6000_generate_compare): Handle TFmode comparisons for E500
2081 (spe_func_has_64bit_regs_p): Check for TFmode for E500 double.
2082 (rs6000_function_value): Handle TFmode and TCmode for E500 double.
2083 (rs6000_libcall_value): Handle TFmode and TCmode for E500 double.
2084 * config/rs6000/rs6000.h (CANNOT_CHANGE_MODE_CLASS): Check for
2085 TFmode for E500 double.
2086 * config/rs6000/rs6000.md (FP): Allow TF for E500 double.
2087 (floatsidf2): Enable for E500 double.
2088 (movtf_softfloat): Use rs6000_nonimmediate_operand.
2089 (extenddftf2): Change to extenddftf2_fprs.
2090 (extenddftf2): Call gen_spe_extenddftf2 or gen_extenddftf2_fprs
2091 depending on TARGET_E500_DOUBLE.
2092 (extendsftf2): Enable for E500 double.
2093 (trunctfdf2): Enable for E500 double.
2094 (trunctfsf2): Change to trunctfsf2_fprs.
2095 (trunctfsf2): Call gen_spe_trunctfsf2 or gen_trunctfsf2_fprs
2096 depending on TARGET_E500_DOUBLE.
2097 (floatsitf2): Enable for E500 double.
2098 (fix_trunctfsi2): Change to fix_trunctfsi2_fprs.
2099 (fix_trunctfsi2): Call gen_spe_fix_trunctfsi2 or
2100 gen_fix_trunctfsi2_fprs depending on TARGET_E500_DOUBLE.
2101 (negtf2): Change to negtf2_internal.
2102 (negtf2): New expander.
2103 (abstf2): Enable for E500 double. Call gen_spe_abstf2_tst,
2104 gen_spe_abstf2_cmp or gen_abstf2_internal depending on
2105 TARGET_E500_DOUBLE and flag_unsafe_math_optimizations.
2106 (movdi_internal32): Use rs6000_nonimmediate_operand.
2107 (unnamed splitter): Likewise.
2108 * config/rs6000/spe.md (CMPTFEQ_GPR, TSTTFEQ_GPR, CMPTFGT_GPR,
2109 TSTTFGT_GPR, CMPTFLT_GPR, TSTTFLT_GPR): New unspecs.
2110 (SPE64TF, DITI): New mode macros.
2111 (frob_df_di): Change to frob_<SPE64:mode>_<DITI:mode>; allow more
2114 (frob_<mode>_di_2): New.
2115 (frob_tf_di_8_2): New.
2116 (frob_di_df): Change to frob_di_<mode>; allow more modes.
2118 (frob_di_df_2): Change to frob_<DITI:mode>_<SPE64:mode>_2; allow
2120 (frob_ti_<mode>_8_2): New.
2121 (frob_ti_tf_2): New.
2122 (mov_si<mode>_e500_subreg0, mov_si<mode>_e500_subreg0_2,
2123 mov_si<mode>_e500_subreg4, mov_si<mode>_e500_subreg4_2): Allow
2125 (mov_sitf_e500_subreg8, mov_sitf_e500_subreg8_2,
2126 mov_sitf_e500_subreg12, mov_sitf_e500_subreg12_2): New.
2127 (spe_trunctfdf2_internal1, spe_trunctfsf2, spe_extenddftf2,
2128 spe_fix_trunctfsi2, spe_fix_trunctfsi2_internal,
2129 spe_negtf2_internal, spe_abstf2_cmp, spe_abstf2_tst): New.
2130 (cmptfeq_gpr, tsttfeq_gpr, cmptfgt_gpr, tsttfgt_gpr, cmptflt_gpr,
2133 2007-01-23 Ian Lance Taylor <iant@google.com>
2135 * Makefile.in (OBJS-common): Reformat, alphabetize, but put
2137 (OBJS-archive): Reformat, alphabetize.
2138 (OBJS): Change out_object_file to OBJS-md.
2140 2007-01-23 Uros Bizjak <ubizjak@gmail.com>
2142 * config/i386/i386.md: Use REG_P, MEM_P, CONST_INT_P, LABEL_P,
2143 JUMP_P and CALL_P predicates where applicable.
2144 * config/i386/i386.c: Ditto.
2145 * config/i386/i386.md: Ditto.
2146 * config/i386/mmx.md: Ditto.
2147 * config/i386/predicates.md: Ditto.
2149 2007-01-22 Andreas Schwab <schwab@suse.de>
2151 * config/m68k/m68k.h: Fix comment.
2153 2007-01-22 Jan Hubicka <jh@suse.cz>
2155 * passes.c (init_optimization_passes): Do not rerun
2156 pass_early_warn_uninitialized.
2158 2007-01-22 Richard Guenther <rguenther@suse.de>
2160 PR tree-optimization/30038
2161 * tree-ssa-math-opts.c (maybe_record_sincos): New static helper
2163 (execute_cse_sincos_1): Likewise.
2164 (execute_cse_sincos): Likewise.
2165 (gate_cse_sincos): Likewise.
2166 (pass_cse_sincos): New pass CSEing sin() and cos() calls using
2167 the cexpi() canonicalization of sincos().
2168 * tree-pass.h (pass_cse_sincos): Declare.
2169 * passes.c (init_optimization_passes): New pass pas_cse_sincos.
2171 2007-01-21 Eric Botcazou <ebotcazou@libertysurf.fr>
2173 PR rtl-optimization/29329
2174 * combine.c (replaced_rhs_insn): Rename to i2mod.
2175 (replaced_rhs_value): Rename to i2mod_new_rhs.
2176 (i2mod_old_rhs): New global variable.
2177 (combine_instructions): Adjust for above change. Save a copy of
2178 the old RHS into i2mod_old_rhs when the contents of a REG_EQUAL
2179 note are substituted in the second instruction.
2180 (distribute_notes) <REG_DEAD>: Adjust for above change. Do not
2181 ditch the note if it pertains to the second eliminated register
2182 and this register is mentioned in i2mod_old_rhs.
2185 2006-09-12 Eric Botcazou <ebotcazou@libertysurf.fr>
2187 * combine.c (distribute_notes) <REG_DEAD>: Do not consider SETs past
2188 the insn to which the note was originally attached.
2190 2007-01-21 Jan Hubicka <jh@suse.cz>
2192 * ipa-inline.c (inlining_mode): Comment, move up.
2193 (cgraph_decide_inlining_incrementally): Do not perform inlining
2194 itself; fix handling of flattening of self recursive functions.
2195 (cgraph_find_cycles): Remove.
2196 (cgraph_flatten_node): Remove.
2197 (cgraph_decide_inlining): Use incremental inliner to handle flattening.
2198 (try_inline): New function.
2199 (cgraph_early_inlining): Update call of
2200 cgraph_decide_inlining_incrementally. Apply inlining here.
2201 (apply_inline): Update call of cgraph_decide_inlining_incrementally.
2203 2007-01-21 Dirk Mueller <dmueller@suse.de>
2206 * tree-vrp.c (check_array_bounds): do not warn
2207 about ADDR_EXPR's of ARRAY_REF's which are immediately
2208 used in binary expressions.
2210 2007-01-21 Ira Rosen <irar@il.ibm.com>
2212 * tree-vectorizer.h (struct _stmt_vec_info): Add new field
2213 read_write_dep and macros for its access.
2214 * tree-vectorizer.c (new_stmt_vec_info): Initialize the new field.
2215 * tree-vect-analyze.c (vect_analyze_data_ref_dependence): Remove
2216 argument, call vect_check_interleaving for every independent pair of
2217 data-refs. Mark loads that access the same memory location as a store
2219 (vect_check_dependences): Remove.
2220 (vect_analyze_data_ref_dependences): Remove vect_check_dependences
2221 call, fix the call to vect_analyze_data_ref_dependence.
2222 (vect_analyze_data_ref_access): For statements that access the same
2223 data-ref, check that they are not stores; for loads, check that there
2224 is no store that access the same location.
2226 2007-01-20 Roger Sayle <roger@eyesopen.com>
2227 Joseph Myers <joseph@codesourcery.com>
2229 * doc/invoke.texi (-fdump-rtl-combine): Fix under/overfull hbox.
2230 (-fdump-rtl-stack): Likewise.
2231 (-fno-signed-zeros): Use @minus{} for a minus sign. Correct typo.
2232 (-mcheck-zero-division, -mcpu): Fix under/overfull hbox.
2233 (-mpt-fixed): Use @minus{} for minus sign.
2234 (Using Precompiled Headers): Fix under/overfull hbox.
2236 2007-01-20 Jan Hubicka <jh@suse.cz>
2238 * tree-flow.h (struct stmt_ann_d): Move references_memory to proper
2239 place within annotation.
2241 2007-01-20 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
2243 * pa.c (output_move_double): Change array size of xoperands to 4.
2245 2007-01-20 Jan Hubicka <jh@suse.cz>
2247 * tree-tailcall.c (adjust_return_value): Do not use RESULT_DECL
2250 2007-01-19 Ian Lance Taylor <iant@google.com>
2252 * expmed.c (expand_divmod) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]: Cast
2253 constant to unsigned HOST_WIDE_INT before negating.
2255 2007-01-19 Ian Lance Taylor <iant@google.com>
2257 * tree-ssa-operands.h (struct vuse_vec_d): Change num_vuse field
2259 (VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Use unsigned comparison.
2260 (VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
2261 (SET_VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Likewise.
2262 (SET_VUSE_ELEMENT_VAR) [ENABLE_CHECKING]: Likewise.
2263 (SET_VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
2264 (realloc_vdef, realloc_vuse): Change second parameter to
2266 (ssa_operand_iterator_d): Change vuse_index and mayuse_index
2268 * tree-ssa-operands.c (realloc_vop): Change num_elem parameter to
2269 unsigned. Change x and lim locals to unsigned.
2270 (realloc_vdef, realloc_vuse): Change num_elem parameter to
2272 (finalize_ssa_vuse_ops): Change old_i local to unsigned.
2273 (copy_virtual_operands): Change i and n locals to unsigned.
2275 2007-01-19 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
2278 * builtins.c (fold_builtin_1): Handle builtin fdim.
2280 2007-01-20 Jan Hubicka <jh@suse.cz>
2282 * tree-ssa.c (init_tree_ssa): Do not call init_alias_heapvars.
2283 * tree-ssa-structalias.c (compute_points_to_sets): Do call
2284 init_alias_heapvars.
2285 (init_alias_heapvars): Initialize only when not already initialized.
2286 (delete_alias_heapvars): Set heapvar_for_stmt to NULL.
2288 2007-01-19 Roger Sayle <roger@eyesopen.com>
2290 * common.opt (fsigned-zeros): New command line option.
2291 * flags.h (HONOR_SIGNED_ZEROS): Control via flag_signed_zeros instead
2292 of flag_unsafe_math_optimizations.
2293 * opts.c (set_fast_math_flags): The -ffast-math command line option
2294 implies -fno-signed-zeros.
2295 (fast_math_flags_set_p): Likewise.
2297 * doc/invoke.texi: Document new -fno-signed-zeros option, and update
2298 the documentation of -ffast-math appropriately. Wrap long lines.
2300 2007-01-19 Steve Ellcey <sje@cup.hp.com>
2302 * system.h (ASM_MAKE_LABEL_LINKONCE): Poison.
2303 * varasm.c (globalize_decl): Remove ASM_MAKE_LABEL_LINKONCE ifdef.
2305 2007-01-19 Tomas Bily <tbily@suse.cz>
2307 * cgraphunit.c (cgraph_finalize_function): Updating of pid
2309 (tree_init_ic_make_global_vars): New function
2310 (tree_init_edge_profiler): call of tree_init_ic_make_global_vars
2311 (tree_gen_ic_profiler): New function
2312 (tree_gen_ic_func_profiler): New function
2313 (tree_profiling): Added calling of tree_gen_ic_func_profiler
2314 (tree_profile_hooks): Added hook for indirec/virtual calls
2315 * value-prof.c (tree_find_values_to_profile): New case for
2317 (tree_values_to_profile): Call for determining indirect/virtual
2319 (tree_indirect_call_to_profile): New function
2320 (tree_ic_transform): New function
2321 (tree_ic): New function
2322 (find_func_by_pid): New function
2323 (init_pid_map): New function
2324 (tree_value_profile_transformations): Added check for
2325 indirect/virtual call transformation
2326 * value-prof.h (enum hist_type): New counter type for
2327 indirect/virtual calls
2328 (profile_hooks): Added new hook for profiling indirect/virtual
2330 * profile.c (instrument_values): New case for indirect/virtual
2332 * gcov-io.h (GCOV_LAST_VALUE_COUNTER): Changed to 6
2333 (GCOV_COUNTER_V_INDIR): New counter type
2334 (GCOV_COUNTER_NAMES): New name of counter "indirect" added
2335 (GCOV_MERGE_FUNCTIONS): New merge function for indirect/virtual
2337 * cgraph.c: Definition of cgraph_max_pid
2338 (cgraph_create_node): Default init of pid attribute
2339 * cgraph.h: Declaration of cgraph_max_pid
2340 (struct cgraph_node): Added pid attribute
2341 * libgcov.c (__gcov_indirect_call_profiler): New function
2342 (__gcov_one_value_profiler_body): New function
2343 (__gcov_one_value_profiler): Body was moved to
2344 __gcov_one_value_profiler_body and calls it
2346 2007-01-19 Basile Starynkevitch <basile@starynkevitch.net>
2348 * doc/gty.texi (Options): Document the mark_hook option to GTY.
2349 * gengtype.c (write_types_data, write_func_for_structure,
2350 write_types, ggc_wtd, pch_wtd): Add skip_hooks to
2351 write_types_data, ggc_wtd, pch_wtd for processing mark_hook.
2352 (walk_type, write_func_for_structure): Generate the mark_hook if
2355 2007-01-19 Jan Hubicka <jh@suse.cz>
2357 * ipa-inline.c (cgraph_decide_inlining_incrementally): Instead of
2358 'early' argument take inlining mode argument specifying whether to
2359 inline for size/speeed or all functions; add support for flattening;
2361 (cgraph_early_inlining): Update call of decide_inlining_incrementally.
2363 2007-01-19 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
2366 * toplev.c (warn_deprecated_use): Use %qD instead of %qs to print
2367 the name of the declared identifier.
2369 2007-01-19 Dirk Mueller <dmueller@suse.de>
2371 * config/i386.h (CONDITIONAL_REGISTER_USAGE): Store
2372 result of PIC_OFFSET_TABLE_REGNUM in temporary variable to avoid
2373 duplicate evaluation.
2375 2007-01-19 Uros Bizjak <ubizjak@gmail.com>
2377 * config/i386/i386.md (acos<mode>2): Rename from acossf2 and acosdf2.
2378 Macroize expander using X87MODEF12 mode macro. Extend operand 1
2379 to XFMode, use acosxf2 and truncate result to requested mode.
2380 Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
2381 (asin<mode>2): Similarly, with asin expanders.
2382 (*fscalexf4_i387): Rename from *fscalexf4.
2383 (expNcorexf3): New expander.
2384 (expxf2, exp10xf2, exp2xf2): Use expNcorexf3 expander.
2385 (exp<mode>2): Rename from expsf2 and expdf2. Macroize expander using
2386 X87MODEF12 mode macro. Extend operand 1 to XFMode, use expxf2 and
2387 truncate result to requested mode. Use SSE_FLOAT_MODE_P to disable
2388 patterns for SSE math.
2389 (exp10<mode>2): Similarly, with exp10 expanders.
2390 (exp2<mode>2): Similarly, with exp2 expanders.
2391 (expm1<mode>2): Similarly, with expm1 expanders.
2392 (ldexp<mode>3): Similarly, with ldexp expanders.
2393 (log<mode>2, log10<mode>2, log2<mode>2, log1p<mode>2, logb<mode>2):
2394 Use gen_truncxf<mode>2_i387_noop to truncate result.
2396 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2398 * config/m68k/m68k.h (PREFERRED_STACK_BOUNDARY): Define to 32
2399 for ColdFire targets.
2401 2007-01-19 Nathan Sidwell <nathan@codesourcery.com>
2402 Richard Sandiford <richard@codesourcery.com>
2404 * config/m68k/m68k.h (M68K_STATIC_CHAIN_REG_NAME): New macro.
2405 (INITIALIZE_TRAMPOLINE): Use STATIC_CHAIN_REGNUM.
2406 (__transfer_from_trampoline): Use M68K_STATIC_CHAIN_REG_NAME.
2407 * config/m68k/m68kelf.h (STATIC_CHAIN_REGNUM): Override.
2408 (M68K_STATIC_CHAIN_REG_NAME): Likewise.
2409 * config/m68k/netbsd-elf.h (M68K_STATIC_CHAIN_REG_NAME): Likewise.
2411 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2413 * config/m68k/m68k.md (adddi_dilshr32): Rename to...
2414 (*adddi_dilshr32): ...this. Fix formatting. Remove commented-out
2415 non-canonical pattern. Restrict to !TARGET_COLDFIRE.
2416 (*adddi_dilshr32_cf): New pattern.
2417 (adddi3, subdi3): Remove first alternatives.
2419 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2421 * config/m68k/m68k.c (notice_update_cc): If an SFmode move is
2422 implemented using move.l, do not use its cc result for floating-point
2425 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2427 * config/m68k/m68k.h (EXTRA_CONSTRAINT): Stop the 'T' constraint
2428 from accepting 's' constraints if flag_pic.
2430 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2432 * config/m68k/m68k.md (bordered, bunordered, buneq, bunge, bungt)
2433 (bunle, bunlt, bltgt, bordered_rev, bunordered_rev, buneq_rev)
2434 (bunge_rev, bungt_rev, bunle_rev, bunlt_rev, bltgt_rev): Change
2435 condition from TARGET_68881 to TARGET_HARD_FLOAT.
2437 2007-01-19 Sandra Loosemore <sandra@codesourcery.com>
2439 * longlong.h (count_leading_zeros, COUNT_LEADING_ZEROS_0): Add
2440 ColdFire alternatives.
2441 * config/m68k/m68k.h (CLZ_DEFINED_VALUE_AT_ZERO): New macro.
2442 * config/m68k/m68k.md (clzsi2): Define for ColdFire
2443 architectures that support ff1 instruction.
2445 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2446 Julian Brown <julian@codesourcery.com>
2448 * config/m68k/m68k.h (CONST_OK_FOR_LETTER_P): Add an 'R' case.
2449 * config/m68k/m68k.md (*movsi_cfv4): Fold into...
2450 (*movsi_cf): ...here. Remove unnecessary 'R' from 'Rg'.
2453 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2455 * config/m68k/m68k-protos.h (valid_mov3q_const): Take a HOST_WIDE_INT
2457 (output_move_const_into_data_reg, output_move_simode_const): Delete.
2458 * config/m68k/m68k.c (const_method, const_int_cost): Take a
2459 HOST_WIDE_INT instead of an rtx.
2460 (m68k_rtx_costs): Update call accordingly.
2461 (output_move_const_into_data_reg): Likewise. Fix formatting.
2462 (valid_mov3q_const): Take a HOST_WIDE_INT instead of an rtx.
2464 (output_move_simode_const): Update calls after above changes.
2465 Rework to use automatic variables and predicates like MEM_P.
2466 * config/m68k/m68k.md (pushexthisi_const): Update call to
2469 2007-01-19 Dirk Mueller <dmueller@suse.de>
2471 * tree-ssa-alias.c (perform_var_substitution): Fix typo
2474 2007-01-19 Richard Guenther <rguenther@suse.de>
2476 * builtins.c (expand_builtin_cexpi): Fall back to expanding
2477 via cexp in case sincos is not available.
2479 2007-01-19 Richard Guenther <rguenther@suse.de>
2481 * doc/tm.texi (TARGET_HAS_SINCOS): Document new target macro.
2482 * defaults.h (TARGET_HAS_SINCOS): Default to off.
2483 * config/linux.h (TARGET_HAS_SINCOS): Set to on if we have glibc.
2484 * config/alpha/linux.h (TARGET_HAS_SINCOS): Likewise.
2485 * config/sparc/linux.h (TARGET_HAS_SINCOS): Likewise.
2486 * config/sparc/linux64.h (TARGET_HAS_SINCOS): Likewise.
2487 * config/rs6000/linux.h (TARGET_HAS_SINCOS): Likewise.
2488 * config/rs6000/linux64.h (TARGET_HAS_SINCOS): Likewise.
2490 2007-01-19 Uros Bizjak <ubizjak@gmail.com>
2492 * config/i386/i386.md (*fpatanxf3_i387, fpatan_extend<mode>xf3_i387):
2494 (atan2sf3_1, atan2df3_1, atan2xf3_1): Remove insn patterns.
2495 (atan2xf3): Directly generate RTL pattern.
2496 (atan2<mode>3): Rename from atan2sf3 and atan2df3 and macroize insn
2497 patterns using X87MODEF12 mode macro. Use fpatan_extend<mode>xf3_i387
2498 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
2499 disable patterns for SSE math.
2500 (atan<mode>2): Rename from atansf2 and atandf2 and macroize insn
2501 patterns using X87MODEF12 mode macro. Use fpatan_extend<mode>xf3_i387
2502 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
2503 disable patterns for SSE math.
2505 2007-01-19 Alexandre Oliva <aoliva@redhat.com>
2507 * libgcc-std.ver: Fix typo in %inherit for GCC_4.3.0.
2509 2007-01-18 Roger Sayle <roger@eyesopen.com>
2511 * fold-const.c (fold_unary) <VIEW_CONVERT_EXPR>: Optimize away a
2512 VIEW_CONVERT_EXPR to the same type as it's operand.
2514 2007-01-18 David Edelsohn <edelsohn@gnu.org>
2516 * config/rs6000/darwin-ldouble.c: Only build _SOFT_FLOAT if
2517 configured for long double 128.
2519 2007-01-18 Mike Stump <mrs@apple.com>
2521 * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Fix build
2524 2007-01-18 Michael Meissner <michael.meissner@amd.com>
2526 * i386.c (ix86_compute_frame_layout): Make fprintf's in #if 0 code
2529 2007-01-18 Jan Hubicka <jh@suse.cz>
2531 * tree-ssa-operands.c (vop_free_bucket_size): Never return value
2532 greater than NUM_VOP_FREE_BUCKETS.
2534 2007-01-18 Daniel Berlin <dberlin@dberlin.org>
2536 * tree-ssa-structalias.c: Update comments.
2537 (ptabitmap_obstack): Removed.
2539 (oldpta_obstack): Ditto.
2540 (stats): Add a few members.
2541 (struct variable_info): Remove node, complex, address_taken, and
2542 indirect_target members. Add oldsolution member.
2543 (new_var_info): Do not initialize removed members.
2544 (constraint_expr_type): Remove INCLUDES.
2545 (constraint_graph): Add size, implicit_preds, rep,
2546 indirect_cycles, eq_rep, label, direct_nodes, and complex members.
2547 (FIRST_REF_NODE): New macro.
2548 (LAST_REF_NODE): Ditto.
2549 (FIRST_ADDR_NODE): Ditto.
2550 (find): New function.
2552 (dump_constraint): Do not handle INCLUDES.
2553 (insert_into_complex): Do not insert duplicate constraints.
2554 (condense_varmap_nodes): Renamed and rewritten into ...
2555 (merge_node_constraints): This. Also fix bug in handling of
2556 offseted copy constraints.
2557 (clear_edges_for_node): No longer need to deal with preds at all,
2558 or removing associated preds/succs.
2559 (merge_graph_nodes): Deal with indirect_cycles.
2560 Don't deal with predecessors.
2561 (add_implicit_graph_edge): New function.
2562 (add_pred_graph_edge): Ditto.
2563 (add_graph_edge): Don't deal with predecessors.
2564 (build_constraint_graph): Removed.
2565 (build_pred_graph): New function.
2566 (build_succ_graph): Ditto.
2567 (struct scc_info): Removed in_component. Added roots, dfs, and
2568 node_mapping. Remove visited_index, unification_queue.
2569 (scc_visit): Deal with union-find we do now.
2570 Deal with cycles with REF nodes.
2571 (collapse_nodes): Renamed and rewritten to ...
2572 (unify_nodes): This.
2573 (process_unification_queue): Removed.
2574 (topo_visit): Cleanup
2575 (do_da_constraint): Use find.
2576 (do_sd_constraint): Ditto.
2577 (do_ds_constraint): Ditto.
2578 (do_complex_constraint): Ditto.
2579 (init_scc_info): Update for removed and added members.
2580 (find_and_collapse_graph_cycles): Renamed and rewritten into ...
2581 (find_indirect_cycles): This.
2582 (equivalence_class): New variable.
2583 (label_visit): New function.
2584 (perform_variable_substitution): Rewritten.
2585 (free_var_substitution_info): New function.
2586 (find_equivalent_node): Ditto.
2587 (move_complex_constraints): Ditto.
2588 (eliminate_indirect_cycles): Ditto.
2589 (solve_graph): Only propagate changed bits.
2590 Use indirect cycle elimination.
2592 (tree_id_t): Rename to tree_vi_t, delete id member, add vi member.
2593 (tree_id_eq): Renamed to ...
2594 (tree_vi_eq): This. Update for member change
2595 (insert_id_for_tree): Renamed and rewritten to ...
2596 (insert_vi_for_tree): This.
2597 (lookup_id_for_tree): Renamed and rewritten to ...
2598 (lookup_vi_for_tree): This.
2599 (get_id_for_tree): Renamed and rewritten to ...
2600 (get_vi_for_tree): Ditto.
2601 (get_constraint_exp_from_ssa_var): Update to use get_vi_for_tree.
2602 (process_constraint): Don't handle INCLUDES.
2603 Remove special ADDRESSOF case.
2604 (find_func_aliases): Rewrite to use vi functions instead of id
2606 (create_function_info_for): Ditto.
2607 (create_variable_info_for): Ditto.
2608 (intra_create_variable_infos): Ditto.
2609 (merge_smts_into): Ditto.
2610 (find_what_p_points_to): Ditto.
2611 (init_base_vars): Ditto.
2612 (init_alias_vars): Ditto.
2613 (remove_preds_and_fake_succs): New function.
2614 (dump_sa_points_to_info): Dump new stats.
2615 (dump_solution_for_var): Use find.
2616 (set_used_smts): Fix formatting.
2617 (compute_points_to_sets): Updated for new functions.
2618 (ipa_pta_execute): Ditto.
2620 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
2621 Richard Sandiford <richard@codesourcery.com>
2623 * doc/tm.texi (TARGET_FUNCTION_VALUE): Expand documentation of
2625 * calls.c (expand_call): If the return value is a PARALLEL,
2626 extract its first member.
2627 * config/m68k/linux.h (FUNCTION_EXTRA_EPILOGUE): Remove.
2628 * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
2629 use FUNCTION_EXTRA_EPILOGUE.
2630 (m68k_function_value): Return a PARALLEL if the return value
2631 is of a pointer type.
2632 * config/m68k/netbsd-elf.h (current_function_returns_pointer)
2633 (FUNCTION_EXTRA_EPILOGUE): Remove.
2634 * config/m68k/m68k.md (D0_REG): New constant.
2636 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
2638 * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
2639 output a NOP for empty epilogues.
2641 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2643 * config/m68k/m68k.c (m68k_use_return_insn): Update comments
2644 before function. Extend register save check to include all
2645 registers, not just integer ones.
2647 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
2649 * config/m68k/m68k-protos.h (use_return_insn): Rename to...
2650 (m68k_use_return_insn): ...this.
2651 * config/m68k/m68k.h (USE_RETURN_INSN): Delete.
2652 * config/m68k/m68k.c (use_return_insn): Rename to...
2653 (m68k_use_return_insn): ...this.
2654 * config/m68k/m68k.md (return): Use m68k_use_return_insn instead
2657 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
2659 * config/m68k/fpgnulib.c (__truncdfsf2): Implement round to
2660 nearest even, fix denormal rounding overflow.
2662 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2664 * config/m68k/m68k.md (movsf_cf_hard): Use fsmove instead of
2665 f%$smove and f%$move.
2666 (movdf_cf_hard): Use fdmove for cases 0 and 3 and fmove for case 1.
2667 (extendsfdf2_cf): Use fdmove instead of f%&move.
2668 (truncdfsf2_cf): Use fsmove instead of f%$smove.
2669 (add<mode>3_cf, sub<mode>3_cf): Use <FP:prec> instead of <FP:round>.
2671 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2673 * config/m68k/m68k.md (movdf_cf_hard): Use output_move_double for
2676 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2678 * real.h (real_format): Add a canonical_nan_lsbs_set field.
2679 (coldfire_single_format): Declare.
2680 (coldfire_double_format): Likewise.
2681 * real.c (encode_ieee_single): Use canonical_nan_lsbs_set instead
2682 of qnan_msb_set to determine the lower bits of a canonical
2684 (encode_ieee_double): Likewise.
2685 (encode_ieee_quad): Likewise.
2686 (ieee_single_format): Initialize canonical_nan_lsbs_set.
2687 (mips_single_format): Likewise.
2688 (ieee_double_format): Likewise.
2689 (mips_double_format): Likewise.
2690 (ieee_extended_motorola_format): Likewise.
2691 (ieee_extended_intel_96_format): Likewise.
2692 (ieee_extended_intel_128_format): Likewise.
2693 (ieee_extended_intel_96_round_53_format): Likewise.
2694 (ibm_extended_format): Likewise.
2695 (mips_extended_format): Likewise.
2696 (ieee_quad_format): Likewise.
2697 (mips_quad_format): Likewise.
2698 (vax_f_format): Likewise.
2699 (vax_d_format): Likewise.
2700 (vax_g_format): Likewise.
2701 (i370_single_format): Likewise.
2702 (i370_double_format): Likewise.
2703 (decimal_single_format): Likewise.
2704 (decimal_double_format): Likewise.
2705 (decimal_quad_format): Likewise.
2706 (c4x_single_format): Likewise.
2707 (c4x_extended_format): Likewise.
2708 (real_internal_format): Likewise.
2709 (coldfire_single_format): New real_format.
2710 (coldfire_double_format): Likewise.
2711 * config/pdp11/pdp11.c (pdp11_f_format): Initialize
2712 canonical_nan_lsbs_set.
2713 (pdp11_d_format): Likewise.
2714 * config/m68k/m68k.c (override_options): Override REAL_FORMAT_MODE
2715 if TARGET_COLDFIRE_CPU.
2717 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2719 * config/m68k/m68k-protos.h (m68k_output_pic_call): Delete.
2720 (output_call, m68k_legitimize_call_address): Declare.
2721 * config/m68k/m68k.h (EXTRA_CONSTRAINT): Remove unnecessary
2722 parenthesees. Add support for a 'W' constraint.
2723 (LEGITIMATE_PIC_OPERAND_P): Remove SYMBOL_REF_FLAG handling.
2724 (PRINT_OPERAND_PUNCT_VALID_P): Remove comment about 'o'.
2725 (m68k_symbolic_call, m68k_symbolic_jump): Declare.
2726 * config/m68k/m68k.c (m68k_symbolic_call, m68k_symbolic_jump): New
2728 (override_options): Initialize them. Do not set flag_no_function_cse
2729 for TARGET_ID_SHARED_LIBRARY.
2730 (m68k_output_pic_call): Delete.
2731 (m68k_legitimize_call_address): New function.
2732 (print_operand): Remove the %o prefix. Handle the %p prefix.
2733 (output_call): New function.
2734 (m68k_output_mi_thunk): Use m68k_symbolic_jump. Always load the
2735 target address from the GOT if symbolic jumps are not allowed.
2736 * config/m68k/m68k.md (call, general_operand): Do not set
2737 SYMBOL_REF_FLAG. Use m68k_legitimize_call_address instead.
2738 Merge separate flag_pic and !flag_pic define_insns into...
2739 (*call, *call_value): ...these new patterns. Match the address
2740 rather than the containing MEM and require it to be a call_operand.
2741 Use output_call to generate the asm template.
2742 * config/m68k/predicates.md (const_call_operand): New predicate.
2743 (call_operand): Likewise.
2745 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
2747 * config/m68k/m68k.h (REGISTER_MOVE_COST): Simplify definition.
2748 (STACK_GROWS_DOWNWARD): Define to 1.
2749 (FUNCTION_VALUE, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P): Equivocate
2750 comments, emphasizing that these values are only defaults.
2751 * config/m68k/linux.h (LINK_SPEC): Fix formatting in #undef.
2752 * config/m68k/m68k.c (const_method): Remove trailing whitespace.
2754 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2756 * config/m68k/m68k.md (cmpsi): Remove outdated flag_pic handling.
2758 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
2759 Richard Sandiford <richard@codesourcery.com>
2761 * config/m68k/m68k.h (DATA_REGNO_P, ADDRESS_REGNO_P, INT_REGNO_P)
2762 (FP_REGNO_P): New macros.
2763 (REGNO_OK_FOR_INDEX_P, REGNO_OK_FOR_BASE_P, REGNO_OK_FOR_DATA_P)
2764 (REGNO_OK_FOR_FP_P, REG_OK_FOR_INDEX_P, REG_OK_FOR_BASE_P): Use them.
2765 (EH_RETURN_STACKADJ_RTX): Use A0_REG.
2766 * config/m68k/m68k.c (m68k_regno_mode_ok): Use the new REGNO macros.
2768 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
2770 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
2771 (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-rtems*): Add t-floatlib
2773 * config/m68k/t-floatlib: New file, extracting common code from...
2774 * config/m68k/t-m68kbare, config/m68k/t-m68kelf,
2775 * config/m68k/t-uclinux: Here.
2776 * config/m68k/fpgnulib.c: Do not compile extendeed precision
2777 routines on ColdFire targets.
2779 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
2781 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
2782 (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*)
2783 (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-linux*)
2784 (m68k-*-rtems*): Use tm_file rather than m68k/m68k.h and
2785 explicitly set MOTOROLA to 1.
2786 * config/m68k/m68k.h (MOTOROLA): Simplify definition accordingly.
2788 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2789 Nathan Sidwell <nathan@codesourcery.com>
2791 * config/m68k/m68k.h (PCC_STATIC_STRUCT_RETURN): Delete.
2792 (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Add MOTOROLA cases.
2793 * config/m68k/coff.h (REGISTER_PREFIX_MD): Delete.
2794 * config/m68k/m68020-elf.h (LIB_SPEC): Delete.
2795 * config/m68k/m68k-none.h (CC1_SPEC, CPP_SUBTARGET_SPEC): Delete.
2796 * config/m68k/m68kelf.h (IMMEDIATE_PREFIX, REGISTER_PREFIX_MD)
2797 (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Delete.
2798 (NO_DOLLAR_IN_LABEL, PCC_STATIC_STRUCT_RETURN): Don't undefine.
2799 (BSS_ASM_OP, ASM_OUTPUT_SKIP, ASM_OUTPUT_ASCII): Delete.
2800 * config/m68k/m68kemb.h (PCC_STATIC_STRUCT_RETURN): Don't undefine.
2801 * config/m68k/linux.h (SIZE_TYPE, PTRDIFF_TYPE, WCHAR_TYPE)
2802 (WCHAR_TYPE_SIZE, TARGET_OBJFMT_CPP_BUILTINS): Delete.
2803 (TARGET_OS_CPP_BUILTINS): Don't define mc68000 and mc68020 here.
2804 (DBX_REGISTER_NUMBER): Delete.
2805 * config/m68k/netbsd-elf.h (IMMEDIATE_PREFIX): Delete.
2806 (PCC_STATIC_STRUCT_RETURN): Don't undefine.
2807 * config/m68k/openbsd.h (PCC_STATIC_STRUCT_RETURN): Define.
2809 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2811 * config.gcc (m68k-*-uclinux*): Add flat.h to $tm_file.
2812 * config/flat.h: New file.
2813 * crtstuff.c (USE_PT_GNU_EH_FRAME): Don't define if
2815 * config/m68k/m68k.h (ASM_PREFERRED_EH_DATA_FORMAT): Do not use
2816 indirect references for -msep-data or -mid-shared-library.
2817 Do not use PC-relative code addresses either.
2818 * config/m68k/m68k.c (override_options): Restrict -fPIC error
2820 * config/m68k/uclinux.h (STARTFILE_SPEC): Define. Use Scrt1.o
2821 for shared libraries and crt1.o for executables. Use crti.o and
2823 (ENDFILE_SPEC): Use crtend.o and crtn.o.
2824 (LIB_SPEC): Suppress -Rlibc.gdb if -static-libc is given.
2825 Do not add -elf2flt or -shared-lib-id options here.
2826 (LINK_SPEC): Define. Pass -elf2flt if no -elf2flt option is given.
2827 Pass -shared-lib-id if -mid-shared-library, taking the library
2828 identifier from -mshared-library-id if given, otherwise
2830 (EH_FRAME_IN_DATA_SECTION): Do not undefine.
2831 (INIT_SECTION_ASM_OP, FINI_SECTION_ASM_OP): Likewise.
2832 (TARGET_OS_CPP_BUILTINS): Define __GXX_MERGED_TYPEINFO_NAMES=0
2833 and __GXX_TYPEINFO_EQUALITY_INLINE=0 if -mid-shared-library.
2834 (DRIVER_SELF_SPECS): Map unadorned PIC options to -msep-data.
2835 * config/m68k/t-uclinux (EXTRA_MULTILIB_PARTS): Add crtbegin.o
2837 * config/m68k/lb1sf68.asm (PICCALL): Use an lea and pc-relative
2838 jump sequence for ISA A and ISA A+.
2839 (PICJUMP): Likewise.
2841 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2843 * config.gcc (m68k-*-uclinux*): Base the port on the common
2844 and m68k GNU/Linux files rather than on the generic ELF ones.
2845 * config/m68k/uclinux.h (TARGET_VERSION): Override.
2846 (TARGET_OS_CPP_BUILTINS): Use LINUX_TARGET_OS_CPP_BUILTINS.
2848 2007-01-18 Julian Brown <julian@codesourcery.com>
2849 Richard Sandiford <richard@codesourcery.com>
2851 * config/m68k/m68k.h (LONG_DOUBLE_TYPE_SIZE): Make 64-bit on ColdFire.
2852 (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
2853 * config/m68k/netbsd-elf.h (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Undefine
2855 * config/m68k/uclinux-oldabi.h (LONG_DOUBLE_TYPE_SIZE): Redefine to
2857 (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
2859 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2861 * doc/install.texi: Document m68k-uclinuxoldabi.
2862 * config.gcc (m68k-*-uclinuxoldabi): New configuration.
2863 * config/m68k/uclinux-oldabi.h: New file, copied from
2864 config/m68k/uclinux.h.
2866 2007-01-18 Kaz Kojima <kkojima@rr.iij4u.or.jp>
2868 * config/m32r/m32r-protos.h (m32r_expand_epilogue): Declare it.
2869 * config/m32r/m32r.c (m32r_setup_incoming_varargs): Use gen_frame_mem.
2870 (m32r_compute_frame_size): Use unsigned for regno.
2871 (m32r_reload_lr): Use gen_frame_mem.
2873 (m32r_output_function_epilogue): Don't output the function epilogue
2875 (m32r_expand_epilogue): New.
2876 (direct_return): Return false if the function has the interrupt
2878 (m32r_hard_regno_rename_ok): Remove code for the textual epilogue.
2879 * config/m32r/m32r.md (epilogue): New expander.
2880 (return_lr, return_rte): New insns.
2881 (return): Make it expander.
2882 (return_normal): New expander.
2884 2007-01-18 Josh Conner <jconner@apple.com>
2887 * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Add
2888 support for UNLE, UNLT, UNGE, and UNGT.
2890 2007-01-18 Jan Hubicka <jh@suse.cz>
2892 * tree-vrp.c (finalize_jump_threads): Do not call cleanup_cfg by hand.
2893 * tree-tailcall (add_virtual_phis): Likewise.
2894 (optimize_tail_call): Return TODOs.
2895 (execute_tail_calls): Return TODOs.
2896 * tree-ssa-ccp (execute_fold_all_builtins): Do cleanup_cfg via TODO.
2897 * tree-cfgcleanup.c (cleanup_tree_cfg_loop): Return if something
2899 * tree-ssa-forwprop.c (tree_ssa_forward_propagate_single_use_value):
2900 Cleanup cfg using TODO.
2901 * tree-flow.h (cleanup_tree_cfg_loop): Update prototype.
2902 * passes.c (execute_function_todo): When cleanup did something, remove
2904 * tree-cfg.c (pass_build_cfg): Add cleanup_cfg TODO.
2905 (make_edges): Don't cleanup_cfg.
2907 2007-01-18 Uros Bizjak <ubizjak@gmail.com>
2909 * reg-stack.c (subst_stack_regs_pat) [UNSPEC_SINCOS_COS,
2910 UNSPEC_XTRACT_FRACT]: Use generic code for instructions that
2911 operate on the top of stack.
2912 [UNSPEC_SINCOS_SIN, UNSPEC_XTRACT_EXP, UNSPEC_TAN]: Rewrite
2913 register handling of instructions that output to the second
2915 [UNSPEC_TAN_ONE, UNSPEC_TAN_TAN]: Remove.
2916 (move_for_stack_reg): Special-case check for dead destination
2917 stack slot for constant load of 1.0 inside UNSPEC_TAN.
2919 * config/i386/i386.md (UNSPEC_TAN): New constant.
2920 (UNSPEC_TAN_ONE, UNSPEC_TAN_TAN): Remove.
2921 (fptanxf4_i387, fptan_extend<mode>xf4_i387): New patterns
2922 to correctly model move of constant 1.0 to top stack slot.
2923 (*tandf3_1, *tansf3_1, *tanxf3_1): Remove insn patterns.
2924 (unnamed peephole2 pattern): Remove corresponding peephole2
2925 pattern that optimizes tan insn and loading of constant 1.0.
2926 (tanxf2): Use fptanxf4_i387.
2927 (tan<mode>2): Rename from tansf2 and tandf2 and macroize insn
2928 patterns using X87MODEF12 mode macro. Use fptan_extend<mode>xf4_i387
2929 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
2930 disable patterns for SSE math.
2931 (sincos<mode>3): Use truncxf<mode>2_i387_noop for truncation.
2932 (fyl2x_extend<mode>xf3_i387): Use X87MODEF12 for operand 1.
2934 2007-01-18 Dirk Mueller <dmueller@suse.de>
2935 Richard Guenther <rguenther@suse.de>
2938 * doc/invoke.texi (Warray-bounds): Document -Warray-bounds.
2939 * common.opt (Warray-bounds): Add new warning option.
2940 * c-opts.c (c_common_handle_option): Define -Warray-bounds
2942 * Makefile.in: make tree-vrp.o depend on toplev.h
2943 * tree-vrp.c (vrp_finalize): Call check_array_refs if -Warray-bounds
2945 (check_array_refs, check_array_bounds, check_array_ref): New.
2947 2007-01-18 Jan Hubicka <jh@suse.cz>
2949 * tree-ssa-ccp.c (ccp_finalize): Return if something changed.
2950 (execute_ssa_ccp): Return flags conditionally.
2951 * tree-ssa-propagate.c (substitue_and_fold): Return if something was
2953 * tree-ssa-propagate.h (substitute_and_fold): Update prototype.
2955 2007-01-18 Steven Bosscher <steven@gcc.gnu.org>
2957 * cfgcleanup.c (cleanup_cfg): Detect cfglayout mode and set
2958 the CLEANUP_CFGLAYOUT flag when in cfglayout mode.
2960 * Makefile.c (GTFILES): Add cfglayout.h.
2961 * gengtype.c (open_base_files): Likewise.
2962 * cfglayout.c (cfg_layout_function_footer,
2963 cfg_layout_function_header) Reindent to make gengtype happy.
2964 * cfglayout.h (cfg_layout_function_footer,
2965 cfg_layout_function_header): Add GTY(()) marker.
2967 * ifcvt.c (noce_try_sign_mask): Make sure INSN_B is non-null.
2969 2007-01-18 Ben Elliston <bje@au.ibm.com>
2971 * genautomata.c (write_automata): Include xstrerror output in the
2972 error message if writing the DFA description file fails.
2974 2007-01-17 H.J. Lu <hongjiu.lu@intel.com>
2976 * config/mips/mips-protos.h (mips_output_external): Make it
2978 * config/mips/iris.h (TARGET_ASM_EXTERNAL_LIBCALL): Removed.
2979 * config/mips/mips.c (irix_output_external_libcall): Likewise.
2980 (extern_list): Likewise.
2981 (extern_head): Likewise.
2982 (TARGET_ASM_FILE_END): Likewise.
2983 (mips_file_end): Likewise.
2984 (mips_output_external): Rewritten.
2986 2007-01-18 Ben Elliston <bje@au.ibm.com>
2988 * genpreds.c (write_insn_preds_c): Only write out the function
2989 body for regclass_for_constraint if we have register constraints.
2991 2007-01-17 Tom Tromey <tromey@redhat.com>
2993 * doc/sourcebuild.texi (libgcj Tests): Use sourceware.org.
2994 * doc/install.texi (Testing): Use sourceware.org.
2995 (Binaries): Likewise.
2996 (Specific): Likewise.
2997 * doc/contrib.texi (Contributors): Use sourceware.org.
2999 2007-01-17 Anatoly Sokolov <aesok@post.ru>
3001 * config/avr/avr.h (AVR_HAVE_LPMX): New macro.
3002 (AVR_ENHANCED): Rename to ...
3003 (AVR_HAVE_MUL): ... new.
3004 (avr_enhanced_p): Rename to ...
3005 (avr_have_mul_p): ... new.
3006 (TARGET_CPU_CPP_BUILTINS): Use 'avr_have_mul_p' instead of
3007 'avr_enhanced_p' for "__AVR_ENHANCED__". Define "__AVR_HAVE_MUL__".
3008 * config/avr/avr.c (avr_enhanced_p): Rename to ...
3009 (avr_have_mul_p): ... new.
3010 (base_arch_s): Rename 'enhanced' to 'have_mul'.
3011 (avr_override_options): Use 'avr_have_mul_p' and 'have_mul' instead of
3012 'avr_enhanced_p' and 'enhanced'.
3013 (ashlhi3_out, ashrhi3_out, lshrhi3_out, avr_rtx_costs): Use
3014 AVR_HAVE_MUL instead of AVR_ENHANCED.
3015 * avr.md (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
3016 (mulqi3, *mulqi3_enh, *mulqi3_call, mulqihi3, umulqihi3, mulhi3,
3017 *mulhi3_enh, *mulhi3_call, mulsi3, *mulsi3_call): Use AVR_HAVE_MUL
3018 instead of AVR_ENHANCED.
3019 (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
3020 * libgcc.S: Use __AVR_HAVE_MUL__ instead of __AVR_ENHANCED__.
3021 (__tablejump__): Use __AVR_HAVE_LPMX__ instead of __AVR_ENHANCED__.
3023 2007-01-17 Ian Lance Taylor <iant@google.com>
3025 * vec.h (VEC_reserve_exact): Define.
3026 (vec_gc_p_reserve_exact): Declare.
3027 (vec_gc_o_reserve_exact): Declare.
3028 (vec_heap_p_reserve_exact): Declare.
3029 (vec_heap_o_reserve_exact): Declare.
3030 (VEC_OP (T,A,reserve_exact)): New static inline function, three
3032 (VEC_OP (T,A,reserve)) [all versions]: Remove handling of
3034 (VEC_OP (T,A,alloc)) [all versions]: Call ...reserve_exact.
3035 (VEC_OP (T,A,copy)) [all versions]: Likewise.
3036 (VEC_OP (T,a,safe_grow)) [all versions]: Likewise.
3037 * vec.c (calculate_allocation): Add exact parameter. Change all
3039 (vec_gc_o_reserve_1): New static function, from vec_gc_o_reserve.
3040 (vec_gc_p_reserve, vec_gc_o_reserve): Call vec_gc_o_reserve_1.
3041 (vec_gc_p_reserve_exact, vec_gc_o_reserve_exact): New functions.
3042 (vec_heap_o_reserve_1): New static function, from vec_heap_o_reserve.
3043 (vec_heap_p_reserve, vec_heap_o_reserve): Call vec_heap_o_reserve_1.
3044 (vec_heap_p_reserve_exact): New function.
3045 (vec_heap_o_reserve_exact): New function.
3047 2007-01-17 Jan Hubicka <jh@suse.cz>
3049 * ipa-type-escape.c (look_for_casts): Revamp using handled_component_p.
3051 2007-01-17 Eric Christopher <echristo@apple.com>
3053 * config.gcc: Support core2 processor.
3055 2007-01-16 Jan Hubicka <jh@suse.cz>
3057 * tree-ssanames.c (release_dead_ssa_names): Instead of ggc_freeing
3058 the names, just unlink the chain so we don't crash on dangling pointers
3061 2007-01-16 Jan Hubicka <jh@suse.cz>
3063 * cgraph.h (cgraph_decide_inlining_incrementally): Kill.
3064 * tree-pass.h: Reorder to make IPA passes appear toegher.
3065 (pass_early_inline, pass_inline_parameters, pass_apply_inline): Declare.
3066 * cgraphunit.c (cgraph_finalize_function): Do not compute inling
3067 parameters, do not call early inliner.
3068 * ipa-inline.c: Update comments. Include tree-flow.h
3069 (cgraph_decide_inlining): Do not compute inlining parameters.
3070 (cgraph_decide_inlining_incrementally): Return TODOs; assume to
3071 be called with function context set up.
3072 (pass_ipa_inline): Remove unreachable functions before pass.
3073 (cgraph_early_inlining): Simplify assuming to be called from the
3075 (pass_early_inline): New pass.
3076 (cgraph_gate_ipa_early_inlining): New gate.
3077 (pass_ipa_early_inline): Turn into simple wrapper.
3078 (compute_inline_parameters): New function.
3079 (gate_inline_passes): New gate.
3080 (pass_inline_parameters): New pass.
3081 (apply_inline): Move here from tree-optimize.c
3082 (pass_apply_inline): New pass.
3083 * ipa.c (cgraph_remove_unreachable_nodes): Verify cgraph after
3085 * tree-inline.c (optimize_inline_calls): Return TODOs rather than
3087 (tree_function_versioning): Do not allocate dummy struct function.
3088 * tree-inline.h (optimize_inline_calls): Update prototype.
3089 * tree-optimize.c (execute_fixup_cfg): Export.
3090 (pass_fixup_cfg): Remove
3091 (tree_rest_of_compilation): Do not apply inlines.
3092 * tree-flow.h (execute_fixup_cfg): Declare.
3093 * Makefile.in (gt-passes.c): New.
3094 * passes.c: Include gt-passes.h
3095 (init_optimization_passes): New passes.
3096 (nnodes, order): New static vars.
3097 (do_per_function_toporder): New function.
3098 (execute_one_pass): Dump current pass here.
3099 (execute_ipa_pass_list): Don't dump current pass here.
3101 2007-01-16 Janis Johnson <janis187@us.ibm.com>
3103 * config/dfp-bit.c (dfp_compare_op): Return separate value for NaN.
3104 (DFP_NE, DFP_LE, DFP_GE): Return false for NaN.
3106 2007-01-16 David Edelsohn <edelsohn@gnu.org>
3108 * config/rs6000/darwin-ldouble.c: Build file for SOFT_FLOAT.
3109 (strong_alias): Define.
3110 (__gcc_qmul): Provide non-FMA for soft-float.
3116 (__gcc_qunord): New.
3126 * config/rs6000/rs6000.c (rs6000_init_libfuncs): Initialize
3127 soft-float functions.
3128 * config/rs6000/libgcc-ppc-glibc.ver: Version soft-float symbols.
3129 * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Do not warn
3130 about long double soft float.
3132 2007-01-16 Dorit Nuzman <dorit@il.ibm.com>
3133 Tehila Meyzels <tehila@il.ibm.com>
3135 * tree-vectorizer.h (is_pattern_stmt_p): New.
3136 * tree-vect-analyze.c (vect_determine_vectorization_factor): Fix
3137 formatting (tabs instead of spaces). Cleanup and clarify setting
3138 of STMT_VINFO_VECTYPE. Call is_pattern_stmt_p.
3139 * tree-vect-transform.c (vect_get_vec_def_for_operand): Fix typo.
3140 (vectorizable_type_demotion): Check that types are integral.
3141 (vectorizable_type_promotion): Likewise.
3142 (vectorizable_store): Fix typo. Eliminate new-line at end of
3145 2007-01-16 Jan Hubicka <jh@suse.cz>
3147 * tree-ssanames.c (release_dead_ssa_names): Remove invalidated
3150 2007-01-15 Eric Christopher <echristo@apple.com>
3152 * ifcvt.c: Include vec.h, vecprim.h.
3153 (check_cond_move_block): New argument regs.
3154 Reorganize. Add registers used to regs.
3155 (cond_move_process_if_block): Use regs set above as
3158 2007-01-15 Eric Christopher <echristo@apple.com>
3160 * config/darwin.h: Update copyright.
3161 (TARGET_OPTION_TRANSLATE_TABLE): Add umbrella.
3162 (LINK_COMMAND_SPEC): Add -u.
3163 (LINK_SPEC): Fix umbrella for above.
3165 2007-01-15 Joseph S. Myers <joseph@codesourcery.com>
3167 * config/soft-fp/op-common.h, config/soft-fp/op-4.h: Update from
3170 2007-01-15 Tom Tromey <tromey@redhat.com>
3172 * doc/sourcebuild.texi (libgcj Tests): Don't mention jacks.
3173 * doc/install.texi (Testing): Don't mention jacks.
3174 (Configuration): Document --enable-java-maintainer-mode. Move
3175 --with-java-home to libgcj-specific section. Document
3177 (Prerequisites): Mention --enable-java-maintainer-mode, ecj1.
3179 2007-01-15 Jan Hubicka <jh@suse.cz>
3181 * tree-ssa-dce.c (DCE_TODOs): New.
3182 (propagate_necessity): Return if something changed.
3183 (eliminate_unnecessary_stmts): Likewise.
3184 (perform_tree_ssa_dce): Return TODO flags when needed.
3185 (pass_dce, pass_dce_loop, pass_cd_dce): Remove TODO flags.
3187 2007-01-15 Uros Bizjak <ubizjak@gmail.com>
3189 * config/i386/i386.md (fyl2xxf3_i387): Rename from fyl2x_xf3.
3190 (fyl2x_extend<mode>xf3_i387): New insn pattern.
3191 (log<mode>2): Rename from logsf2 and logdf2 and macroize insn
3192 patterns using X87MODEF12 mode macro. Extend operand 1
3193 to XFmode. Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
3194 (log10<mode>2): Ditto.
3195 (log2<mode>2): Ditto.
3196 (log1p<mode>2): Ditto.
3197 (logb<mode>2): Ditto.
3198 (fyl2xp1xf3_i387): Rename from fyl2xp1_xf3.
3199 (fyl2xp1_extend<mode>xf3_i387): New insn pattern.
3200 (*fxtractxf3_i387): Rename from *fxtractxf3.
3201 (fxtract_extend<mode>xf3_i387): New insn pattern.
3202 (ilogbsi2): Use match_dup 3, not match_operand:XF 3.
3203 * config/i386/i386.c (ix86_emit_i387_log1p): Use gen_fyl2xp1xf3_i387()
3204 and gen_fyl2xxf3_i387().
3206 2007-01-14 Zdenek Dvorak <dvorakz@suse.cz>
3208 * loop-unswitch.c (unswitch_loop): Do not call fix_loop_placement.
3209 * cfgloopmanip.c (fix_loop_placement): Made static. Use
3210 get_loop_exit_edges. Changed return type to bool.
3211 * cfgloop.h (fix_loop_placement): Declaration removed.
3213 2007-01-14 Dorit Nuzman <dorit@il.ibm.com>
3215 * param.h (MIN_VECT_LOOP_BOUND): New.
3216 * params.def (MIN_VECT_LOOP_BOUND): New.
3217 * tree-vectorizer.c (slpeel_tree_peel_loop_to_edge): Takes another
3218 argument - minimum threshold for number of iterations.
3219 * tree-vectorizer.h (slpeel_tree_peel_loop_to_edge): Add another
3220 argument to declaration.
3221 * tree-vect-analyze.c (vect_analyze_operations): Check value of
3222 MIN_VECT_LOOP_BOUND.
3223 * tree-vect-transform.c (vect_do_peeling_for_loop_bound): Call
3224 slpeel_tree_peel_loop_to_edge with additional argument.
3225 (vect_do_peeling_for_alignment): Likewise.
3226 * doc/invoke.texi (min-vect-loop-bound): Document new param option.
3228 2007-01-14 Uros Bizjak <ubizjak@gmail.com>
3231 * config/i386/i386.c (print_operand) ['z']: Output 'b' for
3234 2007-01-14 Jan Hubicka <jh@suse.cz>
3236 * tree-dfa.c (remove_referenced_var): New function.
3237 * tree-ssa-live.c (remove_unused_locals): Walk referenced vars and
3238 prune referenced vars list too.
3239 * tree-flow.h (remove_referenced_var): Declare.
3241 2007-01-14 Jan Hubicka <jh@suse.cz>
3243 * tree-eh.c (add_stmt_to_eh_region_fn): Do not add call_exprs
3245 (remove_stmt_from_eh_region_fn): Do not remove call_exprs.
3246 (verify_eh_throw_stmt_node, verify_eh_throw_table_statements): Kill.
3247 * except.h (verify_eh_throw_table_statements): Kill prototype.
3248 * cfgexpand.c (expand_gimple_basic_block): Propagate Eh regions
3249 into call exrepssions.
3250 * tree-optimize.c (execute_free_cfg_annotatiosn): Do not call
3252 * tree-cfg.c: Include pointer-set.h.
3253 (verify_node_sharing): Work on pointer set.
3254 (verify_eh_throw_stmt_node): New.
3255 (verify_stmts): Use pointers sets, verify throw_stmt.
3257 2007-01-13 Zdenek Dvorak <dvorakz@suse.cz>
3259 * ipa-reference.c (analyze_function): Consider also addresses taken
3262 2007-01-12 Roger Sayle <roger@eyesopen.com>
3264 * c-typeck.c (null_pointer_constant_p): Replace use of
3265 TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
3266 (build_c_cast): Likewise.
3268 2007-01-12 Roger Sayle <roger@eyesopen.com>
3270 * tree.h (force_fit_type_double): Remove unused final argument.
3271 * c-common.c (constant_expression_warning): Replace use of
3272 TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
3273 (convert_and_check): Likewise.
3274 (shorten_compare): Update call to force_fit_type_double.
3275 (c_common_truthvalue_conversion) <INTEGER_CST>: Use integer_zerop.
3276 * convert.c (convert_to_pointer): Update call to
3277 force_fit_type_double.
3278 * fold-const.c (force_fit_type_double): Remove overflowed_const
3280 (int_const_binop, fold_convert_const_int_from_int,
3281 fold_convert_const_int_from_real, fold_div_compare,
3282 fold_sign_changed_comparison, fold_unary, fold_negate_const,
3283 fold_abs_const, fold_not_const): Remove the final argument from
3284 calls to force_fit_type_double.
3286 2007-01-12 Andrew Pinski <andrew_pinski@playstation.sony.com>
3288 * configure.ac: Set insn to "nop" for spu-*-* also.
3289 * configure: Regenerate.
3291 2007-01-12 Olga Golovanevsky <olga@il.ibm.com>
3293 * builtins.def : Add BUILT_IN_FREE.
3295 2007-01-12 Jan Hubicka <jh@suse.cz>
3297 PR tree-optimization/30443
3298 * tree-inline.c (tree_function_versioning): Do not optimize when
3299 cloning for inlining.
3301 2007-01-12 Zdenek Dvorak <dvorakz@suse.cz>
3303 * doc/loop.texi: Document recording of loop exits.
3304 * cfgloopmanip.c (loopify, duplicate_loop): Use alloc_loop.
3305 (update_single_exits_after_duplication,
3306 update_single_exit_for_duplicated_loop,
3307 update_single_exit_for_duplicated_loops): Removed.
3308 (duplicate_loop_to_header_edge): Do not call
3309 update_single_exits_after_duplication and
3310 update_single_exit_for_duplicated_loops.
3311 (loop_version): Do not update single_exit information.
3312 (fix_loop_structure): Use record_loop_exits instead of
3313 mark_single_exit_loops.
3314 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update
3315 the lists of loop exits.
3316 * cfghooks.c (redirect_edge_and_branch, redirect_edge_and_branch_force,
3317 split_edge, merge_blocks): Update the lists of loop exits.
3318 * modulo-sched.c (sms_schedule): Pass LOOPS_HAVE_RECORDED_EXITS to
3319 loop_optimizer_init.
3320 * loop-init.c (loop_optimizer_init): Call record_loop_exits instead
3321 of mark_single_exit_loops.
3322 (loop_optimizer_finalize): Call release_recorded_exits.
3323 * tree-ssa-loop.c (tree_loop_optimizer_init): Pass
3324 LOOPS_HAVE_RECORDED_EXITS to loop_optimizer_init.
3325 * tree-vectorizer.c (slpeel_tree_duplicate_loop_to_edge_cfg): Do not
3326 update single exit information.
3327 * lambda-code.c (perfect_nestify): Ditto.
3328 * cfgloop.c (flow_loop_free): Destroy the list of exits of the loop.
3329 (mark_single_exit_loops): Removed.
3330 (alloc_loop, loop_exit_hash, loop_exit_eq, loop_exit_free,
3331 get_exit_descriptions, rescan_loop_exit, record_loop_exits,
3332 dump_recorded_exit, dump_recorded_exits, release_recorded_exits): New
3334 (get_loop_exit_edges, single_exit): Use recorded exit lists.
3335 (add_bb_to_loop, remove_bb_from_loops): Update the lists of loop exits.
3336 (verify_loop_structure): Verify consistency of the exit lists.
3337 (flow_loops_find): Use alloc_loop. Initialize exits hash.
3338 (set_single_exit): Removed.
3339 * cfgloop.h (struct loop_exit): New function.
3340 (struct loop): single_exit_ field replaced by exits field.
3341 (LOOPS_HAVE_MARKED_SINGLE_EXITS): Replaced by LOOPS_HAVE_RECORDED_EXITS.
3342 (struct loops): Added exits hash.
3343 (mark_single_exit_loops, set_single_exit): Declaration removed.
3344 (release_recorded_exits, record_loop_exits, rescan_loop_exit): Declare.
3346 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3348 * doc/invoke.texi: Avoid use of @headitem.
3350 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3352 * cse.c (cse_insn): Move HAVE_CC0 code after declarations.
3354 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3356 * doc/install.texi: Fix m68k-*-* anchor and add m68k-*-* to the
3359 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
3360 Richard Sandiford <richard@codesourcery.com>
3362 * doc/invoke.texi: Document -mno-bitfield, -mno-rtd and -mno-short.
3363 * config/m68k/m68k.opt: Resort options.
3364 (mbitfield, mrtd, mshort): Remove RejectNegative properties.
3366 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
3367 Richard Sandiford <richard@codesourcery.com>
3369 * doc/invoke.texi: Document the macros that are defined by
3370 m68k's -mtune and -mhard-float options.
3371 * config/m68k/m68k-protos.h (m68k_cpp_cpu_ident) Declare.
3372 (m68k_cpp_cpu_family): Likewise.
3373 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Add a full set
3374 of __ucfv*__ macros. Define __mcffpu__ if generating code for
3375 ColdFire FPUs. Define __mcf_cpu_* and __mcf_family_* macros.
3376 * config/m68k/m68k.c (m68k_cpp_cpu_ident): New function.
3377 (m68k_cpp_cpu_family): Likewise.
3379 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3381 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Treat all mc68*
3382 macros besides mc68000 as tuning macros. Use a switch statement
3383 to set them and mcpu32.
3385 2007-01-12 Julian Brown <julian@codesourcery.com>
3387 * config/m68k/m68k.h: Use TARGET_68040 instead of TARGET_68040_ONLY.
3388 (TARGET_68040_ONLY): Rename to...
3389 (TARGET_68040): ...this.
3390 * config/m68k/m68k.c: Use TARGET_68040 instead of TARGET_68040_ONLY.
3391 * config/m68k/m68k.md: Likewise.
3393 2007-01-12 Julian Brown <julian@codesourcery.com>
3394 Nathan Sidwell <nathan@codesourcery.com>
3395 Richard Sandiford <richard@codesourcery.com>
3397 * config.gcc (m680[012]0-*-*, m68k*-*-*): Set m68k_cpu_ident to
3398 the -mcpu= argument associated with the --with-cpu setting.
3399 Define M68K_DEFAULT_TUNE to the default -mtune= option,
3400 if different from the one implied by the -mcpu setting.
3401 Accept --with-cpu=FOO if FOO is listed in m68k-devices.def,
3402 using mcpu=FOO as the default CPU option. Set target_cpu_default2.
3403 * doc/invoke.texi: Mention ColdFire in the introduction to the
3404 m68k options. Document the new -march, -mcpu, -mtune, -mdiv,
3405 -mno-div and -mhard-float options. Make -m68881 a synonym for
3406 -mhard-float. Document the previously-undocumented -m5206e,
3407 -m528x, -m5307 and -m5407 options. Tweak the existing option
3408 documentation for consistency.
3409 * doc/install.texi: Mention new --with-cpu arguments.
3410 * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Only use the
3411 default CPU if neither -mcpu nor -march are specified.
3412 (ASM_CPU_SPEC): Pass down -mcpu and -march options.
3413 (TARGET_CPU_CPP_BUILTINS): Set __mcfisa*__ macros from
3414 TARGET_ISA*. Set the legacy __mcf*__ cpu macros in the same way,
3415 using m68k_tune to decide between families that implement the
3416 same ISA. Use m68k_tune to set __mcfv4e__.
3417 (FL_BITFIELD, FL_68881, FL_COLDFIRE, FL_CF_HWDIV, FL_CF_MAC)
3418 (FL_CF_EMAC, FL_CF_EMAC_B, FL_CF_USP, FL_CF_FPU, FL_ISA_68000)
3419 (FL_ISA_68010, FL_ISA_68020, FL_ISA_68040, FL_ISA_A, FL_ISA_B)
3420 (FL_ISA_C, FL_ISA_MMU): New macros.
3421 (MASK_COLDFIRE): Delete.
3422 (TARGET_68010, TARGET_68020, TARGET_68040_ONLY, TARGET_COLDFIRE)
3423 (TARGET_ISAB): Redefine in terms of m68k_cpu_flags.
3424 (TARGET_68881, TARGET_COLDFIRE_FPU): Redefine in terms of m68k_fpu.
3425 (TARGET_HARD_FLOAT): Do not define here.
3426 (TARGET_ISAAPLUS, TARGET_ISAC): New macros.
3427 (TUNE_68000): New macro.
3428 (TUNE_68000_10): Redefine in terms of TUNE_68000 and TUNE_68010.
3429 (TUNE_68010, TUNE_68030, TUNE_68040, TUNE_68060, TUNE_CPU32)
3430 (TUNE_CFV2): Redefine in terms of m68k_tune.
3431 (uarch_type, target_device, fpu_type): New enums.
3432 (m68k_cpu, m68k_tune, m68k_fpu, m68k_cpu_flags): Declare.
3433 * config/m68k/m68k.c (TARGET_DEFAULT): Remove MASK_68881.
3434 (FL_FOR_isa_00, FL_FOR_isa_10, FL_FOR_isa_20, FL_FOR_isa_40)
3435 (FL_FOR_isa_cpu32, FL_FOR_isa_a, FL_FOR_isa_aplus, FL_FOR_isa_b)
3436 (FL_FOR_isa_c): New macros.
3437 (m68k_isa): New enum.
3438 (m68k_target_selection): New structure.
3439 (all_devices, all_isas, all_microarchs): New tables.
3440 (m68k_cpu_entry, m68k_arch_entry, m68k_tune_entry, m68k_cpu)
3441 (m68k_tune, m68k_fpu, m68k_cpu_flags): New variables.
3442 (MASK_ALL_CPU_BITS): Delete.
3443 (m68k_find_selection): New function.
3444 (m68k_handle_option): Handle -mcpu=, -march= and -mtune=.
3445 Map the legacy target options to a combination of the new ones.
3446 (override_options): Set m68k_cpu, m68k_tune, m68k_fpu and
3447 m68k_cpu_flags. Handle M68K_DEFAULT_TUNE. Use m68k_cpu_flags
3448 to derive default MASK_BITFIELD, MASK_CF_HWDIV and MASK_HARD_FLOAT
3450 * config/m68k/m68k.opt (m5200, m5206e, m528x, m5307, m5407, mcfv4e)
3451 (m68010, m68020, m68020-40, m68020-60, m68030, m68040): Remove Mask
3453 (m68881, msoft-float): Change mask from 68881 to HARD_FLOAT.
3454 (march=, mcpu=, mdiv, mhard-float, mtune=): New options.
3455 * config/m68k/m68k-devices.def: New file.
3457 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3458 Nathan Sidwell <nathan@codesourcery.com>
3460 * config/m68k/m68k.h (ASM_CPU_SPEC, ASM_SPEC, EXTRA_SPECS)
3461 (SUBTARGET_EXTRA_SPECS): New macros.
3462 * config/m68k/linux.h (ASM_SPEC): Remove CPU flags;
3463 use %(asm_cpu_spec) instead.
3464 * config/m68k/m68k-none.h (ASM_SPEC): Likewise.
3465 * config/m68k/openbsd.h (ASM_SPEC): Likewise.
3466 * config/m68k/netbsd-elf.h (ASM_SPEC): Likewise.
3467 (EXTRA_SPECS): Rename to...
3468 (SUBTARGET_EXTRA_SPECS): ...this.
3470 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
3471 Richard Sandiford <richard@codesourcery.com>
3472 Julian Brown <julian@codesourcery.com>
3474 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
3475 (m68k-*-uclinux*, m68k-*-linux*, m68k-*-rtems*): Set default_m68k_cpu
3476 to the configuration's default CPU.
3477 (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*): Likewise.
3478 Remove default masks.
3479 (m680[012]0-*-*): Set the default with_cpu to the first part of
3481 (m68k*-*-*): Set the default with_cpu to m$default_m68k_cpu.
3482 (m68k*-*-linux): Extend the --with-cpu handling to...
3483 (m680[012]0-*-*, m68k*-*-*): ...these configurations. Allow m68000
3484 and m68010. Don't set target_cpu_default2.
3485 * doc/install.texi: Document --with-cpu for m68k.
3486 * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Define.
3487 * config/m68k/m68k-none.h (TARGET_CPU_DEFAULT, M68K_CPU_m68k)
3488 (M68K_CPU_m68000, M68K_CPU_m68010, M68K_CPU_m68020, M68K_CPU_m68030)
3489 (M68K_CPU_m68040, M68K_CPU_m68302, M68K_CPU_m68332, TARGET_DEFAULT)
3490 (ASM_CPU_DEFAULT_SPEC, CC1_CPU_DEFAULT_SPEC): Delete.
3491 (ASM_SPEC): Remove use of %(asm_cpu_default).
3492 (EXTRA_SPECS, SUBTARGET_EXTRA_SPECS, MULTILIB_DEFAULTS): Delete.
3493 * config/m68k/linux.h (TARGET_DEFAULT): Delete.
3494 (CPP_SPEC): Merge definitions. Do not handle __HAVE_68881__ here.
3495 * config/m68k/netbsd-elf.h (TARGET_OS_CPP_BUILTINS): Define
3496 __HAVE_FPU__ if TARGET_HARD_FLOAT.
3497 (TARGET_DEFAULT): Delete.
3498 (EXTRA_SPECS): Delete cpp_cpu_default_spec, cpp_cpu_spec,
3499 cpp_fpu_spec, asm_default_spec and netbsd_cpp_spec.
3500 (CPP_CPU_SPEC): Delete.
3501 (TARGET_VERSION): Merge definitions, using TARGET_68010 to pick
3502 the appropriate string.
3503 (CPP_CPU_DEFAULT_SPEC, ASM_DEFAULT_SPEC, CPP_FPU_SPEC): Delete.
3504 (CPP_SPEC): Define to NETBSD_CPP_SPEC.
3505 (ASM_SPEC): Don't use %(asm_default_spec).
3506 * config/m68k/m68k.c (TARGET_DEFAULT_TARGET_FLAGS): Remove
3507 TARGET_DEFAULT and add MASK_68881.
3508 * config/m68k/m68k.md: Remove mention of TARGET_DEFAULT from comments.
3510 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3512 * config.gcc (m68010-*-netbsdelf*): Add MASK_68010.
3513 (m68k*-*-netbsdelf*, m68k*-*-openbsd*, m68k*-linux*): Add
3514 MASK_68010 alongside MASK_68020.
3515 * doc/invoke.texi: Document -m68010.
3516 * config/m68k/m68k.opt (m68010): New.
3517 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Define mc68010
3519 (TUNE_68010): New macro.
3520 * config/m68k/m68k-none.h (M68K_CPU_m68k, M68K_CPU_m68010)
3521 (M68K_CPU_m68020, M68K_CPU_m68030, M68K_CPU_m68040)
3522 (M68K_CPU_m68332): Add MASK_68010.
3523 * config/m68k/linux.h (TARGET_DEFAULT): Add MASK_68010 to
3524 fallback definition.
3525 * config/m68k/netbsd-elf.h (CPP_CPU_SPEC): Remove now-redundant
3527 * config/m68k/m68k.c (MASK_ALL_CPU_BITS): Add MASK_68010.
3528 (m68k_handle_option): Handle OPT_m68010. Add MASK_68010
3529 to all entries that use MASK_68020.
3530 (output_move_simode_const, output_move_himode, output_move_qimode)
3531 (output_move_stricthi, output_move_strictqi): Use TARGET_68010
3532 instead of TARGET_68020 to select clr behavior. Remove comment
3533 about there being no TARGET_68010.
3534 * config/m68k/m68k.md: Likewise throughout.
3536 2007-01-12 Julian Brown <julian@codesourcery.com>
3538 * config/m68k/m68k.h (TARGET_ISAB): New macro.
3539 * config/m68k/m68k.c: Use TARGET_ISAB rather than TARGET_CFV4.
3540 * config/m68k/m68k.md: Likewise.
3542 2007-01-12 Julian Brown <julian@codesourcery.com>
3544 * config/m68k/m68k.h (LEGITIMATE_INDEX_P, LEGITIMIZE_ADDRESS): Use
3545 TARGET_COLDFIRE_FPU instead of TARGET_CFV4E.
3547 2007-01-12 Julian Brown <julian@codesourcery.com>
3549 * config/m68k/m68k.h (TUNE_68040_60): New macro.
3550 * config/m68k/m68k.c (standard_68881_constant_p): Use it.
3551 * config/m68k/m68k.md: Likewise.
3553 2007-01-12 Julian Brown <julian@codesourcery.com>
3554 Richard Sandiford <richard@codesourcery.com>
3556 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Use TUNE_68030
3557 instead of TARGET_68030, TUNE_68040 instead of TARGET_68040,
3558 TUNE_68060 instead of TARGET_68060 and TUNE_CPU32 instead of
3560 (TARGET_CPU32): Rename to...
3561 (TUNE_CPU32): ...this.
3562 (TUNE_68000_10, TUNE_68030, TUNE_68040, TUNE_68060)
3563 (TUNE_CFV2): New macros.
3564 * config/m68k/netbsd-elf.h (LONG_DOUBLE_TYPE_SIZE): Simplify;
3565 remove conditions that are implied by TARGET_68020.
3566 * config/m68k/m68k.c (m68k_output_function_prologue): Use TUNE_68040
3567 instead of TARGET_68040 and TUNE_CPU32 instead of TARGET_CPU32.
3568 (m68k_output_function_epilogue): Likewise.
3569 (m68k_rtx_costs): Likewise. Use TUNE_68060 instead of TARGET_68060
3570 and TUNE_CFV2 instead of TARGET_5200. Use TUNE_68000_10 instead of
3571 "!TARGET_68020 && !TARGET_COLDFIRE" to choose between 68000 and
3572 non-68000 timings. Refactor multiplication and division costs.
3573 (output_addsi3): Use TUNE_68040 instead of TARGET_68040 and
3574 TUNE_CPU32 instead of TARGET_CPU32.
3575 (standard_68881_constant_p): Use TUNE_68040 instead of TARGET_68040
3576 and TUNE_68060 instead of TARGET_68060.
3577 * config/m68k/m68k.md: Use TUNE_68040 instead of TARGET_68040,
3578 TUNE_68060 instead of TARGET_68060, and TUNE_CPU32 instead of
3580 (movsi_const0): Use TUNE_68000_10 rather than "!TARGET_68020
3581 && !TARGET_COLDFIRE" to choose between moveq and clr.
3582 Likewise in the unnamed movsf pattern.
3583 (ashlsi_17_24, lshrsi_17_24): Guard with TUNE_68000_10 rather than
3584 "!TARGET_68020 && !TARGET_COLDFIRE". Likewise the unnamed
3587 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3589 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Increase amount
3590 of tabbing before backslashes.
3592 2007-01-11 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
3594 * pa-linux.h (ASM_OUTPUT_INTERNAL_LABEL): Undefine.
3595 * pa.h (ASM_OUTPUT_LABEL): Output colon when using GAS.
3596 (ASM_OUTPUT_INTERNAL_LABEL): Define.
3598 2007-01-11 Zdenek Dvorak <dvorakz@suse.cz>
3600 * tree-ssa-loop-ivopts.c (extract_cond_operands): Split from
3601 find_interesting_uses_cond.
3602 (find_interesting_uses_cond): Use extract_cond_operands.
3603 (rewrite_use_compare): Use extract_cond_operands and
3604 force_gimple_operand_bsi. Do not call update_stmt.
3605 (determine_use_iv_cost_condition): Use extract_cond_operands.
3606 Return cheaper of using original bound and changing the exit bound.
3608 2007-01-11 Zdenek Dvorak <dvorakz@suse.cz>
3610 PR tree-optimization/29516
3611 * tree-ssa-address.c (tree_mem_ref_addr, add_to_parts,
3612 most_expensive_mult_to_index, addr_to_parts,
3613 create_mem_ref, maybe_fold_tmr): Make the type of
3614 fields of TARGET_MEM_REF sizetype.
3615 (move_fixed_address_to_symbol, move_pointer_to_base):
3617 * tree.def (TARGET_MEM_REF): Add comment on types of
3620 2007-01-11 Joseph Myers <joseph@codesourcery.com>
3622 * c-common.c (vector_types_convertible_p): Treat opaque types as
3623 always convertible if they have the same size, but not otherwise.
3625 2007-01-11 Steven Bosscher <steven@gcc.gnu.org>
3627 * ifcvt.c (struct noce_if_info): Add comments to the fields.
3628 Remove the b_unconditional field.
3629 (noce_try_sign_mask): Do not look at b_unconditional.
3630 (noce_process_if_block): Do not use merge_if_blocks. Update
3631 the CFG here. Do not set b_unconditional.
3632 (cond_move_process_if_block): Likewise.
3633 (find_cond_trap): Likewise.
3634 (check_cond_move_block): Require simple jump insns at the end
3637 2007-01-11 Jan Hubicka <jh@suse.cz>
3639 PR tree-optimization/1046
3640 * tree-tailcall.c (suitable_for_tail_call_opt_p): Use TREE_ADDRESSABLE
3641 when alias info is not ready.
3642 (pass_tail_recursion): Do not require aliasing.
3643 * tree-ssa-copyrename.c (pass_rename_ssa_cop): Likewise.
3644 * tree-ssa-ccp.c (pass_ccp, pass_fold_builtins): Likewise.
3645 * tree-ssa-copy.c (pass_copy_prop): Likewise.
3646 * tree-ssa-forwprop.c (pass_forwprop): Likewise.
3647 * tree-ssa-dce.c (pass_dce, pass_dce_loop, pass_cd_dce): Likewise.
3648 * passes.c (init_optimization_passes): Execute rename_ssa_copies,
3649 ccp, forwprop, copy_prop, merge_phi, copy_prop, dce and tail recursion
3651 * tree-ssa-operands.c (add_virtual_operand, get_indirect_ref_operand):
3652 When aliasing is not build, mark statement as volatile.
3654 2007-01-11 Tom Tromey <tromey@redhat.com>
3656 PR preprocessor/15185, PR preprocessor/20989:
3657 * doc/cppopts.texi <-MT>: Update description of algorithm for
3658 computing default target.
3659 <-M, -MD>: Reword "basename" text.
3661 2007-01-11 Roger Sayle <roger@eyesopen.com>
3663 * builtins.c (expand_builtin_pow, expand_builtin_powi,
3664 fold_builtin_cabs, fold_builtin_sqrt, fold_builtin_trunc,
3665 fold_builtin_floor, fold_builtin_ceil, fold_builtin_round,
3666 fold_builtin_int_int_roundingfn, fold_builtin_bitop,
3667 fold_builtin_bswap, real_constp, fold_builtin_pow,
3668 fold_builtin_powi, fold_builtin_signbit, fold_builtin_copysign,
3669 do_mpfr_arg1, do_mpfr_arg2, do_mpfr_arg3, do_mpfr_sincos): Replace
3670 uses of the macro TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
3671 * convert.c (convert_to_pointer): Likewise.
3672 * expr.c (highest_pow2_factor, expand_expr_real_1): Likewise.
3673 * fold-const.c (force_fit_type, fold_negate_expr, int_const_binop,
3674 const_binop, fold_convert_const_int_from_int,
3675 fold_convert_const_int_from_real,
3676 fold_convert_const_real_from_real, sign_bit_p,
3677 optimize_minmax_comparison, extract_muldiv_1, fold_div_compare,
3678 fold_sign_changed_comparison, fold_unary, fold_comparison,
3679 fold_binary, multiple_of_p, tree_Expr_non_zero_p,
3680 fold_negate_const, fold_abs_const, fold_not_const): Likewise.
3681 * print-tree.c (print_node_brief, print_node): Likewise.
3682 * stor-layout.c (place_field, layout_type): Likewise.
3683 * tree-chrec.c (keep_cast): Likewise.
3684 * tree.c (build_vector, build_real, build_real_from_int_cst,
3685 build_complex): Likewise.
3687 2007-01-11 Roger Sayle <roger@eyesopen.com>
3689 * tree.h (TREE_CONSTANT_OVERFLOW): Obsolete. For the time being,
3690 treat TREE_CONSTANT_OVERFLOW as a synonym of TREE_OVERFLOW.
3692 2007-01-11 Paolo Bonzini <bonzini@gnu.org>
3694 * configure.ac (strict1_warn): Rename to strict_warn.
3695 (WERROR, --enable-werror, symlink hacks, stage1_cflags,
3696 cc_set_by_configure, quoted_cc_set_by_configure,
3697 stage_prefix_set_by_configure, quoted_stage_prefix_set_by_configure,
3698 all_boot_languages, all_stagestuff): Remove.
3699 (target_list): Remove bootstrap targets.
3700 * Makefile.in (quickstrap): Unconditionally make a synonym of all.
3701 (BOOT_LANGUAGES, STAGE1_CFLAGS, STAGE1_CHECKING,
3702 REMAKEFLAGS, FLAGS_TO_PASS, PREPEND_DOTDOT_TO_RELATIVE_PATHS,
3703 SUBDIR_FLAGS_TO_PASS, WERROR_FLAGS, STRICT2_WARN, LANG_STAGESTUFF,
3704 VOL_FILES, POSTSTAGE1_FLAGS_TO_PASS, STAGE2_FLAGS_TO_PASS,
3705 STAGEPROFILE_FLAGS_TO_PASS, STAGEFEEDBACK_FLAGS_TO_PASS, stage1_build,
3706 stage1_copy, stage2_build, stage2_copy, stageprofile_build,
3707 stageprofile_copy, stage3_build, stage3_copy, stagefeedback_build,
3708 stagefeedback_copy, stage4_build, clean_s1, clean_sw, bootstrap,
3709 bootstrap-lean, bootstrap2, bootstrap2-lean, bootstrap3,
3710 bootstrap3-lean, bootstrap4, bootstrap4-lean, unstage1, unstage2,
3711 unstage3, unstage4, unstageprofile, unstagefeedback, restage, restage2,
3712 restage3, restage4, restageprofile, restagefeedback, bubbleestrap,
3713 cleanstrap, unstrap, restrap, *compare, *compare3, *compare4,
3714 *compare-lean, *compare3-lean, *compare4-lean, stage1-start, stage1,
3715 stage2-start, stage2, stage3-start, stage3, stage4-start, stage4,
3716 stageprofile-start, stageprofile, stagefeedback-start, stagefeedback,
3717 risky-stage1, risky-stage2, risky-stage3, risky-stage4): Remove.
3718 (ORDINARY_FLAGS_TO_PASS): Rename to FLAGS_TO_PASS.
3719 (STAGECOPYSTUFF, STAGEMOVESTUFF): Consolidate into MOSTLYCLEANFILES.
3720 (mostlyclean): Adjust.
3721 (clean, distclean): Don't mention bootstrap stuff.
3722 * configure: Regenerate.
3723 * ada/config-lang.in, cp/config-lang.in, forttran/config-lang.in,
3724 java/config-lang.in, objc/config-lang.in, objcp/config-lang.in,
3725 treelang/config-lang.in (stagestuff): Remove.
3726 * doc/sourcebuild.texi (stage1, stage2, stage3, stage4,
3727 stageprofile, stagefeedback, stagestuff): Remove mention.
3729 2007-01-11 Nick Clifton <nickc@redhat.com>
3731 * config/mcore/predicates.md (mcore_general_movesrc_operand):
3733 (mcore_general_movdst_operand): Do not accept CONST_INTs.
3734 (mcore_arith_K_S_operand): Run the test for the S constraint not
3735 the test for the M constraint.
3736 (mcore_addsub_operand): Do not accept integer values that are
3737 larger than 32 bits.
3738 * config/mcore/mcore.md: Remove unused constraints from split.
3739 (andsi3): Use HOST_WIDE_INT instead of int to hold an INTVAL.
3741 (allocate_stack): Likewise.
3742 * config/mcore/mcore.c (mcore_print_operand): Restrict output of P
3743 operands to 32 bits.
3744 (mcore_const_costs): Use HOST_WIDE_INT instead of int to hold an
3746 (mcore_and_cost, mcore_modify_comparison, const_ok_for_mcore,
3747 mcore_const_ok_for_inline, mcore_const_trick_uses_not,
3748 try_constant_tricks, mcore_num_ones, mcore_num_zeros,
3749 mcore_output_bclri, mcore_output_andn, output_inline_const,
3750 mcore_output_move, mcore_output_movedouble): Likewise.
3751 (mcore_output_cmov): Use CONST_OK_FOR_M and CONST_OK_FOR_N.
3752 (output_inline_const): Likewise.
3753 (output_inline_const): Fix format strings used in sprintf
3755 * config/mcore/mcore-protos.h: Update prototypes for changed
3756 functions in mcore.c.
3757 * config/mcore/mcore.h (CONST_OK_FOR_I): Cast values to
3758 HOST_WIDE_INT and not int.
3759 (CONST_OK_FOR_J, CONST_OK_FOR_K, CONST_OK_FOR_L, CONST_OK_FOR_M,
3760 CONST_OK_FOR_N): Likewise.
3761 (LEGITIMATE_CONSTANT_P): Also check CONSTANT_P.
3762 (GO_IF_LEGITIMATE_INDEX): Use HOST_WIDE_INT instead of int to hold
3765 2007-01-10 Jan Hubicka <jh@suse.cz>
3767 * tree-vrp.c (remove_range_assertions): Release defs.
3768 * tree-ssa-loop-ivopts.c (rmeove_statement): Likewise.
3769 * tree-ssa-dom.c (remove_stmt_or_phi): Likewise.
3771 2007-01-10 Paul Brook <paul@codesourcery.com>
3773 * config/arm/arm.c (arm_rtx_costs_1): Handle mutiply-subtract.
3774 * config/arm/arm.md (mulsi3subsi): New insn.
3776 2007-01-10 Zdenek Dvorak <dvorakz@suse.cz>
3778 * tree-ssa-loop-manip.c (tree_unroll_loop): Make it a wrapper over ...
3779 (tree_transform_and_unroll_loop): New.
3780 * tree-flow.h (transform_callback, tree_transform_and_unroll_loop):
3783 2007-01-10 Robert Kennedy <jimbob@google.com>
3785 * fold-const.c (fold_comparison): Fold comparisons like (x *
3786 1000 < 0) to (x < 0).
3788 2007-01-10 Ian Lance Taylor <iant@google.com>
3790 * tree-pretty-print.c (dump_generic_node): Print parentheses when
3791 operands have the same priority.
3793 2007-01-10 Tom Tromey <tromey@redhat.com>
3795 * fold-const.c (fold_truthop): Don't check can_use_bit_fields_p.
3796 (fold_binary): Likewise.
3797 * langhooks.c (lhd_can_use_bit_fields_p): Removed.
3798 * langhooks-def.h (lhd_can_use_bit_fields_p): Removed.
3799 (LANG_HOOKS_CAN_USE_BIT_FIELDS_P): Removed.
3800 (LANG_HOOKS_INITIALIZER): Remove LANG_HOOKS_CAN_USE_BIT_FIELDS_P.
3801 * langhooks.h (struct lang_hooks): Removed field
3802 'can_use_bit_fields_p'.
3804 2007-01-10 Ralf Corsépius <ralf.corsepius@rtems.org>
3806 * config/bfin/t-bfin, config/bfin/t-bfin-elf: Remove GCC_CFLAGS.
3808 2007-01-10 Razya Ladelsky <razya@il.ibm.com>
3810 * function.c (get_last_funcdef_no): New function.
3811 * function.h (get_last_funcdef_no): Declare.
3812 * tree-inline.c (initialize_cfun): Add initialization.
3813 (tree_function_versioning): Cleanup.
3815 2007-01-10 Jan Hubicka <jh@suse.cz>
3817 * tree-inline.c (setup_one_parameter): Do not propagate into abnormal
3820 2007-01-10 Sa Liu <saliu@de.ibm.com>
3821 Ben Elliston <bje@au.ibm.com>
3823 * spu.h (STACK_SAVE_AREA): Use VOIDmode for SAVE_FUNCTION, SImode
3824 for SAVE_NONLOCAL and Pmode for any other save level.
3825 * spu-protos.h (spu_restore_stack_block): Declare.
3826 * spu.md (save_stack_block): Remove.
3827 (restore_stack_block): Call spu_restore_stack_block.
3828 * spu.c (spu_restore_stack_block): New function.
3829 (spu_expand_epilogue): Remove old comment.
3831 2007-01-09 Zdenek Dvorak <dvorakz@suse.cz>
3833 PR tree-optimization/30322
3834 * tree-ssa-loop-ivopts.c (fold_affine_expr, iv_value): Removed.
3835 (cand_value_at): Return the value as aff_tree.
3836 (may_eliminate_iv): Convert the bound from aff_tree to tree.
3837 * tree-affine.c (aff_combination_add_cst, aff_combination_add_product,
3838 aff_combination_mult): New functions.
3839 (aff_combination_add): Use aff_combination_add_cst.
3840 (aff_combination_convert): Allow conversions to a wider type.
3841 (tree_to_aff_combination): Handle BIT_NOT_EXPR.
3842 * tree-affine.h (aff_combination_mult): Declare.
3844 2007-01-09 Carlos O'Donell <carlos@codesourcery.com>
3846 * doc/tm.texi: Update documentation to reflect reality of exec
3847 and start file search behaviours. Update copyright year.
3848 * doc/invoke.texi: Explain how GCC_EXEC_PREFIX is used to find
3849 header file directories.
3851 2007-01-09 Uros Bizjak <ubizjak@gmail.com>
3853 * config/i386/i386.md (*sinxf2): Rename to *sinxf2_i387.
3854 (*cosxf2): Rename to cosxf2_i387.
3855 (*sindf2, *sinsf2): Extend operand 1 to XFmode. Macroize patterns
3856 using X87MODEF12 mode macro. Rename patterns to
3857 *sin_extend<mode>xf2_i387. Use SSE_FLOAT_MODE_P to disable patterns
3859 (*cosdf2, *cossf2): Ditto.
3860 (sincosdf3, sincossf3): Ditto. Rewrite corresponding splitters
3861 to match extended input operands.
3862 (sincos<mode>3): New expander.
3863 (*sinextendsfdf2, *cosextendsfdf2, *sincosextendsfdf3): Remove
3864 insn patterns and corresponding splitters.
3866 2007-01-09 Kaz Kojima <kkojima@gcc.gnu.org>
3868 * config/sh/t-linux (TARGET_LIBGCC2_CFLAGS): Delete.
3869 (SHLIB_MAPFILES, SHLIB_LINK, SHLIB_INSTALL): Likewise.
3871 2007-01-09 Nicolas Pitre <nico@cam.org>
3874 * arm/ieee754-df.S (Lad_s): Also test the low word of X for zero.
3876 2007-01-08 Geoffrey Keating <geoffk@apple.com>
3878 * target.h (struct gcc_target): New field library_rtti_comdat.
3879 * target-def.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): New.
3880 (TARGET_CXX): Add TARGET_CXX_LIBRARY_RTTI_COMDAT.
3881 * doc/tm.texi (C++ ABI): Document TARGET_CXX_LIBRARY_RTTI_COMDAT.
3882 * config/darwin.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
3884 2007-01-08 Geoffrey Keating <geoffk@apple.com>
3886 * doc/invoke.texi (Optimize Options): Correct description of -O0.
3888 2007-01-08 Richard Guenther <rguenther@suse.de>
3890 * tree.h (force_fit_type_double): Export.
3891 (force_fit_type): Remove.
3892 * fold-const.c (force_fit_type_double): New function.
3893 (force_fit_type): Remove.
3894 (int_const_binop): Use it.
3895 (fold_convert_const_int_from_int): Likewise.
3896 (fold_convert_const_int_from_real): Likewise.
3897 (fold_div_compare): Likewise.
3898 (fold_sign_changed_comparison): Likewise.
3899 (fold_unary): Likewise.
3900 (fold_negate_const): Likewise.
3901 (fold_abs_const): Likewise.
3902 (fold_not_const): Likewise.
3903 * c-common.c (shorten_compare): Use force_fit_type_double.
3904 * convert.c (convert_to_pointer): Likewise.
3906 2007-01-08 Richard Guenther <rguenther@suse.de>
3908 * tree.h (build_int_cst_wide_type): Export.
3909 * tree.c (build_int_cst_wide_type): New function.
3910 (build_int_cst_wide): Fix comment.
3911 * builtins.c (fold_builtin_object_size): Use build_int_cst
3912 to build -1 or 0 of the correct type. Use fit_double_type
3913 to check for overflow.
3914 * fold-const.c (optimize_bit_field_compare): Use build_int_cst_type
3916 (decode_field_reference): Likewise.
3917 (all_ones_mask_p): Likewise.
3918 (native_interpret_int): Use build_int_cst_wide_type.
3919 (fold_binary): Use build_int_cst_type to build an all-ones
3921 * stor-layout.c (set_sizetype): Use build_int_cst_wide_type.
3923 2007-01-08 Daniel Jacobowitz <dan@codesourcery.com>
3925 * config/pa/t-pa64 (libgcc_stub.a): Use $(T).
3927 2007-01-09 Ben Elliston <bje@au.ibm.com>
3929 * genautomata.c (STATS_OPTION): New option.
3930 (stats_flag): New flag.
3931 (gen_automata_option): Handle it.
3932 (initiate_automaton_gen): Ditto.
3933 (write_automata): Output statistics only if stats_flag is
3934 set. Likewise, output time statistics only if time_flag is set.
3935 * doc/md.texi (Processor pipeline description): Document new flag.
3937 2007-01-08 Richard Guenther <rguenther@suse.de>
3939 * builtins.c (fold_builtin_int_roundingfn): Use fit_double_type.
3940 * tree.c (build_int_cst_type): Likewise.
3941 (size_in_bytes): Don't call force_fit_type on the result.
3942 (int_fits_type_p): Use fit_double_type.
3943 * fold-const.c (fit_double_type): New function.
3944 (force_fit_type): Use it.
3945 * tree.h (fit_double_type): Export.
3947 2007-01-08 Jan Hubicka <jh@suse.cz>
3949 * tree-vectorizer.c (gate_increase_alignment): Fix return type.
3950 * ipa.c (function_and_variable_visibility): Fix return type.
3952 2007-01-08 Richard Guenther <rguenther@suse.de>
3954 * tree-ssa-ccp.c (maybe_fold_offset_to_array_ref): Use type
3955 of offset to build the index.
3956 * tree-pretty-print.c (dump_generic_node): Don't build negated
3957 const just for printing.
3958 * c-pretty-print.c (pp_c_integer_constant): Likewise.
3959 * builtins.c (fold_builtin_int_roundingfn): Check if result
3960 fits the type by using force_fit_type and comparing the result.
3961 * predict.c (predict_loops): Use compare_tree_int for comparison.
3962 * tree.c (build_int_cst): Fall back to integer_type_node for
3964 (build_int_cst_wide): Assert type is non-null.
3966 2007-01-08 Roberto Costa <roberto.costa@st.com>
3968 * tree-vrp.c (extract_range_from_cond_expr): New.
3969 (extract_range_from_expr): Handle COND_EXPR nodes used as expressions.
3970 * tree-ssa-ccp.c (get_maxval_strlen): Handle COND_EXPR nodes used
3972 (fold_stmt): Bug fix, avoid infinite recursion when folding COND_EXPRs.
3973 * tree-ssa-forwprop.c (simplify_cond, forward_propagate_into_cond,
3974 tree_ssa_forward_propagate_single_use_vars): Handle COND_EXPR nodes
3975 used as expressions.
3976 * tree-object-size.c (cond_expr_object_size): New.
3977 (collect_object_sizes_for): Handle COND_EXPR nodes used as expressions.
3979 2007-01-08 Jan Hubicka <jh@suse.cz>
3981 * tree-ssa-forwprop.c (forward_propagate_into_cond,
3982 tree_ssa_forward_propagate_single_use_va): Release defs of propagated
3985 2007-01-08 Richard Guenther <rguenther@suse.de>
3987 PR tree-optimization/23603
3988 * tree-vrp.c (set_value_range_to_truthvalue): New function.
3989 (extract_range_from_binary): Fall back to truthvalue instead of
3990 varying for TRUTH_*_EXPR.
3991 (extract_range_from_comparison): Fall back to truthvalue instead of
3993 (vrp_visit_phi_node): Don't adjust new range bounds to +INF/-INF
3994 if all visited PHI values were constant.
3996 2007-01-08 Jan Hubicka <jh@suse.cz>
3998 * cgraphunit.c (cgraph_process_new_functions): Reset reachable flag.
3999 (cgraph_analyze_function): break out from ...
4000 (cgraph_finalize_compilation_unit): ... here.
4001 (cgraph_expand_function): Remove forgoten commented out line.
4002 (cgraph_optimize): Analyze functions.
4004 2007-01-08 Jan Hubicka <jh@suse.cz>
4006 * tree-pas.h (TODO_remove_function): New flag.
4007 (TODO_update*): Renumber.
4008 (pass_ipa_increase_alignment,
4009 pass_ipa_function_and_variable_visibility): New passes.
4010 * cgraphunit.c (cgraph_increase_alignment): Move to tree-vectorizer.c
4011 (cgraph_function_and_variable_visibility): Move to ipa.c
4012 (cgraph_optimize): Don't call cgraph_function_and_variable_visibility,
4013 cgraph_increase_alignment.
4014 * ipa-inline.c (cgraph_decide_inlining): Don't push timevar.
4015 (cgraph_decide_inlining_incrementally): Push TV_INTEGRATION before
4016 calling tree-inline.
4017 (cgraph_early_inlining): Do not call cgraph_remove_unreachable_nodes.
4018 (pass_ipa_inline, pass_early_ipa_inlining): Set TODO_remove_functions
4019 * tree-vectorizer.c (increase_alignment): Move here from cgraphunit.c
4020 (gate_increase_alignment): New function.
4021 (pass_ipa_increase_alignment): New pass.
4022 * ipa.c: Inline tree-pass.h and timevar.h
4023 (function_and_variable_visibility): Move here from cgraphunit.c
4024 * tree-optimize.c (pass_early_local_passes): Add TODO_remove_functions.
4025 * passes.c (init_optimization_passes): Add the two new passes.
4026 (execute_todo): Handle cgraph_remove_functions.
4028 2007-01-08 Nick Clifton <nickc@redhat.com>
4030 * config/frv/predicates.md (reg_or_0_operand): Accept
4033 2007-01-08 Ralf Corsépius <ralf.corsepius@rtems.org>
4035 * config/bfin/rtems.h, config/bfin/t-rtems: New.
4036 * config.gcc: Add bfin*-rtems*.
4038 2007-01-08 Mark Shinwell <shinwell@codesourcery.com>
4040 * c.opt: Add -flax-vector-conversions.
4041 * c-typeck.c (convert_for_assignment): Pass flag to
4042 vector_types_convertible_p to allow emission of note.
4043 (digest_init): Likewise.
4044 * c-opts.c: Handle -flax-vector-conversions.
4045 * c-common.c (flag_lax_vector_conversions): New.
4046 (vector_types_convertible_p): Unless -flax-vector conversions
4047 has been passed, disallow conversions between vectors with
4048 differing numbers of subparts and/or element types. If such
4049 a conversion is disallowed, possibly emit a note on the first
4050 occasion only to inform the user of -flax-vector-conversions.
4051 The new last argument specifies this.
4052 * c-common.h (flag_lax_vector_conversions): New.
4053 (vector_types_convertible_p): Add extra argument.
4054 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Use
4055 char_type_node for V*QI type vectors.
4056 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins):
4057 Update to satisfy new typechecking rules.
4058 * config/rs6000/altivec.h (vec_cmple): Use vec_cmpge, for both
4060 * doc/invoke.texi (C Dialect Options): Document
4061 -flax-vector-conversions.
4063 2007-01-08 Mark Shinwell <shinwell@codesourcery.com>
4065 PR tree-optimization/29877
4066 * tree-ssa-ter.c (is_replaceable_p): Deem assignments with
4067 a register variable on the RHS to not be replaceable.
4069 2007-01-08 Chen Liqin <liqin@sunnorth.com.cn>
4070 * config/score/t-score-elf (MULTILIB_OPTIONS): Change.
4071 * config/score/predicates.md (const_uimm5, sr0_operand, const_simm12,
4072 const_simm15, const_pow2, const_npow2): Added.
4073 * config/score/misc.md (insv, extv, extzv, movmemsi,
4074 move_lbu_a/b, mov_lhu_a/b etc): Added and fix some bug.
4075 * config/score/score.c (score_address_cost, score_select_cc_mode):
4077 Change CONST_OK_FOR_LETTER_P/EXTRA_CONSTRAINT define.
4078 Update score_rtx_costs for MACRO TARGET_RTX_COSTS.
4079 Update score_print_operand.
4080 * config/score/score.h (DATA_ALIGNMENT, SELECT_CC_MODE): Added.
4081 Adjust register allocate order and update some macro define.
4082 * config/score/score-mdaux.c (mdx_unaligned_load, mdx_unsigned_store,
4083 mdx_block_move_straight, mdx_block_move_loop_head,
4084 mdx_block_move_loop_body, mdx_block_move_loop_foot, mdx_block_move_loop,
4085 mdx_block_move): Added.
4086 (mdx_movsicc, mdp_select_add_imm, mdp_select, mds_zero_extract_andi,
4087 mdp_limm): Updated and fix some bug and typo.
4088 * config/score/score.md (movqi/hi/si, add/sub/zero/ext): Updated.
4089 (movsf, movdf, doloop_end): Added.
4091 2007-01-08 Kazu Hirata <kazu@codesourcery.com>
4093 * config/arm/arm.c, config/arm/arm.h, config/arm/arm.md,
4094 config/arm/thumb2.md: Fix comment typos.
4095 * doc/extend.texi: Fix a typo.
4097 2007-01-07 Eric Christopher <echristo@apple.com>
4099 * configure.ac: Check for __stack_chk_fail for darwin.
4100 * configure: Regenerate.
4102 2007-01-07 Richard Guenther <rguenther@suse.de>
4104 * tree-vrp.c (extract_range_from_assert): CSE calls to
4105 compare_values where possible.
4106 (extract_range_from_unary_expr): Likewise.
4108 2007-01-07 Anatoly Sokolov <aesok@post.ru>
4110 * config/avr/avr-protos.h (call_insn_operand): Delete prototype.
4111 * config/avr/avr.c (call_insn_operand): Delete function.
4112 * config/avr/avr.md (*pushqi, *pushhi, *pushsi, *pushsf): Use REG_SP
4113 instead of register number. Use predicates.
4114 * config/avr/predicates.md (const0_operand, reg_or_0_operand,
4115 call_insn_operand): Add.
4117 2007-01-06 Jan Hubicka <jh@suse.cz>
4119 * tree-pass.h (pass_build_cgraph_edges): Declare.
4120 * cgraphunit.c (record_refernece): Move to cgraphbuild.c
4121 (visited_nodes): Remove.
4122 (cgraph_create_edges): Move to cgraphbuild.c; rename to
4123 build_cgrpah_edges; make visited_nodes local.
4124 (cgraph_process_new_functions): DO not call initialize_inline_failed.
4125 (record_references_in_initializer): Move to cgraphbuild.c
4126 (initialize_inline_failed, rebuild_cgraph_edges,
4127 pass_rebuild_cgraph_edges): Move to cgraphbuild.c.
4128 (verify_cgraph_node): Make visited_nodes local.
4129 (cgraph_analyze_function): Do not call cgraph_create_edges and
4130 initialize_inline_failed.
4131 (cgraph_expand_function): Do not call cgraph_lower_function;
4132 assert that function is already lowered.
4133 * Makefile.in (cgraphbuild.o): New.
4134 * passes.c (init_optimization_passes): Add pass_build_cgraph_edges
4135 at the end of lowering passes.
4137 2007-01-06 Steven Bosscher <steven@gcc.gnu.org>
4139 * ifcvt.c (cond_move_convert_if_block): New function, code
4140 factored out from...
4141 (cond_move_process_if_block): ...here. Call the new function
4142 on the THEN and ELSE blocks.
4143 (merge_if_block): Do not copy global_live_at_end, merge_blocks
4144 already takes care of this.
4146 2007-01-05 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
4149 * tree.h (TREE_OVERFLOW_P): New.
4150 * c-typeck.c (parser_build_unary_op): Warn only if result
4151 overflowed and operands did not.
4152 (parser_build_binary_op): Likewise.
4153 (convert_for_assignment): Remove redundant overflow_warning.
4154 * c-common.c (overflow_warning): Don't check or set TREE_OVERFLOW.
4156 2007-01-05 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
4158 * c-typeck.c (store_init_value): Split over two lines to follow
4159 the GNU coding style.
4161 2007-01-05 Benjamin Kosnik <bkoz@redhat.com>
4163 * c-cppbuiltin.c (c_cpp_builtins): __GXX_EXPERIMENTAL_CPP0X__ to
4164 __GXX_EXPERIMENTAL_CXX0X__.
4165 * doc/cpp.texi: Same.
4167 2007-01-05 Richard Guenther <rguenther@suse.de>
4170 * tree.c (get_narrower): Do not construct COMPONENT_REFs
4171 with mismatched types. Instead explicitly build a
4172 conversion NOP_EXPR.
4174 2007-01-05 Ian Lance Taylor <iant@google.com>
4176 * c-common.c (decl_with_nonnull_addr_p): New function.
4177 (c_common_truthvalue_conversion): Call it.
4178 * c-typeck.c (build_binary_op): Likewise.
4179 * c-common.h (decl_with_nonnull_addr_p): Declare.
4181 2007-01-05 Jakub Jelinek <jakub@redhat.com>
4184 * libgcc2.c (__divdc3): Compare c and d against 0.0 instead of
4187 2007-01-05 Joel Brobecker <brobecker@adacore.com>
4189 * doc/install.texi (Final install): Document the fact that
4190 the GNAT runtime should not be stripped.
4192 2007-01-04 Jan Hubicka <jh@suse.cz>
4194 * tree-inline.c (fold_marked_statements): Update operand caches
4195 and EH after folding
4197 2007-01-04 Ian Lance Taylor <iant@google.com>
4199 * c-common.c (check_function_nonnull): Whitespace fix.
4201 2007-01-04 Jan Hubicka <jh@suse.cz>
4203 * tree-optimize.c (execute_fixup_cfg): Correct previously mistakely
4204 comitted older version of patch.
4205 (pass_fixup_cfg): Add TODOs to verify flow and statements, dump
4206 function, celanup cfg and collect garbage.
4208 2007-01-04 Mike Stump <mrs@apple.com>
4210 * Makefile.in (mostlyclean): Don't remove libgcc anymore.
4213 2007-01-04 Eric Christopher <echristo@apple.com>
4215 * libgcc2.c (__bswapsi2): Use SItype.
4216 (__bswapdi2): Use DItype.
4217 * libgcc2.h: Update for above.
4219 2007-01-04 Paul Brook <paul@codesourcery.com>
4221 * config/arm/arm.md (arm_mulsi3, thumb_mulsi3, mulsi3_compare0,
4222 mulsi_compare0_scratch, mulsi3addsi, mulsi3addsi_compare0,
4223 mulsi3addsi_compare0_scratch, mulsidi3adddi, mulsidi3,
4224 umulsidi3, umulsidi3adddi, smulsi3_highpart,
4225 umulsi3_highpart): Make conditional on !arm_arch6.
4226 (arm_mulsi3_v6, thumb_mulsi3_v6, mulsi3_compare0_v6,
4227 mulsi_compare0_scratch_v6, mulsi3addsi_v6, mulsi3addsi_compare0_v6,
4228 mulsi3addsi_compare0_scratch_v6, mulsidi3adddi_v6, mulsidi3_v6,
4229 umulsidi3_v6, umulsidi3adddi_v6, smulsi3_highpart_v6,
4230 umulsi3_highpart_v6): New insns.
4232 2007-01-04 Roger Sayle <roger@eyesopen.com>
4234 * fold-const.c (fold_convert): When casting an expression to void,
4235 fold_ignored_result may discover a GIMPLE_MODIFY_STMT which doesn't
4236 have a type. Instead of attempting to build a NOP_EXPR, return
4237 these "special" trees directly.
4239 2007-01-04 Joseph Myers <joseph@codesourcery.com>
4241 * config/rs6000/rs6000.c (rs6000_rtx_costs): Make adjustment for
4242 MULT inside MINUS as either argument. Use rs6000_cost->dmul -
4243 rs6000_cost->fp not 0 as adjustment for outer NEG.
4245 2007-01-04 Jan Hubicka <jh@suse.cz>
4247 * cgraph.c (cgraph_release_function_body): New function.
4248 (cgraph_remove_node): Use it.
4249 * cgraph.h (cgraph_release_function_body): Declare.
4250 * cgraphunit.c (cgraph_expand_function): Use it.
4251 * ipa.c (cgraph_remove_unreahchable_nodes): Use it.
4252 * tree-ssa.c (delete_tree_ssa): Allow to be called before aliasing
4253 is initialized and while compilation of other function is running.
4254 * tree-optimize.c (execute_free_cfg_annotations): Move code to clear
4255 statement CFG annotations from here to ...
4256 * tree-cfg.c (delete_tree_cfg_annotations): ... here.
4258 2007-01-04 Zdenek Dvorak <dvorakz@suse.cz>
4260 * cfgloop.h (enum li_flags): Make the constants powers of two.
4262 2007-01-04 Jan Hubicka <jh@suse.cz>
4264 * tree-inline.c (copy_bb): Insert new statements to statements_to_fold
4266 (fold_marked_statements): New function.
4267 (optimize_inline_calls, tree_function_versioning): Fold new statements.
4268 * tree-inline.h (copy_body_data): Add statements_to_fold.
4270 2007-01-03 Daniel Jacobowitz <dan@codesourcery.com>
4272 * config.gcc: Mention libgcc/config.host.
4273 * Makefile.in: Update comments mentioning libgcc.
4274 (LIBGCC, INSTALL_LIBGCC, GCC_PARTS, mklibgcc): Delete.
4275 (all.cross, start.encap, rest.encap, rest.cross): Update
4276 dependencies for libgcc move.
4277 (libgcc.mk, LIBGCC_DEPS, libgcov.a, libgcc.a, stmp-multilib)
4278 (clean-target, clean-target-libgcc): Delete.
4279 (srcdirify, GCC_EXTRA_PARTS): New macros.
4280 (libgcc-support, libgcc.mvars): New rules.
4281 (distclean): Remove mention of mklibgcc.
4282 (install): Don't reference INSTALL_LIBGCC.
4283 (install-common): Don't reference EXTRA_PARTS.
4284 (install-libgcc, install-multilib): Delete rules.
4285 * mklibgcc.in: Delete file.
4286 * doc/configfiles.texi: Don't mention mklibgcc.
4288 * config/i386/t-darwin (SHLIB_VERPFX): Delete (moved to libgcc).
4289 * config/i386/t-darwin64 (SHLIB_VERPFX): Likewise.
4290 * config/rs6000/t-darwin (SHLIB_VERPFX): Likewise.
4291 * config/rs6000/t-ppccomm (TARGET_LIBGCC2_CFLAGS, SHLIB_MAPFILES)
4292 (mklibgcc, ldblspecs): Likewise.
4294 * config/i386/t-nwld (libgcc.def, libc.def, libpcre.def)
4295 (posixpre.def): Use $(T).
4296 (SHLIB_EXT, SHLIB_NAME, SHLIB_SLIBDIR_QUAL, SHLIB_DEF, SHLIB_MAP)
4297 (SHLIB_SRC, SHLIB_INSTALL): Delete.
4298 (SHLIB_LINK): Make dummy.
4299 * config/t-slibgcc-darwin: Delete contents except for dummy SHLIB_LINK.
4301 * config/frv/t-linux (EXTRA_MULTILIB_PARTS): Clear.
4303 * config/alpha/t-crtfm: Use $(T) in rules for EXTRA_PARTS.
4304 * config/alpha/t-vms, config/alpha/t-vms64, config/fr30/t-fr30,
4305 config/i386/t-rtems-i386, config/ia64/t-ia64, config/rs6000/t-beos,
4306 config/rs6000/t-newas, config/sparc/t-elf: Likewise.
4308 * configure.ac (all_outputs): Remove mklibgcc.
4309 * configure: Regenerated.
4311 2007-01-03 Josh Conner <jconner@apple.com>
4314 * calls.c (compute_argument_addresses): Set stack and stack_slot
4315 for partial args, too.
4316 (store_one_arg): Use locate.size.constant for the size when
4317 generating a save_area.
4319 2007-01-03 Robert Kennedy <jimbob@google.com>
4321 * tree-cfg.c (tree_merge_blocks): Release SSA_NAME phi results
4322 whose definitions are deleted due to basic block merging.
4324 2007-01-03 Paul Brook <paul@codesourcery.com>
4327 * config/arm/arm.c (output_return_instruction): Pop PC in interrupt
4329 (use_return_insn): Return 0 for Thumb interrupt functions.
4330 (print_multi_reg): Add rfe argument for IRQ returns.
4331 (arm_output_epilogue): Pop interrupt return address directly into PC.
4332 (arm_expand_prologue): Only adjust IRQ return address in Arm mode.
4334 2007-01-03 Paul Brook <paul@codesourcery.com>
4336 Merge from sourcerygxx-4_1.
4337 * config/arm/thumb2.md: New file.
4338 * config/arm/elf.h (JUMP_TABLES_IN_TEXT_SECTION): Return True for
4340 * config/arm/coff.h (JUMP_TABLES_IN_TEXT_SECTION): Ditto.
4341 * config/arm/aout.h (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
4342 (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump tables.
4343 * config/arm/aof.h (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump
4345 (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
4346 * config/arm/ieee754-df.S: Use macros for Thumb-2/Unified asm
4348 * config/arm/ieee754-sf.S: Ditto.
4349 * config/arm/arm.c (thumb_base_register_rtx_p): Rename...
4350 (thumb1_base_register_rtx_p): ... to this.
4351 (thumb_index_register_rtx_p): Rename...
4352 (thumb1_index_register_rtx_p): ... to this.
4353 (thumb_output_function_prologue): Rename...
4354 (thumb1_output_function_prologue): ... to this.
4355 (thumb_legitimate_address_p): Rename...
4356 (thumb1_legitimate_address_p): ... to this.
4357 (thumb_rtx_costs): Rename...
4358 (thumb1_rtx_costs): ... to this.
4359 (thumb_compute_save_reg_mask): Rename...
4360 (thumb1_compute_save_reg_mask): ... to this.
4361 (thumb_final_prescan_insn): Rename...
4362 (thumb1_final_prescan_insn): ... to this.
4363 (thumb_expand_epilogue): Rename...
4364 (thumb1_expand_epilogue): ... to this.
4365 (arm_unwind_emit_stm): Rename...
4366 (arm_unwind_emit_sequence): ... to this.
4367 (thumb2_legitimate_index_p, thumb2_legitimate_address_p,
4368 thumb1_compute_save_reg_mask, arm_dwarf_handle_frame_unspec,
4369 thumb2_index_mul_operand, output_move_vfp, arm_shift_nmem,
4370 arm_save_coproc_regs, thumb_set_frame_pointer, arm_print_condition,
4371 thumb2_final_prescan_insn, thumb2_asm_output_opcode, arm_output_shift,
4372 thumb2_output_casesi): New functions.
4373 (TARGET_DWARF_HANDLE_FRAME_UNSPEC): Define.
4374 (FL_THUMB2, FL_NOTM, FL_DIV, FL_FOR_ARCH6T2, FL_FOR_ARCH7,
4375 FL_FOR_ARCH7A, FL_FOR_ARCH7R, FL_FOR_ARCH7M, ARM_LSL_NAME,
4376 THUMB2_WORK_REGS): Define.
4377 (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv, arm_condexec_count,
4378 arm_condexec_mask, arm_condexec_masklen)): New variables.
4379 (all_architectures): Add armv6t2, armv7, armv7a, armv7r and armv7m.
4380 (arm_override_options): Check new CPU capabilities.
4381 Set new architecture flag variables.
4382 (arm_isr_value): Handle v7m interrupt functions.
4383 (user_return_insn): Return 0 for v7m interrupt functions. Handle
4385 (const_ok_for_arm): Handle Thumb-2 constants.
4386 (arm_gen_constant): Ditto. Use movw when available.
4387 (arm_function_ok_for_sibcall): Return false for v7m interrupt
4389 (legitimize_pic_address, arm_call_tls_get_addr): Handle Thumb-2.
4390 (thumb_find_work_register, arm_load_pic_register,
4391 legitimize_tls_address, arm_address_cost, load_multiple_sequence,
4392 emit_ldm_seq, emit_stm_seq, arm_select_cc_mode, get_jump_table_size,
4393 print_multi_reg, output_mov_long_double_fpa_from_arm,
4394 output_mov_long_double_arm_from_fpa, output_mov_double_fpa_from_arm,
4395 output_mov_double_fpa_from_arm, output_move_double,
4396 arm_compute_save_reg_mask, arm_compute_save_reg0_reg12_mask,
4397 output_return_instruction, arm_output_function_prologue,
4398 arm_output_epilogue, arm_get_frame_offsets, arm_regno_class,
4399 arm_output_mi_thunk, thumb_set_return_address): Ditto.
4400 (arm_expand_prologue): Handle Thumb-2. Use arm_save_coproc_regs.
4401 (arm_coproc_mem_operand): Allow POST_INC/PRE_DEC.
4402 (arithmetic_instr, shift_op): Use arm_shift_nmem.
4403 (arm_print_operand): Use arm_print_condition. Handle '(', ')', '.',
4405 (arm_final_prescan_insn): Use extract_constrain_insn_cached.
4406 (thumb_expand_prologue): Use thumb_set_frame_pointer.
4407 (arm_file_start): Output directive for unified syntax.
4408 (arm_unwind_emit_set): Handle stack alignment instruction.
4409 * config/arm/lib1funcs.asm: Remove default for __ARM_ARCH__.
4410 Add v6t2, v7, v7a, v7r and v7m.
4411 (RETLDM): Add Thumb-2 code.
4412 (do_it, shift1, do_push, do_pop, COND, THUMB_SYNTAX): New macros.
4413 * config/arm/arm.h (TARGET_CPU_CPP_BUILTINS): Define __thumb2__.
4414 (TARGET_THUMB1, TARGET_32BIT, TARGET_THUMB2, TARGET_DSP_MULTIPLY,
4415 TARGET_INT_SIMD, TARGET_UNIFIED_ASM, ARM_FT_STACKALIGN, IS_STACKALIGN,
4416 THUMB2_TRAMPOLINE_TEMPLATE, TRAMPOLINE_ADJUST_ADDRESS,
4417 ASM_OUTPUT_OPCODE, THUMB2_GO_IF_LEGITIMATE_ADDRESS,
4418 THUMB2_LEGITIMIZE_ADDRESS, CASE_VECTOR_PC_RELATIVE,
4419 CASE_VECTOR_SHORTEN_MODE, ADDR_VEC_ALIGN, ASM_OUTPUT_CASE_END,
4420 ADJUST_INSN_LENGTH): Define.
4421 (TARGET_REALLY_IWMMXT, TARGET_IWMMXT_ABI, CONDITIONAL_REGISTER_USAGE,
4422 STATIC_CHAIN_REGNUM, HARD_REGNO_NREGS, INDEX_REG_CLASS,
4423 BASE_REG_CLASS, MODE_BASE_REG_CLASS, SMALL_REGISTER_CLASSES,
4424 PREFERRED_RELOAD_CLASS, SECONDARY_OUTPUT_RELOAD_CLASS,
4425 SECONDARY_INPUT_RELOAD_CLASS, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P,
4426 TRAMPOLINE_SIZE, INITIALIZE_TRAMPOLINE, HAVE_PRE_INCREMENT,
4427 HAVE_POST_DECREMENT, HAVE_PRE_DECREMENT, HAVE_PRE_MODIFY_DISP,
4428 HAVE_POST_MODIFY_DISP, HAVE_PRE_MODIFY_REG, HAVE_POST_MODIFY_REG,
4429 REGNO_MODE_OK_FOR_BASE_P, LEGITIMATE_CONSTANT_P,
4430 REG_MODE_OK_FOR_BASE_P, REG_OK_FOR_INDEX_P, GO_IF_LEGITIMATE_ADDRESS,
4431 LEGITIMIZE_ADDRESS, THUMB2_LEGITIMIZE_ADDRESS,
4432 GO_IF_MODE_DEPENDENT_ADDRESS, MEMORY_MOVE_COST, BRANCH_COST,
4433 ASM_APP_OFF, ASM_OUTPUT_CASE_LABEL, ARM_DECLARE_FUNCTION_NAME,
4434 FINAL_PRESCAN_INSN, PRINT_OPERAND_PUNCT_VALID_P,
4435 PRINT_OPERAND_ADDRESS): Adjust for Thumb-2.
4436 (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv): New declarations.
4437 * config/arm/arm-cores.def: Add arm1156t2-s, cortex-a8, cortex-r4 and
4439 * config/arm/arm-tune.md: Regenerate.
4440 * config/arm/arm-protos.h: Update prototypes.
4441 * config/arm/vfp.md: Enable patterns for Thumb-2.
4442 (arm_movsi_vfp): Add movw alternative. Use output_move_vfp.
4443 (arm_movdi_vfp, movsf_vfp, movdf_vfp): Use output_move_vfp.
4444 (thumb2_movsi_vfp, thumb2_movdi_vfp, thumb2_movsf_vfp,
4445 thumb2_movdf_vfp, thumb2_movsfcc_vfp, thumb2_movdfcc_vfp): New.
4446 * config/arm/libunwind.S: Add Thumb-2 code.
4447 * config/arm/constraints.md: Update include Thumb-2.
4448 * config/arm/ieee754-sf.S: Add Thumb-2/Unified asm support.
4449 * config/arm/ieee754-df.S: Ditto.
4450 * config/arm/bpabi.S: Ditto.
4451 * config/arm/t-arm (MD_INCLUDES): Add thumb2.md.
4452 * config/arm/predicates.md (low_register_operand,
4453 low_reg_or_int_operand, thumb_16bit_operator): New.
4454 (thumb_cmp_operand, thumb_cmpneg_operand): Rename...
4455 (thumb1_cmp_operand, thumb1_cmpneg_operand): ... to this.
4456 * config/arm/t-arm-elf: Add armv7 multilib.
4457 * config/arm/arm.md: Update patterns for Thumb-2 and Unified asm.
4459 (UNSPEC_STACK_ALIGN, ce_count): New.
4460 (arm_incscc, arm_decscc, arm_umaxsi3, arm_uminsi3,
4461 arm_zero_extendsidi2, arm_zero_extendqidi2): New
4463 * config/arm/fpa.md: Update patterns for Thumb-2 and Unified asm.
4464 (thumb2_movsf_fpa, thumb2_movdf_fpa, thumb2_movxf_fpa,
4465 thumb2_movsfcc_fpa, thumb2_movdfcc_fpa): New insns.
4466 * config/arm/cirrus.md: Update patterns for Thumb-2 and Unified asm.
4467 (cirrus_thumb2_movdi, cirrus_thumb2_movsi_insn,
4468 thumb2_cirrus_movsf_hard_insn, thumb2_cirrus_movdf_hard_insn): New
4470 * doc/extend.texi: Document ARMv7-M interrupt functions.
4471 * doc/invoke.texi: Document Thumb-2 new cores+architectures.
4473 2007-01-03 Jakub Jelinek <jakub@redhat.com>
4475 * unwind-dw2.c (SIGNAL_FRAME_BIT, EXTENDED_CONTEXT_BIT): Define.
4476 (struct _Unwind_Context): Rename args_size to flags, remove
4477 signal_frame field, add a new args_size field and version field.
4478 (_Unwind_IsSignalFrame, _Unwind_SetSignalFrame,
4479 _Unwind_IsExtendedContext): New inline functions.
4480 (_Unwind_GetGR, _Unwind_SetGR, _Unwind_GetGRPtr, _Unwind_SetGRPtr):
4481 Assume by_value array is only present if _Unwind_IsExtendedContext.
4482 (_Unwind_GetIPInfo, execute_cfa_program, uw_frame_state_for): Use
4483 _Unwind_IsSignalFrame.
4484 (__frame_state_for): Initialize context.flags to EXTENDED_CONTEXT_BIT.
4485 (uw_update_context_1): Use _Unwind_SetSignalFrame.
4486 (uw_init_context_1): Initialize context->flags to
4487 EXTENDED_CONTEXT_BIT.
4488 * config/rs6000/linux-unwind.h (frob_update_context): Use
4489 _Unwind_SetSignalFrame.
4491 2007-01-03 Andrew Pinski <andrew_pinski@playstation.sony.com>
4494 * gimplify.c (gimplify_modify_expr_complex_part): Move below
4495 tree_to_gimple_tuple. Call tree_to_gimple_tuple when we need
4498 2007-01-03 Kazu Hirata <kazu@codesourcery.com>
4500 * config/i386/i386.h (NON_STACK_REG_P, REGNO_OK_FOR_SIREG_P,
4501 REGNO_OK_FOR_DIREG_P, REWRITE_ADDRESS, ASM_OPERAND_LETTER,
4502 RET, AT_SP): Remove.
4503 * config/i386/i386.md (*sse_prologue_save_insn): Use return
4506 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
4507 * cfgbuild.c (find_basic_blocks): Likewise.
4508 * cfgrtl.c (rtl_create_basic_block): Likewise.
4509 * function.c (temp_slots_at_level): Likewise.
4510 * reg-stack.c (stack_regs_mentioned): Likewise.
4511 * regclass.c (allocate_reg_info): Likewise.
4512 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
4513 set_bb_for_stmt, move_block_to_fn): Likewise.
4514 * tree-complex.c (tree_lower_complex): Likewise.
4515 * vec.h (VEC_safe_grow_cleared): New.
4517 * cgraphunit.c, tree-ssa-alias.c: Fix comment typos.
4519 2007-01-03 Zdenek Dvorak <dvorakz@suse.cz>
4521 * loop-unswitch.c (unswitch_loop): Pass probabilities to loopify.
4522 * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Pass probabilities
4524 * cfgloopmanip.c (scale_loop_frequencies): Export.
4525 (loopify): Scale the frequencies by prescribed coefficients.
4526 (set_zero_probability): New function.
4527 (duplicate_loop_to_header_edge): Improve updating of frequencies.
4528 (lv_adjust_loop_entry_edge, loop_version): Set probabilities
4529 and frequencies according to arguments.
4530 * tree-ssa-loop-manip.c (tree_unroll_loop): Set probabilities
4532 * cfg.c (scale_bbs_frequencies_int): Allow scaling the frequencies up.
4533 * modulo-sched.c (sms_schedule): Set probabilities for entering
4534 versioned loop correctly.
4535 * tree-vect-transform.c (vect_transform_loop): Ditto.
4536 * cfgloop.h (loopify, loop_version): Declaration changed.
4537 (scale_loop_frequencies): Declared.
4539 2007-01-02 Jan Hubicka <jh@suse.cz>
4541 * cgraph.c: Include tree-flow.h
4542 (cgraph_add_new-function): Handle IPA_SSA mode; execute
4544 * cgraph.h (enum cgraph_state): Add CGRAPH_STATE_IPA_SSA.
4545 * tree-pass.h (pass_all_early_optimizations): Declare.
4546 * cgraphunit.c (cgraph_process_new_functions): Add IPA_SSA; execute
4548 (cgraph_analyze_function): Do early_local_passes.
4549 * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
4550 Do not add referenced vars.
4551 * tree-optimize.c (gate_all_optimizations): Do not execute when not in
4553 (gate_all_early_local_passes): New gate.
4554 (pass_early_local_passes): Use new gate.
4555 (execute_early_local_optimizations): New functions.
4556 (gate_all_early_optimizations): New gate.
4557 (pass_all_early_optimizations): New pass.
4558 (execute_free_datastructures): Free SSA only when initialized.
4559 (gate_init_datastructures): Init only when optimizing.
4560 (tree_lowering_passes): Do early local passes when called late.
4561 * tree-profile.c (do_tree_profiling): Don't profile functions added
4563 (do_early_tree_profiling, pass_early_tree_profile): Kill.
4564 * tree-cfg.c (update_modified_stmts): Do not update when operands are
4566 * passes.c (init_optimizations_passes): Reorder so we go into SSA
4567 during early_local_passes.
4568 * Makefile.in (cgraph.o): Add dependency on tree-flow.h.
4571 2007-01-02 Carlos O'Donell <carlos@codesourcery.com>
4573 * Makefile.in: Update copyright year.
4575 2007-01-02 Carlos O'Donell <carlos@codesourcery.com>
4577 * Makefile.in: Export GCC_EXEC_PREFIX before calling $(RUNTEST)
4578 in $(lang_checks) and check-consistency targets.
4580 2007-01-02 Jan Hubicka <jh@suse.cz>
4582 * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
4583 Do not add referenced vars.
4584 * tree-cfg.c (update_modified_stmts): Do not update when SSA operands
4586 * passes.c (init_optimization_passes): Put mudflap_2 after
4587 free_datastructures.
4589 2007-01-02 Jan Hubicka <jh@suse.cz>
4591 * tree-optimize (execute_fixup_cfg): Set after_inlining flag.
4592 Set NOTHROW flag on call statements proved to be nothrow.
4593 Update statement of local calls so new pure/const functions are
4594 updated. Update_ssa when in ssa form. Mark PHI nodes of nonlocal
4596 (tree_rest_of_compilation): Register hooks and initialize bitmap
4597 early. Do not set after_inlining flag.
4599 2007-01-02 Steve Ellcey <sje@cup.hp.com>
4601 * sbitmap.c (HOST_BITS_PER_LONG_LONG): Change to
4602 HOST_BITS_PER_LONGLONG
4604 2007-01-02 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
4607 * c-typeck.c (store_init_value): Don't emit pedantic overflow
4608 warning for non-static initializers.
4610 2007-01-02 Steven Bosscher <steven@gcc.gnu.org>
4612 * config/alpha/alpha.md, arm/arm.c, darwin.c, frv/frv.md,
4613 m32r/m32r.c, m32r/m32r.c, mn10300/mn10300.md, pa/pa.c,
4614 rs6000/rs6000.c, s390/s390.md, sh/sh.md, sparc/sparc.c:
4615 Always use set_unique_reg_note to add REG_EQUAL notes.
4617 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
4620 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
4622 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
4623 * cfgbuild.c (find_basic_blocks): Likewise.
4624 * cfgrtl.c (rtl_create_basic_block): Likewise.
4625 * function.c (temp_slots_at_level): Likewise.
4626 * reg-stack.c (stack_regs_mentioned): Likewise.
4627 * regclass.c (allocate_reg_info): Likewise.
4628 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
4629 set_bb_for_stmt, move_block_to_fn): Likewise.
4630 * tree-complex.c (tree_lower_complex): Likewise.
4631 * vec.h (VEC_safe_grow_cleared): New.
4633 2007-01-02 Ian Lance Taylor <iant@google.com>
4635 * c-common.c (c_common_truthvalue_conversion): When warning about
4636 using an assignment as a truth value, set TREE_NO_WARNING.
4638 2007-01-02 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
4641 * c.opt (Wold-style-declaration): New.
4642 * doc/invoke.texi (C-only Warning Options): New.
4643 (Wold-style-declaration): Document it.
4644 (Wextra): Enabled by -Wextra.
4645 * c-opts.c (c_common_post_options): Enabled by -Wextra.
4646 * c-decl.c (declspecs_add_scspec): Replace -Wextra with
4647 -Wold-style-declaration.
4649 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
4651 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
4652 * cfgbuild.c (find_basic_blocks): Likewise.
4653 * cfgrtl.c (rtl_create_basic_block): Likewise.
4654 * function.c (temp_slots_at_level): Likewise.
4655 * reg-stack.c (stack_regs_mentioned): Likewise.
4656 * regclass.c (allocate_reg_info): Likewise.
4657 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
4658 set_bb_for_stmt, move_block_to_fn): Likewise.
4659 * tree-complex.c (tree_lower_complex): Likewise.
4660 * vec.h (VEC_safe_grow_cleared): New.
4662 2007-01-02 Douglas Gregor <doug.gregor@gmail.com>
4664 * c-common.c (c_common_nodes_and_builtins): Since variants of
4665 void_type_node get built before it is given a name, we need to
4666 give those variants the name, too.
4667 (complete_array_type): We need to work with the canonical main
4668 type of the array, from which we will build the qualified version.
4669 * params.def (PARAM_VERIFY_CANONICAL_TYPES): New.
4670 * print-tree.c (print_node): Display canonical type information
4672 * stor-layout.c (layout_type): When we don't know the
4673 alignment of a type for which we're building an array, we end up
4674 guessing wrong, so make the type require structural equality.
4675 * tree.c (make_node_stat): When we build a new type, it is its
4677 (build_type_attribute_qual_variant): When building an attribute
4678 variant, its canonical type is the non-attribute variant. However,
4679 if the attributes are target-dependent and they differ, we need to
4680 use structural equality checks for this type.
4681 (build_qualified_type): A qualified type is not equivalent to its
4682 unqualified variant; set the canonical type appropriately.
4683 (build_distinct_type_copy): When building a distinct type from
4684 another type, the new type is its own canonical type.
4685 (build_variant_type_copy): When building a new type variant, we
4686 assume that it is equivalent to the original type.
4687 (build_pointer_type_for_mode): When building a pointer type, also
4688 build a canonical type pointer.
4689 (build_reference_type_for_mode): When building a reference type,
4690 also build a canonical type reference.
4691 (build_index_type): When we can't hash an index type (e.g.,
4692 because its maximum value is negative), the index type requires
4693 structural equality tests.
4694 (build_array_type): Build the canonical form of an array type.
4695 (build_function_type): Function types require structural equality,
4696 because they contain default arguments, attributes, etc.
4697 (build_method_type_directly): Ditto for method types.
4698 (build_offset_type): Build the canonical offset type.
4699 (build_complex_type): Build the canonical vector type.
4700 (make_vector_type): Build the canonical vector type.
4701 * tree.h (TYPE_CANONICAL): New.
4702 (TYPE_STRUCTURAL_EQUALITY_P): New.
4703 (SET_TYPE_STRUCTURAL_EQUALITY): New.
4704 (struct tree_type): Added "canonical" field.
4705 * params.h (VERIFY_CANONICAL_TYPES): New.
4706 * doc/c-tree.texi (TYPE_CANONICAL): Document.
4707 (TYPE_STRUCTURAL_EQUALITY_P): Document.
4708 (SET_TYPE_STRUCTURAL_EQUALITY): Document.
4709 * doc/invoke.texi (verify-canonical-types): Document --param
4710 parameter for verifying canonical types.
4712 2007-01-02 Joseph Myers <joseph@codesourcery.com>
4714 * config.gcc (powerpc-*-eabispe*, powerpc-*-eabisimaltivec*,
4715 powerpc-*-eabisim*, powerpc-*-eabialtivec*, powerpc-*-eabi*,
4716 powerpc-*-rtems*, powerpc-wrs-vxworks, powerpc-wrs-vxworksae,
4717 powerpcle-*-eabisim*, powerpcle-*-eabi*): Add rs6000/e500.h to
4719 * config/rs6000/e500.h: New.
4720 * config/rs6000/eabi.h (TARGET_SPE_ABI, TARGET_SPE, TARGET_E500,
4721 TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE, TARGET_E500_DOUBLE):
4723 * config/rs6000/linuxspe.h (TARGET_SPE_ABI, TARGET_SPE,
4724 TARGET_E500, TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE,
4725 TARGET_E500_DOUBLE): Remove.
4726 * config/rs6000/vxworks.h (TARGET_SPE_ABI, TARGET_SPE,
4727 TARGET_E500, TARGET_ISEL, TARGET_FPRS): Remove.
4728 * config/rs6000/rs6000.h (CHECK_E500_OPTIONS): Define.
4729 * config/rs6000/rs6000.c (rs6000_override_options): Use
4732 2007-01-02 Joseph Myers <joseph@codesourcery.com>
4734 * config/rs6000/rs6000.c (print_operand): Check (TARGET_SPE ||
4735 TARGET_E500_DOUBLE), not TARGET_E500, for %y.
4736 (rs6000_generate_compare, rs6000_emit_sCOND, output_cbranch,
4737 rs6000_emit_cmove): Don't check TARGET_E500.
4738 * config/rs6000/rs6000.md (bunordered, bordered, sunordered,
4739 sordered): Don't check TARGET_E500.
4741 2007-01-01 Eric Christopher <echristo@apple.com>
4743 * config/mips/mips.c (mips_regno_mode_ok_for_base_p): Use
4744 HARD_REGISTER_NUM_P.
4746 2007-01-01 Roger Sayle <roger@eyesopen.com>
4748 * fold-const.c (fold_binary) <EQ_EXPR>: Fold "(X^C1) eq/ne C2" into
4749 "X eq/ne (C1^C2)". Fold "(X^Z) eq/ne (Y^Z)" as "X eq/ne Y" when Z
4750 has no side-effects. Fold "(X^C1) eq/ne (Y^C2)" as "(X^(C1^C2))
4753 2007-01-01 Mike Stump <mrs@apple.com>
4755 * configure.ac: Remove support for building with Apple's gcc-3.1.
4757 2007-01-02 Joseph Myers <joseph@codesourcery.com>
4760 * caller-save.c (add_stored_regs): Only handle SUBREGs if inner
4761 REG is a hard register. Do not modify REG before calling
4763 * rtlanal.c (subreg_get_info): Don't assert size of XMODE is a
4764 multiple of the size of YMODE for certain lowpart cases.
4766 2007-01-01 Andrew Pinski <pinskia@gmail.com>
4769 * gimplify (voidify_wrapper_expr): Update for
4772 2007-01-01 Andreas Schwab <schwab@suse.de>
4775 * config/ia64/ia64.c (ia64_compute_frame_size): Account space for
4776 save of BR0 in extra_spill_size instead of spill_size.
4777 (ia64_expand_prologue): Save BR0 outside of the gr/br/fr spill
4779 (ia64_expand_epilogue): Restore BR0 from its new location.
4781 2007-01-01 Andrew Pinski <pinskia@gmail.com>
4783 * gimplify.c (gimplify_init_constructor <case VECTOR_TYPE>):
4784 Use a temporary variable if the left hand side is not a gimple
4787 2007-01-01 Andrew Pinski <pinskia@gmail.com>
4789 * gimplify.c (gimplify_return_expr): Make the temporary variable
4790 for the return expression, a gimple register variable.
4792 2007-01-01 Jan Hubicka <jh@suse.cz>
4794 * emit-rtl.c (emit_copy_of_insn_after): Do not call copy_insn_1 for
4797 2007-01-01 Mike Stump <mrs@apple.com>
4799 * configure.ac (HAVE_GAS_LITERAL16): Add autoconf check for
4801 * config/darwin.c (machopic_select_rtx_section): Use
4803 (darwin_mergeable_constant_section): Likewise.
4804 * configure: Regenerate.
4805 * config.in: Regenerate.
4807 2007-01-01 Jan Hubicka <jh@suse.cz>
4808 Andrew Pinski <pinskia@gmail.com>
4810 * cgraphunit.c (cgraph_optimize): Call cgraph_add_new_functions
4811 before starting IPA passes.