1 2007-02-13 Ian Lance Taylor <iant@google.com>
4 * lower-subreg.c (resolve_simple_move): Decompose subregs in
7 2007-02-13 Stuart Hastings <stuart@apple.com>
9 * gcc/config/i386/i386.md (fixuns_truncdfhi2): Require SSE2.
11 2007-02-13 Richard Henderson <rth@redhat.com>
13 * config/alpha/alpha.c (alpha_stdarg_optimize_hook): Strip
14 handled_component_p before looking for the indirect_ref.
16 2007-02-13 Richard Henderson <rth@redhat.com>
18 * config/i386/i386.md (bswapsi_1): Rename from bswapsi2,
20 (bswapsi2): New expander, emit code for !TARGET_BSWAP.
21 (bswaphi_lowpart): New.
22 (bswapdi2): Rename from bswapdi2_rex, remove flags clobber,
23 remove TARGET_BSWAP test. Delete expander of the same name.
25 * optabs.c (widen_bswap, expand_doubleword_bswap): New.
26 (expand_unop): Use them.
28 2007-02-13 Uros Bizjak <ubizjak@gmail.com>
30 * config/i386/i386.md (cmpdi_ccno_1_rex64, *cmpsi_ccno_1,
31 *cmphi_ccno_1, *cmpqi_ccno_1, *movsi_xor, *movstricthi_xor,
32 *movstrictqi_xor, *movdi_xor_rex64, *ashldi3_1_rex64,
33 *ashldi3_cmp_rex64, *ashldi3_cconly_rex64, ashlsi3, *ashlsi3_1_zext,
34 *ashlsi3_cmp, *ashlsi3_cconly, *ashlsi3_cmp_zext, *ashlhi3_1_lea,
35 *ashlhi3_1, *ashlhi3_cmp, *ashlhi3_cconly, *ashlqi3_1_lea,
36 *ashlqi3_1, *ashlqi3_cmp, *ashlqi3_cconly): Remove equivalent
37 assembler dialect choice from asm templates.
39 2007-02-12 Richard Henderson <rth@redhat.com>
41 * config/i386/i386.md (fixuns_trunc<SSEMODEF>si_1): New insn.
42 (fixuns_trunc<SSEMODEF>si2): Use it.
43 * config/i386/sse.md (vec_setv4sf_0): Export.
44 * config/i386/i386.c (ix86_build_const_vector): Export.
45 (ix86_split_convert_uns_si_sse): Rename from
46 ix86_expand_convert_uns_si_sse and rewrite as a splitter.
47 * config/i386/i386-protos.h: Update.
49 2007-02-13 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
52 * c-typeck.c (c_finish_return): Improve warning message.
54 2007-02-12 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
56 * alias.c (find_symbolic_term): Delete unused function.
58 2007-02-12 Uros Bizjak <ubizjak@gmail.com>
60 * config/i386/i386.md (paritydi2, paritysi2): New expanders.
61 (paritydi2_cmp, paritydi2_cmp): New insn and split patterns.
62 (*parityhi2_cmp, *parityqi2_cmp): New insn patterns.
64 2007-02-12 Eric Botcazou <ebotcazou@adacore.com>
66 * tree.h (DECL_IGNORED_P): Document further effect for FUNCTION_DECL.
67 * cgraphunit.c (cgraph_expand_function): If DECL_IGNORED_P is set on
68 the function, temporarily point the debug interface to the null one.
70 2007-02-12 Eric Botcazou <ebotcazou@adacore.com>
72 * dwarf2out.c (round_up_to_align): New static function.
73 (field_byte_offset): Use it to round the offset.
75 2007-02-12 Richard Henderson <rth@redhat.com>
77 * config/alpha/alpha.md (bswapsi2, bswapdi2): New.
78 (inswl_const): Export.
80 2007-02-12 Richard Henderson <rth@redhat.com>
82 * calls.c (emit_library_call_value_1): If PROMOTE_MODE modifed the
83 result mode of the libcall, convert back to outmode.
85 2007-02-12 Roger Sayle <roger@eyesopen.com>
87 * config/i386/i386.md (*bswapdi2_rex): Renamed from bswapdi2.
88 (bswapdi2): New define_expand to implement 32-bit implementation.
90 2007-02-12 Nick Clifton <nickc@redhat.com>
92 * doc/invoke.texi (Overall Options): Document --help=.
93 * gcc.c (target_help_flag): Rename to print_subprocess_flag.
94 (cc1_options): Pass --help= on to cc1.
95 (display_help): Add description of --help=.
96 (process_command): Add code to handle --help=. Allow translated
97 --help and --target-help switches to be passed on to compiler
99 (main): Remove unused if statement.
100 * opts.c (columns): Remove.
101 (LEFT_COLUMN): Define.
102 (wrap_help): Add columns argument.
103 (print_filtered_help): Change parameters to be an include bitmask,
104 an exclude bitmask, an any bitmask and the column width. Move the
105 code to display the params list here. Add code to display the
106 status of options rather than their descriptions if the quiet flag
108 (print_specific_help): Change parameters to be an include bitmask,
109 an exclude bitmask and an any bitmask. Move code to look up the
110 column width here. Decide upon the title for an options listing.
111 (common_handle_options): Add code to handle --help=. Adapt code
112 for --help and --target-help to use the revised form of the
113 print_specific_help function.
114 (print_help): Delete.
115 (print_param_help): Delete.
116 (print_switch): Delete.
117 * opts.h (cl_lang_count): Add prototype.
118 (CL_PARAMS, CL_WARNING, CL_OPTIMIZATION, CL_MIN_OPTION_CLASS,
119 CL_MAX_OPTION_CLASS): New defines.
120 * optc-gen.awk: Add construction of cl_lang_count.
121 * c.opt: Add Warning attribute to warning options and Optimization
122 attribute to optimization options.
123 * common.opt: Likewise.
125 Add -fhelp and -ftarget-help as aliases for the transformed --help
126 and --target-help options.
127 * opt-functions.awk: Add code to handle Warning and Optimization
130 2007-02-12 Richard Henderson <rth@redhat.com>
132 * config/alpha/constraints.md: New file.
133 * config/alpha/alpha.c: Include tm-constrs.h.
134 (alpha_const_ok_for_letter_p, alpha_const_double_ok_for_letter_p,
135 alpha_extra_constraint): Remove.
136 (alpha_emit_conditional_branch): Use satisfies_constraint_*.
137 * config/alpha/alpha-protos.h: Update.
138 * config/alpha/alpha.h (REG_CLASS_FROM_LETTER): Remove.
139 (CONST_OK_FOR_LETTER_P, CONST_DOUBLE_OK_FOR_LETTER_P): Remove.
140 (EXTRA_CONSTRAINT): Remove.
141 * config/alpha/alpha.md: Include constraints.md.
142 (adddi splitter): Use satisfies_constraint_*.
143 * config/alpha/predicates.md (add_operand): Likewise.
144 (sext_add_operand, addition_operation): Likewise.
146 2007-02-12 Dorit Nuzman <dorit@il.ibm.com>
148 PR tree-optimization/29145
149 * tree-data-ref.c (base_addr_differ_p): Make us more conservative
150 in our handling of restrict qualified pointers.
152 2007-02-12 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
155 * doc/invoke.texi (Wunused-value): Update description.
156 (Wextra): Delete item.
157 * opts.c (set_Wextra): Don't use the value of Wextra to set the
158 value of Wunused-value.
159 * c-typeck.c (c_process_expr_stmt): Don't check extra_warnings.
160 (c_finish_stmt_expr): Don't check extra_warnings.
161 (emit_side_effect_warnings): The caller is responsible to check
164 2007-02-11 Roger Sayle <roger@eyesopen.com>
165 Matt Thomas <matt@3am-software.com>
167 * simplify-rtx.c (simplify_relational_operation_1): Correct typo.
169 2007-02-11 Roger Sayle <roger@eyesopen.com>
171 * simplify-rtx.c (simplify_relational_operation_1): Optimize
172 comparisons of POPCOUNT against zero.
173 (simplify_const_relational_operation): Likewise.
175 2007-02-11 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
177 * doc/invoke.texi (Wextra): Delete outdated paragraph.
179 2007-02-11 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
181 * dwarf2out.c (root_type): Delete unused function.
183 2007-02-11 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
185 * genattrtab.c (contained_in_p): Delete unused function.
186 (write_expr_attr_cache): Likewise.
188 2007-02-11 Jan Hubicka <jh@suse.cz>
190 * ipa-inline.c (cgraph_edge_badness): Add "else" missing in the
193 2007-02-11 Steven Bosscher <steven@gcc.gnu.org>
195 * fwprop.c (try_fwprop_subst): Use set_unique_reg_note
196 to add the REG_EQ* note.
197 * see.c (see_merge_one_use_extension): Likewise.
198 * local-alloc.c (update_equiv_regs): Likewise. Also don't
199 turn REG_EQUAL notes into REG_EQUIV notes if the target
200 register may have more than one set.
201 * function.c (assign_parm_setup_reg): Use set_unique_reg_note.
202 * gcse.c (try_replace_reg): Likewise.
203 * alias.c (init_alias_analysis): Use find_reg_equal_equiv_note.
204 * calls.c (fixup_tail_calls): Likewise. Abort if there is
205 more than one REG_EQUIV note.
206 * reload1.c (gen_reload): Use set_unique_reg_note.
208 2007-02-11 Uros Bizjak <ubizjak@gmail.com>
210 * config/i386/i386.c (TARGET_VECTORIZE_BUILTIN_CONVERSION): Define.
211 (ix86_builtin_conversion): New function.
213 2007-02-06 Mark Mitchell <mark@codesourcery.com>
216 * tree.h (DECL_REPLACEABLE_P): New macro.
217 * except.c (set_nothrow_function_flags): Likewise.
219 2007-02-11 Tehila Meyzels <tehila@il.ibm.com>
220 Ira Rosen <irar@il.ibm.com>
221 Dorit Nuzman <dorit@il.ibm.com>
223 * doc/tm.texi (TARGET_VECTORIZE_BUILTIN_CONVERSION): New target hook.
224 * targhooks.c (default_builtin_vectorized_conversion): New.
225 * targhooks.h (default_builtin_vectorized_function): New declaration.
226 * target.h (struct vectorize): Add builtin_conversion field.
227 * tree-vectorizer.h (type_conversion_vec_info_type): New enum
228 stmt_vec_info_type value.
229 (vectorizable_conversion): New declaration.
230 * tree-vect-analyze.c (vect_analyze_operations): Add
231 vectorizable_conversion call.
232 * target-def.h (TARGET_VECTORIZE_BUILTIN_CONVERSION): New.
233 * tree-vect-transform.c (vectorizable_conversion): New function.
234 (vect_transform_stmt): Add case for type_conversion_vec_info_type.
235 * tree-vect-generic.c (expand_vector_operations_1): Consider correct
237 * config/rs6000/rs6000.c (rs6000_builtin_conversion): New.
238 (TARGET_VECTORIZE_BUILTIN_CONVERSION): Defined.
239 (rs6000_expand_builtin): Add handling a case of ALTIVEC_BUILTIN_VCFUX
240 or ALTIVEC_BUILTIN_VCFSX.
242 2007-02-10 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
245 * pa.md (movdf): For 64-bit target, fail if operand 1 is a non-zero
246 CONST_DOUBLE and operand 0 is a hard register.
247 (movdi): For 64-bit target, remove code to force CONST_DOUBLE to
248 memory. Fail if operand 1 is a non-zero CONST_INT and operand 0
249 is a hard floating-point register.
251 2007-02-10 Richard Henderson <rth@redhat.com>
252 Jakub Jelinek <jakub@redhat.com>
253 Alexandre Oliva <aoliva@redhat.com>
255 * Makefile.in (libgcc-support, libgcc.mvars): Add emutls.c.
256 * builtin-types.def (BT_WORD): Make unsigned.
257 (BT_FN_VOID_PTR_WORD_WORD_PTR): New.
258 * builtins.def (BUILT_IN_EMUTLS_GET_ADDRESS): New.
259 (BUILT_IN_EMUTLS_REGISTER_COMMON): New.
260 * c-decl.c (grokdeclarator): Don't error if !have_tls.
261 * c-parser.c (c_parser_omp_threadprivate): Likewise.
262 * dwarf2out.c (loc_descriptor_from_tree_1): Don't do anything for
264 * expr.c (emutls_var_address): New.
265 (expand_expr_real_1): Expand emulated tls.
266 (expand_expr_addr_expr_1): Likewise.
267 * libgcc-std.ver: Add __emutls_get_address, __emutls_register_common.
268 * output.h (emutls_finish): Declare.
269 * toplev.c (compile_file): Call it.
270 * tree-ssa-address.c (gen_addr_rtx): Check for const-ness of the
271 address before wrapping in CONST.
272 * varasm.c (emutls_htab, emutls_object_type): New.
273 (EMUTLS_VAR_PREFIX, EMUTLS_TMPL_PREFIX): New.
274 (get_emutls_object_name, get_emutls_object_type): New.
275 (get_emutls_init_templ_addr, emutls_decl): New.
276 (emutls_common_1, emutls_finish): New.
277 (assemble_variable): When emulating tls, swap decls; generate
278 constructor for the emutls objects.
279 (do_assemble_alias): When emulating tls, swap decl and target name.
280 (default_encode_section_info): Don't add SYMBOL_FLAG_TLS_SHIFT
282 * varpool.c (decide_is_variable_needed): Look at force_output.
283 Recurse for emulated tls.
284 (cgraph_varpool_remove_unreferenced_decls): Remove checks redundant
285 with decide_is_variable_needed.
286 * emutls.c: New file.
287 * config/sparc/sol2.h (ASM_DECLARE_OBJECT_NAME): Only emit
288 tls_object for real tls.
290 2007-02-10 Kaz Kojima <kkojima@gcc.gnu.org>
292 PR rtl-optimization/29599
293 * reload1.c (eliminate_regs_in_insn): Take the destination
294 mode into account when computing the offset.
296 2007-02-09 Stuart Hastings <stuart@apple.com>
297 Richard Henderson <rth@redhat.com>
299 * gcc/config/i386/i386.h (TARGET_KEEPS_VECTOR_ALIGNED_STACK): New.
300 * gcc/config/i386/darwin.h: (TARGET_KEEPS_VECTOR_ALIGNED_STACK): New.
301 * gcc/config/i386/i386.md (fixuns_trunc<mode>si2, fixuns_truncsfhi2,
302 fixuns_truncdfhi2): New.
303 (fix_truncsfdi_sse): Call ix86_expand_convert_sign_didf_sse.
304 (floatunsdidf2): Call ix86_expand_convert_uns_didf_sse.
305 (floatunssisf2): Add call to ix86_expand_convert_uns_sisf_sse.
306 (floatunssidf2): Allow nonimmediate source.
307 * gcc/config/i386/sse.md (movdi_to_sse): New.
308 (vec_concatv2di): Drop '*'.
309 * gcc/config/i386/i386-protos.h (ix86_expand_convert_uns_si_sse,
310 ix86_expand_convert_uns_didf_sse, ix86_expand_convert_uns_sidf_sse,
311 ix86_expand_convert_uns_sisf_sse, ix86_expand_convert_sign_didf_sse):
313 * gcc/config/i386/i386.c (ix86_expand_convert_uns_si_sse,
314 ix86_expand_convert_uns_didf_sse, ix86_expand_convert_uns_sidf_sse,
315 ix86_expand_convert_uns_sisf_sse, ix86_expand_convert_sign_didf_sse,
316 ix86_build_const_vector, ix86_expand_vector_init_one_nonzero): New.
317 (ix86_build_signbit_mask): Fix decl of v, refactor to call
318 ix86_build_const_vector.
319 (x86_emit_floatuns): Rewrite.
321 2007-02-10 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
323 * genautomata.c (longest_path_length): Delete unused function.
324 (struct state): Delete unused longest_path_length.
325 (UNDEFINED_LONGEST_PATH_LENGTH): Delete unused macro.
326 (get_free_state): Delete unused.
328 2007-02-09 Jan Hubicka <jh@suse.cz>
330 * params.def (PARAM_INLINE_UNIT_GROWTH): Set to 30.
331 * doc/invoke.texi (inline-unit-growth): Update default value.
333 * Makefile.in (passes.o, ipa-inline.o): Add dependencies.
334 * cgraphbuild.c (build_cgraph_edges): Compute frequencies.
335 (rebuild_cgraph_edges): Likewise.
336 * cgraph.c (cgraph_set_call_stmt): Add new argument frequency.
337 (dump_cgraph_node): Dump frequencies.
338 (cgraph_clone_edge): Add frequency scales.
339 (cgraph_clone_node): Add freuqnecy.
340 * cgraph.h (cgraph_edge): Add freuqnecy argument.
341 (CGRAPH_FREQ_BASE, CGRAPH_FREQ_MAX): New constants.
342 (cgraph_create_edge, cgraph_clone_edge, cgraph_clone_node): Update.
343 * tree-pass.h (TODO_rebuild_frequencies): New constant.
344 * cgraphunit.c (verify_cgraph_node): Verify frequencies.
345 (cgraph_copy_node_for_versioning): Update call of cgraph_clone_edge.
346 (save_inline_function_body): Likewise.
347 * ipa-inline.c: inluce rtl.h
348 (cgraph_clone_inlined_nods): Update call of cgraph_clone_node.
349 (cgraph_edge_badness): Use frequencies.
350 (cgraph_decide_recursive_inlining): Update clonning.
351 (cgraph_decide_inlining_of_small_function): Dump frequency.
352 * predict.c (estimate_bb_frequencies): Export.
353 * predict.h (estimate_bb_frequencies): Declare.
354 * tree-inline.c (copy_bb): Watch overflows.
355 (expand_call_inline): Update call of cgraph_create_edge.
356 (optimize_inline_calls): Use TODO flags to update frequnecies.
357 * passes.h: Include predict.h
358 (init_optimization_passes): Move profile ahead.
359 (execute_function_todo): Handle TODO_rebuild_frequencies.
361 2007-02-09 Roger Sayle <roger@eyesopen.com>
363 * config/alpha/alpha.c (emit_insxl): Force the first operand of
364 the insbl or inswl pattern into a register.
366 2007-02-09 Roger Sayle <roger@eyesopen.com>
368 * config/ia64/ia64.md (bswapdi2): New define_insn.
370 2007-02-09 Richard Henderson <rth@redhat.com>
372 * config/i386/constraints.md (Ym): New constraint.
373 * config/i386/i386.md (movsi_1): Change Y2 to Yi constraints.
374 (movdi_1_rex64): Split sse and xmm general register moves from
375 memory move alternatives. Use conditional register constraints.
376 (movsf_1, movdf_integer): Likewise.
377 (zero_extendsidi2_32, zero_extendsidi2_rex64): Likewise.
378 (movdf_integer_rex64): New.
379 (pushsf_rex64): Fix output constraints.
380 * config/i386/sse.md (sse2_loadld): Split rm alternative, use Yi.
381 (sse2_stored): Likewise.
382 (sse2_storeq_rex64): New.
383 * config/i386/i386.c (x86_inter_unit_moves): Enable for not
385 (ix86_secondary_memory_needed): Don't bypass TARGET_INTER_UNIT_MOVES
386 for optimize_size. Remove SF/DFmode hack.
388 2007-02-09 Dwarakanath Rajagopal <dwarak.rajagopal@amd.com>
390 * config/i386/driver-i386.c: Turn on -mtune=native for AMDFAM10.
393 2007-02-09 Nathan Sidwell <nathan@codesourcery.com>
394 Richard Sandiford <richard@codesourcery.com>
396 * config.gcc (m68010-*-netbsdelf*, m68k*-*-netbsdelf*)
397 (m68k*-*-openbsd*, m68k-*-linux*): Set default_cf_cpu.
398 (m68k-*-aout*, m68k-*-coff*, m68k-*-uclinux*, m68k-*-rtems*): Add
399 m68k/t-mlib to tmake_file.
400 (m68020-*-elf*, m68k-*-elf*): Likewise. Add t-m68kbare as well.
401 (m68k*-*-*): Use --with-arch to pick a default for --with-cpu.
402 (m680[012]0-*-*, m68k*-*-*): Add support for --with-arch.
403 Allow it to be cf or m68k. Set m68k_arch_family. If that
404 variable is not empty, add t-$m68k_arch_family to tmake_file.
405 Add t-mlibs to tmake_file.
406 * doc/install.texi: Document --with-arch=m68k and --with-arch=cf.
407 * config/m68k/t-cf: New file.
408 * config/m68k/t-m68k: Likewise.
409 * config/m68k/t-mlibs: Likewise.
410 * config/m68k/t-m68kbare (MULTILIB_OPTIONS, MULTILIB_DIRNAMES)
411 (MULTILIB_MATCHES, MULTILIB_EXCEPTIONS): Delete.
412 (M68K_MLIB_DIRNAMES, M68K_MLIB_OPTIONS): Define.
413 * config/m68k/t-m68kelf (MULTILIB_OPTIONS, MULTILIB_DIRNAMES)
414 (MULTILIB_MATCHES, MULTILIB_EXCEPTIONS, LIBGCC, INSTALL_LIBGCC):
416 * config/m68k/t-openbsd (MULTILIB_OPTIONS, LIBGCC): Delete.
417 (INSTALL_LIBGCC): Delete.
418 (M68K_MLIB_DIRNAMES, M68K_MLIB_OPTIONS): Define.
419 * config/m68k/t-rtems (MULTILIB_OPTIONS, MULTILIB_DIRNAMES)
420 (MULTILIB_MATCHES, MULTILIB_EXCEPTIONS): Delete.
421 (M68K_MLIB_CPU): Define.
422 * config/m68k/t-uclinux (MULTILIB_OPTIONS, MULTILIB_DIRNAMES)
423 (MULTILIB_MATCHES, MULTILIB_EXCEPTIONS): Delete.
424 (M68K_MLIB_CPU, M68K_MLIB_OPTIONS, M68K_MLIB_DIRNAMES): Define.
426 2007-02-09 Zdenek Dvorak <dvorakz@suse.cz>
427 Richard Guenther <rguenther@suse.de>
430 * fold-const.c (fold_comparison): Handle obfuscated comparisons
431 against INT_MIN/INT_MAX.
432 * tree-ssa-loop-ivcanon.c (remove_empty_loop): Print to dump
433 file if a loop is removed.
435 2007-02-09 Joseph Myers <joseph@codesourcery.com>
437 * calls.c (store_one_arg): Pass correct alignment to
438 emit_push_insn for non-BLKmode values.
439 * expr.c (emit_push_insn): If STRICT_ALIGNMENT, copy to an
440 unaligned stack slot via a suitably aligned slot.
442 2007-02-08 DJ Delorie <dj@redhat.com>
444 * config/m32c/m32c.c (m32c_unpend_compare): Add default to silence
446 (legal_subregs): Use unsigned char, make const.
447 (m32c_illegal_subreg_p): Use ARRAY_SIZE. Delete unused variables.
449 2007-02-08 Paul Brook <paul@codesourcery.com>
451 * config/arm/lib1funcs.asm (RETLDM): Pop directly into PC when no
452 special interworking needed.
454 2007-02-08 Harsha Jagasia <harsha.jagasia@amd.com>
456 * config/i386/xmmintrin.h: Make inclusion of emmintrin.h
457 conditional to __SSE2__.
458 (Entries below should have been added to first ChangeLog
459 entry for amdfam10 dated 2007-02-05)
460 * config/i386/emmintrin.h: Generate #error if __SSE2__ is not
462 * config/i386/pmmintrin.h: Generate #error if __SSE3__ is not
464 * config/i386/tmmintrin.h: Generate #error if __SSSE3__ is not
467 2007-02-08 DJ Delorie <dj@redhat.com>
469 * config/m32c/m32c-protos.h (m32c_illegal_subreg_p): New.
470 * config/m32c/m32c.c (legal_subregs): New.
471 (m32c_illegal_subreg_p): New.
472 * config/m32c/predicates.md (m32c_any_operand): Use it to reject
473 unsupported subregs of hard regs.
475 2007-02-08 Jan Hubicka <jh@suse.cz>
477 * tree-cfg.c (bsi_replace): Shortcut when replacing the statement with
478 the same one; always update histograms.
480 2007-02-08 Diego Novillo <dnovillo@redhat.com>
482 * passes.c (init_optimization_passes): Tidy comment.
484 2007-02-08 Roger Sayle <roger@eyesopen.com>
486 * simplify-rtx.c (simplify_unary_operation_1) <POPCOUNT>: We can
487 strip zero_extend, bswap and rotates from POCOUNT's argument.
488 <PARITY>: Likewise, we can strip not, bswap, sign_extend,
489 zero_extend and rotates from PARITY's argument.
490 <BSWAP>: A byte-swap followed by a byte-swap is an identity.
491 (simplify_const_unary_operation) <BSWAP>: Evaluate the byte-swap
492 of an integer constant at compile-time.
494 2007-02-08 Diego Novillo <dnovillo@redhat.com>
497 * tree-flow.h (struct var_ann_d): Remove field 'is_used'.
499 * tree-ssa-alias.c (compute_is_aliased): Remove. Update all
502 * tree-ssa-live.c (remove_unused_locals): Do not remove
503 TREE_ADDRESSABLE variables.
504 * tree-ssa-structalias.c (compute_points_to_sets): Tidy.
505 * tree-ssa-operands.c (add_virtual_operand): Remove argument
506 FOR_CLOBBER. Update all users.
507 If VAR has an associated alias set, add a virtual operand for
508 it if no alias is found to conflict with the memory reference.
510 2007-02-07 Jan Hubicka <jh@suse.cz>
511 Robert Kidd <rkidd@crhc.uiuc.edu>
513 * value-prof.c (visit_hist, free_hist): Return 1 instead of 0.
515 2007-02-07 Ian Lance Taylor <iant@google.com>
517 * lower-subreg.c (simple_move): Reject PARTIAL_INT modes.
519 2007-02-07 Roger Sayle <roger@eyesopen.com>
521 * config/rs6000/rs6000.md (ctz<mode>2, ffs<mode>2, popcount<mode>2,
522 parity<mode>2, smulsi3_highpart, abstf2_internal, allocate_stack,
523 tablejumpdi, movsi_to_cr_one): Remove constraints from
524 define_expand's match_operands.
526 2007-02-07 Roger Sayle <roger@eyesopen.com>
528 * global.c (compute_regsets): Move declatation of "i" inside of
529 #ifdef ELIMINABLE_REGS to avoid unused variable bootstrap failure.
531 2007-02-07 Jakub Jelinek <jakub@redhat.com>
534 * gimplify.c (gimplify_scan_omp_clauses): Remove special casing
535 of INDIRECT_REF <RESULT_DECL>.
537 * config/i386/i386.c (override_options): Set PTA_SSSE3 for core2.
539 2007-02-06 J"orn Rennecke <joern.rennecke@arc.com>
540 Kaz Kojima <kkojima@gcc.gnu.org>
543 * config/sh/sh.c (expand_cbranchdi4): Use scratch register
545 (sh_initialize_trampoline): Add parentheses to avoid a warning.
547 2007-02-06 Zdenek Dvorak <dvorakz@suse.cz>
549 * doc/loop.texi: Document possibility not to perform disambiguation
550 of loops with multiple latches.
551 * cfgloopmanip.c (alp_enum_p): Removed.
552 (add_loop): Handle subloops. Use get_loop_body_with_size.
553 (create_preheader): Do not allow ENTRY_BLOCK_PTR to be preheader.
554 * cfghooks.c (redirect_edge_and_branch_force): Set dominator for
555 the new forwarder block.
556 (make_forwarder_block): Only call new_bb_cbk if it is not NULL.
557 Handle the case latch is NULL.
558 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Avoid cfg modifications
559 when marking loop exits.
560 * ifcvt.c (if_convert): Ditto. Mark loop exits even if cfg cannot
562 * loop-init.c (loop_optimizer_init): Do not modify cfg. Call
563 disambiguate_loops_with_multiple_latches.
564 * tree-cfgcleanup.c (cleanup_tree_cfg_loop): Calculate dominators
565 before fix_loop_structure.
566 * cfgloop.c: Include pointer-set.h and output.h.
567 (canonicalize_loop_headers, HEADER_BLOCK, LATCH_EDGE,
568 update_latch_info, mfb_keep_just, mfb_keep_nonlatch): Removed.
569 (get_loop_latch_edges, find_subloop_latch_edge_by_profile,
570 find_subloop_latch_edge_by_ivs, find_subloop_latch_edge,
571 mfb_redirect_edges_in_set, form_subloop, merge_latch_edges,
572 disambiguate_multiple_latches, get_loop_body_with_size,
573 disambiguate_loops_with_multiple_latches): New functions.
574 (flow_loop_dump): Dump multiple latch edges.
575 (flow_loop_nodes_find): Handle loops with multiple latches.
576 (flow_loops_find): Ditto. Do not call canonicalize_loop_headers.
577 (glb_enum_p): Modified.
578 (get_loop_body): Use get_loop_body_with_size.
579 * cfgloop.h (LOOPS_HAVE_RECORDED_EXITS): New flag.
580 (AVOID_CFG_MODIFICATIONS): New constant.
581 (disambiguate_loops_with_multiple_latches, add_loop,
582 get_loop_body_with_size): Declare.
583 * Makefile.in (cfgloop.o): Add pointer-set.h and output.h.
585 2007-02-06 Seongbae Park <seongbae.park@gmail.com>
588 * global.c (compute_regsets): New function.
589 (global_alloc): Refactored ELIMINABLE_REGSET
590 and NO_GLOBAL_ALLOC_REGS computation out.
591 (rest_of_handle_global_alloc): Call compute_regsets()
592 for non-optimizing case.
594 2007-02-06 Richard Henderson <rth@redhat.com>
596 * config/i386/constraints.md (Y2): Rename from Y.
597 (Yi): New constraint.
598 * config/i386/i386.md (movsi_1, movdi_2, pushdf_nointeger,
599 pushdf_integer, movdf_nointeger, movdf_integer, zero_extendsidi2_32,
600 zero_extendsidi2_rex64, truncxfdf2_mixed): Change Y constraints to Y2.
601 (extendsfdf2_mixed, extendsfdf2_sse, truncdfsf_fast_mixed,
602 truncdfsf_fast_sse, truncdfsf_mixed, fix_truncdfdi_sse,
603 fix_truncdfsi_sse, floatsidf2_mixed, floatsidf2_sse,
604 floatdidf2_mixed, floatdidf2_sse, absnegdf2_mixed,
605 absnegdf2_sse, sse_setccdf, fop_df_comm_mixed, fop_df_comm_sse,
606 fop_df_1_mixed, fop_df_1_sse): Change Y constraints to x.
607 * config/i386/mmx.md (mov<MMXMODEI>_internal_rex64,
608 mov<MMXMODEI>_internal, movv2sf_internal_rex64, movv2sf_internal,
609 vec_extractv2si_1): Change Y constraints to Y2.
610 * config/i386/sse.md (vec_setv4sf_0, vec_concatv2df, vec_dupv4si,
611 vec_dupv2di, sse2_concatv2si, vec_concatv4si_1, vec_concatv2di):
612 Change Y constraints to Y2.
613 (sse2_loadld): Change Y constraints to x.
615 2007-02-06 Roger Sayle <roger@eyesopen.com>
617 * config/rs6000/rs6000.md (popcount<mode>2): Rewrite.
618 (parity<mode>2): New define_expand using rs6000_emit_parity.
619 * config/rs6000/rs6000.c (rs6000_emit_popcount,
620 rs6000_emit_parity): New functions.
621 * config/rs6000/rs6000-protos.h (rs6000_emit_popcount,
622 rs6000_emit_parity): Prototype here.
624 2007-02-06 Ian Lance Taylor <iant@google.com>
626 * lower-subreg.c (simple_move_operand): Reject CONST.
627 (resolve_clobber): Call validate_change rather than directly
628 assigning to XEXP (pat, 0).
630 2006-02-06 Paolo Bonzini <bonzini@gnu.org>
632 * Makefile.in (tree-ssa-loop-ivopts.o): Add pointer-set.h dependency.
633 (tree-ssa-reassoc.o): Add pointer-set.h dependency.
634 (tree-cfg.o): Remove hashtab.h dependency.
636 * tree-ssa-loop-ivopts.c: Include pointer-set.h.
637 (struct ivopts_data): Change niters to pointer_map_t.
638 (struct nfe_cache_elt, nfe_hash, nfe_eq): Delete.
639 (niter_for_exit): Create pointer_map on demand. Change for
641 (tree_ssa_iv_optimize_init): Initialize data->niters to NULL.
642 (free_loop_data): Destroy data->niters if created and reset field.
643 (tree_ssa_iv_optimize_finalize): Don't delete data->niters here.
644 (tree_ssa_iv_optimize_loop): Check for presence of stale data.
646 * tree-ssa-reassoc.c: Include pointer-set.h.
647 (bb_rank): Change to long *.
648 (operand_rank): Change to pointer_map_t.
649 (find_operand_rank): Return long, -1 if not found. Declare as inline.
650 (insert_operand_rank): Accept long.
651 (operand_entry_hash, operand_entry_eq): Remove.
652 (get_rank): Return long. Adjust for changes above.
653 (init_reassoc): Change rank type to long. Adjust creation of bb_rank
655 (fini_reassoc): Delete operand_rank with pointer_map_destroy.
657 * tree-ssa-structalias.c (vi_for_tree): Change to pointer_map.
658 (struct tree_vi, tree_vi_t, tree_vi_hash, tree_vi_eq): Delete.
659 (insert_vi_for_tree): Rewrite for pointer_map API. Assert argument
661 (lookup_vi_for_tree): Rewrite for pointer_map API. Return varinfo_t
662 directly since it cannot be NULL.
663 (get_vi_for_tree): Rewrite for pointer_map API.
664 (find_what_p_points_to): Adjust for change to lookup_vi_for_tree.
665 (init_alias_vars): Create vi_for_tree as pointer_map.
666 (delete_points_to_sets): Delete vi_for_tree using pointer_map_destroy.
668 * tree-cfg.c: Don't include hashtab.h.
669 (edge_to_cases): Declare as pointer_map.
670 (struct edge_to_cases_elt, edge_to_cases_hash, edge_to_cases_eq):
672 (edge_to_cases_cleanup): Rewrite as pointer_map_traverse callback.
673 (start_recording_case_labels): Create edge_to_cases as pointer_map.
674 (end_recoding_case_labels): Cleanup edge_to_cases manually before
676 (record_switch_edge): Delete.
677 (get_cases_for_edge): Adjust for pointer_map API, inline
678 record_switch_edge (rewritten for new API), remove goto.
680 2006-02-06 Paolo Bonzini <bonzini@gnu.org>
682 * Makefile.in (tree-nested.o): Add pointer-set.h dependency.
683 * tree-nested.c: Include pointer-set.h.
684 (var_map_elt, var_map_eq, var_map_hash): Delete.
685 (struct nesting_info): Remove GTY marker. Change the two htab_t's
687 (nesting_info_bitmap_obstack): New.
688 (lookup_field_for_decl): Adjust for pointer_map API.
689 (lookup_tramp_for_decl): Adjust for pointer_map API.
690 (get_nonlocal_debug_decl): Adjust for pointer_map API.
691 (get_local_debug_decl): Adjust for pointer_map API.
692 (convert_nl_goto_reference): Adjust for pointer_map API.
693 (convert_nl_goto_receiver): Adjust for pointer_map API.
694 (create_nesting_tree): Create outside GGC space. Create bitmap on
695 the new obstack. Create field_map and var_map as pointer_maps.
696 (free_nesting_tree): Adjust for changes to create_nesting_tree.
698 (lower_nested_functions): Move root here, no need to NULL it.
699 Initialize and release the obstack.
701 2007-02-06 Paolo Bonzini <bonzini@gnu.org>
703 * tree.c (tree_int_map_hash, tree_int_map_eq, tree_int_map_marked_p):
704 Remove prototypes and make them non-static.
705 (struct tree_int_map): Remove.
706 * tree.h (struct tree_int_map): Move here, turning TO into an
708 (tree_int_map_hash, tree_int_map_eq, tree_int_map_marked_p): Declare.
710 * tree.h (TREE_COMPLEXITY): Remove.
711 (struct tree_exp): Remove complexity field.
712 * tree.c (build1_stat): Don't set it.
714 2007-02-06 Dorit Nuzman <dorit@il.ibm.com>
715 Victor Kaplansky <victork@il.ibm.com>
717 * tree-vectorizer.c (vect_is_simple_use): Support induction.
718 (vect_is_simple_reduction): Support reduction with induction as
720 (vect_is_simple_iv_evolution): Fix formatting.
721 * tree-vect-analyze.c (vect_mark_stmts_to_be_vectorized): Fix
722 formatting. Don't mark induction phis for vectorization.
723 (vect_analyze_scalar_cycles): Analyze all inductions, then reductions.
724 * tree-vect-transform.c (get_initial_def_for_induction): New function.
725 (vect_get_vec_def_for_operand): Support induction.
726 (vect_get_vec_def_for_stmt_copy): Fix formatting and add check for
728 (vectorizable_reduction): Support reduction with induction as one of
730 (vectorizable_type_demotion): Use def-type of stmt argument rather
733 * tree-ssa-loop.c (gate_scev_const_prop): Return the value of
734 flag_tree_scev_cprop.
735 * common.opt (tree-scev-cprop): New flag.
737 * tree-vect-transform.c (vect_create_destination_var): Use 'kind' in
738 call to vect_get_new_vect_var.
740 2007-02-06 Ira Rosen <irar@il.ibm.com>
742 * tree-vect-patterns.c (vect_recog_widen_mult_pattern): Check that
744 (vect_pattern_recog_1): Likewise.
746 2007-02-05 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
748 * fold-const.c (negate_expr_p): Handle CONJ_EXPR.
749 (fold_negate_expr): Likewise.
751 2007-02-05 Alexandre Oliva <aoliva@redhat.com>
754 * dwarf2out.c (modified_type_die): Follow DECL_ORIGINAL_TYPE
755 even if cv-qualification is the same.
757 2007-02-05 Geoffrey Keating <geoffk@apple.com>
759 * config/rs6000/darwin-tramp.asm (__trampoline_setup): Call
760 __enable_execute_stack on completion.
762 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
764 * config/i386/athlon.md (athlon_fldxf_k8, athlon_fld_k8,
765 athlon_fstxf_k8, athlon_fst_k8, athlon_fist, athlon_fmov,
766 athlon_fadd_load, athlon_fadd_load_k8, athlon_fadd, athlon_fmul,
767 athlon_fmul_load, athlon_fmul_load_k8, athlon_fsgn,
768 athlon_fdiv_load, athlon_fdiv_load_k8, athlon_fdiv_k8,
769 athlon_fpspc_load, athlon_fpspc, athlon_fcmov_load,
770 athlon_fcmov_load_k8, athlon_fcmov_k8, athlon_fcomi_load_k8,
771 athlon_fcomi, athlon_fcom_load_k8, athlon_fcom): Added amdfam10.
773 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
775 * config/i386/i386.md (x86_sahf_1, cmpfp_i_mixed, cmpfp_i_sse,
776 cmpfp_i_i387, cmpfp_iu_mixed, cmpfp_iu_sse, cmpfp_iu_387,
777 swapsi, swaphi_1, swapqi_1, swapdi_rex64, fix_truncsfdi_sse,
778 fix_truncdfdi_sse, fix_truncsfsi_sse, fix_truncdfsi_sse,
779 x86_fldcw_1, floatsisf2_mixed, floatsisf2_sse, floatdisf2_mixed,
780 floatdisf2_sse, floatsidf2_mixed, floatsidf2_sse,
781 floatdidf2_mixed, floatdidf2_sse, muldi3_1_rex64, mulsi3_1,
782 mulsi3_1_zext, mulhi3_1, mulqi3_1, umulqihi3_1, mulqihi3_insn,
783 umulditi3_insn, umulsidi3_insn, mulditi3_insn, mulsidi3_insn,
784 umuldi3_highpart_rex64, umulsi3_highpart_insn,
785 umulsi3_highpart_zext, smuldi3_highpart_rex64,
786 smulsi3_highpart_insn, smulsi3_highpart_zext, x86_64_shld,
787 x86_shld_1, x86_64_shrd, sqrtsf2_mixed, sqrtsf2_sse,
788 sqrtsf2_i387, sqrtdf2_mixed, sqrtdf2_sse, sqrtdf2_i387,
789 sqrtextendsfdf2_i387, sqrtxf2, sqrtextendsfxf2_i387,
790 sqrtextenddfxf2_i387): Added amdfam10_decode.
792 * config/i386/athlon.md (athlon_idirect_amdfam10,
793 athlon_ivector_amdfam10, athlon_idirect_load_amdfam10,
794 athlon_ivector_load_amdfam10, athlon_idirect_both_amdfam10,
795 athlon_ivector_both_amdfam10, athlon_idirect_store_amdfam10,
796 athlon_ivector_store_amdfam10): New define_insn_reservation.
797 (athlon_idirect_loadmov, athlon_idirect_movstore): Added
800 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
802 * config/i386/athlon.md (athlon_call_amdfam10,
803 athlon_pop_amdfam10, athlon_lea_amdfam10): New
804 define_insn_reservation.
805 (athlon_branch, athlon_push, athlon_leave_k8, athlon_imul_k8,
806 athlon_imul_k8_DI, athlon_imul_mem_k8, athlon_imul_mem_k8_DI,
807 athlon_idiv, athlon_idiv_mem, athlon_str): Added amdfam10.
809 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
811 * config/i386/athlon.md (athlon_sseld_amdfam10,
812 athlon_mmxld_amdfam10, athlon_ssest_amdfam10,
813 athlon_mmxssest_short_amdfam10): New define_insn_reservation.
815 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
817 * config/i386/athlon.md (athlon_sseins_amdfam10): New
818 define_insn_reservation.
819 * config/i386/i386.md (sseins): Added sseins to define_attr type
820 and define_attr unit.
821 * config/i386/sse.md: Set type attribute to sseins for insertq
824 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
826 * config/i386/athlon.md (sselog_load_amdfam10, sselog_amdfam10,
827 ssecmpvector_load_amdfam10, ssecmpvector_amdfam10,
828 ssecomi_load_amdfam10, ssecomi_amdfam10,
829 sseaddvector_load_amdfam10, sseaddvector_amdfam10): New
830 define_insn_reservation.
831 (ssecmp_load_k8, ssecmp, sseadd_load_k8, seadd): Added amdfam10.
833 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
835 * config/i386/athlon.md (cvtss2sd_load_amdfam10,
836 cvtss2sd_amdfam10, cvtps2pd_load_amdfam10, cvtps2pd_amdfam10,
837 cvtsi2sd_load_amdfam10, cvtsi2ss_load_amdfam10,
838 cvtsi2sd_amdfam10, cvtsi2ss_amdfam10, cvtsd2ss_load_amdfam10,
839 cvtsd2ss_amdfam10, cvtpd2ps_load_amdfam10, cvtpd2ps_amdfam10,
840 cvtsX2si_load_amdfam10, cvtsX2si_amdfam10): New
841 define_insn_reservation.
843 * config/i386/sse.md (cvtsi2ss, cvtsi2ssq, cvtss2si,
844 cvtss2siq, cvttss2si, cvttss2siq, cvtsi2sd, cvtsi2sdq,
845 cvtsd2si, cvtsd2siq, cvttsd2si, cvttsd2siq,
846 cvtpd2dq, cvttpd2dq, cvtsd2ss, cvtss2sd,
847 cvtpd2ps, cvtps2pd): Added amdfam10_decode attribute.
849 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
851 * config/i386/athlon.md (athlon_ssedivvector_amdfam10,
852 athlon_ssedivvector_load_amdfam10, athlon_ssemulvector_amdfam10,
853 athlon_ssemulvector_load_amdfam10): New define_insn_reservation.
854 (athlon_ssediv, athlon_ssediv_load_k8, athlon_ssemul,
855 athlon_ssemul_load_k8): Added amdfam10.
857 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
859 * config/i386/i386.h (TARGET_SSE_UNALIGNED_MOVE_OPTIMAL): New macro.
860 (x86_sse_unaligned_move_optimal): New variable.
862 * config/i386/i386.c (x86_sse_unaligned_move_optimal): Enable for
864 (ix86_expand_vector_move_misalign): Add code to generate movupd/movups
865 for unaligned vector SSE double/single precision loads for AMDFAM10.
867 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
869 * config/i386/i386.h (TARGET_AMDFAM10): New macro.
870 (TARGET_CPU_CPP_BUILTINS): Add code for amdfam10.
871 Define TARGET_CPU_DEFAULT_amdfam10.
872 (TARGET_CPU_DEFAULT_NAMES): Add amdfam10.
873 (processor_type): Add PROCESSOR_AMDFAM10.
875 * config/i386/i386.md: Add amdfam10 as a new cpu attribute to match
876 processor_type in config/i386/i386.h.
877 Enable imul peepholes for TARGET_AMDFAM10.
879 * config.gcc: Add support for --with-cpu option for amdfam10.
881 * config/i386/i386.c (amdfam10_cost): New variable.
882 (m_AMDFAM10): New macro.
883 (m_ATHLON_K8_AMDFAM10): New macro.
884 (x86_use_leave, x86_push_memory, x86_movx, x86_unroll_strlen,
885 x86_cmove, x86_3dnow_a, x86_deep_branch, x86_use_simode_fiop,
886 x86_promote_QImode, x86_integer_DFmode_moves,
887 x86_partial_reg_dependency, x86_memory_mismatch_stall,
888 x86_accumulate_outgoing_args, x86_arch_always_fancy_math_387,
889 x86_sse_partial_reg_dependency, x86_sse_typeless_stores,
890 x86_use_ffreep, x86_use_incdec, x86_four_jump_limit,
891 x86_schedule, x86_use_bt, x86_cmpxchg16b, x86_pad_returns):
892 Enable/disable for amdfam10.
893 (override_options): Add amdfam10_cost to processor_target_table.
894 Set up PROCESSOR_AMDFAM10 for amdfam10 entry in
895 processor_alias_table.
896 (ix86_issue_rate): Add PROCESSOR_AMDFAM10.
897 (ix86_adjust_cost): Add code for amdfam10.
899 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
901 * config/i386/i386.opt: Add new Advanced Bit Manipulation (-mabm)
902 instruction set feature flag. Add new (-mpopcnt) flag for popcnt
903 instruction. Add new SSE4A (-msse4a) instruction set feature flag.
904 * config/i386/i386.h: Add builtin definition for SSE4A.
905 * config/i386/i386.md: Add support for ABM instructions
907 * config/i386/sse.md: Add support for SSE4A instructions
908 (movntss, movntsd, extrq, insertq).
909 * config/i386/i386.c: Add support for ABM and SSE4A builtins.
910 Add -march=amdfam10 flag.
911 * config/i386/ammintrin.h: Add support for SSE4A intrinsics.
912 * doc/invoke.texi: Add documentation on flags for sse4a, abm, popcnt
914 * doc/extend.texi: Add documentation for SSE4A builtins.
916 2007-02-05 Bob Wilson <bob.wilson@acm.org>
918 * config/xtensa/xtensa.c (constantpool_mem_p): Skip over SUBREGs.
920 2007-02-05 Richard Guenther <rguenther@suse.de>
922 * tree-vectorizer.h (vectorizable_function): Add argument type
923 argument, change return type.
924 * tree-vect-patterns.c (vect_recog_pow_pattern): Adjust caller.
925 * tree-vect-transform.c (vectorizable_function): Handle extra
926 argument, return vectorized function decl.
927 (build_vectorized_function_call): Remove.
928 (vectorizable_call): Handle calls with result and argument types
929 differing. Handle loop vectorization factor correctly.
930 * targhooks.c (default_builtin_vectorized_function): Adjust for
932 * targhooks.h (default_builtin_vectorized_function): Likewise.
933 * target.h (builtin_vectorized_function): Add argument type
935 * config/i386/i386.c (ix86_builtin_vectorized_function): Handle
936 extra argument, allow vectorizing of lrintf.
937 * doc/tm.texi (TARGET_VECTORIZE_BUILTIN_VECTORIZED_FUNCTION): Adjust
938 documentation of target hook.
940 2007-02-05 Hans-Peter Nilsson <hp@axis.com>
943 * config/cris/cris.md ("*andsi_movu", "*andsi_clear", "*andhi_movu")
944 ("*andhi_clear", andu (casesi+45)): For size-changed operand where
945 memory is allowed, require !side_effects_p, not just !MEM_VOLATILE_P.
947 2007-02-05 Roger Sayle <roger@eyesopen.com>
949 * fold-const.c (fold_unary) <REAL_PART>: Test for availability of
950 BUILT_IN_COS before simplifying REAL_PART(CEXPI)) to COS.
951 <IMAG_PART>: Likewise, check for availability of BUILT_IN_SIN.
952 * builtins.c (fold_builtin_sincos): Check for TARGET_C99_FUNCTIONS
953 before canonicalizing sincos to cexpi.
954 (fold_builtin_cexp): Likewise, for canonicalizing cexp to cexpi.
956 2007-02-05 Roger Sayle <roger@eyesopen.com>
958 * config/alpha/alpha.c (alpha_add_builtins): New Helper function.
959 Set TREE_READONLY and TREE_NOTHROW directly, not via attributes.
960 (alpha_init_builtins): Use alpha_add_builtins to process tables.
962 2007-02-05 Roger Sayle <roger@eyesopen.com>
964 * mips-tfile.c (initialize_init_file): Correct endianness test.
966 2007-02-05 Kazu Hirata <kazu@codesourcery.com>
968 * config/m68k/m68k.md (pushdi-1, pushdi, movsi+1): Don't use
971 2007-02-05 Richard Sandiford <richard@codesourcery.com>
973 * dwarf2out.c (dwarf2out_frame_debug_expr): Record the register
974 saves in a PARALLEL before the register assignments.
976 2007-02-05 Richard Sandiford <richard@codesourcery.com>
978 * doc/tm.texi (DWARF_ALT_FRAME_RETURN_COLUMN): Do not require
979 DWARF_FRAME_RETURN_COLUMN to be a general register.
980 * dwarf2out.c (init_return_column_size): New function, split from...
981 (expand_builtin_init_dwarf_reg_sizes): ...here. Allow both
982 DWARF_FRAME_RETURN_COLUMN and DWARF_ALT_FRAME_RETURN_COLUMN
983 to be nongeneral registers.
984 * config/m68k/m68k.h (DWARF_FRAME_REGNUM): Only map FP and
986 (DWARF_FRAME_REGISTERS, DWARF_FRAME_RETURN_COLUMN): Define.
987 (DWARF_ALT_FRAME_RETURN_COLUMN): Define.
989 2007-02-04 Zdenek Dvorak <dvorakz@suse.cz>
991 * cfgcleanup.c (try_optimize_cfg): Avoid removing ENTRY_BLOCK_PTR.
993 2007-02-04 Zdenek Dvorak <dvorakz@suse.cz>
995 * cfgloopmanip.c (loop_delete_branch_edge): Removed.
996 (remove_path): Use can_remove_branch_p and remove_branch instead
997 of loop_delete_branch_edge.
998 * tree-ssa-loop-manip.c (scale_dominated_blocks_in_loop): New function.
999 (tree_transform_and_unroll_loop): Remove dead branches immediately.
1000 Update profile using scale_dominated_blocks_in_loop.
1001 * cfghooks.c (can_remove_branch_p, remove_branch): New functions.
1002 * cfghooks.h (struct cfg_hooks): Add can_remove_branch_p.
1003 (can_remove_branch_p, remove_branch): Declare.
1004 * tree-cfg.c (tree_can_remove_branch_p): New function.
1005 (tree_cfg_hooks): Add tree_can_remove_branch_p.
1006 * cfgrtl.c (rtl_can_remove_branch_p): New function.
1007 (rtl_cfg_hooks, cfg_layout_rtl_cfg_hook): Add rtl_can_remove_branch_p.
1009 2007-02-05 Jan Hubicka <jh@suse.cz>
1012 * ipa-inline.c (cgraph_clone_inlined_nodes): When there are unanalyzed
1013 nodes in cgraph, don't remove offline copy of the function.
1015 2007-02-04 Jan Hubicka <jh@suse.cz>
1017 * tree-sra.c (sra_walk_expr): Add linebreaks. BITFIELD_REFs into
1018 vectors might cause maybe_lookup_element_for_expr to be called
1019 on non-sra-candidate.
1021 2007-02-04 Kazu Hirata <kazu@codesourcery.com>
1023 * config/bfin/bfin-modes.def, config/bfin/bfin.c,
1024 config/bfin/bfin.md, config/bfin/predicates.md: Follow
1025 spelling conventions.
1027 2007-02-04 Richard Guenther <rguenther@suse.de>
1030 * fold-const.c (try_move_mult_to_index): Make sure to not
1031 overflow one dimension of a multi-dimensional array access.
1033 2007-02-04 Jan Hubicka <jh@suse.cz>
1035 * passes.c (init_optimization_passes): Reindent.
1037 2007-02-04 Jan Hubicka <jh@suse.cz>
1038 Eric Botcazou <ebotcazou@adacore.com>
1040 * tree-optimize.c (has_abnormal_outgoing_edge_p): Move to...
1041 (execute_fixup_cfg): Break out the abnormal goto code.
1042 * tree-inline.c (has_abnormal_outgoing_edge_p): ...here.
1043 (make_nonlocal_label_edges): Move here from execute_fixup_cfg.
1044 (optimize_inline_calls): Call make_nonlocal_label_edges.
1046 2007-02-04 Jan Hubicka <jh@suse.cz>
1048 * tree-ssa-copyrename.c (copy_rename_partition_coalesce): Return
1049 true when something was changed.
1050 (rename_ssa_copies): When something was changed, do
1051 TODO_remove_unused_locals.
1052 * tree-ssa-forwprop.c (tree_ssa_forward_propagate_single_use_value):
1053 add TODO_remove_unused_locals when instruction was removed.
1055 2007-02-04 Jan Hubicka <jh@suse.cz>
1057 * ipa-inline.c (try_inline): Improve debug output; work on already
1059 (cgraph_decide_inlining_incrementally): Indent; improve debug output;
1060 call try_inline for already inlined edges too when flattening;
1061 inline also functions that make callee growth but overall unit size
1064 2007-02-04 Kazu Hirata <kazu@codesourcery.com>
1066 * config/m32c/bitops.md, config/m32c/jump.md,
1067 config/m32c/m32c.c, config/m32c/m32c.h, config/m32r/m32r.c,
1068 config/m32r/m32r.h, config/m32r/m32r.md,
1069 config/m32r/predicates.md, config/m68hc11/larith.asm,
1070 config/m68hc11/m68hc11.c, config/m68hc11/m68hc11.h,
1071 config/m68k/m68k.h, config/mcore/mcore.md, config/mips/4k.md,
1072 config/mips/mips-protos.h, config/mips/mips.c,
1073 config/mips/mips.h, config/mips/mips.md, config/mips/mips16.S,
1074 config/mn10300/mn10300.h, config/mn10300/predicates.md,
1075 config/mt/mt.c, config/mt/mt.h, config/mt/mt.md: Follow
1076 spelling conventions.
1078 * config/v850/v850.c, config/v850/v850.h, config/v850/v850.md:
1079 Follow spelling conventions.
1081 2007-02-03 Douglas Gregor <doug.gregor@gmail.com>
1083 * c-opts.c (c_common_post_options): If C++0x mode is enabled, don't
1084 warn about C++0x compatibility.
1086 2007-02-04 Kazu Hirata <kazu@codesourcery.com>
1088 * config/h8300/h8300.c, config/h8300/h8300.h,
1089 config/h8300/h8300.md: Follow spelling conventions.
1091 2007-02-03 Uros Bizjak <ubizjak@gmail.com>
1094 * combine.c (try_combine): Do not substitute source operand
1095 with constants wider than 2 * HOST_BITS_PER_WIDE_INT.
1097 2007-02-03 Jan Hubicka <jh@suse.cz>
1099 PR gcov-profile/30650
1100 * value-prof.c (stringop_block_profile): Fix handling of size counter;
1101 do not divide by zero for never executed counters.
1102 (tree_find_values_to_profile): Fix counters.
1103 * gcov-ui.h (GCOV_COUNTER_AVERAGE, GCOV_COUNTER_IOR): Fix comments.
1105 2007-02-03 Ian Lance Taylor <iant@google.com>
1107 * lower-subreg.c (simple_move_operand): New static function,
1108 broken out of simple_move. Reject LABEL_REF, SYMBOL_REF, and HIGH
1110 (simple_move): Call simple_move_operand.
1111 (find_decomposable_subregs): Add special handling of MEMs.
1112 (can_decompose_p): Rename from cannot_decompose_p. Reverse
1113 meaning of return value. If we see a hard register, test whether
1114 it can store a word_mode value. Change all callers.
1116 2007-02-03 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
1118 * pa.md (addvdi3, addvsi3, subvdi3, subvsi3, negvdi2, negvsi2): New
1119 ftrapv insns and expanders.
1120 (subdi3): Change define_expand operand 1 to arith11_operand, and
1121 operand 2 to reg_or_0_operand. Change constraints of 64-bit insn
1122 pattern to handle reg_or_0 operands. Revise 32-bit insn pattern to
1123 handle 11-bit constants and reg_or_0 operands in operands 1 and 2,
1127 * varasm.c (notice_global_symbol): Treat global objects as weak when
1130 2007-02-03 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1132 * emit-rtl.c (dconstpi): Delete.
1134 (init_emit_once): Delete dconstpi and init dconstsqrt2.
1135 * real.h (dconstpi): Delete.
1137 * builtins.c (fold_builtin_cabs): Use dconstsqrt2.
1138 (fold_builtin_hypot): Likewise.
1140 2007-02-03 Tom Tromey <tromey@redhat.com>
1143 * gcc.c (cpp_unique_options): Any of -ggdb3, -gstabs3,
1144 -gcoff3, -gxcoff3, -gvms3 implies -dD.
1146 2007-02-03 Kazu Hirata <kazu@codesourcery.com>
1148 * c-decl.c, config/avr/avr.c, config/avr/avr.h,
1149 config/m68k/m68k.c, config/m68k/netbsd-elf.h,
1150 config/mn10300/mn10300.c, config/pdp11/pdp11.h,
1151 config/rs6000/cell.md, config/rs6000/darwin.h,
1152 config/sh/sh.md, config/sh/sh4-300.md, config/spu/spu.c,
1153 config/spu/spu.md, cselib.c, expr.c, haifa-sched.c, hwint.h,
1154 jump.c, reload.c, sched-deps.c, sched-int.h, tree-inline.c,
1155 tree-profile.c, tree-ssa-live.h, tree-vrp.c: Fix comment
1156 typos. Follow spelling conventions.
1157 * doc/invoke.texi: Follow spelling conventions.
1159 2007-02-03 Roger Sayle <roger@eyesopen.com>
1161 * simplify-rtx.c (simplify_relational_operation_1): Implement some
1162 canonicalization transformations that attempt to simplify integer
1163 constant comparisons to become comparisons against zero.
1165 2007-02-03 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1167 * builtins.c (fold_builtin_cabs): Fold cabs(x+xi) into
1169 * fold-const.c (fold_binary): Fix comment typos. Fold complex
1170 (x,0)-(0,y) into (x,-y). Likewise (0,y)-(x,0) into (-x,y).
1172 2007-02-02 Mike Stump <mrs@apple.com>
1174 * config/darwin9.h (DARWIN_LINKER_GENERATES_ISLANDS): Add.
1175 * config/rs6000/rs6000.c (DARWIN_GENERATE_ISLANDS): Add.
1176 (output_call): Use DARWIN_GENERATE_ISLANDS to decide when to
1177 generate a branch island.
1179 2007-02-02 Bob Wilson <bob.wilson@acm.org>
1181 * config/xtensa/xtensa.c (smalloffset_mem_p): Use BASE_REG_P.
1182 (xtensa_legitimate_address_p): New.
1183 (xtensa_legitimize_address): New.
1184 (xtensa_output_addr_const_extra): New.
1185 * config/xtensa/xtensa.h (REG_OK_STRICT_FLAG): Define.
1187 (REG_OK_FOR_BASE_P): Use BASE_REG_P.
1188 (GO_IF_LEGITIMATE_ADDRESS): Move code to xtensa_legitimate_address_p.
1189 (LEGITIMIZE_ADDRESS): Move code to xtensa_legitimize_address.
1190 (OUTPUT_ADDR_CONST_EXTRA): Move code to xtensa_output_addr_const_extra.
1191 * config/xtensa/xtensa-protos.h (xtensa_legitimate_address_p): New.
1192 (xtensa_legitimize_address): New.
1193 (xtensa_output_addr_const_extra): New.
1195 2007-02-02 Steve Ellcey <sje@cup.hp.com>
1197 * config/ia64/ia64.c (ia64_print_operand): Fix compare strings.
1199 2007-02-02 Ian Lance Taylor <iant@google.com>
1201 * expmed.c (expand_divmod): Add comment.
1203 2007-02-02 Kazu Hirata <kazu@codesourcery.com>
1205 * emit-rtl.c (renumber_insns): Remove.
1206 * flags.h: Remove the extern for flag_renumber_insns.
1207 * rtl.h: Remove the prototype for renumber_insns.
1208 * toplev.c (flag_renumber_insns): Remove.
1210 2007-02-02 Hui-May Chang <hm.chang@apple.com>
1212 Revert for x86 darwin:
1213 2005-06-19 Uros Bizjak <uros@kss-loka.si>
1215 * config/i386/i386.c (ix86_function_arg_regno_p): Put back the
1216 code before the following patch under TARGET_MACHO.
1217 (ix86_function_value_regno_p): Likewise.
1219 2007-02-02 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1221 * fold-const.c (negate_expr_p, fold_negate_expr): Handle
1224 2007-02-02 Maxim Kuvyrkov <mkuvyrkov@ispras.ru>
1227 * config/ia64/ia64.c (ia64_speculate_insn): Restrict to memory loads to
1228 general or fp registers. Add comments.
1229 * config/ia64/ia64.md (reg_pred_prefix): Add comment.
1231 2007-02-02 Paolo Bonzini <bonzini@gnu.org>
1233 * pointer-set.c (insert_aux): Only return insertion slot.
1234 (pointer_set_insert): Adjust.
1235 (pointer_set_traverse, struct pointer_map_t, pointer_map_create,
1236 pointer_map_destroy, pointer_map_insert, pointer_map_contains,
1237 pointer_map_traverse): New.
1238 * pointer-set.h (pointer_set_traverse, struct pointer_map_t,
1239 pointer_map_create, pointer_map_destroy, pointer_map_insert,
1240 pointer_map_contains, pointer_map_traverse): Declare.
1242 2007-02-02 Jakub Jelinek <jakub@redhat.com>
1245 * builtins.c (fold_builtin_sprintf): Do not attempt to optimize
1246 sprintf (str, "%s"). Do not optimize sprintf (str, "nopercent", p++).
1248 2007-02-02 Maxim Kuvyrkov <mkuvyrkov@ispras.ru>
1250 * sched-int.h (ds_to_dk, dk_to_ds): Declare functions.
1252 (struct _dep): New type.
1253 (dep_t): New typedef.
1254 (DEP_PRO, DEP_CON, DEP_KIND): New access macros.
1255 (DEP_STATUS): New access macro. The macro with the same name was
1256 renamed to DEP_LINK_STATUS.
1257 (dep_init): Declare function
1259 (struct _dep_link): New type.
1260 (dep_link_t): New typedef.
1261 (DEP_LINK_NODE, DEP_LINK_NEXT, DEP_LINK_PREV_NEXTP): New access macros.
1262 (DEP_LINK_DEP, DEP_LINK_PRO, DEP_LINK_CON, DEP_LINK_KIND): New macros.
1263 (DEP_LINK_STATUS): New macro.
1264 (debug_dep_links): New debug function.
1266 (struct _deps_list): New type.
1267 (deps_list_t): New typedef.
1268 (DEPS_LIST_FIRST): New access macro.
1269 (FOR_EACH_DEP_LINK): New cycle macro.
1270 (create_deps_list, free_deps_list, delete_deps_list): Declare
1272 (deps_list_empty_p, debug_deps_list, add_back_dep_to_deps_list): Ditto.
1273 (find_link_by_pro_in_deps_list, find_link_by_con_in_deps_list): Ditto.
1274 (copy_deps_list_change_con): Ditto.
1276 (move_dep_link): Declare function.
1278 (struct _dep_node): New type.
1279 (dep_node_t): New typedef.
1280 (DEP_NODE_BACK, DEP_NODE_DEP, DEP_NODE_FORW): New access macros.
1282 (struct haifa_insn_data.back_deps): New field to hold backward
1283 dependencies of the insn.
1284 (struct haifa_insn_data.depend): Rename to forw_deps. Change its type
1286 (struct haifa_insn_data.resolved_deps): Rename to resolved_back_deps.
1287 Change its type to deps_list_t.
1288 (INSN_BACK_DEPS): New access macro to use instead of LOG_LINKS.
1289 (INSN_DEPEND): Rename to INSN_FORW_DEPS.
1290 (RESOLVED_DEPS): Rename to INSN_RESOLVED_BACK_DEPS.
1292 (INSN_COST): Move to haifa-sched.c. Use insn_cost () instead.
1294 (DEP_STATUS): Rename to DEP_LINK_STATUS. Fix typo in the comment.
1296 (add_forw_dep, delete_back_forw_dep, insn_cost): Update declaration and
1298 (dep_cost): Declare.
1300 * sched-deps.c (CHECK): New macro to (en/dis)able sanity checks.
1301 (ds_to_dk, dk_to_ds): New functions.
1303 (init_dep_1): New static function.
1304 (init_dep): New function.
1305 (copy_dep): New static function.
1307 (dep_link_consistent_p, attach_dep_link, add_to_deps_list): New static
1309 (detach_dep_link): New static function.
1310 (move_dep_link): New function.
1312 (dep_links_consistent_p, dump_dep_links): New static functions.
1313 (debug_dep_links): New debugging function.
1315 (deps_obstack, dl_obstack, dn_obstack): New static variables.
1317 (alloc_deps_list, init_deps_list): New static functions.
1318 (create_deps_list): New function.
1319 (clear_deps_list): New static function.
1320 (free_deps_list, delete_deps_list, deps_list_empty_p): New functions.
1321 (deps_list_consistent_p, dump_deps_list): New static functions.
1322 (debug_deps_list): New function.
1323 (add_back_dep_to_deps_list, find_link_by_pro_in_deps_list): New
1325 (find_link_by_con_in_deps_list, copy_deps_list_change_con): Ditto.
1327 (maybe_add_or_update_back_dep_1, add_or_update_back_dep_1): Update to
1328 use new scheduler dependencies lists.
1329 (add_back_dep, delete_all_dependences, fixup_sched_groups): Ditto.
1330 (sched_analyze): Ditto. Initialize dependencies lists.
1331 (add_forw_dep, compute_forward_dependences): Update to use new
1332 scheduler dependencies lists.
1334 (init_dependency_caches): Init deps_obstack.
1335 (free_dependency_caches): Free deps_obstack.
1337 (adjust_add_sorted_back_dep, adjust_back_add_forw_dep): Update to use
1338 new scheduler dependencies lists.
1339 (delete_forw_dep, add_or_update_back_forw_dep): Ditto.
1340 (add_back_forw_dep, delete_back_forw_dep): Ditto.
1342 * sched-rgn.c (set_spec_fed, find_conditional_protection, is_pfree):
1343 Update to use new scheduler dependencies lists.
1344 (is_conditionally_protected, is_prisky, add_branch_dependences): Ditto.
1345 (debug_dependencies): Ditto.
1346 (schedule_region): Update comments.
1348 * sched-ebb.c (earliest_block_with_similiar_load): Update to use new
1349 scheduler dependencies lists.
1350 (schedule_ebb): Update comments.
1352 * rtl.def (DEPS_LIST): Remove.
1354 * lists.c (unused_deps_list): Remove.
1355 (free_list): Update assertions.
1357 (alloc_DEPS_LIST, free_DEPS_LIST_list, free_DEPS_LIST_node): Remove.
1358 (remove_free_DEPS_LIST_elem, copy_DEPS_LIST_list): Ditto.
1360 * rtl.h (free_DEPS_LIST_list, alloc_DEPS_LIST): Remove declarations.
1361 (remove_free_DEPS_LIST_elem, copy_DEPS_LIST_list): Ditto.
1363 * haifa-sched.c (comments): Update.
1364 (insn_cost1): Remove. Inline the code into insn_cost ().
1365 (insn_cost): Update to use new scheduler dependencies lists. Move
1366 processing of the dependency cost to dep_cost ().
1367 (dep_cost): New function. Use it instead of insn_cost () when
1368 evaluating cost of the dependency. Use compatible interface to
1369 interact with the target.
1370 (priority): Update to use new scheduler dependencies lists.
1371 (rank_for_schedule): Ditto. Optimize heuristic that prefers the insn
1372 with greater number of insns that depend on the insn.
1373 (schedule_insn): Update to use new scheduler dependencies lists. Add
1374 code to free backward dependencies lists. Inline and optimize code
1375 from resolve_dep () - see PR28071.
1376 (ok_for_early_queue_removal): Update to use new scheduler dependencies
1377 lists. Update call to targetm.sched.is_costly_dependence hook.
1379 (fix_inter_tick, try_ready, fix_tick_ready): Update to use new
1380 scheduler dependencies lists.
1382 (resolve_dep): Remove. Move the logic to schedule_insn ().
1383 (init_h_i_d): Initialize dependencies lists.
1385 (process_insn_depend_be_in_spec): Rename to
1386 process_insn_forw_deps_be_in_spec. Update to use new scheduler
1388 (add_to_speculative_block, create_check_block_twin, fix_recovery_deps):
1389 Update to use new scheduler dependencies lists.
1390 (clear_priorities, calc_priorities, add_jump_dependencies): Ditto.
1392 * ddg.c (create_ddg_dependence, create_ddg_dep_no_link): Update to use
1393 new scheduler dependencies lists.
1394 (build_intra_loop_deps): Ditto.
1396 * target.h (struct _dep): Declare to use in
1397 gcc_target.sched.is_costly_dependence.
1398 (struct gcc_target.sched.adjust_cost): Fix typo.
1399 (struct gcc_target.sched.is_costly_dependence): Change signature to use
1400 single dep_t parameter instead of an equivalent triad.
1401 (struct gcc_target.sched.adjust_cost_2): Remove.
1403 * target-def.h (TARGET_SCHED_ADJUST_COST_2): Remove.
1405 * reg-notes.def (DEP_TRUE, DEP_OUTPUT, DEP_ANTI): Update comments.
1407 * doc/tm.texi (TARGET_SCHED_IS_COSTLY_DEPENDENCE): Update
1409 (TARGET_SCHED_ADJUST_COST_2): Remove documentation.
1411 * doc/rtl.texi (LOG_LINKS): Remove part about instruction scheduler.
1412 (REG_DEP_TRUE): Document.
1414 * config/ia64/ia64.c (ia64_adjust_cost_2): Rename to ia64_adjust_cost.
1415 Change signature to correspond to the targetm.sched.adjust_cost hook.
1416 Update use in TARGET_SCHED_ADJUST_COST_2.
1417 (TARGET_SCHED_ADJUST_COST_2): Rename to TARGET_SCHED_ADJUST_COST.
1418 (ia64_dependencies_evaluation_hook, ia64_dfa_new_cycle): Update to use
1419 new scheduler dependencies lists.
1420 (ia64_gen_check): Ditto.
1422 * config/mips/mips.c (vr4130_swap_insns_p): Update to use new scheduler
1425 * config/rs6000/rs6000.c (rs6000_is_costly_dependence): Change
1426 signature to correspond to the targetm.sched.is_costly_dependence hook.
1427 (is_costly_group): Update to use new scheduler dependencies lists.
1429 * config/spu/spu.c (spu_sched_adjust_cost): Use insn_cost () function
1430 instead of INSN_COST () macro.
1432 2007-02-01 Ian Lance Taylor <iant@google.com>
1434 * lower-subreg.c (resolve_clobber): Handle a subreg of a concatn.
1436 2007-02-01 Guy Martin <gmsoft@gentoo.org>
1438 * pa.md (tp_load): Correct mfctl instruction syntax.
1440 2007-02-01 Geoffrey Keating <geoffk@apple.com>
1442 * config/rs6000/rs6000.c (rs6000_stack_info): Correct
1443 altivec_padding_size calculation on AIX. Improve comment, add
1444 assert to verify that it's right.
1446 * config/rs6000/darwin.h (MD_UNWIND_SUPPORT): Don't define for
1449 2007-2-01 Seongbae Park <seongbae.park@gmail.com>
1452 * global.c (global_alloc): Add mising initialization of
1455 2007-02-01 Roger Sayle <roger@eyesopen.com>
1457 * alias.c (init_alias_analysis): Correct whitespace.
1458 * bb-reorder.c (fix_edges_for_rarely_executed_code,
1459 partition_hot_cold_basic_blocks): Likewise.
1460 * builtins.c (expand_builtin_printf, expand_builtin_fprintf,
1461 expand_builtin_sprintf, fold_builtin_carg, fold_builtin_sprintf,
1462 maybe_emit_sprintf_chk_warning, fold_builtin_sprintf_chk,
1463 fold_builtin_snprintf_chk, fold_builtin_printf,
1464 fold_builtin_fprintf, do_mpfr_ckconv, do_mpfr_arg1, do_mpfr_arg2,
1465 do_mpfr_arg3, do_mpfr_sincos): Likewise.
1466 * cfgcleanup.c (cleanup_cfg): Likewise.
1467 * cfgexpand.c (tree_expand_cfg): Likewise.
1468 * fold-const.c (fold_binary) <RDIV_EXPR>: Likewise.
1469 * function.c (get_next_funcdef_no): Likewise.
1470 * gengtype.c (main): Likewise.
1471 * genmodes.c (main): Likewise.
1472 * gcse.c (bypass_conditional_jumps, print_ldst_list): Likewise.
1473 * haifa-sched.c (schedule_block, extend_h_i_d): Likewise.
1474 * ifcvt.c (noce_emit_move_insn): Likewise.
1475 * modulo-sched.c (generate_prolog_epilog, sms_schedule_by_order):
1477 * stor-layout.c (get_best_mode): Likewise.
1478 * tree-ssa-loop-niter.c (get_val_for): Likewise.
1479 * tree-ssa-structalias.c (get_varinfo, get_varinfo_fc,
1480 scc_visit, do_ds_constraint, do_complex_constraint, label_visit,
1481 perform_var_substitution, solve_graph): Likewise.
1482 * tree-vrp.c (vrp_finalize): Likewise.
1484 2007-02-01 Ian Lance Taylor <iant@google.com>
1486 * lower-subreg.c (simplify_gen_subreg_concatn): If we ask for the
1487 high part of a paradoxical subreg, return a constant zero.
1489 2007-02-01 Zdenek Dvorak <dvorakz@suse.cz>
1491 * toplev.c (lang_dependent_init): Call init_set_costs.
1492 * loop-init.c (loop_optimizer_init): Do not call init_set_costs.
1494 2007-02-01 Richard Guenther <rguenther@suse.de>
1497 * fold-const.c (fold_negate_expr): Allow negating a
1498 constant if overflow does not change.
1500 2007-02-01 Ralf Wildenhues <Ralf.Wildenhues@gmx.de>
1502 * doc/c-tree.texi (Expression trees): Improve markup.
1503 * doc/tm.texi (Register Classes, Addressing Modes)
1504 (Floating Point): Fix spacing after abbreviations. Fix some
1507 2007-02-01 Ben Elliston <bje@au.ibm.com>
1509 * doc/invoke.texi: Replace "bugfix" with "bug fix" throughout.
1510 * doc/contrib.texi: Likewise.
1511 * doc/install.texi: Likewise.
1513 2007-01-31 Richard Henderson <rth@redhat.com>
1514 Ian Lance Taylor <iant@google.com>
1516 * lower-subreg.c: New file.
1517 * rtl.def (CONCATN): Define.
1518 * passes.c (init_optimization_passes): Add pass_lower_subreg and
1520 * emit-rtl.c (update_reg_offset): New static function, broken out
1521 of gen_rtx_REG_offset.
1522 (gen_rtx_REG_offset): Call update_reg_offset.
1523 (gen_reg_rtx_offset): New function.
1524 * regclass.c: Revert patch of 2006-03-05, restoring
1526 (clear_reg_info_regno): New function.
1527 * dwarf2out.c (concatn_loc_descriptor): New static function.
1528 (loc_descriptor): Handle CONCATN.
1529 * common.opt (fsplit_wide_types): New option.
1530 * opts.c (decode_options): Set flag_split_wide_types when
1532 * timevar.def (TV_LOWER_SUBREG): Define.
1533 * rtl.h (gen_reg_rtx_offset): Declare.
1534 (reg_scan_update): Declare.
1535 * regs.h (clear_reg_info_regno): Declare.
1536 * tree-pass.h (pass_lower_subreg): Declare.
1537 (pass_lower_subreg2): Declare.
1538 * doc/invoke.texi (Option Summary): List -fno-split-wide-types.
1539 (Optimize Options): Add -fsplit-wide-types to -O1 list. Document
1541 * doc/rtl.texi (Regs and Memory): Document concat and concatn.
1542 * Makefile.in (OBJS-common): Add lower-subreg.o.
1543 (lower-subreg.o): New target.
1545 2007-01-31 Kazu Hirata <kazu@codesourcery.com>
1547 * config/sh/sh.h (HAVE_SECONDARY_RELOADS): Remove.
1549 2007-01-31 Anatoly Sokolov <aesok@post.ru>
1552 * config/avr/avr.c (DWARF2_ADDR_SIZE): Define.
1554 2007-01-31 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1557 * builtins.c (fold_builtin_sqrt): Use MPFR for constant args.
1559 2007-01-31 Zdenek Dvorak <dvorakz@suse.cz>
1561 * cfgloop.h: Include vec-prim.h.
1562 (enum li_flags): Remove LI_ONLY_OLD.
1563 (loop_iterator): Changed.
1564 (fel_next, fel_init): Iterate over loop tree.
1565 (FOR_EACH_LOOP_BREAK): New macro.
1566 * loop-unswitch.c (unswitch_loops): Do not pass LI_ONLY_OLD to
1568 * tree-ssa-loop-unswitch.c (tree_ssa_unswitch_loops): Ditto.
1569 * modulo-sched.c (sms_schedule): Ditto.
1570 * tree-vectorizer.c (vectorize_loops): Ditto.
1571 * doc/loop.texi: Update information on loop numbering and behavior of
1572 FOR_EACH_LOOP wrto new loops.
1573 * tree-scalar-evolution.c (compute_overall_effect_of_inner_loop,
1574 add_to_evolution_1): Test nestedness of loops instead of comparing
1576 * tree-chrec.c (chrec_fold_plus_poly_poly,
1577 chrec_fold_multiply_poly_poly, chrec_evaluate,
1578 hide_evolution_in_other_loops_than_loop, chrec_component_in_loop_num,
1579 reset_evolution_in_loop): Ditto.
1580 * Makefile.in (CFGLOOP_H): Add vecprim.h dependency.
1582 2007-01-31 Dirk Mueller <dmueller@suse.de>
1584 * c-common.c (warn_about_parentheses): Separate warning about
1585 un-parenthized sequence of comparison operators from the one
1586 which is supposed to warn about x <= y <= z.
1588 2007-01-31 Uros Bizjak <ubizjak@gmail.com>
1590 * optabs.h (enum optab_index): Add new OTI_isinf.
1591 (isinf_optab): Define corresponding macro.
1592 * optabs.c (init_optabs): Initialize isinf_optab.
1593 * genopinit.c (optabs): Implement isinf_optab using isinf?f2
1595 * builtins.c (mathfn_built_in): Handle BUILT_IN_ISINF{,F,L}.
1596 (expand_builtin_interclass_mathfn): Expand BUILT_IN_ISINF{,F,L}
1598 (expand_builtin): Expand BUILT_IN_ISINF{,F,L} using
1599 expand_builtin_interclass_mathfn.
1600 * reg_stack.c (subst_stack_regs_pat): Handle UNSPEC_FXAM.
1601 * config/i386/i386.md (UNSPEC_FXAM): New constant.
1602 (fxam<mode>2_i387): New insn pattern.
1603 (isinf<mode>2) New expander to implement isinf, isinff and isinfl
1604 built-in functions as x87 inline asm.
1606 2007-01-31 Kazu Hirata <kazu@codesourcery.com>
1608 * gcc/config/arm/unwind-arm.h (_sleb128_t, _uleb128_t): New.
1610 2007-01-30 Eric Christopher <echristo@apple.com>
1612 * config.gcc: Add geode.
1614 2007-01-31 Kazu Hirata <kazu@codesourcery.com>
1616 * cgraphunit.c, config/arm/arm.c, config/m68k/m68k.c,
1617 ipa-inline.c, tree-profile.c, tree-ssa-live.c,
1618 tree-ssa-math-opts.c, tree-ssanames.c, tree-vect-analyze.c,
1619 value-prof.c: Fix comment typos.
1621 2007-01-30 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
1624 * doc/invoke.texi (Wpointer-arith): Document warning.
1626 2007-01-30 Janis Johnson <janis187@us.ibm.com>
1628 * doc/extend.texi (Decimal Floating Types): Remove decfloat.h from
1629 the list of discrepancies from the draft TR.
1631 2007-01-30 Dirk Mueller <dmueller@suse.de>
1634 * doc/invoke.texi (-Wreturn-type): Update description to
1637 2007-01-30 Richard Sandiford <richard@codesourcery.com>
1639 * cfgrtl.c (try_redirect_by_replacing_jump): Check only_sets_cc0_p.
1641 2007-01-30 Uros Bizjak <ubizjak@gmail.com>
1643 * builtins.c (expand_builtin_int_interclass_roundingfn): New function
1644 to handle optabs that operate on floating point input argument and
1645 output to integer output.
1646 (expand_builtin_mathfn) [BUILT_IN_ILOGB]: Move from here ...
1647 (expand_builtin_interclass_mathfn) [BUILT_IN_ILOGB]: ... to here.
1648 (expand_builtin): Expand BUILT_IN_ILOGB{,F,L} using
1649 expand_builtin_interclass_mathfn ().
1650 * config/i386/i386.md (fxtractxf3_i387): Rename from *fxtractxf3_i387.
1652 (ilogbxf2, ilogb<mode>2): New expanders to implement ilogb, ilogbf and
1653 ilogbl built-in functions as x87 intrinsics.
1655 2007-01-30 Richard Guenther <rguenther@suse.de>
1658 * dwarf2out.c (reference_to_unused): Query varpool if the
1659 variable was output.
1661 2007-01-30 Richard Guenther <rguenther@suse.de>
1664 * passes.c (execute_one_pass): Reset in_gimple_form to not
1665 confuse non-unit-at-a-time mode.
1667 2007-01-29 Roger Sayle <roger@eyesopen.com>
1668 Richard Guenther <rguenther@suse.de>
1670 * fold-const.c (round_up): Make HIGH an unsigned HOST_WIDE_INT to
1671 avoid undefined behaviour on overflow. Use force_fit_type_double
1672 to construct the constant with the specified TREE_OVERFLOW.
1674 2007-01-29 Janis Johnson <janis187@us.ibm.com>
1676 * config/dfp-bit.c: Add parameterized support for fp exceptions.
1677 * config/dfp-bit.h: Ditto.
1679 2007-01-29 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
1681 * c-decl.c (pop_scope): Replace warnings with call to
1682 warn_for_unused_label.
1683 * c-common.h (warn_for_unused_label): Declare.
1684 * c-common.c (warn_for_unused_label): Define.
1686 2007-01-29 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
1688 * tree-optimize.c (update_inlined_to_pointers): Delete unused
1691 2007-01-29 Janis Johnson <janis187@us.ibm.com>
1693 * Makefile.in (USER_H): Remove decfloat.h.
1694 * ginclude/decfloat.h: Delete, moving contents to ...
1695 * ginclude/float.h: Add support for decimal floating point,
1696 guarded by __STDC_WANT_DEC_FP__.
1698 2007-01-29 Mike Stump <mrs@apple.com>
1700 * doc/gccint.texi (Top): Rename Loop Representation to Loop
1701 Analysis and Representation to resolve case insensitive conflict.
1702 * doc/loop.texi (Loop Analysis and Representation): Likewise.
1704 2007-01-28 Daniel Berlin <dberlin@dberlin.org>
1706 * tree.h (struct tree_memory_tag): Add aliases member.
1707 (MTAG_ALIASES): New macro.
1708 * tree-ssa-alias.c (alias_bitmap_obstack): New variable.
1709 (add_may_alias): Remove pointer-set. Update for may_aliases being
1711 (mark_aliases_call_clobbered): Update for may_aliases being a
1713 (compute_tag_properties): Ditto.
1714 (create_partition_for): Ditto.
1715 (compute_memory_partitions): Ditto.
1716 (dump_may_aliases_for): Ditto.
1717 (is_aliased_with): Ditto.
1718 (add_may_alias_for_new_tag): Ditto.
1719 (rewrite_alias_set_for): Rewrite for may_aliases being a bitmap.
1720 (compute_is_aliased): New function.
1721 (compute_may_aliases): Call compute_is_aliased).
1722 (init_alias_info): Initialize alias_bitmap_obstack.
1723 (union_alias_set_into): New function.
1724 (compute_flow_sensitive_aliasing): Use union_aliases_into.
1725 (have_common_aliases_p): Rewrite to take two bitmaps and use
1727 (compute_flow_insensitive_aliasing): Stop using pointer-sets.
1729 (finalize_ref_all_pointers): Update for add_may_alias changes.
1730 (new_type_alias): Ditto.
1731 * tree-flow-inline.h (may_aliases): Return a bitmap.
1732 * tree-dfa.c (dump_variable): Check for MTAG_P'ness.
1733 * tree-ssa.c (verify_flow_insensitive_alias_info): Update for
1734 may_aliases being a bitmap.
1735 * tree-flow.h (struct var_ann_d): Remove may_aliases member.
1736 may_aliases now returns a bitmap.
1737 * tree-ssa-structalias.c (merge_smts_into): Update for may_aliases
1739 * tree-ssa-operands.c (add_virtual_operand): Update for
1740 may_aliases being a bitmap.
1742 2007-01-29 Daniel Berlin <dberlin@dberlin.org>
1744 PR tree-optimization/30630
1745 * tree-ssa-structalias.c (do_complex_constraint): Mark correct
1746 variable as changed.
1748 2007-01-29 Simon Martin <simartin@users.sourceforge.net>
1751 * gimplify.c (gimplify_target_expr): Make sure that the TARGET_EXPR is
1752 expanded only once even if an error occurs.
1754 2007-01-29 Ben Elliston <bje@au.ibm.com>
1756 * gcov-io.h (__gcov_indirect_call_profiler): Declare.
1757 (__gcov_average_profiler): Likewise.
1758 (__gcov_ior_profiler): Likewise.
1759 (__gcov_merge_ior): Likewise.
1761 2007-01-28 Jan Hubicka <jh@suse.cz>
1763 * builtins.c (expand_builtin_memset): Fix typo in my last patch.
1764 * value-prof.c (stringop_block_profile): Likewise.
1766 2007-01-28 Jan Hubicka <jh@suse.cz>
1768 * expr.c (emit_block_move_via_movmem, emit_block_move_via_libcall): Add
1769 variant handling histograms; add wrapper.
1770 (clear_storage_via_libcall): Export.
1771 (emit_block_move_hints): Break out from ...; add histograms.
1772 (emit_block_move): ... this one.
1773 (clear_storage_hints): Break out from ...; add histograms.
1774 (clear_storage): ... this one.
1775 (set_storage_via_memset): Handle histogram.
1776 * expr.h (emit_block_move_via_libcall, emit_block_move_hints): Declare.
1777 (clear_storage_hints, clear_storage_via_libcall): Declare.
1778 (set_storage_via_setmem): Update prototype.
1779 * doc/md.texi (movmem, setmem): Document new arguments.
1781 * value-prof.c (dump_histogram_value, tree_find_values_to_profile): Add
1783 (stringop_block_profile): New global function.
1784 (tree_stringops_values_to_profile): Profile block size and alignment.
1785 * value-prof.h (enum hist_type): add HIST_TYPE_AVERAGE and
1787 (struct profile_hooks): Add gen_average_profiler and gen_ior_profiler.
1788 (stringop_block_profile): Declare.
1789 * builtins.c: Include value-prof.h.
1790 (expand_builtin_memcpy, expand_builtin_memset): Pass block profile.
1791 * gcov-ui.h (GCOV_COUNTER_NAMES): Add new counter.
1792 (GCOV_COUNTER_AVERAGE, GCOV_COUNTER_IOR): New constants.
1793 (GCOV_COUNTERS, GCOV_LAST_VALUE_COUNTER): Update.
1794 * profile.c (instrument_values): Add new counters.
1795 * cfgexpand.c (expand_gimple_basic_block): Propagate histograms to
1797 * tree-profile.c (tree_average_profiler_fn, tree_ior_profiler_fn): New.
1798 (tree_init_edge_profiler): Build new profilers.
1799 (tree_gen_average_profiler, tree_gen_ior_profiler): New.
1800 (pass_tree_profile): Add dump.
1801 (tree_profile_hooks): Update.
1802 * Makefile.in (LIBGCOV): Add new constants.
1803 * libgcov.c (__gcov_merge_ior, __gcov_average_profiler,
1804 __gcov_ior_profiler): New.
1805 * i386.md (movmem/setmem expanders): Add new optional arguments.
1807 2007-01-28 David Edelsohn <edelsohn@gnu.org>
1809 * doc/md.texi (Standard Pattern Names): Document blockage pattern.
1811 2007-01-28 Zdenek Dvorak <dvorakz@suse.cz>
1813 * tree-ssa-loop-unswitch.c: Include tree-inline.h.
1814 (tree_unswitch_single_loop): Pass eni_size_weights to
1815 tree_num_loop_insns.
1816 * tree-ssa-loop-manip.c: Include tree-inline.h.
1817 (can_unroll_loop_p): Pass eni_size_weights to
1818 tree_num_loop_insns.
1819 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p):
1820 Pass eni_size_weights to estimate_num_insns.
1821 * tree.h (init_inline_once): Export.
1822 * toplev.c (backend_init): Call init_inline_once.
1823 * cgraphunit.c (cgraph_process_new_functions,
1824 cgraph_analyze_function): Pass eni_inlining_weights to
1826 * ipa-inline.c (compute_inline_parameters): Ditto.
1827 * tree-ssa-loop-ivcanon.c (tree_num_loop_insns): Pass weights
1828 to estimate_num_insns.
1829 (try_unroll_loop_completely): Pass eni_size_weights to
1830 tree_num_loop_insns.
1831 * tree-eh.c (decide_copy_try_finally): Pass eni_size_weights
1832 ot estimate_num_insns.
1833 * tree-ssa-loop-prefetch.c: Include tree-inline.h.
1834 (loop_prefetch_arrays): Pass eni_time_weights to tree_num_loop_insns.
1835 * tree-inline.c (eni_inlining_weights, eni_size_weights,
1836 eni_time_weights): New variables.
1837 (init_inline_once): Initialize them.
1838 (struct eni_data): Mew.
1839 (estimate_num_insns_1, estimate_num_insns): Use weights.
1840 * tree-inline.h (struct eni_weights_d): New.
1841 (eni_inlining_weights, eni_size_weights, eni_time_weights): Declare.
1842 (estimate_num_insns): Declaration changed.
1843 * cfgloop.h (tree_num_loop_insns): Declaration changed.
1844 * Makefile.in (tree-ssa-loop-unswitch.o, tree-ssa-loop-prefetch.o,
1845 tree-ssa-loop-manip.o): Add TREE_INLINE_H dependency.
1847 2007-01-28 Zdenek Dvorak <dvorakz@suse.cz>
1849 * tree-data-ref.c (conflict_fn): Assert that the number of affine
1850 relations in the conflict function is valid.
1852 2007-01-27 Ian Lance Taylor <iant@google.com>
1854 * common.opt: Add fstrict-overflow.
1855 * opts.c (decode_options): Set flag_strict_overflow if -O2.
1856 * flags.h (TYPE_OVERFLOW_WRAPS): Define.
1857 (TYPE_OVERFLOW_UNDEFINED): Define.
1858 (TYPE_OVERFLOW_TRAPS): Define. This replaces TYPE_TRAP_SIGNED.
1860 * tree.h (TYPE_TRAP_SIGNED): Don't define.
1861 * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_UNDEFINED.
1862 (fold_negate_expr): Likewise.
1863 (make_range): Likewise.
1864 (extract_muldiv_1): Likewise.
1865 (maybe_canonicalize_comparison): Likewise.
1866 (fold_comparison): Likewise.
1867 (fold_binary): Likewise.
1868 (tree_expr_nonnegative_p): Likewise.
1869 (tree_expr_nonzero_p): Likewise.
1870 * tree-vrp.c (compare_values): Likewise.
1871 (extract_range_from_binary_expr): Likewise.
1872 (extract_range_from_unary_expr): Likewise.
1873 * tree-ssa-loop-niter.c (infer_loop_bounds_from_signedness):
1875 (nowrap_type_p): Likewise.
1876 * tree-scalar-evolution.c (simple_iv): Likewise.
1877 * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_WRAPS.
1878 (build_range_check): Likewise.
1879 (extract_muldiv_1): Likewise.
1880 (fold_comparison): Likewise.
1881 * tree-vrp.c (vrp_int_const_binop): Likewise.
1882 (extract_range_from_unary_expr): Likewise.
1883 * convert.c (convert_to_integer): Likewise.
1884 * fold-const.c (fold_negate_expr): Use TYPE_OVERFLOW_TRAPS.
1885 (fold_comparison): Likewise.
1886 (fold_binary): Likewise.
1887 * optabs.c (optab_for_tree_code): Likewise.
1888 * tree-vectorizer.c (vect_is_simple_reduction): Likewise.
1889 * simplify-rtx.c (simplify_const_relational_operation): Check
1890 flag_strict_overflow and flag_trapv.
1891 (simplify_const_relational_operation): Likewise.
1892 * doc/invoke.texi (Option Summary): Mention -fstrict-overflow.
1893 (Optimize Options): Add -fstrict-overflow to -O2 list. Document
1896 2007-01-27 Roger Sayle <roger@eyesopen.com>
1898 * tree.c (tree_fold_gcd): Delete.
1899 * tree.h (tree_fold_gcd): Remove prototype.
1900 * tree-data-ref.c (tree_fold_divides_p): Don't use tree_fold_gcd to
1901 test whether one constant integer is a multiple of another. Instead
1902 call int_const_binop with TRUNC_MOD_EXPR and test for a zero result.
1903 * fold-const.c (multiple_of_p): We've determined both TOP and
1904 BOTTOM are integer constants so we can call int_const_binop directly
1905 instead of the more generic const_binop.
1907 2007-01-27 Roger Sayle <roger@eyesopen.com>
1909 * fold-const.c (size_binop): In the fast-paths for X+0, 0+X, X-0 and
1910 1*X check that the constant hasn't overflowed, to preserve the
1912 (round_up): Provide an efficient implementation when rouding-up an
1913 INTEGER_CST to a power-of-two.
1915 2007-01-28 Ralf Wildenhues <Ralf.Wildenhues@gmx.de>
1917 * doc/sourcebuild.texi: Add comma for clarity.
1918 * doc/extend.texi: Fix some typos.
1919 * doc/passes.texi: Likewise.
1920 * doc/cppinternals.texi: Likewise.
1921 * doc/c-tree.texi: Likewise.
1922 * doc/tree-ssa.texi: Likewise.
1923 * doc/install.texi: Likewise.
1925 2007-01-27 Jan Hubicka <jh@suse.cz>
1927 * tree-sra.c (sra_walk_function): Don't rely on aliases being build.
1928 (pass_sra): Do not require alias information.
1929 * passes.c (init_optimization_passes): Add SRA
1931 2007-01-27 Steven Bosscher <steven@gcc.gnu.org>
1933 * tracer.c (rest_of_handle_tracer): We already cleaned
1934 up the CFG in tracer() so don't do it here again.
1935 * cfgcleanup.c (rest_of_handle_jump2): Don't repeat
1936 cleanup_cfg here, either. And don't call renumber_insns.
1938 * cfgrtl.c (rtl_verify_flow_info_1): Don't verify that BB_END
1939 and BB_HEAD are in the insn stream here. Instead make sure
1940 that BB_INSN is valid on all insns. Also, do check here that
1941 there are no pending branch predictions...
1942 (rtl_verify_flow_info): ...instead of doing it here. Checks
1943 for BB_END and BB_HEAD moved from rtl_verify_flow_info_1 to
1946 2007-01-26 Roger Sayle <roger@eyesopen.com>
1948 * config/i386/i386.c (ix86_swap_binary_operands_p): New helper
1949 function to simplify/factorize operand order canonicalization.
1950 (ix86_fixup_binary_operands): Reorganize using the above function.
1951 (ix86_binary_operator_ok): Likewise.
1953 2007-01-27 Jakub Jelinek <jakub@redhat.com>
1955 * genattrtab.c (struct attr_value_list, insn_code_values): Move to
1956 file scope from optimize_attrs.
1957 (simplify_test_exp): If insn_code_values is not NULL, use it to speed
1959 (optimize_attrs): Clear insn_code_values after freeing it.
1961 2007-01-26 Zdenek Dvorak <dvorakz@suse.cz>
1963 * tree-ssa-address.c (create_mem_ref): Remove ", bsi" from
1964 a parts.base assignment.
1966 2007-01-26 Zdenek Dvorak <dvorakz@suse.cz>
1968 * tree-data-ref.c (dump_subscript): Use dump_conflict_function.
1969 (compute_subscript_distance, initialize_data_dependence_relation,
1970 finalize_ddr_dependent, analyze_ziv_subscript,
1971 analyze_siv_subscript_cst_affine,
1972 compute_overlap_steps_for_affine_univar,
1973 compute_overlap_steps_for_affine_1_2, analyze_subscript_affine_affine,
1974 analyze_siv_subscript, analyze_miv_subscript,
1975 analyze_overlapping_iterations, subscript_dependence_tester_1,
1976 compute_self_dependence, free_dependence_relation): Work
1977 with affine_fn instead of chrecs.
1978 (dump_affine_function, dump_conflict_function, affine_function_equal_p,
1979 common_affine_function, affine_function_base,
1980 affine_function_constant_p, affine_fn_op, affine_fn_plus,
1981 affine_fn_minus, affine_fn_free, conflict_fn_not_known,
1982 conflict_fn_no_dependence, free_conflict_function, free_subscripts,
1983 conflict_fn, affine_fn_cst, affine_fn_univar): New functions.
1984 (all_chrecs_equal_p): Removed.
1985 * tree-data-ref.h (affine_fn, conflict_function): New types.
1986 (struct subscript): Change type of conflicting_iterations_in_a
1987 and conflicting_iterations_in_b.
1989 2007-01-26 Steve Ellcey <sje@cup.hp.com>
1992 * config/pa/pa.h (TARGET_HPUX_11): New.
1993 * config/pa/pa-hpux11.h (TARGET_HPUX_11): New.
1994 * config/pa/pa.c (pa_init_builtins): Use TARGET_HPUX_11.
1996 2007-01-26 Daniel Berlin <dberlin@dberlin.org>
1997 Richard Guenther <rguenther@suse.de>
1999 * tree-ssa-structalias.c (solve_graph): Handle case
2000 we merged the variable to another.
2002 2007-01-25 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
2004 * builtins.c (fold_builtin_1): Treat ccos and ccosh as 'even'
2007 * fold-const.c (negate_mathfn_p): Treat casin, casinh, catan,
2008 catanh, cproj, csin, csinh, ctan and ctanh as 'odd' functions.
2010 2007-01-25 DJ Delorie <dj@redhat.com>
2012 * config/m32c/m32c.c (m32c_cannot_change_mode_class): We don't
2013 allow changes to modes which don't fit in those registers.
2015 * reload1.c (choose_reload_regs): Check for invalid subregs before
2016 computing their locations, not after.
2018 2007-01-25 Geoffrey Keating <geoffk@apple.com>
2021 * config/rs6000/rs6000.c (first_altivec_reg_to_save): On Darwin,
2022 save Altivec registers in an eh_return function.
2023 (compute_vrsave_mask): Likewise.
2024 (rs6000_stack_info): Correct AIX/Darwin stack alignment computation
2025 for saving Altivec registers.
2026 (rs6000_emit_prologue): Don't allocate stack twice in
2027 eh_return function. Correct expected value of altivec_save_offset
2028 when using save_world. Describe save of R0 to stack when using
2029 save_world. Describe stack pointer adjustment when using
2030 save_world. Remove duplicated eh_return parameter register saving.
2031 Update sp_offset variable after save_world.
2032 * config/rs6000/t-darwin (LIB2FUNCS_STATIC_EXTRA): Remove
2034 (LIB2FUNCS_EXTRA): Add darwin-world.asm.
2035 * config/rs6000/darwin.h (SUBTARGET_OVERRIDE_OPTIONS): -m64
2038 2007-01-25 Steve Ellcey <sje@cup.hp.com>
2040 * config.gcc (ia64*-*-hpux*): Make posix threads the default.
2042 2007-01-25 Steve Ellcey <sje@cup.hp.com>
2045 * config/pa/pa.c (pa_init_builtins): Set asm names for finite routines.
2046 * config/ia64/ia64.c (ia64_init_builtins): Ditto.
2048 2007-01-25 Richard Guenther <rguenther@suse.de>
2050 * doc/invoke.texi (-Wcoverage-mismatch): Document.
2051 * common.opt (-Wcoverage-mismatch): New warning option.
2052 * coverage.c (get_coverage_counts): Ignore coverage mismatch
2053 if -Wcoverage-mismatch is given.
2055 2007-01-25 Razya Ladelsky <razya@il.ibm.com>
2057 * ipa-cp.c (ipcp_insert_stage, ipcp_driver): Support for SSA.
2058 (ipcp_driver): Change to static definition.
2059 Add dumping of the ifunctions.
2060 (constant_val_insert): Remove unused parameter. Support for SSA.
2061 (ipcp_propagate_const): Support for SSA.
2062 (ipcp_profile_bb_print): Print only analyzed nodes.
2063 (ipcp_replace_map_create): Remove support for Fortran constant
2065 * ipa-prop.c (ipa_method_modify_stmt,
2066 ipa_callsite_compute_param): Support for SSA.
2067 * ipa-prop.h (ipcp_driver): Remove declaration.
2068 (IS_VALID_TREE_MAP_INDEX): Add define.
2070 2007-01-24 Geoffrey Keating <geoffk@apple.com>
2072 * unwind-dw2.c (execute_stack_op): Handle DW_OP_swap.
2074 2007-01-24 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
2077 * builtins.c (fold_builtin_cabs): Use MPFR to evaluate a
2078 constant argument to cabs and do it without checking for
2079 -funsafe-math-optimizations.
2081 2007-01-24 Douglas Gregor <dgregor@osl.iu.edu>
2083 * c-common.h (RID_FIRST_CXX0X): New.
2084 (RID_LAST_CXX0X): New.
2085 * c-opts.c (c_common_handle_option): -Wc++0x-compat is triggered
2087 * c.opt (Wc++0x-compat): New.
2088 * doc/invoke.texi (-Wc++0x-compat): Document.
2090 2007-01-24 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
2092 * builtins.c (fold_builtin_carg): New.
2093 (fold_builtin_1): Use it.
2095 2007-01-24 Jan Hubicka <jh@suse.cz>
2097 * ipa-inline.c (cgraph_decide_inlining): Initialize initial_insns.
2099 * ipa-inline.c (initial_insns, max_insns): Delete.
2100 (compute_max_insns): New function.
2101 (cgraph_decide_inlining_of_small_function): Use it; take minimal amount
2102 of insns as base for code growth.
2103 (cgraph_decide_inlining): Make initial_insns local; do not compute
2105 * params.def (PARAM_INLINE_UNIT_GROWTH): Set to 60.
2106 * doc/invoke.texi (inline-unit-growth): Update docs.
2108 2007-01-24 Jakub Jelinek <jakub@redhat.com>
2110 * config/i386/i386.h (x86_cmpxchg16b): Remove const.
2111 (TARGET_CMPXCHG16B): Define to x86_cmpxchg16b.
2112 * config/i386/i386.c (x86_cmpxchg16b): Remove const.
2113 (override_options): Add PTA_CX16 flag. Set x86_cmpxchg16b
2114 for CPUs that have PTA_CX16 set.
2117 * gimplify.c (omp_check_private): New function.
2118 (gimplify_scan_omp_clauses): Use it for
2119 firstprivate/lastprivate/reduction.
2122 * gimplify.c (omp_add_variable): Don't call omp_notice_variable
2123 on TYPE_SIZE_UNIT for GOVD_LOCAL VLAs.
2126 * omp-low.c (lower_omp_for_lastprivate): Add dlist argument.
2127 If lower_lastprivate_clauses emits some statements, append them
2128 to dlist rather than body_p and to body_p append an initializer.
2129 (lower_omp_for): Adjust caller.
2131 2007-01-24 Steve Ellcey <sje@cup.hp.com>
2133 * target.h (globalize_decl_name): New.
2134 * target-def.h (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
2135 * output.h (default_globalize_decl_name): New.
2136 * varasm.c (asm_output_bss): Use globalize_decl_name instead of
2138 (globalize_decl): Ditto.
2139 (default_globalize_decl_name): New.
2140 * config/ia64/ia64.c (ia64_globalize_decl_name): New.
2141 (ia64_handle_version_id_attribute): New.
2142 (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
2143 (ia64_asm_output_external): Use globalize_decl_name instead
2145 * doc/extend.texi (version_id): New pragma.
2146 * doc/tm.texi (ARGET_ASM_GLOBALIZE_DECL_NAME): New target hook.
2148 2007-01-24 Andreas Krebbel <krebbel1@de.ibm.com>
2150 * unwind-dw2-fde.c (get_cie_encoding): Replaced _Unwind_Word with
2151 _uleb128_t and _Unwind_SWord with _sleb128_t.
2152 * unwind-dw2.c (extract_cie_info, execute_stack_op, execute_cfa_program,
2153 uw_frame_state_for, uw_update_context_1): Likewise.
2154 * unwind-c.c (parse_lsda_header, PERSONALITY_FUNCTION): Likewise.
2155 * unwind-pe.h (read_uleb128, read_sleb128,
2156 read_encoded_value_with_base): Likewise.
2157 * unwind-generic.h: Define _sleb128_t and _uleb128_t types.
2159 2007-01-24 Richard Guenther <rguenther@suse.de>
2161 * builtins.c (expand_builtin_cexpi): Get the fndecl
2162 for cexp in the correct way.
2164 2007-01-24 Jan Hubicka <jh@suse.cz>
2166 * tree-ssa-dce.c (eliminate_unnecesary_stmts): Remove dead LHS of calls.
2168 2007-01-24 Andreas Krebbel <krebbel1@de.ibm.com>
2170 * c-cppbuiltin.c (builtin_define_type_sizeof): New function.
2171 (c_cpp_builtins): New builtin macros: __SIZEOF_INT__, __SIZEOF_LONG__,
2172 __SIZEOF_LONG_LONG__, __SIZEOF_SHORT__, __SIZEOF_POINTER__,
2173 __SIZEOF_FLOAT__, __SIZEOF_DOUBLE__, __SIZEOF_LONG_DOUBLE__,
2174 __SIZEOF_SIZE_T__, __SIZEOF_WCHAR_T__, __SIZEOF_WINT_T__ and
2175 __SIZEOF_PTRDIFF_T__.
2176 * doc/cpp.texi: Documentation for the new builtin macros added.
2178 2007-01-24 Uros Bizjak <ubizjak@gmail.com>
2180 * config/i386/i386.md (tanxf2, tan<mode>2, atan<mode>2, log<mode>2,
2181 log10<mode>2, log2<mode>2, expxf2, exp10xf2, exp2xf2): Use op2
2182 instead of operands[2] to avoid access past the end of array.
2184 2007-01-24 Richard Sandiford <richard@codesourcery.com>
2186 * reload1.c (emit_reload_insns): Pass the reload register
2187 for a non-spill output reload through forget_old_reloads_1.
2189 2007-01-23 Joseph Myers <joseph@codesourcery.com>
2191 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Define
2192 _SOFT_DOUBLE if doubles use software floating-point.
2193 * config/rs6000/libgcc-ppc-glibc.ver: Export additional long
2194 double functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
2195 * config/rs6000/darwin-ldouble.c: Also compile functions for
2196 hard-float without FPRs. Use fmsub function for all __NO_FPRS__
2197 cases. Compile extra functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
2198 * config/rs6000/linuxspe.h (SUBSUBTARGET_OVERRIDE_OPTIONS): Remove
2199 commented-out long double override.
2200 (CPP_LONGDOUBLE_DEFAULT_SPEC): Likewise.
2201 * config/rs6000/eabispe.h: Likewise.
2202 * config/rs6000/rs6000.c (rs6000_override_options): Don't override
2203 long double for non-SPE.
2204 (rs6000_handle_option): Likewise.
2205 (invalid_e500_subreg): Disallow more subregs involding DImode,
2206 DFmode, TImode or TFmode.
2207 (rs6000_legitimate_offset_address_p): Check TFmode offsets for
2209 (legitimate_lo_sum_address_p): Also check for TFmode for E500
2211 (rs6000_legitimize_address): Also handle TFmode for E500 double.
2212 (rs6000_legitimize_reload_address): Also handle TFmode for E500
2214 (rs6000_legitimate_address): Also check for TFmode for E500
2216 (rs6000_emit_move): Use DFmode subregs of TFmode for E500 double.
2217 (spe_build_register_parallel): Handle TFmode and TCmode.
2218 (rs6000_spe_function_arg): Handle TFmode and TCmode for E500
2220 (function_arg): Handle TFmode and TCmode for E500 double.
2221 (rs6000_init_libfuncs): Initialize extra libfuncs for soft double
2223 (print_operand): Handle TFmode and TImode for %y.
2224 (rs6000_generate_compare): Handle TFmode comparisons for E500
2226 (spe_func_has_64bit_regs_p): Check for TFmode for E500 double.
2227 (rs6000_function_value): Handle TFmode and TCmode for E500 double.
2228 (rs6000_libcall_value): Handle TFmode and TCmode for E500 double.
2229 * config/rs6000/rs6000.h (CANNOT_CHANGE_MODE_CLASS): Check for
2230 TFmode for E500 double.
2231 * config/rs6000/rs6000.md (FP): Allow TF for E500 double.
2232 (floatsidf2): Enable for E500 double.
2233 (movtf_softfloat): Use rs6000_nonimmediate_operand.
2234 (extenddftf2): Change to extenddftf2_fprs.
2235 (extenddftf2): Call gen_spe_extenddftf2 or gen_extenddftf2_fprs
2236 depending on TARGET_E500_DOUBLE.
2237 (extendsftf2): Enable for E500 double.
2238 (trunctfdf2): Enable for E500 double.
2239 (trunctfsf2): Change to trunctfsf2_fprs.
2240 (trunctfsf2): Call gen_spe_trunctfsf2 or gen_trunctfsf2_fprs
2241 depending on TARGET_E500_DOUBLE.
2242 (floatsitf2): Enable for E500 double.
2243 (fix_trunctfsi2): Change to fix_trunctfsi2_fprs.
2244 (fix_trunctfsi2): Call gen_spe_fix_trunctfsi2 or
2245 gen_fix_trunctfsi2_fprs depending on TARGET_E500_DOUBLE.
2246 (negtf2): Change to negtf2_internal.
2247 (negtf2): New expander.
2248 (abstf2): Enable for E500 double. Call gen_spe_abstf2_tst,
2249 gen_spe_abstf2_cmp or gen_abstf2_internal depending on
2250 TARGET_E500_DOUBLE and flag_unsafe_math_optimizations.
2251 (movdi_internal32): Use rs6000_nonimmediate_operand.
2252 (unnamed splitter): Likewise.
2253 * config/rs6000/spe.md (CMPTFEQ_GPR, TSTTFEQ_GPR, CMPTFGT_GPR,
2254 TSTTFGT_GPR, CMPTFLT_GPR, TSTTFLT_GPR): New unspecs.
2255 (SPE64TF, DITI): New mode macros.
2256 (frob_df_di): Change to frob_<SPE64:mode>_<DITI:mode>; allow more
2259 (frob_<mode>_di_2): New.
2260 (frob_tf_di_8_2): New.
2261 (frob_di_df): Change to frob_di_<mode>; allow more modes.
2263 (frob_di_df_2): Change to frob_<DITI:mode>_<SPE64:mode>_2; allow
2265 (frob_ti_<mode>_8_2): New.
2266 (frob_ti_tf_2): New.
2267 (mov_si<mode>_e500_subreg0, mov_si<mode>_e500_subreg0_2,
2268 mov_si<mode>_e500_subreg4, mov_si<mode>_e500_subreg4_2): Allow
2270 (mov_sitf_e500_subreg8, mov_sitf_e500_subreg8_2,
2271 mov_sitf_e500_subreg12, mov_sitf_e500_subreg12_2): New.
2272 (spe_trunctfdf2_internal1, spe_trunctfsf2, spe_extenddftf2,
2273 spe_fix_trunctfsi2, spe_fix_trunctfsi2_internal,
2274 spe_negtf2_internal, spe_abstf2_cmp, spe_abstf2_tst): New.
2275 (cmptfeq_gpr, tsttfeq_gpr, cmptfgt_gpr, tsttfgt_gpr, cmptflt_gpr,
2278 2007-01-23 Ian Lance Taylor <iant@google.com>
2280 * Makefile.in (OBJS-common): Reformat, alphabetize, but put
2282 (OBJS-archive): Reformat, alphabetize.
2283 (OBJS): Change out_object_file to OBJS-md.
2285 2007-01-23 Uros Bizjak <ubizjak@gmail.com>
2287 * config/i386/i386.md: Use REG_P, MEM_P, CONST_INT_P, LABEL_P,
2288 JUMP_P and CALL_P predicates where applicable.
2289 * config/i386/i386.c: Ditto.
2290 * config/i386/i386.md: Ditto.
2291 * config/i386/mmx.md: Ditto.
2292 * config/i386/predicates.md: Ditto.
2294 2007-01-22 Andreas Schwab <schwab@suse.de>
2296 * config/m68k/m68k.h: Fix comment.
2298 2007-01-22 Jan Hubicka <jh@suse.cz>
2300 * passes.c (init_optimization_passes): Do not rerun
2301 pass_early_warn_uninitialized.
2303 2007-01-22 Richard Guenther <rguenther@suse.de>
2305 PR tree-optimization/30038
2306 * tree-ssa-math-opts.c (maybe_record_sincos): New static helper
2308 (execute_cse_sincos_1): Likewise.
2309 (execute_cse_sincos): Likewise.
2310 (gate_cse_sincos): Likewise.
2311 (pass_cse_sincos): New pass CSEing sin() and cos() calls using
2312 the cexpi() canonicalization of sincos().
2313 * tree-pass.h (pass_cse_sincos): Declare.
2314 * passes.c (init_optimization_passes): New pass pas_cse_sincos.
2316 2007-01-21 Eric Botcazou <ebotcazou@libertysurf.fr>
2318 PR rtl-optimization/29329
2319 * combine.c (replaced_rhs_insn): Rename to i2mod.
2320 (replaced_rhs_value): Rename to i2mod_new_rhs.
2321 (i2mod_old_rhs): New global variable.
2322 (combine_instructions): Adjust for above change. Save a copy of
2323 the old RHS into i2mod_old_rhs when the contents of a REG_EQUAL
2324 note are substituted in the second instruction.
2325 (distribute_notes) <REG_DEAD>: Adjust for above change. Do not
2326 ditch the note if it pertains to the second eliminated register
2327 and this register is mentioned in i2mod_old_rhs.
2330 2006-09-12 Eric Botcazou <ebotcazou@libertysurf.fr>
2332 * combine.c (distribute_notes) <REG_DEAD>: Do not consider SETs past
2333 the insn to which the note was originally attached.
2335 2007-01-21 Jan Hubicka <jh@suse.cz>
2337 * ipa-inline.c (inlining_mode): Comment, move up.
2338 (cgraph_decide_inlining_incrementally): Do not perform inlining
2339 itself; fix handling of flattening of self recursive functions.
2340 (cgraph_find_cycles): Remove.
2341 (cgraph_flatten_node): Remove.
2342 (cgraph_decide_inlining): Use incremental inliner to handle flattening.
2343 (try_inline): New function.
2344 (cgraph_early_inlining): Update call of
2345 cgraph_decide_inlining_incrementally. Apply inlining here.
2346 (apply_inline): Update call of cgraph_decide_inlining_incrementally.
2348 2007-01-21 Dirk Mueller <dmueller@suse.de>
2351 * tree-vrp.c (check_array_bounds): do not warn
2352 about ADDR_EXPR's of ARRAY_REF's which are immediately
2353 used in binary expressions.
2355 2007-01-21 Ira Rosen <irar@il.ibm.com>
2357 * tree-vectorizer.h (struct _stmt_vec_info): Add new field
2358 read_write_dep and macros for its access.
2359 * tree-vectorizer.c (new_stmt_vec_info): Initialize the new field.
2360 * tree-vect-analyze.c (vect_analyze_data_ref_dependence): Remove
2361 argument, call vect_check_interleaving for every independent pair of
2362 data-refs. Mark loads that access the same memory location as a store
2364 (vect_check_dependences): Remove.
2365 (vect_analyze_data_ref_dependences): Remove vect_check_dependences
2366 call, fix the call to vect_analyze_data_ref_dependence.
2367 (vect_analyze_data_ref_access): For statements that access the same
2368 data-ref, check that they are not stores; for loads, check that there
2369 is no store that access the same location.
2371 2007-01-20 Roger Sayle <roger@eyesopen.com>
2372 Joseph Myers <joseph@codesourcery.com>
2374 * doc/invoke.texi (-fdump-rtl-combine): Fix under/overfull hbox.
2375 (-fdump-rtl-stack): Likewise.
2376 (-fno-signed-zeros): Use @minus{} for a minus sign. Correct typo.
2377 (-mcheck-zero-division, -mcpu): Fix under/overfull hbox.
2378 (-mpt-fixed): Use @minus{} for minus sign.
2379 (Using Precompiled Headers): Fix under/overfull hbox.
2381 2007-01-20 Jan Hubicka <jh@suse.cz>
2383 * tree-flow.h (struct stmt_ann_d): Move references_memory to proper
2384 place within annotation.
2386 2007-01-20 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
2388 * pa.c (output_move_double): Change array size of xoperands to 4.
2390 2007-01-20 Jan Hubicka <jh@suse.cz>
2392 * tree-tailcall.c (adjust_return_value): Do not use RESULT_DECL
2395 2007-01-19 Ian Lance Taylor <iant@google.com>
2397 * expmed.c (expand_divmod) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]: Cast
2398 constant to unsigned HOST_WIDE_INT before negating.
2400 2007-01-19 Ian Lance Taylor <iant@google.com>
2402 * tree-ssa-operands.h (struct vuse_vec_d): Change num_vuse field
2404 (VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Use unsigned comparison.
2405 (VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
2406 (SET_VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Likewise.
2407 (SET_VUSE_ELEMENT_VAR) [ENABLE_CHECKING]: Likewise.
2408 (SET_VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
2409 (realloc_vdef, realloc_vuse): Change second parameter to
2411 (ssa_operand_iterator_d): Change vuse_index and mayuse_index
2413 * tree-ssa-operands.c (realloc_vop): Change num_elem parameter to
2414 unsigned. Change x and lim locals to unsigned.
2415 (realloc_vdef, realloc_vuse): Change num_elem parameter to
2417 (finalize_ssa_vuse_ops): Change old_i local to unsigned.
2418 (copy_virtual_operands): Change i and n locals to unsigned.
2420 2007-01-19 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
2423 * builtins.c (fold_builtin_1): Handle builtin fdim.
2425 2007-01-20 Jan Hubicka <jh@suse.cz>
2427 * tree-ssa.c (init_tree_ssa): Do not call init_alias_heapvars.
2428 * tree-ssa-structalias.c (compute_points_to_sets): Do call
2429 init_alias_heapvars.
2430 (init_alias_heapvars): Initialize only when not already initialized.
2431 (delete_alias_heapvars): Set heapvar_for_stmt to NULL.
2433 2007-01-19 Roger Sayle <roger@eyesopen.com>
2435 * common.opt (fsigned-zeros): New command line option.
2436 * flags.h (HONOR_SIGNED_ZEROS): Control via flag_signed_zeros instead
2437 of flag_unsafe_math_optimizations.
2438 * opts.c (set_fast_math_flags): The -ffast-math command line option
2439 implies -fno-signed-zeros.
2440 (fast_math_flags_set_p): Likewise.
2442 * doc/invoke.texi: Document new -fno-signed-zeros option, and update
2443 the documentation of -ffast-math appropriately. Wrap long lines.
2445 2007-01-19 Steve Ellcey <sje@cup.hp.com>
2447 * system.h (ASM_MAKE_LABEL_LINKONCE): Poison.
2448 * varasm.c (globalize_decl): Remove ASM_MAKE_LABEL_LINKONCE ifdef.
2450 2007-01-19 Tomas Bily <tbily@suse.cz>
2452 * cgraphunit.c (cgraph_finalize_function): Updating of pid
2454 (tree_init_ic_make_global_vars): New function
2455 (tree_init_edge_profiler): call of tree_init_ic_make_global_vars
2456 (tree_gen_ic_profiler): New function
2457 (tree_gen_ic_func_profiler): New function
2458 (tree_profiling): Added calling of tree_gen_ic_func_profiler
2459 (tree_profile_hooks): Added hook for indirec/virtual calls
2460 * value-prof.c (tree_find_values_to_profile): New case for
2462 (tree_values_to_profile): Call for determining indirect/virtual
2464 (tree_indirect_call_to_profile): New function
2465 (tree_ic_transform): New function
2466 (tree_ic): New function
2467 (find_func_by_pid): New function
2468 (init_pid_map): New function
2469 (tree_value_profile_transformations): Added check for
2470 indirect/virtual call transformation
2471 * value-prof.h (enum hist_type): New counter type for
2472 indirect/virtual calls
2473 (profile_hooks): Added new hook for profiling indirect/virtual
2475 * profile.c (instrument_values): New case for indirect/virtual
2477 * gcov-io.h (GCOV_LAST_VALUE_COUNTER): Changed to 6
2478 (GCOV_COUNTER_V_INDIR): New counter type
2479 (GCOV_COUNTER_NAMES): New name of counter "indirect" added
2480 (GCOV_MERGE_FUNCTIONS): New merge function for indirect/virtual
2482 * cgraph.c: Definition of cgraph_max_pid
2483 (cgraph_create_node): Default init of pid attribute
2484 * cgraph.h: Declaration of cgraph_max_pid
2485 (struct cgraph_node): Added pid attribute
2486 * libgcov.c (__gcov_indirect_call_profiler): New function
2487 (__gcov_one_value_profiler_body): New function
2488 (__gcov_one_value_profiler): Body was moved to
2489 __gcov_one_value_profiler_body and calls it
2491 2007-01-19 Basile Starynkevitch <basile@starynkevitch.net>
2493 * doc/gty.texi (Options): Document the mark_hook option to GTY.
2494 * gengtype.c (write_types_data, write_func_for_structure,
2495 write_types, ggc_wtd, pch_wtd): Add skip_hooks to
2496 write_types_data, ggc_wtd, pch_wtd for processing mark_hook.
2497 (walk_type, write_func_for_structure): Generate the mark_hook if
2500 2007-01-19 Jan Hubicka <jh@suse.cz>
2502 * ipa-inline.c (cgraph_decide_inlining_incrementally): Instead of
2503 'early' argument take inlining mode argument specifying whether to
2504 inline for size/speeed or all functions; add support for flattening;
2506 (cgraph_early_inlining): Update call of decide_inlining_incrementally.
2508 2007-01-19 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
2511 * toplev.c (warn_deprecated_use): Use %qD instead of %qs to print
2512 the name of the declared identifier.
2514 2007-01-19 Dirk Mueller <dmueller@suse.de>
2516 * config/i386.h (CONDITIONAL_REGISTER_USAGE): Store
2517 result of PIC_OFFSET_TABLE_REGNUM in temporary variable to avoid
2518 duplicate evaluation.
2520 2007-01-19 Uros Bizjak <ubizjak@gmail.com>
2522 * config/i386/i386.md (acos<mode>2): Rename from acossf2 and acosdf2.
2523 Macroize expander using X87MODEF12 mode macro. Extend operand 1
2524 to XFMode, use acosxf2 and truncate result to requested mode.
2525 Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
2526 (asin<mode>2): Similarly, with asin expanders.
2527 (*fscalexf4_i387): Rename from *fscalexf4.
2528 (expNcorexf3): New expander.
2529 (expxf2, exp10xf2, exp2xf2): Use expNcorexf3 expander.
2530 (exp<mode>2): Rename from expsf2 and expdf2. Macroize expander using
2531 X87MODEF12 mode macro. Extend operand 1 to XFMode, use expxf2 and
2532 truncate result to requested mode. Use SSE_FLOAT_MODE_P to disable
2533 patterns for SSE math.
2534 (exp10<mode>2): Similarly, with exp10 expanders.
2535 (exp2<mode>2): Similarly, with exp2 expanders.
2536 (expm1<mode>2): Similarly, with expm1 expanders.
2537 (ldexp<mode>3): Similarly, with ldexp expanders.
2538 (log<mode>2, log10<mode>2, log2<mode>2, log1p<mode>2, logb<mode>2):
2539 Use gen_truncxf<mode>2_i387_noop to truncate result.
2541 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2543 * config/m68k/m68k.h (PREFERRED_STACK_BOUNDARY): Define to 32
2544 for ColdFire targets.
2546 2007-01-19 Nathan Sidwell <nathan@codesourcery.com>
2547 Richard Sandiford <richard@codesourcery.com>
2549 * config/m68k/m68k.h (M68K_STATIC_CHAIN_REG_NAME): New macro.
2550 (INITIALIZE_TRAMPOLINE): Use STATIC_CHAIN_REGNUM.
2551 (__transfer_from_trampoline): Use M68K_STATIC_CHAIN_REG_NAME.
2552 * config/m68k/m68kelf.h (STATIC_CHAIN_REGNUM): Override.
2553 (M68K_STATIC_CHAIN_REG_NAME): Likewise.
2554 * config/m68k/netbsd-elf.h (M68K_STATIC_CHAIN_REG_NAME): Likewise.
2556 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2558 * config/m68k/m68k.md (adddi_dilshr32): Rename to...
2559 (*adddi_dilshr32): ...this. Fix formatting. Remove commented-out
2560 non-canonical pattern. Restrict to !TARGET_COLDFIRE.
2561 (*adddi_dilshr32_cf): New pattern.
2562 (adddi3, subdi3): Remove first alternatives.
2564 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2566 * config/m68k/m68k.c (notice_update_cc): If an SFmode move is
2567 implemented using move.l, do not use its cc result for floating-point
2570 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2572 * config/m68k/m68k.h (EXTRA_CONSTRAINT): Stop the 'T' constraint
2573 from accepting 's' constraints if flag_pic.
2575 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2577 * config/m68k/m68k.md (bordered, bunordered, buneq, bunge, bungt)
2578 (bunle, bunlt, bltgt, bordered_rev, bunordered_rev, buneq_rev)
2579 (bunge_rev, bungt_rev, bunle_rev, bunlt_rev, bltgt_rev): Change
2580 condition from TARGET_68881 to TARGET_HARD_FLOAT.
2582 2007-01-19 Sandra Loosemore <sandra@codesourcery.com>
2584 * longlong.h (count_leading_zeros, COUNT_LEADING_ZEROS_0): Add
2585 ColdFire alternatives.
2586 * config/m68k/m68k.h (CLZ_DEFINED_VALUE_AT_ZERO): New macro.
2587 * config/m68k/m68k.md (clzsi2): Define for ColdFire
2588 architectures that support ff1 instruction.
2590 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2591 Julian Brown <julian@codesourcery.com>
2593 * config/m68k/m68k.h (CONST_OK_FOR_LETTER_P): Add an 'R' case.
2594 * config/m68k/m68k.md (*movsi_cfv4): Fold into...
2595 (*movsi_cf): ...here. Remove unnecessary 'R' from 'Rg'.
2598 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2600 * config/m68k/m68k-protos.h (valid_mov3q_const): Take a HOST_WIDE_INT
2602 (output_move_const_into_data_reg, output_move_simode_const): Delete.
2603 * config/m68k/m68k.c (const_method, const_int_cost): Take a
2604 HOST_WIDE_INT instead of an rtx.
2605 (m68k_rtx_costs): Update call accordingly.
2606 (output_move_const_into_data_reg): Likewise. Fix formatting.
2607 (valid_mov3q_const): Take a HOST_WIDE_INT instead of an rtx.
2609 (output_move_simode_const): Update calls after above changes.
2610 Rework to use automatic variables and predicates like MEM_P.
2611 * config/m68k/m68k.md (pushexthisi_const): Update call to
2614 2007-01-19 Dirk Mueller <dmueller@suse.de>
2616 * tree-ssa-alias.c (perform_var_substitution): Fix typo
2619 2007-01-19 Richard Guenther <rguenther@suse.de>
2621 * builtins.c (expand_builtin_cexpi): Fall back to expanding
2622 via cexp in case sincos is not available.
2624 2007-01-19 Richard Guenther <rguenther@suse.de>
2626 * doc/tm.texi (TARGET_HAS_SINCOS): Document new target macro.
2627 * defaults.h (TARGET_HAS_SINCOS): Default to off.
2628 * config/linux.h (TARGET_HAS_SINCOS): Set to on if we have glibc.
2629 * config/alpha/linux.h (TARGET_HAS_SINCOS): Likewise.
2630 * config/sparc/linux.h (TARGET_HAS_SINCOS): Likewise.
2631 * config/sparc/linux64.h (TARGET_HAS_SINCOS): Likewise.
2632 * config/rs6000/linux.h (TARGET_HAS_SINCOS): Likewise.
2633 * config/rs6000/linux64.h (TARGET_HAS_SINCOS): Likewise.
2635 2007-01-19 Uros Bizjak <ubizjak@gmail.com>
2637 * config/i386/i386.md (*fpatanxf3_i387, fpatan_extend<mode>xf3_i387):
2639 (atan2sf3_1, atan2df3_1, atan2xf3_1): Remove insn patterns.
2640 (atan2xf3): Directly generate RTL pattern.
2641 (atan2<mode>3): Rename from atan2sf3 and atan2df3 and macroize insn
2642 patterns using X87MODEF12 mode macro. Use fpatan_extend<mode>xf3_i387
2643 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
2644 disable patterns for SSE math.
2645 (atan<mode>2): Rename from atansf2 and atandf2 and macroize insn
2646 patterns using X87MODEF12 mode macro. Use fpatan_extend<mode>xf3_i387
2647 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
2648 disable patterns for SSE math.
2650 2007-01-19 Alexandre Oliva <aoliva@redhat.com>
2652 * libgcc-std.ver: Fix typo in %inherit for GCC_4.3.0.
2654 2007-01-18 Roger Sayle <roger@eyesopen.com>
2656 * fold-const.c (fold_unary) <VIEW_CONVERT_EXPR>: Optimize away a
2657 VIEW_CONVERT_EXPR to the same type as it's operand.
2659 2007-01-18 David Edelsohn <edelsohn@gnu.org>
2661 * config/rs6000/darwin-ldouble.c: Only build _SOFT_FLOAT if
2662 configured for long double 128.
2664 2007-01-18 Mike Stump <mrs@apple.com>
2666 * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Fix build
2669 2007-01-18 Michael Meissner <michael.meissner@amd.com>
2671 * i386.c (ix86_compute_frame_layout): Make fprintf's in #if 0 code
2674 2007-01-18 Jan Hubicka <jh@suse.cz>
2676 * tree-ssa-operands.c (vop_free_bucket_size): Never return value
2677 greater than NUM_VOP_FREE_BUCKETS.
2679 2007-01-18 Daniel Berlin <dberlin@dberlin.org>
2681 * tree-ssa-structalias.c: Update comments.
2682 (ptabitmap_obstack): Removed.
2684 (oldpta_obstack): Ditto.
2685 (stats): Add a few members.
2686 (struct variable_info): Remove node, complex, address_taken, and
2687 indirect_target members. Add oldsolution member.
2688 (new_var_info): Do not initialize removed members.
2689 (constraint_expr_type): Remove INCLUDES.
2690 (constraint_graph): Add size, implicit_preds, rep,
2691 indirect_cycles, eq_rep, label, direct_nodes, and complex members.
2692 (FIRST_REF_NODE): New macro.
2693 (LAST_REF_NODE): Ditto.
2694 (FIRST_ADDR_NODE): Ditto.
2695 (find): New function.
2697 (dump_constraint): Do not handle INCLUDES.
2698 (insert_into_complex): Do not insert duplicate constraints.
2699 (condense_varmap_nodes): Renamed and rewritten into ...
2700 (merge_node_constraints): This. Also fix bug in handling of
2701 offseted copy constraints.
2702 (clear_edges_for_node): No longer need to deal with preds at all,
2703 or removing associated preds/succs.
2704 (merge_graph_nodes): Deal with indirect_cycles.
2705 Don't deal with predecessors.
2706 (add_implicit_graph_edge): New function.
2707 (add_pred_graph_edge): Ditto.
2708 (add_graph_edge): Don't deal with predecessors.
2709 (build_constraint_graph): Removed.
2710 (build_pred_graph): New function.
2711 (build_succ_graph): Ditto.
2712 (struct scc_info): Removed in_component. Added roots, dfs, and
2713 node_mapping. Remove visited_index, unification_queue.
2714 (scc_visit): Deal with union-find we do now.
2715 Deal with cycles with REF nodes.
2716 (collapse_nodes): Renamed and rewritten to ...
2717 (unify_nodes): This.
2718 (process_unification_queue): Removed.
2719 (topo_visit): Cleanup
2720 (do_da_constraint): Use find.
2721 (do_sd_constraint): Ditto.
2722 (do_ds_constraint): Ditto.
2723 (do_complex_constraint): Ditto.
2724 (init_scc_info): Update for removed and added members.
2725 (find_and_collapse_graph_cycles): Renamed and rewritten into ...
2726 (find_indirect_cycles): This.
2727 (equivalence_class): New variable.
2728 (label_visit): New function.
2729 (perform_variable_substitution): Rewritten.
2730 (free_var_substitution_info): New function.
2731 (find_equivalent_node): Ditto.
2732 (move_complex_constraints): Ditto.
2733 (eliminate_indirect_cycles): Ditto.
2734 (solve_graph): Only propagate changed bits.
2735 Use indirect cycle elimination.
2737 (tree_id_t): Rename to tree_vi_t, delete id member, add vi member.
2738 (tree_id_eq): Renamed to ...
2739 (tree_vi_eq): This. Update for member change
2740 (insert_id_for_tree): Renamed and rewritten to ...
2741 (insert_vi_for_tree): This.
2742 (lookup_id_for_tree): Renamed and rewritten to ...
2743 (lookup_vi_for_tree): This.
2744 (get_id_for_tree): Renamed and rewritten to ...
2745 (get_vi_for_tree): Ditto.
2746 (get_constraint_exp_from_ssa_var): Update to use get_vi_for_tree.
2747 (process_constraint): Don't handle INCLUDES.
2748 Remove special ADDRESSOF case.
2749 (find_func_aliases): Rewrite to use vi functions instead of id
2751 (create_function_info_for): Ditto.
2752 (create_variable_info_for): Ditto.
2753 (intra_create_variable_infos): Ditto.
2754 (merge_smts_into): Ditto.
2755 (find_what_p_points_to): Ditto.
2756 (init_base_vars): Ditto.
2757 (init_alias_vars): Ditto.
2758 (remove_preds_and_fake_succs): New function.
2759 (dump_sa_points_to_info): Dump new stats.
2760 (dump_solution_for_var): Use find.
2761 (set_used_smts): Fix formatting.
2762 (compute_points_to_sets): Updated for new functions.
2763 (ipa_pta_execute): Ditto.
2765 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
2766 Richard Sandiford <richard@codesourcery.com>
2768 * doc/tm.texi (TARGET_FUNCTION_VALUE): Expand documentation of
2770 * calls.c (expand_call): If the return value is a PARALLEL,
2771 extract its first member.
2772 * config/m68k/linux.h (FUNCTION_EXTRA_EPILOGUE): Remove.
2773 * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
2774 use FUNCTION_EXTRA_EPILOGUE.
2775 (m68k_function_value): Return a PARALLEL if the return value
2776 is of a pointer type.
2777 * config/m68k/netbsd-elf.h (current_function_returns_pointer)
2778 (FUNCTION_EXTRA_EPILOGUE): Remove.
2779 * config/m68k/m68k.md (D0_REG): New constant.
2781 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
2783 * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
2784 output a NOP for empty epilogues.
2786 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2788 * config/m68k/m68k.c (m68k_use_return_insn): Update comments
2789 before function. Extend register save check to include all
2790 registers, not just integer ones.
2792 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
2794 * config/m68k/m68k-protos.h (use_return_insn): Rename to...
2795 (m68k_use_return_insn): ...this.
2796 * config/m68k/m68k.h (USE_RETURN_INSN): Delete.
2797 * config/m68k/m68k.c (use_return_insn): Rename to...
2798 (m68k_use_return_insn): ...this.
2799 * config/m68k/m68k.md (return): Use m68k_use_return_insn instead
2802 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
2804 * config/m68k/fpgnulib.c (__truncdfsf2): Implement round to
2805 nearest even, fix denormal rounding overflow.
2807 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2809 * config/m68k/m68k.md (movsf_cf_hard): Use fsmove instead of
2810 f%$smove and f%$move.
2811 (movdf_cf_hard): Use fdmove for cases 0 and 3 and fmove for case 1.
2812 (extendsfdf2_cf): Use fdmove instead of f%&move.
2813 (truncdfsf2_cf): Use fsmove instead of f%$smove.
2814 (add<mode>3_cf, sub<mode>3_cf): Use <FP:prec> instead of <FP:round>.
2816 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2818 * config/m68k/m68k.md (movdf_cf_hard): Use output_move_double for
2821 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2823 * real.h (real_format): Add a canonical_nan_lsbs_set field.
2824 (coldfire_single_format): Declare.
2825 (coldfire_double_format): Likewise.
2826 * real.c (encode_ieee_single): Use canonical_nan_lsbs_set instead
2827 of qnan_msb_set to determine the lower bits of a canonical
2829 (encode_ieee_double): Likewise.
2830 (encode_ieee_quad): Likewise.
2831 (ieee_single_format): Initialize canonical_nan_lsbs_set.
2832 (mips_single_format): Likewise.
2833 (ieee_double_format): Likewise.
2834 (mips_double_format): Likewise.
2835 (ieee_extended_motorola_format): Likewise.
2836 (ieee_extended_intel_96_format): Likewise.
2837 (ieee_extended_intel_128_format): Likewise.
2838 (ieee_extended_intel_96_round_53_format): Likewise.
2839 (ibm_extended_format): Likewise.
2840 (mips_extended_format): Likewise.
2841 (ieee_quad_format): Likewise.
2842 (mips_quad_format): Likewise.
2843 (vax_f_format): Likewise.
2844 (vax_d_format): Likewise.
2845 (vax_g_format): Likewise.
2846 (i370_single_format): Likewise.
2847 (i370_double_format): Likewise.
2848 (decimal_single_format): Likewise.
2849 (decimal_double_format): Likewise.
2850 (decimal_quad_format): Likewise.
2851 (c4x_single_format): Likewise.
2852 (c4x_extended_format): Likewise.
2853 (real_internal_format): Likewise.
2854 (coldfire_single_format): New real_format.
2855 (coldfire_double_format): Likewise.
2856 * config/pdp11/pdp11.c (pdp11_f_format): Initialize
2857 canonical_nan_lsbs_set.
2858 (pdp11_d_format): Likewise.
2859 * config/m68k/m68k.c (override_options): Override REAL_FORMAT_MODE
2860 if TARGET_COLDFIRE_CPU.
2862 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2864 * config/m68k/m68k-protos.h (m68k_output_pic_call): Delete.
2865 (output_call, m68k_legitimize_call_address): Declare.
2866 * config/m68k/m68k.h (EXTRA_CONSTRAINT): Remove unnecessary
2867 parenthesees. Add support for a 'W' constraint.
2868 (LEGITIMATE_PIC_OPERAND_P): Remove SYMBOL_REF_FLAG handling.
2869 (PRINT_OPERAND_PUNCT_VALID_P): Remove comment about 'o'.
2870 (m68k_symbolic_call, m68k_symbolic_jump): Declare.
2871 * config/m68k/m68k.c (m68k_symbolic_call, m68k_symbolic_jump): New
2873 (override_options): Initialize them. Do not set flag_no_function_cse
2874 for TARGET_ID_SHARED_LIBRARY.
2875 (m68k_output_pic_call): Delete.
2876 (m68k_legitimize_call_address): New function.
2877 (print_operand): Remove the %o prefix. Handle the %p prefix.
2878 (output_call): New function.
2879 (m68k_output_mi_thunk): Use m68k_symbolic_jump. Always load the
2880 target address from the GOT if symbolic jumps are not allowed.
2881 * config/m68k/m68k.md (call, general_operand): Do not set
2882 SYMBOL_REF_FLAG. Use m68k_legitimize_call_address instead.
2883 Merge separate flag_pic and !flag_pic define_insns into...
2884 (*call, *call_value): ...these new patterns. Match the address
2885 rather than the containing MEM and require it to be a call_operand.
2886 Use output_call to generate the asm template.
2887 * config/m68k/predicates.md (const_call_operand): New predicate.
2888 (call_operand): Likewise.
2890 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
2892 * config/m68k/m68k.h (REGISTER_MOVE_COST): Simplify definition.
2893 (STACK_GROWS_DOWNWARD): Define to 1.
2894 (FUNCTION_VALUE, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P): Equivocate
2895 comments, emphasizing that these values are only defaults.
2896 * config/m68k/linux.h (LINK_SPEC): Fix formatting in #undef.
2897 * config/m68k/m68k.c (const_method): Remove trailing whitespace.
2899 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2901 * config/m68k/m68k.md (cmpsi): Remove outdated flag_pic handling.
2903 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
2904 Richard Sandiford <richard@codesourcery.com>
2906 * config/m68k/m68k.h (DATA_REGNO_P, ADDRESS_REGNO_P, INT_REGNO_P)
2907 (FP_REGNO_P): New macros.
2908 (REGNO_OK_FOR_INDEX_P, REGNO_OK_FOR_BASE_P, REGNO_OK_FOR_DATA_P)
2909 (REGNO_OK_FOR_FP_P, REG_OK_FOR_INDEX_P, REG_OK_FOR_BASE_P): Use them.
2910 (EH_RETURN_STACKADJ_RTX): Use A0_REG.
2911 * config/m68k/m68k.c (m68k_regno_mode_ok): Use the new REGNO macros.
2913 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
2915 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
2916 (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-rtems*): Add t-floatlib
2918 * config/m68k/t-floatlib: New file, extracting common code from...
2919 * config/m68k/t-m68kbare, config/m68k/t-m68kelf,
2920 * config/m68k/t-uclinux: Here.
2921 * config/m68k/fpgnulib.c: Do not compile extendeed precision
2922 routines on ColdFire targets.
2924 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
2926 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
2927 (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*)
2928 (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-linux*)
2929 (m68k-*-rtems*): Use tm_file rather than m68k/m68k.h and
2930 explicitly set MOTOROLA to 1.
2931 * config/m68k/m68k.h (MOTOROLA): Simplify definition accordingly.
2933 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2934 Nathan Sidwell <nathan@codesourcery.com>
2936 * config/m68k/m68k.h (PCC_STATIC_STRUCT_RETURN): Delete.
2937 (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Add MOTOROLA cases.
2938 * config/m68k/coff.h (REGISTER_PREFIX_MD): Delete.
2939 * config/m68k/m68020-elf.h (LIB_SPEC): Delete.
2940 * config/m68k/m68k-none.h (CC1_SPEC, CPP_SUBTARGET_SPEC): Delete.
2941 * config/m68k/m68kelf.h (IMMEDIATE_PREFIX, REGISTER_PREFIX_MD)
2942 (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Delete.
2943 (NO_DOLLAR_IN_LABEL, PCC_STATIC_STRUCT_RETURN): Don't undefine.
2944 (BSS_ASM_OP, ASM_OUTPUT_SKIP, ASM_OUTPUT_ASCII): Delete.
2945 * config/m68k/m68kemb.h (PCC_STATIC_STRUCT_RETURN): Don't undefine.
2946 * config/m68k/linux.h (SIZE_TYPE, PTRDIFF_TYPE, WCHAR_TYPE)
2947 (WCHAR_TYPE_SIZE, TARGET_OBJFMT_CPP_BUILTINS): Delete.
2948 (TARGET_OS_CPP_BUILTINS): Don't define mc68000 and mc68020 here.
2949 (DBX_REGISTER_NUMBER): Delete.
2950 * config/m68k/netbsd-elf.h (IMMEDIATE_PREFIX): Delete.
2951 (PCC_STATIC_STRUCT_RETURN): Don't undefine.
2952 * config/m68k/openbsd.h (PCC_STATIC_STRUCT_RETURN): Define.
2954 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2956 * config.gcc (m68k-*-uclinux*): Add flat.h to $tm_file.
2957 * config/flat.h: New file.
2958 * crtstuff.c (USE_PT_GNU_EH_FRAME): Don't define if
2960 * config/m68k/m68k.h (ASM_PREFERRED_EH_DATA_FORMAT): Do not use
2961 indirect references for -msep-data or -mid-shared-library.
2962 Do not use PC-relative code addresses either.
2963 * config/m68k/m68k.c (override_options): Restrict -fPIC error
2965 * config/m68k/uclinux.h (STARTFILE_SPEC): Define. Use Scrt1.o
2966 for shared libraries and crt1.o for executables. Use crti.o and
2968 (ENDFILE_SPEC): Use crtend.o and crtn.o.
2969 (LIB_SPEC): Suppress -Rlibc.gdb if -static-libc is given.
2970 Do not add -elf2flt or -shared-lib-id options here.
2971 (LINK_SPEC): Define. Pass -elf2flt if no -elf2flt option is given.
2972 Pass -shared-lib-id if -mid-shared-library, taking the library
2973 identifier from -mshared-library-id if given, otherwise
2975 (EH_FRAME_IN_DATA_SECTION): Do not undefine.
2976 (INIT_SECTION_ASM_OP, FINI_SECTION_ASM_OP): Likewise.
2977 (TARGET_OS_CPP_BUILTINS): Define __GXX_MERGED_TYPEINFO_NAMES=0
2978 and __GXX_TYPEINFO_EQUALITY_INLINE=0 if -mid-shared-library.
2979 (DRIVER_SELF_SPECS): Map unadorned PIC options to -msep-data.
2980 * config/m68k/t-uclinux (EXTRA_MULTILIB_PARTS): Add crtbegin.o
2982 * config/m68k/lb1sf68.asm (PICCALL): Use an lea and pc-relative
2983 jump sequence for ISA A and ISA A+.
2984 (PICJUMP): Likewise.
2986 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2988 * config.gcc (m68k-*-uclinux*): Base the port on the common
2989 and m68k GNU/Linux files rather than on the generic ELF ones.
2990 * config/m68k/uclinux.h (TARGET_VERSION): Override.
2991 (TARGET_OS_CPP_BUILTINS): Use LINUX_TARGET_OS_CPP_BUILTINS.
2993 2007-01-18 Julian Brown <julian@codesourcery.com>
2994 Richard Sandiford <richard@codesourcery.com>
2996 * config/m68k/m68k.h (LONG_DOUBLE_TYPE_SIZE): Make 64-bit on ColdFire.
2997 (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
2998 * config/m68k/netbsd-elf.h (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Undefine
3000 * config/m68k/uclinux-oldabi.h (LONG_DOUBLE_TYPE_SIZE): Redefine to
3002 (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
3004 2007-01-18 Richard Sandiford <richard@codesourcery.com>
3006 * doc/install.texi: Document m68k-uclinuxoldabi.
3007 * config.gcc (m68k-*-uclinuxoldabi): New configuration.
3008 * config/m68k/uclinux-oldabi.h: New file, copied from
3009 config/m68k/uclinux.h.
3011 2007-01-18 Kaz Kojima <kkojima@rr.iij4u.or.jp>
3013 * config/m32r/m32r-protos.h (m32r_expand_epilogue): Declare it.
3014 * config/m32r/m32r.c (m32r_setup_incoming_varargs): Use gen_frame_mem.
3015 (m32r_compute_frame_size): Use unsigned for regno.
3016 (m32r_reload_lr): Use gen_frame_mem.
3018 (m32r_output_function_epilogue): Don't output the function epilogue
3020 (m32r_expand_epilogue): New.
3021 (direct_return): Return false if the function has the interrupt
3023 (m32r_hard_regno_rename_ok): Remove code for the textual epilogue.
3024 * config/m32r/m32r.md (epilogue): New expander.
3025 (return_lr, return_rte): New insns.
3026 (return): Make it expander.
3027 (return_normal): New expander.
3029 2007-01-18 Josh Conner <jconner@apple.com>
3032 * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Add
3033 support for UNLE, UNLT, UNGE, and UNGT.
3035 2007-01-18 Jan Hubicka <jh@suse.cz>
3037 * tree-vrp.c (finalize_jump_threads): Do not call cleanup_cfg by hand.
3038 * tree-tailcall (add_virtual_phis): Likewise.
3039 (optimize_tail_call): Return TODOs.
3040 (execute_tail_calls): Return TODOs.
3041 * tree-ssa-ccp (execute_fold_all_builtins): Do cleanup_cfg via TODO.
3042 * tree-cfgcleanup.c (cleanup_tree_cfg_loop): Return if something
3044 * tree-ssa-forwprop.c (tree_ssa_forward_propagate_single_use_value):
3045 Cleanup cfg using TODO.
3046 * tree-flow.h (cleanup_tree_cfg_loop): Update prototype.
3047 * passes.c (execute_function_todo): When cleanup did something, remove
3049 * tree-cfg.c (pass_build_cfg): Add cleanup_cfg TODO.
3050 (make_edges): Don't cleanup_cfg.
3052 2007-01-18 Uros Bizjak <ubizjak@gmail.com>
3054 * reg-stack.c (subst_stack_regs_pat) [UNSPEC_SINCOS_COS,
3055 UNSPEC_XTRACT_FRACT]: Use generic code for instructions that
3056 operate on the top of stack.
3057 [UNSPEC_SINCOS_SIN, UNSPEC_XTRACT_EXP, UNSPEC_TAN]: Rewrite
3058 register handling of instructions that output to the second
3060 [UNSPEC_TAN_ONE, UNSPEC_TAN_TAN]: Remove.
3061 (move_for_stack_reg): Special-case check for dead destination
3062 stack slot for constant load of 1.0 inside UNSPEC_TAN.
3064 * config/i386/i386.md (UNSPEC_TAN): New constant.
3065 (UNSPEC_TAN_ONE, UNSPEC_TAN_TAN): Remove.
3066 (fptanxf4_i387, fptan_extend<mode>xf4_i387): New patterns
3067 to correctly model move of constant 1.0 to top stack slot.
3068 (*tandf3_1, *tansf3_1, *tanxf3_1): Remove insn patterns.
3069 (unnamed peephole2 pattern): Remove corresponding peephole2
3070 pattern that optimizes tan insn and loading of constant 1.0.
3071 (tanxf2): Use fptanxf4_i387.
3072 (tan<mode>2): Rename from tansf2 and tandf2 and macroize insn
3073 patterns using X87MODEF12 mode macro. Use fptan_extend<mode>xf4_i387
3074 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
3075 disable patterns for SSE math.
3076 (sincos<mode>3): Use truncxf<mode>2_i387_noop for truncation.
3077 (fyl2x_extend<mode>xf3_i387): Use X87MODEF12 for operand 1.
3079 2007-01-18 Dirk Mueller <dmueller@suse.de>
3080 Richard Guenther <rguenther@suse.de>
3083 * doc/invoke.texi (Warray-bounds): Document -Warray-bounds.
3084 * common.opt (Warray-bounds): Add new warning option.
3085 * c-opts.c (c_common_handle_option): Define -Warray-bounds
3087 * Makefile.in: make tree-vrp.o depend on toplev.h
3088 * tree-vrp.c (vrp_finalize): Call check_array_refs if -Warray-bounds
3090 (check_array_refs, check_array_bounds, check_array_ref): New.
3092 2007-01-18 Jan Hubicka <jh@suse.cz>
3094 * tree-ssa-ccp.c (ccp_finalize): Return if something changed.
3095 (execute_ssa_ccp): Return flags conditionally.
3096 * tree-ssa-propagate.c (substitue_and_fold): Return if something was
3098 * tree-ssa-propagate.h (substitute_and_fold): Update prototype.
3100 2007-01-18 Steven Bosscher <steven@gcc.gnu.org>
3102 * cfgcleanup.c (cleanup_cfg): Detect cfglayout mode and set
3103 the CLEANUP_CFGLAYOUT flag when in cfglayout mode.
3105 * Makefile.c (GTFILES): Add cfglayout.h.
3106 * gengtype.c (open_base_files): Likewise.
3107 * cfglayout.c (cfg_layout_function_footer,
3108 cfg_layout_function_header) Reindent to make gengtype happy.
3109 * cfglayout.h (cfg_layout_function_footer,
3110 cfg_layout_function_header): Add GTY(()) marker.
3112 * ifcvt.c (noce_try_sign_mask): Make sure INSN_B is non-null.
3114 2007-01-18 Ben Elliston <bje@au.ibm.com>
3116 * genautomata.c (write_automata): Include xstrerror output in the
3117 error message if writing the DFA description file fails.
3119 2007-01-17 H.J. Lu <hongjiu.lu@intel.com>
3121 * config/mips/mips-protos.h (mips_output_external): Make it
3123 * config/mips/iris.h (TARGET_ASM_EXTERNAL_LIBCALL): Removed.
3124 * config/mips/mips.c (irix_output_external_libcall): Likewise.
3125 (extern_list): Likewise.
3126 (extern_head): Likewise.
3127 (TARGET_ASM_FILE_END): Likewise.
3128 (mips_file_end): Likewise.
3129 (mips_output_external): Rewritten.
3131 2007-01-18 Ben Elliston <bje@au.ibm.com>
3133 * genpreds.c (write_insn_preds_c): Only write out the function
3134 body for regclass_for_constraint if we have register constraints.
3136 2007-01-17 Tom Tromey <tromey@redhat.com>
3138 * doc/sourcebuild.texi (libgcj Tests): Use sourceware.org.
3139 * doc/install.texi (Testing): Use sourceware.org.
3140 (Binaries): Likewise.
3141 (Specific): Likewise.
3142 * doc/contrib.texi (Contributors): Use sourceware.org.
3144 2007-01-17 Anatoly Sokolov <aesok@post.ru>
3146 * config/avr/avr.h (AVR_HAVE_LPMX): New macro.
3147 (AVR_ENHANCED): Rename to ...
3148 (AVR_HAVE_MUL): ... new.
3149 (avr_enhanced_p): Rename to ...
3150 (avr_have_mul_p): ... new.
3151 (TARGET_CPU_CPP_BUILTINS): Use 'avr_have_mul_p' instead of
3152 'avr_enhanced_p' for "__AVR_ENHANCED__". Define "__AVR_HAVE_MUL__".
3153 * config/avr/avr.c (avr_enhanced_p): Rename to ...
3154 (avr_have_mul_p): ... new.
3155 (base_arch_s): Rename 'enhanced' to 'have_mul'.
3156 (avr_override_options): Use 'avr_have_mul_p' and 'have_mul' instead of
3157 'avr_enhanced_p' and 'enhanced'.
3158 (ashlhi3_out, ashrhi3_out, lshrhi3_out, avr_rtx_costs): Use
3159 AVR_HAVE_MUL instead of AVR_ENHANCED.
3160 * avr.md (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
3161 (mulqi3, *mulqi3_enh, *mulqi3_call, mulqihi3, umulqihi3, mulhi3,
3162 *mulhi3_enh, *mulhi3_call, mulsi3, *mulsi3_call): Use AVR_HAVE_MUL
3163 instead of AVR_ENHANCED.
3164 (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
3165 * libgcc.S: Use __AVR_HAVE_MUL__ instead of __AVR_ENHANCED__.
3166 (__tablejump__): Use __AVR_HAVE_LPMX__ instead of __AVR_ENHANCED__.
3168 2007-01-17 Ian Lance Taylor <iant@google.com>
3170 * vec.h (VEC_reserve_exact): Define.
3171 (vec_gc_p_reserve_exact): Declare.
3172 (vec_gc_o_reserve_exact): Declare.
3173 (vec_heap_p_reserve_exact): Declare.
3174 (vec_heap_o_reserve_exact): Declare.
3175 (VEC_OP (T,A,reserve_exact)): New static inline function, three
3177 (VEC_OP (T,A,reserve)) [all versions]: Remove handling of
3179 (VEC_OP (T,A,alloc)) [all versions]: Call ...reserve_exact.
3180 (VEC_OP (T,A,copy)) [all versions]: Likewise.
3181 (VEC_OP (T,a,safe_grow)) [all versions]: Likewise.
3182 * vec.c (calculate_allocation): Add exact parameter. Change all
3184 (vec_gc_o_reserve_1): New static function, from vec_gc_o_reserve.
3185 (vec_gc_p_reserve, vec_gc_o_reserve): Call vec_gc_o_reserve_1.
3186 (vec_gc_p_reserve_exact, vec_gc_o_reserve_exact): New functions.
3187 (vec_heap_o_reserve_1): New static function, from vec_heap_o_reserve.
3188 (vec_heap_p_reserve, vec_heap_o_reserve): Call vec_heap_o_reserve_1.
3189 (vec_heap_p_reserve_exact): New function.
3190 (vec_heap_o_reserve_exact): New function.
3192 2007-01-17 Jan Hubicka <jh@suse.cz>
3194 * ipa-type-escape.c (look_for_casts): Revamp using handled_component_p.
3196 2007-01-17 Eric Christopher <echristo@apple.com>
3198 * config.gcc: Support core2 processor.
3200 2007-01-16 Jan Hubicka <jh@suse.cz>
3202 * tree-ssanames.c (release_dead_ssa_names): Instead of ggc_freeing
3203 the names, just unlink the chain so we don't crash on dangling pointers
3206 2007-01-16 Jan Hubicka <jh@suse.cz>
3208 * cgraph.h (cgraph_decide_inlining_incrementally): Kill.
3209 * tree-pass.h: Reorder to make IPA passes appear toegher.
3210 (pass_early_inline, pass_inline_parameters, pass_apply_inline): Declare.
3211 * cgraphunit.c (cgraph_finalize_function): Do not compute inling
3212 parameters, do not call early inliner.
3213 * ipa-inline.c: Update comments. Include tree-flow.h
3214 (cgraph_decide_inlining): Do not compute inlining parameters.
3215 (cgraph_decide_inlining_incrementally): Return TODOs; assume to
3216 be called with function context set up.
3217 (pass_ipa_inline): Remove unreachable functions before pass.
3218 (cgraph_early_inlining): Simplify assuming to be called from the
3220 (pass_early_inline): New pass.
3221 (cgraph_gate_ipa_early_inlining): New gate.
3222 (pass_ipa_early_inline): Turn into simple wrapper.
3223 (compute_inline_parameters): New function.
3224 (gate_inline_passes): New gate.
3225 (pass_inline_parameters): New pass.
3226 (apply_inline): Move here from tree-optimize.c
3227 (pass_apply_inline): New pass.
3228 * ipa.c (cgraph_remove_unreachable_nodes): Verify cgraph after
3230 * tree-inline.c (optimize_inline_calls): Return TODOs rather than
3232 (tree_function_versioning): Do not allocate dummy struct function.
3233 * tree-inline.h (optimize_inline_calls): Update prototype.
3234 * tree-optimize.c (execute_fixup_cfg): Export.
3235 (pass_fixup_cfg): Remove
3236 (tree_rest_of_compilation): Do not apply inlines.
3237 * tree-flow.h (execute_fixup_cfg): Declare.
3238 * Makefile.in (gt-passes.c): New.
3239 * passes.c: Include gt-passes.h
3240 (init_optimization_passes): New passes.
3241 (nnodes, order): New static vars.
3242 (do_per_function_toporder): New function.
3243 (execute_one_pass): Dump current pass here.
3244 (execute_ipa_pass_list): Don't dump current pass here.
3246 2007-01-16 Janis Johnson <janis187@us.ibm.com>
3248 * config/dfp-bit.c (dfp_compare_op): Return separate value for NaN.
3249 (DFP_NE, DFP_LE, DFP_GE): Return false for NaN.
3251 2007-01-16 David Edelsohn <edelsohn@gnu.org>
3253 * config/rs6000/darwin-ldouble.c: Build file for SOFT_FLOAT.
3254 (strong_alias): Define.
3255 (__gcc_qmul): Provide non-FMA for soft-float.
3261 (__gcc_qunord): New.
3271 * config/rs6000/rs6000.c (rs6000_init_libfuncs): Initialize
3272 soft-float functions.
3273 * config/rs6000/libgcc-ppc-glibc.ver: Version soft-float symbols.
3274 * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Do not warn
3275 about long double soft float.
3277 2007-01-16 Dorit Nuzman <dorit@il.ibm.com>
3278 Tehila Meyzels <tehila@il.ibm.com>
3280 * tree-vectorizer.h (is_pattern_stmt_p): New.
3281 * tree-vect-analyze.c (vect_determine_vectorization_factor): Fix
3282 formatting (tabs instead of spaces). Cleanup and clarify setting
3283 of STMT_VINFO_VECTYPE. Call is_pattern_stmt_p.
3284 * tree-vect-transform.c (vect_get_vec_def_for_operand): Fix typo.
3285 (vectorizable_type_demotion): Check that types are integral.
3286 (vectorizable_type_promotion): Likewise.
3287 (vectorizable_store): Fix typo. Eliminate new-line at end of
3290 2007-01-16 Jan Hubicka <jh@suse.cz>
3292 * tree-ssanames.c (release_dead_ssa_names): Remove invalidated
3295 2007-01-15 Eric Christopher <echristo@apple.com>
3297 * ifcvt.c: Include vec.h, vecprim.h.
3298 (check_cond_move_block): New argument regs.
3299 Reorganize. Add registers used to regs.
3300 (cond_move_process_if_block): Use regs set above as
3303 2007-01-15 Eric Christopher <echristo@apple.com>
3305 * config/darwin.h: Update copyright.
3306 (TARGET_OPTION_TRANSLATE_TABLE): Add umbrella.
3307 (LINK_COMMAND_SPEC): Add -u.
3308 (LINK_SPEC): Fix umbrella for above.
3310 2007-01-15 Joseph S. Myers <joseph@codesourcery.com>
3312 * config/soft-fp/op-common.h, config/soft-fp/op-4.h: Update from
3315 2007-01-15 Tom Tromey <tromey@redhat.com>
3317 * doc/sourcebuild.texi (libgcj Tests): Don't mention jacks.
3318 * doc/install.texi (Testing): Don't mention jacks.
3319 (Configuration): Document --enable-java-maintainer-mode. Move
3320 --with-java-home to libgcj-specific section. Document
3322 (Prerequisites): Mention --enable-java-maintainer-mode, ecj1.
3324 2007-01-15 Jan Hubicka <jh@suse.cz>
3326 * tree-ssa-dce.c (DCE_TODOs): New.
3327 (propagate_necessity): Return if something changed.
3328 (eliminate_unnecessary_stmts): Likewise.
3329 (perform_tree_ssa_dce): Return TODO flags when needed.
3330 (pass_dce, pass_dce_loop, pass_cd_dce): Remove TODO flags.
3332 2007-01-15 Uros Bizjak <ubizjak@gmail.com>
3334 * config/i386/i386.md (fyl2xxf3_i387): Rename from fyl2x_xf3.
3335 (fyl2x_extend<mode>xf3_i387): New insn pattern.
3336 (log<mode>2): Rename from logsf2 and logdf2 and macroize insn
3337 patterns using X87MODEF12 mode macro. Extend operand 1
3338 to XFmode. Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
3339 (log10<mode>2): Ditto.
3340 (log2<mode>2): Ditto.
3341 (log1p<mode>2): Ditto.
3342 (logb<mode>2): Ditto.
3343 (fyl2xp1xf3_i387): Rename from fyl2xp1_xf3.
3344 (fyl2xp1_extend<mode>xf3_i387): New insn pattern.
3345 (*fxtractxf3_i387): Rename from *fxtractxf3.
3346 (fxtract_extend<mode>xf3_i387): New insn pattern.
3347 (ilogbsi2): Use match_dup 3, not match_operand:XF 3.
3348 * config/i386/i386.c (ix86_emit_i387_log1p): Use gen_fyl2xp1xf3_i387()
3349 and gen_fyl2xxf3_i387().
3351 2007-01-14 Zdenek Dvorak <dvorakz@suse.cz>
3353 * loop-unswitch.c (unswitch_loop): Do not call fix_loop_placement.
3354 * cfgloopmanip.c (fix_loop_placement): Made static. Use
3355 get_loop_exit_edges. Changed return type to bool.
3356 * cfgloop.h (fix_loop_placement): Declaration removed.
3358 2007-01-14 Dorit Nuzman <dorit@il.ibm.com>
3360 * param.h (MIN_VECT_LOOP_BOUND): New.
3361 * params.def (MIN_VECT_LOOP_BOUND): New.
3362 * tree-vectorizer.c (slpeel_tree_peel_loop_to_edge): Takes another
3363 argument - minimum threshold for number of iterations.
3364 * tree-vectorizer.h (slpeel_tree_peel_loop_to_edge): Add another
3365 argument to declaration.
3366 * tree-vect-analyze.c (vect_analyze_operations): Check value of
3367 MIN_VECT_LOOP_BOUND.
3368 * tree-vect-transform.c (vect_do_peeling_for_loop_bound): Call
3369 slpeel_tree_peel_loop_to_edge with additional argument.
3370 (vect_do_peeling_for_alignment): Likewise.
3371 * doc/invoke.texi (min-vect-loop-bound): Document new param option.
3373 2007-01-14 Uros Bizjak <ubizjak@gmail.com>
3376 * config/i386/i386.c (print_operand) ['z']: Output 'b' for
3379 2007-01-14 Jan Hubicka <jh@suse.cz>
3381 * tree-dfa.c (remove_referenced_var): New function.
3382 * tree-ssa-live.c (remove_unused_locals): Walk referenced vars and
3383 prune referenced vars list too.
3384 * tree-flow.h (remove_referenced_var): Declare.
3386 2007-01-14 Jan Hubicka <jh@suse.cz>
3388 * tree-eh.c (add_stmt_to_eh_region_fn): Do not add call_exprs
3390 (remove_stmt_from_eh_region_fn): Do not remove call_exprs.
3391 (verify_eh_throw_stmt_node, verify_eh_throw_table_statements): Kill.
3392 * except.h (verify_eh_throw_table_statements): Kill prototype.
3393 * cfgexpand.c (expand_gimple_basic_block): Propagate Eh regions
3394 into call exrepssions.
3395 * tree-optimize.c (execute_free_cfg_annotatiosn): Do not call
3397 * tree-cfg.c: Include pointer-set.h.
3398 (verify_node_sharing): Work on pointer set.
3399 (verify_eh_throw_stmt_node): New.
3400 (verify_stmts): Use pointers sets, verify throw_stmt.
3402 2007-01-13 Zdenek Dvorak <dvorakz@suse.cz>
3404 * ipa-reference.c (analyze_function): Consider also addresses taken
3407 2007-01-12 Roger Sayle <roger@eyesopen.com>
3409 * c-typeck.c (null_pointer_constant_p): Replace use of
3410 TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
3411 (build_c_cast): Likewise.
3413 2007-01-12 Roger Sayle <roger@eyesopen.com>
3415 * tree.h (force_fit_type_double): Remove unused final argument.
3416 * c-common.c (constant_expression_warning): Replace use of
3417 TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
3418 (convert_and_check): Likewise.
3419 (shorten_compare): Update call to force_fit_type_double.
3420 (c_common_truthvalue_conversion) <INTEGER_CST>: Use integer_zerop.
3421 * convert.c (convert_to_pointer): Update call to
3422 force_fit_type_double.
3423 * fold-const.c (force_fit_type_double): Remove overflowed_const
3425 (int_const_binop, fold_convert_const_int_from_int,
3426 fold_convert_const_int_from_real, fold_div_compare,
3427 fold_sign_changed_comparison, fold_unary, fold_negate_const,
3428 fold_abs_const, fold_not_const): Remove the final argument from
3429 calls to force_fit_type_double.
3431 2007-01-12 Andrew Pinski <andrew_pinski@playstation.sony.com>
3433 * configure.ac: Set insn to "nop" for spu-*-* also.
3434 * configure: Regenerate.
3436 2007-01-12 Olga Golovanevsky <olga@il.ibm.com>
3438 * builtins.def : Add BUILT_IN_FREE.
3440 2007-01-12 Jan Hubicka <jh@suse.cz>
3442 PR tree-optimization/30443
3443 * tree-inline.c (tree_function_versioning): Do not optimize when
3444 cloning for inlining.
3446 2007-01-12 Zdenek Dvorak <dvorakz@suse.cz>
3448 * doc/loop.texi: Document recording of loop exits.
3449 * cfgloopmanip.c (loopify, duplicate_loop): Use alloc_loop.
3450 (update_single_exits_after_duplication,
3451 update_single_exit_for_duplicated_loop,
3452 update_single_exit_for_duplicated_loops): Removed.
3453 (duplicate_loop_to_header_edge): Do not call
3454 update_single_exits_after_duplication and
3455 update_single_exit_for_duplicated_loops.
3456 (loop_version): Do not update single_exit information.
3457 (fix_loop_structure): Use record_loop_exits instead of
3458 mark_single_exit_loops.
3459 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update
3460 the lists of loop exits.
3461 * cfghooks.c (redirect_edge_and_branch, redirect_edge_and_branch_force,
3462 split_edge, merge_blocks): Update the lists of loop exits.
3463 * modulo-sched.c (sms_schedule): Pass LOOPS_HAVE_RECORDED_EXITS to
3464 loop_optimizer_init.
3465 * loop-init.c (loop_optimizer_init): Call record_loop_exits instead
3466 of mark_single_exit_loops.
3467 (loop_optimizer_finalize): Call release_recorded_exits.
3468 * tree-ssa-loop.c (tree_loop_optimizer_init): Pass
3469 LOOPS_HAVE_RECORDED_EXITS to loop_optimizer_init.
3470 * tree-vectorizer.c (slpeel_tree_duplicate_loop_to_edge_cfg): Do not
3471 update single exit information.
3472 * lambda-code.c (perfect_nestify): Ditto.
3473 * cfgloop.c (flow_loop_free): Destroy the list of exits of the loop.
3474 (mark_single_exit_loops): Removed.
3475 (alloc_loop, loop_exit_hash, loop_exit_eq, loop_exit_free,
3476 get_exit_descriptions, rescan_loop_exit, record_loop_exits,
3477 dump_recorded_exit, dump_recorded_exits, release_recorded_exits): New
3479 (get_loop_exit_edges, single_exit): Use recorded exit lists.
3480 (add_bb_to_loop, remove_bb_from_loops): Update the lists of loop exits.
3481 (verify_loop_structure): Verify consistency of the exit lists.
3482 (flow_loops_find): Use alloc_loop. Initialize exits hash.
3483 (set_single_exit): Removed.
3484 * cfgloop.h (struct loop_exit): New function.
3485 (struct loop): single_exit_ field replaced by exits field.
3486 (LOOPS_HAVE_MARKED_SINGLE_EXITS): Replaced by LOOPS_HAVE_RECORDED_EXITS.
3487 (struct loops): Added exits hash.
3488 (mark_single_exit_loops, set_single_exit): Declaration removed.
3489 (release_recorded_exits, record_loop_exits, rescan_loop_exit): Declare.
3491 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3493 * doc/invoke.texi: Avoid use of @headitem.
3495 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3497 * cse.c (cse_insn): Move HAVE_CC0 code after declarations.
3499 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3501 * doc/install.texi: Fix m68k-*-* anchor and add m68k-*-* to the
3504 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
3505 Richard Sandiford <richard@codesourcery.com>
3507 * doc/invoke.texi: Document -mno-bitfield, -mno-rtd and -mno-short.
3508 * config/m68k/m68k.opt: Resort options.
3509 (mbitfield, mrtd, mshort): Remove RejectNegative properties.
3511 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
3512 Richard Sandiford <richard@codesourcery.com>
3514 * doc/invoke.texi: Document the macros that are defined by
3515 m68k's -mtune and -mhard-float options.
3516 * config/m68k/m68k-protos.h (m68k_cpp_cpu_ident) Declare.
3517 (m68k_cpp_cpu_family): Likewise.
3518 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Add a full set
3519 of __ucfv*__ macros. Define __mcffpu__ if generating code for
3520 ColdFire FPUs. Define __mcf_cpu_* and __mcf_family_* macros.
3521 * config/m68k/m68k.c (m68k_cpp_cpu_ident): New function.
3522 (m68k_cpp_cpu_family): Likewise.
3524 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3526 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Treat all mc68*
3527 macros besides mc68000 as tuning macros. Use a switch statement
3528 to set them and mcpu32.
3530 2007-01-12 Julian Brown <julian@codesourcery.com>
3532 * config/m68k/m68k.h: Use TARGET_68040 instead of TARGET_68040_ONLY.
3533 (TARGET_68040_ONLY): Rename to...
3534 (TARGET_68040): ...this.
3535 * config/m68k/m68k.c: Use TARGET_68040 instead of TARGET_68040_ONLY.
3536 * config/m68k/m68k.md: Likewise.
3538 2007-01-12 Julian Brown <julian@codesourcery.com>
3539 Nathan Sidwell <nathan@codesourcery.com>
3540 Richard Sandiford <richard@codesourcery.com>
3542 * config.gcc (m680[012]0-*-*, m68k*-*-*): Set m68k_cpu_ident to
3543 the -mcpu= argument associated with the --with-cpu setting.
3544 Define M68K_DEFAULT_TUNE to the default -mtune= option,
3545 if different from the one implied by the -mcpu setting.
3546 Accept --with-cpu=FOO if FOO is listed in m68k-devices.def,
3547 using mcpu=FOO as the default CPU option. Set target_cpu_default2.
3548 * doc/invoke.texi: Mention ColdFire in the introduction to the
3549 m68k options. Document the new -march, -mcpu, -mtune, -mdiv,
3550 -mno-div and -mhard-float options. Make -m68881 a synonym for
3551 -mhard-float. Document the previously-undocumented -m5206e,
3552 -m528x, -m5307 and -m5407 options. Tweak the existing option
3553 documentation for consistency.
3554 * doc/install.texi: Mention new --with-cpu arguments.
3555 * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Only use the
3556 default CPU if neither -mcpu nor -march are specified.
3557 (ASM_CPU_SPEC): Pass down -mcpu and -march options.
3558 (TARGET_CPU_CPP_BUILTINS): Set __mcfisa*__ macros from
3559 TARGET_ISA*. Set the legacy __mcf*__ cpu macros in the same way,
3560 using m68k_tune to decide between families that implement the
3561 same ISA. Use m68k_tune to set __mcfv4e__.
3562 (FL_BITFIELD, FL_68881, FL_COLDFIRE, FL_CF_HWDIV, FL_CF_MAC)
3563 (FL_CF_EMAC, FL_CF_EMAC_B, FL_CF_USP, FL_CF_FPU, FL_ISA_68000)
3564 (FL_ISA_68010, FL_ISA_68020, FL_ISA_68040, FL_ISA_A, FL_ISA_B)
3565 (FL_ISA_C, FL_ISA_MMU): New macros.
3566 (MASK_COLDFIRE): Delete.
3567 (TARGET_68010, TARGET_68020, TARGET_68040_ONLY, TARGET_COLDFIRE)
3568 (TARGET_ISAB): Redefine in terms of m68k_cpu_flags.
3569 (TARGET_68881, TARGET_COLDFIRE_FPU): Redefine in terms of m68k_fpu.
3570 (TARGET_HARD_FLOAT): Do not define here.
3571 (TARGET_ISAAPLUS, TARGET_ISAC): New macros.
3572 (TUNE_68000): New macro.
3573 (TUNE_68000_10): Redefine in terms of TUNE_68000 and TUNE_68010.
3574 (TUNE_68010, TUNE_68030, TUNE_68040, TUNE_68060, TUNE_CPU32)
3575 (TUNE_CFV2): Redefine in terms of m68k_tune.
3576 (uarch_type, target_device, fpu_type): New enums.
3577 (m68k_cpu, m68k_tune, m68k_fpu, m68k_cpu_flags): Declare.
3578 * config/m68k/m68k.c (TARGET_DEFAULT): Remove MASK_68881.
3579 (FL_FOR_isa_00, FL_FOR_isa_10, FL_FOR_isa_20, FL_FOR_isa_40)
3580 (FL_FOR_isa_cpu32, FL_FOR_isa_a, FL_FOR_isa_aplus, FL_FOR_isa_b)
3581 (FL_FOR_isa_c): New macros.
3582 (m68k_isa): New enum.
3583 (m68k_target_selection): New structure.
3584 (all_devices, all_isas, all_microarchs): New tables.
3585 (m68k_cpu_entry, m68k_arch_entry, m68k_tune_entry, m68k_cpu)
3586 (m68k_tune, m68k_fpu, m68k_cpu_flags): New variables.
3587 (MASK_ALL_CPU_BITS): Delete.
3588 (m68k_find_selection): New function.
3589 (m68k_handle_option): Handle -mcpu=, -march= and -mtune=.
3590 Map the legacy target options to a combination of the new ones.
3591 (override_options): Set m68k_cpu, m68k_tune, m68k_fpu and
3592 m68k_cpu_flags. Handle M68K_DEFAULT_TUNE. Use m68k_cpu_flags
3593 to derive default MASK_BITFIELD, MASK_CF_HWDIV and MASK_HARD_FLOAT
3595 * config/m68k/m68k.opt (m5200, m5206e, m528x, m5307, m5407, mcfv4e)
3596 (m68010, m68020, m68020-40, m68020-60, m68030, m68040): Remove Mask
3598 (m68881, msoft-float): Change mask from 68881 to HARD_FLOAT.
3599 (march=, mcpu=, mdiv, mhard-float, mtune=): New options.
3600 * config/m68k/m68k-devices.def: New file.
3602 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3603 Nathan Sidwell <nathan@codesourcery.com>
3605 * config/m68k/m68k.h (ASM_CPU_SPEC, ASM_SPEC, EXTRA_SPECS)
3606 (SUBTARGET_EXTRA_SPECS): New macros.
3607 * config/m68k/linux.h (ASM_SPEC): Remove CPU flags;
3608 use %(asm_cpu_spec) instead.
3609 * config/m68k/m68k-none.h (ASM_SPEC): Likewise.
3610 * config/m68k/openbsd.h (ASM_SPEC): Likewise.
3611 * config/m68k/netbsd-elf.h (ASM_SPEC): Likewise.
3612 (EXTRA_SPECS): Rename to...
3613 (SUBTARGET_EXTRA_SPECS): ...this.
3615 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
3616 Richard Sandiford <richard@codesourcery.com>
3617 Julian Brown <julian@codesourcery.com>
3619 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
3620 (m68k-*-uclinux*, m68k-*-linux*, m68k-*-rtems*): Set default_m68k_cpu
3621 to the configuration's default CPU.
3622 (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*): Likewise.
3623 Remove default masks.
3624 (m680[012]0-*-*): Set the default with_cpu to the first part of
3626 (m68k*-*-*): Set the default with_cpu to m$default_m68k_cpu.
3627 (m68k*-*-linux): Extend the --with-cpu handling to...
3628 (m680[012]0-*-*, m68k*-*-*): ...these configurations. Allow m68000
3629 and m68010. Don't set target_cpu_default2.
3630 * doc/install.texi: Document --with-cpu for m68k.
3631 * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Define.
3632 * config/m68k/m68k-none.h (TARGET_CPU_DEFAULT, M68K_CPU_m68k)
3633 (M68K_CPU_m68000, M68K_CPU_m68010, M68K_CPU_m68020, M68K_CPU_m68030)
3634 (M68K_CPU_m68040, M68K_CPU_m68302, M68K_CPU_m68332, TARGET_DEFAULT)
3635 (ASM_CPU_DEFAULT_SPEC, CC1_CPU_DEFAULT_SPEC): Delete.
3636 (ASM_SPEC): Remove use of %(asm_cpu_default).
3637 (EXTRA_SPECS, SUBTARGET_EXTRA_SPECS, MULTILIB_DEFAULTS): Delete.
3638 * config/m68k/linux.h (TARGET_DEFAULT): Delete.
3639 (CPP_SPEC): Merge definitions. Do not handle __HAVE_68881__ here.
3640 * config/m68k/netbsd-elf.h (TARGET_OS_CPP_BUILTINS): Define
3641 __HAVE_FPU__ if TARGET_HARD_FLOAT.
3642 (TARGET_DEFAULT): Delete.
3643 (EXTRA_SPECS): Delete cpp_cpu_default_spec, cpp_cpu_spec,
3644 cpp_fpu_spec, asm_default_spec and netbsd_cpp_spec.
3645 (CPP_CPU_SPEC): Delete.
3646 (TARGET_VERSION): Merge definitions, using TARGET_68010 to pick
3647 the appropriate string.
3648 (CPP_CPU_DEFAULT_SPEC, ASM_DEFAULT_SPEC, CPP_FPU_SPEC): Delete.
3649 (CPP_SPEC): Define to NETBSD_CPP_SPEC.
3650 (ASM_SPEC): Don't use %(asm_default_spec).
3651 * config/m68k/m68k.c (TARGET_DEFAULT_TARGET_FLAGS): Remove
3652 TARGET_DEFAULT and add MASK_68881.
3653 * config/m68k/m68k.md: Remove mention of TARGET_DEFAULT from comments.
3655 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3657 * config.gcc (m68010-*-netbsdelf*): Add MASK_68010.
3658 (m68k*-*-netbsdelf*, m68k*-*-openbsd*, m68k*-linux*): Add
3659 MASK_68010 alongside MASK_68020.
3660 * doc/invoke.texi: Document -m68010.
3661 * config/m68k/m68k.opt (m68010): New.
3662 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Define mc68010
3664 (TUNE_68010): New macro.
3665 * config/m68k/m68k-none.h (M68K_CPU_m68k, M68K_CPU_m68010)
3666 (M68K_CPU_m68020, M68K_CPU_m68030, M68K_CPU_m68040)
3667 (M68K_CPU_m68332): Add MASK_68010.
3668 * config/m68k/linux.h (TARGET_DEFAULT): Add MASK_68010 to
3669 fallback definition.
3670 * config/m68k/netbsd-elf.h (CPP_CPU_SPEC): Remove now-redundant
3672 * config/m68k/m68k.c (MASK_ALL_CPU_BITS): Add MASK_68010.
3673 (m68k_handle_option): Handle OPT_m68010. Add MASK_68010
3674 to all entries that use MASK_68020.
3675 (output_move_simode_const, output_move_himode, output_move_qimode)
3676 (output_move_stricthi, output_move_strictqi): Use TARGET_68010
3677 instead of TARGET_68020 to select clr behavior. Remove comment
3678 about there being no TARGET_68010.
3679 * config/m68k/m68k.md: Likewise throughout.
3681 2007-01-12 Julian Brown <julian@codesourcery.com>
3683 * config/m68k/m68k.h (TARGET_ISAB): New macro.
3684 * config/m68k/m68k.c: Use TARGET_ISAB rather than TARGET_CFV4.
3685 * config/m68k/m68k.md: Likewise.
3687 2007-01-12 Julian Brown <julian@codesourcery.com>
3689 * config/m68k/m68k.h (LEGITIMATE_INDEX_P, LEGITIMIZE_ADDRESS): Use
3690 TARGET_COLDFIRE_FPU instead of TARGET_CFV4E.
3692 2007-01-12 Julian Brown <julian@codesourcery.com>
3694 * config/m68k/m68k.h (TUNE_68040_60): New macro.
3695 * config/m68k/m68k.c (standard_68881_constant_p): Use it.
3696 * config/m68k/m68k.md: Likewise.
3698 2007-01-12 Julian Brown <julian@codesourcery.com>
3699 Richard Sandiford <richard@codesourcery.com>
3701 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Use TUNE_68030
3702 instead of TARGET_68030, TUNE_68040 instead of TARGET_68040,
3703 TUNE_68060 instead of TARGET_68060 and TUNE_CPU32 instead of
3705 (TARGET_CPU32): Rename to...
3706 (TUNE_CPU32): ...this.
3707 (TUNE_68000_10, TUNE_68030, TUNE_68040, TUNE_68060)
3708 (TUNE_CFV2): New macros.
3709 * config/m68k/netbsd-elf.h (LONG_DOUBLE_TYPE_SIZE): Simplify;
3710 remove conditions that are implied by TARGET_68020.
3711 * config/m68k/m68k.c (m68k_output_function_prologue): Use TUNE_68040
3712 instead of TARGET_68040 and TUNE_CPU32 instead of TARGET_CPU32.
3713 (m68k_output_function_epilogue): Likewise.
3714 (m68k_rtx_costs): Likewise. Use TUNE_68060 instead of TARGET_68060
3715 and TUNE_CFV2 instead of TARGET_5200. Use TUNE_68000_10 instead of
3716 "!TARGET_68020 && !TARGET_COLDFIRE" to choose between 68000 and
3717 non-68000 timings. Refactor multiplication and division costs.
3718 (output_addsi3): Use TUNE_68040 instead of TARGET_68040 and
3719 TUNE_CPU32 instead of TARGET_CPU32.
3720 (standard_68881_constant_p): Use TUNE_68040 instead of TARGET_68040
3721 and TUNE_68060 instead of TARGET_68060.
3722 * config/m68k/m68k.md: Use TUNE_68040 instead of TARGET_68040,
3723 TUNE_68060 instead of TARGET_68060, and TUNE_CPU32 instead of
3725 (movsi_const0): Use TUNE_68000_10 rather than "!TARGET_68020
3726 && !TARGET_COLDFIRE" to choose between moveq and clr.
3727 Likewise in the unnamed movsf pattern.
3728 (ashlsi_17_24, lshrsi_17_24): Guard with TUNE_68000_10 rather than
3729 "!TARGET_68020 && !TARGET_COLDFIRE". Likewise the unnamed
3732 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3734 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Increase amount
3735 of tabbing before backslashes.
3737 2007-01-11 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
3739 * pa-linux.h (ASM_OUTPUT_INTERNAL_LABEL): Undefine.
3740 * pa.h (ASM_OUTPUT_LABEL): Output colon when using GAS.
3741 (ASM_OUTPUT_INTERNAL_LABEL): Define.
3743 2007-01-11 Zdenek Dvorak <dvorakz@suse.cz>
3745 * tree-ssa-loop-ivopts.c (extract_cond_operands): Split from
3746 find_interesting_uses_cond.
3747 (find_interesting_uses_cond): Use extract_cond_operands.
3748 (rewrite_use_compare): Use extract_cond_operands and
3749 force_gimple_operand_bsi. Do not call update_stmt.
3750 (determine_use_iv_cost_condition): Use extract_cond_operands.
3751 Return cheaper of using original bound and changing the exit bound.
3753 2007-01-11 Zdenek Dvorak <dvorakz@suse.cz>
3755 PR tree-optimization/29516
3756 * tree-ssa-address.c (tree_mem_ref_addr, add_to_parts,
3757 most_expensive_mult_to_index, addr_to_parts,
3758 create_mem_ref, maybe_fold_tmr): Make the type of
3759 fields of TARGET_MEM_REF sizetype.
3760 (move_fixed_address_to_symbol, move_pointer_to_base):
3762 * tree.def (TARGET_MEM_REF): Add comment on types of
3765 2007-01-11 Joseph Myers <joseph@codesourcery.com>
3767 * c-common.c (vector_types_convertible_p): Treat opaque types as
3768 always convertible if they have the same size, but not otherwise.
3770 2007-01-11 Steven Bosscher <steven@gcc.gnu.org>
3772 * ifcvt.c (struct noce_if_info): Add comments to the fields.
3773 Remove the b_unconditional field.
3774 (noce_try_sign_mask): Do not look at b_unconditional.
3775 (noce_process_if_block): Do not use merge_if_blocks. Update
3776 the CFG here. Do not set b_unconditional.
3777 (cond_move_process_if_block): Likewise.
3778 (find_cond_trap): Likewise.
3779 (check_cond_move_block): Require simple jump insns at the end
3782 2007-01-11 Jan Hubicka <jh@suse.cz>
3784 PR tree-optimization/1046
3785 * tree-tailcall.c (suitable_for_tail_call_opt_p): Use TREE_ADDRESSABLE
3786 when alias info is not ready.
3787 (pass_tail_recursion): Do not require aliasing.
3788 * tree-ssa-copyrename.c (pass_rename_ssa_cop): Likewise.
3789 * tree-ssa-ccp.c (pass_ccp, pass_fold_builtins): Likewise.
3790 * tree-ssa-copy.c (pass_copy_prop): Likewise.
3791 * tree-ssa-forwprop.c (pass_forwprop): Likewise.
3792 * tree-ssa-dce.c (pass_dce, pass_dce_loop, pass_cd_dce): Likewise.
3793 * passes.c (init_optimization_passes): Execute rename_ssa_copies,
3794 ccp, forwprop, copy_prop, merge_phi, copy_prop, dce and tail recursion
3796 * tree-ssa-operands.c (add_virtual_operand, get_indirect_ref_operand):
3797 When aliasing is not build, mark statement as volatile.
3799 2007-01-11 Tom Tromey <tromey@redhat.com>
3801 PR preprocessor/15185, PR preprocessor/20989:
3802 * doc/cppopts.texi <-MT>: Update description of algorithm for
3803 computing default target.
3804 <-M, -MD>: Reword "basename" text.
3806 2007-01-11 Roger Sayle <roger@eyesopen.com>
3808 * builtins.c (expand_builtin_pow, expand_builtin_powi,
3809 fold_builtin_cabs, fold_builtin_sqrt, fold_builtin_trunc,
3810 fold_builtin_floor, fold_builtin_ceil, fold_builtin_round,
3811 fold_builtin_int_int_roundingfn, fold_builtin_bitop,
3812 fold_builtin_bswap, real_constp, fold_builtin_pow,
3813 fold_builtin_powi, fold_builtin_signbit, fold_builtin_copysign,
3814 do_mpfr_arg1, do_mpfr_arg2, do_mpfr_arg3, do_mpfr_sincos): Replace
3815 uses of the macro TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
3816 * convert.c (convert_to_pointer): Likewise.
3817 * expr.c (highest_pow2_factor, expand_expr_real_1): Likewise.
3818 * fold-const.c (force_fit_type, fold_negate_expr, int_const_binop,
3819 const_binop, fold_convert_const_int_from_int,
3820 fold_convert_const_int_from_real,
3821 fold_convert_const_real_from_real, sign_bit_p,
3822 optimize_minmax_comparison, extract_muldiv_1, fold_div_compare,
3823 fold_sign_changed_comparison, fold_unary, fold_comparison,
3824 fold_binary, multiple_of_p, tree_Expr_non_zero_p,
3825 fold_negate_const, fold_abs_const, fold_not_const): Likewise.
3826 * print-tree.c (print_node_brief, print_node): Likewise.
3827 * stor-layout.c (place_field, layout_type): Likewise.
3828 * tree-chrec.c (keep_cast): Likewise.
3829 * tree.c (build_vector, build_real, build_real_from_int_cst,
3830 build_complex): Likewise.
3832 2007-01-11 Roger Sayle <roger@eyesopen.com>
3834 * tree.h (TREE_CONSTANT_OVERFLOW): Obsolete. For the time being,
3835 treat TREE_CONSTANT_OVERFLOW as a synonym of TREE_OVERFLOW.
3837 2007-01-11 Paolo Bonzini <bonzini@gnu.org>
3839 * configure.ac (strict1_warn): Rename to strict_warn.
3840 (WERROR, --enable-werror, symlink hacks, stage1_cflags,
3841 cc_set_by_configure, quoted_cc_set_by_configure,
3842 stage_prefix_set_by_configure, quoted_stage_prefix_set_by_configure,
3843 all_boot_languages, all_stagestuff): Remove.
3844 (target_list): Remove bootstrap targets.
3845 * Makefile.in (quickstrap): Unconditionally make a synonym of all.
3846 (BOOT_LANGUAGES, STAGE1_CFLAGS, STAGE1_CHECKING,
3847 REMAKEFLAGS, FLAGS_TO_PASS, PREPEND_DOTDOT_TO_RELATIVE_PATHS,
3848 SUBDIR_FLAGS_TO_PASS, WERROR_FLAGS, STRICT2_WARN, LANG_STAGESTUFF,
3849 VOL_FILES, POSTSTAGE1_FLAGS_TO_PASS, STAGE2_FLAGS_TO_PASS,
3850 STAGEPROFILE_FLAGS_TO_PASS, STAGEFEEDBACK_FLAGS_TO_PASS, stage1_build,
3851 stage1_copy, stage2_build, stage2_copy, stageprofile_build,
3852 stageprofile_copy, stage3_build, stage3_copy, stagefeedback_build,
3853 stagefeedback_copy, stage4_build, clean_s1, clean_sw, bootstrap,
3854 bootstrap-lean, bootstrap2, bootstrap2-lean, bootstrap3,
3855 bootstrap3-lean, bootstrap4, bootstrap4-lean, unstage1, unstage2,
3856 unstage3, unstage4, unstageprofile, unstagefeedback, restage, restage2,
3857 restage3, restage4, restageprofile, restagefeedback, bubbleestrap,
3858 cleanstrap, unstrap, restrap, *compare, *compare3, *compare4,
3859 *compare-lean, *compare3-lean, *compare4-lean, stage1-start, stage1,
3860 stage2-start, stage2, stage3-start, stage3, stage4-start, stage4,
3861 stageprofile-start, stageprofile, stagefeedback-start, stagefeedback,
3862 risky-stage1, risky-stage2, risky-stage3, risky-stage4): Remove.
3863 (ORDINARY_FLAGS_TO_PASS): Rename to FLAGS_TO_PASS.
3864 (STAGECOPYSTUFF, STAGEMOVESTUFF): Consolidate into MOSTLYCLEANFILES.
3865 (mostlyclean): Adjust.
3866 (clean, distclean): Don't mention bootstrap stuff.
3867 * configure: Regenerate.
3868 * ada/config-lang.in, cp/config-lang.in, forttran/config-lang.in,
3869 java/config-lang.in, objc/config-lang.in, objcp/config-lang.in,
3870 treelang/config-lang.in (stagestuff): Remove.
3871 * doc/sourcebuild.texi (stage1, stage2, stage3, stage4,
3872 stageprofile, stagefeedback, stagestuff): Remove mention.
3874 2007-01-11 Nick Clifton <nickc@redhat.com>
3876 * config/mcore/predicates.md (mcore_general_movesrc_operand):
3878 (mcore_general_movdst_operand): Do not accept CONST_INTs.
3879 (mcore_arith_K_S_operand): Run the test for the S constraint not
3880 the test for the M constraint.
3881 (mcore_addsub_operand): Do not accept integer values that are
3882 larger than 32 bits.
3883 * config/mcore/mcore.md: Remove unused constraints from split.
3884 (andsi3): Use HOST_WIDE_INT instead of int to hold an INTVAL.
3886 (allocate_stack): Likewise.
3887 * config/mcore/mcore.c (mcore_print_operand): Restrict output of P
3888 operands to 32 bits.
3889 (mcore_const_costs): Use HOST_WIDE_INT instead of int to hold an
3891 (mcore_and_cost, mcore_modify_comparison, const_ok_for_mcore,
3892 mcore_const_ok_for_inline, mcore_const_trick_uses_not,
3893 try_constant_tricks, mcore_num_ones, mcore_num_zeros,
3894 mcore_output_bclri, mcore_output_andn, output_inline_const,
3895 mcore_output_move, mcore_output_movedouble): Likewise.
3896 (mcore_output_cmov): Use CONST_OK_FOR_M and CONST_OK_FOR_N.
3897 (output_inline_const): Likewise.
3898 (output_inline_const): Fix format strings used in sprintf
3900 * config/mcore/mcore-protos.h: Update prototypes for changed
3901 functions in mcore.c.
3902 * config/mcore/mcore.h (CONST_OK_FOR_I): Cast values to
3903 HOST_WIDE_INT and not int.
3904 (CONST_OK_FOR_J, CONST_OK_FOR_K, CONST_OK_FOR_L, CONST_OK_FOR_M,
3905 CONST_OK_FOR_N): Likewise.
3906 (LEGITIMATE_CONSTANT_P): Also check CONSTANT_P.
3907 (GO_IF_LEGITIMATE_INDEX): Use HOST_WIDE_INT instead of int to hold
3910 2007-01-10 Jan Hubicka <jh@suse.cz>
3912 * tree-vrp.c (remove_range_assertions): Release defs.
3913 * tree-ssa-loop-ivopts.c (rmeove_statement): Likewise.
3914 * tree-ssa-dom.c (remove_stmt_or_phi): Likewise.
3916 2007-01-10 Paul Brook <paul@codesourcery.com>
3918 * config/arm/arm.c (arm_rtx_costs_1): Handle mutiply-subtract.
3919 * config/arm/arm.md (mulsi3subsi): New insn.
3921 2007-01-10 Zdenek Dvorak <dvorakz@suse.cz>
3923 * tree-ssa-loop-manip.c (tree_unroll_loop): Make it a wrapper over ...
3924 (tree_transform_and_unroll_loop): New.
3925 * tree-flow.h (transform_callback, tree_transform_and_unroll_loop):
3928 2007-01-10 Robert Kennedy <jimbob@google.com>
3930 * fold-const.c (fold_comparison): Fold comparisons like (x *
3931 1000 < 0) to (x < 0).
3933 2007-01-10 Ian Lance Taylor <iant@google.com>
3935 * tree-pretty-print.c (dump_generic_node): Print parentheses when
3936 operands have the same priority.
3938 2007-01-10 Tom Tromey <tromey@redhat.com>
3940 * fold-const.c (fold_truthop): Don't check can_use_bit_fields_p.
3941 (fold_binary): Likewise.
3942 * langhooks.c (lhd_can_use_bit_fields_p): Removed.
3943 * langhooks-def.h (lhd_can_use_bit_fields_p): Removed.
3944 (LANG_HOOKS_CAN_USE_BIT_FIELDS_P): Removed.
3945 (LANG_HOOKS_INITIALIZER): Remove LANG_HOOKS_CAN_USE_BIT_FIELDS_P.
3946 * langhooks.h (struct lang_hooks): Removed field
3947 'can_use_bit_fields_p'.
3949 2007-01-10 Ralf Corsépius <ralf.corsepius@rtems.org>
3951 * config/bfin/t-bfin, config/bfin/t-bfin-elf: Remove GCC_CFLAGS.
3953 2007-01-10 Razya Ladelsky <razya@il.ibm.com>
3955 * function.c (get_last_funcdef_no): New function.
3956 * function.h (get_last_funcdef_no): Declare.
3957 * tree-inline.c (initialize_cfun): Add initialization.
3958 (tree_function_versioning): Cleanup.
3960 2007-01-10 Jan Hubicka <jh@suse.cz>
3962 * tree-inline.c (setup_one_parameter): Do not propagate into abnormal
3965 2007-01-10 Sa Liu <saliu@de.ibm.com>
3966 Ben Elliston <bje@au.ibm.com>
3968 * spu.h (STACK_SAVE_AREA): Use VOIDmode for SAVE_FUNCTION, SImode
3969 for SAVE_NONLOCAL and Pmode for any other save level.
3970 * spu-protos.h (spu_restore_stack_block): Declare.
3971 * spu.md (save_stack_block): Remove.
3972 (restore_stack_block): Call spu_restore_stack_block.
3973 * spu.c (spu_restore_stack_block): New function.
3974 (spu_expand_epilogue): Remove old comment.
3976 2007-01-09 Zdenek Dvorak <dvorakz@suse.cz>
3978 PR tree-optimization/30322
3979 * tree-ssa-loop-ivopts.c (fold_affine_expr, iv_value): Removed.
3980 (cand_value_at): Return the value as aff_tree.
3981 (may_eliminate_iv): Convert the bound from aff_tree to tree.
3982 * tree-affine.c (aff_combination_add_cst, aff_combination_add_product,
3983 aff_combination_mult): New functions.
3984 (aff_combination_add): Use aff_combination_add_cst.
3985 (aff_combination_convert): Allow conversions to a wider type.
3986 (tree_to_aff_combination): Handle BIT_NOT_EXPR.
3987 * tree-affine.h (aff_combination_mult): Declare.
3989 2007-01-09 Carlos O'Donell <carlos@codesourcery.com>
3991 * doc/tm.texi: Update documentation to reflect reality of exec
3992 and start file search behaviours. Update copyright year.
3993 * doc/invoke.texi: Explain how GCC_EXEC_PREFIX is used to find
3994 header file directories.
3996 2007-01-09 Uros Bizjak <ubizjak@gmail.com>
3998 * config/i386/i386.md (*sinxf2): Rename to *sinxf2_i387.
3999 (*cosxf2): Rename to cosxf2_i387.
4000 (*sindf2, *sinsf2): Extend operand 1 to XFmode. Macroize patterns
4001 using X87MODEF12 mode macro. Rename patterns to
4002 *sin_extend<mode>xf2_i387. Use SSE_FLOAT_MODE_P to disable patterns
4004 (*cosdf2, *cossf2): Ditto.
4005 (sincosdf3, sincossf3): Ditto. Rewrite corresponding splitters
4006 to match extended input operands.
4007 (sincos<mode>3): New expander.
4008 (*sinextendsfdf2, *cosextendsfdf2, *sincosextendsfdf3): Remove
4009 insn patterns and corresponding splitters.
4011 2007-01-09 Kaz Kojima <kkojima@gcc.gnu.org>
4013 * config/sh/t-linux (TARGET_LIBGCC2_CFLAGS): Delete.
4014 (SHLIB_MAPFILES, SHLIB_LINK, SHLIB_INSTALL): Likewise.
4016 2007-01-09 Nicolas Pitre <nico@cam.org>
4019 * arm/ieee754-df.S (Lad_s): Also test the low word of X for zero.
4021 2007-01-08 Geoffrey Keating <geoffk@apple.com>
4023 * target.h (struct gcc_target): New field library_rtti_comdat.
4024 * target-def.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): New.
4025 (TARGET_CXX): Add TARGET_CXX_LIBRARY_RTTI_COMDAT.
4026 * doc/tm.texi (C++ ABI): Document TARGET_CXX_LIBRARY_RTTI_COMDAT.
4027 * config/darwin.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
4029 2007-01-08 Geoffrey Keating <geoffk@apple.com>
4031 * doc/invoke.texi (Optimize Options): Correct description of -O0.
4033 2007-01-08 Richard Guenther <rguenther@suse.de>
4035 * tree.h (force_fit_type_double): Export.
4036 (force_fit_type): Remove.
4037 * fold-const.c (force_fit_type_double): New function.
4038 (force_fit_type): Remove.
4039 (int_const_binop): Use it.
4040 (fold_convert_const_int_from_int): Likewise.
4041 (fold_convert_const_int_from_real): Likewise.
4042 (fold_div_compare): Likewise.
4043 (fold_sign_changed_comparison): Likewise.
4044 (fold_unary): Likewise.
4045 (fold_negate_const): Likewise.
4046 (fold_abs_const): Likewise.
4047 (fold_not_const): Likewise.
4048 * c-common.c (shorten_compare): Use force_fit_type_double.
4049 * convert.c (convert_to_pointer): Likewise.
4051 2007-01-08 Richard Guenther <rguenther@suse.de>
4053 * tree.h (build_int_cst_wide_type): Export.
4054 * tree.c (build_int_cst_wide_type): New function.
4055 (build_int_cst_wide): Fix comment.
4056 * builtins.c (fold_builtin_object_size): Use build_int_cst
4057 to build -1 or 0 of the correct type. Use fit_double_type
4058 to check for overflow.
4059 * fold-const.c (optimize_bit_field_compare): Use build_int_cst_type
4061 (decode_field_reference): Likewise.
4062 (all_ones_mask_p): Likewise.
4063 (native_interpret_int): Use build_int_cst_wide_type.
4064 (fold_binary): Use build_int_cst_type to build an all-ones
4066 * stor-layout.c (set_sizetype): Use build_int_cst_wide_type.
4068 2007-01-08 Daniel Jacobowitz <dan@codesourcery.com>
4070 * config/pa/t-pa64 (libgcc_stub.a): Use $(T).
4072 2007-01-09 Ben Elliston <bje@au.ibm.com>
4074 * genautomata.c (STATS_OPTION): New option.
4075 (stats_flag): New flag.
4076 (gen_automata_option): Handle it.
4077 (initiate_automaton_gen): Ditto.
4078 (write_automata): Output statistics only if stats_flag is
4079 set. Likewise, output time statistics only if time_flag is set.
4080 * doc/md.texi (Processor pipeline description): Document new flag.
4082 2007-01-08 Richard Guenther <rguenther@suse.de>
4084 * builtins.c (fold_builtin_int_roundingfn): Use fit_double_type.
4085 * tree.c (build_int_cst_type): Likewise.
4086 (size_in_bytes): Don't call force_fit_type on the result.
4087 (int_fits_type_p): Use fit_double_type.
4088 * fold-const.c (fit_double_type): New function.
4089 (force_fit_type): Use it.
4090 * tree.h (fit_double_type): Export.
4092 2007-01-08 Jan Hubicka <jh@suse.cz>
4094 * tree-vectorizer.c (gate_increase_alignment): Fix return type.
4095 * ipa.c (function_and_variable_visibility): Fix return type.
4097 2007-01-08 Richard Guenther <rguenther@suse.de>
4099 * tree-ssa-ccp.c (maybe_fold_offset_to_array_ref): Use type
4100 of offset to build the index.
4101 * tree-pretty-print.c (dump_generic_node): Don't build negated
4102 const just for printing.
4103 * c-pretty-print.c (pp_c_integer_constant): Likewise.
4104 * builtins.c (fold_builtin_int_roundingfn): Check if result
4105 fits the type by using force_fit_type and comparing the result.
4106 * predict.c (predict_loops): Use compare_tree_int for comparison.
4107 * tree.c (build_int_cst): Fall back to integer_type_node for
4109 (build_int_cst_wide): Assert type is non-null.
4111 2007-01-08 Roberto Costa <roberto.costa@st.com>
4113 * tree-vrp.c (extract_range_from_cond_expr): New.
4114 (extract_range_from_expr): Handle COND_EXPR nodes used as expressions.
4115 * tree-ssa-ccp.c (get_maxval_strlen): Handle COND_EXPR nodes used
4117 (fold_stmt): Bug fix, avoid infinite recursion when folding COND_EXPRs.
4118 * tree-ssa-forwprop.c (simplify_cond, forward_propagate_into_cond,
4119 tree_ssa_forward_propagate_single_use_vars): Handle COND_EXPR nodes
4120 used as expressions.
4121 * tree-object-size.c (cond_expr_object_size): New.
4122 (collect_object_sizes_for): Handle COND_EXPR nodes used as expressions.
4124 2007-01-08 Jan Hubicka <jh@suse.cz>
4126 * tree-ssa-forwprop.c (forward_propagate_into_cond,
4127 tree_ssa_forward_propagate_single_use_va): Release defs of propagated
4130 2007-01-08 Richard Guenther <rguenther@suse.de>
4132 PR tree-optimization/23603
4133 * tree-vrp.c (set_value_range_to_truthvalue): New function.
4134 (extract_range_from_binary): Fall back to truthvalue instead of
4135 varying for TRUTH_*_EXPR.
4136 (extract_range_from_comparison): Fall back to truthvalue instead of
4138 (vrp_visit_phi_node): Don't adjust new range bounds to +INF/-INF
4139 if all visited PHI values were constant.
4141 2007-01-08 Jan Hubicka <jh@suse.cz>
4143 * cgraphunit.c (cgraph_process_new_functions): Reset reachable flag.
4144 (cgraph_analyze_function): break out from ...
4145 (cgraph_finalize_compilation_unit): ... here.
4146 (cgraph_expand_function): Remove forgoten commented out line.
4147 (cgraph_optimize): Analyze functions.
4149 2007-01-08 Jan Hubicka <jh@suse.cz>
4151 * tree-pas.h (TODO_remove_function): New flag.
4152 (TODO_update*): Renumber.
4153 (pass_ipa_increase_alignment,
4154 pass_ipa_function_and_variable_visibility): New passes.
4155 * cgraphunit.c (cgraph_increase_alignment): Move to tree-vectorizer.c
4156 (cgraph_function_and_variable_visibility): Move to ipa.c
4157 (cgraph_optimize): Don't call cgraph_function_and_variable_visibility,
4158 cgraph_increase_alignment.
4159 * ipa-inline.c (cgraph_decide_inlining): Don't push timevar.
4160 (cgraph_decide_inlining_incrementally): Push TV_INTEGRATION before
4161 calling tree-inline.
4162 (cgraph_early_inlining): Do not call cgraph_remove_unreachable_nodes.
4163 (pass_ipa_inline, pass_early_ipa_inlining): Set TODO_remove_functions
4164 * tree-vectorizer.c (increase_alignment): Move here from cgraphunit.c
4165 (gate_increase_alignment): New function.
4166 (pass_ipa_increase_alignment): New pass.
4167 * ipa.c: Inline tree-pass.h and timevar.h
4168 (function_and_variable_visibility): Move here from cgraphunit.c
4169 * tree-optimize.c (pass_early_local_passes): Add TODO_remove_functions.
4170 * passes.c (init_optimization_passes): Add the two new passes.
4171 (execute_todo): Handle cgraph_remove_functions.
4173 2007-01-08 Nick Clifton <nickc@redhat.com>
4175 * config/frv/predicates.md (reg_or_0_operand): Accept
4178 2007-01-08 Ralf Corsépius <ralf.corsepius@rtems.org>
4180 * config/bfin/rtems.h, config/bfin/t-rtems: New.
4181 * config.gcc: Add bfin*-rtems*.
4183 2007-01-08 Mark Shinwell <shinwell@codesourcery.com>
4185 * c.opt: Add -flax-vector-conversions.
4186 * c-typeck.c (convert_for_assignment): Pass flag to
4187 vector_types_convertible_p to allow emission of note.
4188 (digest_init): Likewise.
4189 * c-opts.c: Handle -flax-vector-conversions.
4190 * c-common.c (flag_lax_vector_conversions): New.
4191 (vector_types_convertible_p): Unless -flax-vector conversions
4192 has been passed, disallow conversions between vectors with
4193 differing numbers of subparts and/or element types. If such
4194 a conversion is disallowed, possibly emit a note on the first
4195 occasion only to inform the user of -flax-vector-conversions.
4196 The new last argument specifies this.
4197 * c-common.h (flag_lax_vector_conversions): New.
4198 (vector_types_convertible_p): Add extra argument.
4199 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Use
4200 char_type_node for V*QI type vectors.
4201 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins):
4202 Update to satisfy new typechecking rules.
4203 * config/rs6000/altivec.h (vec_cmple): Use vec_cmpge, for both
4205 * doc/invoke.texi (C Dialect Options): Document
4206 -flax-vector-conversions.
4208 2007-01-08 Mark Shinwell <shinwell@codesourcery.com>
4210 PR tree-optimization/29877
4211 * tree-ssa-ter.c (is_replaceable_p): Deem assignments with
4212 a register variable on the RHS to not be replaceable.
4214 2007-01-08 Chen Liqin <liqin@sunnorth.com.cn>
4215 * config/score/t-score-elf (MULTILIB_OPTIONS): Change.
4216 * config/score/predicates.md (const_uimm5, sr0_operand, const_simm12,
4217 const_simm15, const_pow2, const_npow2): Added.
4218 * config/score/misc.md (insv, extv, extzv, movmemsi,
4219 move_lbu_a/b, mov_lhu_a/b etc): Added and fix some bug.
4220 * config/score/score.c (score_address_cost, score_select_cc_mode):
4222 Change CONST_OK_FOR_LETTER_P/EXTRA_CONSTRAINT define.
4223 Update score_rtx_costs for MACRO TARGET_RTX_COSTS.
4224 Update score_print_operand.
4225 * config/score/score.h (DATA_ALIGNMENT, SELECT_CC_MODE): Added.
4226 Adjust register allocate order and update some macro define.
4227 * config/score/score-mdaux.c (mdx_unaligned_load, mdx_unsigned_store,
4228 mdx_block_move_straight, mdx_block_move_loop_head,
4229 mdx_block_move_loop_body, mdx_block_move_loop_foot, mdx_block_move_loop,
4230 mdx_block_move): Added.
4231 (mdx_movsicc, mdp_select_add_imm, mdp_select, mds_zero_extract_andi,
4232 mdp_limm): Updated and fix some bug and typo.
4233 * config/score/score.md (movqi/hi/si, add/sub/zero/ext): Updated.
4234 (movsf, movdf, doloop_end): Added.
4236 2007-01-08 Kazu Hirata <kazu@codesourcery.com>
4238 * config/arm/arm.c, config/arm/arm.h, config/arm/arm.md,
4239 config/arm/thumb2.md: Fix comment typos.
4240 * doc/extend.texi: Fix a typo.
4242 2007-01-07 Eric Christopher <echristo@apple.com>
4244 * configure.ac: Check for __stack_chk_fail for darwin.
4245 * configure: Regenerate.
4247 2007-01-07 Richard Guenther <rguenther@suse.de>
4249 * tree-vrp.c (extract_range_from_assert): CSE calls to
4250 compare_values where possible.
4251 (extract_range_from_unary_expr): Likewise.
4253 2007-01-07 Anatoly Sokolov <aesok@post.ru>
4255 * config/avr/avr-protos.h (call_insn_operand): Delete prototype.
4256 * config/avr/avr.c (call_insn_operand): Delete function.
4257 * config/avr/avr.md (*pushqi, *pushhi, *pushsi, *pushsf): Use REG_SP
4258 instead of register number. Use predicates.
4259 * config/avr/predicates.md (const0_operand, reg_or_0_operand,
4260 call_insn_operand): Add.
4262 2007-01-06 Jan Hubicka <jh@suse.cz>
4264 * tree-pass.h (pass_build_cgraph_edges): Declare.
4265 * cgraphunit.c (record_refernece): Move to cgraphbuild.c
4266 (visited_nodes): Remove.
4267 (cgraph_create_edges): Move to cgraphbuild.c; rename to
4268 build_cgrpah_edges; make visited_nodes local.
4269 (cgraph_process_new_functions): DO not call initialize_inline_failed.
4270 (record_references_in_initializer): Move to cgraphbuild.c
4271 (initialize_inline_failed, rebuild_cgraph_edges,
4272 pass_rebuild_cgraph_edges): Move to cgraphbuild.c.
4273 (verify_cgraph_node): Make visited_nodes local.
4274 (cgraph_analyze_function): Do not call cgraph_create_edges and
4275 initialize_inline_failed.
4276 (cgraph_expand_function): Do not call cgraph_lower_function;
4277 assert that function is already lowered.
4278 * Makefile.in (cgraphbuild.o): New.
4279 * passes.c (init_optimization_passes): Add pass_build_cgraph_edges
4280 at the end of lowering passes.
4282 2007-01-06 Steven Bosscher <steven@gcc.gnu.org>
4284 * ifcvt.c (cond_move_convert_if_block): New function, code
4285 factored out from...
4286 (cond_move_process_if_block): ...here. Call the new function
4287 on the THEN and ELSE blocks.
4288 (merge_if_block): Do not copy global_live_at_end, merge_blocks
4289 already takes care of this.
4291 2007-01-05 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
4294 * tree.h (TREE_OVERFLOW_P): New.
4295 * c-typeck.c (parser_build_unary_op): Warn only if result
4296 overflowed and operands did not.
4297 (parser_build_binary_op): Likewise.
4298 (convert_for_assignment): Remove redundant overflow_warning.
4299 * c-common.c (overflow_warning): Don't check or set TREE_OVERFLOW.
4301 2007-01-05 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
4303 * c-typeck.c (store_init_value): Split over two lines to follow
4304 the GNU coding style.
4306 2007-01-05 Benjamin Kosnik <bkoz@redhat.com>
4308 * c-cppbuiltin.c (c_cpp_builtins): __GXX_EXPERIMENTAL_CPP0X__ to
4309 __GXX_EXPERIMENTAL_CXX0X__.
4310 * doc/cpp.texi: Same.
4312 2007-01-05 Richard Guenther <rguenther@suse.de>
4315 * tree.c (get_narrower): Do not construct COMPONENT_REFs
4316 with mismatched types. Instead explicitly build a
4317 conversion NOP_EXPR.
4319 2007-01-05 Ian Lance Taylor <iant@google.com>
4321 * c-common.c (decl_with_nonnull_addr_p): New function.
4322 (c_common_truthvalue_conversion): Call it.
4323 * c-typeck.c (build_binary_op): Likewise.
4324 * c-common.h (decl_with_nonnull_addr_p): Declare.
4326 2007-01-05 Jakub Jelinek <jakub@redhat.com>
4329 * libgcc2.c (__divdc3): Compare c and d against 0.0 instead of
4332 2007-01-05 Joel Brobecker <brobecker@adacore.com>
4334 * doc/install.texi (Final install): Document the fact that
4335 the GNAT runtime should not be stripped.
4337 2007-01-04 Jan Hubicka <jh@suse.cz>
4339 * tree-inline.c (fold_marked_statements): Update operand caches
4340 and EH after folding
4342 2007-01-04 Ian Lance Taylor <iant@google.com>
4344 * c-common.c (check_function_nonnull): Whitespace fix.
4346 2007-01-04 Jan Hubicka <jh@suse.cz>
4348 * tree-optimize.c (execute_fixup_cfg): Correct previously mistakely
4349 comitted older version of patch.
4350 (pass_fixup_cfg): Add TODOs to verify flow and statements, dump
4351 function, celanup cfg and collect garbage.
4353 2007-01-04 Mike Stump <mrs@apple.com>
4355 * Makefile.in (mostlyclean): Don't remove libgcc anymore.
4358 2007-01-04 Eric Christopher <echristo@apple.com>
4360 * libgcc2.c (__bswapsi2): Use SItype.
4361 (__bswapdi2): Use DItype.
4362 * libgcc2.h: Update for above.
4364 2007-01-04 Paul Brook <paul@codesourcery.com>
4366 * config/arm/arm.md (arm_mulsi3, thumb_mulsi3, mulsi3_compare0,
4367 mulsi_compare0_scratch, mulsi3addsi, mulsi3addsi_compare0,
4368 mulsi3addsi_compare0_scratch, mulsidi3adddi, mulsidi3,
4369 umulsidi3, umulsidi3adddi, smulsi3_highpart,
4370 umulsi3_highpart): Make conditional on !arm_arch6.
4371 (arm_mulsi3_v6, thumb_mulsi3_v6, mulsi3_compare0_v6,
4372 mulsi_compare0_scratch_v6, mulsi3addsi_v6, mulsi3addsi_compare0_v6,
4373 mulsi3addsi_compare0_scratch_v6, mulsidi3adddi_v6, mulsidi3_v6,
4374 umulsidi3_v6, umulsidi3adddi_v6, smulsi3_highpart_v6,
4375 umulsi3_highpart_v6): New insns.
4377 2007-01-04 Roger Sayle <roger@eyesopen.com>
4379 * fold-const.c (fold_convert): When casting an expression to void,
4380 fold_ignored_result may discover a GIMPLE_MODIFY_STMT which doesn't
4381 have a type. Instead of attempting to build a NOP_EXPR, return
4382 these "special" trees directly.
4384 2007-01-04 Joseph Myers <joseph@codesourcery.com>
4386 * config/rs6000/rs6000.c (rs6000_rtx_costs): Make adjustment for
4387 MULT inside MINUS as either argument. Use rs6000_cost->dmul -
4388 rs6000_cost->fp not 0 as adjustment for outer NEG.
4390 2007-01-04 Jan Hubicka <jh@suse.cz>
4392 * cgraph.c (cgraph_release_function_body): New function.
4393 (cgraph_remove_node): Use it.
4394 * cgraph.h (cgraph_release_function_body): Declare.
4395 * cgraphunit.c (cgraph_expand_function): Use it.
4396 * ipa.c (cgraph_remove_unreahchable_nodes): Use it.
4397 * tree-ssa.c (delete_tree_ssa): Allow to be called before aliasing
4398 is initialized and while compilation of other function is running.
4399 * tree-optimize.c (execute_free_cfg_annotations): Move code to clear
4400 statement CFG annotations from here to ...
4401 * tree-cfg.c (delete_tree_cfg_annotations): ... here.
4403 2007-01-04 Zdenek Dvorak <dvorakz@suse.cz>
4405 * cfgloop.h (enum li_flags): Make the constants powers of two.
4407 2007-01-04 Jan Hubicka <jh@suse.cz>
4409 * tree-inline.c (copy_bb): Insert new statements to statements_to_fold
4411 (fold_marked_statements): New function.
4412 (optimize_inline_calls, tree_function_versioning): Fold new statements.
4413 * tree-inline.h (copy_body_data): Add statements_to_fold.
4415 2007-01-03 Daniel Jacobowitz <dan@codesourcery.com>
4417 * config.gcc: Mention libgcc/config.host.
4418 * Makefile.in: Update comments mentioning libgcc.
4419 (LIBGCC, INSTALL_LIBGCC, GCC_PARTS, mklibgcc): Delete.
4420 (all.cross, start.encap, rest.encap, rest.cross): Update
4421 dependencies for libgcc move.
4422 (libgcc.mk, LIBGCC_DEPS, libgcov.a, libgcc.a, stmp-multilib)
4423 (clean-target, clean-target-libgcc): Delete.
4424 (srcdirify, GCC_EXTRA_PARTS): New macros.
4425 (libgcc-support, libgcc.mvars): New rules.
4426 (distclean): Remove mention of mklibgcc.
4427 (install): Don't reference INSTALL_LIBGCC.
4428 (install-common): Don't reference EXTRA_PARTS.
4429 (install-libgcc, install-multilib): Delete rules.
4430 * mklibgcc.in: Delete file.
4431 * doc/configfiles.texi: Don't mention mklibgcc.
4433 * config/i386/t-darwin (SHLIB_VERPFX): Delete (moved to libgcc).
4434 * config/i386/t-darwin64 (SHLIB_VERPFX): Likewise.
4435 * config/rs6000/t-darwin (SHLIB_VERPFX): Likewise.
4436 * config/rs6000/t-ppccomm (TARGET_LIBGCC2_CFLAGS, SHLIB_MAPFILES)
4437 (mklibgcc, ldblspecs): Likewise.
4439 * config/i386/t-nwld (libgcc.def, libc.def, libpcre.def)
4440 (posixpre.def): Use $(T).
4441 (SHLIB_EXT, SHLIB_NAME, SHLIB_SLIBDIR_QUAL, SHLIB_DEF, SHLIB_MAP)
4442 (SHLIB_SRC, SHLIB_INSTALL): Delete.
4443 (SHLIB_LINK): Make dummy.
4444 * config/t-slibgcc-darwin: Delete contents except for dummy SHLIB_LINK.
4446 * config/frv/t-linux (EXTRA_MULTILIB_PARTS): Clear.
4448 * config/alpha/t-crtfm: Use $(T) in rules for EXTRA_PARTS.
4449 * config/alpha/t-vms, config/alpha/t-vms64, config/fr30/t-fr30,
4450 config/i386/t-rtems-i386, config/ia64/t-ia64, config/rs6000/t-beos,
4451 config/rs6000/t-newas, config/sparc/t-elf: Likewise.
4453 * configure.ac (all_outputs): Remove mklibgcc.
4454 * configure: Regenerated.
4456 2007-01-03 Josh Conner <jconner@apple.com>
4459 * calls.c (compute_argument_addresses): Set stack and stack_slot
4460 for partial args, too.
4461 (store_one_arg): Use locate.size.constant for the size when
4462 generating a save_area.
4464 2007-01-03 Robert Kennedy <jimbob@google.com>
4466 * tree-cfg.c (tree_merge_blocks): Release SSA_NAME phi results
4467 whose definitions are deleted due to basic block merging.
4469 2007-01-03 Paul Brook <paul@codesourcery.com>
4472 * config/arm/arm.c (output_return_instruction): Pop PC in interrupt
4474 (use_return_insn): Return 0 for Thumb interrupt functions.
4475 (print_multi_reg): Add rfe argument for IRQ returns.
4476 (arm_output_epilogue): Pop interrupt return address directly into PC.
4477 (arm_expand_prologue): Only adjust IRQ return address in Arm mode.
4479 2007-01-03 Paul Brook <paul@codesourcery.com>
4481 Merge from sourcerygxx-4_1.
4482 * config/arm/thumb2.md: New file.
4483 * config/arm/elf.h (JUMP_TABLES_IN_TEXT_SECTION): Return True for
4485 * config/arm/coff.h (JUMP_TABLES_IN_TEXT_SECTION): Ditto.
4486 * config/arm/aout.h (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
4487 (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump tables.
4488 * config/arm/aof.h (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump
4490 (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
4491 * config/arm/ieee754-df.S: Use macros for Thumb-2/Unified asm
4493 * config/arm/ieee754-sf.S: Ditto.
4494 * config/arm/arm.c (thumb_base_register_rtx_p): Rename...
4495 (thumb1_base_register_rtx_p): ... to this.
4496 (thumb_index_register_rtx_p): Rename...
4497 (thumb1_index_register_rtx_p): ... to this.
4498 (thumb_output_function_prologue): Rename...
4499 (thumb1_output_function_prologue): ... to this.
4500 (thumb_legitimate_address_p): Rename...
4501 (thumb1_legitimate_address_p): ... to this.
4502 (thumb_rtx_costs): Rename...
4503 (thumb1_rtx_costs): ... to this.
4504 (thumb_compute_save_reg_mask): Rename...
4505 (thumb1_compute_save_reg_mask): ... to this.
4506 (thumb_final_prescan_insn): Rename...
4507 (thumb1_final_prescan_insn): ... to this.
4508 (thumb_expand_epilogue): Rename...
4509 (thumb1_expand_epilogue): ... to this.
4510 (arm_unwind_emit_stm): Rename...
4511 (arm_unwind_emit_sequence): ... to this.
4512 (thumb2_legitimate_index_p, thumb2_legitimate_address_p,
4513 thumb1_compute_save_reg_mask, arm_dwarf_handle_frame_unspec,
4514 thumb2_index_mul_operand, output_move_vfp, arm_shift_nmem,
4515 arm_save_coproc_regs, thumb_set_frame_pointer, arm_print_condition,
4516 thumb2_final_prescan_insn, thumb2_asm_output_opcode, arm_output_shift,
4517 thumb2_output_casesi): New functions.
4518 (TARGET_DWARF_HANDLE_FRAME_UNSPEC): Define.
4519 (FL_THUMB2, FL_NOTM, FL_DIV, FL_FOR_ARCH6T2, FL_FOR_ARCH7,
4520 FL_FOR_ARCH7A, FL_FOR_ARCH7R, FL_FOR_ARCH7M, ARM_LSL_NAME,
4521 THUMB2_WORK_REGS): Define.
4522 (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv, arm_condexec_count,
4523 arm_condexec_mask, arm_condexec_masklen)): New variables.
4524 (all_architectures): Add armv6t2, armv7, armv7a, armv7r and armv7m.
4525 (arm_override_options): Check new CPU capabilities.
4526 Set new architecture flag variables.
4527 (arm_isr_value): Handle v7m interrupt functions.
4528 (user_return_insn): Return 0 for v7m interrupt functions. Handle
4530 (const_ok_for_arm): Handle Thumb-2 constants.
4531 (arm_gen_constant): Ditto. Use movw when available.
4532 (arm_function_ok_for_sibcall): Return false for v7m interrupt
4534 (legitimize_pic_address, arm_call_tls_get_addr): Handle Thumb-2.
4535 (thumb_find_work_register, arm_load_pic_register,
4536 legitimize_tls_address, arm_address_cost, load_multiple_sequence,
4537 emit_ldm_seq, emit_stm_seq, arm_select_cc_mode, get_jump_table_size,
4538 print_multi_reg, output_mov_long_double_fpa_from_arm,
4539 output_mov_long_double_arm_from_fpa, output_mov_double_fpa_from_arm,
4540 output_mov_double_fpa_from_arm, output_move_double,
4541 arm_compute_save_reg_mask, arm_compute_save_reg0_reg12_mask,
4542 output_return_instruction, arm_output_function_prologue,
4543 arm_output_epilogue, arm_get_frame_offsets, arm_regno_class,
4544 arm_output_mi_thunk, thumb_set_return_address): Ditto.
4545 (arm_expand_prologue): Handle Thumb-2. Use arm_save_coproc_regs.
4546 (arm_coproc_mem_operand): Allow POST_INC/PRE_DEC.
4547 (arithmetic_instr, shift_op): Use arm_shift_nmem.
4548 (arm_print_operand): Use arm_print_condition. Handle '(', ')', '.',
4550 (arm_final_prescan_insn): Use extract_constrain_insn_cached.
4551 (thumb_expand_prologue): Use thumb_set_frame_pointer.
4552 (arm_file_start): Output directive for unified syntax.
4553 (arm_unwind_emit_set): Handle stack alignment instruction.
4554 * config/arm/lib1funcs.asm: Remove default for __ARM_ARCH__.
4555 Add v6t2, v7, v7a, v7r and v7m.
4556 (RETLDM): Add Thumb-2 code.
4557 (do_it, shift1, do_push, do_pop, COND, THUMB_SYNTAX): New macros.
4558 * config/arm/arm.h (TARGET_CPU_CPP_BUILTINS): Define __thumb2__.
4559 (TARGET_THUMB1, TARGET_32BIT, TARGET_THUMB2, TARGET_DSP_MULTIPLY,
4560 TARGET_INT_SIMD, TARGET_UNIFIED_ASM, ARM_FT_STACKALIGN, IS_STACKALIGN,
4561 THUMB2_TRAMPOLINE_TEMPLATE, TRAMPOLINE_ADJUST_ADDRESS,
4562 ASM_OUTPUT_OPCODE, THUMB2_GO_IF_LEGITIMATE_ADDRESS,
4563 THUMB2_LEGITIMIZE_ADDRESS, CASE_VECTOR_PC_RELATIVE,
4564 CASE_VECTOR_SHORTEN_MODE, ADDR_VEC_ALIGN, ASM_OUTPUT_CASE_END,
4565 ADJUST_INSN_LENGTH): Define.
4566 (TARGET_REALLY_IWMMXT, TARGET_IWMMXT_ABI, CONDITIONAL_REGISTER_USAGE,
4567 STATIC_CHAIN_REGNUM, HARD_REGNO_NREGS, INDEX_REG_CLASS,
4568 BASE_REG_CLASS, MODE_BASE_REG_CLASS, SMALL_REGISTER_CLASSES,
4569 PREFERRED_RELOAD_CLASS, SECONDARY_OUTPUT_RELOAD_CLASS,
4570 SECONDARY_INPUT_RELOAD_CLASS, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P,
4571 TRAMPOLINE_SIZE, INITIALIZE_TRAMPOLINE, HAVE_PRE_INCREMENT,
4572 HAVE_POST_DECREMENT, HAVE_PRE_DECREMENT, HAVE_PRE_MODIFY_DISP,
4573 HAVE_POST_MODIFY_DISP, HAVE_PRE_MODIFY_REG, HAVE_POST_MODIFY_REG,
4574 REGNO_MODE_OK_FOR_BASE_P, LEGITIMATE_CONSTANT_P,
4575 REG_MODE_OK_FOR_BASE_P, REG_OK_FOR_INDEX_P, GO_IF_LEGITIMATE_ADDRESS,
4576 LEGITIMIZE_ADDRESS, THUMB2_LEGITIMIZE_ADDRESS,
4577 GO_IF_MODE_DEPENDENT_ADDRESS, MEMORY_MOVE_COST, BRANCH_COST,
4578 ASM_APP_OFF, ASM_OUTPUT_CASE_LABEL, ARM_DECLARE_FUNCTION_NAME,
4579 FINAL_PRESCAN_INSN, PRINT_OPERAND_PUNCT_VALID_P,
4580 PRINT_OPERAND_ADDRESS): Adjust for Thumb-2.
4581 (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv): New declarations.
4582 * config/arm/arm-cores.def: Add arm1156t2-s, cortex-a8, cortex-r4 and
4584 * config/arm/arm-tune.md: Regenerate.
4585 * config/arm/arm-protos.h: Update prototypes.
4586 * config/arm/vfp.md: Enable patterns for Thumb-2.
4587 (arm_movsi_vfp): Add movw alternative. Use output_move_vfp.
4588 (arm_movdi_vfp, movsf_vfp, movdf_vfp): Use output_move_vfp.
4589 (thumb2_movsi_vfp, thumb2_movdi_vfp, thumb2_movsf_vfp,
4590 thumb2_movdf_vfp, thumb2_movsfcc_vfp, thumb2_movdfcc_vfp): New.
4591 * config/arm/libunwind.S: Add Thumb-2 code.
4592 * config/arm/constraints.md: Update include Thumb-2.
4593 * config/arm/ieee754-sf.S: Add Thumb-2/Unified asm support.
4594 * config/arm/ieee754-df.S: Ditto.
4595 * config/arm/bpabi.S: Ditto.
4596 * config/arm/t-arm (MD_INCLUDES): Add thumb2.md.
4597 * config/arm/predicates.md (low_register_operand,
4598 low_reg_or_int_operand, thumb_16bit_operator): New.
4599 (thumb_cmp_operand, thumb_cmpneg_operand): Rename...
4600 (thumb1_cmp_operand, thumb1_cmpneg_operand): ... to this.
4601 * config/arm/t-arm-elf: Add armv7 multilib.
4602 * config/arm/arm.md: Update patterns for Thumb-2 and Unified asm.
4604 (UNSPEC_STACK_ALIGN, ce_count): New.
4605 (arm_incscc, arm_decscc, arm_umaxsi3, arm_uminsi3,
4606 arm_zero_extendsidi2, arm_zero_extendqidi2): New
4608 * config/arm/fpa.md: Update patterns for Thumb-2 and Unified asm.
4609 (thumb2_movsf_fpa, thumb2_movdf_fpa, thumb2_movxf_fpa,
4610 thumb2_movsfcc_fpa, thumb2_movdfcc_fpa): New insns.
4611 * config/arm/cirrus.md: Update patterns for Thumb-2 and Unified asm.
4612 (cirrus_thumb2_movdi, cirrus_thumb2_movsi_insn,
4613 thumb2_cirrus_movsf_hard_insn, thumb2_cirrus_movdf_hard_insn): New
4615 * doc/extend.texi: Document ARMv7-M interrupt functions.
4616 * doc/invoke.texi: Document Thumb-2 new cores+architectures.
4618 2007-01-03 Jakub Jelinek <jakub@redhat.com>
4620 * unwind-dw2.c (SIGNAL_FRAME_BIT, EXTENDED_CONTEXT_BIT): Define.
4621 (struct _Unwind_Context): Rename args_size to flags, remove
4622 signal_frame field, add a new args_size field and version field.
4623 (_Unwind_IsSignalFrame, _Unwind_SetSignalFrame,
4624 _Unwind_IsExtendedContext): New inline functions.
4625 (_Unwind_GetGR, _Unwind_SetGR, _Unwind_GetGRPtr, _Unwind_SetGRPtr):
4626 Assume by_value array is only present if _Unwind_IsExtendedContext.
4627 (_Unwind_GetIPInfo, execute_cfa_program, uw_frame_state_for): Use
4628 _Unwind_IsSignalFrame.
4629 (__frame_state_for): Initialize context.flags to EXTENDED_CONTEXT_BIT.
4630 (uw_update_context_1): Use _Unwind_SetSignalFrame.
4631 (uw_init_context_1): Initialize context->flags to
4632 EXTENDED_CONTEXT_BIT.
4633 * config/rs6000/linux-unwind.h (frob_update_context): Use
4634 _Unwind_SetSignalFrame.
4636 2007-01-03 Andrew Pinski <andrew_pinski@playstation.sony.com>
4639 * gimplify.c (gimplify_modify_expr_complex_part): Move below
4640 tree_to_gimple_tuple. Call tree_to_gimple_tuple when we need
4643 2007-01-03 Kazu Hirata <kazu@codesourcery.com>
4645 * config/i386/i386.h (NON_STACK_REG_P, REGNO_OK_FOR_SIREG_P,
4646 REGNO_OK_FOR_DIREG_P, REWRITE_ADDRESS, ASM_OPERAND_LETTER,
4647 RET, AT_SP): Remove.
4648 * config/i386/i386.md (*sse_prologue_save_insn): Use return
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 * cgraphunit.c, tree-ssa-alias.c: Fix comment typos.
4664 2007-01-03 Zdenek Dvorak <dvorakz@suse.cz>
4666 * loop-unswitch.c (unswitch_loop): Pass probabilities to loopify.
4667 * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Pass probabilities
4669 * cfgloopmanip.c (scale_loop_frequencies): Export.
4670 (loopify): Scale the frequencies by prescribed coefficients.
4671 (set_zero_probability): New function.
4672 (duplicate_loop_to_header_edge): Improve updating of frequencies.
4673 (lv_adjust_loop_entry_edge, loop_version): Set probabilities
4674 and frequencies according to arguments.
4675 * tree-ssa-loop-manip.c (tree_unroll_loop): Set probabilities
4677 * cfg.c (scale_bbs_frequencies_int): Allow scaling the frequencies up.
4678 * modulo-sched.c (sms_schedule): Set probabilities for entering
4679 versioned loop correctly.
4680 * tree-vect-transform.c (vect_transform_loop): Ditto.
4681 * cfgloop.h (loopify, loop_version): Declaration changed.
4682 (scale_loop_frequencies): Declared.
4684 2007-01-02 Jan Hubicka <jh@suse.cz>
4686 * cgraph.c: Include tree-flow.h
4687 (cgraph_add_new-function): Handle IPA_SSA mode; execute
4689 * cgraph.h (enum cgraph_state): Add CGRAPH_STATE_IPA_SSA.
4690 * tree-pass.h (pass_all_early_optimizations): Declare.
4691 * cgraphunit.c (cgraph_process_new_functions): Add IPA_SSA; execute
4693 (cgraph_analyze_function): Do early_local_passes.
4694 * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
4695 Do not add referenced vars.
4696 * tree-optimize.c (gate_all_optimizations): Do not execute when not in
4698 (gate_all_early_local_passes): New gate.
4699 (pass_early_local_passes): Use new gate.
4700 (execute_early_local_optimizations): New functions.
4701 (gate_all_early_optimizations): New gate.
4702 (pass_all_early_optimizations): New pass.
4703 (execute_free_datastructures): Free SSA only when initialized.
4704 (gate_init_datastructures): Init only when optimizing.
4705 (tree_lowering_passes): Do early local passes when called late.
4706 * tree-profile.c (do_tree_profiling): Don't profile functions added
4708 (do_early_tree_profiling, pass_early_tree_profile): Kill.
4709 * tree-cfg.c (update_modified_stmts): Do not update when operands are
4711 * passes.c (init_optimizations_passes): Reorder so we go into SSA
4712 during early_local_passes.
4713 * Makefile.in (cgraph.o): Add dependency on tree-flow.h.
4716 2007-01-02 Carlos O'Donell <carlos@codesourcery.com>
4718 * Makefile.in: Update copyright year.
4720 2007-01-02 Carlos O'Donell <carlos@codesourcery.com>
4722 * Makefile.in: Export GCC_EXEC_PREFIX before calling $(RUNTEST)
4723 in $(lang_checks) and check-consistency targets.
4725 2007-01-02 Jan Hubicka <jh@suse.cz>
4727 * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
4728 Do not add referenced vars.
4729 * tree-cfg.c (update_modified_stmts): Do not update when SSA operands
4731 * passes.c (init_optimization_passes): Put mudflap_2 after
4732 free_datastructures.
4734 2007-01-02 Jan Hubicka <jh@suse.cz>
4736 * tree-optimize (execute_fixup_cfg): Set after_inlining flag.
4737 Set NOTHROW flag on call statements proved to be nothrow.
4738 Update statement of local calls so new pure/const functions are
4739 updated. Update_ssa when in ssa form. Mark PHI nodes of nonlocal
4741 (tree_rest_of_compilation): Register hooks and initialize bitmap
4742 early. Do not set after_inlining flag.
4744 2007-01-02 Steve Ellcey <sje@cup.hp.com>
4746 * sbitmap.c (HOST_BITS_PER_LONG_LONG): Change to
4747 HOST_BITS_PER_LONGLONG
4749 2007-01-02 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
4752 * c-typeck.c (store_init_value): Don't emit pedantic overflow
4753 warning for non-static initializers.
4755 2007-01-02 Steven Bosscher <steven@gcc.gnu.org>
4757 * config/alpha/alpha.md, arm/arm.c, darwin.c, frv/frv.md,
4758 m32r/m32r.c, m32r/m32r.c, mn10300/mn10300.md, pa/pa.c,
4759 rs6000/rs6000.c, s390/s390.md, sh/sh.md, sparc/sparc.c:
4760 Always use set_unique_reg_note to add REG_EQUAL notes.
4762 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
4765 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
4767 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
4768 * cfgbuild.c (find_basic_blocks): Likewise.
4769 * cfgrtl.c (rtl_create_basic_block): Likewise.
4770 * function.c (temp_slots_at_level): Likewise.
4771 * reg-stack.c (stack_regs_mentioned): Likewise.
4772 * regclass.c (allocate_reg_info): Likewise.
4773 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
4774 set_bb_for_stmt, move_block_to_fn): Likewise.
4775 * tree-complex.c (tree_lower_complex): Likewise.
4776 * vec.h (VEC_safe_grow_cleared): New.
4778 2007-01-02 Ian Lance Taylor <iant@google.com>
4780 * c-common.c (c_common_truthvalue_conversion): When warning about
4781 using an assignment as a truth value, set TREE_NO_WARNING.
4783 2007-01-02 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
4786 * c.opt (Wold-style-declaration): New.
4787 * doc/invoke.texi (C-only Warning Options): New.
4788 (Wold-style-declaration): Document it.
4789 (Wextra): Enabled by -Wextra.
4790 * c-opts.c (c_common_post_options): Enabled by -Wextra.
4791 * c-decl.c (declspecs_add_scspec): Replace -Wextra with
4792 -Wold-style-declaration.
4794 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
4796 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
4797 * cfgbuild.c (find_basic_blocks): Likewise.
4798 * cfgrtl.c (rtl_create_basic_block): Likewise.
4799 * function.c (temp_slots_at_level): Likewise.
4800 * reg-stack.c (stack_regs_mentioned): Likewise.
4801 * regclass.c (allocate_reg_info): Likewise.
4802 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
4803 set_bb_for_stmt, move_block_to_fn): Likewise.
4804 * tree-complex.c (tree_lower_complex): Likewise.
4805 * vec.h (VEC_safe_grow_cleared): New.
4807 2007-01-02 Douglas Gregor <doug.gregor@gmail.com>
4809 * c-common.c (c_common_nodes_and_builtins): Since variants of
4810 void_type_node get built before it is given a name, we need to
4811 give those variants the name, too.
4812 (complete_array_type): We need to work with the canonical main
4813 type of the array, from which we will build the qualified version.
4814 * params.def (PARAM_VERIFY_CANONICAL_TYPES): New.
4815 * print-tree.c (print_node): Display canonical type information
4817 * stor-layout.c (layout_type): When we don't know the
4818 alignment of a type for which we're building an array, we end up
4819 guessing wrong, so make the type require structural equality.
4820 * tree.c (make_node_stat): When we build a new type, it is its
4822 (build_type_attribute_qual_variant): When building an attribute
4823 variant, its canonical type is the non-attribute variant. However,
4824 if the attributes are target-dependent and they differ, we need to
4825 use structural equality checks for this type.
4826 (build_qualified_type): A qualified type is not equivalent to its
4827 unqualified variant; set the canonical type appropriately.
4828 (build_distinct_type_copy): When building a distinct type from
4829 another type, the new type is its own canonical type.
4830 (build_variant_type_copy): When building a new type variant, we
4831 assume that it is equivalent to the original type.
4832 (build_pointer_type_for_mode): When building a pointer type, also
4833 build a canonical type pointer.
4834 (build_reference_type_for_mode): When building a reference type,
4835 also build a canonical type reference.
4836 (build_index_type): When we can't hash an index type (e.g.,
4837 because its maximum value is negative), the index type requires
4838 structural equality tests.
4839 (build_array_type): Build the canonical form of an array type.
4840 (build_function_type): Function types require structural equality,
4841 because they contain default arguments, attributes, etc.
4842 (build_method_type_directly): Ditto for method types.
4843 (build_offset_type): Build the canonical offset type.
4844 (build_complex_type): Build the canonical vector type.
4845 (make_vector_type): Build the canonical vector type.
4846 * tree.h (TYPE_CANONICAL): New.
4847 (TYPE_STRUCTURAL_EQUALITY_P): New.
4848 (SET_TYPE_STRUCTURAL_EQUALITY): New.
4849 (struct tree_type): Added "canonical" field.
4850 * params.h (VERIFY_CANONICAL_TYPES): New.
4851 * doc/c-tree.texi (TYPE_CANONICAL): Document.
4852 (TYPE_STRUCTURAL_EQUALITY_P): Document.
4853 (SET_TYPE_STRUCTURAL_EQUALITY): Document.
4854 * doc/invoke.texi (verify-canonical-types): Document --param
4855 parameter for verifying canonical types.
4857 2007-01-02 Joseph Myers <joseph@codesourcery.com>
4859 * config.gcc (powerpc-*-eabispe*, powerpc-*-eabisimaltivec*,
4860 powerpc-*-eabisim*, powerpc-*-eabialtivec*, powerpc-*-eabi*,
4861 powerpc-*-rtems*, powerpc-wrs-vxworks, powerpc-wrs-vxworksae,
4862 powerpcle-*-eabisim*, powerpcle-*-eabi*): Add rs6000/e500.h to
4864 * config/rs6000/e500.h: New.
4865 * config/rs6000/eabi.h (TARGET_SPE_ABI, TARGET_SPE, TARGET_E500,
4866 TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE, TARGET_E500_DOUBLE):
4868 * config/rs6000/linuxspe.h (TARGET_SPE_ABI, TARGET_SPE,
4869 TARGET_E500, TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE,
4870 TARGET_E500_DOUBLE): Remove.
4871 * config/rs6000/vxworks.h (TARGET_SPE_ABI, TARGET_SPE,
4872 TARGET_E500, TARGET_ISEL, TARGET_FPRS): Remove.
4873 * config/rs6000/rs6000.h (CHECK_E500_OPTIONS): Define.
4874 * config/rs6000/rs6000.c (rs6000_override_options): Use
4877 2007-01-02 Joseph Myers <joseph@codesourcery.com>
4879 * config/rs6000/rs6000.c (print_operand): Check (TARGET_SPE ||
4880 TARGET_E500_DOUBLE), not TARGET_E500, for %y.
4881 (rs6000_generate_compare, rs6000_emit_sCOND, output_cbranch,
4882 rs6000_emit_cmove): Don't check TARGET_E500.
4883 * config/rs6000/rs6000.md (bunordered, bordered, sunordered,
4884 sordered): Don't check TARGET_E500.
4886 2007-01-01 Eric Christopher <echristo@apple.com>
4888 * config/mips/mips.c (mips_regno_mode_ok_for_base_p): Use
4889 HARD_REGISTER_NUM_P.
4891 2007-01-01 Roger Sayle <roger@eyesopen.com>
4893 * fold-const.c (fold_binary) <EQ_EXPR>: Fold "(X^C1) eq/ne C2" into
4894 "X eq/ne (C1^C2)". Fold "(X^Z) eq/ne (Y^Z)" as "X eq/ne Y" when Z
4895 has no side-effects. Fold "(X^C1) eq/ne (Y^C2)" as "(X^(C1^C2))
4898 2007-01-01 Mike Stump <mrs@apple.com>
4900 * configure.ac: Remove support for building with Apple's gcc-3.1.
4902 2007-01-02 Joseph Myers <joseph@codesourcery.com>
4905 * caller-save.c (add_stored_regs): Only handle SUBREGs if inner
4906 REG is a hard register. Do not modify REG before calling
4908 * rtlanal.c (subreg_get_info): Don't assert size of XMODE is a
4909 multiple of the size of YMODE for certain lowpart cases.
4911 2007-01-01 Andrew Pinski <pinskia@gmail.com>
4914 * gimplify (voidify_wrapper_expr): Update for
4917 2007-01-01 Andreas Schwab <schwab@suse.de>
4920 * config/ia64/ia64.c (ia64_compute_frame_size): Account space for
4921 save of BR0 in extra_spill_size instead of spill_size.
4922 (ia64_expand_prologue): Save BR0 outside of the gr/br/fr spill
4924 (ia64_expand_epilogue): Restore BR0 from its new location.
4926 2007-01-01 Andrew Pinski <pinskia@gmail.com>
4928 * gimplify.c (gimplify_init_constructor <case VECTOR_TYPE>):
4929 Use a temporary variable if the left hand side is not a gimple
4932 2007-01-01 Andrew Pinski <pinskia@gmail.com>
4934 * gimplify.c (gimplify_return_expr): Make the temporary variable
4935 for the return expression, a gimple register variable.
4937 2007-01-01 Jan Hubicka <jh@suse.cz>
4939 * emit-rtl.c (emit_copy_of_insn_after): Do not call copy_insn_1 for
4942 2007-01-01 Mike Stump <mrs@apple.com>
4944 * configure.ac (HAVE_GAS_LITERAL16): Add autoconf check for
4946 * config/darwin.c (machopic_select_rtx_section): Use
4948 (darwin_mergeable_constant_section): Likewise.
4949 * configure: Regenerate.
4950 * config.in: Regenerate.
4952 2007-01-01 Jan Hubicka <jh@suse.cz>
4953 Andrew Pinski <pinskia@gmail.com>
4955 * cgraphunit.c (cgraph_optimize): Call cgraph_add_new_functions
4956 before starting IPA passes.